threads
listlengths 1
2.99k
|
|---|
[
{
"msg_contents": "> Or do you really want to commit the previous statements in the\n> transaction rather than throw them away? If so, you didn't really need\n> the transaction block there anyway, since you are willing to accept a\n> partial result.\n\nActually the postgresql behavior on error has been nagging on me also (OK, only a little bit). \nYes, the usual behavior is for the application to decide if it wants to roll back \nall other work on error or go on and correct the error and then commit all work \neven if one part did not succeed.\n\nThis is what is usually done, and actually helps performance since it saves one select:\n\n\nbegin work;\n\tdo arbitrary work\n\tinsert /* violates primary key constraint --> ERROR */ \n\ton error do instead update the offending row\n\tdo other work\ncommit work;\n\nThe decision on whether to do the update or the insert first is based on frequency of \nsuccessful events for the first statement. \n(more inserts --> insert first, more updates --> update first)\nAll commercial DBMS's handle it this way, it is probably even in ANSI.\n\nAndreas\n\t\n\n",
"msg_date": "Tue, 22 Sep 1998 08:46:34 +0200",
"msg_from": "Andreas Zeugswetter <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Errors inside transactions"
}
] |
[
{
"msg_contents": "I was working on docs and notice that Postgres no longer responds with a\nwarning if a database is dropped multiple times:\n\npostgres=> drop database yahoo;\nDESTROYDB\npostgres=> drop database yahoo;\nDESTROYDB\npostgres=> \\c yahoo\nconnecting to new database: yahoo\nFATAL 1: Database 'yahoo' does not exist.\ngolem$ psql\npostgres=> drop database yahoo;\nDESTROYDB\n\nThe destroydb command behaves similarly. The behavior documented by\nJose' was:\n\npostgres=> drop database yahoo;\nDESTROYDB\npostgres=> drop database yahoo;\nWARN: destroydb: database \"yahoo\" does not exist.\n\nBruce, this should probably go on the open issues list and be fixed for\nv6.4...\n\n - Tom\n",
"msg_date": "Tue, 22 Sep 1998 13:46:33 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Problem dropping databases"
},
{
"msg_contents": "> I was working on docs and notice that Postgres no longer responds with \n> a warning if a database is dropped multiple times: \n> Bruce, this should probably go on the open issues list and be fixed \n> for v6.4...\n\nOops. There isn't a problem. I had left a line of diagnostic code in the\nbackend, and it turns out to have been important :*(\n\nSorry for the false alarm...\n\n - Tom\n",
"msg_date": "Tue, 22 Sep 1998 16:20:37 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Problem dropping databases"
},
{
"msg_contents": "> I was working on docs and notice that Postgres no longer responds with a\n> warning if a database is dropped multiple times:\n> \n> postgres=> drop database yahoo;\n> DESTROYDB\n> postgres=> drop database yahoo;\n> DESTROYDB\n> postgres=> \\c yahoo\n> connecting to new database: yahoo\n> FATAL 1: Database 'yahoo' does not exist.\n> golem$ psql\n> postgres=> drop database yahoo;\n> DESTROYDB\n> \n> The destroydb command behaves similarly. The behavior documented by\n> Jose' was:\n> \n> postgres=> drop database yahoo;\n> DESTROYDB\n> postgres=> drop database yahoo;\n> WARN: destroydb: database \"yahoo\" does not exist.\n> \n> Bruce, this should probably go on the open issues list and be fixed for\n> v6.4...\n\nYep.\n\n-- \nBruce Momjian | 830 Blythe Avenue\[email protected] | Drexel Hill, Pennsylvania 19026\nhttp://www.op.net/~candle | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. |\n",
"msg_date": "Tue, 22 Sep 1998 15:43:05 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Problem dropping databases"
},
{
"msg_contents": "Hello Bruce,\n\nmarted�, 22 settembre 98, you wrote:\n\n>> I was working on docs and notice that Postgres no longer responds with a\n>> warning if a database is dropped multiple times:\n>> \n>> postgres=> drop database yahoo;\n>> DESTROYDB\n>> postgres=> drop database yahoo;\n>> DESTROYDB\n>> postgres=> \\c yahoo\n>> connecting to new database: yahoo\n>> FATAL 1: Database 'yahoo' does not exist.\n>> golem$ psql\n>> postgres=> drop database yahoo;\n>> DESTROYDB\n>> \n>> The destroydb command behaves similarly. The behavior documented by\n>> Jose' was:\n\nSorry for interference, I'm Jose'.\nSeems that my problem is a little bit different. I've tried almost all\nPostgreSQL commands and the result is that only the following commands\nkill the backend, i.e.:\n\n - DELETE\n - DROP DATABASE\n - DROP USER\n - LOCK TABLE\n\nPQexec() -- Request was sent to backend, but backend closed the channel before\nresponding.\n This probably means the backend terminated abnormally before or while\nprocessing the request.\n\n-------------------------------------------------------------\nI tried also with success the following commands:\n\nalter table\nalter user\nbegin\ncopy\ncreate aggregate\ncreate database\ncreate function\ncreate index\ncreate language\ncreate operator\ncreate sequence\ncreate table\ncreate trigger\ncreate type\ncreate user\ncreate view\ndeclare\ndrop aggregate\ndrop function\ndrop index\ndrop language\ndrop operator\ndrop sequence\ndrop table\ndrop trigger\ndrop type\ndrop view\nfetch\ngrant\ninsert\nmove\nrevoke\nrollback\nselect\nupdate\nvacuum\n---------------------------------------------\nAny ideas ?\n\n Jose' \n\n\n",
"msg_date": "Thu, 24 Sep 1998 14:35:52 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Problem dropping databases"
},
{
"msg_contents": "> >> I was working on docs and notice that Postgres no longer responds \n> >> with a warning if a database is dropped multiple times:\n> Sorry for interference, I'm Jose'.\n> Seems that my problem is a little bit different.\n\nI hope so :) My problem was entirely my fault; I'd left some debugging\ncode in the backend which affected most queries...\n\n> only the following commands\n> kill the backend, i.e.:\n> \n> - DELETE\n> - DROP DATABASE\n> - DROP USER\n> - LOCK TABLE\n> Any ideas ?\n\nJust guessing: is there an underlying (Unix) permissions problem with\nthe existing tables? That doesn't seem to fit exactly with the symptoms\nthough.\n\n - Tom\n",
"msg_date": "Thu, 24 Sep 1998 14:17:27 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: Problem dropping databases"
},
{
"msg_contents": "> >> Seems that my problem is a little bit different.\n> >> only the following commands\n> >> kill the backend, i.e.:\n> >>\n> >> - DELETE\n> >> - DROP DATABASE\n> >> - DROP USER\n> >> - LOCK TABLE\n> I dont' think that I have Linux permission problems, I can create a \n> new database but I can't destroy it, but I can remove the database\n> directory.\n> \n> $ createdb jose\n> $ destroydb jose\n> PQexec() -- Request was sent to backend, but backend closed the \n> channel before responding.\n> $ rm -r /usr/local/pgsql/data/base/jose/\n> Just guessing, seems this problem is connected with DELETE statement \n> because every one of the incriminate commands use it:\n> - DELETE: itself\n> - DROP DATABASE: delete from pg_database where datname = \n> 'my_db';\n> - DROP USER: delete from pg_user where usename = 'my_name';\n> - LOCK TABLE: mydb=> lock cities;\n> DELETE 0\n> Any ideas?\n\nNo. The destroydb script uses several environment variables, but so does\nthe createdb script, which is not failing for you.\n\n1) I might check the environment variables, just to be sure.\n2) Run the backend with \"-d99\" and see if the debugging output gives you\na clue\n3) Recompile the backend with \"-g -O0\" (includes symbols in the image)\nand then try running your query through gdb. It is pretty easy (now that\nBruce showed me how :):\n $ gdb postgres\n (gdb) run -D<your data path>\n > drop database jose\n <segfault message>\n (gdb) where\n\nThis will show you the backtrace of where it is crashing.\n\nHave you tried a clean install? It's a bit perplexing because we don't\nget very many reports of this kind of trouble...\n\n - Tom\n",
"msg_date": "Fri, 25 Sep 1998 14:40:13 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: Problem dropping databases"
}
] |
[
{
"msg_contents": "> Dear Bruce,\n> \n> > \n> > > At this point the Irix port is nearly complete. There are two problems\n> > > that remain. \n> > > \n> > > First, the snprintf function does not process the %lld format\n> > > correctly. Frankly, there is no reason to use it in\n> > > src/backend/utils/adt/int8.c because there is not chance of exceeding\n> > > the supplied buffer. sprintf is safe in this circumstance and works\n> > > correctly.\n> > \n> > Are you talking about the Irix snprintf, our our ports/snprintf.c?\n> \n> Irix doesn't have snprintf, so I'm talking about your ports. They\n> are missing %lld formats.\n\nOK, we need our ports snprintf() to understand %lld. Marc?\n\n-- \nBruce Momjian | 830 Blythe Avenue\[email protected] | Drexel Hill, Pennsylvania 19026\nhttp://www.op.net/~candle | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. |\n",
"msg_date": "Tue, 22 Sep 1998 12:41:59 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Results of port of Sept 18 port of PostgreSQL"
},
{
"msg_contents": "On Tue, 22 Sep 1998, Bruce Momjian wrote:\n\n> > Irix doesn't have snprintf, so I'm talking about your ports. They\n> > are missing %lld formats.\n> \n> OK, we need our ports snprintf() to understand %lld. Marc?\n\n\tAck...I just borrowed what was in sendmail...are you saying that\nsendmail doesn't work under Irix either? :(\n\n\tAh, okay, just looking at the source code itself...neat, I think\n:) How does this look? '%lld == long long', correct?\n\n*** snprintf.c.orig Tue Sep 22 22:41:46 1998\n--- snprintf.c Tue Sep 22 22:44:26 1998\n***************\n*** 126,131 ****\n--- 126,132 ----\n int ch;\n long value;\n int longflag = 0;\n+ int longlongflag = 0;\n int pointflag = 0;\n int maxwidth = 0;\n char *strvalue;\n***************\n*** 167,173 ****\n len = va_arg( args, int );\n goto nextch;\n case '.': pointflag = 1; goto nextch;\n! case 'l': longflag = 1; goto nextch;\n case 'u': case 'U':\n /*fmtnum(value,base,dosign,ljust,len,zpad) */\n if( longflag ){\n--- 168,178 ----\n len = va_arg( args, int );\n goto nextch;\n case '.': pointflag = 1; goto nextch;\n! case 'l': if(longflag) {\n! longlongflag = 1; goto nextch;\n! } else {\n! longflag = 1; goto nextch;\n! }\n case 'u': case 'U':\n /*fmtnum(value,base,dosign,ljust,len,zpad) */\n if( longflag ){\n***************\n*** 186,192 ****\n--- 191,201 ----\n fmtnum( value, 8,0, ljust, len, zpad ); break;\n case 'd': case 'D':\n if( longflag ){\n+ if( longlongflag ) {\n+ value = va_arg( args, long long );\n+ } else {\n value = va_arg( args, long );\n+ }\n } else {\n value = va_arg( args, int );\n }\n\n",
"msg_date": "Tue, 22 Sep 1998 22:46:14 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Results of port of Sept 18 port of PostgreSQL"
},
{
"msg_contents": "> On Tue, 22 Sep 1998, Bruce Momjian wrote:\n> \n> > > Irix doesn't have snprintf, so I'm talking about your ports. They\n> > > are missing %lld formats.\n> > \n> > OK, we need our ports snprintf() to understand %lld. Marc?\n> \n> \tAck...I just borrowed what was in sendmail...are you saying that\n> sendmail doesn't work under Irix either? :(\n\nNot for long long's, no.\n\n> \n> \tAh, okay, just looking at the source code itself...neat, I think\n> :) How does this look? '%lld == long long', correct?\n\nYes.\n\nSure, throw it in and let him test it.\n\n> \n> *** snprintf.c.orig Tue Sep 22 22:41:46 1998\n> --- snprintf.c Tue Sep 22 22:44:26 1998\n> ***************\n> *** 126,131 ****\n> --- 126,132 ----\n> int ch;\n> long value;\n> int longflag = 0;\n> + int longlongflag = 0;\n> int pointflag = 0;\n> int maxwidth = 0;\n> char *strvalue;\n> ***************\n> *** 167,173 ****\n> len = va_arg( args, int );\n> goto nextch;\n> case '.': pointflag = 1; goto nextch;\n> ! case 'l': longflag = 1; goto nextch;\n> case 'u': case 'U':\n> /*fmtnum(value,base,dosign,ljust,len,zpad) */\n> if( longflag ){\n> --- 168,178 ----\n> len = va_arg( args, int );\n> goto nextch;\n> case '.': pointflag = 1; goto nextch;\n> ! case 'l': if(longflag) {\n> ! longlongflag = 1; goto nextch;\n> ! } else {\n> ! longflag = 1; goto nextch;\n> ! }\n> case 'u': case 'U':\n> /*fmtnum(value,base,dosign,ljust,len,zpad) */\n> if( longflag ){\n> ***************\n> *** 186,192 ****\n> --- 191,201 ----\n> fmtnum( value, 8,0, ljust, len, zpad ); break;\n> case 'd': case 'D':\n> if( longflag ){\n> + if( longlongflag ) {\n> + value = va_arg( args, long long );\n> + } else {\n> value = va_arg( args, long );\n> + }\n> } else {\n> value = va_arg( args, int );\n> }\n> \n> \n\n\n-- \nBruce Momjian | 830 Blythe Avenue\[email protected] | Drexel Hill, Pennsylvania 19026\nhttp://www.op.net/~candle | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. |\n",
"msg_date": "Tue, 22 Sep 1998 22:32:09 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: Results of port of Sept 18 port of PostgreSQL"
},
{
"msg_contents": "On Tue, 22 Sep 1998, Bruce Momjian wrote:\n\n> > \tAh, okay, just looking at the source code itself...neat, I think\n> > :) How does this look? '%lld == long long', correct?\n> \n> Yes.\n> \n> Sure, throw it in and let him test it.\n\n\tDone...\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Wed, 23 Sep 1998 00:11:49 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Results of port of Sept 18 port of PostgreSQL"
},
{
"msg_contents": "> long value;\n> + value = va_arg( args, \n> long long );\n\nUm, no.\n\nTry \"long long value;\"\n\nJP Sugarbroad\n",
"msg_date": "Wed, 23 Sep 1998 20:09:49 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Re: Results of port of Sept 18 port of PostgreSQL"
}
] |
[
{
"msg_contents": "Is there a strlen function in sql?\n\nCan you do a multiple comparison on the same field. I would like to be\nable to find entries that have either a null entry or a blank entry as a\nboolean result and sort by the boolean result.\n\n(ie: select field='' or field is null as x from table order by x;)\n\nThe above statement core dumps. The current solution is to:\n\nupdate table set field='' where field is null; \n\nbefore running the query. There must be a better way? Ideas?\n\nThadvancenks!!!!\n\n--\nColin Dick\nOn Call Internet Services\[email protected]\n\n\n",
"msg_date": "Tue, 22 Sep 1998 10:01:30 -0700 (PDT)",
"msg_from": "Colin Dick <[email protected]>",
"msg_from_op": true,
"msg_subject": "2 questions."
},
{
"msg_contents": "ALERT - cc'd to pgsql-hackers since this looks like a bug in the parser to\nme.\n\nOn Tue, 22 Sep 1998, Colin Dick wrote:\n\n> Can you do a multiple comparison on the same field. I would like to be\n> able to find entries that have either a null entry or a blank entry as a\n> boolean result and sort by the boolean result.\n\nIn a word Yes - it looks like the parser simply doesn't understand your\nSQL and craps out on it.\n\n> (ie: select field='' or field is null as x from table order by x;)\n\nI write this as:\n\nselect field as x from table where field='' or field is null order by x;\n\nwhich works properly (even tested it on 6.3.2). Either you aren't\nsupposed to write queries as you did or the parser should figure out what\nyou are trying to do. Either way I think it's a bug. You can resolve it\nyourself by just writing the query as I did.\n\nChris\n\n",
"msg_date": "Wed, 23 Sep 1998 08:44:49 -0400 (EDT)",
"msg_from": "Chris Johnson <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [SQL] 2 questions."
},
{
"msg_contents": "Hello Colin,\n\nmarted�, 22 settembre 98, you wrote:\n\nCD> Is there a strlen function in sql?\n\nThe SQL function is:\n CHAR_LENGTH\n CHARACTER_LENGTH\n \nunfortunately it doesn't work with NULLs you have to exclude them\nlike this:\n SELECT * FROM table\n WHERE CHAR_LENGTH(field) = 0 AND field IS NOT NULL OR field IS NULL;\n\nThe above command list only the rows where \"field\" is equal '' or\nNULL.\n\nCD> Can you do a multiple comparison on the same field. I would like to be\nCD> able to find entries that have either a null entry or a blank entry as a\nCD> boolean result and sort by the boolean result.\n\nCD> (ie: select field='' or field is null as x from table order by x;)\n\nSELECT field\n FROM table\n WHERE field='' OR field IS NULL\n ORDER BY field;\n\nThis should work, except for ORDER BY because NULL's will be all the\ntime, at bottom of list even if you specify [DESC]endent order.\nI'm sorry but I don't know how to transform NULL's to '', this is also\none of my unsolved problems. We need a function like Oracle NVL() or\nANSI COALESCE() but we don't have this on PostgreSQL.\n\n-----\nPS:\n\nFor hackers only. IMHO I think we need to do something for NULLs. ;)\nWhat do you think about?\n\n Jose'\n\n\n",
"msg_date": "Wed, 23 Sep 1998 15:45:53 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [SQL] 2 questions."
},
{
"msg_contents": "> CD> I would like to...\n> CD> ... find entries that have either a null entry or a blank entry as \n> CD> a boolean result and sort by the boolean result.\n> CD> (ie: select field='' or field is null as x from table order by x;)\n> SELECT field\n> FROM table\n> WHERE field='' OR field IS NULL\n> ORDER BY field;\n\nThe problem statement isn't really clear, but Colin's prototype query\nsuggests that he wants all fields back, with a boolean column \"true\" if\nthe field is zero-length or if the field is NULL.\n\nSo I think his original proposal is the one he wants:\n\n SELECT (field = '' OR field IS NULL)\n FROM table\n ORDER by 1;\n\n> For hackers only. IMHO I think we need to do something for NULLs. ;)\n> What do you think about?\n\nThe only open issues on NULLs afaik (I'm doing this from memory, so if\nthere are other things on the ToDo don't take this as having removed\nthem :) are:\n\n1) functions taking or returning \"pass by value\" arguments, as is\ntypical for functions returning int4 (and other 4-byte or shorter data\ntypes except for float4), cannot signal that the returned value should\nactually be NULL. So they must throw an error instead. That's why\nchar_length() doesn't behave gracefully with nulls.\n\n2) NULL fields don't move from front-to-back or back-to-front when\nchanging the sort order from ascending to descending. I don't speak for\nthe Postgres team on this, but frankly I don't see this as a big issue.\nI know that SQL92 specifies that nulls *should* switch ends of the\nreturned list, but the standard also says that *which* end of the list\nthey are on for, say, ascending order, is implementation-dependent. So,\nit seems like portable code really needs to handle both cases anyway...\nimho it is a place where the standard probably should have said less\n(or, depending on your preference, more) and is inadequate as-is. If you\nare getting back a mix of nulls and values, check each returned row for\nwhether the value is null and you won't get in trouble...\n\n - Tom\n\nbtw, Thanks again Jose' for those great SQL reference pages. Oliver and\nI have got them transcribed to sgml, and they will make a nice addition\nto the v6.4 release of the docs.\n",
"msg_date": "Thu, 24 Sep 1998 02:34:24 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: [SQL] 2 questions."
},
{
"msg_contents": "Thomas G. Lockhart wrote:\n> \n> > CD> I would like to...\n> > CD> ... find entries that have either a null entry or a blank entry as\n> > CD> a boolean result and sort by the boolean result.\n> > CD> (ie: select field='' or field is null as x from table order by x;)\n> > SELECT field\n> > FROM table\n> > WHERE field='' OR field IS NULL\n> > ORDER BY field;\n> \n> The problem statement isn't really clear, but Colin's prototype query\n> suggests that he wants all fields back, with a boolean column \"true\" if\n> the field is zero-length or if the field is NULL.\n> \n> So I think his original proposal is the one he wants:\n> \n> SELECT (field = '' OR field IS NULL)\n> FROM table\n> ORDER by 1;\n> \n> > For hackers only. IMHO I think we need to do something for NULLs. ;)\n> > What do you think about?\n> \n> The only open issues on NULLs afaik (I'm doing this from memory, so if\n> there are other things on the ToDo don't take this as having removed\n> them :) are:\n> \n> 1) functions taking or returning \"pass by value\" arguments, as is\n> typical for functions returning int4 (and other 4-byte or shorter data\n> types except for float4), cannot signal that the returned value should\n> actually be NULL. So they must throw an error instead. That's why\n> char_length() doesn't behave gracefully with nulls.\n> \n> 2) NULL fields don't move from front-to-back or back-to-front when\n> changing the sort order from ascending to descending. I don't speak for\n> the Postgres team on this, but frankly I don't see this as a big issue.\n> I know that SQL92 specifies that nulls *should* switch ends of the\n> returned list, but the standard also says that *which* end of the list\n> they are on for, say, ascending order, is implementation-dependent. So,\n> it seems like portable code really needs to handle both cases anyway...\n> imho it is a place where the standard probably should have said less\n> (or, depending on your preference, more) and is inadequate as-is. If you\n> are getting back a mix of nulls and values, check each returned row for\n> whether the value is null and you won't get in trouble...\n\nYes the problem with NULLS is so big that SQL92 leaves it to\nimplementation-dependent.\n\nHowever SQL92 has some functions to help with nulls and I think we NEED\nto have\nthese on PostgreSQL.\n\nFor example, I don't know how to compute fields containing nulls,\nbecause the result\nis always null.\n\nAFAIK SQL92 has at least two functions that one can use to decide what\nto do with nulls. (i.e.:)\n\n\to CASE WHEN ... ELSE ... END\nand\n\to COALESCE()\n\nother databases have similar functions, for example Oracle has the\nfunction\nNVL() to translate nulls to an arbitrary value.\n\nI tried to write a function to emulate NVL() but it doesn't work because\nnulls\nhave a strange behavior. \n\nint nvl(int arg)\n{\n if(arg) return arg;\n return 0;\n}\n\nAny ideas?\n\n> \n> - Tom -\n> \n> btw, Thanks again Jose' for those great SQL reference pages. Oliver and\n> I have got them transcribed to sgml, and they will make a nice addition\n> to the v6.4 release of the docs.\n\nDon't mention it Tom, it was a very pleasure to me to share a bit of my\ntime \nfor this great work and I intend to co-operate again if you need help.\n \n Jose'\n",
"msg_date": "Thu, 24 Sep 1998 12:46:59 +0200",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: [SQL] 2 questions."
},
{
"msg_contents": "> > 1) functions taking or returning \"pass by value\" arguments\n> > cannot signal that the returned value should\n> > actually be NULL. So they must throw an error instead. That's why\n> > char_length() doesn't behave gracefully with nulls.\n\n1a) functions taking only \"pass by value\" arguments are never called\nwith NULL arguments, since Postgres cannot indicate to the routine that\nthere is a NULL value (it uses a null pointer to indicate this with\n\"pass by reference\" arguments, such as float8 or text).\n\n> However SQL92 has some functions to help with nulls and I think we \n> NEED to have these on PostgreSQL.\n> For example, I don't know how to compute fields containing nulls,\n> because the result is always null.\n> \n> AFAIK SQL92 has at least two functions that one can use to decide what\n> to do with nulls. (i.e.:)\n> o CASE WHEN ... ELSE ... END\n> and\n> o COALESCE()\n> other databases have similar functions, for example Oracle has the\n> function NVL() to translate nulls to an arbitrary value.\n\nSo Oracle does not have the CASE construct? That seems to be a rich area\nfor work in v6.5...\n\n> I tried to write a function to emulate NVL() but it doesn't work \n> because nulls have a strange behavior.\n> \n> int nvl(int arg)\n> {\n> if(arg) return arg;\n> return 0;\n> }\n> \n> Any ideas?\n\nYou are running into the \"pass by value\" item (1a) above. And anyway,\nyou would not be able to do what you want since there is currently no\nway to signal back that the input is NULL rather than zero.\n\nIn the long run, we would need to either convert all user-accessible\ntypes to be \"pass by reference\", or would need to implement another\nmechanism for signalling NULLness as a return condition (perhaps with a\nglobal variable since each Postgres backend is single-threaded).\n\nThis would be a good topic to revisit after v6.4 is released...\n\n - Tom\n",
"msg_date": "Thu, 24 Sep 1998 13:50:22 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: [SQL] 2 questions."
}
] |
[
{
"msg_contents": "I have written a function that will return an integer based on the length\nof a given field defined as text.\nIt seems that if the given field has not yet been populated (ie: it is\nnull), my function returns nothing. I have tried expressly sending back a\nvalue as such to no avail:\n\nint isempty(text*)\n{\n return (100);\n}\n\nIf the text* is null, it does not return 100 though it does seem to\nprocess the function and does not core dump. (ie: returns blank entries\nwith the proper number of tuples at the end). Any ideas? Thanks again.\n\n--\nColin Dick\nOn Call Internet Services\[email protected]\n\n\n",
"msg_date": "Tue, 22 Sep 1998 10:37:13 -0700 (PDT)",
"msg_from": "Colin Dick <[email protected]>",
"msg_from_op": true,
"msg_subject": "Question re: null"
},
{
"msg_contents": "Hello Colin,\n\nmarted�, 22 settembre 98, you wrote:\n\nCD> I have written a function that will return an integer based on the length\nCD> of a given field defined as text.\nCD> It seems that if the given field has not yet been populated (ie: it is\nCD> null), my function returns nothing. I have tried expressly sending back a\nCD> value as such to no avail:\n\nCD> int isempty(text*)\nCD> {\nCD> return (100);\nCD> }\n\nCD> If the text* is null, it does not return 100 though it does seem to\nCD> process the function and does not core dump. (ie: returns blank entries\nCD> with the proper number of tuples at the end). Any ideas? Thanks again.\n\nI'm very interesting on this subject too.\nDoes anybody knows how to solve this question ?\nI can't belive that we are the only people having problems\nwith NULLs. ;)\n\n Jose'\n\n\n",
"msg_date": "Wed, 23 Sep 1998 17:25:52 +0200",
"msg_from": "Sferacarta Software <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [SQL] Question re: null"
}
] |
[
{
"msg_contents": "Here is a comment from primnodes.h. I don't know who 'sp' is, but I\nwould like to meet him.\n\n---------------------------------------------------------------------------\n\n * ----\n * NOTE: in the good old days 'opno' used to be both (or either, or\n * neither) the pg_operator oid, and/or the pg_proc oid depending\n * on the postgres module in question (parser->pg_operator,\n * executor->pg_proc, planner->both), the mood of the programmer,\n * and the phase of the moon (rumors that it was also depending on the day\n * of the week are probably false). To make things even more postgres-like\n * (i.e. a mess) some comments were referring to 'opno' using the name\n * 'opid'. Anyway, now we have two separate fields, and of course that\n * immediately removes all bugs from the code... [ sp :-) ].\n * ----------------\n */\ntypedef struct Oper\n{\n NodeTag type;\n\n-- \nBruce Momjian | 830 Blythe Avenue\[email protected] | Drexel Hill, Pennsylvania 19026\nhttp://www.op.net/~candle | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. |\n",
"msg_date": "Tue, 22 Sep 1998 16:45:25 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "funny comment"
}
] |
[
{
"msg_contents": "I just scanned through the TODO list in the actual source tree. It seems\nthis one isn't up-to-date since the ecpg entry is still old. Also I found\nthis bug:\n\n* select upper(usename), count(usesysid) from pg_shadow group by 1 fails\n\nHowever, actual version says:\n\nmm=> select upper(usename), count(usesysid) from pg_shadow group by 1;\nupper |count\n--------+-----\nMESKES | 1\nPOSTGRES| 1\n(2 rows)\n\nSeems to be okay, doesn't it?\n\nMichael\n\n-- \nDr. Michael Meskes | Th.-Heuss-Str. 61, D-41812 Erkelenz | Go SF49ers!\nSenior-Consultant | business: [email protected] | Go Rhein Fire!\nMummert+Partner | private: [email protected] | Use Debian\nUnternehmensberatung AG | [email protected] | GNU/Linux!\n",
"msg_date": "Wed, 23 Sep 1998 09:12:19 +0200",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "Problem on TODO list"
},
{
"msg_contents": "> I just scanned through the TODO list in the actual source tree. It seems\n> this one isn't up-to-date since the ecpg entry is still old. Also I found\n> this bug:\n\nI have removed the actual version numbers of ecpg from the TODO list.\n\n> \n> * select upper(usename), count(usesysid) from pg_shadow group by 1 fails\n> \n> However, actual version says:\n> \n> mm=> select upper(usename), count(usesysid) from pg_shadow group by 1;\n> upper |count\n> --------+-----\n> MESKES | 1\n> POSTGRES| 1\n> (2 rows)\n> \n> Seems to be okay, doesn't it?\n\nYep. Removed.\n\nThe current TODO list is always on our web site. The distribution gets\nupdated as part of the start of beta, and the final release.\n\n\n-- \n Bruce Momjian | [email protected]\n 830 Blythe Avenue | http://www.op.net/~candle\n Drexel Hill, Pennsylvania 19026 | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. | \n",
"msg_date": "Wed, 23 Sep 1998 13:48:50 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Problem on TODO list"
},
{
"msg_contents": "On Wed, Sep 23, 1998 at 01:48:50PM -0400, Bruce Momjian wrote:\n> I have removed the actual version numbers of ecpg from the TODO list.\n\nI know. That's why I though this TODO was old. :-)\n\n> Yep. Removed.\n> \n> The current TODO list is always on our web site. The distribution gets\n> updated as part of the start of beta, and the final release.\n\nI see. I will download that one and do some more checking.\n\nMichael\n-- \nDr. Michael Meskes | Th.-Heuss-Str. 61, D-41812 Erkelenz | Go SF49ers!\nSenior-Consultant | business: [email protected] | Go Rhein Fire!\nMummert+Partner | private: [email protected] | Use Debian\nUnternehmensberatung AG | [email protected] | GNU/Linux!\n",
"msg_date": "Wed, 23 Sep 1998 20:20:10 +0200",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Problem on TODO list"
}
] |
[
{
"msg_contents": "Hi All,\n\nI can't confirm this against the current latest CVS as my \nlocal tree has undergone some changes, but we seem to have\na problem with \"pg_dump\" and user defined types.\n\nIf I create a type like so:-\n\nCREATE FUNCTION sotime_in (opaque ) RETURNS opaque AS\n'/usr/local/pgsql/shlibs/obj/sotimes.so' LANGUAGE 'C';\n\nCREATE FUNCTION sotime_out (opaque ) RETURNS opaque AS\n'/usr/local/pgsql/shlibs/obj/sotimes.so' LANGUAGE 'C';\n\nCREATE TYPE sotime ( internallength = 2, externallength = 5, \ninput = sotime_in, output = sotime_out, send = sotime_out, \nreceive = sotime_in, default = '-',passedbyvalue); \n\npg_dump gives me back :-\n\nCREATE TYPE sotime ( internallength = 2, externallength = 5, \ninput = sotime_in_18272, output = sotime_out_18304, send = sotime_out_18304, \nreceive = sotime_in_18272, default = '-',passedbyvalue);\n\nCREATE TYPE _sotime ( internallength = -1, externallength = -1, \ninput = array_in_750, output = array_out_751, send = array_out_751, \nreceive = array_in_750, default = '-'); \n.\n<table creates>\n.\nCREATE FUNCTION sotime_in (opaque ) RETURNS opaque AS\n'/usr/local/pgsql/shlibs/obj/sotimes.so' LANGUAGE 'C';\nCREATE FUNCTION sotime_out (opaque ) RETURNS opaque AS\n'/usr/local/pgsql/shlibs/obj/sotimes.so' LANGUAGE 'C';\n\n\nLooks like the FUNCTION creates are OK but the TYPE creation\nis a little confused.\n\nI remember some discussion about name mangling for the functions\nbut it looks like pg_dump can't cope with this.\n\nAnother thing I can't understand is pg_dump dumping the _sotime\ntype, which seems to be created automatically when I create\nthe sotime type.\n\nNeedless to say, I can't dump/restore a database with user defined\ntypes at all.\n\nKeith.\n\nBTW: I was looking as pg_dump'ing VIEWS.\n\n",
"msg_date": "Wed, 23 Sep 1998 11:01:47 +0100 (BST)",
"msg_from": "Keith Parks <[email protected]>",
"msg_from_op": true,
"msg_subject": "pg_dump, problem with user defined types?"
},
{
"msg_contents": "> Hi All,\n> \n> I can't confirm this against the current latest CVS as my \n> local tree has undergone some changes, but we seem to have\n> a problem with \"pg_dump\" and user defined types.\n> \n> If I create a type like so:-\n> \n> CREATE FUNCTION sotime_in (opaque ) RETURNS opaque AS\n> '/usr/local/pgsql/shlibs/obj/sotimes.so' LANGUAGE 'C';\n> \n> CREATE FUNCTION sotime_out (opaque ) RETURNS opaque AS\n> '/usr/local/pgsql/shlibs/obj/sotimes.so' LANGUAGE 'C';\n> \n> CREATE TYPE sotime ( internallength = 2, externallength = 5, \n> input = sotime_in, output = sotime_out, send = sotime_out, \n> receive = sotime_in, default = '-',passedbyvalue); \n> \n> pg_dump gives me back :-\n> \n> CREATE TYPE sotime ( internallength = 2, externallength = 5, \n> input = sotime_in_18272, output = sotime_out_18304, send = sotime_out_18304, \n> receive = sotime_in_18272, default = '-',passedbyvalue);\n> \n> CREATE TYPE _sotime ( internallength = -1, externallength = -1, \n> input = array_in_750, output = array_out_751, send = array_out_751, \n> receive = array_in_750, default = '-'); \n> .\n> <table creates>\n> .\n> CREATE FUNCTION sotime_in (opaque ) RETURNS opaque AS\n> '/usr/local/pgsql/shlibs/obj/sotimes.so' LANGUAGE 'C';\n> CREATE FUNCTION sotime_out (opaque ) RETURNS opaque AS\n> '/usr/local/pgsql/shlibs/obj/sotimes.so' LANGUAGE 'C';\n> \n> \n> Looks like the FUNCTION creates are OK but the TYPE creation\n> is a little confused.\n> \n> I remember some discussion about name mangling for the functions\n> but it looks like pg_dump can't cope with this.\n> \n> Another thing I can't understand is pg_dump dumping the _sotime\n> type, which seems to be created automatically when I create\n> the sotime type.\n> \n> Needless to say, I can't dump/restore a database with user defined\n> types at all.\n\nYes, we have a problem with user-defined types because of the changes I\nmade to make the function name unique by adding the pg_proc oid to the\nname.\n\nI can make the fix, but no one seems to have an opinion on how yet. I\ncan e-mail the pertinent postings if you want to discuss solutions.\n\n-- \n Bruce Momjian | [email protected]\n 830 Blythe Avenue | http://www.op.net/~candle\n Drexel Hill, Pennsylvania 19026 | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. | \n",
"msg_date": "Wed, 23 Sep 1998 12:46:00 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
},
{
"msg_contents": "> > pg_dump gives me back :-\n> > CREATE TYPE sotime ( internallength = 2, externallength = 5,\n> > input = sotime_in_18272, output = sotime_out_18304, send = sotime_out_18304,\n> > receive = sotime_in_18272, default = '-',passedbyvalue);\n> > CREATE TYPE _sotime ( internallength = -1, externallength = -1,\n> > input = array_in_750, output = array_out_751, send = array_out_751,\n> > receive = array_in_750, default = '-');\n> Yes, we have a problem with user-defined types because of the changes \n> I made to make the function name unique by adding the pg_proc oid to \n> the name.\n> I can make the fix, but no one seems to have an opinion on how yet.\n\nIt would seem that for Keith's cases things worked OK originally? There\nwas trouble with other overloaded functions?\n\nAt worst, we should revert to the non-mangled names. On a related note,\nhas anyone had time to try testing pg_upgrade? That would be a nice\nfeature for v6.4 if it worked on a wide range of databases (haven't\ntried it myself yet).\n\n - Tom\n",
"msg_date": "Thu, 24 Sep 1998 02:57:37 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
}
] |
[
{
"msg_contents": "I just noticed that SQL3 has recursive unions but couldn't find an example\nfor it, just the theoretical definitions. Does anyone have such an example,\njust to get a better feel for the definition?\n\nMichael\n-- \nDr. Michael Meskes | Th.-Heuss-Str. 61, D-41812 Erkelenz | Go SF49ers!\nSenior-Consultant | business: [email protected] | Go Rhein Fire!\nMummert+Partner | private: [email protected] | Use Debian\nUnternehmensberatung AG | [email protected] | GNU/Linux!\n",
"msg_date": "Wed, 23 Sep 1998 13:57:36 +0200",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "Recursive Unions"
}
] |
[
{
"msg_contents": "i would set the default rights of new users.\nFor example a new user is able to create class and i want to forbidden\nthis. \ni try to revoke all on pg_class but it doesn't take effect.\nthanks. \n\n\n--\na+\nmaloune\n",
"msg_date": "Wed, 23 Sep 1998 16:27:41 +0200",
"msg_from": "maloune <[email protected]>",
"msg_from_op": true,
"msg_subject": "default rights of new users"
}
] |
[
{
"msg_contents": " ---- you wrote: \n> On Wed, Sep 16, 1998 at 06:35:53PM -0300, Robson Miranda wrote:\n> > \tI was thinking in a major rewrite of the PostrgreSQL transaction\n> > system, in order to provide less tuple overhead and recoverabilty.\n\n> I do not have much of an idea how postgres handles stuff right now, so\n> forgive me if I'm asking stupid questions.\nOk...\n\n> > \tMy first goal is to reduce tuple overhead, getting rid of xmin/xman and\n> > cmin/cmax. To provide this functionality, I'm planning to keep only a\n> > flag indicating if the transaction is in curse or not. If, during a\n> > transaction, a certain tuple is affected, this flag will store the\n> > current transaction id. Thus, if this tuple is commited, an invalid OID\n> > (say, 0), will be written to this flag.\n\n> That means you store one flag per tuple? Does this happen only in memory?\n\nThere will be something to indicate that the tuple is valid or not. An NULL, for example, can indicate if this tuple is part of a commited transaction. Any other number indicates that this is an in-progress transaction (an valid transaction ID). I was thinking in use a unsigned 31 bit for transaction ID and the other bit to flag if this tuple is valid or not. Thus, an NULL indicates that this line was inserted by a commited xaction, and a number with only the 32nd bit setted will indicate the this line was deleted by a commited xaction. \nThis way, if the xaction rollbacks, the lines inserted will have only the 32nd bit setted, and the delted lines will have NULL in this value. But, at commit or rollback, the pages with deleted tuples will have this space added to the free space of the page.\nActualy, there are 2 fields (xmin/xmax) and 4 flags (HEAP_XMIN_COMMITTED, HEAP_XMAX_COMMITTED, HEAP_XMIN_INVALID, HEAP_XMAX_INVALID) that are used for determine the visibility of a tuple. If a tuple doesn't have these flags setted, then a test in log relation will be done to determine the correct status of the transaction and then the flags will be adjusted. These flags operates like a \"cache\" for the transaction status. Look that these flags can be changed during selects, and these pages will have to be written to disk after a query.\n\n> > \tThe only problem a saw using this approach is if some pages got flushed\n> > during the transaction, because these pages will have to be reload from> disk.\n\n> Ah yes, it seems to be in memory only. And you exactly point to one problem.\n> Any idea how to solve this?\n\nAfter looking more deepely in the current transaction model, I think my approach will have a similar performance, since these pages will have to be reload only one time (at the end of transaction). Currently, the log pages could be loaded from disk even during a select.\n\n> > \tTo keep track of current transactions, there will have a list of tuples\n> > affected by this transaction, and the operation executed. This way,\n> > during commit, we only confirm these operations in relations (writing an\n> > invalid OID in current xid of each tuple affected). To rollback, we\n> > delete the new tuples (and mark this operation as a commit) and mark the\n> > old tuples affected as \"live\" (and leave these commited).\n\n> That means we always have both in the relation? That is we write the new\n> tuple in and keep the old one? Is this done the same way in the actual\n> version? I'd prefer to have a clean cut with new and old not being in the\n> same table at the same time. \nNo, the invalid tuples will be deleted at the end of transaction, freeing the space on the page. I'm thinking in keep the page free space in some special blocks inside the table. With pagesize of 8k, I can keep one table of free-space at every 4096 blocks. This way, each table will have the free-space entrys for the 4096 next blocks.\n\n> > \tFor recovery (my second goal), I pretend to, at startup of postmaster,\n> > to rollback all marked in-curse transactions. After that, I'm thinking\n> > about a redo log, but I'm still searching a way to keep it with the\n> > minimum size possible. \n\nWhere's the problem with a redo log?\n\nI'm still researching how store this information, in order to, during recover, do the right things in the right time, and to provides on-line backup capability.\n\n\n> \n> Michael\n> \n\n Robson.\n\n\n---------------------------------------------------\nGet free personalized email at http://www.iname.com\n",
"msg_date": "Wed, 23 Sep 1998 10:45:54 -0400 (EDT)",
"msg_from": "[email protected]",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Transaction system (proposal for 6.5)"
}
] |
[
{
"msg_contents": "Bruce,\n\nIf you have the relevant discussions in a neat bundle I'd like\nto have a read through them to see what the issues are.\n\nThe other curiosity is the creation of an _<typname> type, which\nI believe is an array type, at the time of type creation.\n\nI can't remember seeing this in previous versions.\n\nThanks,\nKeith.\n\n> > Needless to say, I can't dump/restore a database with user defined\n> > types at all.\n> \n> Yes, we have a problem with user-defined types because of the changes I\n> made to make the function name unique by adding the pg_proc oid to the\n> name.\n> \n> I can make the fix, but no one seems to have an opinion on how yet. I\n> can e-mail the pertinent postings if you want to discuss solutions.\n\n\n",
"msg_date": "Wed, 23 Sep 1998 22:01:30 +0100 (BST)",
"msg_from": "Keith Parks <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
},
{
"msg_contents": "> Bruce,\n> \n> If you have the relevant discussions in a neat bundle I'd like\n> to have a read through them to see what the issues are.\n> \n> The other curiosity is the creation of an _<typname> type, which\n> I believe is an array type, at the time of type creation.\n> \n> I can't remember seeing this in previous versions.\n> \n> Thanks,\n> Keith.\n\nOK, here are the relivant postings. Please make a suggestion for a fix.\n\n\n-- \n Bruce Momjian | [email protected]\n 830 Blythe Avenue | http://www.op.net/~candle\n Drexel Hill, Pennsylvania 19026 | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. |",
"msg_date": "Wed, 23 Sep 1998 17:18:23 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
}
] |
[
{
"msg_contents": "I've got a production application (currently running on 6.3.2 + Btree\npatch) in which several clients concurrently update the same set of\ntables. The client requests are pretty simple --- SELECTs using\nbtree-indexed fields, INSERTs, and UPDATEs. Updates are always of the\nform \"UPDATE table SET some-fields WHERE id = constant\", where id\nis a unique identifier field with a btree index. (Some of the tables\nuse OID for this, some have an ID field generated off a SEQUENCE\nobject.) We never delete any tuples. We use BEGIN TRANSACTION / LOCK /\nEND TRANSACTION to protect groups of related updates.\n\nWhat we're seeing is that under load, the tables occasionally get\ncorrupted by duplication of tuples. It's clearly a backend bug, because\nthe duplicates are duplicate right down to the OID. There's no way that\na client could request creation of two tuples with the same OID, right?\n\nI speculate that the backend is updating a tuple (or rearranging a page\nto update another one) and neglecting to mark the old copy dead.\nI don't know whether having multiple backends running in parallel is\nnecessary to cause the bug, but it seems likely --- if this kind of\nthing happened with only one backend, surely it'd have been fixed long\nago.\n\nThe interesting thing is that both tuples are getting inserted into\nthe table's indexes, because the way we generally find out about it\nis that a client fails with \"Cannot insert a duplicate key into a unique\nindex\" when it tries to UPDATE one of the duplicated tuples. Since the\nUPDATE is \"... WHERE id = constant\", presumably this indicates that both\ntuples are getting found by the index search. (But there are duplicate\ntuples, not duplicate index records pointing at the same tuple, because\nI see the duplicates if I just dump out the table with COPY. Also,\nonce or twice I have seen near-duplicates in which the OID is the same\nbut one or two of the other fields disagree. Possibly these were formed\nat the instant of updating the original tuple to modify those fields,\nwith the original tuple not getting cleared?)\n\nI rooted through the archives and found mention of this sort of thing\nfrom last year sometime, as well as a current thread on pgsql-admin\nthat looks like it might be related.\n\nCan anyone give me guidance on dealing with this? Any chance that a\nfix is already present in 6.4-beta? I'm going to start out by trying\nto develop a repeatable test case, but I don't really know enough about\nthe backend innards to debug it competently...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Wed, 23 Sep 1998 17:04:21 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Yipes, I'm getting bit by duplicate tuples"
},
{
"msg_contents": "> I've got a production application (currently running on 6.3.2 + Btree\n> patch) in which several clients concurrently update the same set of\n> tables. The client requests are pretty simple --- SELECTs using\n> btree-indexed fields, INSERTs, and UPDATEs. Updates are always of the\n> form \"UPDATE table SET some-fields WHERE id = constant\", where id\n> is a unique identifier field with a btree index. (Some of the tables\n> use OID for this, some have an ID field generated off a SEQUENCE\n> object.) We never delete any tuples. We use BEGIN TRANSACTION / LOCK /\n> END TRANSACTION to protect groups of related updates.\n> \n> What we're seeing is that under load, the tables occasionally get\n> corrupted by duplication of tuples. It's clearly a backend bug, because\n> the duplicates are duplicate right down to the OID. There's no way that\n> a client could request creation of two tuples with the same OID, right?\n> \n> I speculate that the backend is updating a tuple (or rearranging a page\n> to update another one) and neglecting to mark the old copy dead.\n> I don't know whether having multiple backends running in parallel is\n> necessary to cause the bug, but it seems likely --- if this kind of\n> thing happened with only one backend, surely it'd have been fixed long\n> ago.\n> \n> The interesting thing is that both tuples are getting inserted into\n> the table's indexes, because the way we generally find out about it\n> is that a client fails with \"Cannot insert a duplicate key into a unique\n> index\" when it tries to UPDATE one of the duplicated tuples. Since the\n> UPDATE is \"... WHERE id = constant\", presumably this indicates that both\n> tuples are getting found by the index search. (But there are duplicate\n> tuples, not duplicate index records pointing at the same tuple, because\n> I see the duplicates if I just dump out the table with COPY. Also,\n> once or twice I have seen near-duplicates in which the OID is the same\n> but one or two of the other fields disagree. Possibly these were formed\n> at the instant of updating the original tuple to modify those fields,\n> with the original tuple not getting cleared?)\n> \n> I rooted through the archives and found mention of this sort of thing\n> from last year sometime, as well as a current thread on pgsql-admin\n> that looks like it might be related.\n> \n> Can anyone give me guidance on dealing with this? Any chance that a\n> fix is already present in 6.4-beta? I'm going to start out by trying\n> to develop a repeatable test case, but I don't really know enough about\n> the backend innards to debug it competently...\n\nI fixed some very strange indexing code that could have been the cause. \nIt was one of those, \"Gee, wonder why things are working with this\ncode.\"\n\nIt is part of 6.4, so it may already be fixed. Also, the locking of\nbuffer pages in the pre-6.4 code was kind of messed up, so it is\npossible locks were not being kept. I believe to have fixed that too,\nso running it though a test with 6.4 may show it is fixed.\n\n-- \n Bruce Momjian | [email protected]\n 830 Blythe Avenue | http://www.op.net/~candle\n Drexel Hill, Pennsylvania 19026 | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. | \n",
"msg_date": "Wed, 23 Sep 1998 18:21:44 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Yipes, I'm getting bit by duplicate tuples"
},
{
"msg_contents": "The good news: after much pain, I was able to develop a semi-repeatable\ntest for that duplicate-tuple problem. (It's a script of about 6300 SQL\ncommands for five concurrent backends, which takes about 10 minutes to\nrun and produces a corrupted table about one run out of two...) And\nthis test indicates that the current CVS sources don't have the bug.\nSo, rather than expending a lot of effort trying to figure out just\nwhat the bug is in 6.3.2, we are going to cross our fingers and put our\nproduction application on 6.4beta.\n\nThe bad news: this same script exposes a different bug in the current\nsources (and perhaps older releases too). *Very* rarely, like less\nthan one run out of ten, the test driver gets wedged or fails with an\n\"out of memory\" error. I just traced this to its cause, and the cause\nis that a SELECT reply coming from the backend is corrupt. In fact,\nwhat I see in libpq's input buffer is that a \"NOTIFY\" message has been\ninserted into the middle of the tuple data :-(. So the interlock that\nsupposedly prevents Async_NotifyFrontEnd() from being invoked during\nanother command does not work reliably.\n\nI will look into this, but I could use advice from anyone who\nunderstands how that stuff is supposed to work.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sat, 26 Sep 1998 17:18:07 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "NOTIFY interlock broken (was Yipes,\n\tI'm getting bit by duplicate tuples)"
},
{
"msg_contents": "> The good news: after much pain, I was able to develop a semi-repeatable\n> test for that duplicate-tuple problem. (It's a script of about 6300 SQL\n> commands for five concurrent backends, which takes about 10 minutes to\n> run and produces a corrupted table about one run out of two...) And\n> this test indicates that the current CVS sources don't have the bug.\n> So, rather than expending a lot of effort trying to figure out just\n> what the bug is in 6.3.2, we are going to cross our fingers and put our\n> production application on 6.4beta.\n\nThat's good news.\n\n> The bad news: this same script exposes a different bug in the current\n> sources (and perhaps older releases too). *Very* rarely, like less\n> than one run out of ten, the test driver gets wedged or fails with an\n> \"out of memory\" error. I just traced this to its cause, and the cause\n> is that a SELECT reply coming from the backend is corrupt. In fact,\n> what I see in libpq's input buffer is that a \"NOTIFY\" message has been\n> inserted into the middle of the tuple data :-(. So the interlock that\n> supposedly prevents Async_NotifyFrontEnd() from being invoked during\n> another command does not work reliably.\n> \n> I will look into this, but I could use advice from anyone who\n> understands how that stuff is supposed to work.\n\nTom, I can't think of anyone who understands it better than you.\n\nHowever, if you find something in the backend and need help, let me\nknow.\n\nI will be on vacation from Sunday until Wednesday. \n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sat, 26 Sep 1998 17:53:42 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] NOTIFY interlock broken (was Yipes,\n\tI'm getting bit by duplicate tuples)"
}
] |
[
{
"msg_contents": "Bruce,\n\nWhat I can't see from the postings, is what incident or problem\nprompted the change.\n\nWas there test or query that failed because it chose the wrong\nprocedure?\n\nDo you have an example which failed with the old code and needed\nthe changes to make it work?\n\nI see pg_proc has \"pronargs\" and \"proargtypes\" columns, is this\nnot enough to ensure that the correct procedure is called?\n\nOr am I off at a tangent again!!\n\nKeith.\n\n> Bruce Momjian <[email protected]>\n> > Bruce,\n> > \n> > If you have the relevant discussions in a neat bundle I'd like\n> > to have a read through them to see what the issues are.\n> > \n> > The other curiosity is the creation of an _<typname> type, which\n> > I believe is an array type, at the time of type creation.\n> > \n> > I can't remember seeing this in previous versions.\n> > \n> > Thanks,\n> > Keith.\n> \n> OK, here are the relivant postings. Please make a suggestion for a fix.\n> \n\n",
"msg_date": "Thu, 24 Sep 1998 00:20:04 +0100 (BST)",
"msg_from": "Keith Parks <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
},
{
"msg_contents": "> Bruce,\n> \n> What I can't see from the postings, is what incident or problem\n> prompted the change.\n> \n> Was there test or query that failed because it chose the wrong\n> procedure?\n> \n> Do you have an example which failed with the old code and needed\n> the changes to make it work?\n> \n> I see pg_proc has \"pronargs\" and \"proargtypes\" columns, is this\n> not enough to ensure that the correct procedure is called?\n> \n> Or am I off at a tangent again!!\n\nI found that the regprocin routine was doing a sequential scan of\npg_proc, looking for a function matching the supplied name, and using\nthe first matching entry of pg_proc.proname.\n\nTwo problems: You can't use the system cache to look up the pg_proc\nvalue, because the cache only does unique lookups. Second, we have many\nfunctions that have multiple entries in the pg_proc table with the same\nname, but different arguments, so it was not really accurate.\n\nWe certainly need to change what I have done, but I am not sure how to\nchange it.\n\nWe can put it back to the old code, or move everything to use only an\noid, with no name, or we can somehow allow the user to supply the\npg_proc.proname, and the argument types, and do a match that way. We\ncould allow just the proname if there is only one entry with that\nproname. (Sequtial scan required, but not too bad. We could even use\nthe existing index. It is not done that much.) If it is not unique, we\nwould require the oid.\n\nI hope there is some good solution.\n\nThe code currently supports input of oid, or proname_oid, and outputs\nproname_oid.\n\n-- \n Bruce Momjian | [email protected]\n 830 Blythe Avenue | http://www.op.net/~candle\n Drexel Hill, Pennsylvania 19026 | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. | \n",
"msg_date": "Wed, 23 Sep 1998 22:10:23 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
},
{
"msg_contents": "> > What I can't see from the postings, is what incident or problem\n> > prompted the change.\n> > Was there test or query that failed because it chose the wrong\n> > procedure?\n> > Do you have an example which failed with the old code and needed\n> > the changes to make it work?\n> I found that the regprocin routine was doing a sequential scan of\n> pg_proc, looking for a function matching the supplied name, and using\n> the first matching entry of pg_proc.proname.\n> Two problems: You can't use the system cache to look up the pg_proc\n> value, because the cache only does unique lookups. Second, we have \n> many functions that have multiple entries in the pg_proc table with \n> the same name, but different arguments, so it was not really accurate.\n> We certainly need to change what I have done, but I am not sure how to\n> change it.\n\nI'm sorry Bruce for not keeping up, but could you please send me the old\npostings which summarize the symptoms of the problem? And any private\nresponse to Keith which answered his initial questions above?\n\nI understand that the regprocin support routine was apparently not using\nenough keys in the pg_proc lookup to find the unique entry match, but\nwhere is regprocin used? I assume that it is support code, but is it\nused internally only, or are you using it for some of the psql/pg_dump\nconstruction?\n\n>From the code used in the parser to support functions and operators, I\nwould have guessed that you can define multiple key fields for cache\nlookups; do those techniques fail in this case or are they not being\nused?\n\nAnyway, I'm concerned about the name mangling and the proximity to the\nv6.4 release, and would like to help work out the issues if I can...\n\n - Tom\n",
"msg_date": "Thu, 24 Sep 1998 06:03:56 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
},
{
"msg_contents": "For varchar(n)/char(n) type, input string is silently truncated if it\nis longer than n. A multi-byte letter consists of several bytes and\nthey should not be divided into pieces. Unconditional truncating\nmulti-byte letters would make partial multi-byte bytes.\n\nAttached patches should fix the problem.\n\nIndex: backend/utils/adt/varchar.c\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/backend/utils/adt/varchar.c,v\nretrieving revision 1.39\ndiff -c -r1.39 varchar.c\n*** varchar.c\t1998/09/01 04:32:53\t1.39\n--- varchar.c\t1998/09/24 09:03:37\n***************\n*** 147,153 ****\n--- 147,160 ----\n \tif ((len == -1) || (len == VARSIZE(s)))\n \t\treturn s;\n \n+ #ifdef MULTIBYTE\n+ \t/* truncate multi-byte string in a way not to break\n+ \t multi-byte boundary */\n+ \trlen = pg_mbcliplen(VARDATA(s), len - VARHDRSZ, len - VARHDRSZ);\n+ \tlen = rlen + VARHDRSZ;\n+ #else\n \trlen = len - VARHDRSZ;\n+ #endif\n \n \tif (rlen > 4096)\n \t\telog(ERROR, \"bpchar: length of char() must be less than 4096\");\n***************\n*** 367,373 ****\n--- 374,387 ----\n \n \t/* only reach here if we need to truncate string... */\n \n+ #ifdef MULTIBYTE\n+ \t/* truncate multi-byte string in a way not to break\n+ \t multi-byte boundary */\n+ \tlen = pg_mbcliplen(VARDATA(s), slen - VARHDRSZ, slen - VARHDRSZ);\n+ \tslen = len + VARHDRSZ;\n+ #else\n \tlen = slen - VARHDRSZ;\n+ #endif\n \n \tif (len > 4096)\n \t\telog(ERROR, \"varchar: length of varchar() must be less than 4096\");\nIndex: backend/utils/mb/mbutils.c\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/backend/utils/mb/mbutils.c,v\nretrieving revision 1.3\ndiff -c -r1.3 mbutils.c\n*** mbutils.c\t1998/09/01 04:33:22\t1.3\n--- mbutils.c\t1998/09/24 09:03:38\n***************\n*** 202,207 ****\n--- 202,235 ----\n }\n \n /*\n+ * returns the length of a multi-byte string\n+ * (not necessarily NULL terminated)\n+ * that is not longer than limit.\n+ * this function does not break multi-byte word boundary.\n+ */\n+ int\n+ pg_mbcliplen(const unsigned char *mbstr, int len, int limit)\n+ {\n+ \tint\t\t\tclen = 0;\n+ \tint\t\t\tl;\n+ \n+ \twhile (*mbstr && len > 0)\n+ \t{\n+ \t\tl = pg_mblen(mbstr);\n+ \t\tif ((clen + l) > limit) {\n+ \t\t\tbreak;\n+ \t\t}\n+ \t\tclen += l;\n+ \t\tif (clen == limit) {\n+ \t\t\tbreak;\n+ \t\t}\n+ \t\tlen -= l;\n+ \t\tmbstr += l;\n+ \t}\n+ \treturn (clen);\n+ }\n+ \n+ /*\n * fuctions for utils/init\n */\n static int\tDatabaseEncoding = MULTIBYTE;\nIndex: include/mb/pg_wchar.h\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/include/mb/pg_wchar.h,v\nretrieving revision 1.4\ndiff -c -r1.4 pg_wchar.h\n*** pg_wchar.h\t1998/09/01 04:36:34\t1.4\n--- pg_wchar.h\t1998/09/24 09:03:42\n***************\n*** 103,108 ****\n--- 103,109 ----\n extern int\tpg_mic_mblen(const unsigned char *);\n extern int\tpg_mbstrlen(const unsigned char *);\n extern int\tpg_mbstrlen_with_len(const unsigned char *, int);\n+ extern int\tpg_mbcliplen(const unsigned char *, int, int);\n extern pg_encoding_conv_tbl *pg_get_encent_by_encoding(int);\n extern bool show_client_encoding(void);\n extern bool reset_client_encoding(void);\n",
"msg_date": "Thu, 24 Sep 1998 18:20:08 +0900",
"msg_from": "Tatsuo Ishii <[email protected]>",
"msg_from_op": false,
"msg_subject": "fix for multi-byte partial truncating"
},
{
"msg_contents": "Applied, but for some reason patch did not like the normal cvs/rcs diff\nformat. Not sure why. Please check to see it is OK. Looks OK here.\n\n\n> For varchar(n)/char(n) type, input string is silently truncated if it\n> is longer than n. A multi-byte letter consists of several bytes and\n> they should not be divided into pieces. Unconditional truncating\n> multi-byte letters would make partial multi-byte bytes.\n> \n> Attached patches should fix the problem.\n> \n> Index: backend/utils/adt/varchar.c\n> ===================================================================\n> RCS file: /usr/local/cvsroot/pgsql/src/backend/utils/adt/varchar.c,v\n> retrieving revision 1.39\n> diff -c -r1.39 varchar.c\n> *** varchar.c\t1998/09/01 04:32:53\t1.39\n> --- varchar.c\t1998/09/24 09:03:37\n> ***************\n> *** 147,153 ****\n> --- 147,160 ----\n> \tif ((len == -1) || (len == VARSIZE(s)))\n> \t\treturn s;\n> \n> + #ifdef MULTIBYTE\n> + \t/* truncate multi-byte string in a way not to break\n> + \t multi-byte boundary */\n> + \trlen = pg_mbcliplen(VARDATA(s), len - VARHDRSZ, len - VARHDRSZ);\n> + \tlen = rlen + VARHDRSZ;\n> + #else\n> \trlen = len - VARHDRSZ;\n> + #endif\n> \n> \tif (rlen > 4096)\n> \t\telog(ERROR, \"bpchar: length of char() must be less than 4096\");\n> ***************\n> *** 367,373 ****\n> --- 374,387 ----\n> \n> \t/* only reach here if we need to truncate string... */\n> \n> + #ifdef MULTIBYTE\n> + \t/* truncate multi-byte string in a way not to break\n> + \t multi-byte boundary */\n> + \tlen = pg_mbcliplen(VARDATA(s), slen - VARHDRSZ, slen - VARHDRSZ);\n> + \tslen = len + VARHDRSZ;\n> + #else\n> \tlen = slen - VARHDRSZ;\n> + #endif\n> \n> \tif (len > 4096)\n> \t\telog(ERROR, \"varchar: length of varchar() must be less than 4096\");\n> Index: backend/utils/mb/mbutils.c\n> ===================================================================\n> RCS file: /usr/local/cvsroot/pgsql/src/backend/utils/mb/mbutils.c,v\n> retrieving revision 1.3\n> diff -c -r1.3 mbutils.c\n> *** mbutils.c\t1998/09/01 04:33:22\t1.3\n> --- mbutils.c\t1998/09/24 09:03:38\n> ***************\n> *** 202,207 ****\n> --- 202,235 ----\n> }\n> \n> /*\n> + * returns the length of a multi-byte string\n> + * (not necessarily NULL terminated)\n> + * that is not longer than limit.\n> + * this function does not break multi-byte word boundary.\n> + */\n> + int\n> + pg_mbcliplen(const unsigned char *mbstr, int len, int limit)\n> + {\n> + \tint\t\t\tclen = 0;\n> + \tint\t\t\tl;\n> + \n> + \twhile (*mbstr && len > 0)\n> + \t{\n> + \t\tl = pg_mblen(mbstr);\n> + \t\tif ((clen + l) > limit) {\n> + \t\t\tbreak;\n> + \t\t}\n> + \t\tclen += l;\n> + \t\tif (clen == limit) {\n> + \t\t\tbreak;\n> + \t\t}\n> + \t\tlen -= l;\n> + \t\tmbstr += l;\n> + \t}\n> + \treturn (clen);\n> + }\n> + \n> + /*\n> * fuctions for utils/init\n> */\n> static int\tDatabaseEncoding = MULTIBYTE;\n> Index: include/mb/pg_wchar.h\n> ===================================================================\n> RCS file: /usr/local/cvsroot/pgsql/src/include/mb/pg_wchar.h,v\n> retrieving revision 1.4\n> diff -c -r1.4 pg_wchar.h\n> *** pg_wchar.h\t1998/09/01 04:36:34\t1.4\n> --- pg_wchar.h\t1998/09/24 09:03:42\n> ***************\n> *** 103,108 ****\n> --- 103,109 ----\n> extern int\tpg_mic_mblen(const unsigned char *);\n> extern int\tpg_mbstrlen(const unsigned char *);\n> extern int\tpg_mbstrlen_with_len(const unsigned char *, int);\n> + extern int\tpg_mbcliplen(const unsigned char *, int, int);\n> extern pg_encoding_conv_tbl *pg_get_encent_by_encoding(int);\n> extern bool show_client_encoding(void);\n> extern bool reset_client_encoding(void);\n> \n> \n\n\n-- \n Bruce Momjian | [email protected]\n 830 Blythe Avenue | http://www.op.net/~candle\n Drexel Hill, Pennsylvania 19026 | (610) 353-9879(w)\n + If your life is a hard drive, | (610) 853-3000(h)\n + Christ can be your backup. | \n",
"msg_date": "Thu, 24 Sep 1998 21:47:15 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] fix for multi-byte partial truncating"
},
{
"msg_contents": ">Applied, but for some reason patch did not like the normal cvs/rcs diff\n>format. Not sure why. Please check to see it is OK. Looks OK here.\n\nThank you, Bruce. Everything seems OK too.\n\nBut I found a mistake with my patches. bpchar does not pad blanks\nanymore! Could you apply following patches to\nbackend/utils/adt/varchar.c? (the diff is against the current source\ntree)\n\n*** varchar.c.orig\tFri Sep 25 15:12:34 1998\n--- varchar.c\tFri Sep 25 17:59:47 1998\n***************\n*** 147,160 ****\n \tif ((len == -1) || (len == VARSIZE(s)))\n \t\treturn s;\n \n- #ifdef MULTIBYTE\n- \t/* truncate multi-byte string in a way not to break\n- \t multi-byte boundary */\n- \trlen = pg_mbcliplen(VARDATA(s), len - VARHDRSZ, len - VARHDRSZ);\n- \tlen = rlen + VARHDRSZ;\n- #else\n \trlen = len - VARHDRSZ;\n- #endif\n \n \tif (rlen > 4096)\n \t\telog(ERROR, \"bpchar: length of char() must be less than 4096\");\n--- 147,153 ----\n***************\n*** 167,173 ****\n--- 160,172 ----\n \tresult = (char *) palloc(len);\n \tVARSIZE(result) = len;\n \tr = VARDATA(result);\n+ #ifdef MULTIBYTE\n+ \t/* truncate multi-byte string in a way not to break\n+ \t multi-byte boundary */\n+ \tslen = pg_mbcliplen(VARDATA(s), rlen, rlen);\n+ #else\n \tslen = VARSIZE(s) - VARHDRSZ;\n+ #endif\n \ts = VARDATA(s);\n \n #ifdef STRINGDEBUG\n",
"msg_date": "Fri, 25 Sep 1998 18:04:22 +0900",
"msg_from": "Tatsuo Ishii <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] fix for multi-byte partial truncating "
},
{
"msg_contents": "Applied.\n\n\n\n> >Applied, but for some reason patch did not like the normal cvs/rcs diff\n> >format. Not sure why. Please check to see it is OK. Looks OK here.\n> \n> Thank you, Bruce. Everything seems OK too.\n> \n> But I found a mistake with my patches. bpchar does not pad blanks\n> anymore! Could you apply following patches to\n> backend/utils/adt/varchar.c? (the diff is against the current source\n> tree)\n> \n> *** varchar.c.orig\tFri Sep 25 15:12:34 1998\n> --- varchar.c\tFri Sep 25 17:59:47 1998\n> ***************\n> *** 147,160 ****\n> \tif ((len == -1) || (len == VARSIZE(s)))\n> \t\treturn s;\n> \n> - #ifdef MULTIBYTE\n> - \t/* truncate multi-byte string in a way not to break\n> - \t multi-byte boundary */\n> - \trlen = pg_mbcliplen(VARDATA(s), len - VARHDRSZ, len - VARHDRSZ);\n> - \tlen = rlen + VARHDRSZ;\n> - #else\n> \trlen = len - VARHDRSZ;\n> - #endif\n> \n> \tif (rlen > 4096)\n> \t\telog(ERROR, \"bpchar: length of char() must be less than 4096\");\n> --- 147,153 ----\n> ***************\n> *** 167,173 ****\n> --- 160,172 ----\n> \tresult = (char *) palloc(len);\n> \tVARSIZE(result) = len;\n> \tr = VARDATA(result);\n> + #ifdef MULTIBYTE\n> + \t/* truncate multi-byte string in a way not to break\n> + \t multi-byte boundary */\n> + \tslen = pg_mbcliplen(VARDATA(s), rlen, rlen);\n> + #else\n> \tslen = VARSIZE(s) - VARHDRSZ;\n> + #endif\n> \ts = VARDATA(s);\n> \n> #ifdef STRINGDEBUG\n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 25 Sep 1998 11:51:24 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] fix for multi-byte partial truncating"
}
] |
[
{
"msg_contents": "Hi all,\n\nSeems that CHAR_LENGTH function has a bug.\nIt doesn't work if you try to have the length of columns with nulls.\nExample:\n\ninsert into c values('foo');\ninsert into c values(NULL);\ninsert into c values('');\n\nselect b from c;\nb\n-------\nfoo\n <-- this is a NULL\n <-- this is a ''\n(3 rows)\n\nselect char_length(b) from c;\nERROR: Bad (null) varchar() external representation\n\n-- Why this error? \n-- I think it should be:\n\nlength\n------\n 3\n <-- length of NULL should be \"unknown\" this is the behavior of\nother databases\n 0\n(3 rows)\n \nWhat do you think about ?\n Jose'\n",
"msg_date": "Thu, 24 Sep 1998 09:39:46 +0200",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "bug on CHAR_LENGTH function ?"
},
{
"msg_contents": "> select char_length(b) from c;\n> ERROR: Bad (null) varchar() external representation\n\nThis is covered in our thread with Colin...\n\n - Tom\n",
"msg_date": "Thu, 24 Sep 1998 13:52:11 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] bug on CHAR_LENGTH function ?"
}
] |
[
{
"msg_contents": "subscribe\n\n--\n--------\nRegards\nTheo\n\n\n\n",
"msg_date": "Thu, 24 Sep 1998 15:20:56 +0200",
"msg_from": "Theo Kramer <[email protected]>",
"msg_from_op": true,
"msg_subject": "None"
}
] |
[
{
"msg_contents": "I cvsuped the source this morning and configure is broken.\nThe enclosed patch fixes it, but I do not know enough about\nautoconf to be able to fix it correctly. \n\nSYSTEM:\tUlatra 5\nOS:\tSolaris 2.6\n\nThanks,\n\nMatt\n----------\nMatthew C. Aycock\nOperating Systems Analyst/Admin, Senior\nDept Math/CS\nEmory University, Atlanta, GA \nInternet: [email protected] \t\t\n\n\n",
"msg_date": "Thu, 24 Sep 1998 10:56:22 -0400 (EDT)",
"msg_from": "\"Matthew C. Aycock\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "configure..."
}
] |
[
{
"msg_contents": "I'm about halfway convinced that the database corruption problem I\nreported yesterday is a result of interlock failure among multiple\nbackends. I have a trace of the frontend/backend interactions that\nwere happening at the time the table got corrupted, and let me tell\nyou they are peculiar. Four clients were simultaneously trying to\naccess two tables, using BEGIN TRANSACTION / LOCK / END TRANSACTION\nto ensure consistency. Works fine 99% of the time. This particular\ntime, not only was the table corrupted but the clients got logically\ninconsistent results: one transaction saw some but not all of the\nupdates committed by a previous transaction. Moreover, the timestamps\nshow that one client successfully executed several begin/lock/update/\nend transaction cycles on one of the tables *while another client\nbelieved it was holding a lock on that table*. The timestamps also\nindicate that the bogus transactions took about ten times longer to\nexecute than they normally would've.\n\nGiven this evidence, I am strongly inclined to think that spinlocking\n(S_LOCK and friends) is not working right on my platform ... which is\nHPUX 9. I've eyeballed the HP-PA assembly implementation of tas(),\nand the only thing potentially wrong with it that I can see is that\nthe 16-byte slock_t object had better be aligned at least on a 4-byte\nboundary. If it happened to be placed at an odd byte address, the\ntas() code would overwrite one to three bytes beyond the end of the\nslock_t object.\n\nCan anyone say whether that's possible? Is slock_t ever part of a\ntuple that might be packed to strange boundaries?\n\nAnother thing that would kill this implementation is if someone tried\nto copy an slock_t around while it is in the locked state --- the\nassembly code is actually using whichever word of the 16-byte object\nis aligned on a 16-byte boundary, because that's what HP-PA's semaphore\nlock instruction requires. Move the slock_t to a different address,\nand the active word within it probably changes. So is there any place\nin the system where structures containing slock_t's might be shifted\naround?\n\nI think I will try modding tas.s to arrange a coredump if the passed\naddress isn't adequately aligned, and then start testing things...\nbut if anyone can tell me exactly where slock_t's usually live, it\nmight save me some time.\n\nThe next possibility is that it's not spin-locking but a higher level\nof lock code that is broken. If anyone can give me an idea where to\nlook, I'd appreciate it.\n\nBTW, I have only seen these failures with a 6.3.2 server, not with\ncurrent sources ... but I haven't stressed my development server very\nmuch with multiple clients. The bug could still be in 6.4beta.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 24 Sep 1998 14:26:07 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Can slock_t ever be unaligned?"
}
] |
[
{
"msg_contents": "> Looks like I am going to need some help here.\n> The old code dumped out regproc fields as the pg_proc.proname.\n> There is a problem with this. First, you can have multiple proname \n> entries with the same proname. The differ in their argument \n> number/types. The old code, when reading in a regproc name, would do \n> a sequential scan of the pg_proc table, and find the first entry that \n> matches the given proname.\n> If that is not the one you wanted, too bad. No way to change it.\n\nHi Bruce. I'm sorry again for being so slow, but I'm still not\nunderstanding the initial conditions which prompted these changes. Are\nyou fixing something proactively, or was there a specific example of\nmisbehavior? The example I see in your mail with Tatsuo which now causes\ntrouble is for type input and output routine names, which _are_ likely\nto be unique. \n\nWould it be possible for you to bracket the code in the cvs tree so that\nwe can enable/disable the old behavior? That way we can see what has\nchanged and how it used to behave. I suppose that would involve\nbracketing code in regprocin/out and in pg_dump??\n\nTIA\n\n - Tom\n",
"msg_date": "Fri, 25 Sep 1998 02:30:51 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
},
{
"msg_contents": "> > Looks like I am going to need some help here.\n> > The old code dumped out regproc fields as the pg_proc.proname.\n> > There is a problem with this. First, you can have multiple proname \n> > entries with the same proname. The differ in their argument \n> > number/types. The old code, when reading in a regproc name, would do \n> > a sequential scan of the pg_proc table, and find the first entry that \n> > matches the given proname.\n> > If that is not the one you wanted, too bad. No way to change it.\n> \n> Hi Bruce. I'm sorry again for being so slow, but I'm still not\n> understanding the initial conditions which prompted these changes. Are\n> you fixing something proactively, or was there a specific example of\n> misbehavior? The example I see in your mail with Tatsuo which now causes\n> trouble is for type input and output routine names, which _are_ likely\n> to be unique. \n> \n> Would it be possible for you to bracket the code in the cvs tree so that\n> we can enable/disable the old behavior? That way we can see what has\n> changed and how it used to behave. I suppose that would involve\n> bracketing code in regprocin/out and in pg_dump??\n\nProactive fix. See my other posting today that has an idea to roll back\nthe old behavour, while making sure the regproc name is unique.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 24 Sep 1998 23:38:40 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_dump, problem with user defined types?"
}
] |
[
{
"msg_contents": "I don't like implementation of referential integrity constraints \nin contrib/spi/refint.* and implementation of ... unique indices:\nbackend checks constraints when tuple is changed.\n\nOracle checks constraints after entire _statement_ is done.\nDo you remember this:\n\nvac=> create table x (a int unique);\nNOTICE: CREATE TABLE/UNIQUE will create implicit index x_a_key for table x\nCREATE\nvac=> insert into x values (1);\nINSERT 143852 1\nvac=> insert into x values (2);\nINSERT 143853 1\nvac=> update x set a = a + 1;\nERROR: Cannot insert a duplicate key into a unique index\n\nThis is simplest case. Keeping in mind triggers\nI would like to see Oracle behaviour some day and so\nI decided to do not implement FOREIGN keys in 6.4 -:((\n\nSorry.\n\nVadim\n",
"msg_date": "Fri, 25 Sep 1998 11:53:02 +0800",
"msg_from": "Vadim Mikheev <[email protected]>",
"msg_from_op": true,
"msg_subject": "unfortunately..."
},
{
"msg_contents": "> This is simplest case. Keeping in mind triggers\n> I would like to see Oracle behaviour some day and so\n> I decided to do not implement FOREIGN keys in 6.4 -:((\n> Sorry.\n\nNo problem. We have a nice feature set for v6.4 anyway, and it leaves at\nleast one thing to do for v6.5 :)\n\n - Tom\n",
"msg_date": "Fri, 25 Sep 1998 04:50:08 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] unfortunately..."
},
{
"msg_contents": "On Fri, 25 Sep 1998, Vadim Mikheev wrote:\n\n> I don't like implementation of referential integrity constraints \n> in contrib/spi/refint.* and implementation of ... unique indices:\n> backend checks constraints when tuple is changed.\n> \n> Oracle checks constraints after entire _statement_ is done.\n> Do you remember this:\n> \n> vac=> create table x (a int unique);\n> NOTICE: CREATE TABLE/UNIQUE will create implicit index x_a_key for table x\n> CREATE\n> vac=> insert into x values (1);\n> INSERT 143852 1\n> vac=> insert into x values (2);\n> INSERT 143853 1\n> vac=> update x set a = a + 1;\n> ERROR: Cannot insert a duplicate key into a unique index\n> \n> This is simplest case. Keeping in mind triggers\n> I would like to see Oracle behaviour some day and so\n> I decided to do not implement FOREIGN keys in 6.4 -:((\n\n\tSounds reasonable to me...all things considered, I think that v6.4\nhas turned out, once more, to be one giant leap forward :)\n\n> Sorry.\n\n\tDon't be...I think if Bruce had his way, we'd be funding a cloning\nproject just so that we could get a dozen of you :) Its nice to be\nreminded once in a while that you are \"only human\" *grin*\n\n\n\n",
"msg_date": "Fri, 25 Sep 1998 07:18:08 -0400 (EDT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] unfortunately..."
},
{
"msg_contents": "> I don't like implementation of referential integrity constraints \n> in contrib/spi/refint.* and implementation of ... unique indices:\n> backend checks constraints when tuple is changed.\n> \n> Oracle checks constraints after entire _statement_ is done.\n> Do you remember this:\n> \n> vac=> create table x (a int unique);\n> NOTICE: CREATE TABLE/UNIQUE will create implicit index x_a_key for table x\n> CREATE\n> vac=> insert into x values (1);\n> INSERT 143852 1\n> vac=> insert into x values (2);\n> INSERT 143853 1\n> vac=> update x set a = a + 1;\n> ERROR: Cannot insert a duplicate key into a unique index\n> \n> This is simplest case. Keeping in mind triggers\n> I would like to see Oracle behaviour some day and so\n> I decided to do not implement FOREIGN keys in 6.4 -:((\n\nOK. Removed from open 6.4 items list.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 25 Sep 1998 11:48:30 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] unfortunately..."
},
{
"msg_contents": "Thus spake Vadim Mikheev\n> I decided to do not implement FOREIGN keys in 6.4 -:((\n\nDoes this mean that we also won't have full primary key support? It\nseems to me like it could be implemented without worrying about foreign\nkeys but you know better. I can work around it given the support\nalready there but it would be nice if making a field primary would\nautomatically set the relevant indisprimary field right now.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Fri, 25 Sep 1998 13:07:59 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: unfortunately..."
}
] |
[
{
"msg_contents": "Hi Bruce. I'd like to propose that we use the sgml sources for the\ninstallation instructions. That will give us one set of instructions\nwhich can appear in postscript, html, and plain text. I'll generate a\nplain text version which we can put into the top-level directory as\nusual.\n\nThe plain-text version can include the accurate porting info and release\nnotes also, since those are in sgml.\n\nI notice that INSTALL was updated on Aug. 30, but that the sgml version\nof the same instructions (install.sgml) was not, so we would have to\nremerge again anyway...\n\n - Tom\n",
"msg_date": "Fri, 25 Sep 1998 14:50:08 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Installation instructions"
},
{
"msg_contents": "> Hi Bruce. I'd like to propose that we use the sgml sources for the\n> installation instructions. That will give us one set of instructions\n> which can appear in postscript, html, and plain text. I'll generate a\n> plain text version which we can put into the top-level directory as\n> usual.\n> \n> The plain-text version can include the accurate porting info and release\n> notes also, since those are in sgml.\n> \n> I notice that INSTALL was updated on Aug. 30, but that the sgml version\n> of the same instructions (install.sgml) was not, so we would have to\n> remerge again anyway...\n\nOK. Makes sense.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 25 Sep 1998 11:53:52 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [DOCS] Installation instructions"
}
] |
[
{
"msg_contents": "CC'd to the PostgreSQL Hackers list.\n\nI performed the same test as Brian on the cvs version of 6.4 and it\nexhibits the same behavior. could we get a fix in for the release.\n\t\t-DEJ\n\n> Hehehe... \n> \n> Try using the \"*\", as I posted to your early send... see if it makes\n> any\n> difference.\nDoing a like '*' will search for the character '*';\n\n> > er sorry about that last \"early send\" message...\n> > \n> > well, at first I though you were correct, but it turns out that\n> > postgresql\n> > is also inconsistent. Consider a table with a field username. 3\n> records:\n> > \n> > \"Brian % Schaffner\"\n> > \"Brian T Schaffner\"\n> > \"%\"\n> > \n> > select * from table where username='%' gets all rows (expected)\n> > select * from table where username='%%' gets the row with \"%\"\n> (expected)\n> > select * from table where username='%%%' gets the row with \"%\"\n> > (expected)\n> > select * from table where username='%%%%' gets no rows (expected)\n> > select * from table where username='% %' gets the 2 name rows\n> (expected)\n> > select * from table where username='% %%' gets the 2 name rows (NOT\n> > expected)\n> > select * from table where username='%% %' gets no rows (NOT\n> expected)\n> > select * from table where username='% % %' gets the 2 name rows\n> > (expected)\n> > select * from table where username='% %% %' gets the 2 name rows\n> (NOT\n> > expected)\n> > select * from table where username='% % % %' gets no rows (expected)\n> > \n> > so, if %% is the LIKE representation for a literal %, then why does\n> '%\n> > %%' return\n> > the 2 name rows, and '%% %' return no rows, and '% %% %' not return\n> the\n> > row with the\n> > embedded literal %?\n> > \n> > I could not get postgres to single out the row with the embedded %\n> using\n> > LIKE.\n> > \n> > why is this not getting any easier to define?\n> > \n> > -----Original Message-----\n> > From: Daniel J. Lashua [mailto:[email protected]]\n> > Sent: Friday, September 25, 1998 10:58 AM\n> > To: Brian Schaffner\n> > Cc: 'Rasmus Lerdorf'; '[email protected]';\n> > [email protected]\n> > Subject: RE: [PHP3] ARGH!! strstr() changed?\n> > \n> > \n> > On Fri, 25 Sep 1998, Brian Schaffner wrote:\n> > \n> > > in PostgreSQL (6.3.2 on FreeBSD 2.2.6):\n> > > \n> > > select * from table where field like '%%';\n> > > \n> > > returns NO rows;\n> > > \n> > > select * from table where field like '%';\n> > > \n> > > returns ALL rows;\n> > > \n> > > -brian-\n> > > \n> > \n> > \n> > I am not in any way doubting what you say... but that doesn't seem\n> > right.\n> > They should both reuturn all rows. Maybe in Postgres %% is the way\n> to\n> > state you actually want to search for ONE \"%\"?\n> > \n> > Out of curiosity, if you have time, could you test that. Make a\n> table\n> > with\n> > a field and in one row of the table in the field insert \"%\". Then do\n> > your\n> > select * from table where somefield LIKE '%%' and see if it returns\n> the\n> > one row?\n> > \n> > Am I offbase, or does that sound like incorrect behavior to you too?\n> > \n> > Daniel\n> > \n> > \n> \n> \n> --\n> PHP 3 Mailing List http://www.php.net/\n> To unsubscribe send an empty message to [email protected]\n> To subscribe to the digest list: [email protected]\n> For help: [email protected] Archive:\n> http://www.php.net/mailsearch.php3\n> \n",
"msg_date": "Fri, 25 Sep 1998 13:55:51 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [PHP3] [OFF-TOPIC] POSTGRES"
},
{
"msg_contents": "I am just getting to this message. I have overhauled the LIKE handling\nfor the upcoming 6.5 beta(soon to start).\n\nI am attaching a patch that represents my changes.\n\nFYI, the SQL standard say LIKE '%%' is the same as LIKE '%'. To match a\nliteral '%', you have to do '\\%'.\n\nLet me know if you see any LIKE problems. Thanks.\n\n\n> CC'd to the PostgreSQL Hackers list.\n> \n> I performed the same test as Brian on the cvs version of 6.4 and it\n> exhibits the same behavior. could we get a fix in for the release.\n> \t\t-DEJ\n> \n> > Hehehe... \n> > \n> > Try using the \"*\", as I posted to your early send... see if it makes\n> > any\n> > difference.\n> Doing a like '*' will search for the character '*';\n> \n> > > er sorry about that last \"early send\" message...\n> > > \n> > > well, at first I though you were correct, but it turns out that\n> > > postgresql\n> > > is also inconsistent. Consider a table with a field username. 3\n> > records:\n> > > \n> > > \"Brian % Schaffner\"\n> > > \"Brian T Schaffner\"\n> > > \"%\"\n> > > \n> > > select * from table where username='%' gets all rows (expected)\n> > > select * from table where username='%%' gets the row with \"%\"\n> > (expected)\n> > > select * from table where username='%%%' gets the row with \"%\"\n> > > (expected)\n> > > select * from table where username='%%%%' gets no rows (expected)\n> > > select * from table where username='% %' gets the 2 name rows\n> > (expected)\n> > > select * from table where username='% %%' gets the 2 name rows (NOT\n> > > expected)\n> > > select * from table where username='%% %' gets no rows (NOT\n> > expected)\n> > > select * from table where username='% % %' gets the 2 name rows\n> > > (expected)\n> > > select * from table where username='% %% %' gets the 2 name rows\n> > (NOT\n> > > expected)\n> > > select * from table where username='% % % %' gets no rows (expected)\n> > > \n> > > so, if %% is the LIKE representation for a literal %, then why does\n> > '%\n> > > %%' return\n> > > the 2 name rows, and '%% %' return no rows, and '% %% %' not return\n> > the\n> > > row with the\n> > > embedded literal %?\n> > > \n> > > I could not get postgres to single out the row with the embedded %\n> > using\n> > > LIKE.\n> > > \n> > > why is this not getting any easier to define?\n> > > \n> > > -----Original Message-----\n> > > From: Daniel J. Lashua [mailto:[email protected]]\n> > > Sent: Friday, September 25, 1998 10:58 AM\n> > > To: Brian Schaffner\n> > > Cc: 'Rasmus Lerdorf'; '[email protected]';\n> > > [email protected]\n> > > Subject: RE: [PHP3] ARGH!! strstr() changed?\n> > > \n> > > \n> > > On Fri, 25 Sep 1998, Brian Schaffner wrote:\n> > > \n> > > > in PostgreSQL (6.3.2 on FreeBSD 2.2.6):\n> > > > \n> > > > select * from table where field like '%%';\n> > > > \n> > > > returns NO rows;\n> > > > \n> > > > select * from table where field like '%';\n> > > > \n> > > > returns ALL rows;\n> > > > \n> > > > -brian-\n> > > > \n> > > \n> > > \n> > > I am not in any way doubting what you say... but that doesn't seem\n> > > right.\n> > > They should both reuturn all rows. Maybe in Postgres %% is the way\n> > to\n> > > state you actually want to search for ONE \"%\"?\n> > > \n> > > Out of curiosity, if you have time, could you test that. Make a\n> > table\n> > > with\n> > > a field and in one row of the table in the field insert \"%\". Then do\n> > > your\n> > > select * from table where somefield LIKE '%%' and see if it returns\n> > the\n> > > one row?\n> > > \n> > > Am I offbase, or does that sound like incorrect behavior to you too?\n> > > \n> > > Daniel\n> > > \n> > > \n> > \n> > \n> > --\n> > PHP 3 Mailing List http://www.php.net/\n> > To unsubscribe send an empty message to [email protected]\n> > To subscribe to the digest list: [email protected]\n> > For help: [email protected] Archive:\n> > http://www.php.net/mailsearch.php3\n> > \n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\nIndex: like.c\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/backend/utils/adt/like.c,v\nretrieving revision 1.22\nretrieving revision 1.23\ndiff -c -r1.22 -r1.23\n*** like.c\t1999/03/15 02:18:36\t1.22\n--- like.c\t1999/03/15 13:45:09\t1.23\n***************\n*** 150,159 ****\n {\n \tint\t\t\tmatched;\n \n! \tfor (; *p; text ++, p++)\n \t{\n- \t\tif (*text == '\\0' && *p != '%')\n- \t\t\treturn LIKE_ABORT;\n \t\tswitch (*p)\n \t\t{\n \t\t\tcase '\\\\':\n--- 150,157 ----\n {\n \tint\t\t\tmatched;\n \n! \tfor (; *p && *text; text++, p++)\n \t{\n \t\tswitch (*p)\n \t\t{\n \t\t\tcase '\\\\':\n***************\n*** 161,187 ****\n \t\t\t\tp++;\n \t\t\t\t/* FALLTHROUGH */\n \t\t\tdefault:\n! \t\t\t\tif (*text !=*p)\n \t\t\t\t\treturn LIKE_FALSE;\n! \t\t\t\tcontinue;\n \t\t\tcase '_':\n \t\t\t\t/* Match anything. */\n! \t\t\t\tcontinue;\n \t\t\tcase '%':\n! \t\t\t\twhile (*++p == '%')\n! \t\t\t\t\t/* Consecutive percents act just like one. */\n! \t\t\t\t\tcontinue;\n \t\t\t\tif (*p == '\\0')\n \t\t\t\t\t/* Trailing percent matches everything. */\n \t\t\t\t\treturn LIKE_TRUE;\n \t\t\t\twhile (*text)\n! \t\t\t\t\tif ((matched = DoMatch(text ++, p)) != LIKE_FALSE)\n \t\t\t\t\t\treturn matched;\n \t\t\t\treturn LIKE_ABORT;\n \t\t}\n \t}\n \n! \treturn *text == '\\0';\n }\n \n \n--- 159,201 ----\n \t\t\t\tp++;\n \t\t\t\t/* FALLTHROUGH */\n \t\t\tdefault:\n! \t\t\t\tif (*text != *p)\n \t\t\t\t\treturn LIKE_FALSE;\n! \t\t\t\tbreak;\n \t\t\tcase '_':\n \t\t\t\t/* Match anything. */\n! \t\t\t\tbreak;\n \t\t\tcase '%':\n! \t\t\t\t/* %% is the same as % according to the SQL standard */\n! \t\t\t\t/* Advance past all %'s */\n! \t\t\t\twhile (*p == '%')\n! \t\t\t\t\tp++;\n \t\t\t\tif (*p == '\\0')\n \t\t\t\t\t/* Trailing percent matches everything. */\n \t\t\t\t\treturn LIKE_TRUE;\n \t\t\t\twhile (*text)\n! \t\t\t\t{\n! \t\t\t\t\t/* Optimization to prevent most recursion */\n! \t\t\t\t\tif ((*text == *p ||\n! \t\t\t\t\t *p == '\\\\' || *p == '%' || *p == '_') &&\n! \t\t\t\t\t\t(matched = DoMatch(text, p)) != LIKE_FALSE)\n \t\t\t\t\t\treturn matched;\n+ \t\t\t\t\ttext++;\n+ \t\t\t\t}\n \t\t\t\treturn LIKE_ABORT;\n \t\t}\n \t}\n \n! \tif (*text != '\\0')\n! \t\treturn LIKE_ABORT;\n! \telse\n! \t{\n! \t\t/* End of input string. Do we have matching string remaining? */\n! \t\tif (p[0] == '\\0' || (p[0] == '%' && p[1] == '\\0'))\n! \t\t\treturn LIKE_TRUE;\n! \t\telse\n! \t\t\treturn LIKE_ABORT;\n! \t}\n }",
"msg_date": "Mon, 15 Mar 1999 08:48:17 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [PHP3] [OFF-TOPIC] POSTGRES"
}
] |
[
{
"msg_contents": "It's easy: just destroydb one database that has active backends\nstill running in it. Then stand back and watch the carnage.\n(Hint: back up all your *other* databases first, because you'll\nbe lucky if you can still access them.)\n\nThere really, really needs to be an interlock to prevent this mistake\n:-(\n\n\t\t\tregards, tom lane\n",
"msg_date": "Fri, 25 Sep 1998 18:18:12 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "How to destroy your entire Postgres installation"
},
{
"msg_contents": "Nick Bastin <[email protected]> writes:\n> Now I'm no programming neophyte, but can somebody explain to me\n> why this happened? What exactly is destroydb doing, or am I missing\n> something obvious here?\n\nWell, the backends for the destroyed database are *still running*;\nyou didn't kill them off by deleting their current working directory\nfrom under them. (In fact, the database's top level directory is still\nthere, because those processes still have it open ... it just has no\nlinks left in the filesystem and will be deleted when the last process\nholding it open exits. Some of the member files of the dead database\nare likely still on disk for the same reason.)\n\nThat means those backends are still participating in the shared memory\nbuffer arena used by all the backends. And, very possibly, have dirty\nbuffers that should have been written out to files of the destroyed DB.\n\nWhen I did this I got messages like\n\t\"mdblindwrt: oid of db XYZ is not NNNNN\"\nwhich a quick 'glimpse' traces to a routine with this header comment:\n\n/*\n *\tmdblindwrt() -- Write a block to disk blind.\n *\n *\t\tWe have to be able to do this using only the name and OID of\n *\t\tthe database and relation in which the block belongs. This\n *\t\tis a synchronous write.\n */\n\nThe error message is fairly misleading, because it's actually used for\n*any* failure to look up the database's info ... like, say, the database\nhaving been deleted.\n\nI got this even from backends that had nothing to do with the dead\ndatabase and had been started after it was destroyed. Killing all the\nbackends belonging to the dead database didn't help. I surmise that the\nbackends communally take responsibility for writing dirty buffers out to\nthe files where they belong, and thus any backend might try to write out\nsuch an orphaned buffer --- and when it fails, it treats that as a fatal\nerror.\n\nPerhaps someone with a better understanding of the backend can say more.\n\nAnyway, I felt very lucky that I was able to extract the data I needed\nfrom my non-toy databases. (BTW, a hint for anyone else who makes the\nsame mistake: try creating the dead database again. That seems to be\nenough to prevent mdblindwrt from deciding that it has a fatal error\non its hands.)\n\nBut, as I said, there ought to be some interlocks in there. You should\nnot be able to destroy a database that has connected backends --- and\nit'd be a good idea to scan the buffer pool and make darn sure it has\nno associated buffers, either.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Fri, 25 Sep 1998 19:36:12 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] How to destroy your entire Postgres installation "
},
{
"msg_contents": "Tom Lane wrote:\n> \n> It's easy: just destroydb one database that has active backends\n> still running in it. Then stand back and watch the carnage.\n> (Hint: back up all your *other* databases first, because you'll\n> be lucky if you can still access them.)\n\nSo just for fun I tried this.. ;-) I did a destroydb on a database with 2\nactive backends and only 90 megs of data, and it managed to munge through the\nother running database, which is fast approaching 2 gigs, in short order. The\nfirst time I did it, I was able to rescue the second database just using the\naudit trails that I'd built into it, but the second time it completely trashed\nit. Now I'm no programming neophyte, but can somebody explain to me why this\nhappened? What exactly is destroydb doing, or am I missing something obvious here?\n\n--\nNick Bastin\nRBB Systems, Inc.\n",
"msg_date": "Fri, 25 Sep 1998 19:07:43 -0600",
"msg_from": "Nick Bastin <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] How to destroy your entire Postgres installation"
},
{
"msg_contents": "I have modified the destroydb code so it marks all buffers associated\nwith the database as clean _before_ removing the files. The old code\nmarked the buffers as clean after removing the database files. Any\nbackend trying to flush dirty buffers for that database after the files\nwere removed but before the buffers were marked clean would get errors.\n\nThis does not fix the problem of someone else having the database open\nduring the destroy, but should fix the other problem you mentioned.\n\n\n\n\n> Nick Bastin <[email protected]> writes:\n> > Now I'm no programming neophyte, but can somebody explain to me\n> > why this happened? What exactly is destroydb doing, or am I missing\n> > something obvious here?\n> \n> Well, the backends for the destroyed database are *still running*;\n> you didn't kill them off by deleting their current working directory\n> from under them. (In fact, the database's top level directory is still\n> there, because those processes still have it open ... it just has no\n> links left in the filesystem and will be deleted when the last process\n> holding it open exits. Some of the member files of the dead database\n> are likely still on disk for the same reason.)\n> \n> That means those backends are still participating in the shared memory\n> buffer arena used by all the backends. And, very possibly, have dirty\n> buffers that should have been written out to files of the destroyed DB.\n> \n> When I did this I got messages like\n> \t\"mdblindwrt: oid of db XYZ is not NNNNN\"\n> which a quick 'glimpse' traces to a routine with this header comment:\n> \n> /*\n> *\tmdblindwrt() -- Write a block to disk blind.\n> *\n> *\t\tWe have to be able to do this using only the name and OID of\n> *\t\tthe database and relation in which the block belongs. This\n> *\t\tis a synchronous write.\n> */\n> \n> The error message is fairly misleading, because it's actually used for\n> *any* failure to look up the database's info ... like, say, the database\n> having been deleted.\n> \n> I got this even from backends that had nothing to do with the dead\n> database and had been started after it was destroyed. Killing all the\n> backends belonging to the dead database didn't help. I surmise that the\n> backends communally take responsibility for writing dirty buffers out to\n> the files where they belong, and thus any backend might try to write out\n> such an orphaned buffer --- and when it fails, it treats that as a fatal\n> error.\n> \n> Perhaps someone with a better understanding of the backend can say more.\n> \n> Anyway, I felt very lucky that I was able to extract the data I needed\n> from my non-toy databases. (BTW, a hint for anyone else who makes the\n> same mistake: try creating the dead database again. That seems to be\n> enough to prevent mdblindwrt from deciding that it has a fatal error\n> on its hands.)\n> \n> But, as I said, there ought to be some interlocks in there. You should\n> not be able to destroy a database that has connected backends --- and\n> it'd be a good idea to scan the buffer pool and make darn sure it has\n> no associated buffers, either.\n> \n> \t\t\tregards, tom lane\n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Mon, 15 Mar 1999 09:11:28 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] How to destroy your entire Postgres installation"
}
] |
[
{
"msg_contents": "With the current snapshot, there seems to be problem with functional indexes. \nExample:\n\nCREATE TABLE timelog (loginname text, site varchar(16), start_time datetime, \nelapsed timespan);\nCREATE FUNCTION datetime_day (datetime ) RETURNS float8 AS 'select \ndate_part(''day'', $1)' LANGUAGE 'SQL';\nINSERT INTO timelog values ('abc','varna','Thu 01 Jan 23:23:34 1998 EET','@ 10 \nmins');\nINSERT INTO timelog values ('abc','sofia','Thu 01 Jan 23:23:34 1998 EET','@ 2 \nhours 10 mins');\n\n$ psql test\nWelcome to the POSTGRESQL interactive sql monitor:\n Please read the file COPYRIGHT for copyright terms of POSTGRESQL\n\n type \\? for help on slash commands\n type \\q to quit\n type \\g or terminate with semicolon to execute query\n You are currently connected to the database: test\n\ntest=> create index timelog_date_idx on timelog (datetime_day(start_time));\nERROR: DefineIndex: (null) class not found\ntest=> \n\nAlso, using date_part() directly in the index definition:\n\ntest=> create index timelog_date_idx on timelog (datetime_part('day', \nstart_time));\nERROR: parser: parse error at or near \"'\"\n\n\nCan something be done about this before 6.4 is released?\n\nDaniel\n\n\n",
"msg_date": "Sat, 26 Sep 1998 12:53:59 +0300",
"msg_from": "Daniel Kalchev <[email protected]>",
"msg_from_op": true,
"msg_subject": "functional indexes"
},
{
"msg_contents": "Added to open 6.4 bugs list.\n\n\n> With the current snapshot, there seems to be problem with functional indexes. \n> Example:\n> \n> CREATE TABLE timelog (loginname text, site varchar(16), start_time datetime, \n> elapsed timespan);\n> CREATE FUNCTION datetime_day (datetime ) RETURNS float8 AS 'select \n> date_part(''day'', $1)' LANGUAGE 'SQL';\n> INSERT INTO timelog values ('abc','varna','Thu 01 Jan 23:23:34 1998 EET','@ 10 \n> mins');\n> INSERT INTO timelog values ('abc','sofia','Thu 01 Jan 23:23:34 1998 EET','@ 2 \n> hours 10 mins');\n> \n> $ psql test\n> Welcome to the POSTGRESQL interactive sql monitor:\n> Please read the file COPYRIGHT for copyright terms of POSTGRESQL\n> \n> type \\? for help on slash commands\n> type \\q to quit\n> type \\g or terminate with semicolon to execute query\n> You are currently connected to the database: test\n> \n> test=> create index timelog_date_idx on timelog (datetime_day(start_time));\n> ERROR: DefineIndex: (null) class not found\n> test=> \n> \n> Also, using date_part() directly in the index definition:\n> \n> test=> create index timelog_date_idx on timelog (datetime_part('day', \n> start_time));\n> ERROR: parser: parse error at or near \"'\"\n> \n> \n> Can something be done about this before 6.4 is released?\n> \n> Daniel\n> \n> \n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 02:16:08 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] functional indexes"
},
{
"msg_contents": ">\n> Added to open 6.4 bugs list.\n\n Can be removed from the list.\n\n>\n>\n> > With the current snapshot, there seems to be problem with functional indexes.\n> > Example:\n> >\n> > CREATE TABLE timelog (loginname text, site varchar(16), start_time datetime,\n> > elapsed timespan);\n> > CREATE FUNCTION datetime_day (datetime ) RETURNS float8 AS 'select\n> > date_part(''day'', $1)' LANGUAGE 'SQL';\n> > INSERT INTO timelog values ('abc','varna','Thu 01 Jan 23:23:34 1998 EET','@ 10\n> > mins');\n> > INSERT INTO timelog values ('abc','sofia','Thu 01 Jan 23:23:34 1998 EET','@ 2\n> > hours 10 mins');\n> >\n> > $ psql test\n> > Welcome to the POSTGRESQL interactive sql monitor:\n> > Please read the file COPYRIGHT for copyright terms of POSTGRESQL\n> >\n> > type \\? for help on slash commands\n> > type \\q to quit\n> > type \\g or terminate with semicolon to execute query\n> > You are currently connected to the database: test\n> >\n> > test=> create index timelog_date_idx on timelog (datetime_day(start_time));\n> > ERROR: DefineIndex: (null) class not found\n> > test=>\n> >\n> > Also, using date_part() directly in the index definition:\n> >\n> > test=> create index timelog_date_idx on timelog (datetime_part('day',\n> > start_time));\n> > ERROR: parser: parse error at or near \"'\"\n> >\n> >\n> > Can something be done about this before 6.4 is released?\n> >\n> > Daniel\n> >\n> >\n> >\n> >\n\n Problem 1:\n You must specify an operator class for the index.\n\n Solution:\n CREATE INDEX timelog_date_idx ON timelog\n (datetime_day(start_time) float8_ops);\n\n in the above case.\n\n Problem 2:\n Only tuple attributes can be given to a functional\n indexes proc. This is why date_part('day', start_time)\n cannot be used directly. There is no place in pg_index to\n store constant values.\n\n Solution:\n As you did with a wrapper function. But...\n\n Problem 3:\n You cannot use SQL language functions for functional\n index procs. The index access methods call\n fmgr_array_args() to compute the functional index value.\n\tAnd this cannot execute SQL language functions.\n\n Solution:\n Write a C language function for it or use the PL/pgSQL\n function\n\n CREATE FUNCTION datetime_day(datetime) RETURNS float8 AS '\n BEGIN\n RETURN date_part(''day'', $1);\n END;'\n LANGUAGE 'plpgsql';\n\n after installing the procedural language in your\n database. Did I tell that I like PL/pgSQL more and more?\n\tThink it was really time for it.\n\n Problem 4:\n The isNull flag used in access/index/indexam.c function\n GetIndexValue() is not initialized correctly.\n\n Solution:\n Apply the patch at the end.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n*** indexam.c.orig\tFri Oct 2 15:16:49 1998\n--- indexam.c\tFri Oct 2 15:17:24 1998\n***************\n*** 362,368 ****\n \t\t\t bool *attNull)\n {\n \tDatum\t\treturnVal;\n! \tbool\t\tisNull;\n \n \tif (PointerIsValid(fInfo) && FIgetProcOid(fInfo) != InvalidOid)\n \t{\n--- 362,368 ----\n \t\t\t bool *attNull)\n {\n \tDatum\t\treturnVal;\n! \tbool\t\tisNull = FALSE;\n \n \tif (PointerIsValid(fInfo) && FIgetProcOid(fInfo) != InvalidOid)\n \t{\n***************\n*** 375,387 ****\n \t\t\t\t\t\t\t\t\t attrNums[i],\n \t\t\t\t\t\t\t\t\t hTupDesc,\n \t\t\t\t\t\t\t\t\t attNull);\n \t\t}\n \t\treturnVal = (Datum) fmgr_array_args(FIgetProcOid(fInfo),\n \t\t\t\t\t\t\t\t\t\t\tFIgetnArgs(fInfo),\n \t\t\t\t\t\t\t\t\t\t\t(char **) attData,\n \t\t\t\t\t\t\t\t\t\t\t&isNull);\n \t\tpfree(attData);\n! \t\t*attNull = FALSE;\n \t}\n \telse\n \t\treturnVal = heap_getattr(tuple, attrNums[attOff], hTupDesc, attNull);\n--- 375,389 ----\n \t\t\t\t\t\t\t\t\t attrNums[i],\n \t\t\t\t\t\t\t\t\t hTupDesc,\n \t\t\t\t\t\t\t\t\t attNull);\n+ \t\t\tif (*attNull)\n+ \t\t\t\tisNull = TRUE;\n \t\t}\n \t\treturnVal = (Datum) fmgr_array_args(FIgetProcOid(fInfo),\n \t\t\t\t\t\t\t\t\t\t\tFIgetnArgs(fInfo),\n \t\t\t\t\t\t\t\t\t\t\t(char **) attData,\n \t\t\t\t\t\t\t\t\t\t\t&isNull);\n \t\tpfree(attData);\n! \t\t*attNull = isNull;\n \t}\n \telse\n \t\treturnVal = heap_getattr(tuple, attrNums[attOff], hTupDesc, attNull);\n",
"msg_date": "Fri, 2 Oct 1998 15:36:09 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] functional indexes"
}
] |
[
{
"msg_contents": "Well, I've spent the last several hours studying how NOTIFY works, and\nI think it's a mess :-(. We need some significant revisions in there.\n\nI have not been able to trace an exact flow of control that would\nyield the NOTIFY-message-interspersed-in-data result that I saw this\nafternoon. However, I am pretty certain that I know the basic cause\nof the problem. The risky aspect of NOTIFY is that very substantial\namounts of processing (an entire transaction, in fact) are launched from\na signal handler routine. As a rule of thumb, doing more than setting a\nflag variable in a signal handler is a big NO NO, because you don't have\nany idea what bit of code you've interrupted or what you can safely\ncall. (I have a strong suspicion, in fact, that the behavior I saw\narose from the signal interrupt occuring during the backend's fflush()\nof data destined for the frontend at the end of a query cycle. I dunno\nhow re-entrant the stdio library is on my machine, but it seems quite\npossible that it behaves undesirably if one attempts to queue more\noutput data from a routine that interrupts fflush()...)\n\nThe Postgres code tries to protect itself against this sort of problem\nby having the SIGUSR2 handler not do anything except set a flag if the\ncurrent transaction state says that a transaction is in progress. This\nis almost good enough, but evidently not quite. The first issue is that\nsome query initialization and finalization code in PostgresMain happens\noutside the StartTransactionCommand...CommitTransactionCommand zone of\nsafety. (In particular, the final fflush of data bound for the\nfrontend.) The second issue is that async.c itself sometimes starts and\nends transactions, by calling xact.c, and yet xact.c also calls async.c\nas part of transaction commit. At least two levels of recursion are\npossible in there. I have not convinced myself whether the bug I\nencountered is due to some subtly incorrect ordering of events in that\ninteraction, or whether it's just due to non-reentrancy in stdio.\nBut I no longer care. This code is demonstrably buggy, and it is too\ncomplex to understand or maintain anyway. I propose to rewrite it.\n\nI propose that:\n\n1. The link between inbound-notify processing and transaction\nstart/commit should be removed. Having inbound-notify execution called\nfrom transaction commit is too complicated because that notify execution\nitself may have to induce a transaction start/commit. Instead, inbound\nnotify execution ought to be triggered from the PostgresMain main loop.\nThis has the additional advantage that we can have more control over\njust which parts of the main loop allow or disallow notify execution\ndirectly from the signal handler. The main loop should now contain\nsomething like:\n\n\tfor (;;) {\n\n\t\tReadyForQuery();\t// contains fflush() call\n\n\t\tAllowNotifyInterrupt();\n\n\t\tReadCommand();\n\n\t\tDisableNotifyInterrupt();\n\n\t\texecute command, including transaction start/commit\n\t}\n\nso that the SIGUSR2 interrupt is allowed to do something serious\nonly in the *very* tiny window where we are just waiting for or\nreading the next frontend query. (Since the interrupt has no\nreason to touch the state of stdin, this should be safe enough.)\nAn interrupt occuring any other time will just set a flag, which\nwill be read by AllowNotifyInterrupt --- if an interrupt has occured\nsince the last DisableNotifyInterrupt, then AllowNotifyInterrupt will\ncarry out the necessary processing. Thus, all inbound-notify execution\neffectively happens at the outer loop --- either while waiting for\nfrontend input, or just before doing so.\n\n\n2. The innards of NOTIFY execution are unnecessarily complicated and\ninefficient, quite aside from the above problem of controlling interrupt\nsafety. The NOTIFY statement itself (routine Async_Notify) makes one\npass over the pg_listener relation, writing a 1 into the notification\nfield of each tuple with a matching relname. Then at commit time, we\nhave to make another pass over the pg_listener relation to issue signals\n(and/or send a message to our own front end, if we're NOTIFYing\nsomething we also LISTEN on, which is pretty common in my applications\nanyway). This is unnecessary processing; worse, if the NOTIFY is done\nearly in a long transaction block, we end up holding a write lock on\npg_listen for a long time. Since pg_listen is a global resource, this\ncan effectively bring all the rest of the backends to a halt!\n\nThe right way to do this is for NOTIFY to simply add the target relation\nname to a list of pending notifies (as indeed it does already). Then\nwhen we are ready to commit, we scan pg_listen *once*, performing tuple\nupdates as needed and also sending signals and/or frontend messages.\n(Note this is for an *outbound* notify, which we do want to have happen\nduring transaction commit. The real problem with the current async.c is\nthat it tries to combine outbound and inbound notify code. Keeping them\nseparate will be better.) This approach takes less processing, and it\nalso means that we need not do anything so risky as\nRelationUnsetLockForWrite() to gain performance. The pg_listen lock\nwill be held for only a very short time before we release it as a normal\npart of commit. Moreover, a self-notify need never modify the pg_listen\ntable at all, saving cycles and opening the door for another feature\n(see below).\n\nIf the transaction is aborted rather than committed, we need only\ndiscard the list of pending notify names. pg_listen is never touched\nin that case.\n\n\n3. While I am at it, I would like to add a feature that I've been\nwanting for a while. It seems to me that the be_pid field provided in a\nPGnotify message ought not always be your own backend's PID --- what use\nis that? What it ought to be is the PID of the *signaling* backend, or\nat least of one of the signaling backends other than your own. I have\na bunch of applications that all read and write the same tables and let\neach other know about updates by NOTIFY messages. When one of these\napps gets a NOTIFY, it has to re-read the table *even if the notify was\nthe one it just issued*, and the updates in question are the ones it\njust wrote out. This is a waste of cycles, but right now there is no\nchoice because you cannot tell whether the NOTIFY you get is due\n(solely) to your own NOTIFY or (perhaps in part) to someone else's.\n\nWhat I'd like to do is to have cross-backend notifies write the\noriginating backend's PID into the \"notification\" field, rather than\njust a \"1\". The receiving backend can then send that value to its\nfrontend, as opposed to sending its own PID as it will do in the\nself-notify case. In this way the frontend can unambiguously\ndistinguish its own NOTIFY bouncing back from a NOTIFY from another\nclient. (If several other clients send NOTIFY at about the same time,\nyou may get only one message with only one of their PIDs, since there\nis only room for one originating PID in pg_listen. But it will be kept\ndistinct from any message resulting from a NOTIFY of your own.) Note\nthat we already made the backend's own PID available to the frontend as\npart of the CANCEL-related changes, so that part of the info is there\nfor free.\n\nA more radical solution would be to suppress self-notifies altogether,\nbut that is a change in semantics that would likely break some existing\napps. I will be satisfied if the frontend can tell the difference\nbetween a self-notify and an incoming notify.\n\n\nAny objections? If there's something crucial I've missed in all this,\nbetter let me know ASAP!\n\nI know that it's a tad late in the beta cycle to be making such significant\nchanges, but my applications really need NOTIFY to work and work reliably.\nWe have to have this *now*. I hope y'all will bear with me.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sat, 26 Sep 1998 21:05:13 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Think we need major revisions in async.c..."
},
{
"msg_contents": "> What I'd like to do is to have cross-backend notifies write the\n> originating backend's PID into the \"notification\" field, rather than\n> just a \"1\". The receiving backend can then send that value to its\n> frontend, as opposed to sending its own PID as it will do in the\n> self-notify case. In this way the frontend can unambiguously\n> distinguish its own NOTIFY bouncing back from a NOTIFY from another\n> client. (If several other clients send NOTIFY at about the same time,\n> you may get only one message with only one of their PIDs, since there\n> is only room for one originating PID in pg_listen. But it will be kept\n> distinct from any message resulting from a NOTIFY of your own.) Note\n> that we already made the backend's own PID available to the frontend as\n> part of the CANCEL-related changes, so that part of the info is there\n> for free.\n> \n> A more radical solution would be to suppress self-notifies altogether,\n> but that is a change in semantics that would likely break some existing\n> apps. I will be satisfied if the frontend can tell the difference\n> between a self-notify and an incoming notify.\n> \n> \n> Any objections? If there's something crucial I've missed in all this,\n> better let me know ASAP!\n> \n> I know that it's a tad late in the beta cycle to be making such significant\n> changes, but my applications really need NOTIFY to work and work reliably.\n> We have to have this *now*. I hope y'all will bear with me.\n\nSounds good. You have worked with the Notify code more than anyone\nelse, so go ahead. You are fixing a bug, and that is fine at this\nstage. \n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 27 Sep 1998 01:41:32 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Think we need major revisions in async.c..."
},
{
"msg_contents": "> \n> Well, I've spent the last several hours studying how NOTIFY works, and\n> I think it's a mess :-(. We need some significant revisions in there.\n\nI agreee. I tried to improve it but it's really a mess.\n\n> I have not been able to trace an exact flow of control that would\n> yield the NOTIFY-message-interspersed-in-data result that I saw this\n> afternoon. However, I am pretty certain that I know the basic cause\n> of the problem. The risky aspect of NOTIFY is that very substantial\n> amounts of processing (an entire transaction, in fact) are launched from\n> a signal handler routine. As a rule of thumb, doing more than setting a\n> flag variable in a signal handler is a big NO NO, because you don't have\n> any idea what bit of code you've interrupted or what you can safely\n> call. (I have a strong suspicion, in fact, that the behavior I saw\n> arose from the signal interrupt occuring during the backend's fflush()\n\nI suspected this could happen but don't really know how signals are handled.\n\n> of data destined for the frontend at the end of a query cycle. I dunno\n> how re-entrant the stdio library is on my machine, but it seems quite\n> possible that it behaves undesirably if one attempts to queue more\n> output data from a routine that interrupts fflush()...)\n> \n> The Postgres code tries to protect itself against this sort of problem\n> by having the SIGUSR2 handler not do anything except set a flag if the\n> current transaction state says that a transaction is in progress. This\n> is almost good enough, but evidently not quite. The first issue is that\n> some query initialization and finalization code in PostgresMain happens\n> outside the StartTransactionCommand...CommitTransactionCommand zone of\n> safety. (In particular, the final fflush of data bound for the\n> frontend.) The second issue is that async.c itself sometimes starts and\n> ends transactions, by calling xact.c, and yet xact.c also calls async.c\n> as part of transaction commit. At least two levels of recursion are\n> possible in there. I have not convinced myself whether the bug I\n\nI believe that the transaction code can avoid this type of recursion.\n\n> encountered is due to some subtly incorrect ordering of events in that\n> interaction, or whether it's just due to non-reentrancy in stdio.\n> But I no longer care. This code is demonstrably buggy, and it is too\n> complex to understand or maintain anyway. I propose to rewrite it.\n> \n> I propose that:\n> \n> 1. The link between inbound-notify processing and transaction\n> start/commit should be removed. Having inbound-notify execution called\n> from transaction commit is too complicated because that notify execution\n> itself may have to induce a transaction start/commit. Instead, inbound\n> notify execution ought to be triggered from the PostgresMain main loop.\n\nThe reason why it is done at commit is that it is the best place where we\ncan know if the transaction completed or not. If we move the notify code\nin the main loop we must also keep some flag which tell us if and when the\nnotify can be done.\n\n> This has the additional advantage that we can have more control over\n> just which parts of the main loop allow or disallow notify execution\n> directly from the signal handler. The main loop should now contain\n> something like:\n> \n> \tfor (;;) {\n> \n> \t\tReadyForQuery();\t// contains fflush() call\n> \n> \t\tAllowNotifyInterrupt();\n> \n> \t\tReadCommand();\n> \n> \t\tDisableNotifyInterrupt();\n> \n> \t\texecute command, including transaction start/commit\n> \t}\n> \n> so that the SIGUSR2 interrupt is allowed to do something serious\n> only in the *very* tiny window where we are just waiting for or\n> reading the next frontend query. (Since the interrupt has no\n> reason to touch the state of stdin, this should be safe enough.)\n> An interrupt occuring any other time will just set a flag, which\n> will be read by AllowNotifyInterrupt --- if an interrupt has occured\n> since the last DisableNotifyInterrupt, then AllowNotifyInterrupt will\n> carry out the necessary processing. Thus, all inbound-notify execution\n> effectively happens at the outer loop --- either while waiting for\n> frontend input, or just before doing so.\n> \n> \n> 2. The innards of NOTIFY execution are unnecessarily complicated and\n> inefficient, quite aside from the above problem of controlling interrupt\n> safety. The NOTIFY statement itself (routine Async_Notify) makes one\n> pass over the pg_listener relation, writing a 1 into the notification\n> field of each tuple with a matching relname. Then at commit time, we\n> have to make another pass over the pg_listener relation to issue signals\n> (and/or send a message to our own front end, if we're NOTIFYing\n> something we also LISTEN on, which is pretty common in my applications\n> anyway). This is unnecessary processing; worse, if the NOTIFY is done\n> early in a long transaction block, we end up holding a write lock on\n> pg_listen for a long time. Since pg_listen is a global resource, this\n> can effectively bring all the rest of the backends to a halt!\n\nThis is the standard behaviour of the notify code. If you set the flag\nnotifyunlock in .../data/pg_options the write lock is released immediately\nafter updating the table, but I suspect this is the reason of the\nduplicate oids. For this reason I left the flag disabled by default, and\nmaybe this explains why you don't get duplicate oids in 6.4, but I have\nenabled it in a production environment without problems at all.\n\n> The right way to do this is for NOTIFY to simply add the target relation\n> name to a list of pending notifies (as indeed it does already). Then\n> when we are ready to commit, we scan pg_listen *once*, performing tuple\n> updates as needed and also sending signals and/or frontend messages.\n> (Note this is for an *outbound* notify, which we do want to have happen\n> during transaction commit. The real problem with the current async.c is\n> that it tries to combine outbound and inbound notify code. Keeping them\n\nThe real problem is that inbound notifiy can happen at any moment from the\nsignal handler.\n\n> separate will be better.) This approach takes less processing, and it\n> also means that we need not do anything so risky as\n> RelationUnsetLockForWrite() to gain performance. The pg_listen lock\n> will be held for only a very short time before we release it as a normal\n> part of commit. Moreover, a self-notify need never modify the pg_listen\n> table at all, saving cycles and opening the door for another feature\n> (see below).\n\nBut in general you must still get the write lock to prevent deadlocks.\n\n> If the transaction is aborted rather than committed, we need only\n> discard the list of pending notify names. pg_listen is never touched\n> in that case.\n\nThis sounds reasonable.\n\n> 3. While I am at it, I would like to add a feature that I've been\n> wanting for a while. It seems to me that the be_pid field provided in a\n> PGnotify message ought not always be your own backend's PID --- what use\n> is that? What it ought to be is the PID of the *signaling* backend, or\n> at least of one of the signaling backends other than your own. I have\n> a bunch of applications that all read and write the same tables and let\n> each other know about updates by NOTIFY messages. When one of these\n> apps gets a NOTIFY, it has to re-read the table *even if the notify was\n> the one it just issued*, and the updates in question are the ones it\n> just wrote out. This is a waste of cycles, but right now there is no\n> choice because you cannot tell whether the NOTIFY you get is due\n> (solely) to your own NOTIFY or (perhaps in part) to someone else's.\n\nThe problem is that notify has the same semantic of unix signals. You can\nhave more notifies for the same table from many backends but you only know\nthat at least one has happened.\nWhat you propose should be some sort of message queuing between backends\nin which each message is notified separately to the other backend In this\ncase it would also be useful to pass some arbitrary text to the listening\nbackend.\n\n> What I'd like to do is to have cross-backend notifies write the\n> originating backend's PID into the \"notification\" field, rather than\n> just a \"1\". The receiving backend can then send that value to its\n> frontend, as opposed to sending its own PID as it will do in the\n> self-notify case. In this way the frontend can unambiguously\n> distinguish its own NOTIFY bouncing back from a NOTIFY from another\n> client. (If several other clients send NOTIFY at about the same time,\n> you may get only one message with only one of their PIDs, since there\n> is only room for one originating PID in pg_listen. But it will be kept\n> distinct from any message resulting from a NOTIFY of your own.) Note\n> that we already made the backend's own PID available to the frontend as\n> part of the CANCEL-related changes, so that part of the info is there\n> for free.\n> \n> A more radical solution would be to suppress self-notifies altogether,\n> but that is a change in semantics that would likely break some existing\n> apps. I will be satisfied if the frontend can tell the difference\n> between a self-notify and an incoming notify.\n\nI wouldn't change this.\n\n> Any objections? If there's something crucial I've missed in all this,\n> better let me know ASAP!\n> \n> I know that it's a tad late in the beta cycle to be making such significant\n> changes, but my applications really need NOTIFY to work and work reliably.\n> We have to have this *now*. I hope y'all will bear with me.\n> \n> \t\t\tregards, tom lane\n> \n\nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n| Massimo Dal Zotto email: [email protected] |\n| Via Marconi, 141 phone: ++39-461-534251 |\n| 38057 Pergine Valsugana (TN) www: http://www.cs.unitn.it/~dz/ |\n| Italy pgp: finger [email protected] |\n+----------------------------------------------------------------------+\n",
"msg_date": "Mon, 28 Sep 1998 18:39:47 +0200 (MET DST)",
"msg_from": "Massimo Dal Zotto <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Think we need major revisions in async.c..."
},
{
"msg_contents": "Massimo Dal Zotto <[email protected]> writes:\n>> 1. The link between inbound-notify processing and transaction\n>> start/commit should be removed. Having inbound-notify execution called\n>> from transaction commit is too complicated because that notify execution\n>> itself may have to induce a transaction start/commit. Instead, inbound\n>> notify execution ought to be triggered from the PostgresMain main loop.\n\n> The reason why it is done at commit is that it is the best place where we\n> can know if the transaction completed or not. If we move the notify code\n> in the main loop we must also keep some flag which tell us if and when the\n> notify can be done.\n\nNo, you're confusing outbound and inbound notify. Outbound notify ---\nie, actually doing something about a NOTIFY command issued by the user\n--- should indeed be done from transaction commit. If the transaction\nis aborted, we want the NOTIFY to never take effect, just like all the\nother side-effects of the transaction. What I proposed for that was\nthat we could simplify matters by doing both the signal-sending and\npg_listener-table-updating work at the same time during transaction\ncommit, rather than changing the table and issuing the signals in\nseparate passes.\n\nInbound notify, ie, receiving messages from other backends, is a\ndifferent story. There is no question of not accepting the signals;\nthey've already been committed by some other backend. Now for\nimplementation reasons we need to hold off responding to an inbound\nnotify until we are outside of any transaction. (Otherwise, our\nattempt to mark the notify accepted by changing pg_listener could be\nundone by a later transaction abort; not what we want.) But that\ndoesn't mean the response processing has to be called from xact.c.\nBetter to do it at the main loop, with a check to see if we are inside\na transaction block or not. Otherwise we have to depend on stdio\nto be re-entrant. Also, xact.c is already tricky enough without having\nto worry about the exact order of operations in straightline code.\nI spent some time on Saturday trying to see if the variables the signal\nhandler was looking at were being changed before the transaction was\nquite complete --- and I'm still not sure whether it was really safe,\ngiven the potential for mutual recursion between async.c and xact.c\nin the old code. Writing code that interacts correctly with an ISR is\ntricky stuff, so you really want to do it in a place that doesn't have\nmuch of anything else to worry about.\n\n> This is the standard behaviour of the notify code. If you set the flag\n> notifyunlock in .../data/pg_options the write lock is released immediately\n> after updating the table, but I suspect this is the reason of the\n> duplicate oids. For this reason I left the flag disabled by default, and\n> maybe this explains why you don't get duplicate oids in 6.4, but I have\n> enabled it in a production environment without problems at all.\n\nNeither NotifyUnlock nor NotifyHack are necessary in the rewritten\ncode...\n\n>> This is a waste of cycles, but right now there is no\n>> choice because you cannot tell whether the NOTIFY you get is due\n>> (solely) to your own NOTIFY or (perhaps in part) to someone else's.\n\n> What you propose should be some sort of message queuing between backends\n> in which each message is notified separately to the other backend In this\n> case it would also be useful to pass some arbitrary text to the listening\n> backend.\n\nNot really. For the applications I'm interested in, there are only two\ninteresting cases: a NOTIFY from some other client vs. one of your own\nNOTIFYs bouncing back at you. With the code as I now have it, you can\ndifferentiate those cases. You're right that a more general\nmessage-passing facility might be useful. On the other hand, one\ncan be built by users using an ordinary table and NOTIFY to wake up\nsleeping respondents. So I'm not sure we need to put more stuff in\nthe backend for that.\n\nI have been testing this new code and it seems to work fine, except for\nthe question of interlocks against \"select * from pg_listener\" --- and\nif that's a bug, it's one that was there before anyway. So I'm ready\nto check it in. (Marc, are you there?)\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 28 Sep 1998 15:15:06 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Think we need major revisions in async.c... "
}
] |
[
{
"msg_contents": "Hello,\nI have query:\nUPDATE userd_session_stat SET status =1 WHERE status=0 AND ((uid <>627 AND\ntty <>'ttyA03') OR (uid <> 425 AND tty <> 'ttyA05') OR (uid <> 8011 AND tty\n<> 'ttyA09') OR (uid <> 2092 AND tty <> 'ttyA0f') OR (uid <> 249 AND tty <>\n'ttyp3') OR (uid <> 249 AND tty <> 'ttyp4') OR (uid <> 249 AND tty <>\n'ttyp5') OR (uid <> 249 AND tty <> 'ttyp6'))\n\nBut, postgres complains that:\nFATAL 1: palloc failure: memory exhausted\n\n I see, the query must be less than 4kB, and this query is less.\nLong SELECT queries works fine.\nHave any idea? Maybe, I have to change postmaster's settings ? Query\nexecutes from libpg programm.\n\nThanx,\n---------------------------\nSergei Chernev\nInternet: [email protected]\nPhone: +7-3832-397354\n\n",
"msg_date": "Sun, 27 Sep 1998 15:20:35 +0600",
"msg_from": "\"Sergei Chernev\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Long update query ?"
},
{
"msg_contents": "This is caused by a semi-well known weakness in the optimizer. The optimizer\nrewrites the WHERE clause in conjunctive normal form (CNF):\n\n (A and B) or (C and D) ==> (A or C) and (A or D) and (B or C) and (B or D)\n\nTry this with your statement and you will see the expression explodes. Foe\nnow, I would suggest that you break this up into multiple statements.\n\nSergei Chernev wrote:\n\n> Hello,\n> I have query:\n> UPDATE userd_session_stat SET status =1 WHERE status=0 AND ((uid <>627 AND\n> tty <>'ttyA03') OR (uid <> 425 AND tty <> 'ttyA05') OR (uid <> 8011 AND tty\n> <> 'ttyA09') OR (uid <> 2092 AND tty <> 'ttyA0f') OR (uid <> 249 AND tty <>\n> 'ttyp3') OR (uid <> 249 AND tty <> 'ttyp4') OR (uid <> 249 AND tty <>\n> 'ttyp5') OR (uid <> 249 AND tty <> 'ttyp6'))\n>\n> But, postgres complains that:\n> FATAL 1: palloc failure: memory exhausted\n>\n> I see, the query must be less than 4kB, and this query is less.\n> Long SELECT queries works fine.\n> Have any idea? Maybe, I have to change postmaster's settings ? Query\n> executes from libpg programm.\n\n\n\n",
"msg_date": "Mon, 28 Sep 1998 09:21:08 -0400",
"msg_from": "David Hartwig <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [GENERAL] Long update query ?"
},
{
"msg_contents": "> This is caused by a semi-well known weakness in the optimizer.\n> The optimizer\n> rewrites the WHERE clause in conjunctive normal form (CNF):\n>\n> (A and B) or (C and D) ==> (A or C) and (A or D) and (B or C)\n> and (B or D)\n>\n\nWouldn't disjunctive normal form be better, since it can be implemented as\nthe simple union of a set of small queries?\n\nTaral\n\n",
"msg_date": "Mon, 28 Sep 1998 12:15:01 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [GENERAL] Long update query ?"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > This is caused by a semi-well known weakness in the optimizer.\n> > The optimizer\n> > rewrites the WHERE clause in conjunctive normal form (CNF):\n> >\n> > (A and B) or (C and D) ==> (A or C) and (A or D) and (B or C)\n> > and (B or D)\n> >\n> \n> Wouldn't disjunctive normal form be better, since it can be implemented as\n> the simple union of a set of small queries?\n\nPlease tell us more.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 18:54:19 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [GENERAL] Long update query ?"
},
{
"msg_contents": "> > Wouldn't disjunctive normal form be better, since it can be\n> implemented as\n> > the simple union of a set of small queries?\n>\n> Please tell us more.\n\nWell, I don't know how the backend processes queries, but one can imagine\nthis scenario (for DNF):\n\n1) Analyze query and set up columns in result table\n2) Rewrite query into DNF\n3) Split query into subqueries\n4) For each subquery:\n a) Process query\n b) Append matching tuples to result table\n5) Do any post-processing (ORDER BY, etc.)\n6) Return result\n\nHow is the processing currently done (with CNF)?\n\nTaral\n\n",
"msg_date": "Thu, 1 Oct 1998 18:45:33 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > > Wouldn't disjunctive normal form be better, since it can be\n> > implemented as\n> > > the simple union of a set of small queries?\n> >\n> > Please tell us more.\n> \n> Well, I don't know how the backend processes queries, but one can imagine\n> this scenario (for DNF):\n> \n> 1) Analyze query and set up columns in result table\n> 2) Rewrite query into DNF\n> 3) Split query into subqueries\n> 4) For each subquery:\n> a) Process query\n> b) Append matching tuples to result table\n> 5) Do any post-processing (ORDER BY, etc.)\n> 6) Return result\n> \n> How is the processing currently done (with CNF)?\n\nIt currently convert to CNF so it can select the most restrictive\nrestriction and join, and use those first. However, the CNF conversion\nis a memory exploder for some queries, and we certainly need to have\nanother method to split up those queries into UNIONS. I think we need\nto code to identify those queries capable of being converted to UNIONS,\nand do that before the query gets to the CNF section. That would be\ngreat, and David Hartwig has implemented a limited capability of doing\nthis, but we really need a general routine to do this with 100%\nreliability.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 22:18:29 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)"
},
{
"msg_contents": "> It currently convert to CNF so it can select the most restrictive\n> restriction and join, and use those first. However, the CNF conversion\n> is a memory exploder for some queries, and we certainly need to have\n> another method to split up those queries into UNIONS. I think we need\n> to code to identify those queries capable of being converted to UNIONS,\n> and do that before the query gets to the CNF section. That would be\n> great, and David Hartwig has implemented a limited capability of doing\n> this, but we really need a general routine to do this with 100%\n> reliability.\n\nWell, if you're talking about a routine to generate a heuristic for CNF vs.\nDNF, it is possible to precalculate the query sizes for CNF and DNF\nrewrites...\n\nFor conversion to CNF:\n\nAt every node:\n\nif nodeType = AND then f(node) = f(left) + f(right)\nif nodeType = OR then f(node) = f(left) * f(right)\n\nf(root) = a reasonably (but not wonderful) metric\n\nFor DNF just switch AND and OR in the above. You may want to compute both\nmetrics and compare... take the smaller one and use that path.\n\nHow to deal with other operators depends on their implementation...\n\nTaral\n\n",
"msg_date": "Fri, 2 Oct 1998 00:35:12 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > It currently convert to CNF so it can select the most restrictive\n> > restriction and join, and use those first. However, the CNF conversion\n> > is a memory exploder for some queries, and we certainly need to have\n> > another method to split up those queries into UNIONS. I think we need\n> > to code to identify those queries capable of being converted to UNIONS,\n> > and do that before the query gets to the CNF section. That would be\n> > great, and David Hartwig has implemented a limited capability of doing\n> > this, but we really need a general routine to do this with 100%\n> > reliability.\n> \n> Well, if you're talking about a routine to generate a heuristic for CNF vs.\n> DNF, it is possible to precalculate the query sizes for CNF and DNF\n> rewrites...\n> \n> For conversion to CNF:\n> \n> At every node:\n> \n> if nodeType = AND then f(node) = f(left) + f(right)\n> if nodeType = OR then f(node) = f(left) * f(right)\n> \n> f(root) = a reasonably (but not wonderful) metric\n> \n> For DNF just switch AND and OR in the above. You may want to compute both\n> metrics and compare... take the smaller one and use that path.\n> \n> How to deal with other operators depends on their implementation...\n\n[Moved to Hackers list.]\n\nThis is interesting. Check CNF size and DNF size. Choose smallest. \nCNF uses existing code, DNF converts to UNIONs. How do you return the\nproper rows with/without proper duplicates?\n\ni.e.\n\n\tSELECT * FROM tab1 WHERE x > 1 or x > 2\n\nWe need to return all rows where x > 1, even if some there are indentical\nrows in tab1.\n\nWhat I do in the index OR code is to test that rows in index matches\nfound in 2nd and 3rd index scans are false in earlier index scans. I am\nnot sure how to do that with a UNION query, but it may be possible.\n\nWe currently have UNION and UNION ALL, and I think we may need a new\nUNION type internally to prevent 2nd and 3rd queries from returning rows\nreturned by earlier UNION queries.\n\nThis is interesting.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 02:02:48 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)"
},
{
"msg_contents": "> This is interesting. Check CNF size and DNF size. Choose smallest. \n> CNF uses existing code, DNF converts to UNIONs. How do you return the\n> proper rows with/without proper duplicates?\n\nCreate a temporary oid hash? (for each table selected on, I guess)\n\nTaral\n",
"msg_date": "Fri, 2 Oct 1998 11:37:15 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > This is interesting. Check CNF size and DNF size. Choose smallest. \n> > CNF uses existing code, DNF converts to UNIONs. How do you return the\n> > proper rows with/without proper duplicates?\n> \n> Create a temporary oid hash? (for each table selected on, I guess)\n> \n> Taral\n> \n\nWhat I did with indexes was to run the previous OR clause index\nrestrictions through the qualification code, and make sure it failed,\nbut I am not sure how that is going to work with a more complex WHERE\nclause. Perhaps I need to restrict this to just simple cases of\nconstants, which are easy to pick out an run through. Doing this with\njoins would be very hard, I think.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 12:40:56 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)"
},
{
"msg_contents": "> > Create a temporary oid hash? (for each table selected on, I guess)\n>\n> What I did with indexes was to run the previous OR clause index\n> restrictions through the qualification code, and make sure it failed,\n> but I am not sure how that is going to work with a more complex WHERE\n> clause. Perhaps I need to restrict this to just simple cases of\n> constants, which are easy to pick out an run through. Doing this with\n> joins would be very hard, I think.\n\nActually, I was thinking more of an index of returned rows... After each\nsubquery, the backend would check each row to see if it was already in the\nindex... Simple duplicate check, in other words. Of course, I don't know how\nwell this would behave with large tables being returned...\n\nAnyone else have some ideas they want to throw in?\n\nTaral\n\n",
"msg_date": "Fri, 2 Oct 1998 11:47:48 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [GENERAL] Long update query ? (also Re: [GENERAL] CNF vs. DNF)"
},
{
"msg_contents": "> \n> > > Create a temporary oid hash? (for each table selected on, I guess)\n> >\n> > What I did with indexes was to run the previous OR clause index\n> > restrictions through the qualification code, and make sure it failed,\n> > but I am not sure how that is going to work with a more complex WHERE\n> > clause. Perhaps I need to restrict this to just simple cases of\n> > constants, which are easy to pick out an run through. Doing this with\n> > joins would be very hard, I think.\n> \n> Actually, I was thinking more of an index of returned rows... After each\n> subquery, the backend would check each row to see if it was already in the\n> index... Simple duplicate check, in other words. Of course, I don't know how\n> well this would behave with large tables being returned...\n> \n> Anyone else have some ideas they want to throw in?\n> \n> Taral\n> \n\n But what about unions of join queries? Which OID's then should\n be checked against which? And unions from view selects? There\n are no OID's at all after rewriting.\n\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n",
"msg_date": "Fri, 2 Oct 1998 19:31:18 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > > Create a temporary oid hash? (for each table selected on, I guess)\n> >\n> > What I did with indexes was to run the previous OR clause index\n> > restrictions through the qualification code, and make sure it failed,\n> > but I am not sure how that is going to work with a more complex WHERE\n> > clause. Perhaps I need to restrict this to just simple cases of\n> > constants, which are easy to pick out an run through. Doing this with\n> > joins would be very hard, I think.\n> \n> Actually, I was thinking more of an index of returned rows... After each\n> subquery, the backend would check each row to see if it was already in the\n> index... Simple duplicate check, in other words. Of course, I don't know how\n> well this would behave with large tables being returned...\n> \n> Anyone else have some ideas they want to throw in?\n\nI certainly think we are heading in the direction for a good general\nsolution.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 13:40:39 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> > \n> > > > Create a temporary oid hash? (for each table selected on, I guess)\n> > >\n> > > What I did with indexes was to run the previous OR clause index\n> > > restrictions through the qualification code, and make sure it failed,\n> > > but I am not sure how that is going to work with a more complex WHERE\n> > > clause. Perhaps I need to restrict this to just simple cases of\n> > > constants, which are easy to pick out an run through. Doing this with\n> > > joins would be very hard, I think.\n> > \n> > Actually, I was thinking more of an index of returned rows... After each\n> > subquery, the backend would check each row to see if it was already in the\n> > index... Simple duplicate check, in other words. Of course, I don't know how\n> > well this would behave with large tables being returned...\n> > \n> > Anyone else have some ideas they want to throw in?\n> > \n> > Taral\n> > \n> \n> But what about unions of join queries? Which OID's then should\n> be checked against which? And unions from view selects? There\n> are no OID's at all after rewriting.\n\nYep, you can't just use oid's, I think. Joins and specifiying a table\nmultiple times using a table alias would break this anyway.\n\nCNF'ify only goes through the tables once, so we somehow need to\nsimulate this. Perhaps we can restrict the kinds of queries used for\nDNF so we can do this easily. \n\nAnother idea is that we rewrite queries such as:\n\n\tSELECT *\n\tFROM tab\n\tWHERE (a=1 AND b=2 AND c=3) OR\n\t (a=1 AND b=2 AND c=4) OR\n\t (a=1 AND b=2 AND c=5) OR\n\t (a=1 AND b=2 AND c=6)\n\ninto:\n\n\tSELECT *\n\tFROM tab\n\tWHERE (a=1 AND b=2) AND (c=3 OR c=4 OR c=5 OR c=6)\n\nand we do this BEFORE calling cnfify(). How much does this do for us?\n\nSeems this would not be too hard, and would be a good performer.\n\n\nYou could even convert:\n\n\tSELECT *\n\tFROM tab\n\tWHERE (a=1 AND b=2 AND c=3) OR\n\t (a=1 AND b=2 AND c=4) OR\n\t (a=1 AND b=52 AND c=5) OR\n\t (a=1 AND b=52 AND c=6)\n\ninto:\n\n\tSELECT *\n\tFROM tab\n\tWHERE ((a=1 AND b=2) AND (c=3 OR c=4)) OR\n\tWHERE ((a=1 AND b=52) AND (c=5 OR c=6))\n\nThis should work OK too. Someone want to try this? David, is this what\nyour code does?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 13:58:25 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> Another idea is that we rewrite queries such as:\n>\n> \tSELECT *\n> \tFROM tab\n> \tWHERE (a=1 AND b=2 AND c=3) OR\n> \t (a=1 AND b=2 AND c=4) OR\n> \t (a=1 AND b=2 AND c=5) OR\n> \t (a=1 AND b=2 AND c=6)\n>\n> into:\n>\n> \tSELECT *\n> \tFROM tab\n> \tWHERE (a=1 AND b=2) AND (c=3 OR c=4 OR c=5 OR c=6)\n\nVery nice, but that's like trying to code factorization of numbers... not\npretty, and very CPU intensive on complex queries...\n\nTaral\n\n",
"msg_date": "Fri, 2 Oct 1998 16:12:19 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > Another idea is that we rewrite queries such as:\n> >\n> > \tSELECT *\n> > \tFROM tab\n> > \tWHERE (a=1 AND b=2 AND c=3) OR\n> > \t (a=1 AND b=2 AND c=4) OR\n> > \t (a=1 AND b=2 AND c=5) OR\n> > \t (a=1 AND b=2 AND c=6)\n> >\n> > into:\n> >\n> > \tSELECT *\n> > \tFROM tab\n> > \tWHERE (a=1 AND b=2) AND (c=3 OR c=4 OR c=5 OR c=6)\n> \n> Very nice, but that's like trying to code factorization of numbers... not\n> pretty, and very CPU intensive on complex queries...\n\nYes, but how large are the WHERE clauses going to be? Considering the\ncost of cnfify() and UNION, it seems like a clear win. Is it general\nenough to solve our problems?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 17:39:05 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> > Very nice, but that's like trying to code factorization of\n> numbers... not\n> > pretty, and very CPU intensive on complex queries...\n>\n> Yes, but how large are the WHERE clauses going to be? Considering the\n> cost of cnfify() and UNION, it seems like a clear win. Is it general\n> enough to solve our problems?\n\nCould be... the examples I received where the cnfify() was really bad were\ncases where the query was submitted alredy in DNF... and where the UNION was\na simple one. However, I don't know of any algorithms for generic\nsimplification of logical constraints. One problem is resolution/selection\nof factors:\n\nSELECT * FROM a WHERE (a = 1 AND b = 2 AND c = 3) OR (a = 4 AND b = 2 AND c\n= 3) OR (a = 1 AND b = 5 AND c = 3) OR (a = 1 AND b = 2 AND c = 6);\n\nTry that on for size. You can understand why that code gets ugly, fast.\nSomebody could try coding it, but it's not a clear win to me.\n\nMy original heuristic was missing one thing: \"Where the heuristic fails to\nprocess or decide, default to CNF.\" Since that's the current behavior, we're\nless likely to break things.\n\nTaral\n\n",
"msg_date": "Fri, 2 Oct 1998 16:49:26 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > > Very nice, but that's like trying to code factorization of\n> > numbers... not\n> > > pretty, and very CPU intensive on complex queries...\n> >\n> > Yes, but how large are the WHERE clauses going to be? Considering the\n> > cost of cnfify() and UNION, it seems like a clear win. Is it general\n> > enough to solve our problems?\n> \n> Could be... the examples I received where the cnfify() was really bad were\n> cases where the query was submitted alredy in DNF... and where the UNION was\n> a simple one. However, I don't know of any algorithms for generic\n> simplification of logical constraints. One problem is resolution/selection\n> of factors:\n> \n> SELECT * FROM a WHERE (a = 1 AND b = 2 AND c = 3) OR (a = 4 AND b = 2 AND c\n> = 3) OR (a = 1 AND b = 5 AND c = 3) OR (a = 1 AND b = 2 AND c = 6);\n> \n> Try that on for size. You can understand why that code gets ugly, fast.\n> Somebody could try coding it, but it's not a clear win to me.\n> \n> My original heuristic was missing one thing: \"Where the heuristic fails to\n> process or decide, default to CNF.\" Since that's the current behavior, we're\n> less likely to break things.\n\nOK, but if we use UNION, how to we return the proper rows? Is there any\nsolution for that, and we are executing the query over and over again. \nAny factoring would be faster than running those multiple queries,\nwouldn't it?\n\nAlso, I amagine the case where we are doing a join, so we have:\n\n\tSELECT *\n\tFROM tab1, tab2\n\tWHERE tab1.col1 = tab2.col2 AND\n\t ((a=1 and b=2 and c=3) OR\n\t (a=1 and b=2 and c=4))\n\nHow do we do that with UNION, and return the right rows. Seems the\n_join_ happending multiple times would be much worse than the factoring.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 17:57:15 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> How do we do that with UNION, and return the right rows. Seems the\n> _join_ happending multiple times would be much worse than the factoring.\n\nOk... We have two problems:\n\n1) DNF for unjoined queries.\n2) Factorization for the rest.\n\nI have some solutions for (1). Not for (2). Remember that unjoined queries\nare quite common. :)\n\nFor (1), we can always try to parallel the multiple queries... especially in\nthe case where a sequential search is required.\n\nTaral\n\n",
"msg_date": "Fri, 2 Oct 1998 17:04:24 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > How do we do that with UNION, and return the right rows. Seems the\n> > _join_ happending multiple times would be much worse than the factoring.\n> \n> Ok... We have two problems:\n> \n> 1) DNF for unjoined queries.\n> 2) Factorization for the rest.\n> \n> I have some solutions for (1). Not for (2). Remember that unjoined queries\n> are quite common. :)\n> \n> For (1), we can always try to parallel the multiple queries... especially in\n> the case where a sequential search is required.\n\nI don't know how to return the proper rows with UNION.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 23:01:07 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "\nI have another idea.\n\nWhen we cnfify, this:\n\n\t(A AND B) OR (C AND D)\n\nbecomes\n\n\t(A OR C) AND (A OR D) AND (B OR C) AND (B OR D)\n\nhowever if A and C are identical, this could become:\n\n\t(A OR A) AND (A OR D) AND (B OR A) AND (B OR D)\n\nand A OR A is A:\n\n\tA AND (A OR D) AND (B OR A) AND (B OR D)\n\nand since we are now saying A has to be true, we can remove OR's with A:\n\n\tA AND (B OR D)\n\nMuch smaller, and a big win for queries like:\n\n\tSELECT *\n\tFROM tab\n\tWHERE\t(a=1 AND b=2) OR\n\t\t(a=1 AND b=3)\n\nThis becomes:\n\n\t\t(a=1) AND (b=2 OR b=3)\n\nwhich is accurate, and uses our OR indexing.\n\nSeems I could code cnfify() to look for identical qualifications in two\njoined OR clauses and remove the duplicates.\n\nSound like big win, and fairly easy and inexpensive in processing time.\n\nComments?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sat, 3 Oct 1998 21:57:30 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> however if A and C are identical, this could become:\n>\n> \t(A OR A) AND (A OR D) AND (B OR A) AND (B OR D)\n>\n> and A OR A is A:\n>\n> \tA AND (A OR D) AND (B OR A) AND (B OR D)\n>\n> and since we are now saying A has to be true, we can remove OR's with A:\n>\n> \tA AND (B OR D)\n\nVery nice... and you could do that after each iteration of the rewrite,\npreventing the size from getting too big. :)\n\nI have a symbolic expression tree evaluator that would be perfect for\nthis... I'll see if I can't adapt it.\n\nCan someone mail me the structures for expression trees? I don't want to\nhave to excise them from the source. Please?\n\nTaral\n\n",
"msg_date": "Sat, 3 Oct 1998 21:26:20 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> > however if A and C are identical, this could become:\n> >\n> > \t(A OR A) AND (A OR D) AND (B OR A) AND (B OR D)\n> >\n> > and A OR A is A:\n> >\n> > \tA AND (A OR D) AND (B OR A) AND (B OR D)\n> >\n> > and since we are now saying A has to be true, we can remove OR's with A:\n> >\n> > \tA AND (B OR D)\n> \n> Very nice... and you could do that after each iteration of the rewrite,\n> preventing the size from getting too big. :)\n> \n> I have a symbolic expression tree evaluator that would be perfect for\n> this... I'll see if I can't adapt it.\n> \n> Can someone mail me the structures for expression trees? I don't want to\n> have to excise them from the source. Please?\n\nThat is very hard to do. We have lots of structures involved. I\nrecommend you look at backend/optimizer/prep/prepqual.c. That has the\nCNF'ify code, and I am studying it now. There are supporing functions\non backend/nodes that will allow comparisons of many structures.\n\nWe may not be that far off. normalize() does much of the work, and\nqual_cleanup() reomves duplicates using remove_duplicates(), but\nqual_cleanup() is only called after normalize completes, not during the\nnormalization, which seems to be the problem. If we can remove the\nduplicates BEFORE the OR explosion, we are much better off.\n\nYou can then use ctags to jump around to see the supporting structures.\nSee the developers FAQ in the web site or doc directory.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sat, 3 Oct 1998 23:32:14 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "\n\nBruce Momjian wrote:\n\n> I have another idea.\n>\n> When we cnfify, this:\n>\n> (A AND B) OR (C AND D)\n>\n> becomes\n>\n> (A OR C) AND (A OR D) AND (B OR C) AND (B OR D)\n>\n> however if A and C are identical, this could become:\n>\n> (A OR A) AND (A OR D) AND (B OR A) AND (B OR D)\n>\n> and A OR A is A:\n>\n> A AND (A OR D) AND (B OR A) AND (B OR D)\n>\n> and since we are now saying A has to be true, we can remove OR's with A:\n>\n> A AND (B OR D)\n>\n> Much smaller, and a big win for queries like:\n>\n> SELECT *\n> FROM tab\n> WHERE (a=1 AND b=2) OR\n> (a=1 AND b=3)\n>\n> This becomes:\n>\n> (a=1) AND (b=2 OR b=3)\n>\n> which is accurate, and uses our OR indexing.\n>\n> Seems I could code cnfify() to look for identical qualifications in two\n> joined OR clauses and remove the duplicates.\n>\n> Sound like big win, and fairly easy and inexpensive in processing time.\n>\n> Comments?\n\nApologies for not commenting sooner. I have been incognito.\n\nAs to your earlier question, Bruce, the KSQO patch rewrites qualifying\nqueries as UNIONs.\n\n (A AND B) OR (C AND D) ==> (A AND B) UNION (C AND D)\n\nThe rules to qualify are fairly strict. Must be have ANDs; rectangular in\nshape; all (VAR op CONST) type nodes; minimum of 10 nodes; etc. I was\ntargeting the keysets queries generated by ODBC tools.\n\nAs for the current direction this thread is going, (factoring) I have one\nword of caution. PREPARE. If you take this route, you will never be able\nto implement a workable PREPARE statement. I believe that in order for\nPostgrerSQL ever become a industrial strength client/server it must implement\na PREPARE statement with parameters.\n\n\n",
"msg_date": "Mon, 05 Oct 1998 10:33:37 -0400",
"msg_from": "David Hartwig <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> (A AND B) OR (C AND D) ==> (A AND B) UNION (C AND D)\n> \n> The rules to qualify are fairly strict. Must be have ANDs; rectangular in\n> shape; all (VAR op CONST) type nodes; minimum of 10 nodes; etc. I was\n> targeting the keysets queries generated by ODBC tools.\n> \n> As for the current direction this thread is going, (factoring) I have one\n> word of caution. PREPARE. If you take this route, you will never be able\n> to implement a workable PREPARE statement. I believe that in order for\n> PostgrerSQL ever become a industrial strength client/server it must implement\n> a PREPARE statement with parameters.\n\nI see that adding nodes it going to mess up prepare, but we already add\nextra nodes as part of part of \"col in (1, 2, 3).\"\n\nI think the PARAM's we already use will be duplicated/removed and still\nretain their values for substitution. They just may be in a different\norder.\n\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Mon, 5 Oct 1998 22:23:13 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "\n\nBruce Momjian wrote:\n\n> > (A AND B) OR (C AND D) ==> (A AND B) UNION (C AND D)\n> >\n> > The rules to qualify are fairly strict. Must be have ANDs; rectangular in\n> > shape; all (VAR op CONST) type nodes; minimum of 10 nodes; etc. I was\n> > targeting the keysets queries generated by ODBC tools.\n> >\n> > As for the current direction this thread is going, (factoring) I have one\n> > word of caution. PREPARE. If you take this route, you will never be able\n> > to implement a workable PREPARE statement. I believe that in order for\n> > PostgrerSQL ever become a industrial strength client/server it must implement\n> > a PREPARE statement with parameters.\n>\n> I see that adding nodes it going to mess up prepare, but we already add\n> extra nodes as part of part of \"col in (1, 2, 3).\"\n\nIt's not extra nodes I am worried about. It is factored out nodes.\n\n> I think the PARAM's we already use will be duplicated/removed and still\n> retain their values for substitution. They just may be in a different\n> order.\n\nI realize I may be stretching the point, since I brought it up I will complete my\nthoughts. Now, you may understand this, but just to be sure. Here is a typical\nclient/server scenario:\n\n - prepare statement S\n - retrieve result description of S\n - retrieve number of parameters of S\n - retrieve parameter descriptions of S\n - put data into parameters of S\n - execute S\n - retrieve result\n [REPEAT]\n - put different data into parameters of S\n - execute S\n - retrieve result\n [END REPEAT]\n - free statement S\n\nThe problem is that you cannot depend upon factoring to reduce these complex\nstatements. We need to retain a place holder (pointer) for each passed\nparameter. Otherwise we need to re-(parse and plan) the statement before each\nexecution; thus, loosing one of the major benefits of PREPARE.\n\nThe other major benefits are:\n1. Gaining access to the statement result description w/o having to actually\nexecute the statement. Client/server tools live off this stuff.\n2. Smaller statement size. The parameters in the WHERE clause can be sent to that\nbackend in separate chunks.\nBack to the subject at hand.\n\nMy point is that the factoring approach may be a bit short sighted in the long term\nevolution of PostgreSQL.\n\n",
"msg_date": "Tue, 06 Oct 1998 11:28:22 -0400",
"msg_from": "David Hartwig <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> > I see that adding nodes it going to mess up prepare, but we already add\n> > extra nodes as part of part of \"col in (1, 2, 3).\"\n> \n> It's not extra nodes I am worried about. It is factored out nodes.\n> \n> > I think the PARAM's we already use will be duplicated/removed and still\n> > retain their values for substitution. They just may be in a different\n> > order.\n> \n> I realize I may be stretching the point, since I brought it up I will complete my\n> thoughts. Now, you may understand this, but just to be sure. Here is a typical\n> client/server scenario:\n> \n> - prepare statement S\n> - retrieve result description of S\n> - retrieve number of parameters of S\n> - retrieve parameter descriptions of S\n> - put data into parameters of S\n> - execute S\n> - retrieve result\n> [REPEAT]\n> - put different data into parameters of S\n> - execute S\n> - retrieve result\n> [END REPEAT]\n> - free statement S\n> \n> The problem is that you cannot depend upon factoring to reduce these complex\n> statements. We need to retain a place holder (pointer) for each passed\n> parameter. Otherwise we need to re-(parse and plan) the statement before each\n> execution; thus, loosing one of the major benefits of PREPARE.\n> \n> The other major benefits are:\n> 1. Gaining access to the statement result description w/o having to actually\n> execute the statement. Client/server tools live off this stuff.\n> 2. Smaller statement size. The parameters in the WHERE clause can be sent to that\n> backend in separate chunks.\n> Back to the subject at hand.\n> \n> My point is that the factoring approach may be a bit short sighted in the long term\n> evolution of PostgreSQL.\n\nYikes. I see what you mean. The factoring of one query with certain\nconstants will be different than another query. That will certainly be\na problem.\n\nI still haven't had time to look over the cnfify code, to see if calling\nqual_cleanup earlier in the code will help reduce the palloc failures. \nIf it is easy to do, I will implement it, and we can remove it or change\nit once we start looking at prepared queries.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Tue, 6 Oct 1998 22:41:21 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> The problem is that you cannot depend upon factoring to reduce these complex\n> statements. We need to retain a place holder (pointer) for each passed\n> parameter. Otherwise we need to re-(parse and plan) the statement before each\n> execution; thus, loosing one of the major benefits of PREPARE.\n\nI think we already have such a problem. When using optimization\nstatistics, the optimizer checks the value of the constant to determine\nhow many rows will be returned by a \"x > 10\" by looking at the min/max\nvalues for the column. Prepared queries where this value will change\nwould make that a problem.\n\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Wed, 7 Oct 1998 13:56:18 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "\n\nBruce Momjian wrote:\n\n> > The problem is that you cannot depend upon factoring to reduce these complex\n> > statements. We need to retain a place holder (pointer) for each passed\n> > parameter. Otherwise we need to re-(parse and plan) the statement before each\n> > execution; thus, loosing one of the major benefits of PREPARE.\n>\n> I think we already have such a problem. When using optimization\n> statistics, the optimizer checks the value of the constant to determine\n> how many rows will be returned by a \"x > 10\" by looking at the min/max\n> values for the column. Prepared queries where this value will change\n> would make that a problem.\n\nGad Zooks. The future is here. I wonder how Vadim's SPI_Prepare() will respond\nto this. I have not used it much, but I believe it accepts parameters.\n\nFor that matter, I seem to recall some kind of reduction going on in the query\nplan. In 6.3.2 something like:\n\n -- with an index on bar\n EXPLAIN SELECT stuff FROM foo WHERE bar = 1 OR bar = 2;\n -- does not use index; this is expected in 6.3.2\n\n EXPLAIN SELECT stuff FROM foo WHERE bar = 1 OR bar = 1;\n -- uses index; I speculated on some reduction going on here.\n\n...\n\nI just tried it with on out with our corp (6.3.2) database. _day is an indexed field\non dates.\n\ncorp=> explain select * from dates where _day = '1/1/99';\nNOTICE: QUERY PLAN:\n\nIndex Scan on dates (cost=2.05 size=1 width=24)\n\nEXPLAIN\ncorp=> explain select * from dates where _day = '1/1/99' or _day = '1/1/99';\nNOTICE: QUERY PLAN:\n\nIndex Scan on dates (cost=2.05 size=1 width=24)\n\nEXPLAIN\ncorp=> explain select * from dates where _day = '1/1/99' or _day = '1/2/99';\nNOTICE: QUERY PLAN:\n\nSeq Scan on dates (cost=91.27 size=219 width=24)\n\n\nSPI_prepare may need to be tested, along with your example, to see how it responds.\n\n",
"msg_date": "Wed, 07 Oct 1998 16:34:15 -0400",
"msg_from": "David Hartwig <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
},
{
"msg_contents": "> Gad Zooks. The future is here. I wonder how Vadim's SPI_Prepare() will respond\n> to this. I have not used it much, but I believe it accepts parameters.\n> \n> For that matter, I seem to recall some kind of reduction going on in the query\n> plan. In 6.3.2 something like:\n> \n> -- with an index on bar\n> EXPLAIN SELECT stuff FROM foo WHERE bar = 1 OR bar = 2;\n> -- does not use index; this is expected in 6.3.2\n> \n> EXPLAIN SELECT stuff FROM foo WHERE bar = 1 OR bar = 1;\n> -- uses index; I speculated on some reduction going on here.\n> \n> ...\n> \n> I just tried it with on out with our corp (6.3.2) database. _day is an indexed field\n> on dates.\n> \n> corp=> explain select * from dates where _day = '1/1/99';\n> NOTICE: QUERY PLAN:\n> \n> Index Scan on dates (cost=2.05 size=1 width=24)\n> \n> EXPLAIN\n> corp=> explain select * from dates where _day = '1/1/99' or _day = '1/1/99';\n> NOTICE: QUERY PLAN:\n> \n> Index Scan on dates (cost=2.05 size=1 width=24)\n\nI believe this reduction is done by cnfify when it removes duplicates as\nits last step.\n\n\n> \n> EXPLAIN\n> corp=> explain select * from dates where _day = '1/1/99' or _day = '1/2/99';\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on dates (cost=91.27 size=219 width=24)\n> \n\nYes, sure looks like that is what is happening.\n> \n> SPI_prepare may need to be tested, along with your example, to see how it responds.\n\nI don't think it has actual values in the prepare, but just\nplace-holders, so it doesn't do the reduction, and my code wouldn't do\nit either.\n\nIt is only when they use constants, and want to re-run the query with\nnew constants that could cause a problem.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Wed, 7 Oct 1998 17:33:07 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [GENERAL] Long update query ? (also Re: [GENERAL]\n\tCNF vs. DNF)"
}
] |
[
{
"msg_contents": "I've got this new notify code almost working, but...\n\nWhat exactly is the protocol for locking a table that you intend to\nmodify? The old notify code just did RelationSetLockForWrite(),\nbut it's not clear to me that that works correctly --- for one thing,\nit doesn't seem to offer any way of detecting failure to acquire the\nlock. (RelationSetLockForWrite calls MultiLockReln, which *does*\nreturn a boolean, but RelationSetLockForWrite ignores it...) Also,\nit's not at all clear whether I should call RelationUnsetLockForWrite\nat the end of the routine or not; some existing code does, some doesn't.\n\nI'm concerned because interlocking of the specialized NOTIFY-related\nstatements seems to work fine, but they seem not to be interlocked\nagainst user operations on the pg_listener table.\n\nFor example, this works as I'd expect:\n\n\tpsql#1\t\t\t\tpsql#2\n\n\tbegin;\n\tlisten z;\n\n\t\t\t\t\tnotify z;\n\t\t\t\t\t(hangs up until #1 commits)\n\n\tend;\n\nbecause \"listen\" acquires a write lock on the pg_listener table, which\nthe notify has to wait for.\n\nBut this doesn't work as I'd expect:\n\n\tpsql#1\t\t\t\tpsql#2\n\n\tbegin;\n\tselect * from pg_listener;\n\n\t\t\t\t\tnotify z;\n\t\t\t\t\t(completes immediately)\n\n\tend;\n\nSeems to me the \"select\" should acquire a read lock that would prevent\nthe #2 backend from writing pg_listener until the end of #1's transaction.\n\nIs there a bug here, or is there some special definition of user access\nto a system table that means the select isn't acquiring a read lock?\nSelects and updates on ordinary user tables seem to interlock fine...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sun, 27 Sep 1998 17:34:33 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "It sorta works, but I'm confused about locking"
},
{
"msg_contents": "> \n> I've got this new notify code almost working, but...\n\nCould you plase send it to me ?\n\n> What exactly is the protocol for locking a table that you intend to\n> modify? The old notify code just did RelationSetLockForWrite(),\n> but it's not clear to me that that works correctly --- for one thing,\n> it doesn't seem to offer any way of detecting failure to acquire the\n> lock. (RelationSetLockForWrite calls MultiLockReln, which *does*\n> return a boolean, but RelationSetLockForWrite ignores it...) Also,\n> it's not at all clear whether I should call RelationUnsetLockForWrite\n> at the end of the routine or not; some existing code does, some doesn't.\n\nIt is not done where there is an immediate CommitTransactionCommand which\nalready releases the locks.\n\n> I'm concerned because interlocking of the specialized NOTIFY-related\n> statements seems to work fine, but they seem not to be interlocked\n> against user operations on the pg_listener table.\n> \n> For example, this works as I'd expect:\n> \n> \tpsql#1\t\t\t\tpsql#2\n> \n> \tbegin;\n> \tlisten z;\n> \n> \t\t\t\t\tnotify z;\n> \t\t\t\t\t(hangs up until #1 commits)\n> \n> \tend;\n> \n> because \"listen\" acquires a write lock on the pg_listener table, which\n> the notify has to wait for.\n> \n> But this doesn't work as I'd expect:\n> \n> \tpsql#1\t\t\t\tpsql#2\n> \n> \tbegin;\n> \tselect * from pg_listener;\n> \n> \t\t\t\t\tnotify z;\n> \t\t\t\t\t(completes immediately)\n> \n> \tend;\n> \n> Seems to me the \"select\" should acquire a read lock that would prevent\n> the #2 backend from writing pg_listener until the end of #1's transaction.\n> \n> Is there a bug here, or is there some special definition of user access\n> to a system table that means the select isn't acquiring a read lock?\n> Selects and updates on ordinary user tables seem to interlock fine...\n\nVery strange, it seems a bug. But users should not know about pg_listener.\n\n-- \nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n| Massimo Dal Zotto email: [email protected] |\n| Via Marconi, 141 phone: ++39-461-534251 |\n| 38057 Pergine Valsugana (TN) www: http://www.cs.unitn.it/~dz/ |\n| Italy pgp: finger [email protected] |\n+----------------------------------------------------------------------+\n",
"msg_date": "Mon, 28 Sep 1998 18:46:58 +0200 (MET DST)",
"msg_from": "Massimo Dal Zotto <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] It sorta works, but I'm confused about locking"
},
{
"msg_contents": "Massimo Dal Zotto <[email protected]> writes:\n>> it's not at all clear whether I should call RelationUnsetLockForWrite\n>> at the end of the routine or not; some existing code does, some doesn't.\n\n> It is not done where there is an immediate CommitTransactionCommand which\n> already releases the locks.\n\nHmm. OK, but I guess I don't really understand why it's ever valid to\nrelease a lock before the end of the transaction --- and\nCommitTransactionCommand isn't necessarily the end of the transaction\nanyway, if you're inside a transaction block.\n\nBasically: it seems to me it's a bug to call UnsetLock (any flavor)\nfrom *anywhere* except transaction commit. If this is not so, why not?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 28 Sep 1998 14:09:06 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] It sorta works, but I'm confused about locking "
},
{
"msg_contents": "> I've got this new notify code almost working, but...\n> \n> What exactly is the protocol for locking a table that you intend to\n> modify? The old notify code just did RelationSetLockForWrite(),\n> but it's not clear to me that that works correctly --- for one thing,\n> it doesn't seem to offer any way of detecting failure to acquire the\n> lock. (RelationSetLockForWrite calls MultiLockReln, which *does*\n> return a boolean, but RelationSetLockForWrite ignores it...) Also,\n> it's not at all clear whether I should call RelationUnsetLockForWrite\n> at the end of the routine or not; some existing code does, some doesn't.\n\nWelcome to PostgreSQL. This is the type of thing I saw when doing the\nmega-patch, were people were not doing things consistently, because some\ncoders do not understand what is happening inside the code, and just\nwrite something that works, sort of.\n\nI recommend you check out what is currently done properly, and fix the\nones that are incorrect.\n\nI can imagine some cases where you would want to get a lock and keep it\nuntil the end of the transaction, and other times when you would want to\nrelease it before transaction end.\n\n\n> \n> I'm concerned because interlocking of the specialized NOTIFY-related\n> statements seems to work fine, but they seem not to be interlocked\n> against user operations on the pg_listener table.\n> \n> For example, this works as I'd expect:\n> \n> \tpsql#1\t\t\t\tpsql#2\n> \n> \tbegin;\n> \tlisten z;\n> \n> \t\t\t\t\tnotify z;\n> \t\t\t\t\t(hangs up until #1 commits)\n> \n> \tend;\n> \n> because \"listen\" acquires a write lock on the pg_listener table, which\n> the notify has to wait for.\n> \n> But this doesn't work as I'd expect:\n> \n> \tpsql#1\t\t\t\tpsql#2\n> \n> \tbegin;\n> \tselect * from pg_listener;\n> \n> \t\t\t\t\tnotify z;\n> \t\t\t\t\t(completes immediately)\n> \n> \tend;\n> \n> Seems to me the \"select\" should acquire a read lock that would prevent\n> the #2 backend from writing pg_listener until the end of #1's transaction.\n> \n> Is there a bug here, or is there some special definition of user access\n> to a system table that means the select isn't acquiring a read lock?\n> Selects and updates on ordinary user tables seem to interlock fine...\n\nSelect certainly should be locking. Something is wrong, but I am not\nsure what. If you want me to check into it, let me know.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 17:29:42 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] It sorta works, but I'm confused about locking"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> I recommend you check out what is currently done properly, and fix the\n> ones that are incorrect.\n\nWell, yes. The question was how to tell which is which :-)\n\n> I can imagine some cases where you would want to get a lock and keep it\n> until the end of the transaction, and other times when you would want to\n> release it before transaction end.\n\nI guess I'm not understanding something. How can it ever be correct\npractice to release a lock before transaction end? For example, if I\nwrite some changes in a table, and then release the lock, wouldn't that\nallow other backends to see the not-yet-committed changes? What if I\nthen abort my transaction? Now the other backends have acted on\ninformation they should never have seen at all.\n\nReleasing a read lock strikes me as just as dangerous but for more\nsubtle reasons --- once you have read a table, what you have read\nought to look the same until the end of your transaction.\n\nSince there is an unset-write-lock function, I assume it must have\nvalid uses, but I don't see what they are.\n\n>> Is there a bug here, or is there some special definition of user access\n>> to a system table that means the select isn't acquiring a read lock?\n>> Selects and updates on ordinary user tables seem to interlock fine...\n\n> Select certainly should be locking. Something is wrong, but I am not\n> sure what. If you want me to check into it, let me know.\n\nPlease. (Note that I saw this with my revised version of async.c;\nI believe you will see the same behavior with the currently-checked-in\ncode, but do not have the time to rebuild that version to make sure.)\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 01 Oct 1998 20:25:15 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] It sorta works, but I'm confused about locking "
},
{
"msg_contents": "Tom Lane wrote:\n> \n> > I can imagine some cases where you would want to get a lock and keep it\n> > until the end of the transaction, and other times when you would want to\n> > release it before transaction end.\n> \n> I guess I'm not understanding something. How can it ever be correct\n> practice to release a lock before transaction end? For example, if I\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nPostgres does this for system tables only.\n\n> write some changes in a table, and then release the lock, wouldn't that\n> allow other backends to see the not-yet-committed changes? What if I\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nNo. Backend uses xmin/xmax to know is tuple visible or not.\n\n> then abort my transaction? Now the other backends have acted on\n> information they should never have seen at all.\n\nBackend will see that xmin(xmax) is not committed.\n\nVadim\n",
"msg_date": "Fri, 02 Oct 1998 09:20:25 +0800",
"msg_from": "Vadim Mikheev <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] It sorta works, but I'm confused about locking"
},
{
"msg_contents": "Vadim Mikheev <[email protected]> writes:\n> Tom Lane wrote:\n>> I guess I'm not understanding something. How can it ever be correct\n>> practice to release a lock before transaction end? For example, if I\n>> write some changes in a table, and then release the lock, wouldn't that\n>> allow other backends to see the not-yet-committed changes? What if I\n> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n> No. Backend uses xmin/xmax to know is tuple visible or not.\n\nAh, some light dawns. Thanks for the clue!\n\n> Postgres does this for system tables only.\n\nOK, so what is good coding practice? Always release write lock after\noperating on a system table? Or is that only good some of the time,\nand if so what's the consideration?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 01 Oct 1998 21:47:13 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] It sorta works, but I'm confused about locking "
},
{
"msg_contents": "> Bruce Momjian <[email protected]> writes:\n> > I recommend you check out what is currently done properly, and fix the\n> > ones that are incorrect.\n> \n> Well, yes. The question was how to tell which is which :-)\n\nI didn't say it was easy. :-)\n\n> \n> > I can imagine some cases where you would want to get a lock and keep it\n> > until the end of the transaction, and other times when you would want to\n> > release it before transaction end.\n> \n> I guess I'm not understanding something. How can it ever be correct\n> practice to release a lock before transaction end? For example, if I\n> write some changes in a table, and then release the lock, wouldn't that\n> allow other backends to see the not-yet-committed changes? What if I\n> then abort my transaction? Now the other backends have acted on\n> information they should never have seen at all.\n> \n> Releasing a read lock strikes me as just as dangerous but for more\n> subtle reasons --- once you have read a table, what you have read\n> ought to look the same until the end of your transaction.\n> \n> Since there is an unset-write-lock function, I assume it must have\n> valid uses, but I don't see what they are.\n\nSuppose you want to update a table as part of internal query processing.\nYou lock it, update it, and unlock it. You just need the lock while\nyou are updating it.\n\n> \n> >> Is there a bug here, or is there some special definition of user access\n> >> to a system table that means the select isn't acquiring a read lock?\n> >> Selects and updates on ordinary user tables seem to interlock fine...\n> \n> > Select certainly should be locking. Something is wrong, but I am not\n> > sure what. If you want me to check into it, let me know.\n> \n> Please. (Note that I saw this with my revised version of async.c;\n> I believe you will see the same behavior with the currently-checked-in\n> code, but do not have the time to rebuild that version to make sure.)\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 22:24:20 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] It sorta works, but I'm confused about locking"
}
] |
[
{
"msg_contents": "Here's the new diff for getting the NS32K assembly code into the\nspin-locks. Notice that it's now inline assembler in s_lock.h,\nrather than seperate code in s_lock.c. It also shrank a little\nbit... Just rip out the S_LOCK() define and insert the tas() inline\nfunction. Please let me know if there are any problems with it.\n\nJon Buller\n\n\n\nIndex: src/include/storage/s_lock.h\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/include/storage/s_lock.h,v\nretrieving revision 1.50\ndiff -r1.50 s_lock.h\n170a171,186\n> #if defined(NEED_NS32K_TAS_ASM)\n> #define TAS(lock) tas(lock)\n> \n> static __inline__ int\n> tas(volatile slock_t *lock)\n> {\n> register _res;\n> __asm__(\"sbitb 0, %0\n> \tsfsd %1\"\n> \t: \"=m\"(*lock), \"=r\"(_res));\n> return (int) _res; \n> }\n> \n> #endif /* NEED_NS32K_TAS_ASM */\n> \n> \n214,231d229\n< \n< \n< \n< #if defined(NEED_NS32K_TAS_ASM)\n< \n< #define S_LOCK(lock)\t\t\t\t\\\n< {\t\t\t\t\t\t\\\n< \tslock_t res = 1;\t\t\t\t\\\n< \twhile (res) {\t\t\t\t\\\n< \t __asm__(\"movqd 0, r0\");\t\t\t\\\n< \t __asm__(\"sbitd r0, %0\" : \"=m\"(*lock));\t\\\n< \t __asm__(\"sprb us, %0\" : \"=r\" (res));\t\\\n< \t res = ((res >> 5) & 1);\t\t\t\\\n< \t}\t\t\t\t\t\t\\\n< }\n< \n< #endif\t /* NEED_NS32K_TAS_ASM */\n< \n",
"msg_date": "Sun, 27 Sep 1998 22:27:45 -0500",
"msg_from": "Jon Buller <[email protected]>",
"msg_from_op": true,
"msg_subject": "spinlock code for ns32k (again)"
},
{
"msg_contents": "Applied.\n\n> Here's the new diff for getting the NS32K assembly code into the\n> spin-locks. Notice that it's now inline assembler in s_lock.h,\n> rather than seperate code in s_lock.c. It also shrank a little\n> bit... Just rip out the S_LOCK() define and insert the tas() inline\n> function. Please let me know if there are any problems with it.\n> \n> Jon Buller\n> \n> \n> \n> Index: src/include/storage/s_lock.h\n> ===================================================================\n> RCS file: /usr/local/cvsroot/pgsql/src/include/storage/s_lock.h,v\n> retrieving revision 1.50\n> diff -r1.50 s_lock.h\n> 170a171,186\n> > #if defined(NEED_NS32K_TAS_ASM)\n> > #define TAS(lock) tas(lock)\n> > \n> > static __inline__ int\n> > tas(volatile slock_t *lock)\n> > {\n> > register _res;\n> > __asm__(\"sbitb 0, %0\n> > \tsfsd %1\"\n> > \t: \"=m\"(*lock), \"=r\"(_res));\n> > return (int) _res; \n> > }\n> > \n> > #endif /* NEED_NS32K_TAS_ASM */\n> > \n> > \n> 214,231d229\n> < \n> < \n> < \n> < #if defined(NEED_NS32K_TAS_ASM)\n> < \n> < #define S_LOCK(lock)\t\t\t\t\\\n> < {\t\t\t\t\t\t\\\n> < \tslock_t res = 1;\t\t\t\t\\\n> < \twhile (res) {\t\t\t\t\\\n> < \t __asm__(\"movqd 0, r0\");\t\t\t\\\n> < \t __asm__(\"sbitd r0, %0\" : \"=m\"(*lock));\t\\\n> < \t __asm__(\"sprb us, %0\" : \"=r\" (res));\t\\\n> < \t res = ((res >> 5) & 1);\t\t\t\\\n> < \t}\t\t\t\t\t\t\\\n> < }\n> < \n> < #endif\t /* NEED_NS32K_TAS_ASM */\n> < \n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 11:22:27 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] spinlock code for ns32k (again)"
}
] |
[
{
"msg_contents": "Hi All.\nI tested libpq for win32 coming with 6.4-BETA and have a question about\nLIBPQ.\n\nI used LIBPQ from a C program,but coundn't connect PostgreSQL DB.\n\nCalling pg_connect() without WSAStartup() failed because of gethostbyname()\nerror in LIBPQ.\nso I modified DllMain in LIBPQ as follows.\n\nBOOL WINAPI DllMain( HINSTANCE hinstDLL, DWORD fdwReason,\n LPVOID lpReserved ){\n\n\tWSADATA wsaData;\n\tswitch (fdwReason)\n\t\t {\n\t case DLL_PROCESS_ATTACH:\n\t\t WSAStartup( WSAStartup(MAKEWORD(1, 1), &wsaData);\n\t\t break;\n\t case DLL_PROCESS_DETACH:\n\t\t WSACleanup();\n\t\t break;\n\t\t }\n\n\treturn (TRUE);\n}\n\nand the connection to DB was OK.\n\nWhy WSAStartup() is not called in LIBPQ?\n\nThanks.\n\nHiroshi Inoue\[email protected]\n\n",
"msg_date": "Mon, 28 Sep 1998 13:09:47 +0900",
"msg_from": "\"Hiroshi Inoue\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "LIBPQ for WIN32"
}
] |
[
{
"msg_contents": "subscribe\n--------------------------------------------------------\nStan Thomas\nemail: <[email protected]>\n",
"msg_date": "Mon, 28 Sep 1998 00:08:29 -7",
"msg_from": "\"Stan Thomas\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "subscribe"
}
] |
[
{
"msg_contents": "Bruce,\n\n this is a partial new rewrite system. Could you please apply\n the patch?\n\n Changes:\n\n The order how rules get applied has changed. All non RIR\n rules are applied first recursively (resulting in\n parsetree splitting for insert/update/delete rules). All\n RIR rules are then applied on the list of querytrees\n managing for view rules.\n\n Aggregate columns of views used in a queries\n qualification are rewritten into sublinks.\n\n All RIR rules are also applied on queries other than\n SELECT.\n\n Enhancements:\n\n INSERT ... SELECT view ... works now.\n\n Views can have at least one aggregate column.\n\n The aggregate column of a view can be used in the\n qualification. But there are some limitations on it. The\n aggregate column must be on the right side in the qual\n expression.\n\n Views can have computed columns. Example:\n\n Let's have two tables\n table part (part_name char(8), unit char(4), size float);\n table unit (unit_name char(4), unit_factor float);\n\n The parts size is given in different units (cm, inch,\n ...). The table unit contains factors to compute\n centimeters from any unit. Now one can setup a view\n\n CREATE VIEW part_v1 AS SELECT\n P.part_name, (P.size * U.unit_factor) AS size_in_cm\n from part P, unit U\n where U.unit_name = P.unit;\n\n What's next:\n\n Make multiple action and non INSTEAD rules on SELECT\n work.\n\n I'll give the some weeks ago mentioned function code to\n wrap around aggregates in qualification a try.\n\n Limitations:\n\n The rewrite system is located between the parser and the\n planner/optimizer. This requires (up to now), that it\n outputs something the parser could have produced and this\n implies some limitations to what the rewrite system can\n do. I don't know much about the planner/optimizer, so I\n would need some help to improve the rule system more and\n more for 6.5.\n\n Outer joins aren't implemented yet. But SQL functions\n could be used to get around many of the limitations we\n currently have. Someone can setup SQL functions that\n cover aggregates and use them in a view. This view would\n behave as if we have outer joins (return count(*) of\n zero) and can have multiple aggregate columns without any\n grouping. These columns can be used as if they are\n regular value columns with no limits.\n\n Views of UNION and DISTINCT views are another beast. It's\n not such a big problem to rewrite a simple SELECT into a\n SELECT UNION or DISTINCT. But I don't know how to apply\n such a views rule on an\n\n INSERT ... SELECT view ...\n\n or an\n\n UPDATE ... SET col = view.attr WHERE ...\n\n Is it possible at all? And what about joins of tables\n where one is a union?\n\n Maybe I don't see the simple solutions because I'm\n thinking too complex. But all this could be impossible\n with our implementation of views over rewrite rules. At\n least we would need to extend the planner to be able to\n create plans for things that the parser never could\n produce.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\nbegin 644 rewrite.diff\nM9&EF9B`M8W(@<W)C+F]R:6<O8F%C:V5N9\"]R97=R:71E+VQO8VMS+F,@<W)C\nM+V)A8VME;F0O<F5W<FET92]L;V-K<RYC\"BHJ*B!S<F,N;W)I9R]B86-K96YD\nM+W)E=W)I=&4O;&]C:W,N8PE4:'4@4V5P(#(T(#$Q.C4W.C$U(#$Y.3@*+2TM\nM('-R8R]B86-K96YD+W)E=W)I=&4O;&]C:W,N8PE&<FD@4V5P(#(U(#(P.C$R\nM.C(S(#$Y.3@**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`R-\"PS,\"`J*BHJ\"B`@(VEN\nM8VQU9&4@(G5T:6QS+V)U:6QT:6YS+F@B\"B`@(VEN8VQU9&4@(F-A=&%L;V<O\nM<&=?<VAA9&]W+F@B\"B`@\"BT@<W1A=&EC('9O:60@8VAE8VM,;V-K4&5R;7,H\nM3&ES=\"`J;&]C:W,L(%%U97)Y(\"IP87)S971R964L(&EN=\"!R=%]I;F1E>\"D[\nM\"B`@\"B`@+RH*(\"`@*B!4:&ES3&]C:U=A<U1R:6=G97)E9`HM+2T@,C0L,CD@\nM+2TM+0HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#$W,\"PQ-S8@*BHJ*@H@('T*(\"`*\nM(\"`*(2!S=&%T:6,@=F]I9`H@(&-H96-K3&]C:U!E<FUS*$QI<W0@*FQO8VMS\nM+\"!1=65R>2`J<&%R<V5T<F5E+\"!I;G0@<G1?:6YD97@I\"B`@>PH@(`E296QA\nM=&EO;@EE=E]R96P[\"BTM+2`Q-CDL,3<U(\"TM+2T*(\"!]\"B`@\"B`@\"B$@=F]I\nM9`H@(&-H96-K3&]C:U!E<FUS*$QI<W0@*FQO8VMS+\"!1=65R>2`J<&%R<V5T\nM<F5E+\"!I;G0@<G1?:6YD97@I\"B`@>PH@(`E296QA=&EO;@EE=E]R96P[\"F1I\nM9F8@+6-R('-R8RYO<FEG+V)A8VME;F0O<F5W<FET92]R97=R:71E2&%N9&QE\nM<BYC('-R8R]B86-K96YD+W)E=W)I=&4O<F5W<FET94AA;F1L97(N8PHJ*BH@\nM<W)C+F]R:6<O8F%C:V5N9\"]R97=R:71E+W)E=W)I=&5(86YD;&5R+F,)5&AU\nM(%-E<\"`R-\"`Q,3HU-SHQ-R`Q.3DX\"BTM+2!S<F,O8F%C:V5N9\"]R97=R:71E\nM+W)E=W)I=&5(86YD;&5R+F,)1G)I(%-E<\"`R-2`R,3HP.#HT-R`Q.3DX\"BHJ\nM*BHJ*BHJ*BHJ*BHJ*@HJ*BH@-BPQ,B`J*BHJ\"B`@(\"H*(\"`@*@H@(\"`J($E$\nM14Y4249)0T%424].\"B$@(\"H)(\"`D2&5A9&5R.B`O=7-R+VQO8V%L+V-V<W)O\nM;W0O<&=S<6PO<W)C+V)A8VME;F0O<F5W<FET92]R97=R:71E2&%N9&QE<BYC\nM+'8@,2XR,2`Q.3DX+S`Y+S`Q(#`T.C,Q.C,S(&UO;6II86X@17AP(\"0*(\"`@\nM*@H@(\"`J+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM\nM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+0H@(\"`J+PHM+2T@\nM-BPQ,B`M+2TM\"B`@(\"H*(\"`@*@H@(\"`J($E$14Y4249)0T%424].\"B$@(\"H)\nM(\"`D2&5A9&5R.B`O=7-R+VQO8V%L+V-V<W)O;W0O<&=S<6PO<W)C+V)A8VME\nM;F0O<F5W<FET92]R97=R:71E2&%N9&QE<BYC+'8@,2XR,\"`Q.3DX+S`X+S(T\nM(#`Q.C,W.C4Y(&UO;6II86X@17AP(\"0*(\"`@*@H@(\"`J+2TM+2TM+2TM+2TM\nM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM\nM+2TM+2TM+2TM+2TM+2TM+0H@(\"`J+PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#$X\nM+#(S(\"HJ*BH*+2TM(#$X+#(T(\"TM+2T*(\"`C:6YC;'5D92`B=71I;',O<F5L\nM+F@B\"B`@(VEN8VQU9&4@(FYO9&5S+W!G7VQI<W0N:\"(*(\"`C:6YC;'5D92`B\nM;F]D97,O<')I;6YO9&5S+F@B\"BL@(VEN8VQU9&4@(FYO9&5S+W)E;&%T:6]N\nM+F@B\"B`@\"B`@(VEN8VQU9&4@(G!A<G-E<B]P87)S971R964N:\"()+RH@9F]R\nM('!A<G-E=')E92!M86YI<'5L871I;VX@*B\\*(\"`C:6YC;'5D92`B<&%R<V5R\nM+W!A<G-E7W)E;&%T:6]N+F@B\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@,S$L-3`@\nM*BHJ*@H@(\"-I;F-L=61E(\")C;VUM86YD<R]C<F5A=&EN:\"YH(@H@(\"-I;F-L\nM=61E(\")A8V-E<W,O:&5A<&%M+F@B\"B`@\"B`@(VEN8VQU9&4@(G5T:6QS+W-Y\nM<V-A8VAE+F@B\"B`@(VEN8VQU9&4@(G5T:6QS+V%C;\"YH(@H@(\"-I;F-L=61E\nM(\")C871A;&]G+W!G7W-H861O=RYH(@H@(`HM('-T871I8R!V;VED($%P<&QY\nM4F5T<FEE=F52=6QE*%%U97)Y(\"IP87)S971R964L(%)E=W)I=&52=6QE(\"IR\nM=6QE+`HM(`D)\"0D@(&EN=\"!R=%]I;F1E>\"P@:6YT(')E;&%T:6]N7VQE=F5L\nM+`HM(`D)\"0D@(%)E;&%T:6]N(')E;&%T:6]N+\"!I;G0@*FUO9&EF:65D*3L*\nM+2!S=&%T:6,@3&ES=\"`J9FER95)U;&5S*%%U97)Y(\"IP87)S971R964L(&EN\nM=\"!R=%]I;F1E>\"P@0VUD5'EP92!E=F5N=\"P*+2`)\"2`@8F]O;\"`J:6YS=&5A\nM9%]F;&%G+\"!,:7-T(\"IL;V-K<RP@3&ES=\"`J*G%U86Q?<')O9'5C=',I.PHM\nM('-T871I8R!V;VED(%%U97)Y4F5W<FET95-U8DQI;FLH3F]D92`J;F]D92D[\nM\"BT@<W1A=&EC($QI<W0@*E%U97)Y4F5W<FET94]N92A1=65R>2`J<&%R<V5T\nM<F5E*3L*+2!S=&%T:6,@3&ES=\"`J9&5E<%)E=W)I=&51=65R>2A1=65R>2`J\nM<&%R<V5T<F5E*3L*+2!S=&%T:6,@=F]I9\"!297=R:71E4')E<')O8V5S<U%U\nM97)Y*%%U97)Y(\"IP87)S971R964I.PHM('-T871I8R!1=65R>2`J4F5W<FET\nM95!O<W1P<F]C97-S3F]N4V5L96-T*%%U97)Y(\"IP87)S971R964I.PH@(`H@\nM(\"\\J\"B`@(\"H@9V%T:&5R4F5W<FET94UE=&$@+0HM+2T@,S(L-S4@+2TM+0H@\nM(\"-I;F-L=61E(\")C;VUM86YD<R]C<F5A=&EN:\"YH(@H@(\"-I;F-L=61E(\")A\nM8V-E<W,O:&5A<&%M+F@B\"B`@\"BL@(VEN8VQU9&4@(G5T:6QS+VQS>7-C86-H\nM92YH(@H@(\"-I;F-L=61E(\")U=&EL<R]S>7-C86-H92YH(@H@(\"-I;F-L=61E\nM(\")U=&EL<R]A8VPN:\"(*(\"`C:6YC;'5D92`B8V%T86QO9R]P9U]S:&%D;W<N\nM:\"(**R`C:6YC;'5D92`B8V%T86QO9R]P9U]T>7!E+F@B\"BL@\"BL@\"BL@\"BL@\nM<W1A=&EC(%)E=W)I=&5);F9O(\"IG871H97)297=R:71E365T82A1=65R>2`J\nM<&%R<V5T<F5E+`HK(`D)\"0D@(%%U97)Y(\"IR=6QE7V%C=&EO;BP**R`)\"0D)\nM(\"!.;V1E(\"IR=6QE7W%U86PL\"BL@\"0D)\"2`@:6YT(')T7VEN9&5X+`HK(`D)\nM\"0D@($-M9%1Y<&4@979E;G0L\"BL@\"0D)\"2`@8F]O;\"`J:6YS=&5A9%]F;&%G\nM*3L**R!S=&%T:6,@8F]O;\"!R86YG951A8FQE16YT<GE?=7-E9\"A.;V1E(\"IN\nM;V1E+\"!I;G0@<G1?:6YD97@L(&EN=\"!S=6)L979E;'-?=7`I.PHK('-T871I\nM8R!B;V]L(&%T=')I8G5T95]U<V5D*$YO9&4@*FYO9&4L(&EN=\"!R=%]I;F1E\nM>\"P@:6YT(&%T=&YO+\"!I;G0@<W5B;&5V96QS7W5P*3L**R!S=&%T:6,@=F]I\nM9\"!O9F9S971?=F%R;F]D97,H3F]D92`J;F]D92P@:6YT(&]F9G-E=\"P@:6YT\nM('-U8FQE=F5L<U]U<\"D[\"BL@<W1A=&EC('9O:60@8VAA;F=E7W9A<FYO9&5S\nM*$YO9&4@*FYO9&4L(&EN=\"!R=%]I;F1E>\"P@:6YT(&YE=U]I;F1E>\"P@:6YT\nM('-U8FQE=F5L<U]U<\"D[\"BL@<W1A=&EC('9O:60@;6]D:69Y06=G<F5G57!L\nM979E;\"A.;V1E(\"IN;V1E*3L**R!S=&%T:6,@=F]I9\"!M;V1I9GE!9V=R96=#\nM:&%N9V5687)N;V1E<RA.;V1E(\"HJ;F]D95!T<BP@:6YT(')T7VEN9&5X+\"!I\nM;G0@;F5W7VEN9&5X+\"!I;G0@<W5B;&5V96QS7W5P*3L**R!S=&%T:6,@=F]I\nM9\"!M;V1I9GE!9V=R96=$<F]P475A;\"A.;V1E(\"HJ;F]D95!T<BP@3F]D92`J\nM;W)I9VYO9&4L($5X<'(@*F5X<'(I.PHK('-T871I8R!3=6),:6YK(\"IM;V1I\nM9GE!9V=R96=-86ME4W5B;&EN:RA%>'!R(\"IO<FEG97AP+\"!1=65R>2`J<&%R\nM<V5T<F5E*3L**R!S=&%T:6,@=F]I9\"!M;V1I9GE!9V=R96=1=6%L*$YO9&4@\nM*BIN;V1E4'1R+\"!1=65R>2`J<&%R<V5T<F5E*3L**R`**R`**R`**R`**R`*\nM*R`**R`**R`**R`**R`**R`**R`**R`**R!S=&%T:6,@475E<GD@*F9I<F52\nM25)R=6QE<RA1=65R>2`J<&%R<V5T<F5E*3L*(\"`*(\"`*(\"`O*@H@(\"`J(&=A\nM=&AE<E)E=W)I=&5-971A(\"T**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`Q,3@L-#,S\nM(\"HJ*BH*(\"`)<F5T=7)N(&EN9F\\[\"B`@?0H@(`HM('-T871I8R!,:7-T(\"H*\nM+2!/<'1I;6EZ95))4E)U;&5S*$QI<W0@*FQO8VMS*0HM('L*+2`)3&ES=`D@\nM(\"`J871T<E]L979E;\"`]($Y)3\"P*+2`)\"0D@(\"`J:3L*+2`)3&ES=`D@(\"`J\nM<F5L871I;VY?;&5V96P@/2!.24P[\"BT@\"BT@\"69O<F5A8V@H:2P@;&]C:W,I\nM\"BT@\"7L*+2`)\"5)E=W)I=&52=6QE(\"IR=6QE7VQO8VL@/2!L9FER<W0H:2D[\nM\"BT@\"BT@\"0EI9B`H<G5L95]L;V-K+3YA='1R;F\\@/3T@+3$I\"BT@\"0D)<F5L\nM871I;VY?;&5V96P@/2!L87!P96YD*')E;&%T:6]N7VQE=F5L+\"!R=6QE7VQO\nM8VLI.PHM(`D)96QS90HM(`D)\"6%T=')?;&5V96P@/2!L87!P96YD*&%T=')?\nM;&5V96PL(')U;&5?;&]C:RD[\"BT@\"7T*+2`)<F5T=7)N(&YC;VYC*')E;&%T\nM:6]N7VQE=F5L+\"!A='1R7VQE=F5L*3L*+2!]\"B`@\"B`@+RH*(2`@*B!I9&5A\nM(&ES('1O(&9I<F4@<F5G=6QA<B!R=6QE<R!F:7)S=\"P@=&AE;B!Q=6%L:69I\nM960@:6YS=&5A9`HA(\"`J(')U;&5S(&%N9\"!U;G%U86QI9FEE9\"!I;G-T96%D\nM(')U;&5S(&QA<W0N($%N>2!L96UM:6YG(&ES(&-O=6YT960@9F]R+@H@(\"`J\nM+PHA('-T871I8R!,:7-T(\"H*(2!O<F1E<E)U;&5S*$QI<W0@*FQO8VMS*0H@\nM('L*(2`)3&ES=`D@(\"`J<F5G=6QA<B`]($Y)3#L*(2`)3&ES=`D@(\"`J:6YS\nM=&5A9%]R=6QE<R`]($Y)3#L*(2`)3&ES=`D@(\"`J:6YS=&5A9%]Q=6%L:69I\nM960@/2!.24P[\"B$@\"4QI<W0)(\"`@*FD[\"B`@\"B$@\"69O<F5A8V@H:2P@;&]C\nM:W,I\"B$@\"7L*(2`)\"5)E=W)I=&52=6QE(\"IR=6QE7VQO8VL@/2`H4F5W<FET\nM95)U;&4@*BD@;&9I<G-T*&DI.PH@(`HA(`D):68@*')U;&5?;&]C:RT^:7-)\nM;G-T96%D*0HA(`D)>PHA(`D)\"6EF(\"AR=6QE7VQO8VLM/G%U86P@/3T@3E5,\nM3\"D*(2`)\"0D):6YS=&5A9%]R=6QE<R`](&QA<'!E;F0H:6YS=&5A9%]R=6QE\nM<RP@<G5L95]L;V-K*3L*(2`)\"0EE;'-E\"B$@\"0D)\"6EN<W1E861?<75A;&EF\nM:65D(#T@;&%P<&5N9\"AI;G-T96%D7W%U86QI9FEE9\"P@<G5L95]L;V-K*3L*\nM(2`)\"7T*(2`)\"65L<V4*(2`)\"0ER96=U;&%R(#T@;&%P<&5N9\"AR96=U;&%R\nM+\"!R=6QE7VQO8VLI.PHA(`E]\"B$@\"7)E9W5L87(@/2!N8V]N8RAR96=U;&%R\nM+\"!I;G-T96%D7W%U86QI9FEE9\"D[\"B$@\"7)E='5R;B!N8V]N8RAR96=U;&%R\nM+\"!I;G-T96%D7W)U;&5S*3L*(2!]\"B`@\"B$@<W1A=&EC(&EN=`HA($%L;%)E\nM=')I979E*$QI<W0@*F%C=&EO;G,I\"B$@>PHA(`E,:7-T\"2`@(\"IN.PH@(`HA\nM(`EF;W)E86-H*&XL(&%C=&EO;G,I\"B$@\"7L*(2`)\"5%U97)Y\"2`@(\"IP=\"`]\nM(&QF:7)S=\"AN*3L*(\"`*(2`)\"2\\J\"B$@\"0D@*B!I;B!T:&4@;VQD('!O<W1G\nM<F5S(&-O9&4L('=E(&-H96-K('=H971H97(@8V]M;6%N9%]T>7!E(&ES(&$*\nM(2`)\"2`J(&-O;G-P(&]F(\"<H)RHG+F-O;6UA;F14>7!E*2X@8G5T('=E)W9E\nM(&YE=F5R('-U<'!O<G1E9`HA(`D)(\"H@=')A;G-I=&EV92!C;&]S=7)E<RX@\nM2&5N8V4@<F5M;W9E9`D@+2!A>2`Q,\"\\Y-\"X*(2`)\"2`J+PHA(`D):68@*'!T\nM+3YC;VUM86YD5'EP92`A/2!#341?4T5,14-4*0HA(`D)\"7)E='5R;B!F86QS\nM93L*(2`)?0HA(`ER971U<FX@=')U93L*(2!]\"B`@\"B$@<W1A=&EC($QI<W0@\nM*@HA($9I<F52971R:65V95)U;&5S0711=65R>2A1=65R>2`J<&%R<V5T<F5E\nM+`HA(`D)\"0D)\"2!I;G0@<G1?:6YD97@L\"B$@\"0D)\"0D)(%)E;&%T:6]N(')E\nM;&%T:6]N+`HA(`D)\"0D)\"2!B;V]L(\"II;G-T96%D7V9L86<L\"B$@\"0D)\"0D)\nM(&EN=\"!R=6QE7V9L86<I\"B$@>PHA(`E,:7-T\"2`@(\"II+`HA(`D)\"2`@(\"IL\nM;V-K<SL*(2`)4G5L94QO8VL@(\"`J<G1?96YT<GE?;&]C:W,@/2!.54Q,.PHA\nM(`E,:7-T\"2`@(\"IW;W)K(#T@3DE,.PH@(`HA(`EI9B`H*')T7V5N=')Y7VQO\nM8VMS(#T@<F5L871I;VXM/G)D7W)U;&5S*2`]/2!.54Q,*0HA(`D)<F5T=7)N\nM($Y)3#L*(\"`*(2`);&]C:W,@/2!M871C:$QO8VMS*$--1%]314Q%0U0L(')T\nM7V5N=')Y7VQO8VMS+\"!R=%]I;F1E>\"P@<&%R<V5T<F5E*3L*(\"`*(2`)+RH@\nM9FEN9\"!A;&P@<F5T<FEE=F4@:6YS=&5A9\"`J+PHA(`EF;W)E86-H*&DL(&QO\nM8VMS*0HA(`E[\"B$@\"0E297=R:71E4G5L92`J<G5L95]L;V-K(#T@*%)E=W)I\nM=&52=6QE(\"HI(&QF:7)S=\"AI*3L*(\"`*(2`)\"6EF(\"@A<G5L95]L;V-K+3YI\nM<TEN<W1E860I\"B$@\"0D)8V]N=&EN=64[\"B$@\"0EW;W)K(#T@;&%P<&5N9\"AW\nM;W)K+\"!R=6QE7VQO8VLI.PHA(`E]\"B$@\"6EF(\"AW;W)K(\"$]($Y)3\"D*(2`)\nM>PHA(`D)=V]R:R`]($]P=&EM:7IE4DE24G5L97,H;&]C:W,I.PHA(`D)9F]R\nM96%C:\"AI+\"!W;W)K*0HA(`D)>PHA(`D)\"5)E=W)I=&52=6QE(\"IR=6QE7VQO\nM8VL@/2!L9FER<W0H:2D[\"B$@\"0D):6YT\"0D)<F5L871I;VY?;&5V96P[\"B$@\nM\"0D):6YT\"0D);6]D:69I960@/2!&04Q313L*(2`*(2`)\"0ER96QA=&EO;E]L\nM979E;\"`](\"AR=6QE7VQO8VLM/F%T=')N;R`]/2`M,2D[\"B$@\"0D):68@*')U\nM;&5?;&]C:RT^86-T:6]N<R`]/2!.24PI\"B$@\"0D)>PHA(`D)\"0DJ:6YS=&5A\nM9%]F;&%G(#T@5%)513L*(2`)\"0D)<F5T=7)N($Y)3#L*(2`)\"0E]\"B$@\"0D)\nM:68@*\"%R=6QE7V9L86<@)B8*(2`)\"0D);&5N9W1H*')U;&5?;&]C:RT^86-T\nM:6]N<RD@/CT@,B`F)@HA(`D)\"0E!;&Q2971R:65V92AR=6QE7VQO8VLM/F%C\nM=&EO;G,I*0HA(`D)\"7L*(2`)\"0D)*FEN<W1E861?9FQA9R`](%12544[\"B$@\nM\"0D)\"7)E='5R;B!R=6QE7VQO8VLM/F%C=&EO;G,[\"B$@\"0D)?0HA(`D)\"4%P\nM<&QY4F5T<FEE=F52=6QE*'!A<G-E=')E92P@<G5L95]L;V-K+\"!R=%]I;F1E\nM>\"P@<F5L871I;VY?;&5V96PL(')E;&%T:6]N+`HA(`D)\"0D)\"0D@(\"9M;V1I\nM9FEE9\"D[\"B$@\"0D):68@*&UO9&EF:65D*0HA(`D)\"7L*(2`)\"0D)*FEN<W1E\nM861?9FQA9R`](%12544[\"B$@\"0D)\"49I>%)E<V1O;51Y<&5S*'!A<G-E=')E\nM92T^=&%R9V5T3&ES=\"D[\"B$@\"0D)\"7)E='5R;B!L8V]N<RAP87)S971R964L\nM($Y)3\"D[\"B`@\"0D)?0HA(`D)?0H@(`E]\"B$@\"7)E='5R;B!.24P[\"B`@?0H@\nM(`H@(`HA(\"\\J($ED96$@:7,@;&EK92!T:&ES.@HA(\"`J\"B$@(\"H@<F5T<FEE\nM=F4M:6YS=&5A9\"UR971R:65V92!R=6QE<R!H879E(&1I9F9E<F5N=\"!S96UA\nM;G1I8W,@=&AA;B!U<&1A=&4@;F]D97,*(2`@*B!397!A<F%T92!225(@<G5L\nM97,@9G)O;2!O=&AE<G,N\"5!A<W,@;W1H97)S('1O($9I<F52=6QE<RX*(2`@\nM*B!/<F1E<B!225(@<G5L97,@86YD('!R;V-E<W,N\"B$@(\"H*(2`@*B!S:61E\nM(&5F9F5C=#H@<&%R<V5T<F5E)W,@<G1A8FQE(&9I96QD(&UI9VAT(&)E(&-H\nM86YG960*(\"`@*B\\*(2!S=&%T:6,@=F]I9`HA($%P<&QY4F5T<FEE=F52=6QE\nM*%%U97)Y(\"IP87)S971R964L\"B$@\"0D)\"2`@4F5W<FET95)U;&4@*G)U;&4L\nM\"B$@\"0D)\"2`@:6YT(')T7VEN9&5X+`HA(`D)\"0D@(&EN=\"!R96QA=&EO;E]L\nM979E;\"P*(2`)\"0D)(\"!296QA=&EO;B!R96QA=&EO;BP*(2`)\"0D)(\"!I;G0@\nM*FUO9&EF:65D*0H@('L*(2`)475E<GD)(\"`@*G)U;&5?86-T:6]N(#T@3E5,\nM3#L*(2`)3F]D90D@(\"`J<G5L95]Q=6%L.PHA(`E,:7-T\"2`@(\"IR=&%B;&4L\nM\"B$@\"0D)(\"`@*G)T.PHA(`EI;G0)\"0EN;W1H:6YG+`HA(`D)\"0ER=%]L96YG\nM=&@[\"B$@\"6EN=`D)\"6)A9'-Q;\"`]($9!3%-%.PH@(`HA(`ER=6QE7W%U86P@\nM/2!R=6QE+3YQ=6%L.PHA(`EI9B`H<G5L92T^86-T:6]N<RD*(2`)>PHA(`D)\nM:68@*&QE;F=T:\"AR=6QE+3YA8W1I;VYS*2`^(#$I\"2\\J(#\\_/R!B96-A=7-E\nM('=E(&1O;B=T(&AA;F1L90HA(`D)\"0D)\"0D)\"0D@*B!R=6QE<R!W:71H(&UO\nM<F4@=&AA;B!O;F4*(2`)\"0D)\"0D)\"0D)(\"H@86-T:6]N/R`M87D@*B\\*(\"`*\nM(2`)\"0DO*@HA(`D)\"2`J(%=!4DY)3D<A(2$@268@=V4@<V]M971I;65S(&AA\nM;F1L92!R=6QE<R!W:71H(&UO<F4@=&AA;B!O;F4*(2`)\"0D@*B!A8W1I;VXL\nM('1H92!V:65W(&%C;\"!C:&5C:W,@;6EG:'0@9V5T(&)R;VME;BX*(2`)\"0D@\nM*B!V:65W06-L3W9E<G)I9&4@<VAO=6QD(&]N;'D@8F5C;VUE('1R=64@*&)E\nM;&]W*2!I9B!T:&ES(&ES\"B$@\"0D)(\"H@82!R96QA=&EO;E]L979E;\"P@:6YS\nM=&5A9\"P@<V5L96-T('%U97)Y(\"T@2F%N\"B$@\"0D)(\"HO\"B$@\"0D)<F5T=7)N\nM.PHA(`D)<G5L95]A8W1I;VX@/2!C;W!Y3V)J96-T*&QF:7)S=\"AR=6QE+3YA\nM8W1I;VYS*2D[\"B$@\"0EN;W1H:6YG(#T@1D%,4T4[\"B$@\"7T*(2`)96QS90HA\nM(`D);F]T:&EN9R`](%12544[\"B`@\"B$@\"7)T86)L92`](&-O<'E/8FIE8W0H\nM<&%R<V5T<F5E+3YR=&%B;&4I.PHA(`EF;W)E86-H*')T+\"!R=&%B;&4I\"B$@\nM\"7L*(2`)\"5)A;F=E5&)L16YT<GD@*G)T92`](&QF:7)S=\"AR=\"D[\"B`@\"B$@\nM\"0DO*@HA(`D)(\"H@=&AI<R!I<R!T;R!P<F5V96YT(&%D9%]M:7-S:6YG7W9A\nM<G-?=&]?8F%S95]R96QS*\"D@9G)O;2!A9&1I;F<*(2`)\"2`J(&$@8F]G=7,@\nM96YT<GD@=&\\@=&AE(&YE=R!T87)G970@;&ES=\"X*(2`)\"2`J+PHA(`D)<G1E\nM+3YI;D9R;VU#;\"`](&9A;'-E.PHA(`E]\"B$@\"7)T7VQE;F=T:\"`](&QE;F=T\nM:\"AR=&%B;&4I.PH@(`HA(`ER=&%B;&4@/2!N8V]N8RAR=&%B;&4L(&-O<'E/\nM8FIE8W0H<G5L95]A8W1I;VXM/G)T86)L92DI.PHA(`EP87)S971R964M/G)T\nM86)L92`](')T86)L93L*(\"`*(2`)<G5L95]A8W1I;VXM/G)T86)L92`](')T\nM86)L93L*(2`)3V9F<V5T5F%R3F]D97,H<G5L95]A8W1I;VXM/G%U86PL(')T\nM7VQE;F=T:\"D[\"B$@\"4]F9G-E=%9A<DYO9&5S*\"A.;V1E(\"HI(')U;&5?86-T\nM:6]N+3YT87)G971,:7-T+\"!R=%]L96YG=&@I.PHA(`E/9F9S971687).;V1E\nM<RAR=6QE7W%U86PL(')T7VQE;F=T:\"D[\"B`@\"B$@\"4]F9G-E=%9A<DYO9&5S\nM*\"A.;V1E(\"HI(')U;&5?86-T:6]N+3YG<F]U<$-L875S92P@<G1?;&5N9W1H\nM*3L*(2`)3V9F<V5T5F%R3F]D97,H*$YO9&4@*BD@<G5L95]A8W1I;VXM/FAA\nM=FEN9U%U86PL(')T7VQE;F=T:\"D[\"B`@\"B$@\"4-H86YG959A<DYO9&5S*')U\nM;&5?86-T:6]N+3YQ=6%L+`HA(`D)\"0D@(\"!04E,R7T-54E)%3E1?5D%23D\\@\nM*R!R=%]L96YG=&@L(')T7VEN9&5X+\"`P*3L*(2`)0VAA;F=E5F%R3F]D97,H\nM*$YO9&4@*BD@<G5L95]A8W1I;VXM/G1A<F=E=$QI<W0L\"B$@\"0D)\"2`@(%!2\nM4S)?0U524D5.5%]605).3R`K(')T7VQE;F=T:\"P@<G1?:6YD97@L(#`I.PHA\nM(`E#:&%N9V5687).;V1E<RAR=6QE7W%U86PL(%!24S)?0U524D5.5%]605).\nM3R`K(')T7VQE;F=T:\"P@<G1?:6YD97@L(#`I.PH@(`HA(`E#:&%N9V5687).\nM;V1E<R@H3F]D92`J*2!R=6QE7V%C=&EO;BT^9W)O=7!#;&%U<V4L\"B$@\"0D)\nM\"2`@(%!24S)?0U524D5.5%]605).3R`K(')T7VQE;F=T:\"P@<G1?:6YD97@L\nM(#`I.PHA(`E#:&%N9V5687).;V1E<R@H3F]D92`J*2!R=6QE7V%C=&EO;BT^\nM:&%V:6YG475A;\"P*(2`)\"0D)(\"`@4%)3,E]#55)214Y47U9!4DY/(\"L@<G1?\nM;&5N9W1H+\"!R=%]I;F1E>\"P@,\"D[\"B`@\"B$@\"6EF(\"AR96QA=&EO;E]L979E\nM;\"D*(2`)>PHA(`D)2&%N9&QE5FEE=U)U;&4H<&%R<V5T<F5E+\"!R=&%B;&4L\nM(')U;&5?86-T:6]N+3YT87)G971,:7-T+\"!R=%]I;F1E>\"P*(2`)\"0D)\"2`@\nM(&UO9&EF:65D*3L*(2`)?0HA(`EE;'-E\"B$@\"7L*(2`)\"4AA;F1L95))4D%T\nM=')I8G5T95)U;&4H<&%R<V5T<F5E+\"!R=&%B;&4L(')U;&5?86-T:6]N+3YT\nM87)G971,:7-T+`HA(`D)\"0D)\"0D@(\"!R=%]I;F1E>\"P@<G5L92T^871T<FYO\nM+\"!M;V1I9FEE9\"P@)F)A9'-Q;\"D[\"B$@\"7T*(2`):68@*\"IM;V1I9FEE9\"`F\nM)B`A8F%D<W%L*0HA(`E[\"B$@\"0E!9&11=6%L*'!A<G-E=')E92P@<G5L95]A\nM8W1I;VXM/G%U86PI.PH@(`HA(`D)+RH*(2`)\"2`J(%1H:7,@=VEL;\"!O;FQY\nM('=O<FL@:68@=&AE('%U97)Y(&UA9&4@=&\\@=&AE('9I97<@9&5F:6YE9\"!B\nM>0HA(`D)(\"H@=&AE(&9O;&QO=VEN9R!G<F]U<$-L875S92!G<F]U<',@8GD@\nM=&AE('-A;64@871T<FEB=71E<R!O<B!D;V5S\"B$@\"0D@*B!N;W0@=7-E(&=R\nM;W5P(&%T(&%L;\"$*(2`)\"2`J+PHA(`D):68@*'!A<G-E=')E92T^9W)O=7!#\nM;&%U<V4@/3T@3E5,3\"D*(2`)\"0EP87)S971R964M/F=R;W5P0VQA=7-E(#T@\nM<G5L95]A8W1I;VXM/F=R;W5P0VQA=7-E.PHA(`D)061D2&%V:6YG475A;\"AP\nM87)S971R964L(')U;&5?86-T:6]N+3YH879I;F=1=6%L*3L*(2`)\"7!A<G-E\nM=')E92T^:&%S06=G<R`](\"AR=6QE7V%C=&EO;BT^:&%S06=G<R!\\?\"!P87)S\nM971R964M/FAA<T%G9W,I.PHA(`D)<&%R<V5T<F5E+3YH87-3=6),:6YK<R`]\nM(\"AR=6QE7V%C=&EO;BT^:&%S4W5B3&EN:W,@?'P@<&%R<V5T<F5E+3YH87-3\nM=6),:6YK<RD[\"B$@\"7T*(2!]\"B`@\"B$@<W1A=&EC($QI<W0@*@HA(%!R;V-E\nM<W-2971R:65V95%U97)Y*%%U97)Y(\"IP87)S971R964L\"B$@\"0D)\"0D@3&ES\nM=\"`J<G1A8FQE+`HA(`D)\"0D)(&)O;VP@*FEN<W1E861?9FQA9RP*(2`)\"0D)\nM\"2!B;V]L(')U;&4I\"B$@>PHA(`E,:7-T\"2`@(\"IR=#L*(2`)3&ES=`D@(\"`J\nM<')O9'5C=%]Q=65R:65S(#T@3DE,.PHA(`EI;G0)\"0ER=%]I;F1E>\"`](#`[\nM\"B`@\"B`@\"B$@\"69O<F5A8V@H<G0L(')T86)L92D*(2`)>PHA(`D)4F%N9V54\nM8FQ%;G1R>2`J<G1?96YT<GD@/2!L9FER<W0H<G0I.PHA(`D)4F5L871I;VX)\nM<G1?96YT<GE?<F5L871I;VX@/2!.54Q,.PHA(`D)3&ES=`D@(\"`J<F5S=6QT\nM(#T@3DE,.PH@(`HA(`D)<G1?:6YD97@K*SL*(2`)\"7)T7V5N=')Y7W)E;&%T\nM:6]N(#T@:&5A<%]O<&5N<BAR=%]E;G1R>2T^<F5L;F%M92D[\"B`@\"B`@\"B`@\nM\"BT@\"0EI9B`H<G1?96YT<GE?<F5L871I;VXM/G)D7W)U;&5S(\"$]($Y53$PI\nM\"BT@\"0E[\"BT@\"0D)<F5S=6QT(#T*+2`)\"0D)1FER95)E=')I979E4G5L97-!\nM=%%U97)Y*'!A<G-E=')E92P*+2`)\"0D)\"0D)\"0D)(')T7VEN9&5X+`HM(`D)\nM\"0D)\"0D)\"0D@<G1?96YT<GE?<F5L871I;VXL\"BT@\"0D)\"0D)\"0D)\"2!I;G-T\nM96%D7V9L86<L\"BT@\"0D)\"0D)\"0D)\"2!R=6QE*3L*+2`)\"7T*+2`)\"6AE87!?\nM8VQO<V4H<G1?96YT<GE?<F5L871I;VXI.PHM(`D):68@*\"II;G-T96%D7V9L\nM86<I\"BT@\"0D)<F5T=7)N(')E<W5L=#L*(\"`)?0HM(`EI9B`H<G5L92D*+2`)\nM\"7)E='5R;B!.24P[\"B`@\"B$@\"7)T7VEN9&5X(#T@,#L*(\"`*+2`)9F]R96%C\nM:\"AR=\"P@<G1A8FQE*0HM(`E[\"BT@\"0E286YG951B;$5N=')Y(\"IR=%]E;G1R\nM>2`](&QF:7)S=\"AR=\"D[\"BT@\"0E296QA=&EO;@ER=%]E;G1R>5]R96QA=&EO\nM;B`]($Y53$P[\"BT@\"0E2=6QE3&]C:R`@(\"IR=%]E;G1R>5]L;V-K<R`]($Y5\nM3$P[\"BT@\"0E,:7-T\"2`@(\"IR97-U;'0@/2!.24P[\"BT@\"0E,:7-T\"2`@(\"IL\nM;V-K<R`]($Y)3#L*+2`)\"4QI<W0)(\"`@*F1U;6UY7W!R;V1U8W1S.PHM(`HM\nM(`D)<G1?:6YD97@K*SL*+2`)\"7)T7V5N=')Y7W)E;&%T:6]N(#T@:&5A<%]O\nM<&5N<BAR=%]E;G1R>2T^<F5L;F%M92D[\"BT@\"0ER=%]E;G1R>5]L;V-K<R`]\nM(')T7V5N=')Y7W)E;&%T:6]N+3YR9%]R=6QE<SL*+2`)\"6AE87!?8VQO<V4H\nM<G1?96YT<GE?<F5L871I;VXI.PH@(`H@(`HA(`D):68@*')T7V5N=')Y7VQO\nM8VMS*0HA(`D)>PHA(`D)\"6QO8VMS(#T*(2`)\"0D);6%T8VA,;V-K<RA#341?\nM4T5,14-4+\"!R=%]E;G1R>5]L;V-K<RP@<G1?:6YD97@L('!A<G-E=')E92D[\nM\"B`@\"0E]\"B$@\"0EI9B`H;&]C:W,@(3T@3DE,*0H@(`D)>PHA(`D)\"7)E<W5L\nM=\"`](&9I<F52=6QE<RAP87)S971R964L(')T7VEN9&5X+\"!#341?4T5,14-4\nM+`HA(`D)\"0D)\"0D@(\"!I;G-T96%D7V9L86<L(&QO8VMS+\"`F9'5M;7E?<')O\nM9'5C=',I.PHA(`D)\"6EF(\"@J:6YS=&5A9%]F;&%G*0HA(`D)\"0ER971U<FX@\nM;&%P<&5N9\"A.24PL(')E<W5L=\"D[\"B$@\"0D):68@*')E<W5L=\"`A/2!.24PI\nM\"B$@\"0D)\"7!R;V1U8W1?<75E<FEE<R`](&YC;VYC*'!R;V1U8W1?<75E<FEE\nM<RP@<F5S=6QT*3L*(\"`)\"7T*(\"`)?0HA(`ER971U<FX@<')O9'5C=%]Q=65R\nM:65S.PH@('T*(\"`*(\"!S=&%T:6,@475E<GD@*@H@($-O<'E!;F1!9&11=6%L\nM*%%U97)Y(\"IP87)S971R964L\"B`@\"0D)(\"`@3&ES=\"`J86-T:6]N<RP*+2TM\nM(#$T,RPR,C(U(\"TM+2T*(\"`)<F5T=7)N(&EN9F\\[\"B`@?0H@(`H@(`H@(\"\\J\nM\"B$@(\"H@<F%N9V5486)L945N=')Y7W5S960@+0HA(\"`J\"7=E(&YE960@=&\\@\nM<')O8V5S<R!A(%)412!F;W(@4DE2(')U;&5S(&]N;'D@:68@:70@:7,*(2`@\nM*@ER969E<F5N8V5D('-O;65W:&5R92!I;B!V87(@;F]D97,@;V8@=&AE('%U\nM97)Y+@H@(\"`J+PHA('-T871I8R!B;V]L\"B$@<F%N9V5486)L945N=')Y7W5S\nM960H3F]D92`J;F]D92P@:6YT(')T7VEN9&5X+\"!I;G0@<W5B;&5V96QS7W5P\nM*0H@('L*(2`):68@*&YO9&4@/3T@3E5,3\"D*(2`)\"7)E='5R;B!&04Q313L*\nM(\"`*(2`)<W=I=&-H*&YO9&5486<H;F]D92DI('L*(2`)\"6-A<V4@5%]487)G\nM971%;G1R>3H*(2`)\"0E[\"B$@\"0D)\"51A<F=E=$5N=')Y\"2IT;&4@/2`H5&%R\nM9V5T16YT<GD@*BEN;V1E.PH@(`HA(`D)\"0ER971U<FX@<F%N9V5486)L945N\nM=')Y7W5S960H\"B$@\"0D)\"0D)*$YO9&4@*BDH=&QE+3YE>'!R*2P*(2`)\"0D)\nM\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\"7T*(2`)\nM\"0EB<F5A:SL*(\"`*(2`)\"6-A<V4@5%]!9V=R96<Z\"B$@\"0D)>PHA(`D)\"0E!\nM9V=R96<)*F%G9R`](\"A!9V=R96<@*BEN;V1E.PH@(`HA(`D)\"0ER971U<FX@\nM<F%N9V5486)L945N=')Y7W5S960H\"B$@\"0D)\"0D)*$YO9&4@*BDH86=G+3YT\nM87)G970I+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U\nM<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PH@(`HA(`D)8V%S92!47T=R;W5P0VQA\nM=7-E.@HA(`D)\"7L*(2`)\"0D)1W)O=7!#;&%U<V4)*F=R<\"`](\"A'<F]U<$-L\nM875S92`J*6YO9&4[\"B`@\"B$@\"0D)\"7)E='5R;B!R86YG951A8FQE16YT<GE?\nM=7-E9\"@*(2`)\"0D)\"0DH3F]D92`J*2AG<G`M/F5N=')Y*2P*(2`)\"0D)\"0ER\nM=%]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\"7T*(2`)\"0EB\nM<F5A:SL*(\"`*(2`)\"6-A<V4@5%]%>'!R.@HA(`D)\"7L*(2`)\"0D)17AP<@DJ\nM97AP(#T@*$5X<'(@*BEN;V1E.PH@(`HA(`D)\"0ER971U<FX@<F%N9V5486)L\nM945N=')Y7W5S960H\"B$@\"0D)\"0D)*$YO9&4@*BDH97AP+3YA<F=S*2P*(2`)\nM\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\"7T*\nM(2`)\"0EB<F5A:SL*(\"`*(2`)\"6-A<V4@5%])=&5R.@HA(`D)\"7L*(2`)\"0D)\nM271E<@DJ:71E<B`](\"A)=&5R(\"HI;F]D93L*(\"`*(2`)\"0D)<F5T=7)N(')A\nM;F=E5&%B;&5%;G1R>5]U<V5D*`HA(`D)\"0D)\"2A.;V1E(\"HI*&ET97(M/FET\nM97)E>'!R*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?\nM=7`I.PH@(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]!<G)A>5)E\nM9CH*(2`)\"0E[\"B$@\"0D)\"4%R<F%Y4F5F\"2IR968@/2`H07)R87E2968@*BEN\nM;V1E.PHA(`HA(`D)\"0EI9B`H<F%N9V5486)L945N=')Y7W5S960H\"B$@\"0D)\nM\"0D)*$YO9&4@*BDH<F5F+3YR969U<'!E<FEN9&5X<'(I+`HA(`D)\"0D)\"7)T\nM7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"DI\"B$@\"0D)\"0ER971U<FX@\nM5%)513L*(2`)\"0D)\"B$@\"0D)\"6EF(\"AR86YG951A8FQE16YT<GE?=7-E9\"@*\nM(2`)\"0D)\"0DH3F]D92`J*2AR968M/G)E9FQO=V5R:6YD97AP<BDL\"B$@\"0D)\nM\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*2D*(2`)\"0D)\"7)E\nM='5R;B!44E5%.PHA(`D)\"0D*(2`)\"0D):68@*')A;F=E5&%B;&5%;G1R>5]U\nM<V5D*`HA(`D)\"0D)\"2A.;V1E(\"HI*')E9BT^<F5F97AP<BDL\"B$@\"0D)\"0D)\nM<G1?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*2D*(2`)\"0D)\"7)E='5R\nM;B!44E5%.PHA(`D)\"0D*(2`)\"0D):68@*')A;F=E5&%B;&5%;G1R>5]U<V5D\nM*`HA(`D)\"0D)\"2A.;V1E(\"HI*')E9BT^<F5F87-S9VYE>'!R*2P*(2`)\"0D)\nM\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I*0HA(`D)\"0D)<F5T\nM=7)N(%12544[\"B$@\"0D)\"0HA(`D)\"0ER971U<FX@1D%,4T4[\"B$@\"0D)?0HA\nM(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U9A<CH*(2`)\"0E[\"B$@\"0D)\"59A\nM<@DJ=F%R(#T@*%9A<B`J*6YO9&4[\"B$@\"B$@\"0D)\"6EF(\"AV87(M/G9A<FQE\nM=F5L<W5P(#T]('-U8FQE=F5L<U]U<\"D*(2`)\"0D)\"7)E='5R;B!V87(M/G9A\nM<FYO(#T](')T7VEN9&5X.PHA(`D)\"0EE;'-E\"B$@\"0D)\"0ER971U<FX@1D%,\nM4T4[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U!A<F%M.@HA\nM(`D)\"7)E='5R;B!&04Q313L*(2`*(2`)\"6-A<V4@5%]#;VYS=#H*(2`)\"0ER\nM971U<FX@1D%,4T4[\"B$@\"B$@\"0EC87-E(%1?3&ES=#H*(2`)\"0E[\"B$@\"0D)\nM\"4QI<W0)*FP[\"B$@\"B$@\"0D)\"69O<F5A8V@@*&PL(\"A,:7-T(\"HI;F]D92D@\nM>PHA(`D)\"0D):68@*')A;F=E5&%B;&5%;G1R>5]U<V5D*`HA(`D)\"0D)\"0DH\nM3F]D92`J*6QF:7)S=\"AL*2P*(2`)\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)\nM\"7-U8FQE=F5L<U]U<\"DI\"B$@\"0D)\"0D)<F5T=7)N(%12544[\"B$@\"0D)\"7T*\nM(2`)\"0D)<F5T=7)N($9!3%-%.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\nM\"6-A<V4@5%]3=6),:6YK.@HA(`D)\"7L*(2`)\"0D)4W5B3&EN:PDJ<W5B(#T@\nM*%-U8DQI;FL@*BEN;V1E.PHA(`HA(`D)\"0EI9B`H<F%N9V5486)L945N=')Y\nM7W5S960H\"B$@\"0D)\"0D)*$YO9&4@*BDH<W5B+3YL969T:&%N9\"DL\"B$@\"0D)\nM\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*2D*(2`)\"0D)\"7)E\nM='5R;B!44E5%.PHA(`HA(`D)\"0EI9B`H<F%N9V5486)L945N=')Y7W5S960H\nM\"B$@\"0D)\"0D)*$YO9&4@*BDH<W5B+3YS=6)S96QE8W0I+`HA(`D)\"0D)\"7)T\nM7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"`K(#$I*0HA(`D)\"0D)<F5T\nM=7)N(%12544[\"B$@\"B$@\"0D)\"7)E='5R;B!&04Q313L*(2`)\"0E]\"B$@\"0D)\nM8G)E86L[\"B$@\"B$@\"0EC87-E(%1?475E<GDZ\"B$@\"0D)>PHA(`D)\"0E1=65R\nM>0DJ<7)Y(#T@*%%U97)Y(\"HI;F]D93L*(2`*(2`)\"0D):68@*')A;F=E5&%B\nM;&5%;G1R>5]U<V5D*`HA(`D)\"0D)\"2A.;V1E(\"HI*'%R>2T^=&%R9V5T3&ES\nM=\"DL\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*2D*\nM(2`)\"0D)\"7)E='5R;B!44E5%.PHA(`HA(`D)\"0EI9B`H<F%N9V5486)L945N\nM=')Y7W5S960H\"B$@\"0D)\"0D)*$YO9&4@*BDH<7)Y+3YQ=6%L*2P*(2`)\"0D)\nM\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I*0HA(`D)\"0D)<F5T\nM=7)N(%12544[\"B$@\"B$@\"0D)\"6EF(\"AR86YG951A8FQE16YT<GE?=7-E9\"@*\nM(2`)\"0D)\"0DH3F]D92`J*2AQ<GDM/FAA=FEN9U%U86PI+`HA(`D)\"0D)\"7)T\nM7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"DI\"B$@\"0D)\"0ER971U<FX@\nM5%)513L*(2`*(2`)\"0D):68@*')A;F=E5&%B;&5%;G1R>5]U<V5D*`HA(`D)\nM\"0D)\"2A.;V1E(\"HI*'%R>2T^9W)O=7!#;&%U<V4I+`HA(`D)\"0D)\"7)T7VEN\nM9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"DI\"B$@\"0D)\"0ER971U<FX@5%)5\nM13L*(2`*(2`)\"0D)<F5T=7)N($9!3%-%.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*\nM(2`*(2`)\"61E9F%U;'0Z\"B$@\"0D)96QO9RA.3U1)0T4L(\")U;FMN;W=N(&YO\nM9&4@=&%G(\"5D(&EN(')A;F=E5&%B;&5%;G1R>5]U<V5D*\"DB+\"!N;V1E5&%G\nM*&YO9&4I*3L*(2`)\"0EE;&]G*$Y/5$E#12P@(DYO9&4@:7,Z(\"5S(BP@;F]D\nM951O4W1R:6YG*&YO9&4I*3L*(2`)\"0EB<F5A:SL*(2`*(2`*(\"`)?0HA(`HA\nM(`ER971U<FX@1D%,4T4[\"B`@?0H@(`H@(`HA(\"\\J\"B$@(\"H@871T<FEB=71E\nM7W5S960@+0HA(\"`J\"4-H96-K(&EF(&$@<W!E8VEF:6,@871T<FEB=71E(&YU\nM;6)E<B!O9B!A(%)412!I<R!U<V5D\"B$@(\"H)<V]M97=H97)E(&EN('1H92!Q\nM=65R>0H@(\"`J+PHA('-T871I8R!B;V]L\"B$@871T<FEB=71E7W5S960H3F]D\nM92`J;F]D92P@:6YT(')T7VEN9&5X+\"!I;G0@871T;F\\L(&EN=\"!S=6)L979E\nM;'-?=7`I\"B`@>PHA(`EI9B`H;F]D92`]/2!.54Q,*0HA(`D)<F5T=7)N($9!\nM3%-%.PH@(`HA(`ES=VET8V@H;F]D951A9RAN;V1E*2D@>PHA(`D)8V%S92!4\nM7U1A<F=E=$5N=')Y.@HA(`D)\"7L*(2`)\"0D)5&%R9V5T16YT<GD)*G1L92`]\nM(\"A487)G971%;G1R>2`J*6YO9&4[\"B`@\"B$@\"0D)\"7)E='5R;B!A='1R:6)U\nM=&5?=7-E9\"@*(2`)\"0D)\"0DH3F]D92`J*2AT;&4M/F5X<'(I+`HA(`D)\"0D)\nM\"7)T7VEN9&5X+`HA(`D)\"0D)\"6%T=&YO+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U\nM<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PH@(`HA(`D)8V%S92!47T%G9W)E9SH*\nM(2`)\"0E[\"B$@\"0D)\"4%G9W)E9PDJ86=G(#T@*$%G9W)E9R`J*6YO9&4[\"B`@\nM\"B$@\"0D)\"7)E='5R;B!A='1R:6)U=&5?=7-E9\"@*(2`)\"0D)\"0DH3F]D92`J\nM*2AA9V<M/G1A<F=E=\"DL\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)871T\nM;F\\L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\nM\"B`@\"B$@\"0EC87-E(%1?1W)O=7!#;&%U<V4Z\"B$@\"0D)>PHA(`D)\"0E'<F]U\nM<$-L875S90DJ9W)P(#T@*$=R;W5P0VQA=7-E(\"HI;F]D93L*(\"`*(2`)\"0D)\nM<F5T=7)N(&%T=')I8G5T95]U<V5D*`HA(`D)\"0D)\"2A.;V1E(\"HI*&=R<\"T^\nM96YT<GDI+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6%T=&YO+`HA(`D)\nM\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PH@(`HA(`D)\nM8V%S92!47T5X<'(Z\"B$@\"0D)>PHA(`D)\"0E%>'!R\"2IE>'`@/2`H17AP<B`J\nM*6YO9&4[\"B`@\"B$@\"0D)\"7)E='5R;B!A='1R:6)U=&5?=7-E9\"@*(2`)\"0D)\nM\"0DH3F]D92`J*2AE>'`M/F%R9W,I+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\nM\"0D)\"6%T=&YO+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)?0HA(`D)\nM\"6)R96%K.PH@(`HA(`D)8V%S92!47TET97(Z\"B$@\"0D)>PHA(`D)\"0E)=&5R\nM\"2II=&5R(#T@*$ET97(@*BEN;V1E.PH@(`HA(`D)\"0ER971U<FX@871T<FEB\nM=71E7W5S960H\"B$@\"0D)\"0D)*$YO9&4@*BDH:71E<BT^:71E<F5X<'(I+`HA\nM(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6%T=&YO+`HA(`D)\"0D)\"7-U8FQE\nM=F5L<U]U<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PH@(`HA(`D)8V%S92!47T%R\nM<F%Y4F5F.@HA(`D)\"7L*(2`)\"0D)07)R87E2968)*G)E9B`](\"A!<G)A>5)E\nM9B`J*6YO9&4[\"B`@\"B$@\"0D)\"6EF(\"AA='1R:6)U=&5?=7-E9\"@*(2`)\"0D)\nM\"0DH3F]D92`J*2AR968M/G)E9G5P<&5R:6YD97AP<BDL\"B$@\"0D)\"0D)<G1?\nM:6YD97@L\"B$@\"0D)\"0D)871T;F\\L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*2D*\nM(2`)\"0D)\"7)E='5R;B!44E5%.PHA(`HA(`D)\"0EI9B`H871T<FEB=71E7W5S\nM960H\"B$@\"0D)\"0D)*$YO9&4@*BDH<F5F+3YR969L;W=E<FEN9&5X<'(I+`HA\nM(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6%T=&YO+`HA(`D)\"0D)\"7-U8FQE\nM=F5L<U]U<\"DI\"B$@\"0D)\"0ER971U<FX@5%)513L*(2`*(2`)\"0D):68@*&%T\nM=')I8G5T95]U<V5D*`HA(`D)\"0D)\"2A.;V1E(\"HI*')E9BT^<F5F97AP<BDL\nM\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)871T;F\\L\"B$@\"0D)\"0D)<W5B\nM;&5V96QS7W5P*2D*(2`)\"0D)\"7)E='5R;B!44E5%.PHA(`HA(`D)\"0EI9B`H\nM871T<FEB=71E7W5S960H\"B$@\"0D)\"0D)*$YO9&4@*BDH<F5F+3YR969A<W-G\nM;F5X<'(I+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6%T=&YO+`HA(`D)\nM\"0D)\"7-U8FQE=F5L<U]U<\"DI\"B$@\"0D)\"0ER971U<FX@5%)513L*(\"`**R`)\nM\"0D)<F5T=7)N($9!3%-%.PHK(`D)\"7T**R`)\"0EB<F5A:SL*(\"`*(2`)\"6-A\nM<V4@5%]687(Z\"B$@\"0D)>PHA(`D)\"0E687()*G9A<B`](\"A687(@*BEN;V1E\nM.PH@(`HA(`D)\"0EI9B`H=F%R+3YV87)L979E;'-U<\"`]/2!S=6)L979E;'-?\nM=7`I\"B$@\"0D)\"0ER971U<FX@=F%R+3YV87)N;R`]/2!R=%]I;F1E>#L*(2`)\nM\"0D)96QS90HA(`D)\"0D)<F5T=7)N($9!3%-%.PHA(`D)\"7T*(2`)\"0EB<F5A\nM:SL*(\"`**R`)\"6-A<V4@5%]087)A;3H**R`)\"0ER971U<FX@1D%,4T4[\"BL@\nM\"BL@\"0EC87-E(%1?0V]N<W0Z\"BL@\"0D)<F5T=7)N($9!3%-%.PHK(`HK(`D)\nM8V%S92!47TQI<W0Z\"BL@\"0D)>PHK(`D)\"0E,:7-T\"2IL.PHK(`HK(`D)\"0EF\nM;W)E86-H(\"AL+\"`H3&ES=\"`J*6YO9&4I('L**R`)\"0D)\"6EF(\"AA='1R:6)U\nM=&5?=7-E9\"@**R`)\"0D)\"0D)*$YO9&4@*BEL9FER<W0H;\"DL\"BL@\"0D)\"0D)\nM\"7)T7VEN9&5X+`HK(`D)\"0D)\"0EA='1N;RP**R`)\"0D)\"0D)<W5B;&5V96QS\nM7W5P*2D**R`)\"0D)\"0ER971U<FX@5%)513L**R`)\"0D)?0HK(`D)\"0ER971U\nM<FX@1D%,4T4[\"BL@\"0D)?0HK(`D)\"6)R96%K.PHK(`HK(`D)8V%S92!47U-U\nM8DQI;FLZ\"BL@\"0D)>PHK(`D)\"0E3=6),:6YK\"2IS=6(@/2`H4W5B3&EN:R`J\nM*6YO9&4[\"BL@\"BL@\"0D)\"6EF(\"AA='1R:6)U=&5?=7-E9\"@**R`)\"0D)\"0DH\nM3F]D92`J*2AS=6(M/FQE9G1H86YD*2P**R`)\"0D)\"0ER=%]I;F1E>\"P**R`)\nM\"0D)\"0EA='1N;RP**R`)\"0D)\"0ES=6)L979E;'-?=7`I*0HK(`D)\"0D)<F5T\nM=7)N(%12544[\"BL@\"BL@\"0D)\"6EF(\"AA='1R:6)U=&5?=7-E9\"@**R`)\"0D)\nM\"0DH3F]D92`J*2AS=6(M/G-U8G-E;&5C=\"DL\"BL@\"0D)\"0D)<G1?:6YD97@L\nM\"BL@\"0D)\"0D)871T;F\\L\"BL@\"0D)\"0D)<W5B;&5V96QS7W5P(\"L@,2DI\"BL@\nM\"0D)\"0ER971U<FX@5%)513L**R`**R`)\"0D)<F5T=7)N($9!3%-%.PHK(`D)\nM\"7T**R`)\"0EB<F5A:SL**R`**R`)\"6-A<V4@5%]1=65R>3H**R`)\"0E[\"BL@\nM\"0D)\"5%U97)Y\"2IQ<GD@/2`H475E<GD@*BEN;V1E.PHK(`HK(`D)\"0EI9B`H\nM871T<FEB=71E7W5S960H\"BL@\"0D)\"0D)*$YO9&4@*BDH<7)Y+3YT87)G971,\nM:7-T*2P**R`)\"0D)\"0ER=%]I;F1E>\"P**R`)\"0D)\"0EA='1N;RP**R`)\"0D)\nM\"0ES=6)L979E;'-?=7`I*0HK(`D)\"0D)<F5T=7)N(%12544[\"BL@\"BL@\"0D)\nM\"6EF(\"AA='1R:6)U=&5?=7-E9\"@**R`)\"0D)\"0DH3F]D92`J*2AQ<GDM/G%U\nM86PI+`HK(`D)\"0D)\"7)T7VEN9&5X+`HK(`D)\"0D)\"6%T=&YO+`HK(`D)\"0D)\nM\"7-U8FQE=F5L<U]U<\"DI\"BL@\"0D)\"0ER971U<FX@5%)513L**R`**R`)\"0D)\nM:68@*&%T=')I8G5T95]U<V5D*`HK(`D)\"0D)\"2A.;V1E(\"HI*'%R>2T^:&%V\nM:6YG475A;\"DL\"BL@\"0D)\"0D)<G1?:6YD97@L\"BL@\"0D)\"0D)871T;F\\L\"BL@\nM\"0D)\"0D)<W5B;&5V96QS7W5P*2D**R`)\"0D)\"7)E='5R;B!44E5%.PHK(`HK\nM(`D)\"0EI9B`H871T<FEB=71E7W5S960H\"BL@\"0D)\"0D)*$YO9&4@*BDH<7)Y\nM+3YG<F]U<$-L875S92DL\"BL@\"0D)\"0D)<G1?:6YD97@L\"BL@\"0D)\"0D)871T\nM;F\\L\"BL@\"0D)\"0D)<W5B;&5V96QS7W5P*2D**R`)\"0D)\"7)E='5R;B!44E5%\nM.PHK(`HK(`D)\"0ER971U<FX@1D%,4T4[\"BL@\"0D)?0HK(`D)\"6)R96%K.PHK\nM(`HK(`D)9&5F875L=#H**R`)\"0EE;&]G*$Y/5$E#12P@(G5N:VYO=VX@;F]D\nM92!T86<@)60@:6X@871T<FEB=71E7W5S960H*2(L(&YO9&5486<H;F]D92DI\nM.PHK(`D)\"65L;V<H3D]424-%+\"`B3F]D92!I<SH@)7,B+\"!N;V1E5&]3=')I\nM;F<H;F]D92DI.PHK(`D)\"6)R96%K.PH@(`H@(`H@(`E]\"B`@\"B$@\"7)E='5R\nM;B!&04Q313L*(2!]\"B`@\"B`@\"BL@+RH**R`@*B!O9F9S971?=F%R;F]D97,@\nM+0HK(\"`J\"5=E(&YE960@86YO=&AE<B!V97)S:6]N(&]F($]F9G-E=%9A<DYO\nM9&5S*\"D@=VAE;B!P<F]C97-S:6YG\"BL@(\"H)4DE2(')U;&5S\"BL@(\"HO\"BL@\nM<W1A=&EC('9O:60**R!O9F9S971?=F%R;F]D97,H3F]D92`J;F]D92P@:6YT\nM(&]F9G-E=\"P@:6YT('-U8FQE=F5L<U]U<\"D**R![\"BL@\"6EF(\"AN;V1E(#T]\nM($Y53$PI\"BL@\"0ER971U<FX[\"B`@\"B$@\"7-W:71C:\"AN;V1E5&%G*&YO9&4I\nM*2![\"B$@\"0EC87-E(%1?5&%R9V5T16YT<GDZ\"B$@\"0D)>PHA(`D)\"0E487)G\nM971%;G1R>0DJ=&QE(#T@*%1A<F=E=$5N=')Y(\"HI;F]D93L*(2`*(2`)\"0D)\nM;V9F<V5T7W9A<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HI*'1L92T^97AP<BDL\nM\"B$@\"0D)\"0D);V9F<V5T+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)\nM?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T%G9W)E9SH*(2`)\"0E[\"B$@\nM\"0D)\"4%G9W)E9PDJ86=G(#T@*$%G9W)E9R`J*6YO9&4[\"B$@\"B$@\"0D)\"6]F\nM9G-E=%]V87)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*2AA9V<M/G1A<F=E=\"DL\nM\"B$@\"0D)\"0D);V9F<V5T+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)\nM?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T=R;W5P0VQA=7-E.@HA(`D)\nM\"7L*(2`)\"0D)1W)O=7!#;&%U<V4)*F=R<\"`](\"A'<F]U<$-L875S92`J*6YO\nM9&4[\"B$@\"B$@\"0D)\"6]F9G-E=%]V87)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J\nM*2AG<G`M/F5N=')Y*2P*(2`)\"0D)\"0EO9F9S970L\"B$@\"0D)\"0D)<W5B;&5V\nM96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?17AP\nM<CH*(2`)\"0E[\"B$@\"0D)\"45X<'()*F5X<\"`](\"A%>'!R(\"HI;F]D93L*(2`*\nM(2`)\"0D);V9F<V5T7W9A<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HI*&5X<\"T^\nM87)G<RDL\"B$@\"0D)\"0D);V9F<V5T+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\nM\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47TET97(Z\"B$@\"0D)\nM>PHA(`D)\"0E)=&5R\"2II=&5R(#T@*$ET97(@*BEN;V1E.PHA(`HA(`D)\"0EO\nM9F9S971?=F%R;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BDH:71E<BT^:71E<F5X\nM<'(I+`HA(`D)\"0D)\"6]F9G-E=\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA\nM(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]!<G)A>5)E9CH*(2`)\nM\"0E[\"B$@\"0D)\"4%R<F%Y4F5F\"2IR968@/2`H07)R87E2968@*BEN;V1E.PHA\nM(`HA(`D)\"0EO9F9S971?=F%R;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BDH<F5F\nM+3YR969U<'!E<FEN9&5X<'(I+`HA(`D)\"0D)\"6]F9G-E=\"P*(2`)\"0D)\"0ES\nM=6)L979E;'-?=7`I.PHA(`D)\"0EO9F9S971?=F%R;F]D97,H\"B$@\"0D)\"0D)\nM*$YO9&4@*BDH<F5F+3YR969L;W=E<FEN9&5X<'(I+`HA(`D)\"0D)\"6]F9G-E\nM=\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\"0EO9F9S971?=F%R;F]D\nM97,H\"B$@\"0D)\"0D)*$YO9&4@*BDH<F5F+3YR969E>'!R*2P*(2`)\"0D)\"0EO\nM9F9S970L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0D);V9F<V5T7W9A\nM<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HI*')E9BT^<F5F87-S9VYE>'!R*2P*\nM(2`)\"0D)\"0EO9F9S970L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0E]\nM\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?5F%R.@HA(`D)\"7L*(2`)\"0D)\nM5F%R\"2IV87(@/2`H5F%R(\"HI;F]D93L*(2`*(2`)\"0D):68@*'9A<BT^=F%R\nM;&5V96QS=7`@/3T@<W5B;&5V96QS7W5P*2\"0D)\"79A<BT^=F%R;F]O;&0@*ST@;V9F<V5T.PHA\nM(`D)\"0E]\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U!A<F%M\nM.@HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T-O;G-T.@HA(`D)\"6)R96%K\nM.PHA(`HA(`D)8V%S92!47TQI<W0Z\"B$@\"0D)>PHA(`D)\"0E,:7-T\"2IL.PHA\nM(`HA(`D)\"0EF;W)E86-H(\"AL+\"`H3&ES=\"`J*6YO9&4I\"B$@\"0D)\"0EO9F9S\nM971?=F%R;F]D97,H\"B$@\"0D)\"0D)\"2A.;V1E(\"HI;&9I<G-T*&PI+`HA(`D)\nM\"0D)\"0EO9F9S970L\"B$@\"0D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)?0HA\nM(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U-U8DQI;FLZ\"B$@\"0D)>PHA(`D)\nM\"0E3=6),:6YK\"2IS=6(@/2`H4W5B3&EN:R`J*6YO9&4[\"B$@\"B$@\"0D)\"6]F\nM9G-E=%]V87)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*2AS=6(M/FQE9G1H86YD\nM*2P*(2`)\"0D)\"0EO9F9S970L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`*\nM(2`)\"0D);V9F<V5T7W9A<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HI*'-U8BT^\nM<W5B<V5L96-T*2P*(2`)\"0D)\"0EO9F9S970L\"B$@\"0D)\"0D)<W5B;&5V96QS\nM7W5P(\"L@,2D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U%U\nM97)Y.@HA(`D)\"7L*(2`)\"0D)475E<GD)*G%R>2`](\"A1=65R>2`J*6YO9&4[\nM\"B$@\"B$@\"0D)\"6]F9G-E=%]V87)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*2AQ\nM<GDM/G1A<F=E=$QI<W0I+`HA(`D)\"0D)\"6]F9G-E=\"P*(2`)\"0D)\"0ES=6)L\nM979E;'-?=7`I.PHA(`HA(`D)\"0EO9F9S971?=F%R;F]D97,H\"B$@\"0D)\"0D)\nM*$YO9&4@*BDH<7)Y+3YQ=6%L*2P*(2`)\"0D)\"0EO9F9S970L\"B$@\"0D)\"0D)\nM<W5B;&5V96QS7W5P*3L*(2`*(2`)\"0D);V9F<V5T7W9A<FYO9&5S*`HA(`D)\nM\"0D)\"2A.;V1E(\"HI*'%R>2T^:&%V:6YG475A;\"DL\"B$@\"0D)\"0D);V9F<V5T\nM+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"B$@\"0D)\"6]F9G-E=%]V87)N\nM;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*2AQ<GDM/F=R;W5P0VQA=7-E*2P*(2`)\nM\"0D)\"0EO9F9S970L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\nM\"0D)8G)E86L[\"B$@\"B$@\"0ED969A=6QT.@HA(`D)\"65L;V<H3D]424-%+\"`B\nM=6YK;F]W;B!N;V1E('1A9R`E9\"!I;B!O9F9S971?=F%R;F]D97,H*2(L(&YO\nM9&5486<H;F]D92DI.PHA(`D)\"65L;V<H3D]424-%+\"`B3F]D92!I<SH@)7,B\nM+\"!N;V1E5&]3=')I;F<H;F]D92DI.PHA(`D)\"6)R96%K.PHA(`HA(`HA(`E]\nM\"B$@?0HA(`HA(`HA(\"\\J\"B$@(\"H@8VAA;F=E7W9A<FYO9&5S(\"T*(2`@*@EA\nM;F0@86YO=&AE<B!#:&%N9V5687).;V1E<R@I('1O;PHA(\"`J+PHA('-T871I\nM8R!V;VED\"B$@8VAA;F=E7W9A<FYO9&5S*$YO9&4@*FYO9&4L(&EN=\"!R=%]I\nM;F1E>\"P@:6YT(&YE=U]I;F1E>\"P@:6YT('-U8FQE=F5L<U]U<\"D*(2![\"B$@\nM\"6EF(\"AN;V1E(#T]($Y53$PI\"B$@\"0ER971U<FX[\"B$@\"B$@\"7-W:71C:\"AN\nM;V1E5&%G*&YO9&4I*2![\"B$@\"0EC87-E(%1?5&%R9V5T16YT<GDZ\"B$@\"0D)\nM>PHA(`D)\"0E487)G971%;G1R>0DJ=&QE(#T@*%1A<F=E=$5N=')Y(\"HI;F]D\nM93L*(2`*(2`)\"0D)8VAA;F=E7W9A<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HI\nM*'1L92T^97AP<BDL\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D);F5W7VEN\nM9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K\nM.PHA(`HA(`D)8V%S92!47T%G9W)E9SH*(2`)\"0E[\"B$@\"0D)\"4%G9W)E9PDJ\nM86=G(#T@*$%G9W)E9R`J*6YO9&4[\"B$@\"B$@\"0D)\"6-H86YG95]V87)N;V1E\nM<R@*(2`)\"0D)\"0DH3F]D92`J*2AA9V<M/G1A<F=E=\"DL\"B$@\"0D)\"0D)<G1?\nM:6YD97@L\"B$@\"0D)\"0D);F5W7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U\nM<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T=R;W5P0VQA\nM=7-E.@HA(`D)\"7L*(2`)\"0D)1W)O=7!#;&%U<V4)*F=R<\"`](\"A'<F]U<$-L\nM875S92`J*6YO9&4[\"B$@\"B$@\"0D)\"6-H86YG95]V87)N;V1E<R@*(2`)\"0D)\nM\"0DH3F]D92`J*2AG<G`M/F5N=')Y*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\nM\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0E]\nM\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?17AP<CH*(2`)\"0E[\"B$@\"0D)\nM\"45X<'()*F5X<\"`](\"A%>'!R(\"HI;F]D93L*(2`*(2`)\"0D)8VAA;F=E7W9A\nM<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HI*&5X<\"T^87)G<RDL\"B$@\"0D)\"0D)\nM<G1?:6YD97@L\"B$@\"0D)\"0D);F5W7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L\nM<U]U<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47TET97(Z\nM\"B$@\"0D)>PHA(`D)\"0E)=&5R\"2II=&5R(#T@*$ET97(@*BEN;V1E.PHA(`HA\nM(`D)\"0EC:&%N9V5?=F%R;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BDH:71E<BT^\nM:71E<F5X<'(I+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6YE=U]I;F1E\nM>\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*\nM(2`*(2`)\"6-A<V4@5%]!<G)A>5)E9CH*(2`)\"0E[\"B$@\"0D)\"4%R<F%Y4F5F\nM\"2IR968@/2`H07)R87E2968@*BEN;V1E.PHA(`HA(`D)\"0EC:&%N9V5?=F%R\nM;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BDH<F5F+3YR969U<'!E<FEN9&5X<'(I\nM+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6YE=U]I;F1E>\"P*(2`)\"0D)\nM\"0ES=6)L979E;'-?=7`I.PHA(`D)\"0EC:&%N9V5?=F%R;F]D97,H\"B$@\"0D)\nM\"0D)*$YO9&4@*BDH<F5F+3YR969L;W=E<FEN9&5X<'(I+`HA(`D)\"0D)\"7)T\nM7VEN9&5X+`HA(`D)\"0D)\"6YE=U]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?\nM=7`I.PHA(`D)\"0EC:&%N9V5?=F%R;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BDH\nM<F5F+3YR969E>'!R*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?\nM:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0D)8VAA;F=E7W9A\nM<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HI*')E9BT^<F5F87-S9VYE>'!R*2P*\nM(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)\nM<W5B;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E\nM(%1?5F%R.@HA(`D)\"7L*(2`)\"0D)5F%R\"2IV87(@/2`H5F%R(\"HI;F]D93L*\nM(2`*(2`)\"0D):68@*'9A<BT^=F%R;&5V96QS=7`@/3T@<W5B;&5V96QS7W5P\nM(\"8F\"B$@\"0D)\"0D)=F%R+3YV87)N;R`]/2!R=%]I;F1E>\"D@>PHA(`D)\"0D)\nM=F%R+3YV87)N;R`](&YE=U]I;F1E>#L*(2`)\"0D)\"79A<BT^=F%R;F]O;&0@\nM/2!N97=?:6YD97@[\"B$@\"0D)\"7T*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\nM\"0EC87-E(%1?4&%R86TZ\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?0V]N\nM<W0Z\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?3&ES=#H*(2`)\"0E[\"B$@\nM\"0D)\"4QI<W0)*FP[\"B$@\"B$@\"0D)\"69O<F5A8V@@*&PL(\"A,:7-T(\"HI;F]D\nM92D*(2`)\"0D)\"6-H86YG95]V87)N;V1E<R@*(2`)\"0D)\"0D)*$YO9&4@*BEL\nM9FER<W0H;\"DL\"B$@\"0D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"0EN97=?:6YD\nM97@L\"B$@\"0D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K\nM.PHA(`HA(`D)8V%S92!47U-U8DQI;FLZ\"B$@\"0D)>PHA(`D)\"0E3=6),:6YK\nM\"2IS=6(@/2`H4W5B3&EN:R`J*6YO9&4[\"B$@\"B$@\"0D)\"6-H86YG95]V87)N\nM;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*2AS=6(M/FQE9G1H86YD*2P*(2`)\"0D)\nM\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V\nM96QS7W5P*3L*(2`*(2`)\"0D)8VAA;F=E7W9A<FYO9&5S*`HA(`D)\"0D)\"2A.\nM;V1E(\"HI*'-U8BT^<W5B<V5L96-T*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\nM\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P(\"L@,2D[\"B$@\nM\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U%U97)Y.@HA(`D)\"7L*\nM(2`)\"0D)475E<GD)*G%R>2`](\"A1=65R>2`J*6YO9&4[\"B$@\"B$@\"0D)\"6-H\nM86YG95]V87)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*2AQ<GDM/G1A<F=E=$QI\nM<W0I+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6YE=U]I;F1E>\"P*(2`)\nM\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`HA(`D)\"0EC:&%N9V5?=F%R;F]D97,H\nM\"B$@\"0D)\"0D)*$YO9&4@*BDH<7)Y+3YQ=6%L*2P*(2`)\"0D)\"0ER=%]I;F1E\nM>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*\nM(2`*(2`)\"0D)8VAA;F=E7W9A<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HI*'%R\nM>2T^:&%V:6YG475A;\"DL\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D);F5W\nM7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"B$@\"0D)\"6-H86YG\nM95]V87)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*2AQ<GDM/F=R;W5P0VQA=7-E\nM*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\nM\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0ED\nM969A=6QT.@HA(`D)\"65L;V<H3D]424-%+\"`B=6YK;F]W;B!N;V1E('1A9R`E\nM9\"!I;B!C:&%N9V5?=F%R;F]D97,H*2(L(&YO9&5486<H;F]D92DI.PHA(`D)\nM\"65L;V<H3D]424-%+\"`B3F]D92!I<SH@)7,B+\"!N;V1E5&]3=')I;F<H;F]D\nM92DI.PHA(`D)\"6)R96%K.PHA(`HA(`HA(`E]\"B$@?0HA(`HA(`HA(\"\\J\"B$@\nM(\"H@;6]D:69Y06=G<F5G57!L979E;\"`M\"B$@(\"H)26X@=&AE(&YE=VQY(&-R\nM96%T960@<W5B;&EN:R!F;W(@86X@86=G<F5G871E(&-O;'5M;B!U<V5D(&EN\nM\"B$@(\"H)=&AE('%U86QI9FEC871I;VXL('=E(&UU<W0@861J=7-T('1H92!V\nM87)L979E;'-U<\"!I;B!A;&P@=&AE\"B$@(\"H)=F%R(&YO9&5S+@HA(\"`J+PHA\nM('-T871I8R!V;VED\"B$@;6]D:69Y06=G<F5G57!L979E;\"A.;V1E(\"IN;V1E\nM*0HA('L*(2`):68@*&YO9&4@/3T@3E5,3\"D*(2`)\"7)E='5R;CL*(2`*(2`)\nM<W=I=&-H*&YO9&5486<H;F]D92DI('L*(2`)\"6-A<V4@5%]487)G971%;G1R\nM>3H*(2`)\"0E[\"B$@\"0D)\"51A<F=E=$5N=')Y\"2IT;&4@/2`H5&%R9V5T16YT\nM<GD@*BEN;V1E.PHA(`HA(`D)\"0EM;V1I9GE!9V=R96=5<&QE=F5L*`HA(`D)\nM\"0D)\"2A.;V1E(\"HI*'1L92T^97AP<BDI.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*\nM(2`*(2`)\"6-A<V4@5%]!9V=R96<Z\"B$@\"0D)>PHA(`D)\"0E!9V=R96<)*F%G\nM9R`](\"A!9V=R96<@*BEN;V1E.PHA(`HA(`D)\"0EM;V1I9GE!9V=R96=5<&QE\nM=F5L*`HA(`D)\"0D)\"2A.;V1E(\"HI*&%G9RT^=&%R9V5T*2D[\"B$@\"0D)?0HA\nM(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T5X<'(Z\"B$@\"0D)>PHA(`D)\"0E%\nM>'!R\"2IE>'`@/2`H17AP<B`J*6YO9&4[\"B$@\"B$@\"0D)\"6UO9&EF>4%G9W)E\nM9U5P;&5V96PH\"B$@\"0D)\"0D)*$YO9&4@*BDH97AP+3YA<F=S*2D[\"B$@\"0D)\nM?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47TET97(Z\"B$@\"0D)>PHA(`D)\nM\"0E)=&5R\"2II=&5R(#T@*$ET97(@*BEN;V1E.PHA(`HA(`D)\"0EM;V1I9GE!\nM9V=R96=5<&QE=F5L*`HA(`D)\"0D)\"2A.;V1E(\"HI*&ET97(M/FET97)E>'!R\nM*2D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T%R<F%Y4F5F\nM.@HA(`D)\"7L*(2`)\"0D)07)R87E2968)*G)E9B`](\"A!<G)A>5)E9B`J*6YO\nM9&4[\"B$@\"B$@\"0D)\"6UO9&EF>4%G9W)E9U5P;&5V96PH\"B$@\"0D)\"0D)*$YO\nM9&4@*BDH<F5F+3YR969U<'!E<FEN9&5X<'(I*3L*(2`)\"0D);6]D:69Y06=G\nM<F5G57!L979E;\"@*(2`)\"0D)\"0DH3F]D92`J*2AR968M/G)E9FQO=V5R:6YD\nM97AP<BDI.PHA(`D)\"0EM;V1I9GE!9V=R96=5<&QE=F5L*`HA(`D)\"0D)\"2A.\nM;V1E(\"HI*')E9BT^<F5F97AP<BDI.PHA(`D)\"0EM;V1I9GE!9V=R96=5<&QE\nM=F5L*`HA(`D)\"0D)\"2A.;V1E(\"HI*')E9BT^<F5F87-S9VYE>'!R*2D[\"B$@\nM\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U9A<CH*(2`)\"0E[\"B$@\nM\"0D)\"59A<@DJ=F%R(#T@*%9A<B`J*6YO9&4[\"B$@\"B$@\"0D)\"79A<BT^=F%R\nM;&5V96QS=7`K*SL*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?\nM4&%R86TZ\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?0V]N<W0Z\"B$@\"0D)\nM8G)E86L[\"B$@\"B$@\"0EC87-E(%1?3&ES=#H*(2`)\"0E[\"B$@\"0D)\"4QI<W0)\nM*FP[\"B$@\"B$@\"0D)\"69O<F5A8V@@*&PL(\"A,:7-T(\"HI;F]D92D*(2`)\"0D)\nM\"6UO9&EF>4%G9W)E9U5P;&5V96PH\"B$@\"0D)\"0D)\"2A.;V1E(\"HI;&9I<G-T\nM*&PI*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?4W5B3&EN\nM:SH*(2`)\"0E[\"B$@\"0D)\"5-U8DQI;FL)*G-U8B`](\"A3=6),:6YK(\"HI;F]D\nM93L*(2`*(2`)\"0D);6]D:69Y06=G<F5G57!L979E;\"@*(2`)\"0D)\"0DH3F]D\nM92`J*2AS=6(M/FQE9G1H86YD*2D[\"B$@\"B$@\"0D)\"6UO9&EF>4%G9W)E9U5P\nM;&5V96PH\"B$@\"0D)\"0D)*$YO9&4@*BDH<W5B+3YO<&5R*2D[\"B$@\"B$@\"0D)\nM\"6UO9&EF>4%G9W)E9U5P;&5V96PH\"B$@\"0D)\"0D)*$YO9&4@*BDH<W5B+3YS\nM=6)S96QE8W0I*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?\nM475E<GDZ\"B$@\"0D)>PHA(`D)\"0E1=65R>0DJ<7)Y(#T@*%%U97)Y(\"HI;F]D\nM93L*(2`*(2`)\"0D);6]D:69Y06=G<F5G57!L979E;\"@*(2`)\"0D)\"0DH3F]D\nM92`J*2AQ<GDM/G1A<F=E=$QI<W0I*3L*(2`*(2`)\"0D);6]D:69Y06=G<F5G\nM57!L979E;\"@*(2`)\"0D)\"0DH3F]D92`J*2AQ<GDM/G%U86PI*3L*(2`*(2`)\nM\"0D);6]D:69Y06=G<F5G57!L979E;\"@*(2`)\"0D)\"0DH3F]D92`J*2AQ<GDM\nM/FAA=FEN9U%U86PI*3L*(2`*(2`)\"0D);6]D:69Y06=G<F5G57!L979E;\"@*\nM(2`)\"0D)\"0DH3F]D92`J*2AQ<GDM/F=R;W5P0VQA=7-E*2D[\"B$@\"0D)?0HA\nM(`D)\"6)R96%K.PHA(`HA(`D)9&5F875L=#H*(2`)\"0EE;&]G*$Y/5$E#12P@\nM(G5N:VYO=VX@;F]D92!T86<@)60@:6X@;6]D:69Y06=G<F5G57!L979E;\"@I\nM(BP@;F]D951A9RAN;V1E*2D[\"B$@\"0D)96QO9RA.3U1)0T4L(\").;V1E(&ES\nM.B`E<R(L(&YO9&54;U-T<FEN9RAN;V1E*2D[\"B$@\"0D)8G)E86L[\"B$@\"B$@\nM\"B$@\"7T*(2!]\"B$@\"B$@\"B$@+RH*(2`@*B!M;V1I9GE!9V=R96=#:&%N9V56\nM87)N;V1E<R`M\"B$@(\"H)0VAA;F=E('1H92!V87(@;F]D97,@:6X@82!S=6)L\nM:6YK(&-R96%T960@9F]R(&%N(&%G9W)E9V%T92!C;VQU;6X*(2`@*@EU<V5D\nM(&EN('1H92!Q=6%L:69I8V%T:6]N('1H870@:7,@<W5B:F5C=\"!O9B!T:&4@\nM86=G<F5G871E\"B$@(\"H)9G5N8W1I;VX@=&\\@<&]I;G0@=&\\@=&AE(&-O<G)E\nM8W0@;&]C86P@4E1%+@HA(\"`J+PHA('-T871I8R!V;VED\"B$@;6]D:69Y06=G\nM<F5G0VAA;F=E5F%R;F]D97,H3F]D92`J*FYO9&50='(L(&EN=\"!R=%]I;F1E\nM>\"P@:6YT(&YE=U]I;F1E>\"P@:6YT('-U8FQE=F5L<U]U<\"D*(2![\"B$@\"4YO\nM9&4)*FYO9&4@/2`J;F]D95!T<CL*(2`*(2`):68@*&YO9&4@/3T@3E5,3\"D*\nM(2`)\"7)E='5R;CL*(2`*(2`)<W=I=&-H*&YO9&5486<H;F]D92DI('L*(2`)\nM\"6-A<V4@5%]487)G971%;G1R>3H*(2`)\"0E[\"B$@\"0D)\"51A<F=E=$5N=')Y\nM\"2IT;&4@/2`H5&%R9V5T16YT<GD@*BEN;V1E.PHA(`HA(`D)\"0EM;V1I9GE!\nM9V=R96=#:&%N9V5687)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAT;&4M\nM/F5X<'(I*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\nM\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\nM\"B$@\"0EC87-E(%1?06=G<F5G.@HA(`D)\"7L*(2`)\"0D)06=G<F5G\"2IA9V<@\nM/2`H06=G<F5G(\"HI;F]D93L*(2`*(2`)\"0D);6]D:69Y06=G<F5G0VAA;F=E\nM5F%R;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H86=G+3YT87)G970I*2P*\nM(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)\nM<W5B;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E\nM(%1?1W)O=7!#;&%U<V4Z\"B$@\"0D)>PHA(`D)\"0E'<F]U<$-L875S90DJ9W)P\nM(#T@*$=R;W5P0VQA=7-E(\"HI;F]D93L*(2`*(2`)\"0D);6]D:69Y06=G<F5G\nM0VAA;F=E5F%R;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H9W)P+3YE;G1R\nM>2DI+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6YE=U]I;F1E>\"P*(2`)\nM\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\nM\"6-A<V4@5%]%>'!R.@HA(`D)\"7L*(2`)\"0D)17AP<@DJ97AP(#T@*$5X<'(@\nM*BEN;V1E.PHA(`HA(`D)\"0EM;V1I9GE!9V=R96=#:&%N9V5687)N;V1E<R@*\nM(2`)\"0D)\"0DH3F]D92`J*BDH)BAE>'`M/F%R9W,I*2P*(2`)\"0D)\"0ER=%]I\nM;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P\nM*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?271E<CH*(2`)\nM\"0E[\"B$@\"0D)\"4ET97()*FET97(@/2`H271E<B`J*6YO9&4[\"B$@\"B$@\"0D)\nM\"6UO9&EF>4%G9W)E9T-H86YG959A<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HJ\nM*2@F*&ET97(M/FET97)E>'!R*2DL\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\nM\"0D);F5W7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)?0HA\nM(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T%R<F%Y4F5F.@HA(`D)\"7L*(2`)\nM\"0D)07)R87E2968)*G)E9B`](\"A!<G)A>5)E9B`J*6YO9&4[\"B$@\"B$@\"0D)\nM\"6UO9&EF>4%G9W)E9T-H86YG959A<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HJ\nM*2@F*')E9BT^<F5F=7!P97)I;F1E>'!R*2DL\"B$@\"0D)\"0D)<G1?:6YD97@L\nM\"B$@\"0D)\"0D);F5W7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\nM\"0D)\"6UO9&EF>4%G9W)E9T-H86YG959A<FYO9&5S*`HA(`D)\"0D)\"2A.;V1E\nM(\"HJ*2@F*')E9BT^<F5F;&]W97)I;F1E>'!R*2DL\"B$@\"0D)\"0D)<G1?:6YD\nM97@L\"B$@\"0D)\"0D);F5W7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\nM\"B$@\"0D)\"6UO9&EF>4%G9W)E9T-H86YG959A<FYO9&5S*`HA(`D)\"0D)\"2A.\nM;V1E(\"HJ*2@F*')E9BT^<F5F97AP<BDI+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA\nM(`D)\"0D)\"6YE=U]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\nM\"0EM;V1I9GE!9V=R96=#:&%N9V5687)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J\nM*BDH)BAR968M/G)E9F%S<V=N97AP<BDI+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA\nM(`D)\"0D)\"6YE=U]I;F1E>\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\nM\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]687(Z\"B$@\"0D)>PHA(`D)\nM\"0E687()*G9A<B`](\"A687(@*BEN;V1E.PHA(`HA(`D)\"0EI9B`H=F%R+3YV\nM87)L979E;'-U<\"`]/2!S=6)L979E;'-?=7`@)B8*(2`)\"0D)\"0EV87(M/G9A\nM<FYO(#T](')T7VEN9&5X*2![\"B$@\"0D)\"0EV87(@/2!C;W!Y3V)J96-T*'9A\nM<BD[\"B$@\"0D)\"0EV87(M/G9A<FYO(#T@;F5W7VEN9&5X.PHA(`D)\"0D)=F%R\nM+3YV87)N;V]L9\"`](&YE=U]I;F1E>#L*(2`)\"0D)\"79A<BT^=F%R;&5V96QS\nM=7`@/2`P.PHA(`HA(`D)\"0D)*FYO9&50='(@/2`H3F]D92`J*79A<CL*(2`)\nM\"0D)?0HA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]087)A;3H*\nM(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]#;VYS=#H*(2`)\"0EB<F5A:SL*\nM(2`*(2`)\"6-A<V4@5%],:7-T.@HA(`D)\"7L*(2`)\"0D)3&ES=`DJ;#L*(2`*\nM(2`)\"0D)9F]R96%C:\"`H;\"P@*$QI<W0@*BEN;V1E*0HA(`D)\"0D);6]D:69Y\nM06=G<F5G0VAA;F=E5F%R;F]D97,H\"B$@\"0D)\"0D)\"2A.;V1E(\"HJ*2@F;&9I\nM<G-T*&PI*2P*(2`)\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)\"6YE=U]I;F1E\nM>\"P*(2`)\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\nM\"B$@\"B$@\"0EC87-E(%1?4W5B3&EN:SH*(2`)\"0E[\"B$@\"0D)\"5-U8DQI;FL)\nM*G-U8B`](\"A3=6),:6YK(\"HI;F]D93L*(2`*(2`)\"0D);6]D:69Y06=G<F5G\nM0VAA;F=E5F%R;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H<W5B+3YL969T\nM:&%N9\"DI+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6YE=U]I;F1E>\"P*\nM(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`HA(`D)\"0EM;V1I9GE!9V=R96=#\nM:&%N9V5687)N;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAS=6(M/F]P97(I\nM*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\nM\"0D)<W5B;&5V96QS7W5P*3L*(2`*(2`)\"0D);6]D:69Y06=G<F5G0VAA;F=E\nM5F%R;F]D97,H\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H<W5B+3YS=6)S96QE8W0I\nM*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\nM\"0D)<W5B;&5V96QS7W5P(\"L@,2D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA\nM(`D)8V%S92!47U%U97)Y.@HA(`D)\"7L*(2`)\"0D)475E<GD)*G%R>2`](\"A1\nM=65R>2`J*6YO9&4[\"B$@\"B$@\"0D)\"6UO9&EF>4%G9W)E9T-H86YG959A<FYO\nM9&5S*`HA(`D)\"0D)\"2A.;V1E(\"HJ*2@F*'%R>2T^=&%R9V5T3&ES=\"DI+`HA\nM(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"6YE=U]I;F1E>\"P*(2`)\"0D)\"0ES\nM=6)L979E;'-?=7`I.PHA(`HA(`D)\"0EM;V1I9GE!9V=R96=#:&%N9V5687)N\nM;V1E<R@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAQ<GDM/G%U86PI*2P*(2`)\"0D)\nM\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)<W5B;&5V\nM96QS7W5P*3L*(2`*(2`)\"0D);6]D:69Y06=G<F5G0VAA;F=E5F%R;F]D97,H\nM\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H<7)Y+3YH879I;F=1=6%L*2DL\"B$@\"0D)\nM\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D);F5W7VEN9&5X+`HA(`D)\"0D)\"7-U8FQE\nM=F5L<U]U<\"D[\"B$@\"B$@\"0D)\"6UO9&EF>4%G9W)E9T-H86YG959A<FYO9&5S\nM*`HA(`D)\"0D)\"2A.;V1E(\"HJ*2@F*'%R>2T^9W)O=7!#;&%U<V4I*2P*(2`)\nM\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0EN97=?:6YD97@L\"B$@\"0D)\"0D)<W5B\nM;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0ED969A=6QT\nM.@HA(`D)\"65L;V<H3D]424-%+\"`B=6YK;F]W;B!N;V1E('1A9R`E9\"!I;B!M\nM;V1I9GE!9V=R96=#:&%N9V5687)N;V1E<R@I(BP@;F]D951A9RAN;V1E*2D[\nM\"B$@\"0D)96QO9RA.3U1)0T4L(\").;V1E(&ES.B`E<R(L(&YO9&54;U-T<FEN\nM9RAN;V1E*2D[\"B$@\"0D)8G)E86L[\"B$@\"B$@\"B$@\"7T*(2!]\"B$@\"B$@\"B$@\nM+RH*(2`@*B!M;V1I9GE!9V=R96=$<F]P475A;\"`M\"B$@(\"H)<F5M;W9E('1H\nM92!P=7)E(&%G9W)E9R!C;&%S92!F<F]M(&$@<75A;&EF:6-A=&EO;@HA(\"`J\nM+PHA('-T871I8R!V;VED\"B$@;6]D:69Y06=G<F5G1')O<%%U86PH3F]D92`J\nM*FYO9&50='(L($YO9&4@*F]R:6=N;V1E+\"!%>'!R(\"IE>'!R*0HA('L*(2`)\nM3F]D90DJ;F]D92`](\"IN;V1E4'1R.PHA(`HA(`EI9B`H;F]D92`]/2!.54Q,\nM*0HA(`D)<F5T=7)N.PHA(`HA(`ES=VET8V@H;F]D951A9RAN;V1E*2D@>PHA\nM(`D)8V%S92!47U9A<CH*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]!9V=R\nM96<Z\"B$@\"0D)>PHA(`D)\"0E!9V=R96<)*F%G9R`](\"A!9V=R96<@*BEN;V1E\nM.PHA(`D)\"0E!9V=R96<)*F]A9V<@/2`H06=G<F5G(\"HI;W)I9VYO9&4[\"B$@\nM\"B$@\"0D)\"6UO9&EF>4%G9W)E9T1R;W!1=6%L*`HA(`D)\"0D)\"2A.;V1E(\"HJ\nM*2@F*&%G9RT^=&%R9V5T*2DL\"B$@\"0D)\"0D)*$YO9&4@*BDH;V%G9RT^=&%R\nM9V5T*2P*(2`)\"0D)\"0EE>'!R*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\nM\"0EC87-E(%1?4&%R86TZ\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?0V]N\nM<W0Z\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?1W)O=7!#;&%U<V4Z\"B$@\nM\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?17AP<CH*(2`)\"0E[\"B$@\"0D)\"45X\nM<'()*G1H:7-?97AP<B`](\"A%>'!R(\"HI;F]D93L*(2`)\"0D)17AP<@DJ;W)I\nM9U]E>'!R(#T@*$5X<'(@*BEO<FEG;F]D93L*(2`*(2`)\"0D):68@*&]R:6=?\nM97AP<B`]/2!E>'!R*2![\"B$@\"0D)\"0E#;VYS=`DJ8W1R=64[\"B$@\"B$@\"0D)\nM\"0EI9B`H97AP<BT^='EP94]I9\"`A/2!\"3T],3TE$*0HA(`D)\"0D)\"65L;V<H\nM15)23U(L\"B$@\"0D)\"0D)\"2)A9V=R96=A=&4@97AP<F5S<VEO;B!I;B!Q=6%L\nM:69I8V%T:6]N(&ES;B=T(&]F('1Y<&4@8F]O;\"(I.PHA(`D)\"0D)8W1R=64@\nM/2!M86ME3F]D92A#;VYS=\"D[\"B$@\"0D)\"0EC=')U92T^8V]N<W1T>7!E(#T@\nM0D]/3$])1#L*(2`)\"0D)\"6-T<G5E+3YC;VYS=&QE;B`](#$[\"B$@\"0D)\"0EC\nM=')U92T^8V]N<W1I<VYU;&P@/2!&04Q313L*(2`)\"0D)\"6-T<G5E+3YC;VYS\nM='9A;'5E(#T@*$1A='5M*512544[\"B$@\"0D)\"0EC=')U92T^8V]N<W1B>79A\nM;\"`](%12544[\"B$@\"B$@\"0D)\"0DJ;F]D95!T<B`](\"A.;V1E(\"HI8W1R=64[\nM\"B$@\"0D)\"7T*(2`)\"0D)96QS90HA(`D)\"0D);6]D:69Y06=G<F5G1')O<%%U\nM86PH\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H=&AI<U]E>'!R+3YA<F=S*2DL\"B$@\nM\"0D)\"0D)*$YO9&4@*BDH;W)I9U]E>'!R+3YA<F=S*2P*(2`)\"0D)\"0EE>'!R\nM*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?271E<CH*(2`)\nM\"0E[\"B$@\"0D)\"4ET97()*FET97(@/2`H271E<B`J*6YO9&4[\"B$@\"0D)\"4ET\nM97()*F]I=&5R(#T@*$ET97(@*BEO<FEG;F]D93L*(2`*(2`)\"0D);6]D:69Y\nM06=G<F5G1')O<%%U86PH\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H:71E<BT^:71E\nM<F5X<'(I*2P*(2`)\"0D)\"0DH3F]D92`J*2AO:71E<BT^:71E<F5X<'(I+`HA\nM(`D)\"0D)\"65X<'(I.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@\nM5%]!<G)A>5)E9CH*(2`)\"0E[\"B$@\"0D)\"4%R<F%Y4F5F\"2IR968@/2`H07)R\nM87E2968@*BEN;V1E.PHA(`D)\"0E!<G)A>5)E9@DJ;W)E9B`](\"A!<G)A>5)E\nM9B`J*6]R:6=N;V1E.PHA(`HA(`D)\"0EM;V1I9GE!9V=R96=$<F]P475A;\"@*\nM(2`)\"0D)\"0DH3F]D92`J*BDH)BAR968M/G)E9G5P<&5R:6YD97AP<BDI+`HA\nM(`D)\"0D)\"2A.;V1E(\"HI*&]R968M/G)E9G5P<&5R:6YD97AP<BDL\"B$@\"0D)\nM\"0D)97AP<BD[\"B$@\"0D)\"6UO9&EF>4%G9W)E9T1R;W!1=6%L*`HA(`D)\"0D)\nM\"2A.;V1E(\"HJ*2@F*')E9BT^<F5F;&]W97)I;F1E>'!R*2DL\"B$@\"0D)\"0D)\nM*$YO9&4@*BDH;W)E9BT^<F5F;&]W97)I;F1E>'!R*2P*(2`)\"0D)\"0EE>'!R\nM*3L*(2`)\"0D);6]D:69Y06=G<F5G1')O<%%U86PH\"B$@\"0D)\"0D)*$YO9&4@\nM*BHI*\"8H<F5F+3YR969E>'!R*2DL\"B$@\"0D)\"0D)*$YO9&4@*BDH;W)E9BT^\nM<F5F97AP<BDL\"B$@\"0D)\"0D)97AP<BD[\"B$@\"0D)\"6UO9&EF>4%G9W)E9T1R\nM;W!1=6%L*`HA(`D)\"0D)\"2A.;V1E(\"HJ*2@F*')E9BT^<F5F87-S9VYE>'!R\nM*2DL\"B$@\"0D)\"0D)*$YO9&4@*BDH;W)E9BT^<F5F87-S9VYE>'!R*2P*(2`)\nM\"0D)\"0EE>'!R*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?\nM3&ES=#H*(2`)\"0E[\"B$@\"0D)\"4QI<W0)*FP[\"B$@\"0D)\"4QI<W0)*F]L(#T@\nM*$QI<W0@*BEO<FEG;F]D93L*(2`)\"0D):6YT\"6QI(#T@,#L*(2`*(2`)\"0D)\nM9F]R96%C:\"`H;\"P@*$QI<W0@*BEN;V1E*2![\"B$@\"0D)\"0EM;V1I9GE!9V=R\nM96=$<F]P475A;\"@*(2`)\"0D)\"0D)*$YO9&4@*BHI*\"8H;&9I<G-T*&PI*2DL\nM\"B$@\"0D)\"0D)\"2A.;V1E(\"HI;G1H*&QI+\"!O;\"DL\"B$@\"0D)\"0D)\"65X<'(I\nM.PHA(`D)\"0D);&DK*SL*(2`)\"0D)?0HA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*\nM(2`)\"6-A<V4@5%]3=6),:6YK.@HA(`D)\"7L*(2`)\"0D)4W5B3&EN:PDJ<W5B\nM(#T@*%-U8DQI;FL@*BEN;V1E.PHA(`D)\"0E3=6),:6YK\"2IO<W5B(#T@*%-U\nM8DQI;FL@*BEO<FEG;F]D93L*(2`*(2`)\"0D);6]D:69Y06=G<F5G1')O<%%U\nM86PH\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H<W5B+3YS=6)S96QE8W0I*2P*(2`)\nM\"0D)\"0DH3F]D92`J*2AO<W5B+3YS=6)S96QE8W0I+`HA(`D)\"0D)\"65X<'(I\nM.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]1=65R>3H*(2`)\nM\"0E[\"B$@\"0D)\"5%U97)Y\"2IQ<GD@/2`H475E<GD@*BEN;V1E.PHA(`D)\"0E1\nM=65R>0DJ;W%R>2`](\"A1=65R>2`J*6]R:6=N;V1E.PHA(`HA(`D)\"0EM;V1I\nM9GE!9V=R96=$<F]P475A;\"@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAQ<GDM/G%U\nM86PI*2P*(2`)\"0D)\"0DH3F]D92`J*2AO<7)Y+3YQ=6%L*2P*(2`)\"0D)\"0EE\nM>'!R*3L*(2`*(2`)\"0D);6]D:69Y06=G<F5G1')O<%%U86PH\"B$@\"0D)\"0D)\nM*$YO9&4@*BHI*\"8H<7)Y+3YH879I;F=1=6%L*2DL\"B$@\"0D)\"0D)*$YO9&4@\nM*BDH;W%R>2T^:&%V:6YG475A;\"DL\"B$@\"0D)\"0D)97AP<BD[\"B$@\"0D)?0HA\nM(`D)\"6)R96%K.PHA(`HA(`D)9&5F875L=#H*(2`)\"0EE;&]G*$Y/5$E#12P@\nM(G5N:VYO=VX@;F]D92!T86<@)60@:6X@;6]D:69Y06=G<F5G1')O<%%U86PH\nM*2(L(&YO9&5486<H;F]D92DI.PHA(`D)\"65L;V<H3D]424-%+\"`B3F]D92!I\nM<SH@)7,B+\"!N;V1E5&]3=')I;F<H;F]D92DI.PHA(`D)\"6)R96%K.PHA(`HA\nM(`HA(`E]\"B$@?0HA(`HA(`HA(\"\\J\"B$@(\"H@;6]D:69Y06=G<F5G36%K95-U\nM8FQI;FL@+0HA(\"`J\"4-R96%T92!A('-U8FQI;FL@;F]D92!F;W(@82!Q=6%L\nM:69I8V%T:6]N(&5X<')E<W-I;VX@=&AA=`HA(\"`J\"75S97,@86X@86=G<F5G\nM871E(&-O;'5M;B!O9B!A('9I97<*(2`@*B\\*(2!S=&%T:6,@4W5B3&EN:R`J\nM\"B$@;6]D:69Y06=G<F5G36%K95-U8FQI;FLH17AP<B`J;W)I9V5X<\"P@475E\nM<GD@*G!A<G-E=')E92D*(2![\"B$@\"5-U8DQI;FL)\"2IS=6)L:6YK.PHA(`E1\nM=65R>0D)*G-U8G%U97)Y.PHA(`E.;V1E\"0DJ<W5B<75A;#L*(2`)4F%N9V54\nM8FQ%;G1R>0DJ<G1E.PHA(`E!9V=R96<)\"2IA9V=R96<[\"B$@\"59A<@D)*G1A\nM<F=E=#L*(2`)5&%R9V5T16YT<GD)*G1L93L*(2`)4F5S9&]M\"0DJ<F5S9&]M\nM.PHA(`E%>'!R\"0DJ97AP(#T@8V]P>4]B:F5C=\"AO<FEG97AP*3L*(2`*(2`)\nM:68@*&YO9&5486<H;G1H*#`L(&5X<\"T^87)G<RDI(#T](%1?06=G<F5G*0HA\nM(`D):68@*&YO9&5486<H;G1H*#$L(&5X<\"T^87)G<RDI(#T](%1?06=G<F5G\nM*0HA(`D)\"65L;V<H15)23U(L(\")R97=R:71E.B!C;VUP87)I<VEO;B!O9B`R\nM(&%G9W)E9V%T92!C;VQU;6YS(&YO=\"!S=7!P;W)T960B*3L*(2`)\"65L<V4*\nM(2`)\"0EE;&]G*$524D]2+\"`B<F5W<FET93H@86=G<F5G871E(&-O;'5M;B!O\nM9B!V:65W(&UU<W0@8F4@870@<FEG=&@@<VED92!I;B!Q=6%L(BD[\"B$@\"B$@\nM\"6%G9W)E9R`](\"A!9V=R96<@*BEN=&@H,2P@97AP+3YA<F=S*3L*(2`)=&%R\nM9V5T\"3T@*%9A<B`J*2AA9V=R96<M/G1A<F=E=\"D[\"B$@\"7)T90D](\"A286YG\nM951B;$5N=')Y(\"HI;G1H*'1A<F=E=\"T^=F%R;F\\@+2`Q+\"!P87)S971R964M\nM/G)T86)L92D[\"B$@\"71L90D](&UA:V5.;V1E*%1A<F=E=$5N=')Y*3L*(2`)\nM<F5S9&]M\"3T@;6%K94YO9&4H4F5S9&]M*3L*(2`*(2`)86=G<F5G+3YU<V5N\nM=6QL<R`](%12544[\"B$@\"B$@\"7)E<V1O;2T^<F5S;F\\)/2`Q.PHA(`ER97-D\nM;VTM/G)E<W1Y<&4)/2`H*$]P97(@*BDH97AP+3YO<&5R*2DM/F]P<F5S=6QT\nM='EP93L*(2`)<F5S9&]M+3YR97-T>7!M;V0@/2`M,3L*(2`)<F5S9&]M+3YR\nM97-N86UE(#T@<'-T<F1U<\"@B/&YO;F%M93XB*3L*(2`)<F5S9&]M+3YR97-K\nM97D)/2`P.PHA(`ER97-D;VTM/G)E<VME>6]P(#T@,#L*(2`)<F5S9&]M+3YR\nM97-J=6YK\"3T@,#L*(2`*(2`)=&QE+3YR97-D;VT)/2!R97-D;VT[\"B$@\"71L\nM92T^97AP<@D](\"A.;V1E(\"HI86=G<F5G.PHA(`HA(`ES=6)Q=6%L(#T@8V]P\nM>4]B:F5C=\"AP87)S971R964M/G%U86PI.PHA(`EM;V1I9GE!9V=R96=$<F]P\nM475A;\"@H3F]D92`J*BDF<W5B<75A;\"P@*$YO9&4@*BEP87)S971R964M/G%U\nM86PL(&]R:6=E>'`I.PHA(`HA(`ES=6)L:6YK(#T@;6%K94YO9&4H4W5B3&EN\nM:RD[\"B$@\"7-U8FQI;FLM/G-U8DQI;FM4>7!E\"3T@15A04E]354),24Y+.PHA\nM(`ES=6)L:6YK+3YU<V5O<@D)/2!&04Q313L*(2`)<W5B;&EN:RT^;&5F=&AA\nM;F0)/2!L87!P96YD*$Y)3\"P@8V]P>4]B:F5C=\"AL9FER<W0H97AP+3YA<F=S\nM*2DI.PHA(`ES=6)L:6YK+3YO<&5R\"0D](&QA<'!E;F0H3DE,+\"!C;W!Y3V)J\nM96-T*&5X<\"DI.PHA(`ES=6)L:6YK+3YS=6)S96QE8W0)/2!.54Q,.PHA(`HA\nM(`ES=6)Q=65R>0D)/2!M86ME3F]D92A1=65R>2D[\"B$@\"7-U8FQI;FLM/G-U\nM8G-E;&5C=`D](\"A.;V1E(\"HI<W5B<75E<GD[\"B$@\"B$@\"7-U8G%U97)Y+3YC\nM;VUM86YD5'EP90D)/2!#341?4T5,14-4.PHA(`ES=6)Q=65R>2T^=71I;&ET\nM>5-T;70)\"3T@3E5,3#L*(2`)<W5B<75E<GDM/G)E<W5L=%)E;&%T:6]N\"3T@\nM,#L*(2`)<W5B<75E<GDM/FEN=&\\)\"0D]($Y53$P[\"B$@\"7-U8G%U97)Y+3YI\nM<U!O<G1A;`D)/2!&04Q313L*(2`)<W5B<75E<GDM/FES0FEN87)Y\"0D]($9!\nM3%-%.PHA(`ES=6)Q=65R>2T^=6YI;VYA;&P)\"3T@1D%,4T4[\"B$@\"7-U8G%U\nM97)Y+3YU;FEQ=65&;&%G\"0D]($Y53$P[\"B$@\"7-U8G%U97)Y+3YS;W)T0VQA\nM=7-E\"0D]($Y53$P[\"B$@\"7-U8G%U97)Y+3YR=&%B;&4)\"3T@;&%P<&5N9\"A.\nM24PL(')T92D[\"B$@\"7-U8G%U97)Y+3YT87)G971,:7-T\"0D](&QA<'!E;F0H\nM3DE,+\"!T;&4I.PHA(`ES=6)Q=65R>2T^<75A;`D)\"3T@<W5B<75A;#L*(2`)\nM<W5B<75E<GDM/F=R;W5P0VQA=7-E\"0D]($Y)3#L*(2`)<W5B<75E<GDM/FAA\nM=FEN9U%U86P)\"3T@3E5,3#L*(2`)<W5B<75E<GDM/FAA<T%G9W,)\"3T@5%)5\nM13L*(2`)<W5B<75E<GDM/FAA<U-U8DQI;FMS\"0D]($9!3%-%.PHA(`ES=6)Q\nM=65R>2T^=6YI;VY#;&%U<V4)\"3T@3E5,3#L*(2`*(2`*(2`);6]D:69Y06=G\nM<F5G57!L979E;\"@H3F]D92`J*7-U8FQI;FLI.PHA(`HA(`EM;V1I9GE!9V=R\nM96=#:&%N9V5687)N;V1E<R@H3F]D92`J*BDF*'-U8FQI;FLM/FQE9G1H86YD\nM*2P@=&%R9V5T+3YV87)N;RP*(2`)\"0DQ+\"!T87)G970M/G9A<FQE=F5L<W5P\nM*3L*(2`);6]D:69Y06=G<F5G0VAA;F=E5F%R;F]D97,H*$YO9&4@*BHI)BAS\nM=6)L:6YK+3YO<&5R*2P@=&%R9V5T+3YV87)N;RP*(2`)\"0DQ+\"!T87)G970M\nM/G9A<FQE=F5L<W5P*3L*(2`);6]D:69Y06=G<F5G0VAA;F=E5F%R;F]D97,H\nM*$YO9&4@*BHI)BAS=6)L:6YK+3YS=6)S96QE8W0I+\"!T87)G970M/G9A<FYO\nM+`HA(`D)\"3$L('1A<F=E=\"T^=F%R;&5V96QS=7`I.PHA(`HA(`ER971U<FX@\nM<W5B;&EN:SL*(2!]\"B$@\"B$@\"B$@+RH*(2`@*B!M;V1I9GE!9V=R96=1=6%L\nM(\"T*(2`@*@E396%R8V@@9F]R('%U86QI9FEC871I;VX@97AP<F5S<VEO;G,@\nM=&AA=\"!C;VYT86EN(&%G9W)E9V%T90HA(\"`J\"69U;F-T:6]N<R!A;F0@<W5B\nM<W1I='5R92!T:&5M(&)Y('-U8FQI;FMS+B!4:&5S92!E>'!R97-S:6]N<PHA\nM(\"`J\"6]R:6=I;F%L;'D@8V]M92!F<F]M('%U86QI9FEC871I;VYS('1H870@\nM=7-E(&%G9W)E9V%T92!C;VQU;6YS\"B$@(\"H);V8@82!V:65W+@HA(\"`J+PHA\nM('-T871I8R!V;VED\"B$@;6]D:69Y06=G<F5G475A;\"A.;V1E(\"HJ;F]D95!T\nM<BP@475E<GD@*G!A<G-E=')E92D*(2![\"B$@\"4YO9&4)*FYO9&4@/2`J;F]D\nM95!T<CL*(2`*(2`):68@*&YO9&4@/3T@3E5,3\"D*(2`)\"7)E='5R;CL*(2`*\nM(2`)<W=I=&-H*&YO9&5486<H;F]D92DI('L*(2`)\"6-A<V4@5%]687(Z\"B$@\nM\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?4&%R86TZ\"B$@\"0D)8G)E86L[\"B$@\nM\"B$@\"0EC87-E(%1?0V]N<W0Z\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?\nM1W)O=7!#;&%U<V4Z\"B$@\"0D)>PHA(`D)\"0E'<F]U<$-L875S90DJ9W)P(#T@\nM*$=R;W5P0VQA=7-E(\"HI;F]D93L*(2`*(2`)\"0D);6]D:69Y06=G<F5G475A\nM;\"@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAG<G`M/F5N=')Y*2DL\"B$@\"0D)\"0D)\nM<&%R<V5T<F5E*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?\nM17AP<CH*(2`)\"0E[\"B$@\"0D)\"45X<'()*F5X<\"`](\"A%>'!R(\"HI;F]D93L*\nM(2`)\"0D)4W5B3&EN:PDJ<W5B.PHA(`HA(`HA(`D)\"0EI9B`H;&5N9W1H*&5X\nM<\"T^87)G<RD@(3T@,BD@>PHA(`D)\"0D);6]D:69Y06=G<F5G475A;\"@*(2`)\nM\"0D)\"0DH3F]D92`J*BDH)BAE>'`M/F%R9W,I*2P*(2`)\"0D)\"0EP87)S971R\nM964I.PHA(`D)\"0D)8G)E86L[\"B$@\"0D)\"7T*(2`*(2`)\"0D):68@*&YO9&54\nM86<H;G1H*#`L(&5X<\"T^87)G<RDI(\"$](%1?06=G<F5G(\"8F\"B$@\"0D)\"0EN\nM;V1E5&%G*&YT:\"@Q+\"!E>'`M/F%R9W,I*2`A/2!47T%G9W)E9RD@>PHA(`HA\nM(`D)\"0D);6]D:69Y06=G<F5G475A;\"@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAE\nM>'`M/F%R9W,I*2P*(2`)\"0D)\"0EP87)S971R964I.PHA(`D)\"0D)8G)E86L[\nM\"B$@\"0D)\"7T*(2`*(2`)\"0D)<W5B(#T@;6]D:69Y06=G<F5G36%K95-U8FQI\nM;FLH97AP+`HA(`D)\"0D)\"7!A<G-E=')E92D[\"B$@\"B$@\"0D)\"2IN;V1E4'1R\nM(#T@*$YO9&4@*BES=6([\"B$@\"0D)\"7!A<G-E=')E92T^:&%S4W5B3&EN:W,@\nM/2!44E5%.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%])=&5R\nM.@HA(`D)\"7L*(2`)\"0D)271E<@DJ:71E<B`](\"A)=&5R(\"HI;F]D93L*(2`*\nM(2`)\"0D);6]D:69Y06=G<F5G475A;\"@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAI\nM=&5R+3YI=&5R97AP<BDI+`HA(`D)\"0D)\"7!A<G-E=')E92D[\"B$@\"0D)?0HA\nM(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T%R<F%Y4F5F.@HA(`D)\"7L*(2`)\nM\"0D)07)R87E2968)*G)E9B`](\"A!<G)A>5)E9B`J*6YO9&4[\"B$@\"B$@\"0D)\nM\"6UO9&EF>4%G9W)E9U%U86PH\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H<F5F+3YR\nM969U<'!E<FEN9&5X<'(I*2P*(2`)\"0D)\"0EP87)S971R964I.PHA(`D)\"0EM\nM;V1I9GE!9V=R96=1=6%L*`HA(`D)\"0D)\"2A.;V1E(\"HJ*2@F*')E9BT^<F5F\nM;&]W97)I;F1E>'!R*2DL\"B$@\"0D)\"0D)<&%R<V5T<F5E*3L*(2`)\"0D);6]D\nM:69Y06=G<F5G475A;\"@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAR968M/G)E9F5X\nM<'(I*2P*(2`)\"0D)\"0EP87)S971R964I.PHA(`D)\"0EM;V1I9GE!9V=R96=1\nM=6%L*`HA(`D)\"0D)\"2A.;V1E(\"HJ*2@F*')E9BT^<F5F87-S9VYE>'!R*2DL\nM\"B$@\"0D)\"0D)<&%R<V5T<F5E*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\nM\"0EC87-E(%1?3&ES=#H*(2`)\"0E[\"B$@\"0D)\"4QI<W0)*FP[\"B$@\"B$@\"0D)\nM\"69O<F5A8V@@*&PL(\"A,:7-T(\"HI;F]D92D*(2`)\"0D)\"6UO9&EF>4%G9W)E\nM9U%U86PH\"B$@\"0D)\"0D)\"2A.;V1E(\"HJ*2@F*&QF:7)S=\"AL*2DI+`HA(`D)\nM\"0D)\"0EP87)S971R964I.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A\nM<V4@5%]3=6),:6YK.@HA(`D)\"7L*(2`)\"0D)4W5B3&EN:PDJ<W5B(#T@*%-U\nM8DQI;FL@*BEN;V1E.PHA(`HA(`D)\"0EM;V1I9GE!9V=R96=1=6%L*`HA(`D)\nM\"0D)\"2A.;V1E(\"HJ*2@F*'-U8BT^<W5B<V5L96-T*2DL\"B$@\"0D)\"0D)*%%U\nM97)Y(\"HI*'-U8BT^<W5B<V5L96-T*2D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA\nM(`HA(`D)8V%S92!47U%U97)Y.@HA(`D)\"7L*(2`)\"0D)475E<GD)*G%R>2`]\nM(\"A1=65R>2`J*6YO9&4[\"B$@\"B$@\"0D)\"6UO9&EF>4%G9W)E9U%U86PH\"B$@\nM\"0D)\"0D)*$YO9&4@*BHI*\"8H<7)Y+3YQ=6%L*2DL\"B$@\"0D)\"0D)<&%R<V5T\nM<F5E*3L*(2`*(2`)\"0D);6]D:69Y06=G<F5G475A;\"@*(2`)\"0D)\"0DH3F]D\nM92`J*BDH)BAQ<GDM/FAA=FEN9U%U86PI*2P*(2`)\"0D)\"0EP87)S971R964I\nM.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"61E9F%U;'0Z\"B$@\"0D)96QO\nM9RA.3U1)0T4L(\")U;FMN;W=N(&YO9&4@=&%G(\"5D(&EN(&UO9&EF>4%G9W)E\nM9U%U86PH*2(L(&YO9&5486<H;F]D92DI.PHA(`D)\"65L;V<H3D]424-%+\"`B\nM3F]D92!I<SH@)7,B+\"!N;V1E5&]3=')I;F<H;F]D92DI.PHA(`D)\"6)R96%K\nM.PHA(`HA(`HA(`E]\"B$@?0HA(`HA(`HA('-T871I8R!.;V1E(\"H*(2!&:6YD\nM36%T8VAI;F=43$5N=')Y*$QI<W0@*G1L:7-T+\"!C:&%R(\"IE7V%T=&YA;64I\nM\"B$@>PHA(`E,:7-T\"2`@(\"II.PHA(`HA(`EF;W)E86-H*&DL('1L:7-T*0HA\nM(`E[\"B$@\"0E487)G971%;G1R>2`J=&QE(#T@;&9I<G-T*&DI.PHA(`D)8VAA\nM<@D@(\"`J<F5S;F%M93L*(2`*(2`)\"7)E<VYA;64@/2!T;&4M/G)E<V1O;2T^\nM<F5S;F%M93L*(2`)\"6EF(\"@A<W1R8VUP*&5?871T;F%M92P@<F5S;F%M92DI\nM\"B$@\"0D)<F5T=7)N(\"AT;&4M/F5X<'(I.PHA(`E]\"B$@\"7)E='5R;B!.54Q,\nM.PHA('T*(2`*(2`*(2!S=&%T:6,@3F]D92`J\"B$@;6%K95]N=6QL*$]I9\"!T\nM>7!E*0HA('L*(2`)0V]N<W0)(\"`@*F,@/2!M86ME3F]D92A#;VYS=\"D[\"B$@\nM\"B$@\"6,M/F-O;G-T='EP92`]('1Y<&4[\"B$@\"6,M/F-O;G-T;&5N(#T@9V5T\nM7W1Y<&QE;BAT>7!E*3L*(2`)8RT^8V]N<W1V86QU92`](%!O:6YT97)'971$\nM871U;2A.54Q,*3L*(2`)8RT^8V]N<W1I<VYU;&P@/2!T<G5E.PHA(`EC+3YC\nM;VYS=&)Y=F%L(#T@9V5T7W1Y<&)Y=F%L*'1Y<&4I.PHA(`ER971U<FX@*$YO\nM9&4@*BD@8SL*(2!]\"B$@\"B$@\"B$@<W1A=&EC('9O:60*(2!A<'!L>5]225)?\nM=FEE=RA.;V1E(\"HJ;F]D95!T<BP@:6YT(')T7VEN9&5X+\"!286YG951B;$5N\nM=')Y(\"IR=&4L($QI<W0@*G1L:7-T+\"!I;G0@*FUO9&EF:65D+\"!I;G0@<W5B\nM;&5V96QS7W5P*0HA('L*(2`)3F]D90DJ;F]D92`](\"IN;V1E4'1R.PHA(`HA\nM(`EI9B`H;F]D92`]/2!.54Q,*0HA(`D)<F5T=7)N.PHA(`HA(`ES=VET8V@H\nM;F]D951A9RAN;V1E*2D@>PHA(`D)8V%S92!47U1A<F=E=$5N=')Y.@HA(`D)\nM\"7L*(2`)\"0D)5&%R9V5T16YT<GD)*G1L92`](\"A487)G971%;G1R>2`J*6YO\nM9&4[\"B$@\"B$@\"0D)\"6%P<&QY7U))4E]V:65W*`HA(`D)\"0D)\"2A.;V1E(\"HJ\nM*2@F*'1L92T^97AP<BDI+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"7)T\nM92P*(2`)\"0D)\"0ET;&ES=\"P*(2`)\"0D)\"0EM;V1I9FEE9\"P*(2`)\"0D)\"0ES\nM=6)L979E;'-?=7`I.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@\nM5%]!9V=R96<Z\"B$@\"0D)>PHA(`D)\"0E!9V=R96<)*F%G9R`](\"A!9V=R96<@\nM*BEN;V1E.PHA(`HA(`D)\"0EA<'!L>5]225)?=FEE=R@*(2`)\"0D)\"0DH3F]D\nM92`J*BDH)BAA9V<M/G1A<F=E=\"DI+`HA(`D)\"0D)\"7)T7VEN9&5X+`HA(`D)\nM\"0D)\"7)T92P*(2`)\"0D)\"0ET;&ES=\"P*(2`)\"0D)\"0EM;V1I9FEE9\"P*(2`)\nM\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\nM\"6-A<V4@5%]'<F]U<$-L875S93H*(2`)\"0E[\"B$@\"0D)\"4=R;W5P0VQA=7-E\nM\"2IG<G`@/2`H1W)O=7!#;&%U<V4@*BEN;V1E.PHA(`HA(`D)\"0EA<'!L>5]2\nM25)?=FEE=R@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAG<G`M/F5N=')Y*2DL\"B$@\nM\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)<G1E+`HA(`D)\"0D)\"71L:7-T+`HA\nM(`D)\"0D)\"6UO9&EF:65D+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)\nM?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T5X<'(Z\"B$@\"0D)>PHA(`D)\nM\"0E%>'!R\"2IE>'`@/2`H17AP<B`J*6YO9&4[\"B$@\"B$@\"0D)\"6%P<&QY7U))\nM4E]V:65W*`HA(`D)\"0D)\"2A.;V1E(\"HJ*2@F*&5X<\"T^87)G<RDI+`HA(`D)\nM\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)\"7)T92P*(2`)\"0D)\"0ET;&ES=\"P*(2`)\nM\"0D)\"0EM;V1I9FEE9\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`D)\"7T*\nM(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%])=&5R.@HA(`D)\"7L*(2`)\"0D)\nM271E<@DJ:71E<B`](\"A)=&5R(\"HI;F]D93L*(2`*(2`)\"0D)87!P;'E?4DE2\nM7W9I97<H\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H:71E<BT^:71E<F5X<'(I*2P*\nM(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ER=&4L\"B$@\"0D)\"0D)=&QI<W0L\nM\"B$@\"0D)\"0D);6]D:69I960L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\nM\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?07)R87E2968Z\"B$@\"0D)\nM>PHA(`D)\"0E!<G)A>5)E9@DJ<F5F(#T@*$%R<F%Y4F5F(\"HI;F]D93L*(2`*\nM(2`)\"0D)87!P;'E?4DE27W9I97<H\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H<F5F\nM+3YR969U<'!E<FEN9&5X<'(I*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\nM\"0ER=&4L\"B$@\"0D)\"0D)=&QI<W0L\"B$@\"0D)\"0D);6]D:69I960L\"B$@\"0D)\nM\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0D)87!P;'E?4DE27W9I97<H\"B$@\"0D)\nM\"0D)*$YO9&4@*BHI*\"8H<F5F+3YR969L;W=E<FEN9&5X<'(I*2P*(2`)\"0D)\nM\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ER=&4L\"B$@\"0D)\"0D)=&QI<W0L\"B$@\"0D)\nM\"0D);6]D:69I960L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0D)87!P\nM;'E?4DE27W9I97<H\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H<F5F+3YR969E>'!R\nM*2DL\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)<G1E+`HA(`D)\"0D)\"71L\nM:7-T+`HA(`D)\"0D)\"6UO9&EF:65D+`HA(`D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\nM\"B$@\"0D)\"6%P<&QY7U))4E]V:65W*`HA(`D)\"0D)\"2A.;V1E(\"HJ*2@F*')E\nM9BT^<F5F87-S9VYE>'!R*2DL\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)\nM<G1E+`HA(`D)\"0D)\"71L:7-T+`HA(`D)\"0D)\"6UO9&EF:65D+`HA(`D)\"0D)\nM\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S\nM92!47U9A<CH*(2`)\"0E[\"B$@\"0D)\"59A<@DJ=F%R(#T@*%9A<B`J*6YO9&4[\nM\"B$@\"B$@\"0D)\"6EF(\"AV87(M/G9A<FQE=F5L<W5P(#T]('-U8FQE=F5L<U]U\nM<\"`F)@HA(`D)\"0D)\"79A<BT^=F%R;F\\@/3T@<G1?:6YD97@I('L*(2`)\"0D)\nM\"4YO9&4)\"2IE>'`[\"B$@\"B$@\"0D)\"0EE>'`@/2!&:6YD36%T8VAI;F=43$5N\nM=')Y*`HA(`D)\"0D)\"0ET;&ES=\"P*(2`)\"0D)\"0D)9V5T7V%T=&YA;64H<G1E\nM+3YR96QI9\"P*(2`)\"0D)\"0D)\"79A<BT^=F%R871T;F\\I*3L*(2`*(2`)\"0D)\nM\"6EF(\"AE>'`@/3T@3E5,3\"D@>PHA(`D)\"0D)\"2IN;V1E4'1R(#T@;6%K95]N\nM=6QL*'9A<BT^=F%R='EP92D[\"B$@\"0D)\"0D)<F5T=7)N.PHA(`D)\"0D)?0HA\nM(`HA(`D)\"0D):68@*'9A<BT^=F%R;&5V96QS=7`@/B`P(\"8F\"B$@\"0D)\"0D)\nM\"6YO9&5486<H97AP*2`]/2!47U9A<BD@>PHA(`D)\"0D)\"65X<\"`](&-O<'E/\nM8FIE8W0H97AP*3L*(2`)\"0D)\"0DH*%9A<B`J*65X<\"DM/G9A<FQE=F5L<W5P\nM(#T@=F%R+3YV87)L979E;'-U<#L*(2`)\"0D)\"7T*(2`)\"0D)\"2IN;V1E4'1R\nM(#[email protected](`D)\"0D)*FUO9&EF:65D(#T@5%)513L*(2`)\"0D)?0HA(`D)\nM\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]087)A;3H*(2`)\"0EB<F5A\nM:SL*(2`*(2`)\"6-A<V4@5%]#;VYS=#H*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A\nM<V4@5%],:7-T.@HA(`D)\"7L*(2`)\"0D)3&ES=`DJ;#L*(2`*(2`)\"0D)9F]R\nM96%C:\"`H;\"P@*$QI<W0@*BEN;V1E*0HA(`D)\"0D)87!P;'E?4DE27W9I97<H\nM\"B$@\"0D)\"0D)\"2A.;V1E(\"HJ*2@F*&QF:7)S=\"AL*2DI+`HA(`D)\"0D)\"0ER\nM=%]I;F1E>\"P*(2`)\"0D)\"0D)<G1E+`HA(`D)\"0D)\"0ET;&ES=\"P*(2`)\"0D)\nM\"0D);6]D:69I960L\"B$@\"0D)\"0D)\"7-U8FQE=F5L<U]U<\"D[\"B$@\"0D)?0HA\nM(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U-U8DQI;FLZ\"B$@\"0D)>PHA(`D)\nM\"0E3=6),:6YK\"2IS=6(@/2`H4W5B3&EN:R`J*6YO9&4[\"B$@\"B$@\"0D)\"6%P\nM<&QY7U))4E]V:65W*`HA(`D)\"0D)\"2A.;V1E(\"HJ*2@F*'-U8BT^;&5F=&AA\nM;F0I*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ER=&4L\"B$@\"0D)\"0D)\nM=&QI<W0L\"B$@\"0D)\"0D);6]D:69I960L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P\nM*3L*(2`*(2`)\"0D)87!P;'E?4DE27W9I97<H\"B$@\"0D)\"0D)*$YO9&4@*BHI\nM*\"8H<W5B+3YS=6)S96QE8W0I*2P*(2`)\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\nM\"0ER=&4L\"B$@\"0D)\"0D)=&QI<W0L\"B$@\"0D)\"0D);6]D:69I960L\"B$@\"0D)\nM\"0D)<W5B;&5V96QS7W5P(\"L@,2D[\"B$@\"0D)?0HA(`D)\"6)R96%K.PHA(`HA\nM(`D)8V%S92!47U%U97)Y.@HA(`D)\"7L*(2`)\"0D)475E<GD)*G%R>2`](\"A1\nM=65R>2`J*6YO9&4[\"B$@\"B$@\"0D)\"6%P<&QY7U))4E]V:65W*`HA(`D)\"0D)\nM\"2A.;V1E(\"HJ*2@F*'%R>2T^=&%R9V5T3&ES=\"DI+`HA(`D)\"0D)\"7)T7VEN\nM9&5X+`HA(`D)\"0D)\"7)T92P*(2`)\"0D)\"0ET;&ES=\"P*(2`)\"0D)\"0EM;V1I\nM9FEE9\"P*(2`)\"0D)\"0ES=6)L979E;'-?=7`I.PHA(`HA(`D)\"0EA<'!L>5]2\nM25)?=FEE=R@*(2`)\"0D)\"0DH3F]D92`J*BDH)BAQ<GDM/G%U86PI*2P*(2`)\nM\"0D)\"0ER=%]I;F1E>\"P*(2`)\"0D)\"0ER=&4L\"B$@\"0D)\"0D)=&QI<W0L\"B$@\nM\"0D)\"0D);6]D:69I960L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`*(2`)\nM\"0D)87!P;'E?4DE27W9I97<H\"B$@\"0D)\"0D)*$YO9&4@*BHI*\"8H<7)Y+3YH\nM879I;F=1=6%L*2DL\"B$@\"0D)\"0D)<G1?:6YD97@L\"B$@\"0D)\"0D)<G1E+`HA\nM(`D)\"0D)\"71L:7-T+`HA(`D)\"0D)\"6UO9&EF:65D+`HA(`D)\"0D)\"7-U8FQE\nM=F5L<U]U<\"D[\"B$@\"B$@\"0D)\"6%P<&QY7U))4E]V:65W*`HA(`D)\"0D)\"2A.\nM;V1E(\"HJ*2@F*'%R>2T^9W)O=7!#;&%U<V4I*2P*(2`)\"0D)\"0ER=%]I;F1E\nM>\"P*(2`)\"0D)\"0ER=&4L\"B$@\"0D)\"0D)=&QI<W0L\"B$@\"0D)\"0D);6]D:69I\nM960L\"B$@\"0D)\"0D)<W5B;&5V96QS7W5P*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\nM\"B$@\"B$@\"0ED969A=6QT.@HA(`D)\"65L;V<H3D]424-%+\"`B=6YK;F]W;B!N\nM;V1E('1A9R`E9\"!I;B!A<'!L>5]225)?=FEE=R@I(BP@;F]D951A9RAN;V1E\nM*2D[\"B$@\"0D)96QO9RA.3U1)0T4L(\").;V1E(&ES.B`E<R(L(&YO9&54;U-T\nM<FEN9RAN;V1E*2D[\"B$@\"0D)8G)E86L[\"B$@\"B$@\"B$@\"7T*(2!]\"B$@\"B$@\nM\"B$@<W1A=&EC('9O:60*(2!!<'!L>5)E=')I979E4G5L92A1=65R>2`J<&%R\nM<V5T<F5E+`HA(`D)\"0D@(%)E=W)I=&52=6QE(\"IR=6QE+`HA(`D)\"0D@(&EN\nM=\"!R=%]I;F1E>\"P*(2`)\"0D)(\"!I;G0@<F5L871I;VY?;&5V96PL\"B$@\"0D)\nM\"2`@4F5L871I;VX@<F5L871I;VXL\"B$@\"0D)\"2`@:6YT(\"IM;V1I9FEE9\"D*\nM(2![\"B$@\"5%U97)Y\"2`@(\"IR=6QE7V%C=&EO;B`]($Y53$P[\"B$@\"4YO9&4)\nM(\"`@*G)U;&5?<75A;#L*(2`)3&ES=`D@(\"`J<G1A8FQE+`HA(`D)\"2`@(\"IR\nM=#L*(2`):6YT\"0D);F]T:&EN9RP*(2`)\"0D)<G1?;&5N9W1H.PHA(`EI;G0)\nM\"0EB861S<6P@/2!&04Q313L*(2`*(2`)<G5L95]Q=6%L(#T@<G5L92T^<75A\nM;#L*(2`):68@*')U;&4M/F%C=&EO;G,I\"B$@\"7L*(2`)\"6EF(\"AL96YG=&@H\nM<G5L92T^86-T:6]N<RD@/B`Q*0DO*B`_/S\\@8F5C875S92!W92!D;VXG=\"!H\nM86YD;&4*(2`)\"0D)\"0D)\"0D)(\"H@<G5L97,@=VET:\"!M;W)E('1H86X@;VYE\nM\"B$@\"0D)\"0D)\"0D)\"2`J(&%C=&EO;C\\@+6%Y(\"HO\"B$@\"B$@\"0D)<F5T=7)N\nM.PHA(`D)<G5L95]A8W1I;VX@/2!C;W!Y3V)J96-T*&QF:7)S=\"AR=6QE+3YA\nM8W1I;VYS*2D[\"B$@\"0EN;W1H:6YG(#T@1D%,4T4[\"B$@\"7T*(2`)96QS90HA\nM(`D);F]T:&EN9R`](%12544[\"B$@\"B$@\"7)T86)L92`](&-O<'E/8FIE8W0H\nM<&%R<V5T<F5E+3YR=&%B;&4I.PHA(`EF;W)E86-H*')T+\"!R=&%B;&4I\"B$@\nM\"7L*(2`)\"5)A;F=E5&)L16YT<GD@*G)T92`](&QF:7)S=\"AR=\"D[\"B$@\"B$@\nM\"0DO*@HA(`D)(\"H@=&AI<R!I<R!T;R!P<F5V96YT(&%D9%]M:7-S:6YG7W9A\nM<G-?=&]?8F%S95]R96QS*\"D@9G)O;2!A9&1I;F<*(2`)\"2`J(&$@8F]G=7,@\nM96YT<GD@=&\\@=&AE(&YE=R!T87)G970@;&ES=\"X*(2`)\"2`J+PHA(`D)<G1E\nM+3YI;D9R;VU#;\"`](&9A;'-E.PHA(`E]\"B$@\"7)T7VQE;F=T:\"`](&QE;F=T\nM:\"AR=&%B;&4I.PHA(`HA(`ER=&%B;&4@/2!N8V]N8RAR=&%B;&4L(&-O<'E/\nM8FIE8W0H<G5L95]A8W1I;VXM/G)T86)L92DI.PHA(`EP87)S971R964M/G)T\nM86)L92`](')T86)L93L*(2`*(2`)<G5L95]A8W1I;VXM/G)T86)L92`](')T\nM86)L93L*(2`);V9F<V5T7W9A<FYO9&5S*\"A.;V1E(\"HI(')U;&5?<75A;\"P@\nM(\"!R=%]L96YG=&@L(#`I.PHA(`EO9F9S971?=F%R;F]D97,H*$YO9&4@*BD@\nM<G5L95]A8W1I;VXL(')T7VQE;F=T:\"P@,\"D[\"B$@\"B$@\"6-H86YG95]V87)N\nM;V1E<R@H3F]D92`J*2!R=6QE7W%U86PL(`HA(`D)\"0D@(\"!04E,R7T-54E)%\nM3E1?5D%23D\\@*R!R=%]L96YG=&@L(')T7VEN9&5X+\"`P*3L*(2`)8VAA;F=E\nM7W9A<FYO9&5S*\"A.;V1E(\"HI(')U;&5?86-T:6]N+`HA(`D)\"0D@(\"!04E,R\nM7T-54E)%3E1?5D%23D\\@*R!R=%]L96YG=&@L(')T7VEN9&5X+\"`P*3L*(2`*\nM(2`):68@*')E;&%T:6]N7VQE=F5L*0HA(`E[\"B$@\"2`@87!P;'E?4DE27W9I\nM97<H*$YO9&4@*BHI(\"9P87)S971R964L(')T7VEN9&5X+\"`*(2`)(\"`)\"2A2\nM86YG951B;$5N=')Y(\"HI;G1H*')T7VEN9&5X(\"T@,2P@<G1A8FQE*2P*(2`)\nM\"0ER=6QE7V%C=&EO;BT^=&%R9V5T3&ES=\"P@;6]D:69I960L(#`I.PHA(`D@\nM(&%P<&QY7U))4E]V:65W*\"A.;V1E(\"HJ*2`F<G5L95]A8W1I;VXL(')T7VEN\nM9&5X+\"`*(2`)(\"`)\"2A286YG951B;$5N=')Y(\"HI;G1H*')T7VEN9&5X(\"T@\nM,2P@<G1A8FQE*2P*(2`)\"0ER=6QE7V%C=&EO;BT^=&%R9V5T3&ES=\"P@;6]D\nM:69I960L(#`I.PHA(`E]\"B$@\"65L<V4*(2`)>PHA(`D@($AA;F1L95))4D%T\nM=')I8G5T95)U;&4H<&%R<V5T<F5E+\"!R=&%B;&4L(')U;&5?86-T:6]N+3YT\nM87)G971,:7-T+`HA(`D)\"0D@<G1?:6YD97@L(')U;&4M/F%T=')N;RP@;6]D\nM:69I960L(\"9B861S<6PI.PHA(`E]\"B$@\"6EF(\"@J;6]D:69I960@)B8@(6)A\nM9'-Q;\"D@>PHA(`D@($%D9%%U86PH<&%R<V5T<F5E+\"!R=6QE7V%C=&EO;BT^\nM<75A;\"D[\"B$@\"2`@+RH@5&AI<R!W:6QL(&]N;'D@=V]R:R!I9B!T:&4@<75E\nM<GD@;6%D92!T;R!T:&4@=FEE=R!D969I;F5D(&)Y('1H92!F;VQL;W=I;F<*\nM(2`)(\"`@*B!G<F]U<$-L875S92!G<F]U<',@8GD@=&AE('-A;64@871T<FEB\nM=71E<R!O<B!D;V5S(&YO=\"!U<V4@9W)O=7`@870@86QL(2`J+PHA(`D@(&EF\nM(\"AP87)S971R964M/F=R;W5P0VQA=7-E(#T]($Y53$PI\"B$@\"2`@(\"!P87)S\nM971R964M/F=R;W5P0VQA=7-E/7)U;&5?86-T:6]N+3YG<F]U<$-L875S93L*\nM(2`)(\"!!9&1(879I;F=1=6%L*'!A<G-E=')E92P@<G5L95]A8W1I;VXM/FAA\nM=FEN9U%U86PI.PHA(`D@('!A<G-E=')E92T^:&%S06=G<R`](\"AR=6QE7V%C\nM=&EO;BT^:&%S06=G<R!\\?\"!P87)S971R964M/FAA<T%G9W,I.PHA(`D@('!A\nM<G-E=')E92T^:&%S4W5B3&EN:W,@/2`H<G5L95]A8W1I;VXM/FAA<U-U8DQI\nM;FMS('Q\\(\"!P87)S971R964M/FAA<U-U8DQI;FMS*3L*(2`)?0D*(2!]\"B$@\nM\"B$@\"B$@<W1A=&EC('9O:60*(2!F:7)E4DE2;VY3=6)S96QE8W0H3F]D92`J\nM;F]D92D*(2![\"B$@\"6EF(\"AN;V1E(#T]($Y53$PI\"B$@\"0ER971U<FX[\"B$@\nM\"B$@\"7-W:71C:\"AN;V1E5&%G*&YO9&4I*2![\"B$@\"0EC87-E(%1?5&%R9V5T\nM16YT<GDZ\"B$@\"0D)>PHA(`D)\"0E487)G971%;G1R>0DJ=&QE(#T@*%1A<F=E\nM=$5N=')Y(\"HI;F]D93L*(2`*(2`)\"0D)9FER95))4F]N4W5B<V5L96-T*`HA\nM(`D)\"0D)\"2A.;V1E(\"HI*'1L92T^97AP<BDI.PHA(`D)\"7T*(2`)\"0EB<F5A\nM:SL*(2`*(2`)\"6-A<V4@5%]!9V=R96<Z\"B$@\"0D)>PHA(`D)\"0E!9V=R96<)\nM*F%G9R`](\"A!9V=R96<@*BEN;V1E.PHA(`HA(`D)\"0EF:7)E4DE2;VY3=6)S\nM96QE8W0H\"B$@\"0D)\"0D)*$YO9&4@*BDH86=G+3YT87)G970I*3L*(2`)\"0E]\nM\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC87-E(%1?1W)O=7!#;&%U<V4Z\"B$@\"0D)\nM>PHA(`D)\"0E'<F]U<$-L875S90DJ9W)P(#T@*$=R;W5P0VQA=7-E(\"HI;F]D\nM93L*(2`*(2`)\"0D)9FER95))4F]N4W5B<V5L96-T*`HA(`D)\"0D)\"2A.;V1E\nM(\"HI*&=R<\"T^96YT<GDI*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC\nM87-E(%1?17AP<CH*(2`)\"0E[\"B$@\"0D)\"45X<'()*F5X<\"`](\"A%>'!R(\"HI\nM;F]D93L*(2`*(2`)\"0D)9FER95))4F]N4W5B<V5L96-T*`HA(`D)\"0D)\"2A.\nM;V1E(\"HI*&5X<\"T^87)G<RDI.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\nM\"6-A<V4@5%])=&5R.@HA(`D)\"7L*(2`)\"0D)271E<@DJ:71E<B`](\"A)=&5R\nM(\"HI;F]D93L*(2`*(2`)\"0D)9FER95))4F]N4W5B<V5L96-T*`HA(`D)\"0D)\nM\"2A.;V1E(\"HI*&ET97(M/FET97)E>'!R*2D[\"B$@\"0D)?0HA(`D)\"6)R96%K\nM.PHA(`HA(`D)8V%S92!47T%R<F%Y4F5F.@HA(`D)\"7L*(2`)\"0D)07)R87E2\nM968)*G)E9B`](\"A!<G)A>5)E9B`J*6YO9&4[\"B$@\"B$@\"0D)\"69I<F5225)O\nM;E-U8G-E;&5C=\"@*(2`)\"0D)\"0DH3F]D92`J*2AR968M/G)E9G5P<&5R:6YD\nM97AP<BDI.PHA(`D)\"0EF:7)E4DE2;VY3=6)S96QE8W0H\"B$@\"0D)\"0D)*$YO\nM9&4@*BDH<F5F+3YR969L;W=E<FEN9&5X<'(I*3L*(2`)\"0D)9FER95))4F]N\nM4W5B<V5L96-T*`HA(`D)\"0D)\"2A.;V1E(\"HI*')E9BT^<F5F97AP<BDI.PHA\nM(`D)\"0EF:7)E4DE2;VY3=6)S96QE8W0H\"B$@\"0D)\"0D)*$YO9&4@*BDH<F5F\nM+3YR969A<W-G;F5X<'(I*3L*(2`)\"0E]\"B$@\"0D)8G)E86L[\"B$@\"B$@\"0EC\nM87-E(%1?5F%R.@HA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47U!A<F%M.@HA\nM(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!47T-O;G-T.@HA(`D)\"6)R96%K.PHA\nM(`HA(`D)8V%S92!47TQI<W0Z\"B$@\"0D)>PHA(`D)\"0E,:7-T\"2IL.PHA(`HA\nM(`D)\"0EF;W)E86-H(\"AL+\"`H3&ES=\"`J*6YO9&4I\"B$@\"0D)\"0EF:7)E4DE2\nM;VY3=6)S96QE8W0H\"B$@\"0D)\"0D)\"2A.;V1E(\"HI*&QF:7)S=\"AL*2DI.PHA\nM(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]3=6),:6YK.@HA(`D)\nM\"7L*(2`)\"0D)4W5B3&EN:PDJ<W5B(#T@*%-U8DQI;FL@*BEN;V1E.PHA(`D)\nM\"0E1=65R>0DJ<7)Y.PHA(`HA(`D)\"0EF:7)E4DE2;VY3=6)S96QE8W0H\"B$@\nM\"0D)\"0D)*$YO9&4@*BDH<W5B+3YL969T:&%N9\"DI.PHA(`HA(`D)\"0EQ<GD@\nM/2!F:7)E4DE2<G5L97,H*%%U97)Y(\"HI*'-U8BT^<W5B<V5L96-T*2D[\"B$@\nM\"B$@\"0D)\"69I<F5225)O;E-U8G-E;&5C=\"@*(2`)\"0D)\"0DH3F]D92`J*7%R\nM>2D[\"B$@\"B$@\"0D)\"7-U8BT^<W5B<V5L96-T(#T@*$YO9&4@*BD@<7)Y.PHA\nM(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\"6-A<V4@5%]1=65R>3H*(2`)\"0E[\nM\"B$@\"0D)\"5%U97)Y\"2IQ<GD@/2`H475E<GD@*BEN;V1E.PHA(`HA(`D)\"0EF\nM:7)E4DE2;VY3=6)S96QE8W0H\"B$@\"0D)\"0D)*$YO9&4@*BDH<7)Y+3YT87)G\nM971,:7-T*2D[\"B$@\"B$@\"0D)\"69I<F5225)O;E-U8G-E;&5C=\"@*(2`)\"0D)\nM\"0DH3F]D92`J*2AQ<GDM/G%U86PI*3L*(2`*(2`)\"0D)9FER95))4F]N4W5B\nM<V5L96-T*`HA(`D)\"0D)\"2A.;V1E(\"HI*'%R>2T^:&%V:6YG475A;\"DI.PHA\nM(`HA(`D)\"0EF:7)E4DE2;VY3=6)S96QE8W0H\"B$@\"0D)\"0D)*$YO9&4@*BDH\nM<7)Y+3YG<F]U<$-L875S92DI.PHA(`D)\"7T*(2`)\"0EB<F5A:SL*(2`*(2`)\nM\"61E9F%U;'0Z\"B$@\"0D)96QO9RA.3U1)0T4L(\")U;FMN;W=N(&YO9&4@=&%G\nM(\"5D(&EN(&9I<F5225)O;E-U8G-E;&5C=\"@I(BP@;F]D951A9RAN;V1E*2D[\nM\"B$@\"0D)96QO9RA.3U1)0T4L(\").;V1E(&ES.B`E<R(L(&YO9&54;U-T<FEN\nM9RAN;V1E*2D[\"B$@\"0D)8G)E86L[\"B$@\"B$@\"B$@\"7T*(2!]\"B$@\"B$@\"B$@\nM+RH*(2`@*B!F:7)E4DE2<G5L97,@+0HA(\"`J\"4%P<&QY(&%L;\"!225(@<G5L\nM97,@;VX@96%C:\"!R86YG971A8FQE(&5N=')Y(&EN(&$@<75E<GD*(2`@*B\\*\nM(2!S=&%T:6,@475E<GD@*@HA(&9I<F5225)R=6QE<RA1=65R>2`J<&%R<V5T\nM<F5E*0HA('L*(2`):6YT\"0ER=%]I;F1E>#L*(2`)4F%N9V548FQ%;G1R>0DJ\nM<G1E.PHA(`E296QA=&EO;@ER96P[\"B$@\"4QI<W0)\"2IL;V-K<SL*(2`)4G5L\nM94QO8VL)*G)U;&5S.PHA(`E297=R:71E4G5L90DJ<G5L93L*(2`)4F5W<FET\nM95)U;&4)4DE2;VYL>3L*(2`):6YT\"0EM;V1I9FEE9#L*(2`):6YT\"0EI.PHA\nM(`E,:7-T\"0DJ;#L*(2`*(2`)<G1?:6YD97@@/2`P.PHA(`EW:&EL92AR=%]I\nM;F1E>\"`\\(&QE;F=T:\"AP87)S971R964M/G)T86)L92DI('L*(2`)\"2LK<G1?\nM:6YD97@[\"B$@\"B$@\"0EI9B`H(7)A;F=E5&%B;&5%;G1R>5]U<V5D*\"A.;V1E\nM(\"HI<&%R<V5T<F5E+\"!R=%]I;F1E>\"P@,\"DI\"B$@\"0D)8V]N=&EN=64[\"B$@\nM\"0D*(2`)\"7)T92`](&YT:\"AR=%]I;F1E>\"`M(#$L('!A<G-E=')E92T^<G1A\nM8FQE*3L*(2`)\"7)E;\"`](&AE87!?;W!E;G(H<G1E+3YR96QN86UE*3L*(2`)\nM\"6EF(\"AR96PM/G)D7W)U;&5S(#T]($Y53$PI('L*(2`)\"0EH96%P7V-L;W-E\nM*')E;\"D[\"B$@\"0D)8V]N=&EN=64[\"B$@\"0E]\"B$@\"B$@\"0ER=6QE<R`](')E\nM;\"T^<F1?<G5L97,[\"B$@\"0EL;V-K<R`]($Y)3#L*(2`*(2`)\"2\\J\"B$@\"0D@\nM*B!#;VQL96-T('1H92!225(@<G5L97,@=&AA=\"!W92!M=7-T(&%P<&QY\"B$@\nM\"0D@*B\\*(2`)\"69O<B`H:2`](#`[(&D@/\"!R=6QE<RT^;G5M3&]C:W,[(&DK\nM*RD@>PHA(`D)\"7)U;&4@/2!R=6QE<RT^<G5L97-;:5T[\"B$@\"0D):68@*')U\nM;&4M/F5V96YT(\"$]($--1%]314Q%0U0I\"B$@\"0D)\"6-O;G1I;G5E.PHA(`D)\nM\"0HA(`D)\"6EF(\"AR=6QE+3YA='1R;F\\@/B`P(\"8F\"B$@\"0D)\"0DA871T<FEB\nM=71E7W5S960H*$YO9&4@*BEP87)S971R964L\"B$@\"0D)\"0D)\"7)T7VEN9&5X\nM+`HA(`D)\"0D)\"0ER=6QE+3YA='1R;F\\L(#`I*0HA(`D)\"0EC;VYT:6YU93L*\nM(2`*(2`)\"0EL;V-K<R`](&QA<'!E;F0H;&]C:W,L(')U;&4I.PHA(`D)?0HA\nM(`HA(`D)+RH*(2`)\"2`J($-H96-K('!E<FUI<W-I;VYS\"B$@\"0D@*B\\*(2`)\nM\"6-H96-K3&]C:U!E<FUS*&QO8VMS+\"!P87)S971R964L(')T7VEN9&5X*3L*\nM(2`*(2`)\"2\\J\"B$@\"0D@*B!.;W<@87!P;'D@=&AE;0HA(`D)(\"HO\"B$@\"0EF\nM;W)E86-H(\"AL+\"!L;V-K<RD@>PHA(`D)\"7)U;&4@/2!L9FER<W0H;\"D[\"B$@\nM\"B$@\"0D)4DE2;VYL>2YE=F5N=`D](')U;&4M/F5V96YT.PHA(`D)\"5))4F]N\nM;'DN871T<FYO\"3T@<G5L92T^871T<FYO.PHA(`D)\"5))4F]N;'DN<75A;`D]\nM(')U;&4M/G%U86P[\"B$@\"0D)4DE2;VYL>2YA8W1I;VYS\"3T@<G5L92T^86-T\nM:6]N<SL*(2`*(2`)\"0E!<'!L>5)E=')I979E4G5L92AP87)S971R964L\"B$@\nM\"0D)\"0DF4DE2;VYL>2P*(2`)\"0D)\"7)T7VEN9&5X+`HA(`D)\"0D)4DE2;VYL\nM>2YA='1R;F\\@/3T@+3$L\"B$@\"0D)\"0ER96PL\"B$@\"0D)\"0DF;6]D:69I960I\nM.PH@(`D)?0HA(`HA(`D):&5A<%]C;&]S92AR96PI.PHA(`E]\"B$@\"B$@\"69I\nM<F5225)O;E-U8G-E;&5C=\"@H3F]D92`J*2!P87)S971R964I.PHA(`EM;V1I\nM9GE!9V=R96=1=6%L*\"A.;V1E(\"HJ*2`F*'!A<G-E=')E92T^<75A;\"DL('!A\nM<G-E=')E92D[\"B$@\"B$@\"7)E='5R;B!P87)S971R964[\"B$@?0HA(`HA(`HA\nM(\"\\J\"B$@(\"H@:61E82!I<R!T;R!F:7)E(')E9W5L87(@<G5L97,@9FER<W0L\nM('1H96X@<75A;&EF:65D(&EN<W1E860*(2`@*B!R=6QE<R!A;F0@=6YQ=6%L\nM:69I960@:6YS=&5A9\"!R=6QE<R!L87-T+B!!;GD@;&5M;6EN9R!I<R!C;W5N\nM=&5D(&9O<BX*(2`@*B\\*(2!S=&%T:6,@3&ES=\"`J\"B$@;W)D97)2=6QE<RA,\nM:7-T(\"IL;V-K<RD*(2![\"B$@\"4QI<W0)(\"`@*G)E9W5L87(@/2!.24P[\"B$@\nM\"4QI<W0)(\"`@*FEN<W1E861?<G5L97,@/2!.24P[\"B$@\"4QI<W0)(\"`@*FEN\nM<W1E861?<75A;&EF:65D(#T@3DE,.PHA(`E,:7-T\"2`@(\"II.PHA(`HA(`EF\nM;W)E86-H*&DL(&QO8VMS*0HA(`E[\"B$@\"0E297=R:71E4G5L92`J<G5L95]L\nM;V-K(#T@*%)E=W)I=&52=6QE(\"HI(&QF:7)S=\"AI*3L*(2`*(2`)\"6EF(\"AR\nM=6QE7VQO8VLM/FES26YS=&5A9\"D*(\"`)\"7L*(2`)\"0EI9B`H<G5L95]L;V-K\nM+3YQ=6%L(#T]($Y53$PI\"B$@\"0D)\"6EN<W1E861?<G5L97,@/2!L87!P96YD\nM*&EN<W1E861?<G5L97,L(')U;&5?;&]C:RD[\"B$@\"0D)96QS90HA(`D)\"0EI\nM;G-T96%D7W%U86QI9FEE9\"`](&QA<'!E;F0H:6YS=&5A9%]Q=6%L:69I960L\nM(')U;&5?;&]C:RD[\"B`@\"0E]\"BL@\"0EE;'-E\"BL@\"0D)<F5G=6QA<B`](&QA\nM<'!E;F0H<F5G=6QA<BP@<G5L95]L;V-K*3L*(\"`)?0HA(`ER96=U;&%R(#T@\nM;F-O;F,H<F5G=6QA<BP@:6YS=&5A9%]Q=6%L:69I960I.PHA(`ER971U<FX@\nM;F-O;F,H<F5G=6QA<BP@:6YS=&5A9%]R=6QE<RD[\"B`@?0H@(`HK(`HK(`H@\nM('-T871I8R!1=65R>2`J\"B`@0V]P>4%N9$%D9%%U86PH475E<GD@*G!A<G-E\nM=')E92P*(\"`)\"0D@(\"!,:7-T(\"IA8W1I;VYS+`HJ*BHJ*BHJ*BHJ*BHJ*BH*\nM*BHJ(#0V,BPT-C<@*BHJ*@HM+2T@,C(U-\"PR,C8P(\"TM+2T*(\"!]\"B`@\"B`@\nM\"BL@\"B`@+RH*(\"`@*@EF:7)E4G5L97,@+0H@(\"`J\"2`@($ET97)A=&4@=&AR\nM;W5G:\"!R=6QE(&QO8VMS(&%P<&QY:6YG(')U;&5S+@HJ*BHJ*BHJ*BHJ*BHJ\nM*BH**BHJ(#0X.\"PU,#`@*BHJ*@H@(`DO*B!C:&]O<V4@<G5L92!T;R!F:7)E\nM(&9R;VT@;&ES=\"!O9B!R=6QE<R`J+PH@(`EI9B`H;&]C:W,@/3T@3DE,*0H@\nM(`E[\"B$@\"0E0<F]C97-S4F5T<FEE=F51=65R>2AP87)S971R964L\"B$@\"0D)\nM\"0D)\"2!P87)S971R964M/G)T86)L92P*(2`)\"0D)\"0D)(&EN<W1E861?9FQA\nM9RP@5%)512D[\"B$@\"0EI9B`H*FEN<W1E861?9FQA9RD*(2`)\"0ER971U<FX@\nM;&%P<&5N9\"A.24PL('!A<G-E=')E92D[\"B$@\"0EE;'-E\"B$@\"0D)<F5T=7)N\nM($Y)3#L*(\"`)?0H@(`H@(`EL;V-K<R`](&]R9&5R4G5L97,H;&]C:W,I.PDO\nM*B!R96%L(&EN<W1E860@<G5L97,@;&%S=\"`J+PHM+2T@,C(X,2PR,C@W(\"TM\nM+2T*(\"`)+RH@8VAO;W-E(')U;&4@=&\\@9FER92!F<F]M(&QI<W0@;V8@<G5L\nM97,@*B\\*(\"`):68@*&QO8VMS(#T]($Y)3\"D*(\"`)>PHA(`D)<F5T=7)N($Y)\nM3#L*(\"`)?0H@(`H@(`EL;V-K<R`](&]R9&5R4G5L97,H;&]C:W,I.PDO*B!R\nM96%L(&EN<W1E860@<G5L97,@;&%S=\"`J+PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ\nM(#4P-2PU,3$@*BHJ*@H@(`D)\"0D@(\"`J979E;G1?<75A;#L*(\"`)\"4QI<W0)\nM(\"`@*F%C=&EO;G,[\"B`@\"0E,:7-T\"2`@(\"IR.PHM(`D)8F]O;`D);W)I9U]I\nM;G-T96%D7V9L86<@/2`J:6YS=&5A9%]F;&%G.PH@(`H@(`D)+RH*(\"`)\"2`J\nM($EN<W1E860@<G5L97,@8VAA;F=E('1H92!R97-U;'1296QA=&EO;B!O9B!T\nM:&4@<75E<GDN(%-O('1H90HM+2T@,C(Y,BPR,CDW(\"TM+2T**BHJ*BHJ*BHJ\nM*BHJ*BHJ\"BHJ*B`V-#4L-C4R(\"HJ*BH*+2TM(#(T,S$L,C0T,\"`M+2TM\"B`@\nM\"0D)(\"HM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM\nM+2TM+2TM+2TM+0H@(`D)\"2`J+PH@(`D)\"6EN9F\\M/G)U;&5?86-T:6]N+3YR\nM=&%B;&4@/2!I;F9O+3YR=#L**R`)\"0DO*@H@(`D)\"5!R;V-E<W-2971R:65V\nM95%U97)Y*&EN9F\\M/G)U;&5?86-T:6]N+\"!I;F9O+3YR=\"P*(\"`)\"0D)\"0D)\nM\"2`F;W)I9U]I;G-T96%D7V9L86<L(%12544I.PHK(`D)\"2HO\"B`@\"B`@\"0D)\nM+RHM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM+2TM\nM+2TM+2TM+0H@(`D)\"2`J(%-T97`@-`HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#8W\nM,\"PW.3<@*BHJ*@H@(`ER971U<FX@<F5S=6QT<SL*(\"!]\"B`@\"BT@+RH@+2TM\nM+2TM+2TM+0HM(\"`J(%)E=W)I=&50<F5P<F]C97-S475E<GD@+0HM(\"`J\"6%D\nM:G5S=\"!D971A:6QS(&EN('1H92!P87)S971R964L('1H92!R=6QE('-Y<W1E\nM;0HM(\"`J\"61E<&5N9',@;VX*+2`@*B`M+2TM+2TM+2TM\"BT@(\"HO\"BT@<W1A\nM=&EC('9O:60*+2!297=R:71E4')E<')O8V5S<U%U97)Y*%%U97)Y(\"IP87)S\nM971R964I\"BT@>PHM(`DO*B`M+2TM+2TM+2TM\"BT@\"2`J(&EF('1H92!Q=65R\nM>2!H87,@82!R97-U;'1296QA=&EO;BP@<F5A<W-I9VX@=&AE\"BT@\"2`J(')E\nM<W5L=\"!D;VUA:6X@;G5M8F5R<R!T;R!T:&4@871T<FEB=71E(&YU;6)E<G,@\nM:6X@=&AE\"BT@\"2`J('1A<F=E=\"!R96QA=&EO;BX@1FEX3F5W*\"D@9&5P96YD\nM<R!O;B!I=\"!W:&5N(')E<&QA8VEN9PHM(`D@*B`J;F5W*B!R969E<F5N8V5S\nM(&EN(&$@<G5L92!A8W1I;VX@8GD@=&AE(&5X<')E<W-I;VYS\"BT@\"2`J(&9R\nM;VT@=&AE(')E=W)I='1E;B!Q=65R>2X*+2`)(\"H@+2TM+2TM+2TM+0HM(`D@\nM*B\\*+2`):68@*'!A<G-E=')E92T^<F5S=6QT4F5L871I;VX@/B`P*0HM(`E[\nM\"BT@\"0E286YG951B;$5N=')Y(\"IR=&4[\"BT@\"0E296QA=&EO;@ER9#L*+2`)\nM\"4QI<W0)(\"`@*G1L.PHM(`D)5&%R9V5T16YT<GD@*G1L93L*+2`)\"6EN=`D)\nM\"7)E<V1O;6YO.PHM(`HM(`D)<G1E(#T@*%)A;F=E5&)L16YT<GD@*BD@;G1H\nM*'!A<G-E=')E92T^<F5S=6QT4F5L871I;VX@+2`Q+`HM(`D)\"0D)\"0D)\"7!A\nM<G-E=')E92T^<G1A8FQE*3L*+2`)\"7)D(#T@:&5A<%]O<&5N<BAR=&4M/G)E\nM;&YA;64I.PHM(`HM(`D)9F]R96%C:\"AT;\"P@<&%R<V5T<F5E+3YT87)G971,\nM:7-T*0HM(`D)>PHM(`D)\"71L92`](\"A487)G971%;G1R>2`J*2!L9FER<W0H\nM=&PI.PHM(`D)\"7)E<V1O;6YO(#T@871T;F%M94%T=$YU;2AR9\"P@=&QE+3YR\nM97-D;VTM/G)E<VYA;64I.PHM(`D)\"71L92T^<F5S9&]M+3YR97-N;R`](')E\nM<V1O;6YO.PHM(`D)?0HM(`HM(`D):&5A<%]C;&]S92AR9\"D[\"BT@\"7T*+2!]\nM\"BT@\"BT@\"BT@+RH@+2TM+2TM+2TM+0HM(\"`J(%)E=W)I=&50;W-T<')O8V5S\nM<TYO;E-E;&5C=\"`M\"BT@(\"H)87!P;'D@:6YS=&5A9\"!S96QE8W0@<G5L97,@\nM;VX@82!Q=65R>2!F:7)E9\"!I;B!B>0HM(\"`J\"71H92!R97=R:71E('-Y<W1E\nM;0HM(\"`J(\"TM+2TM+2TM+2T*+2`@*B\\*+2!S=&%T:6,@475E<GD@*@HM(%)E\nM=W)I=&50;W-T<')O8V5S<TYO;E-E;&5C=\"A1=65R>2`J<&%R<V5T<F5E*0HM\nM('L*+2`)3&ES=`D@(\"`J<G0[\"BT@\"6EN=`D)\"7)T7VEN9&5X(#T@,#L*+2`)\nM475E<GD)(\"`@*FYE=W1R964@/2!C;W!Y3V)J96-T*'!A<G-E=')E92D[\"BT@\nM\"BT@\"69O<F5A8V@H<G0L('!A<G-E=')E92T^<G1A8FQE*0HM(`E[\"BT@\"0E2\nM86YG951B;$5N=')Y(\"IR=%]E;G1R>2`](&QF:7)S=\"AR=\"D[\"BT@\"0E296QA\nM=&EO;@ER=%]E;G1R>5]R96QA=&EO;B`]($Y53$P[\"BT@\"0E2=6QE3&]C:R`@\nM(\"IR=%]E;G1R>5]L;V-K<R`]($Y53$P[\"BT@\"0E,:7-T\"2`@(\"IL;V-K<R`]\nM($Y)3#L*+2`)\"4QI<W0)(\"`@*FEN<W1E861?;&]C:W,@/2!.24P[\"BT@\"0E,\nM:7-T\"2`@(\"IL;V-K.PHM(`D)4F5W<FET95)U;&4@*G)U;&4[\"BT@\"BT@\"0ER\nM=%]I;F1E>\"LK.PHM(`D)<G1?96YT<GE?<F5L871I;VX@/2!H96%P7V]P96YR\nM*')T7V5N=')Y+3YR96QN86UE*3L*+2`)\"7)T7V5N=')Y7VQO8VMS(#T@<G1?\nM96YT<GE?<F5L871I;VXM/G)D7W)U;&5S.PHM(`HM(`D):68@*')T7V5N=')Y\nM7VQO8VMS*0HM(`D)>PHM(`D)\"6EN=`D)\"6]R:6=C;61T>7!E(#T@;F5W=')E\nM92T^8V]M;6%N9%1Y<&4[\"BT@\"BT@\"0D);F5W=')E92T^8V]M;6%N9%1Y<&4@\nM/2!#341?4T5,14-4.PHM(`D)\"6QO8VMS(#T*+2`)\"0D);6%T8VA,;V-K<RA#\nM341?4T5,14-4+\"!R=%]E;G1R>5]L;V-K<RP@<G1?:6YD97@L(&YE=W1R964I\nM.PHM(`D)\"6YE=W1R964M/F-O;6UA;F14>7!E(#T@;W)I9V-M9'1Y<&4[\"BT@\nM\"0E]\"BT@\"0EI9B`H;&]C:W,@(3T@3DE,*0HM(`D)>PHM(`D)\"69O<F5A8V@H\nM;&]C:RP@;&]C:W,I\"BT@\"0D)>PHM(`D)\"0ER=6QE(#T@*%)E=W)I=&52=6QE\nM(\"HI(&QF:7)S=\"AL;V-K*3L*+2`)\"0D):68@*')U;&4M/FES26YS=&5A9\"D*\nM+2`)\"0D)\"6EN<W1E861?;&]C:W,@/2!N8V]N8RAI;G-T96%D7VQO8VMS+\"!L\nM;V-K*3L*+2`)\"0E]\"BT@\"0E]\"BT@\"0EI9B`H:6YS=&5A9%]L;V-K<R`A/2!.\nM24PI\"BT@\"0E[\"BT@\"0D)9F]R96%C:\"AL;V-K+\"!I;G-T96%D7VQO8VMS*0HM\nM(`D)\"7L*+2`)\"0D):6YT\"0D)<F5L871I;VY?;&5V96P[\"BT@\"0D)\"6EN=`D)\nM\"6UO9&EF:65D(#T@,#L*+2`*+2`)\"0D)<G5L92`](\"A297=R:71E4G5L92`J\nM*2!L9FER<W0H;&]C:RD[\"BT@\"0D)\"7)E;&%T:6]N7VQE=F5L(#T@*')U;&4M\nM/F%T=')N;R`]/2`M,2D[\"B`@\"BT@\"0D)\"4%P<&QY4F5T<FEE=F52=6QE*&YE\nM=W1R964L\"BT@\"0D)\"0D)\"0D@(')U;&4L\"BT@\"0D)\"0D)\"0D@(')T7VEN9&5X\nM+`HM(`D)\"0D)\"0D)(\"!R96QA=&EO;E]L979E;\"P*+2`)\"0D)\"0D)\"2`@<G1?\nM96YT<GE?<F5L871I;VXL\"BT@\"0D)\"0D)\"0D@(\"9M;V1I9FEE9\"D[\"BT@\"0D)\nM?0HM(`D)?0HM(`HM(`D):&5A<%]C;&]S92AR=%]E;G1R>5]R96QA=&EO;BD[\nM\"BT@\"7T*+2`*+2`)<F5T=7)N(&YE=W1R964[\"BT@?0H@(`H@('-T871I8R!,\nM:7-T(\"H*(\"!297=R:71E475E<GDH475E<GD@*G!A<G-E=')E92P@8F]O;\"`J\nM:6YS=&5A9%]F;&%G+\"!,:7-T(\"HJ<75A;%]P<F]D=6-T<RD*(\"![\"B`@\"4-M\nM9%1Y<&4)\"65V96YT.PHA(`E,:7-T\"2`@(\"IP<F]D=6-T7W%U97)I97,@/2!.\nM24P[\"B$@\"6EN=`D)\"7)E<W5L=%]R96QA=&EO;B`](#`[\"B`@\"B`@\"4%S<V5R\nM=\"AP87)S971R964@(3T@3E5,3\"D[\"B`@\"B`@\"65V96YT(#T@<&%R<V5T<F5E\nM+3YC;VUM86YD5'EP93L*(\"`*(\"`):68@*&5V96YT(#T]($--1%]55$E,2519\nM*0H@(`D)<F5T=7)N($Y)3#L*(\"`*+2TM(#(T-3@L,C0X.2`M+2TM\"B`@\"7)E\nM='5R;B!R97-U;'1S.PH@('T*(\"`*(\"`*(\"`*(\"!S=&%T:6,@3&ES=\"`J\"B`@\nM4F5W<FET95%U97)Y*%%U97)Y(\"IP87)S971R964L(&)O;VP@*FEN<W1E861?\nM9FQA9RP@3&ES=\"`J*G%U86Q?<')O9'5C=',I\"B`@>PH@(`E#;614>7!E\"0EE\nM=F5N=#L*(2`)3&ES=`D@(\"`)*G!R;V1U8W1?<75E<FEE<R`]($Y)3#L*(2`)\nM:6YT\"0ER97-U;'1?<F5L871I;VX@/2`P.PHA(`E286YG951B;$5N=')Y\"2IR\nM=%]E;G1R>3L*(2`)4F5L871I;VX)<G1?96YT<GE?<F5L871I;VX@/2!.54Q,\nM.PHA(`E2=6QE3&]C:PDJ<G1?96YT<GE?;&]C:W,@/2!.54Q,.PH@(`H@(`E!\nM<W-E<G0H<&%R<V5T<F5E(\"$]($Y53$PI.PH@(`H@(`EE=F5N=\"`]('!A<G-E\nM=')E92T^8V]M;6%N9%1Y<&4[\"B`@\"BL@\"2\\J\"BL@\"2`J(%-%3$5#5\"!R=6QE\nM<R!A<F4@:&%N9&QE9\"!L871E<B!W:&5N('=E(&AA=F4@86QL('1H90HK(`D@\nM*B!Q=65R:65S('1H870@<VAO=6QD(&=E=\"!E>&5C=71E9`HK(`D@*B\\**R`)\nM:68@*&5V96YT(#T]($--1%]314Q%0U0I\"BL@\"0ER971U<FX@3DE,.PHK(`HK\nM(`DO*@HK(`D@*B!5=&EL:71I97,@87)E;B=T(')E=W)I='1E;B!A=\"!A;&P@\nM+2!W:'D@:7,@=&AI<R!H97)E/PHK(`D@*B\\*(\"`):68@*&5V96YT(#T]($--\nM1%]55$E,2519*0H@(`D)<F5T=7)N($Y)3#L*(\"`**BHJ*BHJ*BHJ*BHJ*BHJ\nM\"BHJ*B`X,#,L.#@Q(\"HJ*BH*(\"`*(\"`)<F5S=6QT7W)E;&%T:6]N(#T@<&%R\nM<V5T<F5E+3YR97-U;'1296QA=&EO;CL*(\"`*(2`):68@*&5V96YT(\"$]($--\nM1%]314Q%0U0I\"B$@\"7L*(2`*(2`)\"2\\J\"B$@\"0D@*B!T:&4@<W1A=&5M96YT\nM(&ES(&%N('5P9&%T92P@:6YS97)T(&]R(&1E;&5T90HA(`D)(\"HO\"B$@\"0E2\nM86YG951B;$5N=')Y(\"IR=%]E;G1R>3L*(2`)\"5)E;&%T:6]N\"7)T7V5N=')Y\nM7W)E;&%T:6]N(#T@3E5,3#L*(2`)\"5)U;&5,;V-K(\"`@*G)T7V5N=')Y7VQO\nM8VMS(#T@3E5,3#L*(2`*(2`)\"7)T7V5N=')Y(#T@<G1?9F5T8V@H<F5S=6QT\nM7W)E;&%T:6]N+\"!P87)S971R964M/G)T86)L92D[\"B$@\"0ER=%]E;G1R>5]R\nM96QA=&EO;B`](&AE87!?;W!E;G(H<G1?96YT<GDM/G)E;&YA;64I.PHA(`D)\nM<G1?96YT<GE?;&]C:W,@/2!R=%]E;G1R>5]R96QA=&EO;BT^<F1?<G5L97,[\nM\"B$@\"0EH96%P7V-L;W-E*')T7V5N=')Y7W)E;&%T:6]N*3L*(2`*(2`)\"6EF\nM(\"AR=%]E;G1R>5]L;V-K<R`A/2!.54Q,*0HA(`D)>PHA(`D)\"4QI<W0)(\"`@\nM*FQO8VMS(#T*(2`)\"0EM871C:$QO8VMS*&5V96YT+\"!R=%]E;G1R>5]L;V-K\nM<RP@<F5S=6QT7W)E;&%T:6]N+\"!P87)S971R964I.PHA(`HA(`D)\"7!R;V1U\nM8W1?<75E<FEE<R`]\"B$@\"0D)\"69I<F52=6QE<RAP87)S971R964L\"B$@\"0D)\nM\"0D)(\"!R97-U;'1?<F5L871I;VXL\"B$@\"0D)\"0D)(\"!E=F5N=\"P*(2`)\"0D)\nM\"0D@(&EN<W1E861?9FQA9RP*(2`)\"0D)\"0D@(&QO8VMS+`HA(`D)\"0D)\"2`@\nM<75A;%]P<F]D=6-T<RD[\"B$@\"0E]\"B$@\"B$@\"0DO*B`M+2TM+2TM+2TM\"B$@\nM\"0D@*B!D965P4F5W<FET95%U97)Y(&1O97,@;F]T(&AA;F1L92!T:&4@<VET\nM=6%T:6]N\"B$@\"0D@*B!W:&5R92!A('%U97)Y(&9I<F5D(&)Y(&$@<G5L92!U\nM<V5S(')E;&%T:6]N<R!T:&%T\"B$@\"0D@*B!H879E(&EN<W1E860@<V5L96-T\nM(')U;&5S(&1E9FEN960@*'9I97=S(&%N9\"!T:&4@;&EK92DN\"B$@\"0D@*B!3\nM;R!W92!C87)E(&9O<B!T:&5M(&AE<F4N\"B$@\"0D@*B`M+2TM+2TM+2TM\"B$@\nM\"0D@*B\\*(2`)\"6EF(\"AP<F]D=6-T7W%U97)I97,@(3T@3DE,*0HA(`D)>PHA\nM(`D)\"4QI<W0)(\"`@*G!Q.PHA(`D)\"5%U97)Y\"2`@(\"IT;7`[\"B$@\"0D)3&ES\nM=`D@(\"`J;F5W7W!R;V1U8W1S(#T@3DE,.PHA(`HA(`D)\"69O<F5A8V@H<'$L\nM('!R;V1U8W1?<75E<FEE<RD*(2`)\"0E[\"B$@\"0D)\"71M<\"`](\"A1=65R>2`J\nM*2!L9FER<W0H<'$I.PHA(`D)\"0ET;7`@/2!297=R:71E4&]S='!R;V-E<W-.\nM;VY396QE8W0H=&UP*3L*(2`)\"0D);F5W7W!R;V1U8W1S(#T@;&%P<&5N9\"AN\nM97=?<')O9'5C=',L('1M<\"D[\"B$@\"0D)?0HA(`D)\"7!R;V1U8W1?<75E<FEE\nM<R`](&YE=U]P<F]D=6-T<SL*(2`)\"7T*(2`*(2`)\"7)E='5R;B!P<F]D=6-T\nM7W%U97)I97,[\"B`@\"7T*+2`)96QS90HM(`E[\"BT@\"BT@\"0DO*@HM(`D)(\"H@\nM=&AE('-T871E;65N=\"!I<R!A('-E;&5C=`HM(`D)(\"HO\"BT@\"0E1=65R>0D@\nM(\"`J;W1H97([\"B`@\"B$@\"0DO*@HA(`D)(\"H@07!P;'E2971R:65V95)U;&4@\nM8VAA;F=E<R!T:&4@<F%N9V4@=&%B;&4@6%A8(%5N:6]N<R!A<F4@8V]P:65D\nM\"B$@\"0D@*B!A9V%I;BX*(2`)\"2`J+PHA(`D);W1H97(@/2!C;W!Y3V)J96-T\nM*'!A<G-E=')E92D[\"B`@\"BT@\"0ER971U<FX*+2`)\"0E0<F]C97-S4F5T<FEE\nM=F51=65R>2AO=&AE<BP@<&%R<V5T<F5E+3YR=&%B;&4L\"BT@\"0D)\"0D)\"0D@\nM:6YS=&5A9%]F;&%G+\"!&04Q312D[\"BT@\"7T*(\"!]\"B`@\"B`@+RH*(\"`@*B!T\nM;R!A=F]I9\"!I;F9I;FET92!R96-U<G-I;VXL('=E(')E<W1R:6-T('1H92!N\nM=6UB97(@;V8@=&EM97,@82!Q=65R>0H@(\"`J(&-A;B!B92!R97=R:71T96XN\nM($1E=&5C=&EN9R!C>6-L97,@:7,@;&5F=\"!F;W(@=&AE(')E861E<B!A<R!A\nM;B!E>&-E<F-I<V4N\"BTM+2`R-#DU+#(U,C@@+2TM+0H@(`H@(`ER97-U;'1?\nM<F5L871I;VX@/2!P87)S971R964M/G)E<W5L=%)E;&%T:6]N.PH@(`HA(`DO\nM*@HA(`D@*B!T:&4@<W1A=&5M96YT(&ES(&%N('5P9&%T92P@:6YS97)T(&]R\nM(&1E;&5T92`M(&9I<F4@<G5L97,*(2`)(\"H@;VX@:70N\"B$@\"2`J+PHA(`ER\nM=%]E;G1R>2`](')T7V9E=&-H*')E<W5L=%]R96QA=&EO;BP@<&%R<V5T<F5E\nM+3YR=&%B;&4I.PHA(`ER=%]E;G1R>5]R96QA=&EO;B`](&AE87!?;W!E;G(H\nM<G1?96YT<GDM/G)E;&YA;64I.PHA(`ER=%]E;G1R>5]L;V-K<R`](')T7V5N\nM=')Y7W)E;&%T:6]N+3YR9%]R=6QE<SL*(2`):&5A<%]C;&]S92AR=%]E;G1R\nM>5]R96QA=&EO;BD[\"B$@\"B$@\"6EF(\"AR=%]E;G1R>5]L;V-K<R`A/2!.54Q,\nM*0HA(`E[\"B$@\"0E,:7-T\"2`@(\"IL;V-K<R`]\"B$@\"0EM871C:$QO8VMS*&5V\nM96YT+\"!R=%]E;G1R>5]L;V-K<RP@<F5S=6QT7W)E;&%T:6]N+\"!P87)S971R\nM964I.PHA(`HA(`D)<')O9'5C=%]Q=65R:65S(#T*(2`)\"0EF:7)E4G5L97,H\nM<&%R<V5T<F5E+`HA(`D)\"0D)(\"!R97-U;'1?<F5L871I;VXL\"B$@\"0D)\"0D@\nM(&5V96YT+`HA(`D)\"0D)(\"!I;G-T96%D7V9L86<L\"B$@\"0D)\"0D@(&QO8VMS\nM+`HA(`D)\"0D)(\"!Q=6%L7W!R;V1U8W1S*3L*(\"`)?0H@(`HA(`ER971U<FX@\nM<')O9'5C=%]Q=65R:65S.PH@(`H@('T*(\"`**R`*(\"`O*@H@(\"`J('1O(&%V\nM;VED(&EN9FEN:71E(')E8W5R<VEO;BP@=V4@<F5S=')I8W0@=&AE(&YU;6)E\nM<B!O9B!T:6UE<R!A('%U97)Y\"B`@(\"H@8V%N(&)E(')E=W)I='1E;BX@1&5T\nM96-T:6YG(&-Y8VQE<R!I<R!L969T(&9O<B!T:&4@<F5A9&5R(&%S(&%N(&5X\nM8V5R8VES92X**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`X.#<L.3@Y(\"HJ*BH*(\"!S\nM=&%T:6,@:6YT\"6YU;5%U97)Y4F5W<FET94EN=F]K960@/2`P.PH@(`H@(\"\\J\nM\"BT@(\"H@475E<GE297=R:71E(\"T*+2`@*@D@(')E=W)I=&4@;VYE('%U97)Y\nM('9I82!1=65R>5)E=W)I=&4@<WES=&5M+\"!P;W-S:6)L>2!R971U<FYI;F<@\nM,\"P@;W(@;6%N>0HM(\"`J\"2`@<75E<FEE<PHM(\"`J+PHM($QI<W0@*@HM(%%U\nM97)Y4F5W<FET92A1=65R>2`J<&%R<V5T<F5E*0HM('L*+2`)4F5W<FET95!R\nM97!R;V-E<W-1=65R>2AP87)S971R964I.PHM(`HM(`E1=65R>5)E=W)I=&53\nM=6),:6YK*'!A<G-E=')E92T^<75A;\"D[\"BT@\"5%U97)Y4F5W<FET95-U8DQI\nM;FLH<&%R<V5T<F5E+3YH879I;F=1=6%L*3L*+2`*+2`)<F5T=7)N(%%U97)Y\nM4F5W<FET94]N92AP87)S971R964I.PHM('T*+2`*+2`O*@HM(\"`J\"5%U97)Y\nM4F5W<FET95-U8DQI;FL*+2`@*@HM(\"`J\"51H:7,@<F5W<FET97,@=&AE(%-U\nM8DQI;FL@<W5B<75E<FEE<R!F:7)S=\"P@9&]I;F<@=&AE(&QO=V5S=\"!O;F5S\nM(&9I<G-T+@HM(\"`J\"5=E(&%L<F5A9'D@:&%V92!C;V1E(&EN('1H92!M86EN\nM(')E=W)I=&4@;&]O<',@=&\\@<')O8V5S<R!C;W)R96QA=&5D\"BT@(\"H)=F%R\nM:6%B;&5S(&9R;VT@=7!P97(@<75E<FEE<R!T:&%T(&5X:7-T(&EN('-U8G%U\nM97)I97,N\"BT@(\"HO\"BT@<W1A=&EC('9O:60*+2!1=65R>5)E=W)I=&53=6),\nM:6YK*$YO9&4@*FYO9&4I\"BT@>PHM(`EI9B`H;F]D92`]/2!.54Q,*0HM(`D)\nM<F5T=7)N.PHM(`HM(`ES=VET8V@@*&YO9&5486<H;F]D92DI\"BT@\"7L*+2`)\nM\"6-A<V4@5%]487)G971%;G1R>3H*+2`)\"0EB<F5A:SL*+2`)\"6-A<V4@5%]!\nM9V=R96<Z\"BT@\"0D)8G)E86L[\"BT@\"0EC87-E(%1?17AP<CH*+2`)\"0E[\"BT@\nM\"0D)\"45X<'()(\"`@*F5X<'(@/2`H17AP<B`J*2!N;V1E.PHM(`HM(`D)\"0E1\nM=65R>5)E=W)I=&53=6),:6YK*\"A.;V1E(\"HI(&5X<'(M/F%R9W,I.PHM(`D)\nM\"7T*+2`)\"0EB<F5A:SL*+2`)\"6-A<V4@5%]687(Z\"BT@\"0D)8G)E86L[\"BT@\nM\"0EC87-E(%1?3&ES=#H*+2`)\"0E[\"BT@\"0D)\"4QI<W0)(\"`@*FP[\"BT@\"BT@\nM\"0D)\"69O<F5A8V@H;\"P@*$QI<W0@*BD@;F]D92D*+2`)\"0D)\"5%U97)Y4F5W\nM<FET95-U8DQI;FLH;&9I<G-T*&PI*3L*+2`)\"0E]\"BT@\"0D)8G)E86L[\"BT@\nM\"0EC87-E(%1?4W5B3&EN:SH*+2`)\"0E[\"BT@\"0D)\"5-U8DQI;FL@(\"`@*G-U\nM8FQI;FL@/2`H4W5B3&EN:R`J*2!N;V1E.PHM(`D)\"0E1=65R>0D@(\"`J<75E\nM<GD@/2`H475E<GD@*BD@<W5B;&EN:RT^<W5B<V5L96-T.PHM(`D)\"0E,:7-T\nM\"2`@(\"IR970[\"BT@\"BT@\"0D)\"2\\J\"BT@\"0D)\"2`J($YE<W0@9&]W;B!F:7)S\nM=\"X@(%=E(&1O('1H:7,@<V\\@:68@82!R97=R:71E(&%D9',@80HM(`D)\"0D@\nM*B!3=6),:6YK('=E(&1O;B=T('!R;V-E<W,@:70@87,@<&%R=\"!O9B!T:&ES\nM(&QO;W`N\"BT@\"0D)\"2`J+PHM(`D)\"0E1=65R>5)E=W)I=&53=6),:6YK*\"A.\nM;V1E(\"HI('%U97)Y+3YQ=6%L*3L*+2`*+2`)\"0D)475E<GE297=R:71E4W5B\nM3&EN:R@H3F]D92`J*2!Q=65R>2T^:&%V:6YG475A;\"D[\"BT@\"BT@\"0D)\"7)E\nM=\"`](%%U97)Y4F5W<FET94]N92AQ=65R>2D[\"BT@\"0D)\"6EF(\"@A<F5T*0HM\nM(`D)\"0D)<W5B;&EN:RT^<W5B<V5L96-T(#T@3E5,3#L*+2`)\"0D)96QS92!I\nM9B`H;&YE>'0H<F5T*2`]/2!.24PI\"BT@\"0D)\"0ES=6)L:6YK+3YS=6)S96QE\nM8W0@/2!L9FER<W0H<F5T*3L*+2`)\"0D)96QS90HM(`D)\"0D)96QO9RA%4E)/\nM4BP@(D1O;B=T(&MN;W<@:&]W('1O('!R;V-E<W,@<W5B<75E<GD@=&AA=\"!R\nM97=R:71E<R!T;R!M=6QT:7!L92!Q=65R:65S+B(I.PHM(`D)\"7T*+2`)\"0EB\nM<F5A:SL*+2`)\"61E9F%U;'0Z\"BT@\"0D)+RH@:6=N;W)E('1H92!O=&AE<G,@\nM*B\\*+2`)\"0EB<F5A:SL*+2`)?0HM(`ER971U<FX[\"BT@?0HM(`HM(\"\\J\"BT@\nM(\"H@475E<GE/;F5297=R:71E(\"T*+2`@*@D@(')E=W)I=&4@;VYE('%U97)Y\nM\"BT@(\"HO\"BT@<W1A=&EC($QI<W0@*@HM(%%U97)Y4F5W<FET94]N92A1=65R\nM>2`J<&%R<V5T<F5E*0HM('L*+2`);G5M475E<GE297=R:71E26YV;VME9\"`]\nM(#`[\"BT@\"BT@\"2\\J\"BT@\"2`J('1A:V4@82!D965P(&)R96%T:\"!A;F0@87!P\nM;'D@86QL('1H92!R97=R:71E(')U;&5S(\"T@87D*+2`)(\"HO\"BT@\"7)E='5R\nM;B!D965P4F5W<FET95%U97)Y*'!A<G-E=')E92D[\"BT@?0HM(`HM(\"\\J\"B`@\nM(\"H@9&5E<%)E=W)I=&51=65R>2`M\"B`@(\"H)(\"!R97=R:71E<R!T:&4@<75E\nM<GD@86YD(&%P<&QY('1H92!R=6QE<R!A9V%I;B!O;B!T:&4@<75E<FEE<R!R\nM97=R:71T96X*(\"`@*B\\*+2TM(#(U,S0L,C4S.2`M+2TM\"BHJ*BHJ*BHJ*BHJ\nM*BHJ*@HJ*BH@,3`T,\"PQ,#0R(\"HJ*BH*+2TM(#(U.3`L,C8Y,R`M+2TM\"B`@\nM\"B`@\"7)E='5R;B!R97=R:71T96X[\"B`@?0HK(`HK(`HK(\"\\J\"BL@(\"H@475E\nM<GE/;F5297=R:71E(\"T**R`@*@D@(')E=W)I=&4@;VYE('%U97)Y\"BL@(\"HO\nM\"BL@<W1A=&EC($QI<W0@*@HK(%%U97)Y4F5W<FET94]N92A1=65R>2`J<&%R\nM<V5T<F5E*0HK('L**R`);G5M475E<GE297=R:71E26YV;VME9\"`](#`[\"BL@\nM\"BL@\"2\\J\"BL@\"2`J('1A:V4@82!D965P(&)R96%T:\"!A;F0@87!P;'D@86QL\nM('1H92!R97=R:71E(')U;&5S(\"T@87D**R`)(\"HO\"BL@\"7)E='5R;B!D965P\nM4F5W<FET95%U97)Y*'!A<G-E=')E92D[\"BL@?0HK(`HK(`HK(\"\\J(\"TM+2TM\nM+2TM+2T**R`@*B!297=R:71E4')E<')O8V5S<U%U97)Y(\"T**R`@*@EA9&IU\nM<W0@9&5T86EL<R!I;B!T:&4@<&%R<V5T<F5E+\"!T:&4@<G5L92!S>7-T96T*\nM*R`@*@ED97!E;F1S(&]N\"BL@(\"H@+2TM+2TM+2TM+0HK(\"`J+PHK('-T871I\nM8R!V;VED\"BL@4F5W<FET95!R97!R;V-E<W-1=65R>2A1=65R>2`J<&%R<V5T\nM<F5E*0HK('L**R`)+RH@+2TM+2TM+2TM+0HK(`D@*B!I9B!T:&4@<75E<GD@\nM:&%S(&$@<F5S=6QT4F5L871I;VXL(')E87-S:6=N('1H90HK(`D@*B!R97-U\nM;'0@9&]M86EN(&YU;6)E<G,@=&\\@=&AE(&%T=')I8G5T92!N=6UB97)S(&EN\nM('1H90HK(`D@*B!T87)G970@<F5L871I;VXN($9I>$YE=R@I(&1E<&5N9',@\nM;VX@:70@=VAE;B!R97!L86-I;F<**R`)(\"H@*FYE=RH@<F5F97)E;F-E<R!I\nM;B!A(')U;&4@86-T:6]N(&)Y('1H92!E>'!R97-S:6]N<PHK(`D@*B!F<F]M\nM('1H92!R97=R:71T96X@<75E<GDN\"BL@\"2`J(\"TM+2TM+2TM+2T**R`)(\"HO\nM\"BL@\"6EF(\"AP87)S971R964M/G)E<W5L=%)E;&%T:6]N(#X@,\"D**R`)>PHK\nM(`D)4F%N9V548FQ%;G1R>2`J<G1E.PHK(`D)4F5L871I;VX)<F0[\"BL@\"0E,\nM:7-T\"2`@(\"IT;#L**R`)\"51A<F=E=$5N=')Y(\"IT;&4[\"BL@\"0EI;G0)\"0ER\nM97-D;VUN;SL**R`**R`)\"7)T92`](\"A286YG951B;$5N=')Y(\"HI(&YT:\"AP\nM87)S971R964M/G)E<W5L=%)E;&%T:6]N(\"T@,2P**R`)\"0D)\"0D)\"0EP87)S\nM971R964M/G)T86)L92D[\"BL@\"0ER9\"`](&AE87!?;W!E;G(H<G1E+3YR96QN\nM86UE*3L**R`**R`)\"69O<F5A8V@H=&PL('!A<G-E=')E92T^=&%R9V5T3&ES\nM=\"D**R`)\"7L**R`)\"0ET;&4@/2`H5&%R9V5T16YT<GD@*BD@;&9I<G-T*'1L\nM*3L**R`)\"0ER97-D;VUN;R`](&%T=&YA;65!='1.=6TH<F0L('1L92T^<F5S\nM9&]M+3YR97-N86UE*3L**R`)\"0ET;&4M/G)E<V1O;2T^<F5S;F\\@/2!R97-D\nM;VUN;SL**R`)\"7T**R`**R`)\"6AE87!?8VQO<V4H<F0I.PHK(`E]\"BL@?0HK\nM(`HK(`HK(\"\\J\"BL@(\"H@475E<GE297=R:71E(\"T**R`@*@D@(')E=W)I=&4@\nM;VYE('%U97)Y('9I82!Q=65R>2!R97=R:71E('-Y<W1E;2P@<&]S<VEB;'D@\nM<F5T=7)N:6YG(#`**R`@*@D@(&]R(&UA;GD@<75E<FEE<PHK(\"`J+PHK($QI\nM<W0@*@HK(%%U97)Y4F5W<FET92A1=65R>2`J<&%R<V5T<F5E*0HK('L**R`)\nM3&ES=`D)*G%U97)Y;&ES=#L**R`)3&ES=`D)*G)E<W5L=',@/2!.24P[\"BL@\nM\"4QI<W0)\"2IL.PHK(`E1=65R>0D)*G%U97)Y.PHK(`HK(`DO*@HK(`D@*B!3\nM=&5P(#$**R`)(\"H**R`)(\"H@5&AE<F4@<W1I;&P@<V5E;7,@<V]M971H:6YG\nM(&)R;VME;B!W:71H('1H92!R97-D;VT@;G5M8F5R<PHK(`D@*B!S;R!W92!R\nM96%S<VEG;B!T:&5M(&9I<G-T+@HK(`D@*B\\**R`)4F5W<FET95!R97!R;V-E\nM<W-1=65R>2AP87)S971R964I.PHK(`HK(`DO*@HK(`D@*B!3=&5P(#(**R`)\nM(\"H**R`)(\"H@07!P;'D@86QL(&YO;BU314Q%0U0@<G5L97,@<&]S<VEB;'D@\nM9V5T=&EN9R`P(&]R(&UA;GD@<75E<FEE<PHK(`D@*B\\**R`)<75E<GEL:7-T\nM(#T@475E<GE297=R:71E3VYE*'!A<G-E=')E92D[\"BL@\"BL@\"2\\J\"BL@\"2`J\nM(%-T97`@,PHK(`D@*@HK(`D@*B!!<'!L>2!A;&P@=&AE(%))4B!R=6QE<R!O\nM;B!E86-H('%U97)Y\"BL@\"2`J+PHK(`EF;W)E86-H(\"AL+\"!Q=65R>6QI<W0I\nM('L**R`)\"7%U97)Y(#T@*%%U97)Y(\"HI;&9I<G-T*&PI.PHK(`D)<F5S=6QT\nM<R`](&QA<'!E;F0H<F5S=6QT<RP@9FER95))4G)U;&5S*'%U97)Y*2D[\"BL@\nM\"7T**R`**R`)<F5T=7)N(')E<W5L=',[\"BL@?0HK(`HK(`ID:69F(\"UC<B!S\nM<F,N;W)I9R]B86-K96YD+W)E=W)I=&4O<F5W<FET94UA;FEP+F,@<W)C+V)A\nM8VME;F0O<F5W<FET92]R97=R:71E36%N:7`N8PHJ*BH@<W)C+F]R:6<O8F%C\nM:V5N9\"]R97=R:71E+W)E=W)I=&5-86YI<\"YC\"51H=2!397`@,C0@,3$Z-3<Z\nM,3@@,3DY.`HM+2T@<W)C+V)A8VME;F0O<F5W<FET92]R97=R:71E36%N:7`N\nM8PE&<FD@4V5P(#(U(#(Q.C$T.C4U(#$Y.3@**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ\nM*B`W,RPW.\"`J*BHJ\"BTM+2`W,RPY-2`M+2TM\"B`@\"0D)\"4]F9G-E=%9A<DYO\nM9&5S*\"A.;V1E(\"HI(&5X<'(M/F%R9W,L(&]F9G-E=\"D[\"B`@\"0D)?0H@(`D)\nM\"6)R96%K.PHK(`D)8V%S92!47TET97(Z\"BL@\"0D)>PHK(`D)\"0E)=&5R\"2`@\nM(\"II=&5R(#T@*$ET97(@*BD@;F]D93L**R`**R`)\"0D)3V9F<V5T5F%R3F]D\nM97,H*$YO9&4@*BD@:71E<BT^:71E<F5X<'(L(&]F9G-E=\"D[\"BL@\"0D)?0HK\nM(`D)\"6)R96%K.PHK(`D)8V%S92!47T%R<F%Y4F5F.@HK(`D)\"7L**R`)\"0D)\nM07)R87E2968)(\"`@*G)E9B`](\"A!<G)A>5)E9B`J*2!N;V1E.PHK(`HK(`D)\nM\"0E/9F9S971687).;V1E<R@H3F]D92`J*2!R968M/G)E9G5P<&5R:6YD97AP\nM<BP@;V9F<V5T*3L**R`)\"0D)3V9F<V5T5F%R3F]D97,H*$YO9&4@*BD@<F5F\nM+3YR969L;W=E<FEN9&5X<'(L(&]F9G-E=\"D[\"BL@\"0D)\"4]F9G-E=%9A<DYO\nM9&5S*\"A.;V1E(\"HI(')E9BT^<F5F97AP<BP@;V9F<V5T*3L**R`)\"0D)3V9F\nM<V5T5F%R3F]D97,H*$YO9&4@*BD@<F5F+3YR969A<W-G;F5X<'(L(&]F9G-E\nM=\"D[\"BL@\"0D)?0HK(`D)\"6)R96%K.PH@(`D)8V%S92!47U9A<CH*(\"`)\"0E[\nM\"B`@\"0D)\"59A<@D)(\"`@*G9A<B`](\"A687(@*BD@;F]D93L**BHJ*BHJ*BHJ\nM*BHJ*BHJ\"BHJ*B`Q-3<L,38R(\"HJ*BH*+2TM(#$W-\"PQ.38@+2TM+0H@(`D)\nM\"0E#:&%N9V5687).;V1E<R@H3F]D92`J*2!E>'!R+3YA<F=S+\"!O;&1?=F%R\nM;F\\L(&YE=U]V87)N;RP@<W5B;&5V96QS7W5P*3L*(\"`)\"0E]\"B`@\"0D)8G)E\nM86L[\"BL@\"0EC87-E(%1?271E<CH**R`)\"0E[\"BL@\"0D)\"4ET97()(\"`@*FET\nM97(@/2`H271E<B`J*2!N;V1E.PHK(`HK(`D)\"0E#:&%N9V5687).;V1E<R@H\nM3F]D92`J*2!I=&5R+3YI=&5R97AP<BP@;VQD7W9A<FYO+\"!N97=?=F%R;F\\L\nM('-U8FQE=F5L<U]U<\"D[\"BL@\"0D)?0HK(`D)\"6)R96%K.PHK(`D)8V%S92!4\nM7T%R<F%Y4F5F.@HK(`D)\"7L**R`)\"0D)07)R87E2968)(\"`@*G)E9B`](\"A!\nM<G)A>5)E9B`J*2!N;V1E.PHK(`HK(`D)\"0E#:&%N9V5687).;V1E<R@H3F]D\nM92`J*2!R968M/G)E9G5P<&5R:6YD97AP<BP@;VQD7W9A<FYO+\"!N97=?=F%R\nM;F\\L('-U8FQE=F5L<U]U<\"D[\"BL@\"0D)\"4-H86YG959A<DYO9&5S*\"A.;V1E\nM(\"HI(')E9BT^<F5F;&]W97)I;F1E>'!R+\"!O;&1?=F%R;F\\L(&YE=U]V87)N\nM;RP@<W5B;&5V96QS7W5P*3L**R`)\"0D)0VAA;F=E5F%R3F]D97,H*$YO9&4@\nM*BD@<F5F+3YR969E>'!R+\"!O;&1?=F%R;F\\L(&YE=U]V87)N;RP@<W5B;&5V\nM96QS7W5P*3L**R`)\"0D)0VAA;F=E5F%R3F]D97,H*$YO9&4@*BD@<F5F+3YR\nM969A<W-G;F5X<'(L(&]L9%]V87)N;RP@;F5W7W9A<FYO+\"!S=6)L979E;'-?\nM=7`I.PHK(`D)\"7T**R`)\"0EB<F5A:SL*(\"`)\"6-A<V4@5%]687(Z\"B`@\"0D)\nM>PH@(`D)\"0E687()\"2`@(\"IV87(@/2`H5F%R(\"HI(&YO9&4[\"BHJ*BHJ*BHJ\nM*BHJ*BHJ*@HJ*BH@,S4S+#,U.\"`J*BHJ\"BTM+2`S.#<L-#`V(\"TM+2T*(\"`)\nM\"0E297-O;'9E3F5W*&EN9F\\L('1A<F=E=&QI<W0L(\"A.;V1E(\"HJ*2`H)B@H\nM*$5X<'(@*BD@;F]D92DM/F%R9W,I*2P*(\"`)\"0D)\"2`@('-U8FQE=F5L<U]U\nM<\"D[\"B`@\"0D)8G)E86L[\"BL@\"0EC87-E(%1?271E<CH**R`)\"0E297-O;'9E\nM3F5W*&EN9F\\L('1A<F=E=&QI<W0L(\"A.;V1E(\"HJ*2`H)B@H*$ET97(@*BD@\nM;F]D92DM/FET97)E>'!R*2DL\"BL@\"0D)\"0D@(\"!S=6)L979E;'-?=7`I.PHK\nM(`D)\"6)R96%K.PHK(`D)8V%S92!47T%R<F%Y4F5F.@HK(`D)\"5)E<V]L=F5.\nM97<H:6YF;RP@=&%R9V5T;&ES=\"P@*$YO9&4@*BHI(\"@F*\"@H07)R87E2968@\nM*BD@;F]D92DM/G)E9G5P<&5R:6YD97AP<BDI+`HK(`D)\"0D)(\"`@<W5B;&5V\nM96QS7W5P*3L**R`)\"0E297-O;'9E3F5W*&EN9F\\L('1A<F=E=&QI<W0L(\"A.\nM;V1E(\"HJ*2`H)B@H*$%R<F%Y4F5F(\"HI(&YO9&4I+3YR969L;W=E<FEN9&5X\nM<'(I*2P**R`)\"0D)\"2`@('-U8FQE=F5L<U]U<\"D[\"BL@\"0D)4F5S;VQV94YE\nM=RAI;F9O+\"!T87)G971L:7-T+\"`H3F]D92`J*BD@*\"8H*\"A!<G)A>5)E9B`J\nM*2!N;V1E*2T^<F5F97AP<BDI+`HK(`D)\"0D)(\"`@<W5B;&5V96QS7W5P*3L*\nM*R`)\"0E297-O;'9E3F5W*&EN9F\\L('1A<F=E=&QI<W0L(\"A.;V1E(\"HJ*2`H\nM)B@H*$%R<F%Y4F5F(\"HI(&YO9&4I+3YR969A<W-G;F5X<'(I*2P**R`)\"0D)\nM\"2`@('-U8FQE=F5L<U]U<\"D[\"BL@\"0D)8G)E86L[\"B`@\"0EC87-E(%1?5F%R\nM.@H@(`D)\"7L*(\"`)\"0D):6YT\"0D)=&AI<U]V87)N;R`](\"AI;G0I(\"@H5F%R\nM(\"HI(&YO9&4I+3YV87)N;SL**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`T-30L-#4Y\nM(\"HJ*BH*+2TM(#4P,BPU,SD@+2TM+0H@(`D)\"0D)\"0D)\"0D@(\"!S=6)L979E\nM;'-?=7`I.PH@(`D)\"7T*(\"`)\"0EB<F5A:SL**R`)\"6-A<V4@5%])=&5R.@HK\nM(`D)\"7L**R`)\"0D)271E<@D@(\"`J:71E<B`](\"A)=&5R(\"HI(&YO9&4[\"BL@\nM\"BL@\"0D)\"6YO9&5(86YD;&5225)!='1R:6)U=&52=6QE*\"A.;V1E(\"HJ*2`H\nM)BAI=&5R+3YI=&5R97AP<BDI+\"!R=&%B;&4L\"BL@\"0D)\"0D)\"0D)\"2`@('1A\nM<F=E=&QI<W0L(')T7VEN9&5X+\"!A='1R7VYU;2P**R`)\"0D)\"0D)\"0D)(\"`@\nM;6]D:69I960L(&)A9'-Q;\"P**R`)\"0D)\"0D)\"0D)(\"`@<W5B;&5V96QS7W5P\nM*3L**R`)\"0E]\"BL@\"0D)8G)E86L[\"BL@\"0EC87-E(%1?07)R87E2968Z\"BL@\nM\"0D)>PHK(`D)\"0E!<G)A>5)E9@D@(\"`J<F5F(#T@*$%R<F%Y4F5F(\"HI(&YO\nM9&4[\"BL@\"BL@\"0D)\"6YO9&5(86YD;&5225)!='1R:6)U=&52=6QE*\"A.;V1E\nM(\"HJ*2`H)BAR968M/G)E9G5P<&5R:6YD97AP<BDI+\"!R=&%B;&4L\"BL@\"0D)\nM\"0D)\"0D)\"2`@('1A<F=E=&QI<W0L(')T7VEN9&5X+\"!A='1R7VYU;2P**R`)\nM\"0D)\"0D)\"0D)(\"`@;6]D:69I960L(&)A9'-Q;\"P**R`)\"0D)\"0D)\"0D)(\"`@\nM<W5B;&5V96QS7W5P*3L**R`)\"0D);F]D94AA;F1L95))4D%T=')I8G5T95)U\nM;&4H*$YO9&4@*BHI(\"@F*')E9BT^<F5F;&]W97)I;F1E>'!R*2DL(')T86)L\nM92P**R`)\"0D)\"0D)\"0D)(\"`@=&%R9V5T;&ES=\"P@<G1?:6YD97@L(&%T=')?\nM;G5M+`HK(`D)\"0D)\"0D)\"0D@(\"!M;V1I9FEE9\"P@8F%D<W%L+`HK(`D)\"0D)\nM\"0D)\"0D@(\"!S=6)L979E;'-?=7`I.PHK(`D)\"0EN;V1E2&%N9&QE4DE2071T\nM<FEB=71E4G5L92@H3F]D92`J*BD@*\"8H<F5F+3YR969E>'!R*2DL(')T86)L\nM92P**R`)\"0D)\"0D)\"0D)(\"`@=&%R9V5T;&ES=\"P@<G1?:6YD97@L(&%T=')?\nM;G5M+`HK(`D)\"0D)\"0D)\"0D@(\"!M;V1I9FEE9\"P@8F%D<W%L+`HK(`D)\"0D)\nM\"0D)\"0D@(\"!S=6)L979E;'-?=7`I.PHK(`D)\"0EN;V1E2&%N9&QE4DE2071T\nM<FEB=71E4G5L92@H3F]D92`J*BD@*\"8H<F5F+3YR969A<W-G;F5X<'(I*2P@\nM<G1A8FQE+`HK(`D)\"0D)\"0D)\"0D@(\"!T87)G971L:7-T+\"!R=%]I;F1E>\"P@\nM871T<E]N=6TL\"BL@\"0D)\"0D)\"0D)\"2`@(&UO9&EF:65D+\"!B861S<6PL\"BL@\nM\"0D)\"0D)\"0D)\"2`@('-U8FQE=F5L<U]U<\"D[\"BL@\"0D)?0HK(`D)\"6)R96%K\nM.PH@(`D)8V%S92!47U9A<CH*(\"`)\"0E[\"B`@\"0D)\"6EN=`D)\"71H:7-?=F%R\nM;F\\@/2`H*%9A<B`J*2!N;V1E*2T^=F%R;F\\[\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ\nM*BH@-3DT+#4Y.2`J*BHJ\"BTM+2`V-S0L-S`V(\"TM+2T*(\"`)\"0D)17AP<@D@\nM(\"`J97AP<B`](\"A%>'!R(\"HI(&YO9&4[\"B`@\"B`@\"0D)\"6YO9&5(86YD;&56\nM:65W4G5L92@H3F]D92`J*BD@*\"8H97AP<BT^87)G<RDI+`HK(`D)\"0D)\"0D)\nM(\"`@<G1A8FQE+\"!T87)G971L:7-T+`HK(`D)\"0D)\"0D)(\"`@<G1?:6YD97@L\nM(&UO9&EF:65D+\"!S=6)L979E;'-?=7`I.PHK(`D)\"7T**R`)\"0EB<F5A:SL*\nM*R`)\"6-A<V4@5%])=&5R.@HK(`D)\"7L**R`)\"0D)271E<@D@(\"`J:71E<B`]\nM(\"A)=&5R(\"HI(&YO9&4[\"BL@\"BL@\"0D)\"6YO9&5(86YD;&56:65W4G5L92@H\nM3F]D92`J*BD@*\"8H:71E<BT^:71E<F5X<'(I*2P**R`)\"0D)\"0D)\"2`@(')T\nM86)L92P@=&%R9V5T;&ES=\"P**R`)\"0D)\"0D)\"2`@(')T7VEN9&5X+\"!M;V1I\nM9FEE9\"P@<W5B;&5V96QS7W5P*3L**R`)\"0E]\"BL@\"0D)8G)E86L[\"BL@\"0EC\nM87-E(%1?07)R87E2968Z\"BL@\"0D)>PHK(`D)\"0E!<G)A>5)E9@D@(\"`J<F5F\nM(#T@*$%R<F%Y4F5F(\"HI(&YO9&4[\"BL@\"BL@\"0D)\"6YO9&5(86YD;&56:65W\nM4G5L92@H3F]D92`J*BD@*\"8H<F5F+3YR969U<'!E<FEN9&5X<'(I*2P**R`)\nM\"0D)\"0D)\"2`@(')T86)L92P@=&%R9V5T;&ES=\"P**R`)\"0D)\"0D)\"2`@(')T\nM7VEN9&5X+\"!M;V1I9FEE9\"P@<W5B;&5V96QS7W5P*3L**R`)\"0D);F]D94AA\nM;F1L959I97=2=6QE*\"A.;V1E(\"HJ*2`H)BAR968M/G)E9FQO=V5R:6YD97AP\nM<BDI+`HK(`D)\"0D)\"0D)(\"`@<G1A8FQE+\"!T87)G971L:7-T+`HK(`D)\"0D)\nM\"0D)(\"`@<G1?:6YD97@L(&UO9&EF:65D+\"!S=6)L979E;'-?=7`I.PHK(`D)\nM\"0EN;V1E2&%N9&QE5FEE=U)U;&4H*$YO9&4@*BHI(\"@F*')E9BT^<F5F97AP\nM<BDI+`HK(`D)\"0D)\"0D)(\"`@<G1A8FQE+\"!T87)G971L:7-T+`HK(`D)\"0D)\nM\"0D)(\"`@<G1?:6YD97@L(&UO9&EF:65D+\"!S=6)L979E;'-?=7`I.PHK(`D)\nM\"0EN;V1E2&%N9&QE5FEE=U)U;&4H*$YO9&4@*BHI(\"@F*')E9BT^<F5F87-S\nM9VYE>'!R*2DL\"B`@\"0D)\"0D)\"0D@(\"!R=&%B;&4L('1A<F=E=&QI<W0L\"B`@\nM\"0D)\"0D)\"0D@(\"!R=%]I;F1E>\"P@;6]D:69I960L('-U8FQE=F5L<U]U<\"D[\nM\"B`@\"0D)?0ID:69F(\"UC<B!S<F,N;W)I9R]I;F-L=61E+W)E=W)I=&4O;&]C\nM:W,N:\"!S<F,O:6YC;'5D92]R97=R:71E+VQO8VMS+F@**BHJ('-R8RYO<FEG\nM+VEN8VQU9&4O<F5W<FET92]L;V-K<RYH\"51H=2!397`@,C0@,3$Z-3@Z-#`@\nM,3DY.`HM+2T@<W)C+VEN8VQU9&4O<F5W<FET92]L;V-K<RYH\"49R:2!397`@\nM,C4@,C`Z,#<Z-3$@,3DY.`HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#$Y+#(S(\"HJ\nM*BH*+2TM(#$Y+#(T(\"TM+2T*(\"`*(\"!E>'1E<FX@3&ES=\"`J;6%T8VA,;V-K\nM<RA#;614>7!E(&5V96YT+\"!2=6QE3&]C:R`J<G5L96QO8VMS+\"!I;G0@=F%R\nM;F\\L\"B`@\"0D@(\"!1=65R>2`J<&%R<V5T<F5E*3L**R!E>'1E<FX@=F]I9\"!C\nM:&5C:TQO8VM097)M<RA,:7-T(\"IL;V-K<RP@475E<GD@*G!A<G-E=')E92P@\nM:6YT(')T7VEN9&5X*3L*(\"`*(\"`C96YD:68)(\"\\J($Q/0TM37T@@*B\\*9&EF\nM9B`M8W(@<W)C+F]R:6<O=&5S=\"]R96=R97-S+V5X<&5C=&5D+W)U;&5S+F]U\nM=\"!S<F,O=&5S=\"]R96=R97-S+V5X<&5C=&5D+W)U;&5S+F]U=`HJ*BH@<W)C\nM+F]R:6<O=&5S=\"]R96=R97-S+V5X<&5C=&5D+W)U;&5S+F]U=`E7960@4V5P\nM(#$V(#$V.C,U.C`V(#$Y.3@*+2TM('-R8R]T97-T+W)E9W)E<W,O97AP96-T\nM960O<G5L97,N;W5T\"49R:2!397`@,C4@,C$Z,S(Z,30@,3DY.`HJ*BHJ*BHJ\nM*BHJ*BHJ*BH**BHJ(#8W,2PV-S,@*BHJ*@HM+2T@-C<Q+#@U,2`M+2TM\"B`@\nM,C`P?$]+\"B`@*#(@<F]W<RD*(\"`**R!1545263H@8W)E871E('1A8FQE(')T\nM97-T7W9I97<Q(\"AA(&EN=#0L(&(@=&5X=\"P@=B!B;V]L*3L**R!1545263H@\nM8W)E871E('1A8FQE(')T97-T7W9I97<R(\"AA(&EN=#0I.PHK(%%515)9.B!C\nM<F5A=&4@=&%B;&4@<G1E<W1?=FEE=S,@*&$@:6YT-\"P@8B!T97AT*3L**R!1\nM545263H@8W)E871E('1A8FQE(')T97-T7W9I97<T(\"AA(&EN=#0L(&(@=&5X\nM=\"P@8R!I;G0T*3L**R!1545263H@8W)E871E('9I97<@<G1E<W1?=G9I97<Q\nM(&%S('-E;&5C=\"!A+\"!B(&9R;VT@<G1E<W1?=FEE=S$@6`HK(`EW:&5R92`P\nM(#P@*'-E;&5C=\"!C;W5N=\"@J*2!F<F]M(')T97-T7W9I97<R(%D@=VAE<F4@\nM62YA(#T@6\"YA*3L**R!1545263H@8W)E871E('9I97<@<G1E<W1?=G9I97<R\nM(&%S('-E;&5C=\"!A+\"!B(&9R;VT@<G1E<W1?=FEE=S$@=VAE<F4@=CL**R!1\nM545263H@8W)E871E('9I97<@<G1E<W1?=G9I97<S(&%S('-E;&5C=\"!A+\"!B\nM(&9R;VT@<G1E<W1?=G9I97<R(%@**R`)=VAE<F4@,\"`\\(\"AS96QE8W0@8V]U\nM;G0H*BD@9G)O;2!R=&5S=%]V:65W,B!9('=H97)E(%DN82`](%@N82D[\"BL@\nM455%4EDZ(&-R96%T92!V:65W(')T97-T7W9V:65W-\"!A<R!S96QE8W0@6\"YA\nM+\"!8+F(L(&-O=6YT*%DN82D@87,@<F5F8V]U;G0**R`)9G)O;2!R=&5S=%]V\nM:65W,2!8+\"!R=&5S=%]V:65W,B!9\"BL@\"7=H97)E(%@N82`](%DN80HK(`EG\nM<F]U<\"!B>2!8+F$L(%@N8CL**R!1545263H@8W)E871E(&9U;F-T:6]N(')T\nM97-T7W9I97=F=6YC,2AI;G0T*2!R971U<FYS(&EN=#0@87,**R`))W-E;&5C\nM=\"!C;W5N=\"@J*2!F<F]M(')T97-T7W9I97<R('=H97)E(&$@/2`D,2<**R`)\nM;&%N9W5A9V4@)W-Q;\"<[\"BL@455%4EDZ(&-R96%T92!V:65W(')T97-T7W9V\nM:65W-2!A<R!S96QE8W0@82P@8BP@<G1E<W1?=FEE=V9U;F,Q*&$I(&%S(')E\nM9F-O=6YT\"BL@\"69R;VT@<G1E<W1?=FEE=S$[\"BL@455%4EDZ(&EN<V5R=\"!I\nM;G1O(')T97-T7W9I97<Q('9A;'5E<R`H,2P@)VET96T@,2<L(\"=T)RD[\"BL@\nM455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<Q('9A;'5E<R`H,BP@)VET\nM96T@,B<L(\"=T)RD[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<Q\nM('9A;'5E<R`H,RP@)VET96T@,R<L(\"=T)RD[\"BL@455%4EDZ(&EN<V5R=\"!I\nM;G1O(')T97-T7W9I97<Q('9A;'5E<R`H-\"P@)VET96T@-\"<L(\"=F)RD[\"BL@\nM455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<Q('9A;'5E<R`H-2P@)VET\nM96T@-2<L(\"=T)RD[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<Q\nM('9A;'5E<R`H-BP@)VET96T@-B<L(\"=F)RD[\"BL@455%4EDZ(&EN<V5R=\"!I\nM;G1O(')T97-T7W9I97<Q('9A;'5E<R`H-RP@)VET96T@-R<L(\"=T)RD[\"BL@\nM455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<Q('9A;'5E<R`H.\"P@)VET\nM96T@.\"<L(\"=T)RD[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R\nM('9A;'5E<R`H,BD[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R\nM('9A;'5E<R`H,BD[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R\nM('9A;'5E<R`H-\"D[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R\nM('9A;'5E<R`H-2D[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R\nM('9A;'5E<R`H-RD[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R\nM('9A;'5E<R`H-RD[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R\nM('9A;'5E<R`H-RD[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R\nM('9A;'5E<R`H-RD[\"BL@455%4EDZ('-E;&5C=\"`J(&9R;VT@<G1E<W1?=G9I\nM97<Q.PHK(&%\\8B`@(\"`@\"BL@+2LM+2TM+2T**R`R?&ET96T@,@HK(#1\\:71E\nM;2`T\"BL@-7QI=&5M(#4**R`W?&ET96T@-PHK(\"@T(')O=W,I\"BL@\"BL@455%\nM4EDZ('-E;&5C=\"`J(&9R;VT@<G1E<W1?=G9I97<R.PHK(&%\\8B`@(\"`@\"BL@\nM+2LM+2TM+2T**R`Q?&ET96T@,0HK(#)\\:71E;2`R\"BL@,WQI=&5M(#,**R`U\nM?&ET96T@-0HK(#=\\:71E;2`W\"BL@.'QI=&5M(#@**R`H-B!R;W=S*0HK(`HK\nM(%%515)9.B!S96QE8W0@*B!F<F]M(')T97-T7W9V:65W,SL**R!A?&(@(\"`@\nM(`HK(\"TK+2TM+2TM\"BL@,GQI=&5M(#(**R`U?&ET96T@-0HK(#=\\:71E;2`W\nM\"BL@*#,@<F]W<RD**R`**R!1545263H@<V5L96-T(\"H@9G)O;2!R=&5S=%]V\nM=FEE=S0[\"BL@87QB(\"`@(\"!\\<F5F8V]U;G0**R`M*RTM+2TM+2LM+2TM+2TM\nM+0HK(#)\\:71E;2`R?\"`@(\"`@(\"`R\"BL@-'QI=&5M(#1\\(\"`@(\"`@(#$**R`U\nM?&ET96T@-7P@(\"`@(\"`@,0HK(#=\\:71E;2`W?\"`@(\"`@(\"`T\"BL@*#0@<F]W\nM<RD**R`**R!1545263H@<V5L96-T(\"H@9G)O;2!R=&5S=%]V=FEE=S4[\"BL@\nM87QB(\"`@(\"!\\<F5F8V]U;G0**R`M*RTM+2TM+2LM+2TM+2TM+0HK(#%\\:71E\nM;2`Q?\"`@(\"`@(\"`P\"BL@,GQI=&5M(#)\\(\"`@(\"`@(#(**R`S?&ET96T@,WP@\nM(\"`@(\"`@,`HK(#1\\:71E;2`T?\"`@(\"`@(\"`Q\"BL@-7QI=&5M(#5\\(\"`@(\"`@\nM(#$**R`V?&ET96T@-GP@(\"`@(\"`@,`HK(#=\\:71E;2`W?\"`@(\"`@(\"`T\"BL@\nM.'QI=&5M(#A\\(\"`@(\"`@(#`**R`H.\"!R;W=S*0HK(`HK(%%515)9.B!I;G-E\nM<G0@:6YT;R!R=&5S=%]V:65W,R!S96QE8W0@*B!F<F]M(')T97-T7W9V:65W\nM,2!W:&5R92!A(#P@-SL**R!1545263H@<V5L96-T(\"H@9G)O;2!R=&5S=%]V\nM:65W,SL**R!A?&(@(\"`@(`HK(\"TK+2TM+2TM\"BL@,GQI=&5M(#(**R`T?&ET\nM96T@-`HK(#5\\:71E;2`U\"BL@*#,@<F]W<RD**R`**R!1545263H@9&5L971E\nM(&9R;VT@<G1E<W1?=FEE=S,[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T\nM7W9I97<S('-E;&5C=\"`J(&9R;VT@<G1E<W1?=G9I97<R('=H97)E(&$@(3T@\nM-2!A;F0@8B`A?B`G,B<[\"BL@455%4EDZ('-E;&5C=\"`J(&9R;VT@<G1E<W1?\nM=FEE=S,[\"BL@87QB(\"`@(\"`**R`M*RTM+2TM+0HK(#%\\:71E;2`Q\"BL@,WQI\nM=&5M(#,**R`W?&ET96T@-PHK(#A\\:71E;2`X\"BL@*#0@<F]W<RD**R`**R!1\nM545263H@9&5L971E(&9R;VT@<G1E<W1?=FEE=S,[\"BL@455%4EDZ(&EN<V5R\nM=\"!I;G1O(')T97-T7W9I97<S('-E;&5C=\"`J(&9R;VT@<G1E<W1?=G9I97<S\nM.PHK(%%515)9.B!S96QE8W0@*B!F<F]M(')T97-T7W9I97<S.PHK(&%\\8B`@\nM(\"`@\"BL@+2LM+2TM+2T**R`R?&ET96T@,@HK(#5\\:71E;2`U\"BL@-WQI=&5M\nM(#<**R`H,R!R;W=S*0HK(`HK(%%515)9.B!D96QE=&4@9G)O;2!R=&5S=%]V\nM:65W,SL**R!1545263H@:6YS97)T(&EN=&\\@<G1E<W1?=FEE=S0@<V5L96-T\nM(\"H@9G)O;2!R=&5S=%]V=FEE=S0@=VAE<F4@,R`^(')E9F-O=6YT.PHK(%%5\nM15)9.B!S96QE8W0@*B!F<F]M(')T97-T7W9I97<T.PHK(&%\\8B`@(\"`@?&,*\nM*R`M*RTM+2TM+2LM\"BL@,GQI=&5M(#)\\,@HK(#1\\:71E;2`T?#$**R`U?&ET\nM96T@-7PQ\"BL@*#,@<F]W<RD**R`**R!1545263H@9&5L971E(&9R;VT@<G1E\nM<W1?=FEE=S0[\"BL@455%4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W9I97<T('-E\nM;&5C=\"`J(&9R;VT@<G1E<W1?=G9I97<U('=H97)E(&$@/B`R(&%N9\"!R969C\nM;W5N=\"`](#`[\"BL@455%4EDZ('-E;&5C=\"`J(&9R;VT@<G1E<W1?=FEE=S0[\nM\"BL@87QB(\"`@(\"!\\8PHK(\"TK+2TM+2TM*RT**R`S?&ET96T@,WPP\"BL@-GQI\nM=&5M(#9\\,`HK(#A\\:71E;2`X?#`**R`H,R!R;W=S*0HK(`HK(%%515)9.B!D\nM96QE=&4@9G)O;2!R=&5S=%]V:65W-#L**R!1545263H@8W)E871E('1A8FQE\nM(')T97-T7V-O;7`@*`HK(`EP87)T\"71E>'0L\"BL@\"75N:70)8VAA<B@T*2P*\nM*R`)<VEZ90EF;&]A=`HK(\"D[\"BL@455%4EDZ(&-R96%T92!T86)L92!R=&5S\nM=%]U;FET9F%C=\"`H\"BL@\"75N:70)8VAA<B@T*2P**R`)9F%C=&]R\"69L;V%T\nM\"BL@*3L**R!1545263H@8W)E871E('9I97<@<G1E<W1?=F-O;7`@87,**R`)\nM<V5L96-T(%@N<&%R=\"P@*%@N<VEZ92`J(%DN9F%C=&]R*2!A<R!S:7IE7VEN\nM7V-M\"BL@\"0D)9G)O;2!R=&5S=%]C;VUP(%@L(')T97-T7W5N:71F86-T(%D*\nM*R`)\"0EW:&5R92!8+G5N:70@/2!9+G5N:70[\"BL@455%4EDZ(&EN<V5R=\"!I\nM;G1O(')T97-T7W5N:71F86-T('9A;'5E<R`H)VTG+\"`Q,#`N,\"D[\"BL@455%\nM4EDZ(&EN<V5R=\"!I;G1O(')T97-T7W5N:71F86-T('9A;'5E<R`H)V-M)RP@\nM,2XP*3L**R!1545263H@:6YS97)T(&EN=&\\@<G1E<W1?=6YI=&9A8W0@=F%L\nM=65S(\"@G:6YC:\"<L(#(N-30I.PHK(%%515)9.B!I;G-E<G0@:6YT;R!R=&5S\nM=%]C;VUP('9A;'5E<R`H)W`Q)RP@)VTG+\"`U+C`I.PHK(%%515)9.B!I;G-E\nM<G0@:6YT;R!R=&5S=%]C;VUP('9A;'5E<R`H)W`R)RP@)VTG+\"`S+C`I.PHK\nM(%%515)9.B!I;G-E<G0@:6YT;R!R=&5S=%]C;VUP('9A;'5E<R`H)W`S)RP@\nM)V-M)RP@-2XP*3L**R!1545263H@:6YS97)T(&EN=&\\@<G1E<W1?8V]M<\"!V\nM86QU97,@*\"=P-\"<L(\"=C;2<L(#$U+C`I.PHK(%%515)9.B!I;G-E<G0@:6YT\nM;R!R=&5S=%]C;VUP('9A;'5E<R`H)W`U)RP@)VEN8V@G+\"`W+C`I.PHK(%%5\nM15)9.B!I;G-E<G0@:6YT;R!R=&5S=%]C;VUP('9A;'5E<R`H)W`V)RP@)VEN\nM8V@G+\"`T+C0I.PHK(%%515)9.B!S96QE8W0@*B!F<F]M(')T97-T7W9C;VUP\nM(&]R9&5R(&)Y('!A<G0[\"BL@<&%R='QS:7IE7VEN7V-M\"BL@+2TM+2LM+2TM\nM+2TM+2TM\"BL@<#$@('P@(\"`@(\"`@-3`P\"BL@<#(@('P@(\"`@(\"`@,S`P\"BL@\nM<#,@('P@(\"`@(\"`@(\"`U\"BL@<#0@('P@(\"`@(\"`@(#$U\"BL@<#4@('P@(\"`@\nM(#$W+C<X\"BL@<#8@('P@(\"`@,3$N,3<V\"BL@*#8@<F]W<RD**R`**R!15452\nM63H@<V5L96-T(\"H@9G)O;2!R=&5S=%]V8V]M<\"!W:&5R92!S:7IE7VEN7V-M\nM(#X@,3`N,\"!O<F1E<B!B>2!S:7IE7VEN7V-M('5S:6YG(#X[\"BL@<&%R='QS\nM:7IE7VEN7V-M\"BL@+2TM+2LM+2TM+2TM+2TM\"BL@<#$@('P@(\"`@(\"`@-3`P\nM\"BL@<#(@('P@(\"`@(\"`@,S`P\"BL@<#4@('P@(\"`@(#$W+C<X\"BL@<#0@('P@\nM(\"`@(\"`@(#$U\"BL@<#8@('P@(\"`@,3$N,3<V\"BL@*#4@<F]W<RD**R`*9&EF\nM9B`M8W(@<W)C+F]R:6<O=&5S=\"]R96=R97-S+W-Q;\"]R=6QE<RYS<6P@<W)C\nM+W1E<W0O<F5G<F5S<R]S<6PO<G5L97,N<W%L\"BHJ*B!S<F,N;W)I9R]T97-T\nM+W)E9W)E<W,O<W%L+W)U;&5S+G-Q;`E7960@4V5P(#$V(#$V.C,U.C,T(#$Y\nM.3@*+2TM('-R8R]T97-T+W)E9W)E<W,O<W%L+W)U;&5S+G-Q;`E&<FD@4V5P\nM(#(U(#(Q.C,Q.C4S(#$Y.3@**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`T,#0L-#`V\nM(\"HJ*BH*+2TM(#0P-\"PU,#,@+2TM+0H@('-E;&5C=\"`J(&9R;VT@<G1E<W1?\nM;F]T:&XR.PH@('-E;&5C=\"`J(&9R;VT@<G1E<W1?;F]T:&XS.PH@(`HK(&-R\nM96%T92!T86)L92!R=&5S=%]V:65W,2`H82!I;G0T+\"!B('1E>'0L('8@8F]O\nM;\"D[\"BL@8W)E871E('1A8FQE(')T97-T7W9I97<R(\"AA(&EN=#0I.PHK(&-R\nM96%T92!T86)L92!R=&5S=%]V:65W,R`H82!I;G0T+\"!B('1E>'0I.PHK(&-R\nM96%T92!T86)L92!R=&5S=%]V:65W-\"`H82!I;G0T+\"!B('1E>'0L(&,@:6YT\nM-\"D[\"BL@8W)E871E('9I97<@<G1E<W1?=G9I97<Q(&%S('-E;&5C=\"!A+\"!B\nM(&9R;VT@<G1E<W1?=FEE=S$@6\"`**R`)=VAE<F4@,\"`\\(\"AS96QE8W0@8V]U\nM;G0H*BD@9G)O;2!R=&5S=%]V:65W,B!9('=H97)E(%DN82`](%@N82D[\"BL@\nM8W)E871E('9I97<@<G1E<W1?=G9I97<R(&%S('-E;&5C=\"!A+\"!B(&9R;VT@\nM<G1E<W1?=FEE=S$@=VAE<F4@=CL**R!C<F5A=&4@=FEE=R!R=&5S=%]V=FEE\nM=S,@87,@<V5L96-T(&$L(&(@9G)O;2!R=&5S=%]V=FEE=S(@6`HK(`EW:&5R\nM92`P(#P@*'-E;&5C=\"!C;W5N=\"@J*2!F<F]M(')T97-T7W9I97<R(%D@=VAE\nM<F4@62YA(#T@6\"YA*3L**R!C<F5A=&4@=FEE=R!R=&5S=%]V=FEE=S0@87,@\nM<V5L96-T(%@N82P@6\"YB+\"!C;W5N=\"A9+F$I(&%S(')E9F-O=6YT\"BL@\"69R\nM;VT@<G1E<W1?=FEE=S$@6\"P@<G1E<W1?=FEE=S(@60HK(`EW:&5R92!8+F$@\nM/2!9+F$**R`)9W)O=7`@8GD@6\"YA+\"!8+F([\"BL@8W)E871E(&9U;F-T:6]N\nM(')T97-T7W9I97=F=6YC,2AI;G0T*2!R971U<FYS(&EN=#0@87,**R`))W-E\nM;&5C=\"!C;W5N=\"@J*2!F<F]M(')T97-T7W9I97<R('=H97)E(&$@/2`D,2<*\nM*R`);&%N9W5A9V4@)W-Q;\"<[\"BL@8W)E871E('9I97<@<G1E<W1?=G9I97<U\nM(&%S('-E;&5C=\"!A+\"!B+\"!R=&5S=%]V:65W9G5N8S$H82D@87,@<F5F8V]U\nM;G0**R`)9G)O;2!R=&5S=%]V:65W,3L**R`**R!I;G-E<G0@:6YT;R!R=&5S\nM=%]V:65W,2!V86QU97,@*#$L(\"=I=&5M(#$G+\"`G=\"<I.PHK(&EN<V5R=\"!I\nM;G1O(')T97-T7W9I97<Q('9A;'5E<R`H,BP@)VET96T@,B<L(\"=T)RD[\"BL@\nM:6YS97)T(&EN=&\\@<G1E<W1?=FEE=S$@=F%L=65S(\"@S+\"`G:71E;2`S)RP@\nM)W0G*3L**R!I;G-E<G0@:6YT;R!R=&5S=%]V:65W,2!V86QU97,@*#0L(\"=I\nM=&5M(#0G+\"`G9B<I.PHK(&EN<V5R=\"!I;G1O(')T97-T7W9I97<Q('9A;'5E\nM<R`H-2P@)VET96T@-2<L(\"=T)RD[\"BL@:6YS97)T(&EN=&\\@<G1E<W1?=FEE\nM=S$@=F%L=65S(\"@V+\"`G:71E;2`V)RP@)V8G*3L**R!I;G-E<G0@:6YT;R!R\nM=&5S=%]V:65W,2!V86QU97,@*#<L(\"=I=&5M(#<G+\"`G=\"<I.PHK(&EN<V5R\nM=\"!I;G1O(')T97-T7W9I97<Q('9A;'5E<R`H.\"P@)VET96T@.\"<L(\"=T)RD[\nM\"BL@\"BL@:6YS97)T(&EN=&\\@<G1E<W1?=FEE=S(@=F%L=65S(\"@R*3L**R!I\nM;G-E<G0@:6YT;R!R=&5S=%]V:65W,B!V86QU97,@*#(I.PHK(&EN<V5R=\"!I\nM;G1O(')T97-T7W9I97<R('9A;'5E<R`H-\"D[\"BL@:6YS97)T(&EN=&\\@<G1E\nM<W1?=FEE=S(@=F%L=65S(\"@U*3L**R!I;G-E<G0@:6YT;R!R=&5S=%]V:65W\nM,B!V86QU97,@*#<I.PHK(&EN<V5R=\"!I;G1O(')T97-T7W9I97<R('9A;'5E\nM<R`H-RD[\"BL@:6YS97)T(&EN=&\\@<G1E<W1?=FEE=S(@=F%L=65S(\"@W*3L*\nM*R!I;G-E<G0@:6YT;R!R=&5S=%]V:65W,B!V86QU97,@*#<I.PHK(`HK('-E\nM;&5C=\"`J(&9R;VT@<G1E<W1?=G9I97<Q.PHK('-E;&5C=\"`J(&9R;VT@<G1E\nM<W1?=G9I97<R.PHK('-E;&5C=\"`J(&9R;VT@<G1E<W1?=G9I97<S.PHK('-E\nM;&5C=\"`J(&9R;VT@<G1E<W1?=G9I97<T.PHK('-E;&5C=\"`J(&9R;VT@<G1E\nM<W1?=G9I97<U.PHK(`HK(&EN<V5R=\"!I;G1O(')T97-T7W9I97<S('-E;&5C\nM=\"`J(&9R;VT@<G1E<W1?=G9I97<Q('=H97)E(&$@/\"`W.PHK('-E;&5C=\"`J\nM(&9R;VT@<G1E<W1?=FEE=S,[\"BL@9&5L971E(&9R;VT@<G1E<W1?=FEE=S,[\nM\"BL@\"BL@:6YS97)T(&EN=&\\@<G1E<W1?=FEE=S,@<V5L96-T(\"H@9G)O;2!R\nM=&5S=%]V=FEE=S(@=VAE<F4@82`A/2`U(&%N9\"!B(\"%^(\"<R)SL**R!S96QE\nM8W0@*B!F<F]M(')T97-T7W9I97<S.PHK(&1E;&5T92!F<F]M(')T97-T7W9I\nM97<S.PHK(`HK(&EN<V5R=\"!I;G1O(')T97-T7W9I97<S('-E;&5C=\"`J(&9R\nM;VT@<G1E<W1?=G9I97<S.PHK('-E;&5C=\"`J(&9R;VT@<G1E<W1?=FEE=S,[\nM\"BL@9&5L971E(&9R;VT@<G1E<W1?=FEE=S,[\"BL@\"BL@:6YS97)T(&EN=&\\@\nM<G1E<W1?=FEE=S0@<V5L96-T(\"H@9G)O;2!R=&5S=%]V=FEE=S0@=VAE<F4@\nM,R`^(')E9F-O=6YT.PHK('-E;&5C=\"`J(&9R;VT@<G1E<W1?=FEE=S0[\"BL@\nM9&5L971E(&9R;VT@<G1E<W1?=FEE=S0[\"BL@\"BL@:6YS97)T(&EN=&\\@<G1E\nM<W1?=FEE=S0@<V5L96-T(\"H@9G)O;2!R=&5S=%]V=FEE=S4@=VAE<F4@82`^\nM(#(@86YD(')E9F-O=6YT(#T@,#L**R!S96QE8W0@*B!F<F]M(')T97-T7W9I\nM97<T.PHK(&1E;&5T92!F<F]M(')T97-T7W9I97<T.PHK(\"TM\"BL@+2T@5&5S\nM=\"!F;W(@8V]M<'5T871I;VYS(&EN('9I97=S\"BL@+2T**R!C<F5A=&4@=&%B\nM;&4@<G1E<W1?8V]M<\"`H\"BL@\"7!A<G0)=&5X=\"P**R`)=6YI=`EC:&%R*#0I\nM+`HK(`ES:7IE\"69L;V%T\"BL@*3L**R`**R`**R!C<F5A=&4@=&%B;&4@<G1E\nM<W1?=6YI=&9A8W0@*`HK(`EU;FET\"6-H87(H-\"DL\"BL@\"69A8W1O<@EF;&]A\nM=`HK(\"D[\"BL@\"BL@8W)E871E('9I97<@<G1E<W1?=F-O;7`@87,@\"BL@\"7-E\nM;&5C=\"!8+G!A<G0L(\"A8+G-I>F4@*B!9+F9A8W1O<BD@87,@<VEZ95]I;E]C\nM;0HK(`D)\"69R;VT@<G1E<W1?8V]M<\"!8+\"!R=&5S=%]U;FET9F%C=\"!9\"BL@\nM\"0D)=VAE<F4@6\"YU;FET(#T@62YU;FET.PHK(`HK(`HK(&EN<V5R=\"!I;G1O\nM(')T97-T7W5N:71F86-T('9A;'5E<R`H)VTG+\"`Q,#`N,\"D[\"BL@:6YS97)T\nM(&EN=&\\@<G1E<W1?=6YI=&9A8W0@=F%L=65S(\"@G8VTG+\"`Q+C`I.PHK(&EN\nM<V5R=\"!I;G1O(')T97-T7W5N:71F86-T('9A;'5E<R`H)VEN8V@G+\"`R+C4T\nM*3L**R`**R!I;G-E<G0@:6YT;R!R=&5S=%]C;VUP('9A;'5E<R`H)W`Q)RP@\nM)VTG+\"`U+C`I.PHK(&EN<V5R=\"!I;G1O(')T97-T7V-O;7`@=F%L=65S(\"@G\nM<#(G+\"`G;2<L(#,N,\"D[\"BL@:6YS97)T(&EN=&\\@<G1E<W1?8V]M<\"!V86QU\nM97,@*\"=P,R<L(\"=C;2<L(#4N,\"D[\"BL@:6YS97)T(&EN=&\\@<G1E<W1?8V]M\nM<\"!V86QU97,@*\"=P-\"<L(\"=C;2<L(#$U+C`I.PHK(&EN<V5R=\"!I;G1O(')T\nM97-T7V-O;7`@=F%L=65S(\"@G<#4G+\"`G:6YC:\"<L(#<N,\"D[\"BL@:6YS97)T\nM(&EN=&\\@<G1E<W1?8V]M<\"!V86QU97,@*\"=P-B<L(\"=I;F-H)RP@-\"XT*3L*\nM*R`**R!S96QE8W0@*B!F<F]M(')T97-T7W9C;VUP(&]R9&5R(&)Y('!A<G0[\nM\"BL@\"BL@<V5L96-T(\"H@9G)O;2!R=&5S=%]V8V]M<\"!W:&5R92!S:7IE7VEN\nH7V-M(#X@,3`N,\"!O<F1E<B!B>2!S:7IE7VEN7V-M('5S:6YG(#X[\"@``\n`\nend\n",
"msg_date": "Mon, 28 Sep 1998 10:01:28 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "rewrite rules patch"
}
] |
[
{
"msg_contents": "> Hi All.\n> I tested libpq for win32 coming with 6.4-BETA and have a \n> question about\n> LIBPQ.\n> \n> I used LIBPQ from a C program,but coundn't connect PostgreSQL DB.\n> \n> Calling pg_connect() without WSAStartup() failed because of \n> gethostbyname()\n> error in LIBPQ.\n> so I modified DllMain in LIBPQ as follows.\n> \n> BOOL WINAPI DllMain( HINSTANCE hinstDLL, DWORD fdwReason,\n> LPVOID lpReserved ){\n> \n> \tWSADATA wsaData;\n> \tswitch (fdwReason)\n> \t\t {\n> \t case DLL_PROCESS_ATTACH:\n> \t\t WSAStartup( WSAStartup(MAKEWORD(1, 1), &wsaData);\n> \t\t break;\n> \t case DLL_PROCESS_DETACH:\n> \t\t WSACleanup();\n> \t\t break;\n> \t\t }\n> \n> \treturn (TRUE);\n> }\n> \n> and the connection to DB was OK.\n> \n> Why WSAStartup() is not called in LIBPQ?\n\nThis is probably a good thing to do :-)\nWhen I wrote the code, I thought that I was only allowed to\na) Call WSAStartup() once or\nb) All calls to WSAStartup() had to have the same version number\nBut after reading the specs a little more careful, I see that this is not\nthe case.\n\nThe specs say:\nAn application or DLL can call WSAStartup more than once if it needs to\nobtain the WSAData structure information more than once. On each such call\nthe application can specify any version number supported by the DLL.\n\nThe question is - what will happen when you call it with different versions.\nSay the application asks for Winsock 2.0, and then the Libpq library goes in\nand asks for Winsock 1.1. Will the application still have access to Winsock\n2.0 functions?\nWith the current implementation, I don't beleive it will make a difference -\nI don't think that any part of the Winsock system is actually hidden if you\nask for a lower version. But it is _permitted_ by the specification that the\nDLL can hide parts that belong to a higher version than the one requested.\n\nSo I'm not 100% sure... Does anybody have access to a Winsock that actually\nhides some details when you ask for the wrong version?\n\n\n//Magnus\n",
"msg_date": "Mon, 28 Sep 1998 10:30:01 +0200",
"msg_from": "Magnus Hagander <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
},
{
"msg_contents": "> > Hi All.\n> > I tested libpq for win32 coming with 6.4-BETA and have a\n> > question about\n> > LIBPQ.\n> >\n> > I used LIBPQ from a C program,but coundn't connect PostgreSQL DB.\n> >\n> > Calling pg_connect() without WSAStartup() failed because of\n> > gethostbyname()\n> > error in LIBPQ.\n> > so I modified DllMain in LIBPQ as follows.\n> >\n> > BOOL WINAPI DllMain( HINSTANCE hinstDLL, DWORD fdwReason,\n> > LPVOID lpReserved ){\n> >\n> > \tWSADATA wsaData;\n> > \tswitch (fdwReason)\n> > \t\t {\n> > \t case DLL_PROCESS_ATTACH:\n> > \t\t WSAStartup( WSAStartup(MAKEWORD(1, 1), &wsaData);\n> > \t\t break;\n> > \t case DLL_PROCESS_DETACH:\n> > \t\t WSACleanup();\n> > \t\t break;\n> > \t\t }\n> >\n> > \treturn (TRUE);\n> > }\n> >\n> > and the connection to DB was OK.\n> >\n> > Why WSAStartup() is not called in LIBPQ?\n>\n> This is probably a good thing to do :-)\n> When I wrote the code, I thought that I was only allowed to\n> a) Call WSAStartup() once or\n> b) All calls to WSAStartup() had to have the same version number\n> But after reading the specs a little more careful, I see that this is not\n> the case.\n>\n> The specs say:\n> An application or DLL can call WSAStartup more than once if it needs to\n> obtain the WSAData structure information more than once. On each such call\n> the application can specify any version number supported by the DLL.\n>\n> The question is - what will happen when you call it with different\nversions.\n> Say the application asks for Winsock 2.0, and then the Libpq library goes\nin\n> and asks for Winsock 1.1. Will the application still have access to\nWinsock\n> 2.0 functions?\n\nI think that current DLL supports version 1.0,1.1,2.0,2.1 and 2.2 and we can\nspecify any version from 1.0 to 2.2 .\nFor example,if we call WSAStartup(MAKEWORD(1.0),&wsaData) from libpq\nand call WSAStartup(MAKEWORD(2,2),&wsaData) from psql,both return OK\nand wsaData.wVersion is 1.0 for libpq and 2.2 for psql.\n\n> With the current implementation, I don't beleive it will make a\ndifference -\n> I don't think that any part of the Winsock system is actually hidden if\nyou\n> ask for a lower version. But it is _permitted_ by the specification that\nthe\n> DLL can hide parts that belong to a higher version than the one requested.\n>\n> So I'm not 100% sure... Does anybody have access to a Winsock that\nactually\n> hides some details when you ask for the wrong version?\n>\n\nBy the specs there may be the DLL that doesn't support lower versions.\nIn that case my code doesn't work well.\nBut we can delay to call WSAStartup() after the first socket call in LIBPQ.\nMy example code in fe-connect.c is such as follows.\n\n\thp = gethostbyname(conn->pghost);\n#ifdef\tWIN32\n\tif ((hp == NULL) && (GetLastError() == WSANOTINITIALISED))\n\t{\n\t\tWSADATA\twsaData;\n\t\tif (WSAStartup(MAKEWORD(1,1),&wsaData))\n\t\t{\n\t\t\tfprintf(stderr, \"Failed to start winsock: %i\\n\", WSAGetLastError());\n\t\t\texit(1);\n\t\t}\n\t\t....\n\t\t???? for WSACleanup() ????\n\t\t....\n\t\thp = gethostbyname(conn->host);\n\t}\n#endif\n\tif ((hp == NULL) || (hp->h_addrtype != AF_INET))\n\t{\n\t\t(void) sprintf(conn->errorMessage,\n\t\t\t\t\t \"connectDB() -- unknown hostname: %s\\n\",\n\t\t\t\t\t conn->pghost);\n\t\tgoto connect_errReturn;\n\t}\n\nHiroshi Inoue\[email protected]\n\n",
"msg_date": "Tue, 29 Sep 1998 08:58:44 +0900",
"msg_from": "\"Hiroshi Inoue\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
},
{
"msg_contents": "\nOn 28-Sep-98 Hiroshi Inoue wrote:\n> I think that current DLL supports version 1.0,1.1,2.0,2.1 and 2.2 and we\n> can\n> specify any version from 1.0 to 2.2 .\n> For example,if we call WSAStartup(MAKEWORD(1.0),&wsaData) from libpq\n> and call WSAStartup(MAKEWORD(2,2),&wsaData) from psql,both return OK\n> and wsaData.wVersion is 1.0 for libpq and 2.2 for psql.\n\nWhichever is done, try not to ever request a version higher than what is\nrequired. It forces upgrades that may not even be necessary. It's been\nawhile since I've done windows code, but I've been bitten by requiring a\nhigher version of a library than necessary - more than once without even\nrealising it. It's not very fun when it comes time to fix it.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n",
"msg_date": "Mon, 28 Sep 1998 20:28:05 -0400 (EDT)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
},
{
"msg_contents": ">\n> On 28-Sep-98 Hiroshi Inoue wrote:\n> > I think that current DLL supports version 1.0,1.1,2.0,2.1 and 2.2 and we\n> > can\n> > specify any version from 1.0 to 2.2 .\n> > For example,if we call WSAStartup(MAKEWORD(1.0),&wsaData) from libpq\n> > and call WSAStartup(MAKEWORD(2,2),&wsaData) from psql,both return OK\n> > and wsaData.wVersion is 1.0 for libpq and 2.2 for psql.\n>\n> Whichever is done, try not to ever request a version higher than what is\n> required. It forces upgrades that may not even be necessary. It's been\n> awhile since I've done windows code, but I've been bitten by requiring a\n> higher version of a library than necessary - more than once without even\n> realising it. It's not very fun when it comes time to fix it.\n>\n\nHow about my sample code at the end of my previous post.\nIn that code,libpq does nothing if main application calls WSAStartup().\n\nHiroshi Inoue\[email protected]\n\n",
"msg_date": "Tue, 29 Sep 1998 10:46:34 +0900",
"msg_from": "\"Hiroshi Inoue\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
},
{
"msg_contents": "On Tue, 29 Sep 1998, Hiroshi Inoue wrote:\n\n> >\n> > On 28-Sep-98 Hiroshi Inoue wrote:\n> > > I think that current DLL supports version 1.0,1.1,2.0,2.1 and 2.2 and we\n> > > can\n> > > specify any version from 1.0 to 2.2 .\n> > > For example,if we call WSAStartup(MAKEWORD(1.0),&wsaData) from libpq\n> > > and call WSAStartup(MAKEWORD(2,2),&wsaData) from psql,both return OK\n> > > and wsaData.wVersion is 1.0 for libpq and 2.2 for psql.\n> >\n> > Whichever is done, try not to ever request a version higher than what is\n> > required. It forces upgrades that may not even be necessary. It's been\n> > awhile since I've done windows code, but I've been bitten by requiring a\n> > higher version of a library than necessary - more than once without even\n> > realising it. It's not very fun when it comes time to fix it.\n> >\n> \n> How about my sample code at the end of my previous post.\n> In that code,libpq does nothing if main application calls WSAStartup().\n\nAccording to the winsock 1.1 spec an intermediate dll should handle it's\nown startup and cleanup functions. Thinking about it, you really have \nno way of knowing if the app is going to stop using and unload winsock\nif you're not done with it. It may have loaded it for any number of \nreasons not necessarily related to the database access.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n\n",
"msg_date": "Tue, 29 Sep 1998 06:09:06 -0400 (EDT)",
"msg_from": "Vince Vielhaber <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
}
] |
[
{
"msg_contents": "I have been looking into the problem reported by Massimo, and also seen\nby me, that backends sometimes exit without cleaning up entries they\nmade in the pg_listener table. I have fixed several contributing bugs\n(having to do with the limited size of the on_shmem_exit table), but\nI still see the problem occasionally.\n\nI just discovered one fairly fundamental reason why there's still a\nproblem:\n\n\ttest=> listen test;\n\tLISTEN\n\ttest=> begin;\n\tBEGIN\n\ttest=> \\q\n\t$\n\n\t(start new psql)\n\n\ttest=> select * from pg_listener;\n\trelname|listenerpid|notification\n\t-------+-----------+------------\n\ttest | 20024| 0\n\t(1 row)\n\nOops. It would seem that async.c is *trying* to clean up, but the\nchanges are thrown away because they are considered part of a\nnever-committed transaction :-(. More generally, I believe that\na backend killed in the middle of any individual SQL statement will\nfail to remove its LISTEN entries, because it will think that any\ndatabase changes made just before exit ought to be discarded.\n\nIf we want a backend to clean out its pg_listen entries at shutdown,\nit looks to me like backend exit will have to work like this:\n\n1. Abort the current transaction, if any, and exit the current\n transaction block, if any.\n\n2. Perform LISTEN cleanup within a new transaction. Commit this\n transaction.\n\n3. Continue with closing up shop.\n\nI don't know whether there are any other things besides LISTEN cleanup\nthat really need to be done, not discarded, during backend exit. If\nthere are, then we probably ought to restructure the exit sequence to\nhandle this cleanly. But I'm worried about doing this much processing\nduring backend exit --- if you're trying to kill a backend that's gone\nwrong, the last thing you want is for it to start doing new\ntransactions.\n\n(BTW, that last thought also says that we *still* won't be able to get\nrid of pg_listen entries 100% reliably. In a panic exit scenario,\nie after one backend has crashed, none of the backends will clean up\ntheir pg_listen entries. We dare not tamper with that. At least\nI dare not.)\n\nI'm half tempted to say that we should rip out the cleanup code and\nnot even try to remove pg_listener entries at backend exit. They'll\ngo away anyway over time, because the next notifier to try to notify\nthat relation name will delete the entries when he discovers he can't\nsignal that backend PID anymore. So cleanup-at-exit is not at all a\ncritical function, and it's hard to justify a bunch of new complexity\njust for this.\n\nComments? This depends a lot on whether you think LISTEN cleanup\nis a unique case, or whether there will be other standard cleanup\nactivities that a backend ought to perform. (Is there such a thing\nas an ON EXIT hook in SQL?)\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 28 Sep 1998 10:28:26 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Proper cleanup at backend exit"
},
{
"msg_contents": "> \n> I have been looking into the problem reported by Massimo, and also seen\n> by me, that backends sometimes exit without cleaning up entries they\n> made in the pg_listener table. I have fixed several contributing bugs\n> (having to do with the limited size of the on_shmem_exit table), but\n\ncould you explain this to us ?\n\n> I still see the problem occasionally.\n> \n> I just discovered one fairly fundamental reason why there's still a\n> problem:\n> \n> \ttest=> listen test;\n> \tLISTEN\n> \ttest=> begin;\n> \tBEGIN\n> \ttest=> \\q\n> \t$\n> \n> \t(start new psql)\n> \n> \ttest=> select * from pg_listener;\n> \trelname|listenerpid|notification\n> \t-------+-----------+------------\n> \ttest | 20024| 0\n> \t(1 row)\n> \n> Oops. It would seem that async.c is *trying* to clean up, but the\n> changes are thrown away because they are considered part of a\n> never-committed transaction :-(. More generally, I believe that\n> a backend killed in the middle of any individual SQL statement will\n> fail to remove its LISTEN entries, because it will think that any\n> database changes made just before exit ought to be discarded.\n\nThis is correct in some sense. Any change made but not commited should\nnever be seen in the table, so the tuple is not deleted.\n\n> If we want a backend to clean out its pg_listen entries at shutdown,\n> it looks to me like backend exit will have to work like this:\n> \n> 1. Abort the current transaction, if any, and exit the current\n> transaction block, if any.\n> \n> 2. Perform LISTEN cleanup within a new transaction. Commit this\n> transaction.\n> \n> 3. Continue with closing up shop.\n> \n> I don't know whether there are any other things besides LISTEN cleanup\n> that really need to be done, not discarded, during backend exit. If\n> there are, then we probably ought to restructure the exit sequence to\n> handle this cleanly. But I'm worried about doing this much processing\n> during backend exit --- if you're trying to kill a backend that's gone\n> wrong, the last thing you want is for it to start doing new\n> transactions.\n> \n> (BTW, that last thought also says that we *still* won't be able to get\n> rid of pg_listen entries 100% reliably. In a panic exit scenario,\n> ie after one backend has crashed, none of the backends will clean up\n> their pg_listen entries. We dare not tamper with that. At least\n> I dare not.)\n> \n> I'm half tempted to say that we should rip out the cleanup code and\n> not even try to remove pg_listener entries at backend exit. They'll\n> go away anyway over time, because the next notifier to try to notify\n> that relation name will delete the entries when he discovers he can't\n> signal that backend PID anymore. So cleanup-at-exit is not at all a\n> critical function, and it's hard to justify a bunch of new complexity\n> just for this.\n> \n> Comments? This depends a lot on whether you think LISTEN cleanup\n> is a unique case, or whether there will be other standard cleanup\n> activities that a backend ought to perform. (Is there such a thing\n> as an ON EXIT hook in SQL?)\n\nThe code should be left if it works in the normal case.\nI believe that some tuples left in pg_listener don't do any harm because\nthey refer to dead backends and the notify code can detect this situation\nand delete the dead tuples. Anyway for safety I always truncate the file\nto 0 in the shell loop that keeps the postmaster alive. Also a vacuum every\nfew hours doesn't hurt.\n\n-- \nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n| Massimo Dal Zotto email: [email protected] |\n| Via Marconi, 141 phone: ++39-461-534251 |\n| 38057 Pergine Valsugana (TN) www: http://www.cs.unitn.it/~dz/ |\n| Italy pgp: finger [email protected] |\n+----------------------------------------------------------------------+\n",
"msg_date": "Mon, 28 Sep 1998 19:00:18 +0200 (MET DST)",
"msg_from": "Massimo Dal Zotto <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Proper cleanup at backend exit"
},
{
"msg_contents": "Massimo Dal Zotto <[email protected]> writes:\n>> I have fixed several contributing bugs\n>> (having to do with the limited size of the on_shmem_exit table),\n\n> could you explain this to us ?\n\nasync.c tries to arrange to clean up its entries in pg_listener by\nscheduling an exit callback routine via on_shmem_exit. (I think\nBruce or someone has been changing that stuff, and that on_shmem_exit\nused to be called something else, but it's still basically an atexit-\nlike list of procedures to call.)\n\nThe trouble is that on_shmem_exit has a fixed-size, and not very large,\narray for the callbacks. Try to register too many exit callbacks, and\nafter a while they stop getting registered.\n\nI saw two problems associated with that. First, async.c was registering\na separate callback for each pg_listen entry it's ever made. That was\neasily dealt with, since we now have an \"unlisten all\" subroutine:\nI got rid of the per-listen callback and now register only a single\ncallback over the life of the backend. It just calls UnlistenAll.\n\nSecond, the main loop in postgres.c was miscoded so that every time you\nhad an error, it added a redundant FlushBufferPool callback to the list.\nMake enough mistakes before you do your first listen, and you're still\nout of luck. A simple reordering of the code fixes that. (While I was\nat it I tried to improve the comments to make it clear where the main\nloop *really* starts, so that future hackers won't misplace code in\nPostgresMain the same way again...)\n\nIn the long term it might be a good idea to make on_shmem_exit construct\na linked list of registered callbacks (eg using Dllist) rather than\nhaving a fixed maximum number of exit callbacks. But just at the moment\nI don't think it's necessary, given these bug fixes --- and if we did\nhave a linked list in there, we'd still want to make these changes\nbecause they would be memory leaks otherwise.\n\n\t\t\tregards, tom lane\n\nPS: I hope to check these changes in shortly.\n",
"msg_date": "Mon, 28 Sep 1998 14:03:31 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Proper cleanup at backend exit "
},
{
"msg_contents": "> I have been looking into the problem reported by Massimo, and also seen\n> by me, that backends sometimes exit without cleaning up entries they\n> made in the pg_listener table. I have fixed several contributing bugs\n> (having to do with the limited size of the on_shmem_exit table), but\n> I still see the problem occasionally.\n> \n> I just discovered one fairly fundamental reason why there's still a\n> problem:\n> \n> \ttest=> listen test;\n> \tLISTEN\n> \ttest=> begin;\n> \tBEGIN\n> \ttest=> \\q\n> \t$\n> \n> \t(start new psql)\n> \n> \ttest=> select * from pg_listener;\n> \trelname|listenerpid|notification\n> \t-------+-----------+------------\n> \ttest | 20024| 0\n> \t(1 row)\n> \n> Oops. It would seem that async.c is *trying* to clean up, but the\n> changes are thrown away because they are considered part of a\n> never-committed transaction :-(. More generally, I believe that\n> a backend killed in the middle of any individual SQL statement will\n> fail to remove its LISTEN entries, because it will think that any\n> database changes made just before exit ought to be discarded.\n> \n> If we want a backend to clean out its pg_listen entries at shutdown,\n> it looks to me like backend exit will have to work like this:\n> \n> 1. Abort the current transaction, if any, and exit the current\n> transaction block, if any.\n> \n> 2. Perform LISTEN cleanup within a new transaction. Commit this\n> transaction.\n> \n> 3. Continue with closing up shop.\n> \n> I don't know whether there are any other things besides LISTEN cleanup\n> that really need to be done, not discarded, during backend exit. If\n> there are, then we probably ought to restructure the exit sequence to\n> handle this cleanly. But I'm worried about doing this much processing\n> during backend exit --- if you're trying to kill a backend that's gone\n> wrong, the last thing you want is for it to start doing new\n> transactions.\n> \n> (BTW, that last thought also says that we *still* won't be able to get\n> rid of pg_listen entries 100% reliably. In a panic exit scenario,\n> ie after one backend has crashed, none of the backends will clean up\n> their pg_listen entries. We dare not tamper with that. At least\n> I dare not.)\n> \n> I'm half tempted to say that we should rip out the cleanup code and\n> not even try to remove pg_listener entries at backend exit. They'll\n> go away anyway over time, because the next notifier to try to notify\n> that relation name will delete the entries when he discovers he can't\n> signal that backend PID anymore. So cleanup-at-exit is not at all a\n> critical function, and it's hard to justify a bunch of new complexity\n> just for this.\n> \n> Comments? This depends a lot on whether you think LISTEN cleanup\n> is a unique case, or whether there will be other standard cleanup\n> activities that a backend ought to perform. (Is there such a thing\n> as an ON EXIT hook in SQL?)\n\nSeems like we should try and clean them out, especially when the backend\nis doing a normal exit. In a panic exit, we don't need to clean it out.\nWe can install something into proc_exit() and shmem_exit() to clean out\nthe table. You can have a global variable that installed the pg_listen\ncleaner the first time a notify() is done so as not to install a new\nexit handler for each notify() call.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 17:38:46 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Proper cleanup at backend exit"
},
{
"msg_contents": "> Massimo Dal Zotto <[email protected]> writes:\n> >> I have fixed several contributing bugs\n> >> (having to do with the limited size of the on_shmem_exit table),\n> \n> > could you explain this to us ?\n> \n> async.c tries to arrange to clean up its entries in pg_listener by\n> scheduling an exit callback routine via on_shmem_exit. (I think\n> Bruce or someone has been changing that stuff, and that on_shmem_exit\n> used to be called something else, but it's still basically an atexit-\n> like list of procedures to call.)\n\nYes. The old code atexit() was called for process exit and\nshared-memory reset, and that caused problems as more and more cleanup\nhad to be done ONLY at shared memory reset time. Having two separate\nqueues and proc names helped.\n\n> \n> The trouble is that on_shmem_exit has a fixed-size, and not very large,\n> array for the callbacks. Try to register too many exit callbacks, and\n> after a while they stop getting registered.\n> \n> I saw two problems associated with that. First, async.c was registering\n> a separate callback for each pg_listen entry it's ever made. That was\n> easily dealt with, since we now have an \"unlisten all\" subroutine:\n> I got rid of the per-listen callback and now register only a single\n> callback over the life of the backend. It just calls UnlistenAll.\n\nThat is why you need some type of global variable that sets the cleanup\nthe first time, and only once. You could just call the proc_exit call,\nand prevent it from scheduling two of the same exit functions. That\nwould be cleaner.\n\n> \n> Second, the main loop in postgres.c was miscoded so that every time you\n> had an error, it added a redundant FlushBufferPool callback to the list.\n> Make enough mistakes before you do your first listen, and you're still\n> out of luck. A simple reordering of the code fixes that. (While I was\n> at it I tried to improve the comments to make it clear where the main\n> loop *really* starts, so that future hackers won't misplace code in\n> PostgresMain the same way again...)\n> \n> In the long term it might be a good idea to make on_shmem_exit construct\n> a linked list of registered callbacks (eg using Dllist) rather than\n> having a fixed maximum number of exit callbacks. But just at the moment\n> I don't think it's necessary, given these bug fixes --- and if we did\n> have a linked list in there, we'd still want to make these changes\n> because they would be memory leaks otherwise.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 18:58:39 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Proper cleanup at backend exit"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n>> [ should we work harder to ensure pg_listener entries get cleaned out? ]\n\n> Seems like we should try and clean them out, especially when the backend\n> is doing a normal exit. In a panic exit, we don't need to clean it out.\n\nOK, I will add some more code to try to cope with the exit-during-\nuncommitted-transaction case.\n\n> We can install something into proc_exit() and shmem_exit() to clean out\n> the table.\n\nThe old code inserted the cleanup action into shmem_exit(). Is that the\nright list to put it in? I'm not clear on the difference between\nshmem_exit and proc_exit lists...\n\n> You can have a global variable that installed the pg_listen\n> cleaner the first time a notify() is done so as not to install a new\n> exit handler for each notify() call.\n\nThat part's already done.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 01 Oct 1998 20:29:17 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Proper cleanup at backend exit "
},
{
"msg_contents": "> Bruce Momjian <[email protected]> writes:\n> >> [ should we work harder to ensure pg_listener entries get cleaned out? ]\n> \n> > Seems like we should try and clean them out, especially when the backend\n> > is doing a normal exit. In a panic exit, we don't need to clean it out.\n> \n> OK, I will add some more code to try to cope with the exit-during-\n> uncommitted-transaction case.\n> \n> > We can install something into proc_exit() and shmem_exit() to clean out\n> > the table.\n> \n> The old code inserted the cleanup action into shmem_exit(). Is that the\n> right list to put it in? I'm not clear on the difference between\n> shmem_exit and proc_exit lists...\n\nproc_exit is triggered on every backend exit. shmem_exit is for\npostmaster exit, or postmaster shared memory reset.\n\n> \n> > You can have a global variable that installed the pg_listen\n> > cleaner the first time a notify() is done so as not to install a new\n> > exit handler for each notify() call.\n> \n> That part's already done.\n\nCool.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 22:25:23 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Proper cleanup at backend exit"
}
] |
[
{
"msg_contents": "Hi,\n\n while fixing the subselect parseback in the new ruleutil\n functions and checking if the output is now what's needed for\n dumping rules/views I came across a little detail in the\n parser I'm confused about.\n\n Having a table\n\n CREATE TABLE t1 (a char(20));\n\n the two statements\n\n INSERT INTO t1 VALUES ('x');\n\n INSERT INTO t1 VALUES ('x'::bpchar);\n\n produce mainly the same parsetree (where the const value 'x'\n of type 1042 is embedded into a call to bpchar(bpchar, int4)\n for the padding).\n\n But in the first case argument 1 is constbyval TRUE and in\n the second one FALSE (where I feel the second one is right\n for a bpchar const). Seems that it doesn't matter later.\n\n The ruleutil functions output the typecasting any time and\n this little detail makes it harder for me to check if their\n output really recreates all the rules as the original CREATE\n statements did.\n\n The output from the ruleutil functions can differ in many\n cases from the original CREATE commands (for 'AS funcname'\n where no one was given and in the subselects area where 'attr\n in (val, val, ...)' will be explicit OR list, 'attr IN\n (SELECT ...' will be output as 'attr = ANY (SELECT ...' and\n so on). But the resulting rules will be the same AFAIK.\n\n For checking the output to be O.K. I took all the pg_rewrite\n content from a regression database (plus some more views with\n subselects etc.) into a temp file, dropped all rules and\n recreated them from what pg_rules printed. Then I took again\n all the pg_rewrite content and expected the same (and got it\n except for the constbyval diffs).\n\n I'll send in the fixes for ruleutils soon.\n\n If someone likes to add dumping rules/views to pg_dump, some\n little background details:\n\n Views could be nested (one view selects another view). So\n restoring by CREATE VIEW depends on the correct order.\n\n But a view could also be recreated by a CREATE TABLE,\n CREATE RULE, UPDATE pg_rewrite sequence. This time, if\n all the CREATE TABLE statements are made first and the\n CREATE RULE ones later, they don't depend on each other\n any more.\n\n The view pg_rules will (as mentioned by Keith) not\n include view rules any longer. So pg_dump should use\n function pg_get_ruledef() directly on pg_rewrite to get\n all the rules definitions.\n\n To reconstruct anything correct it must be executed in\n the following order:\n\n Step 1: Create all tables and views as regular tables.\n\n Step 2: Insert all data (where tables that have an\n INSTEAD rule on SELECT named _RET<tablename> receive no\n data at all).\n\n Step 3: Create all rules (except the ones named _RETpg_*\n from system views) and rename the remaining _RET* ones\n from _ret* to _RET* (CREATE RULE converts the rulename to\n lower case - that's why the UPDATE pg_rewrite is\n required).\n\n Inserting the data before recreating the rules is necessary\n because rules could fire more actions on insert, and that\n isn't wanted at reload time.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 28 Sep 1998 18:56:39 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Little parser question"
},
{
"msg_contents": "> Having a table\n> CREATE TABLE t1 (a char(20));\n> the two statements\n> INSERT INTO t1 VALUES ('x');\n> INSERT INTO t1 VALUES ('x'::bpchar);\n> produce mainly the same parsetree (where the const value 'x'\n> of type 1042 is embedded into a call to bpchar(bpchar, int4)\n> for the padding).\n> \n> But in the first case argument 1 is constbyval TRUE and in\n> the second one FALSE (where I feel the second one is right\n> for a bpchar const). Seems that it doesn't matter later.\n\nWithout actually looking at the code, I'm pretty sure that the\nunspecified constant string in the first insert above is getting\ninterpreted as a string of type UNKNOWN. Apparently it actually tests\nfor the length of the string and tries to do a \"pass by value\" if it can\nfit.\n\nI don't see this as being a big win, and if you would find it easier we\ncould try to track it down and make UNKNOWN a pass by reference always.\n\nbtw, the call to bpchar(bpchar,int4) is part of my new type coersion\ncode; this is the first release that will actually check and truncate\nstrings as they are being stored into a table. In previous releases\nlonger strings which were the result of, for example, a concatenation\nwere not length-checked on storage into the table.\n\nAnyway, let me know what you'd like...\n\n - Tom\n",
"msg_date": "Tue, 29 Sep 1998 04:30:36 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Little parser question"
},
{
"msg_contents": "> Hi,\n> \n> while fixing the subselect parseback in the new ruleutil\n> functions and checking if the output is now what's needed for\n> dumping rules/views I came across a little detail in the\n> parser I'm confused about.\n> \n> Having a table\n> \n> CREATE TABLE t1 (a char(20));\n> \n> the two statements\n> \n> INSERT INTO t1 VALUES ('x');\n> \n> INSERT INTO t1 VALUES ('x'::bpchar);\n> \n> produce mainly the same parsetree (where the const value 'x'\n> of type 1042 is embedded into a call to bpchar(bpchar, int4)\n> for the padding).\n> \n> But in the first case argument 1 is constbyval TRUE and in\n> the second one FALSE (where I feel the second one is right\n> for a bpchar const). Seems that it doesn't matter later.\n\nI have just committed a fix for this. The parse_coerce code was not\nsetting the constbyval dependent on the type. There were a few other\nplaces where this was not set properly, and those are fixed now.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 18:51:49 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Little parser question"
}
] |
[
{
"msg_contents": "Bruce,\n\n please apply the patch at the end.\n\n The patch fixes the broken backparsing of subselects in the\n utility functions for pg_views and pg_rules, changes the\n views definitions a little and adds two new system views.\n\n Even if it might look like, they don't add new features\n except for two new builtin functions (pg_get_userbyid() and\n pg_get_indexdef()). So I think it's still O.K. during BETA.\n\n New layout of pg_views:\n\n viewname name -- Name of the view\n viewowner name -- Owner name from pg_shadow\n definition text -- SELECT query that defines view\n\n Owner name will be 'unknown (UID=n)' if record in\n pg_shadow is missing.\n\n New layout of pg_rules:\n\n tablename name -- Table rule is fired for\n rulename name -- Name of the rule\n definition text -- CREATE statement of rule\n\n The view will no longer output rules which names start\n with '_RET' (these are view rewrite rules).\n\n New view pg_tables:\n\n tablename name -- Table name\n tableowner name -- Owner name from pg_shadow\n hasindexes bool -- True if table has index(es)\n hasrules bool -- True if table has rule(s)\n hastriggers bool -- True if table has trigger(s)\n\n The view does not include index relations or views.\n\n Owner name will be 'unknown (UID=n)' if record in\n pg_shadow is missing.\n\n New view pg_indexes:\n\n tablename name -- Table name index is defined on\n indexname name -- Name of index\n definition text -- CREATE statement of index\n\n If required, I could add the predicates of partial\n indexes (currently omitted) to the index definition (what\n is the final state of partial indexes - will they live on\n or get removed?).\n\n\n Any suggestions about more system views? Should the above\n ones have more fields?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\nbegin 644 ruleutil.diff\nM9&EF9B`M8W(@<W)C+F]R:6<O8F%C:V5N9\"]U=&EL<R]A9'0O<G5L975T:6QS\nM+F,@<W)C+V)A8VME;F0O=71I;',O861T+W)U;&5U=&EL<RYC\"BHJ*B!S<F,N\nM;W)I9R]B86-K96YD+W5T:6QS+V%D=\"]R=6QE=71I;',N8PE-;VX@4V5P(#(X\nM(#$S.C0U.C4T(#$Y.3@*+2TM('-R8R]B86-K96YD+W5T:6QS+V%D=\"]R=6QE\nM=71I;',N8PE-;VX@4V5P(#(X(#(S.C`W.C4P(#$Y.3@**BHJ*BHJ*BHJ*BHJ\nM*BHJ\"BHJ*B`U,BPV,2`J*BHJ\"BTM+2`U,BPW-\"`M+2TM\"B`@(VEN8VQU9&4@\nM(G5T:6QS+VQS>7-C86-H92YH(@H@(\"-I;F-L=61E(\")C871A;&]G+W!G7V-L\nM87-S+F@B\"B`@(VEN8VQU9&4@(F-A=&%L;V<O<&=?='EP92YH(@HK(\"-I;F-L\nM=61E(\")C871A;&]G+W!G7W-H861O=RYH(@HK(\"-I;F-L=61E(\")C871A;&]G\nM+W!G7VEN9&5X+F@B\"BL@(VEN8VQU9&4@(F-A=&%L;V<O<&=?;W!C;&%S<RYH\nM(@H@(\"-I;F-L=61E(\")F;6=R+F@B\"B`@\"B`@\"B`@+RH@+2TM+2TM+2TM+0HK\nM(\"`J($QO8V%L(&1A=&$@='EP97,**R`@*B`M+2TM+2TM+2TM\"BL@(\"HO\"BL@\nM='EP961E9B!S=')U8W0@47)Y2&EE<B![\"BL@\"7-T<G5C=\"!1<GE(:65R\"0DJ\nM<&%R96YT.PHK(`E1=65R>0D)\"2IQ=65R>3L**R!](%%R>4AI97([\"BL@\"BL@\nM\"BL@+RH@+2TM+2TM+2TM+0H@(\"`J($=L;V)A;\"!D871A\"B`@(\"H@+2TM+2TM\nM+2TM+0H@(\"`J+PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#8T+#8Y(\"HJ*BH*+2TM\nM(#<W+#@V(\"TM+2T*(\"!S=&%T:6,@8VAA<B`J<75E<GE?9V5T<G5L92`](\")3\nM14Q%0U0@*B!&4D]-('!G7W)E=W)I=&4@5TA%4D4@<G5L96YA;64@/2`D,2([\nM\"B`@<W1A=&EC('9O:60@*G!L86Y?9V5T=FEE=R`]($Y53$P[\"B`@<W1A=&EC\nM(&-H87(@*G%U97)Y7V=E='9I97<@/2`B4T5,14-4(\"H@1E)/32!P9U]R97=R\nM:71E(%=(15)%(')U;&5N86UE(#T@)#$@;W(@<G5L96YA;64@/2`D,B([\"BL@\nM<W1A=&EC('9O:60@*G!L86Y?9V5T86T@/2!.54Q,.PHK('-T871I8R!C:&%R\nM(\"IQ=65R>5]G971A;2`](\")314Q%0U0@*B!&4D]-('!G7V%M(%=(15)%(&]I\nM9\"`](\"0Q(CL**R!S=&%T:6,@=F]I9\"`J<&QA;E]G971O<&-L87-S(#T@3E5,\nM3#L**R!S=&%T:6,@8VAA<B`J<75E<GE?9V5T;W!C;&%S<R`](\")314Q%0U0@\nM*B!&4D]-('!G7V]P8VQA<W,@5TA%4D4@;VED(#T@)#$B.PH@(`H@(`H@(\"\\J\nM(\"TM+2TM+2TM+2T**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`W,BPW-R`J*BHJ\"BTM\nM+2`X.2PY-B`M+2TM\"B`@(\"HO\"B`@=&5X=`D@(\"`J<&=?9V5T7W)U;&5D968H\nM3F%M941A=&$@*G)N86UE*3L*(\"!T97AT\"2`@(\"IP9U]G971?=FEE=V1E9BA.\nM86UE1&%T82`J<FYA;64I.PHK('1E>'0@(\"`@(\"`@*G!G7V=E=%]I;F1E>&1E\nM9BA/:60@:6YD97AR96QI9\"D[\"BL@3F%M941A=&$@(\"`J<&=?9V5T7W5S97)B\nM>6ED*&EN=#0@=6ED*3L*(\"`*(\"`*(\"`O*B`M+2TM+2TM+2TM\"BHJ*BHJ*BHJ\nM*BHJ*BHJ*@HJ*BH@.#`L.30@*BHJ*@H@(\"`J+PH@('-T871I8R!C:&%R(\"IM\nM86ME7W)U;&5D968H2&5A<%1U<&QE(')U;&5T=7`L(%1U<&QE1&5S8R!R=6QE\nM='1C*3L*(\"!S=&%T:6,@8VAA<B`J;6%K95]V:65W9&5F*$AE87!4=7!L92!R\nM=6QE='5P+\"!4=7!L941E<V,@<G5L971T8RD[\"B$@<W1A=&EC(&-H87(@*F=E\nM=%]Q=65R>5]D968H475E<GD@*G%U97)Y*3L*(2!S=&%T:6,@8VAA<B`J9V5T\nM7W-E;&5C=%]Q=65R>5]D968H475E<GD@*G%U97)Y*3L*(2!S=&%T:6,@8VAA\nM<B`J9V5T7VEN<V5R=%]Q=65R>5]D968H475E<GD@*G%U97)Y*3L*(2!S=&%T\nM:6,@8VAA<B`J9V5T7W5P9&%T95]Q=65R>5]D968H475E<GD@*G%U97)Y*3L*\nM(2!S=&%T:6,@8VAA<B`J9V5T7V1E;&5T95]Q=65R>5]D968H475E<GD@*G%U\nM97)Y*3L*(2!S=&%T:6,@8VAA<B`J9V5T7W)U;&5?97AP<BA,:7-T(\"IR=&%B\nM;&4L(&EN=\"!R=%]I;F1E>\"P@3F]D92`J;F]D92P@8F]O;\"!V87)P<F5F:7@I\nM.PHA('-T871I8R!C:&%R(\"IG971?9G5N8U]E>'!R*$QI<W0@*G)T86)L92P@\nM:6YT(')T7VEN9&5X+\"!%>'!R(\"IE>'!R+\"!B;V]L('9A<G!R969I>\"D[\"B$@\nM<W1A=&EC(&-H87(@*F=E=%]T;&5?97AP<BA,:7-T(\"IR=&%B;&4L(&EN=\"!R\nM=%]I;F1E>\"P@5&%R9V5T16YT<GD@*G1L92P@8F]O;\"!V87)P<F5F:[email protected]@\nM('-T871I8R!C:&%R(\"IG971?8V]N<W1?97AP<BA#;VYS=\"`J8V]N<W1V86PI\nM.PH@('-T871I8R!C:&%R(\"IG971?<F5L871I;VY?;F%M92A/:60@<F5L:60I\nM.PH@('-T871I8R!C:&%R(\"IG971?871T<FEB=71E7VYA;64H3VED(')E;&ED\nM+\"!I;G0R(&%T=&YU;2D[\"B`@<W1A=&EC(&)O;VP@8VAE8VM?:69?<G1E7W5S\nM960H:6YT(')T7VEN9&5X+\"!.;V1E(\"IN;V1E+\"!I;G0@<W5P*3L*+2TM(#DY\nM+#$Q-\"`M+2TM\"B`@(\"HO\"B`@<W1A=&EC(&-H87(@*FUA:V5?<G5L961E9BA(\nM96%P5'5P;&4@<G5L971U<\"P@5'5P;&5$97-C(')U;&5T=&,I.PH@('-T871I\nM8R!C:&%R(\"IM86ME7W9I97=D968H2&5A<%1U<&QE(')U;&5T=7`L(%1U<&QE\nM1&5S8R!R=6QE='1C*3L*(2!S=&%T:6,@8VAA<B`J9V5T7W%U97)Y7V1E9BA1\nM=65R>2`J<75E<GDL(%%R>4AI97(@*G!A<F5N='%H*3L*(2!S=&%T:6,@8VAA\nM<B`J9V5T7W-E;&5C=%]Q=65R>5]D968H475E<GD@*G%U97)Y+\"!1<GE(:65R\nM(\"IQ:\"D[\"B$@<W1A=&EC(&-H87(@*F=E=%]I;G-E<G1?<75E<GE?9&5F*%%U\nM97)Y(\"IQ=65R>2P@47)Y2&EE<B`J<[email protected]('-T871I8R!C:&%R(\"IG971?\nM=7!D871E7W%U97)Y7V1E9BA1=65R>2`J<75E<GDL(%%R>4AI97(@*G%H*3L*\nM(2!S=&%T:6,@8VAA<B`J9V5T7V1E;&5T95]Q=65R>5]D968H475E<GD@*G%U\nM97)Y+\"!1<GE(:65R(\"IQ:\"D[\"B$@<W1A=&EC(&-H87(@*F=E=%]R=6QE7V5X\nM<'(H47)Y2&EE<B`J<6@L(&EN=\"!R=%]I;F1E>\"P@3F]D92`J;F]D92P@8F]O\nM;\"!V87)P<F5F:[email protected]('-T871I8R!C:&%R(\"IG971?9G5N8U]E>'!R*%%R\nM>4AI97(@*G%H+\"!I;G0@<G1?:6YD97@L($5X<'(@*F5X<'(L(&)O;VP@=F%R\nM<')E9FEX*3L*(2!S=&%T:6,@8VAA<B`J9V5T7W1L95]E>'!R*%%R>4AI97(@\nM*G%H+\"!I;G0@<G1?:6YD97@L(%1A<F=E=$5N=')Y(\"IT;&4L(&)O;VP@=F%R\nM<')E9FEX*3L*(\"!S=&%T:6,@8VAA<B`J9V5T7V-O;G-T7V5X<'(H0V]N<W0@\nM*F-O;G-T=F%L*3L**R!S=&%T:6,@8VAA<B`J9V5T7W-U8FQI;FM?97AP<BA1\nM<GE(:65R(\"IQ:\"P@:6YT(')T7VEN9&5X+\"!.;V1E(\"IN;V1E+\"!B;V]L('9A\nM<G!R969I>\"D[\"B`@<W1A=&EC(&-H87(@*F=E=%]R96QA=&EO;E]N86UE*$]I\nM9\"!R96QI9\"D[\"B`@<W1A=&EC(&-H87(@*F=E=%]A='1R:6)U=&5?;F%M92A/\nM:60@<F5L:60L(&EN=#(@871T;G5M*3L*(\"!S=&%T:6,@8F]O;\"!C:&5C:U]I\nM9E]R=&5?=7-E9\"AI;G0@<G1?:6YD97@L($YO9&4@*FYO9&4L(&EN=\"!S=7`I\nM.PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#(X.2PR.30@*BHJ*@HM+2T@,S`Y+#4X\nM,\"`M+2TM\"B`@\"B`@\"B`@+RH@+2TM+2TM+2TM+0HK(\"`J(&=E=%]V:65W9&5F\nM\"0D)+2!-86EN;'D@=&AE('-A;64@=&AI;F<L(&)U=\"!W90HK(\"`J\"0D)\"2`@\nM;VYL>2!R971U<FX@=&AE(%-%3$5#5\"!P87)T(&]F(&$@=FEE=PHK(\"`J(\"TM\nM+2TM+2TM+2T**R`@*B\\**R!T97AT(\"`@(\"`@(\"H**R!P9U]G971?:6YD97AD\nM968H3VED(&EN9&5X<F5L:60I\"BL@>PHK(`ET97AT\"0DJ:6YD97AD968[\"BL@\nM\"4AE87!4=7!L90EH=%]I9'@[\"BL@\"4AE87!4=7!L90EH=%]I9'AR96P[\"BL@\nM\"4AE87!4=7!L90EH=%]I;F1R96P[\"BL@\"4AE87!4=7!L90ES<&E?='5P.PHK\nM(`E4=7!L941E<V,)<W!I7W1T8SL**R`):6YT\"0ES<&E?9FYO.PHK(`E&;W)M\nM7W!G7VEN9&5X\"6ED>')E8SL**R`)1F]R;5]P9U]C;&%S<PEI9'AR96QR96,[\nM\"BL@\"49O<FU?<&=?8VQA<W,):6YD<F5L<F5C.PHK(`E$871U;0D)<W!I7V%R\nM9W-;,5T[\"BL@\"6-H87()\"7-P:5]N=6QL<ULR73L**R`):6YT\"0ES<&ER8SL*\nM*R`):6YT\"0EL96X[\"BL@\"6EN=`D):V5Y;F\\[\"BL@\"6-H87()\"6)U9ELX,3DR\nM73L**R`)8VAA<@D):[email protected])=.PHK(`EC:&%R\"0DJ<V5P.PHK(`HK\nM(`DO*B`M+2TM+2TM+2TM\"BL@\"2`J($-O;FYE8W0@=&\\@4U!)(&UA;F%G97(*\nM*R`)(\"H@+2TM+2TM+2TM+0HK(`D@*B\\**R`):68@*%-025]C;VYN96-T*\"D@\nM(3T@4U!)7T]+7T-/3DY%0U0I\"BL@\"0EE;&]G*$524D]2+\"`B9V5T7VEN9&5X\nM9&5F.B!C86YN;W0@8V]N;F5C=\"!T;R!34$D@;6%N86=E<B(I.PHK(`HK(`DO\nM*B`M+2TM+2TM+2TM\"BL@\"2`J($]N('1H92!F:7)S=\"!C86QL('!R97!A<F4@\nM=&AE('!L86YS('1O(&QO;VMU<\"!P9U]A;0HK(`D@*B!A;F0@<&=?;W!C;&%S\nM<RX**R`)(\"H@+2TM+2TM+2TM+0HK(`D@*B\\**R`):68@*'!L86Y?9V5T86T@\nM/3T@3E5,3\"D**R`)>PHK(`D)3VED\"0D)87)G='EP97-;,5T[\"BL@\"0EV;VED\nM\"2`@(\"IP;&%N.PHK(`HK(`D)87)G='EP97-;,%T@/2!/241/240[\"BL@\"0EP\nM;&%N(#T@4U!)7W!R97!A<F4H<75E<GE?9V5T86TL(#$L(&%R9W1Y<&5S*3L*\nM*R`)\"6EF(\"AP;&%N(#T]($Y53$PI\"BL@\"0D)96QO9RA%4E)/4BP@(E-025]P\nM<F5P87)E*\"D@9F%I;&5D(&9O<B!<(B5S7\"(B+\"!Q=65R>5]G971A;2D[\"BL@\nM\"0EP;&%N7V=E=&%M(#T@4U!)7W-A=F5P;&%N*'!L86XI.PHK(`HK(`D)87)G\nM='EP97-;,%T@/2!/241/240[\"BL@\"0EP;&%N(#T@4U!)7W!R97!A<F4H<75E\nM<GE?9V5T;W!C;&%S<RP@,2P@87)G='EP97,I.PHK(`D):68@*'!L86X@/3T@\nM3E5,3\"D**R`)\"0EE;&]G*$524D]2+\"`B4U!)7W!R97!A<F4H*2!F86EL960@\nM9F]R(%PB)7-<(B(L('%U97)Y7V=E=&]P8VQA<W,I.PHK(`D)<&QA;E]G971O\nM<&-L87-S(#T@4U!)7W-A=F5P;&%N*'!L86XI.PHK(`E]\"BL@\"BL@\"2\\J(\"TM\nM+2TM+2TM+2T**R`)(\"H@1F5T8V@@=&AE('!G7VEN9&5X('1U<&QE(&)Y('1H\nM92!/:60@;V8@=&AE(&EN9&5X\"BL@\"2`J(\"TM+2TM+2TM+2T**R`)(\"HO\"BL@\nM\"6AT7VED>\"`](%-E87)C:%-Y<T-A8VAE5'5P;&4H24Y$15A214Q)1\"P**R`)\nM\"0E/8FIE8W1)9$=E=$1A='5M*&EN9&5X<F5L:60I+\"`P+\"`P+\"`P*3L**R`)\nM:68@*\"%(96%P5'5P;&5)<U9A;&ED*&AT7VED>\"DI\"BL@\"0EE;&]G*$524D]2\nM+\"`B<WES8V%C:&4@;&]O:W5P(&9O<B!I;F1E>\"`E9\"!F86EL960B+\"!I;F1E\nM>')E;&ED*3L**R`):61X<F5C(#T@*$9O<FU?<&=?:6YD97@I1T544U1254-4\nM*&AT7VED>\"D[\"BL@\"BL@\"2\\J(\"TM+2TM+2TM+2T**R`)(\"H@1F5T8V@@=&AE\nM('!G7V-L87-S('1U<&QE(&]F('1H92!I;F1E>\"!R96QA=&EO;@HK(`D@*B`M\nM+2TM+2TM+2TM\"BL@\"2`J+PHK(`EH=%]I9'AR96P@/2!396%R8VA3>7-#86-H\nM951U<&QE*%)%3$])1\"P**R`)\"0E/8FIE8W1)9$=E=$1A='5M*&ED>')E8RT^\nM:6YD97AR96QI9\"DL(#`L(#`L(#`I.PHK(`EI9B`H(4AE87!4=7!L94ES5F%L\nM:60H:'1?:61X<F5L*2D**R`)\"65L;V<H15)23U(L(\")S>7-C86-H92!L;V]K\nM=7`@9F]R(')E;&ED(\"5D(&9A:6QE9\"(L(&ED>')E8RT^:6YD97AR96QI9\"D[\nM\"BL@\"6ED>')E;')E8R`](\"A&;W)M7W!G7V-L87-S*4=%5%-44E5#5\"AH=%]I\nM9'AR96PI.PHK(`HK(`DO*B`M+2TM+2TM+2TM\"BL@\"2`J($9E=&-H('1H92!P\nM9U]C;&%S<R!T=7!L92!O9B!T:&4@:6YD97AE9\"!R96QA=&EO;@HK(`D@*B`M\nM+2TM+2TM+2TM\"BL@\"2`J+PHK(`EH=%]I;F1R96P@/2!396%R8VA3>7-#86-H\nM951U<&QE*%)%3$])1\"P**R`)\"0E/8FIE8W1)9$=E=$1A='5M*&ED>')E8RT^\nM:6YD<F5L:60I+\"`P+\"`P+\"`P*3L**R`):68@*\"%(96%P5'5P;&5)<U9A;&ED\nM*&AT7VEN9')E;\"DI\"BL@\"0EE;&]G*$524D]2+\"`B<WES8V%C:&4@;&]O:W5P\nM(&9O<B!R96QI9\"`E9\"!F86EL960B+\"!I9'AR96,M/FEN9')E;&ED*3L**R`)\nM:6YD<F5L<F5C(#T@*$9O<FU?<&=?8VQA<W,I1T544U1254-4*&AT7VEN9')E\nM;\"D[\"BL@\"BL@\"2\\J(\"TM+2TM+2TM+2T**R`)(\"H@1V5T('1H92!A;2!N86UE\nM(&9O<B!T:&4@:6YD97@@<F5L871I;VX**R`)(\"H@+2TM+2TM+2TM+0HK(`D@\nM*B\\**R`)<W!I7V%R9W-;,%T@/2!/8FIE8W1)9$=E=$1A='5M*&ED>')E;')E\nM8RT^<F5L86TI.PHK(`ES<&E?;G5L;'-;,%T@/2`G(\"<[\"BL@\"7-P:5]N=6QL\nM<ULQ72`](\"=<,\"<[\"BL@\"7-P:7)C(#T@4U!)7V5X96-P*'!L86Y?9V5T86TL\nM('-P:5]A<F=S+\"!S<&E?;G5L;',L(#$I.PHK(`EI9B`H<W!I<F,@(3T@4U!)\nM7T]+7U-%3$5#5\"D**R`)\"65L;V<H15)23U(L(\")F86EL960@=&\\@9V5T('!G\nM7V%M('1U<&QE(&9O<B!I;F1E>\"`E<R(L(&YA;65O=70H)BAI9'AR96QR96,M\nM/G)E;&YA;64I*2D[\"BL@\"6EF(\"A34$E?<')O8V5S<V5D(\"$](#$I\"BL@\"0EE\nM;&]G*$524D]2+\"`B9F%I;&5D('1O(&=E=\"!P9U]A;2!T=7!L92!F;W(@:6YD\nM97@@)7,B+\"!N86UE;W5T*\"8H:61X<F5L<F5C+3YR96QN86UE*2DI.PHK(`ES\nM<&E?='5P(#T@4U!)7W1U<'1A8FQE+3YV86QS6S!=.PHK(`ES<&E?='1C(#T@\nM4U!)7W1U<'1A8FQE+3YT=7!D97-C.PHK(`ES<&E?9FYO(#T@4U!)7V9N=6UB\nM97(H<W!I7W1T8RP@(F%M;F%M92(I.PHK(`HK(`DO*B`M+2TM+2TM+2TM\"BL@\nM\"2`J(%-T87)T('1H92!I;F1E>\"!D969I;FET:6]N\"BL@\"2`J(\"TM+2TM+2TM\nM+2T**R`)(\"HO\"BL@\"7-P<FEN=&8H8G5F+\"`B0U)%051%(\"5S24Y$15@@)7,@\nM3TX@)7,@55-)3D<@)7,@*\"(L\"BL@\"0EI9'AR96,M/FEN9&ES=6YI<75E(#\\@\nM(E5.25%512`B(#H@(B(L\"BL@\"0EN86UE;W5T*\"8H:61X<F5L<F5C+3YR96QN\nM86UE*2DL\"BL@\"0EN86UE;W5T*\"8H:6YD<F5L<F5C+3YR96QN86UE*2DL\"BL@\nM\"0E34$E?9V5T=F%L=64H<W!I7W1U<\"P@<W!I7W1T8RP@<W!I7V9N;RDI.PHK\nM(`D)\"0HK(`DO*B`M+2TM+2TM+2TM\"BL@\"2`J($-O;&QE8W0@=&AE(&EN9&5X\nM960@871T<FEB=71E<PHK(`D@*B`M+2TM+2TM+2TM\"BL@\"2`J+PHK(`ES97`@\nM/2`B(CL**R`):V5Y8G5F6S!=(#T@)UPP)SL**R`)9F]R(\"AK97EN;R`](#`[\nM(&ME>6YO(#P@24Y$15A?34%87TM%65,[(&ME>6YO*RLI\"BL@\"7L**R`)\"6EF\nM(\"AI9'AR96,M/FEN9&ME>5MK97EN;UT@/3T@26YV86QI9$%T=').=6UB97(I\nM\"BL@\"0D)8G)E86L[\"BL@\"BL@\"0ES=')C870H:V5Y8G5F+\"!S97`I.PHK(`D)\nM<V5P(#T@(BP@(CL**R`**R`)\"2\\J(\"TM+2TM+2TM+2T**R`)\"2`J($%D9\"!T\nM:&4@:6YD97AE9\"!F:65L9\"!N86UE\"BL@\"0D@*B`M+2TM+2TM+2TM\"BL@\"0D@\nM*B\\**R`)\"6EF(\"AI9'AR96,M/FEN9&ME>5MK97EN;UT@/3T@3V)J96-T261!\nM='1R:6)U=&5.=6UB97(@+2`Q*0HK(`D)\"7-T<F-A=\"AK97EB=68L(\")O:60B\nM*3L**R`)\"65L<V4**R`)\"0ES=')C870H:V5Y8G5F+\"!G971?871T<FEB=71E\nM7VYA;64H:61X<F5C+3YI;F1R96QI9\"P**R`)\"0D)\"0EI9'AR96,M/FEN9&ME\nM>5MK97EN;UTI*3L**R`**R`)\"2\\J(\"TM+2TM+2TM+2T**R`)\"2`J($EF(&YO\nM=\"!A(&9U;F-T:6]N86P@:6YD97@L(&%D9\"!T:&4@;W!E<F%T;W(@8VQA<W,@\nM;F%M90HK(`D)(\"H@+2TM+2TM+2TM+0HK(`D)(\"HO\"BL@\"0EI9B`H:61X<F5C\nM+3YI;F1P<F]C(#T]($EN=F%L:61/:60I\"BL@\"0E[\"BL@\"0D)<W!I7V%R9W-;\nM,%T@/2!/8FIE8W1)9$=E=$1A='5M*&ED>')E8RT^:6YD8VQA<W-;:V5Y;F]=\nM*3L**R`)\"0ES<&E?;G5L;'-;,%T@/2`G(\"<[\"BL@\"0D)<W!I7VYU;&QS6S%=\nM(#T@)UPP)SL**R`)\"0ES<&ER8R`](%-025]E>&5C<\"AP;&%N7V=E=&]P8VQA\nM<W,L('-P:5]A<F=S+\"!S<&E?;G5L;',L(#$I.PHK(`D)\"6EF(\"AS<&ER8R`A\nM/2!34$E?3TM?4T5,14-4*0HK(`D)\"0EE;&]G*$524D]2+\"`B9F%I;&5D('1O\nM(&=E=\"!P9U]O<&-L87-S('1U<&QE(\"5D(BP@:61X<F5C+3YI;F1C;&%S<UMK\nM97EN;UTI.PHK(`D)\"6EF(\"A34$E?<')O8V5S<V5D(\"$](#$I\"BL@\"0D)\"65L\nM;V<H15)23U(L(\")F86EL960@=&\\@9V5T('!G7V]P8VQA<W,@='5P;&4@)60B\nM+\"!I9'AR96,M/FEN9&-L87-S6VME>6YO72D[\"BL@\"0D)<W!I7W1U<\"`](%-0\nM25]T=7!T86)L92T^=F%L<ULP73L**R`)\"0ES<&E?='1C(#T@4U!)7W1U<'1A\nM8FQE+3YT=7!D97-C.PHK(`D)\"7-P:5]F;F\\@/2!34$E?9FYU;6)E<BAS<&E?\nM='1C+\"`B;W!C;F%M92(I.PHK(`D)\"7-T<F-A=\"AK97EB=68L(\"(@(BD[\"BL@\nM\"0D)<W1R8V%T*&ME>6)U9BP@4U!)7V=E='9A;'5E*'-P:5]T=7`L('-P:5]T\nM=&,L('-P:5]F;F\\I*3L**R`)\"7T**R`)?0HK(`HK(`DO*B`M+2TM+2TM+2TM\nM\"BL@\"2`J($9O<B!F=6YC=&EO;F%L(&EN9&5X('-A>2`G9G5N8R`H871T<G,I\nM(&]P8VQA<W,G\"BL@\"2`J(\"TM+2TM+2TM+2T**R`)(\"HO\"BL@\"6EF(\"AI9'AR\nM96,M/FEN9'!R;V,@(3T@26YV86QI9$]I9\"D**R`)>PHK(`D)2&5A<%1U<&QE\nM\"7!R;V-T=7`[\"BL@\"0E&;W)M7W!G7W!R;V,)<')O8U-T<G5C=#L**R`**R`)\nM\"7!R;V-T=7`@/2!396%R8VA3>7-#86-H951U<&QE*%!23T])1\"P**R`)\"0D)\nM\"0D)\"4]B:F5C=$ED1V5T1&%T=6TH:61X<F5C+3YI;F1P<F]C*2P@,\"P@,\"P@\nM,\"D[\"BL@\"0EI9B`H(4AE87!4=7!L94ES5F%L:60H<')O8W1U<\"DI\"BL@\"0D)\nM96QO9RA%4E)/4BP@(F-A8VAE(&QO;VMU<\"!F;W(@<')O8R`E9\"!F86EL960B\nM+\"!I9'AR96,M/FEN9'!R;V,I.PHK(`HK(`D)<')O8U-T<G5C=\"`](\"A&;W)M\nM7W!G7W!R;V,I($=%5%-44E5#5\"AP<F]C='5P*3L**R`)\"7-T<F-A=\"AB=68L\nM(&YA;65O=70H)BAP<F]C4W1R=6-T+3YP<F]N86UE*2DI.PHK(`D)<W1R8V%T\nM*&)U9BP@(B`H(BD[\"BL@\"0ES=')C870H8G5F+\"!K97EB=68I.PHK(`D)<W1R\nM8V%T*&)U9BP@(BD@(BD[\"BL@\"BL@\"0ES<&E?87)G<ULP72`]($]B:F5C=$ED\nM1V5T1&%T=6TH:61X<F5C+3YI;F1C;&%S<ULP72D[\"BL@\"0ES<&E?;G5L;'-;\nM,%T@/2`G(\"<[\"BL@\"0ES<&E?;G5L;'-;,5T@/2`G7#`G.PHK(`D)<W!I<F,@\nM/2!34$E?97AE8W`H<&QA;E]G971O<&-L87-S+\"!S<&E?87)G<RP@<W!I7VYU\nM;&QS+\"`Q*3L**R`)\"6EF(\"AS<&ER8R`A/2!34$E?3TM?4T5,14-4*0HK(`D)\nM\"65L;V<H15)23U(L(\")F86EL960@=&\\@9V5T('!G7V]P8VQA<W,@='5P;&4@\nM)60B+\"!I9'AR96,M/FEN9&-L87-S6S!=*3L**R`)\"6EF(\"A34$E?<')O8V5S\nM<V5D(\"$](#$I\"BL@\"0D)96QO9RA%4E)/4BP@(F9A:6QE9\"!T;R!G970@<&=?\nM;W!C;&%S<R!T=7!L92`E9\"(L(&ED>')E8RT^:6YD8VQA<W-;,%TI.PHK(`D)\nM<W!I7W1U<\"`](%-025]T=7!T86)L92T^=F%L<ULP73L**R`)\"7-P:5]T=&,@\nM/2!34$E?='5P=&%B;&4M/G1U<&1E<V,[\"BL@\"0ES<&E?9FYO(#T@4U!)7V9N\nM=6UB97(H<W!I7W1T8RP@(F]P8VYA;64B*3L**R`)\"7-T<F-A=\"AB=68L(%-0\nM25]G971V86QU92AS<&E?='5P+\"!S<&E?='1C+\"!S<&E?9FYO*2D[\"BL@\"7T*\nM*R`)96QS90HK(`DO*B`M+2TM+2TM+2TM\"BL@\"2`J($9O<B!T:&4@;W1H97)S\nM('-A>2`G871T<B!O<&-L87-S(%LL(\"XN+ETG\"BL@\"2`J(\"TM+2TM+2TM+2T*\nM*R`)(\"HO\"BL@\"7L**R`)\"7-T<F-A=\"AB=68L(&ME>6)U9BD[\"BL@\"7T**R`*\nM*R`)+RH@+2TM+2TM+2TM+0HK(`D@*B!&:6YI<V@**R`)(\"H@+2TM+2TM+2TM\nM+0HK(`D@*B\\**R`)<W1R8V%T*&)U9BP@(BDB*3L**R`**R`)+RH@+2TM+2TM\nM+2TM+0HK(`D@*B!#<F5A=&4@=&AE(')E<W5L=\"!I;B!U<'!E<B!E>&5C=71O\nM<B!M96UO<GD**R`)(\"H@+2TM+2TM+2TM+0HK(`D@*B\\**R`);&5N(#T@<W1R\nM;&5N*&)U9BD@*R!605)(1%)36CL**R`):6YD97AD968@/2!34$E?<&%L;&]C\nM*&QE;BD[\"BL@\"59!4E-)6D4H:6YD97AD968I(#T@;&5N.PHK(`EM96UC<'DH\nM5D%21$%402AI;F1E>&1E9BDL(&)U9BP@;&5N(\"T@5D%22$124UHI.PHK(`HK\nM(`DO*B`M+2TM+2TM+2TM\"BL@\"2`J($1I<V-O;FYE8W0@9G)O;2!34$D@;6%N\nM86=E<@HK(`D@*B`M+2TM+2TM+2TM\"BL@\"2`J+PHK(`EI9B`H4U!)7V9I;FES\nM:\"@I(\"$](%-025]/2U]&24Y)4T@I\"BL@\"0EE;&]G*$524D]2+\"`B9V5T7W9I\nM97=D968Z(%-025]F:6YI<V@H*2!F86EL960B*3L**R`**R`)<F5T=7)N(&EN\nM9&5X9&5F.PHK('T**R`**R`**R`O*B`M+2TM+2TM+2TM\"BL@(\"H@9V5T7W5S\nM97)B>6ED\"0D)+2!'970@82!U<V5R(&YA;64@8GD@=7-E<WES:60@86YD\"BL@\nM(\"H)\"0D)(\"!F86QL8F%C:R!T;R`G=6YK;F]W;B`H54E$/6XI)PHK(\"`J(\"TM\nM+2TM+2TM+2T**R`@*B\\**R!.86UE1&%T82`@(\"H**R!P9U]G971?=7-E<F)Y\nM:60H:6YT-\"!U:60I\"BL@>PHK(`E(96%P5'5P;&4)=7-E<G1U<#L**R`)1F]R\nM;5]P9U]S:&%D;W<)=7-E<E]R96,[\"BL@\"4YA;65$871A\"2IR97-U;'0[\"BL@\nM\"BL@\"2\\J(\"TM+2TM+2TM+2T**R`)(\"H@06QL;V-A=&4@<W!A8V4@9F]R('1H\nM92!R97-U;'0**R`)(\"H@+2TM+2TM+2TM+0HK(`D@*B\\**R`)<F5S=6QT(#T@\nM*$YA;65$871A(\"HI('!A;&QO8RA.04U%1$%404Q%3BD[\"BL@\"6UE;7-E=\"AR\nM97-U;'0M/F1A=&$L(#`L($Y!345$051!3$5.*3L**R`**R`)+RH@+2TM+2TM\nM+2TM+0HK(`D@*B!'970@=&AE('!G7W-H861O=R!E;G1R>2!A;F0@<')I;G0@\nM=&AE(')E<W5L=`HK(`D@*B`M+2TM+2TM+2TM\"BL@\"2`J+PHK(`EU<V5R='5P\nM(#T@4V5A<F-H4WES0V%C:&54=7!L92A54T5365-)1\"P**R`)\"0E/8FIE8W1)\nM9$=E=$1A='5M*'5I9\"DL(#`L(#`L(#`I.PHK(`EI9B`H2&5A<%1U<&QE27-6\nM86QI9\"AU<V5R='5P*2D**R`)>PHK(`D)=7-E<E]R96,@/2`H1F]R;5]P9U]S\nM:&%D;W<I1T544U1254-4*'5S97)T=7`I.PHK(`D)4W1R3D-P>2AR97-U;'0M\nM/F1A=&$L(\"@F*'5S97)?<F5C+3YU<V5N86UE*2DM/F1A=&$L($Y!345$051!\nM3$5.*3L**R`)?0HK(`EE;'-E\"BL@\"7L**R`)\"7-P<FEN=&8H*&-H87(@*BER\nM97-U;'0L(\")U;FMN;W=N(\"A5240])60I(BP@=6ED*3L**R`)?0HK(`HK(`ER\nM971U<FX@<F5S=6QT.PHK('T**R`**R`**R`O*B`M+2TM+2TM+2TM\"B`@(\"H@\nM;6%K95]R=6QE9&5F\"0D)+2!R96-O;G-T<G5C=\"!T:&4@0U)%051%(%)53$4@\nM8V]M;6%N9`H@(\"`J\"0D)\"2`@9F]R(&$@9VEV96X@<&=?<F5W<FET92!T=7!L\nM90H@(\"`J(\"TM+2TM+2TM+2T**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`S,S$L,S0V\nM(\"HJ*BH*(\"`*(\"`)9FYO(#T@4U!)7V9N=6UB97(H<G5L971T8RP@(F5V7W%U\nM86PB*3L*(\"`)979?<75A;\"`](%-025]G971V86QU92AR=6QE='5P+\"!R=6QE\nM='1C+\"!F;F\\I.PHM(`EI9B`H:7-N=6QL*0HM(`D)979?<75A;\"`]($Y53$P[\nM\"B`@\"B`@\"69N;R`](%-025]F;G5M8F5R*')U;&5T=&,L(\")E=E]A8W1I;VXB\nM*3L*(\"`)979?86-T:6]N(#T@4U!)7V=E='9A;'5E*')U;&5T=7`L(')U;&5T\nM=&,L(&9N;RD[\"BT@\"6EF(\"AI<VYU;&PI\"BT@\"0EE=E]A8W1I;VX@/2!.54Q,\nM.PH@(`EI9B`H979?86-T:6]N(\"$]($Y53$PI\"B`@\"0EA8W1I;VYS(#T@*$QI\nM<W0@*BD@<W1R:6YG5&].;V1E*&5V7V%C=&EO;BD[\"B`@\"B`@\"2\\J(\"TM+2TM\nM+2TM+2T*(\"`)(\"H@0G5I;&0@=&AE(')U;&5S(&1E9FEN:71I;VX@=&5X=`H@\nM(`D@*B`M+2TM+2TM+2TM\"BTM+2`V,3<L-C(Y(\"TM+2T*(\"`*(\"`)9FYO(#T@\nM4U!)7V9N=6UB97(H<G5L971T8RP@(F5V7W%U86PB*3L*(\"`)979?<75A;\"`]\nM(%-025]G971V86QU92AR=6QE='5P+\"!R=6QE='1C+\"!F;F\\I.PH@(`H@(`EF\nM;F\\@/2!34$E?9FYU;6)E<BAR=6QE='1C+\"`B979?86-T:6]N(BD[\"B`@\"65V\nM7V%C=&EO;B`](%-025]G971V86QU92AR=6QE='5P+\"!R=6QE='1C+\"!F;F\\I\nM.PH@(`EI9B`H979?86-T:6]N(\"$]($Y53$PI\"B`@\"0EA8W1I;VYS(#T@*$QI\nM<W0@*BD@<W1R:6YG5&].;V1E*&5V7V%C=&EO;BD[\"B`@\"BL@\"B`@\"2\\J(\"TM\nM+2TM+2TM+2T*(\"`)(\"H@0G5I;&0@=&AE(')U;&5S(&1E9FEN:71I;VX@=&5X\nM=`H@(`D@*B`M+2TM+2TM+2TM\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@,SDQ+#0P\nM,B`J*BHJ\"B`@\"7L*(\"`)\"4YO9&4)(\"`@*G%U86P[\"B`@\"0E1=65R>0D@(\"`J\nM<75E<GD[\"B`@\"B`@\"0EQ=6%L(#T@<W1R:6YG5&].;V1E*&5V7W%U86PI.PH@\nM(`D)<75E<GD@/2`H475E<GD@*BD@;&9I<G-T*&%C=&EO;G,I.PH@(`H@(`D)\nM<W1R8V%T*&)U9BP@(B!72$5212`B*3L*(2`)\"7-T<F-A=\"AB=68L(&=E=%]R\nM=6QE7V5X<'(H<75E<GDM/G)T86)L92P@,\"P@<75A;\"P@5%)512DI.PH@(`E]\nM\"B`@\"B`@\"7-T<F-A=\"AB=68L(\"(@1$\\@(BD[\"BTM+2`V-S0L-C@X(\"TM+2T*\nM(\"`)>PH@(`D)3F]D90D@(\"`J<75A;#L*(\"`)\"5%U97)Y\"2`@(\"IQ=65R>3L*\nM*R`)\"5%R>4AI97(@(\"`@<6@[\"B`@\"B`@\"0EQ=6%L(#T@<W1R:6YG5&].;V1E\nM*&5V7W%U86PI.PH@(`D)<75E<GD@/2`H475E<GD@*BD@;&9I<G-T*&%C=&EO\nM;G,I.PHK(`D)<6@N<&%R96YT(#T@3E5,3#L**R`)\"7%H+G%U97)Y(\"`]('%U\nM97)Y.PH@(`H@(`D)<W1R8V%T*&)U9BP@(B!72$5212`B*3L*(2`)\"7-T<F-A\nM=\"AB=68L(&=E=%]R=6QE7V5X<'(H)G%H+\"`P+\"!Q=6%L+\"!44E5%*2D[\"B`@\nM\"7T*(\"`*(\"`)<W1R8V%T*&)U9BP@(B!$3R`B*3L**BHJ*BHJ*BHJ*BHJ*BHJ\nM\"BHJ*B`T,34L-#(Q(\"HJ*BH*(\"`)\"69O<F5A8V@H86-T:6]N+\"!A8W1I;VYS\nM*0H@(`D)>PH@(`D)\"7%U97)Y(#T@*%%U97)Y(\"HI(&QF:7)S=\"AA8W1I;VXI\nM.PHA(`D)\"7-T<F-A=\"AB=68L(&=E=%]Q=65R>5]D968H<75E<GDI*3L*(\"`)\nM\"0ES=')C870H8G5F+\"`B.R`B*3L*(\"`)\"7T*(\"`)\"7-T<F-A=\"AB=68L(\"(I\nM.R(I.PHM+2T@-S`Q+#<P-R`M+2TM\"B`@\"0EF;W)E86-H*&%C=&EO;BP@86-T\nM:6]N<RD*(\"`)\"7L*(\"`)\"0EQ=65R>2`](\"A1=65R>2`J*2!L9FER<W0H86-T\nM:6]N*3L*(2`)\"0ES=')C870H8G5F+\"!G971?<75E<GE?9&5F*'%U97)Y+\"!.\nM54Q,*2D[\"B`@\"0D)<W1R8V%T*&)U9BP@(CL@(BD[\"B`@\"0E]\"B`@\"0ES=')C\nM870H8G5F+\"`B*3LB*3L**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`T,S$L-#,W(\"HJ\nM*BH*(\"`)\"0E1=65R>0D@(\"`J<75E<GD[\"B`@\"B`@\"0D)<75E<GD@/2`H475E\nM<GD@*BD@;&9I<G-T*&%C=&EO;G,I.PHA(`D)\"7-T<F-A=\"AB=68L(&=E=%]Q\nM=65R>5]D968H<75E<GDI*3L*(\"`)\"0ES=')C870H8G5F+\"`B.R(I.PH@(`D)\nM?0H@(`E]\"BTM+2`W,3<L-S(S(\"TM+2T*(\"`)\"0E1=65R>0D@(\"`J<75E<GD[\nM\"B`@\"B`@\"0D)<75E<GD@/2`H475E<GD@*BD@;&9I<G-T*&%C=&EO;G,I.PHA\nM(`D)\"7-T<F-A=\"AB=68L(&=E=%]Q=65R>5]D968H<75E<GDL($Y53$PI*3L*\nM(\"`)\"0ES=')C870H8G5F+\"`B.R(I.PH@(`D)?0H@(`E]\"BHJ*BHJ*BHJ*BHJ\nM*BHJ*@HJ*BH@-#@R+#0Y-\"`J*BHJ\"B`@\"B`@\"69N;R`](%-025]F;G5M8F5R\nM*')U;&5T=&,L(\")E=E]Q=6%L(BD[\"B`@\"65V7W%U86P@/2!34$E?9V5T=F%L\nM=64H<G5L971U<\"P@<G5L971T8RP@9FYO*3L*+2`):68@*&ES;G5L;\"D*+2`)\nM\"65V7W%U86P@/2`B(CL*(\"`*(\"`)9FYO(#T@4U!)7V9N=6UB97(H<G5L971T\nM8RP@(F5V7V%C=&EO;B(I.PH@(`EE=E]A8W1I;VX@/2!34$E?9V5T=F%L=64H\nM<G5L971U<\"P@<G5L971T8RP@9FYO*3L*+2`):68@*&ES;G5L;\"D*+2`)\"65V\nM7V%C=&EO;B`]($Y53$P[\"B`@\"6EF(\"AE=E]A8W1I;VX@(3T@3E5,3\"D*(\"`)\nM\"6%C=&EO;G,@/2`H3&ES=\"`J*2!S=')I;F=4;TYO9&4H979?86-T:6]N*3L*\nM(\"`*+2TM(#<V.\"PW-S8@+2TM+0HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#4P,\"PU\nM,#8@*BHJ*@H@(`EI9B`H979?='EP92`A/2`G,2<@?'P@979?871T<B`^/2`P\nM('Q\\(\"%I<U]I;G-T96%D('Q\\('-T<F-M<\"AE=E]Q=6%L+\"`B(BDI\"B`@\"0ER\nM971U<FX@(DYO=\"!A('9I97<B.PH@(`HA(`ES=')C<'DH8G5F+\"!G971?<V5L\nM96-T7W%U97)Y7V1E9BAQ=65R>2DI.PH@(`ES=')C870H8G5F+\"`B.R(I.PH@\nM(`H@(`DO*B`M+2TM+2TM+2TM\"BTM+2`W.#(L-S@X(\"TM+2T*(\"`):68@*&5V\nM7W1Y<&4@(3T@)S$G('Q\\(&5V7V%T='(@/CT@,\"!\\?\"`A:7-?:6YS=&5A9\"!\\\nM?\"!S=')C;7`H979?<75A;\"P@(B(I*0H@(`D)<F5T=7)N(\").;W0@82!V:65W\nM(CL*(\"`*(2`)<W1R8W!Y*&)U9BP@9V5T7W%U97)Y7V1E9BAQ=65R>2P@3E5,\nM3\"DI.PH@(`ES=')C870H8G5F+\"`B.R(I.PH@(`H@(`DO*B`M+2TM+2TM+2TM\nM\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@-3$X+#4T,2`J*BHJ\"B`@(\"H@+2TM+2TM\nM+2TM+0H@(\"`J+PH@('-T871I8R!C:&%R(\"H*(2!G971?<75E<GE?9&5F*%%U\nM97)Y(\"IQ=65R>2D*(\"![\"B`@\"7-W:71C:\"`H<75E<GDM/F-O;6UA;F14>7!E\nM*0H@(`E[\"B`@\"0D)8V%S92!#341?4T5,14-4.@HA(`D)\"7)E='5R;B!G971?\nM<V5L96-T7W%U97)Y7V1E9BAQ=65R>2D[\"B`@\"0D)8G)E86L[\"B`@\"B`@\"0EC\nM87-E($--1%]54$1!5$4Z\"B$@\"0D)<F5T=7)N(&=E=%]U<&1A=&5?<75E<GE?\nM9&5F*'%U97)Y*3L*(\"`)\"0EB<F5A:SL*(\"`*(\"`)\"6-A<V4@0TU$7TE.4T52\nM5#H*(2`)\"0ER971U<FX@9V5T7VEN<V5R=%]Q=65R>5]D968H<75E<GDI.PH@\nM(`D)\"6)R96%K.PH@(`H@(`D)8V%S92!#341?1$5,151%.@HA(`D)\"7)E='5R\nM;B!G971?9&5L971E7W%U97)Y7V1E9BAQ=65R>2D[\"B`@\"0D)8G)E86L[\"B`@\nM\"B`@\"0EC87-E($--1%].3U1(24Y'.@HM+2T@.#`P+#@R.\"`M+2TM\"B`@(\"H@\nM+2TM+2TM+2TM+0H@(\"`J+PH@('-T871I8R!C:&%R(\"H*(2!G971?<75E<GE?\nM9&5F*%%U97)Y(\"IQ=65R>2P@47)Y2&EE<B`J<&%R96YT<6@I\"B`@>PHK(`E1\nM<GE(:65R\"0EQ:#L**R`**R`)<6@N<&%R96YT(#T@<&%R96YT<6@[\"BL@\"7%H\nM+G%U97)Y(\"`]('%U97)Y.PHK(`H@(`ES=VET8V@@*'%U97)Y+3YC;VUM86YD\nM5'EP92D*(\"`)>PH@(`D)\"6-A<V4@0TU$7U-%3$5#5#H*(2`)\"0ER971U<FX@\nM9V5T7W-E;&5C=%]Q=65R>5]D968H<75E<GDL(\"9Q:\"D[\"B`@\"0D)8G)E86L[\nM\"B`@\"B`@\"0EC87-E($--1%]54$1!5$4Z\"B$@\"0D)<F5T=7)N(&=E=%]U<&1A\nM=&5?<75E<GE?9&5F*'%U97)Y+\"`F<[email protected]@(`D)\"6)R96%K.PH@(`H@(`D)\nM8V%S92!#341?24Y315)4.@HA(`D)\"7)E='5R;B!G971?:6YS97)T7W%U97)Y\nM7V1E9BAQ=65R>2P@)G%H*3L*(\"`)\"0EB<F5A:SL*(\"`*(\"`)\"6-A<V4@0TU$\nM7T1%3$5413H*(2`)\"0ER971U<FX@9V5T7V1E;&5T95]Q=65R>5]D968H<75E\nM<GDL(\"9Q:\"D[\"B`@\"0D)8G)E86L[\"B`@\"B`@\"0EC87-E($--1%].3U1(24Y'\nM.@HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#4U-RPU-C,@*BHJ*@H@(\"`J(\"TM+2TM\nM+2TM+2T*(\"`@*B\\*(\"!S=&%T:6,@8VAA<B`J\"B$@9V5T7W-E;&5C=%]Q=65R\nM>5]D968H475E<GD@*G%U97)Y*0H@('L*(\"`)8VAA<@D)[email protected])=.PH@\nM(`EC:&%R\"2`@(\"IS97`[\"BTM+2`X-#0L.#4P(\"TM+2T*(\"`@*B`M+2TM+2TM\nM+2TM\"B`@(\"HO\"B`@<W1A=&EC(&-H87(@*@HA(&=E=%]S96QE8W1?<75E<GE?\nM9&5F*%%U97)Y(\"IQ=65R>2P@47)Y2&EE<B`J<6@I\"B`@>PH@(`EC:&%R\"0EB\nM=69;.#$Y,ET[\"B`@\"6-H87()(\"`@*G-E<#L**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ\nM*B`V,S4L-C0Q(\"HJ*BH*(\"`)\"7-T<F-A=\"AB=68L('-E<\"D[\"B`@\"0ES97`@\nM/2`B+\"`B.PH@(`HA(`D)<W1R8V%T*&)U9BP@9V5T7W1L95]E>'!R*'%U97)Y\nM+3YR=&%B;&4L(#`L('1L92P@*')T7VYU;75S960@/B`Q*2DI.PH@(`H@(`D)\nM+RH@0VAE8VL@:68@=V4@;75S=\"!S87D@05,@+BXN(\"HO\"B`@\"0EI9B`H;F]D\nM951A9RAT;&4M/F5X<'(I(\"$](%1?5F%R*[email protected](R+#DR.\"`M+2TM\"B`@\nM\"0ES=')C870H8G5F+\"!S97`I.PH@(`D)<V5P(#T@(BP@(CL*(\"`*(2`)\"7-T\nM<F-A=\"AB=68L(&=E=%]T;&5?97AP<BAQ:\"P@,\"P@=&QE+\"`H<G1?;G5M=7-E\nM9\"`^(#$I*2D[\"B`@\"B`@\"0DO*B!#:&5C:R!I9B!W92!M=7-T('-A>2!!4R`N\nM+BX@*B\\*(\"`)\"6EF(\"AN;V1E5&%G*'1L92T^97AP<BD@(3T@5%]687(I\"BHJ\nM*BHJ*BHJ*BHJ*BHJ*@HJ*BH@-C@Q+#8X-R`J*BHJ\"B`@\"0D)\"7-T<F-A=\"AB\nM=68L('-E<\"D[\"B`@\"0D)\"7-E<\"`](\"(L(\"([\"B`@\"0D)\"7-T<F-A=\"AB=68L\nM(')T92T^<F5L;F%M92D[\"B$@\"0D)\"6EF(\"AR=%]N=6UU<V5D(#X@,2D*(\"`)\nM\"0D)>PH@(`D)\"0D)<W1R8V%T*&)U9BP@(B`B*3L*(\"`)\"0D)\"7-T<F-A=\"AB\nM=68L(')T92T^<F5F;F%M92D[\"BTM+2`[email protected]<T(\"TM+2T*(\"`)\"0D)<W1R\nM8V%T*&)U9BP@<V5P*3L*(\"`)\"0D)<V5P(#T@(BP@(CL*(\"`)\"0D)<W1R8V%T\nM*&)U9BP@<G1E+3YR96QN86UE*3L*(2`)\"0D):68@*'-T<F-M<\"AR=&4M/G)E\nM;&YA;64L(')T92T^<F5F;F%M92D@(3T@,\"D*(\"`)\"0D)>PH@(`D)\"0D)<W1R\nM8V%T*&)U9BP@(B`B*3L*(\"`)\"0D)\"7-T<F-A=\"AB=68L(')T92T^<F5F;F%M\nM92D[\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@-CDT+#<P,\"`J*BHJ\"B`@\"6EF(\"AQ\nM=65R>2T^<75A;\"`A/2!.54Q,*0H@(`E[\"B`@\"0ES=')C870H8G5F+\"`B(%=(\nM15)%(\"(I.PHA(`D)<W1R8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAQ=65R>2T^\nM<G1A8FQE+\"`P+\"!Q=65R>2T^<75A;\"P@*')T7VYU;75S960@/B`Q*2DI.PH@\nM(`E]\"B`@\"B`@\"2\\J($%D9\"!T:&4@1U)/55`@0ED@0TQ!55-%(\"HO\"BTM+2`Y\nM.#$L.3@W(\"TM+2T*(\"`):68@*'%U97)Y+3YQ=6%L(\"$]($Y53$PI\"B`@\"7L*\nM(\"`)\"7-T<F-A=\"AB=68L(\"(@5TA%4D4@(BD[\"B$@\"0ES=')C870H8G5F+\"!G\nM971?<G5L95]E>'!R*'%H+\"`P+\"!Q=65R>2T^<75A;\"P@*')T7VYU;75S960@\nM/B`Q*2DI.PH@(`E]\"B`@\"B`@\"2\\J($%D9\"!T:&4@1U)/55`@0ED@0TQ!55-%\nM(\"HO\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@-S`V+#<Q,B`J*BHJ\"B`@\"0E[\"B`@\nM\"0D)<W1R8V%T*&)U9BP@<V5P*3L*(\"`)\"0ES97`@/2`B+\"`B.PHA(`D)\"7-T\nM<F-A=\"AB=68L(&=E=%]R=6QE7V5X<'(H<75E<GDM/G)T86)L92P@,\"P@;&9I\nM<G-T*&PI+\"`H<G1?;G5M=7-E9\"`^(#$I*2D[\"B`@\"0E]\"B`@\"7T*(\"`*+2TM\nM(#DY,RPY.3D@+2TM+0H@(`D)>PH@(`D)\"7-T<F-A=\"AB=68L('-E<\"D[\"B`@\nM\"0D)<V5P(#T@(BP@(CL*(2`)\"0ES=')C870H8G5F+\"!G971?<G5L95]E>'!R\nM*'%H+\"`P+\"!L9FER<W0H;\"DL(\"AR=%]N=6UU<V5D(#X@,2DI*3L*(\"`)\"7T*\nM(\"`)?0H@(`HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#<R,RPW,CD@*BHJ*@H@(\"`J\nM(\"TM+2TM+2TM+2T*(\"`@*B\\*(\"!S=&%T:6,@8VAA<B`J\"B$@9V5T7VEN<V5R\nM=%]Q=65R>5]D968H475E<GD@*G%U97)Y*0H@('L*(\"`)8VAA<@D)8G5F6S@Q\nM.3)=.PH@(`EC:&%R\"2`@(\"IS97`[\"BTM+2`Q,#$P+#$P,38@+2TM+0H@(\"`J\nM(\"TM+2TM+2TM+2T*(\"`@*B\\*(\"!S=&%T:6,@8VAA<B`J\"B$@9V5T7VEN<V5R\nM=%]Q=65R>5]D968H475E<GD@*G%U97)Y+\"!1<GE(:65R(\"IQ:\"D*(\"![\"B`@\nM\"6-H87()\"6)U9ELX,3DR73L*(\"`)8VAA<@D@(\"`J<V5P.PHJ*BHJ*BHJ*BHJ\nM*BHJ*BH**BHJ(#@Q,\"PX,C$@*BHJ*@H@(`H@(`D)\"7-T<F-A=\"AB=68L('-E\nM<\"D[\"B`@\"0D)<V5P(#T@(BP@(CL*(2`)\"0ES=')C870H8G5F+\"!G971?=&QE\nM7V5X<'(H<75E<GDM/G)T86)L92P@,\"P@=&QE+\"`H<G1?;G5M=7-E9\"`^(#$I\nM*2D[\"B`@\"0E]\"B`@\"0ES=')C870H8G5F+\"`B*2(I.PH@(`E]\"B`@\"65L<V4*\nM(2`)\"7-T<F-A=\"AB=68L(&=E=%]S96QE8W1?<75E<GE?9&5F*'%U97)Y*2D[\nM\"B`@\"B`@\"2\\J(\"TM+2TM+2TM+2T*(\"`)(\"H@0V]P>2!T:&4@<75E<GD@<W1R\nM:6YG(&EN=&\\@86QL;V-A=&5D('-P86-E(&%N9\"!R971U<FX@:70*+2TM(#$P\nM.3<L,3$P.\"`M+2TM\"B`@\"B`@\"0D)<W1R8V%T*&)U9BP@<V5P*3L*(\"`)\"0ES\nM97`@/2`B+\"`B.PHA(`D)\"7-T<F-A=\"AB=68L(&=E=%]T;&5?97AP<BAQ:\"P@\nM,\"P@=&QE+\"`H<G1?;G5M=7-E9\"`^(#$I*2D[\"B`@\"0E]\"B`@\"0ES=')C870H\nM8G5F+\"`B*2(I.PH@(`E]\"B`@\"65L<V4*(2`)\"7-T<F-A=\"AB=68L(&=E=%]Q\nM=65R>5]D968H<75E<GDL('%H*2D[\"B`@\"B`@\"2\\J(\"TM+2TM+2TM+2T*(\"`)\nM(\"H@0V]P>2!T:&4@<75E<GD@<W1R:6YG(&EN=&\\@86QL;V-A=&5D('-P86-E\nM(&%N9\"!R971U<FX@:70**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`X,S`L.#,V(\"HJ\nM*BH*(\"`@*B`M+2TM+2TM+2TM\"B`@(\"HO\"B`@<W1A=&EC(&-H87(@*@HA(&=E\nM=%]U<&1A=&5?<75E<GE?9&5F*%%U97)Y(\"IQ=65R>2D*(\"![\"B`@\"6-H87()\nM\"6)U9ELX,3DR73L*(\"`)8VAA<@D@(\"`J<V5P.PHM+2T@,3$Q-RPQ,3(S(\"TM\nM+2T*(\"`@*B`M+2TM+2TM+2TM\"B`@(\"HO\"B`@<W1A=&EC(&-H87(@*@HA(&=E\nM=%]U<&1A=&5?<75E<GE?9&5F*%%U97)Y(\"IQ=65R>2P@47)Y2&EE<B`J<6@I\nM\"B`@>PH@(`EC:&%R\"0EB=69;.#$Y,ET[\"B`@\"6-H87()(\"`@*G-E<#L**BHJ\nM*BHJ*BHJ*BHJ*BHJ\"BHJ*B`X-3<L.#8S(\"HJ*BH*(\"`)\"7-E<\"`](\"(L(\"([\nM\"B`@\"0ES=')C870H8G5F+\"!T;&4M/G)E<V1O;2T^<F5S;F%M92D[\"B`@\"0ES\nM=')C870H8G5F+\"`B(#T@(BD[\"B$@\"0ES=')C870H8G5F+\"!G971?=&QE7V5X\nM<'(H<75E<GDM/G)T86)L92P@<75E<GDM/G)E<W5L=%)E;&%T:6]N+`H@(`D)\nM\"0D)\"0D)('1L92P@5%)512DI.PH@(`E]\"B`@\"BTM+2`Q,30T+#$Q-3`@+2TM\nM+0H@(`D)<V5P(#T@(BP@(CL*(\"`)\"7-T<F-A=\"AB=68L('1L92T^<F5S9&]M\nM+3YR97-N86UE*3L*(\"`)\"7-T<F-A=\"AB=68L(\"(@/2`B*3L*(2`)\"7-T<F-A\nM=\"AB=68L(&=E=%]T;&5?97AP<BAQ:\"P@<75E<GDM/G)E<W5L=%)E;&%T:6]N\nM+`H@(`D)\"0D)\"0D)('1L92P@5%)512DI.PH@(`E]\"B`@\"BHJ*BHJ*BHJ*BHJ\nM*BHJ*@HJ*BH@.#8U+#@W,2`J*BHJ\"B`@\"6EF(\"AQ=65R>2T^<75A;\"`A/2!.\nM54Q,*0H@(`E[\"B`@\"0ES=')C870H8G5F+\"`B(%=(15)%(\"(I.PHA(`D)<W1R\nM8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAQ=65R>2T^<G1A8FQE+\"!Q=65R>2T^\nM<F5S=6QT4F5L871I;VXL\"B`@\"0D)\"0D)\"0D@('%U97)Y+3YQ=6%L+\"!44E5%\nM*2D[\"B`@\"7T*(\"`*+2TM(#$Q-3(L,3$U.\"`M+2TM\"B`@\"6EF(\"AQ=65R>2T^\nM<75A;\"`A/2!.54Q,*0H@(`E[\"B`@\"0ES=')C870H8G5F+\"`B(%=(15)%(\"(I\nM.PHA(`D)<W1R8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAQ:\"P@<75E<GDM/G)E\nM<W5L=%)E;&%T:6]N+`H@(`D)\"0D)\"0D)(\"!Q=65R>2T^<75A;\"P@5%)512DI\nM.PH@(`E]\"B`@\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@.#@R+#@X.\"`J*BHJ\"B`@\nM(\"H@+2TM+2TM+2TM+0H@(\"`J+PH@('-T871I8R!C:&%R(\"H*(2!G971?9&5L\nM971E7W%U97)Y7V1E9BA1=65R>2`J<75E<GDI\"B`@>PH@(`EC:&%R\"0EB=69;\nM.#$Y,ET[\"B`@\"5)A;F=E5&)L16YT<GD@*G)T93L*+2TM(#$Q-CDL,3$W-2`M\nM+2TM\"B`@(\"H@+2TM+2TM+2TM+0H@(\"`J+PH@('-T871I8R!C:&%R(\"H*(2!G\nM971?9&5L971E7W%U97)Y7V1E9BA1=65R>2`J<75E<GDL(%%R>4AI97(@*G%H\nM*0H@('L*(\"`)8VAA<@D)[email protected])=.PH@(`E286YG951B;$5N=')Y(\"IR\nM=&4[\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@.#DY+#DP-2`J*BHJ\"B`@\"6EF(\"AQ\nM=65R>2T^<75A;\"`A/2!.54Q,*0H@(`E[\"B`@\"0ES=')C870H8G5F+\"`B(%=(\nM15)%(\"(I.PHA(`D)<W1R8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAQ=65R>2T^\nM<G1A8FQE+\"`P+\"!Q=65R>2T^<75A;\"P@1D%,4T4I*3L*(\"`)?0H@(`H@(`DO\nM*B`M+2TM+2TM+2TM\"BTM+2`Q,3@V+#$Q.3(@+2TM+0H@(`EI9B`H<75E<GDM\nM/G%U86P@(3T@3E5,3\"D*(\"`)>PH@(`D)<W1R8V%T*&)U9BP@(B!72$5212`B\nM*3L*(2`)\"7-T<F-A=\"AB=68L(&=E=%]R=6QE7V5X<'(H<6@L(#`L('%U97)Y\nM+3YQ=6%L+\"!&04Q312DI.PH@(`E]\"B`@\"B`@\"2\\J(\"TM+2TM+2TM+2T**BHJ\nM*BHJ*BHJ*BHJ*BHJ\"BHJ*B`Y,34L.3(Q(\"HJ*BH*(\"`@*B`M+2TM+2TM+2TM\nM\"B`@(\"HO\"B`@<W1A=&EC(&-H87(@*@HA(&=E=%]R=6QE7V5X<'(H3&ES=\"`J\nM<G1A8FQE+\"!I;G0@<G1?:6YD97@L($YO9&4@*FYO9&4L(&)O;VP@=F%R<')E\nM9FEX*0H@('L*(\"`)8VAA<@D)[email protected])=.PH@(`HM+2T@,3(P,BPQ,C`X\nM(\"TM+2T*(\"`@*B`M+2TM+2TM+2TM\"B`@(\"HO\"B`@<W1A=&EC(&-H87(@*@HA\nM(&=E=%]R=6QE7V5X<'(H47)Y2&EE<B`J<6@L(&EN=\"!R=%]I;F1E>\"P@3F]D\nM92`J;F]D92P@8F]O;\"!V87)P<F5F:7@I\"B`@>PH@(`EC:&%R\"0EB=69;.#$Y\nM,ET[\"B`@\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*[email protected],V+#DT,B`J*BHJ\"B`@\"0D)\nM>PH@(`D)\"0E487)G971%;G1R>2`J=&QE(#T@*%1A<F=E=$5N=')Y(\"HI(&YO\nM9&4[\"B`@\"B$@\"0D)\"7)E='5R;B!G971?<G5L95]E>'!R*')T86)L92P@<G1?\nM:6YD97@L\"B`@\"0D)\"0D)\"0D)(\"A.;V1E(\"HI(\"AT;&4M/F5X<'(I+\"!V87)P\nM<F5F:[email protected]@(`D)\"7T*(\"`)\"0EB<F5A:SL*+2TM(#$R,C,L,3(R.2`M+2TM\nM\"B`@\"0D)>PH@(`D)\"0E487)G971%;G1R>2`J=&QE(#T@*%1A<F=E=$5N=')Y\nM(\"HI(&YO9&4[\"B`@\"B$@\"0D)\"7)E='5R;B!G971?<G5L95]E>'!R*'%H+\"!R\nM=%]I;F1E>\"P*(\"`)\"0D)\"0D)\"0D@*$YO9&4@*BD@*'1L92T^97AP<BDL('9A\nM<G!R969I>\"D[\"B`@\"0D)?0H@(`D)\"6)R96%K.PHJ*BHJ*BHJ*BHJ*BHJ*BH*\nM*BHJ(#DT-RPY-3,@*BHJ*@H@(`H@(`D)\"0ES=')C870H8G5F+\"!A9V<M/F%G\nM9VYA;64I.PH@(`D)\"0ES=')C870H8G5F+\"`B*\"(I.PHA(`D)\"0ES=')C870H\nM8G5F+\"!G971?<G5L95]E>'!R*')T86)L92P@<G1?:6YD97@L\"B`@\"0D)\"0D)\nM\"0D)(\"A.;V1E(\"HI(\"AA9V<M/G1A<F=E=\"DL('9A<G!R969I>\"DI.PH@(`D)\nM\"0ES=')C870H8G5F+\"`B*2(I.PH@(`D)\"0ER971U<FX@<'-T<F1U<\"AB=68I\nM.PHM+2T@,3(S-\"PQ,C0P(\"TM+2T*(\"`*(\"`)\"0D)<W1R8V%T*&)U9BP@86=G\nM+3YA9V=N86UE*3L*(\"`)\"0D)<W1R8V%T*&)U9BP@(B@B*3L*(2`)\"0D)<W1R\nM8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAQ:\"P@<G1?:6YD97@L\"B`@\"0D)\"0D)\nM\"0D)(\"A.;V1E(\"HI(\"AA9V<M/G1A<F=E=\"DL('9A<G!R969I>\"DI.PH@(`D)\nM\"0ES=')C870H8G5F+\"`B*2(I.PH@(`D)\"0ER971U<FX@<'-T<F1U<\"AB=68I\nM.PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#DU.\"PY-C0@*BHJ*@H@(`D)\"7L*(\"`)\nM\"0D)1W)O=7!#;&%U<V4@*F=R<\"`](\"A'<F]U<$-L875S92`J*2!N;V1E.PH@\nM(`HA(`D)\"0ER971U<FX@9V5T7W)U;&5?97AP<BAR=&%B;&4L(')T7VEN9&5X\nM+`H@(`D)\"0D)\"0D)\"2`H3F]D92`J*2`H9W)P+3YE;G1R>2DL('9A<G!R969I\nM>\"D[\"B`@\"0D)?0H@(`D)\"6)R96%K.PHM+2T@,3(T-2PQ,C4Q(\"TM+2T*(\"`)\nM\"0E[\"B`@\"0D)\"4=R;W5P0VQA=7-E(\"IG<G`@/2`H1W)O=7!#;&%U<V4@*BD@\nM;F]D93L*(\"`*(2`)\"0D)<F5T=7)N(&=E=%]R=6QE7V5X<'(H<6@L(')T7VEN\nM9&5X+`H@(`D)\"0D)\"0D)\"2`H3F]D92`J*2`H9W)P+3YE;G1R>2DL('9A<G!R\nM969I>\"D[\"B`@\"0D)?0H@(`D)\"6)R96%K.PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ\nM(#DW-\"PY.#8@*BHJ*@H@(`D)\"0ES=VET8V@@*&5X<'(M/F]P5'EP92D*(\"`)\nM\"0D)>PH@(`D)\"0D)8V%S92!/4%]%6%!2.@HA(`D)\"0D)\"7-T<F-A=\"AB=68L\nM(&=E=%]R=6QE7V5X<'(H<G1A8FQE+\"!R=%]I;F1E>\"P*(\"`)\"0D)\"0D)\"0D)\nM\"2`@(\"A.;V1E(\"HI(&=E=%]L969T;W`H97AP<BDL\"B`@\"0D)\"0D)\"0D)\"0D)\nM(\"!V87)P<F5F:7@I*3L*(\"`)\"0D)\"0ES=')C870H8G5F+\"`B(\"(I.PH@(`D)\nM\"0D)\"7-T<F-A=\"AB=68L(&=E=%]O<&YA;64H*\"A/<&5R(\"HI(&5X<'(M/F]P\nM97(I+3YO<&YO*2D[\"B`@\"0D)\"0D)<W1R8V%T*&)U9BP@(B`B*3L*(2`)\"0D)\nM\"0ES=')C870H8G5F+\"!G971?<G5L95]E>'!R*')T86)L92P@<G1?:6YD97@L\nM\"B`@\"0D)\"0D)\"0D)\"0D@(\"A.;V1E(\"HI(&=E=%]R:6=H=&]P*&5X<'(I+`H@\nM(`D)\"0D)\"0D)\"0D)\"2`@=F%R<')E9FEX*2D[\"B`@\"0D)\"0D)<F5T=7)N('!S\nM=')D=7`H8G5F*3L*+2TM(#$R-C$L,3(W,R`M+2TM\"B`@\"0D)\"7-W:71C:\"`H\nM97AP<BT^;W!4>7!E*0H@(`D)\"0E[\"B`@\"0D)\"0EC87-E($]07T584%(Z\"B$@\nM\"0D)\"0D)<W1R8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAQ:\"P@<G1?:6YD97@L\nM\"B`@\"0D)\"0D)\"0D)\"0D@(\"`H3F]D92`J*2!G971?;&5F=&]P*&5X<'(I+`H@\nM(`D)\"0D)\"0D)\"0D)\"2`@=F%R<')E9FEX*2D[\"B`@\"0D)\"0D)<W1R8V%T*&)U\nM9BP@(B`B*3L*(\"`)\"0D)\"0ES=')C870H8G5F+\"!G971?;W!N86UE*\"@H3W!E\nM<B`J*2!E>'!R+3YO<&5R*2T^;W!N;RDI.PH@(`D)\"0D)\"7-T<F-A=\"AB=68L\nM(\"(@(BD[\"B$@\"0D)\"0D)<W1R8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAQ:\"P@\nM<G1?:6YD97@L\"B`@\"0D)\"0D)\"0D)\"0D@(\"A.;V1E(\"HI(&=E=%]R:6=H=&]P\nM*&5X<'(I+`H@(`D)\"0D)\"0D)\"0D)\"2`@=F%R<')E9FEX*2D[\"B`@\"0D)\"0D)\nM<F5T=7)N('!S=')D=7`H8G5F*3L**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`Y.#@L\nM.3DX(\"HJ*BH*(\"`*(\"`)\"0D)\"6-A<V4@3U)?15A04CH*(\"`)\"0D)\"0ES=')C\nM870H8G5F+\"`B*\"(I.PHA(`D)\"0D)\"7-T<F-A=\"AB=68L(&=E=%]R=6QE7V5X\nM<'(H<G1A8FQE+\"!R=%]I;F1E>\"P*(\"`)\"0D)\"0D)\"0D)\"2`@(\"A.;V1E(\"HI\nM(&=E=%]L969T;W`H97AP<BDL\"B`@\"0D)\"0D)\"0D)\"0D)(\"!V87)P<F5F:7@I\nM*3L*(\"`)\"0D)\"0ES=')C870H8G5F+\"`B*2!/4B`H(BD[\"B$@\"0D)\"0D)<W1R\nM8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAR=&%B;&4L(')T7VEN9&5X+`H@(`D)\nM\"0D)\"0D)\"0D)(\"`H3F]D92`J*2!G971?<FEG:'1O<\"AE>'!R*2P*(\"`)\"0D)\nM\"0D)\"0D)\"0D@('9A<G!R969I>\"DI.PH@(`D)\"0D)\"7-T<F-A=\"AB=68L(\"(I\nM(BD[\"BTM+2`Q,C<U+#$R.#4@+2TM+0H@(`H@(`D)\"0D)8V%S92!/4E]%6%!2\nM.@H@(`D)\"0D)\"7-T<F-A=\"AB=68L(\"(H(BD[\"B$@\"0D)\"0D)<W1R8V%T*&)U\nM9BP@9V5T7W)U;&5?97AP<BAQ:\"P@<G1?:6YD97@L\"B`@\"0D)\"0D)\"0D)\"0D@\nM(\"`H3F]D92`J*2!G971?;&5F=&]P*&5X<'(I+`H@(`D)\"0D)\"0D)\"0D)\"2`@\nM=F%R<')E9FEX*2D[\"B`@\"0D)\"0D)<W1R8V%T*&)U9BP@(BD@3U(@*\"(I.PHA\nM(`D)\"0D)\"7-T<F-A=\"AB=68L(&=E=%]R=6QE7V5X<'(H<6@L(')T7VEN9&5X\nM+`H@(`D)\"0D)\"0D)\"0D)(\"`H3F]D92`J*2!G971?<FEG:'1O<\"AE>'!R*2P*\nM(\"`)\"0D)\"0D)\"0D)\"0D@('9A<G!R969I>\"DI.PH@(`D)\"0D)\"7-T<F-A=\"AB\nM=68L(\"(I(BD[\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@,3`P,2PQ,#$Q(\"HJ*BH*\nM(\"`*(\"`)\"0D)\"6-A<V4@04Y$7T584%(Z\"B`@\"0D)\"0D)<W1R8V%T*&)U9BP@\nM(B@B*3L*(2`)\"0D)\"0ES=')C870H8G5F+\"!G971?<G5L95]E>'!R*')T86)L\nM92P@<G1?:6YD97@L\"B`@\"0D)\"0D)\"0D)\"0D@(\"`H3F]D92`J*2!G971?;&5F\nM=&]P*&5X<'(I+`H@(`D)\"0D)\"0D)\"0D)\"2`@=F%R<')E9FEX*2D[\"B`@\"0D)\nM\"0D)<W1R8V%T*&)U9BP@(BD@04Y$(\"@B*3L*(2`)\"0D)\"0ES=')C870H8G5F\nM+\"!G971?<G5L95]E>'!R*')T86)L92P@<G1?:6YD97@L\"B`@\"0D)\"0D)\"0D)\nM\"0D@(\"A.;V1E(\"HI(&=E=%]R:6=H=&]P*&5X<'(I+`H@(`D)\"0D)\"0D)\"0D)\nM\"2`@=F%R<')E9FEX*2D[\"B`@\"0D)\"0D)<W1R8V%T*&)U9BP@(BDB*3L*+2TM\nM(#$R.#@L,3(Y.\"`M+2TM\"B`@\"B`@\"0D)\"0EC87-E($%.1%]%6%!2.@H@(`D)\nM\"0D)\"7-T<F-A=\"AB=68L(\"(H(BD[\"B$@\"0D)\"0D)<W1R8V%T*&)U9BP@9V5T\nM7W)U;&5?97AP<BAQ:\"P@<G1?:6YD97@L\"B`@\"0D)\"0D)\"0D)\"0D@(\"`H3F]D\nM92`J*2!G971?;&5F=&]P*&5X<'(I+`H@(`D)\"0D)\"0D)\"0D)\"2`@=F%R<')E\nM9FEX*2D[\"B`@\"0D)\"0D)<W1R8V%T*&)U9BP@(BD@04Y$(\"@B*3L*(2`)\"0D)\nM\"0ES=')C870H8G5F+\"!G971?<G5L95]E>'!R*'%H+\"!R=%]I;F1E>\"P*(\"`)\nM\"0D)\"0D)\"0D)\"2`@*$YO9&4@*BD@9V5T7W)I9VAT;W`H97AP<BDL\"B`@\"0D)\nM\"0D)\"0D)\"0D)(\"!V87)P<F5F:7@I*3L*(\"`)\"0D)\"0ES=')C870H8G5F+\"`B\nM*2(I.PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#$P,30L,3`R,\"`J*BHJ\"B`@\"B`@\nM\"0D)\"0EC87-E($Y/5%]%6%!2.@H@(`D)\"0D)\"7-T<F-A=\"AB=68L(\").3U0@\nM*\"(I.PHA(`D)\"0D)\"7-T<F-A=\"AB=68L(&=E=%]R=6QE7V5X<'(H<G1A8FQE\nM+\"!R=%]I;F1E>\"P*(\"`)\"0D)\"0D)\"0D)\"2`@(\"A.;V1E(\"HI(&=E=%]L969T\nM;W`H97AP<BDL\"B`@\"0D)\"0D)\"0D)\"0D)(\"!V87)P<F5F:7@I*3L*(\"`)\"0D)\nM\"0ES=')C870H8G5F+\"`B*2(I.PHM+2T@,3,P,2PQ,S`W(\"TM+2T*(\"`*(\"`)\nM\"0D)\"6-A<V4@3D]47T584%(Z\"B`@\"0D)\"0D)<W1R8V%T*&)U9BP@(DY/5\"`H\nM(BD[\"B$@\"0D)\"0D)<W1R8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAQ:\"P@<G1?\nM:6YD97@L\"B`@\"0D)\"0D)\"0D)\"0D@(\"`H3F]D92`J*2!G971?;&5F=&]P*&5X\nM<'(I+`H@(`D)\"0D)\"0D)\"0D)\"2`@=F%R<')E9FEX*2D[\"B`@\"0D)\"0D)<W1R\nM8V%T*&)U9BP@(BDB*3L**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`Q,#(R+#$P,C@@\nM*BHJ*@H@(`D)\"0D)\"6)R96%K.PH@(`H@(`D)\"0D)8V%S92!&54Y#7T584%(Z\nM\"B$@\"0D)\"0D)<F5T=7)N(&=E=%]F=6YC7V5X<'(H<G1A8FQE+\"!R=%]I;F1E\nM>\"P*(\"`)\"0D)\"0D)\"0D)\"2`H17AP<B`J*2!N;V1E+`H@(`D)\"0D)\"0D)\"0D)\nM('9A<G!R969I>\"D[\"B`@\"0D)\"0D)8G)E86L[\"BTM+2`Q,S`Y+#$S,34@+2TM\nM+0H@(`D)\"0D)\"6)R96%K.PH@(`H@(`D)\"0D)8V%S92!&54Y#7T584%(Z\"B$@\nM\"0D)\"0D)<F5T=7)N(&=E=%]F=6YC7V5X<'(H<6@L(')T7VEN9&5X+`H@(`D)\nM\"0D)\"0D)\"0D)(\"A%>'!R(\"HI(&YO9&4L\"B`@\"0D)\"0D)\"0D)\"0D@=F%R<')E\nM9FEX*3L*(\"`)\"0D)\"0EB<F5A:SL**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ*B`Q,#,W\nM+#$P-#,@*BHJ*@H@(`D)8V%S92!47U9A<CH*(\"`)\"0E[\"B`@\"0D)\"59A<@D)\nM(\"`@*G9A<B`](\"A687(@*BD@;F]D93L*(2`)\"0D)4F%N9V548FQ%;G1R>2`J\nM<G1E(#T@*%)A;F=E5&)L16YT<GD@*BD@;G1H*'9A<BT^=F%R;F\\@+2`Q+\"!R\nM=&%B;&4I.PH@(`H@(`D)\"0EI9B`H(7-T<F-M<\"AR=&4M/G)E9FYA;64L(\"(J\nM3D57*B(I*0H@(`D)\"0D)<W1R8V%T*&)U9BP@(FYE=RXB*3L*+2TM(#$S,C0L\nM,3,S-R`M+2TM\"B`@\"0EC87-E(%1?5F%R.@H@(`D)\"7L*(\"`)\"0D)5F%R\"0D@\nM(\"`J=F%R(#T@*%9A<B`J*2!N;V1E.PHA(`D)\"0E286YG951B;$5N=')Y(\"IR\nM=&4[\"B$@\"0D)\"6EN=\"!S=7`@/2!V87(M/G9A<FQE=F5L<W5P.PHA(`HA(`D)\nM\"0EW:&EL92AS=7`M+2`^(#`I('%H(#T@<6@M/G!A<F5N=#L*(2`)\"0D)<G1E\nM(#T@*%)A;F=E5&)L16YT<GD@*BD@;G1H*'9A<BT^=F%R;F\\@+2`Q+\"!Q:\"T^\nM<75E<GDM/G)T86)L92D[\"B$@\"B$@\"0D)\"6EF(\"AQ:\"T^<&%R96YT(#T]($Y5\nM3$P@)B8@=F%R+3YV87)L979E;'-U<\"`^(#`I\"B$@\"0D)\"0ER=&4@/2`H4F%N\nM9V548FQ%;G1R>2`J*2!N=&@H=F%R+3YV87)N;R`K(#$L('%H+3YQ=65R>2T^\nM<G1A8FQE*3L*(\"`*(\"`)\"0D):68@*\"%S=')C;7`H<G1E+3YR969N86UE+\"`B\nM*DY%5RHB*2D*(\"`)\"0D)\"7-T<F-A=\"AB=68L(\")N97<N(BD[\"BHJ*BHJ*BHJ\nM*BHJ*BHJ*@HJ*BH@,3`T-RPQ,#4S(\"HJ*BH*(\"`)\"0D)\"0ES=')C870H8G5F\nM+\"`B8W5R<F5N=\"XB*3L*(\"`)\"0D)\"65L<V4*(\"`)\"0D)\"7L*(2`)\"0D)\"0EI\nM9B`H=F%R<')E9FEX(\"8F('9A<BT^=F%R;F\\@(3T@<G1?:6YD97@I\"B`@\"0D)\nM\"0D)>PH@(`D)\"0D)\"0ES=')C870H8G5F+\"!R=&4M/G)E9FYA;64I.PH@(`D)\nM\"0D)\"0ES=')C870H8G5F+\"`B+B(I.PHM+2T@,3,T,2PQ,S0W(\"TM+2T*(\"`)\nM\"0D)\"0ES=')C870H8G5F+\"`B8W5R<F5N=\"XB*3L*(\"`)\"0D)\"65L<V4*(\"`)\nM\"0D)\"7L*(2`)\"0D)\"0EI9B`H<W1R8VUP*')T92T^<F5L;F%M92P@<G1E+3YR\nM969N86UE*2`A/2`P*0H@(`D)\"0D)\"7L*(\"`)\"0D)\"0D)<W1R8V%T*&)U9BP@\nM<G1E+3YR969N86UE*3L*(\"`)\"0D)\"0D)<W1R8V%T*&)U9BP@(BXB*3L**BHJ\nM*BHJ*BHJ*BHJ*BHJ\"BHJ*B`Q,#8Y+#$P.3@@*BHJ*@H@(`H@(`D)8V%S92!4\nM7U-U8DQI;FLZ\"B`@\"0D)>PHA(`D)\"0E3=6),:6YK(\"`@(\"IS=6)L:6YK(#T@\nM*%-U8DQI;FL@*BD@;F]D93L*(2`)\"0D)475E<GD)(\"`@*G%U97)Y(#T@*%%U\nM97)Y(\"HI(\"AS=6)L:6YK+3YS=6)S96QE8W0I.PHA(`D)\"0E,:7-T\"2`@(\"IL\nM.PHA(`D)\"0EC:&%R\"2`@(\"IS97`[\"B$@\"B$@\"0D)\"6EF(\"AS=6)L:6YK+3YL\nM969T:&%N9\"`A/2!.54Q,*0HA(`D)\"0E[\"B$@\"0D)\"0ES=')C870H8G5F+\"`B\nM*\"(I.PHA(`D)\"0D)<V5P(#T@(B([\"B$@\"0D)\"0EF;W)E86-H*&PL('-U8FQI\nM;FLM/FQE9G1H86YD*0HA(`D)\"0D)>PHA(`D)\"0D)\"7-T<F-A=\"AB=68L('-E\nM<\"D[\"B$@\"0D)\"0D)<V5P(#T@(BP@(CL*(2`)\"0D)\"0ES=')C870H8G5F+\"!G\nM971?<G5L95]E>'!R*')T86)L92P@<G1?:6YD97@L\"B$@\"0D)\"0D)\"0D)\"0D)\nM(\"!L9FER<W0H;\"DL('9A<G!R969I>\"DI.PHA(`D)\"0D)?0HA(`D)\"0D)<W1R\nM8V%T*&)U9BP@(BD@24X@(BD[\"B$@\"0D)\"7T*(2`*(2`)\"0D)<W1R8V%T*&)U\nM9BP@(B@B*3L*(2`)\"0D)<W1R8V%T*&)U9BP@9V5T7W%U97)Y7V1E9BAQ=65R\nM>2DI.PHA(`D)\"0ES=')C870H8G5F+\"`B*2(I.PHA(`HA(`D)\"0ER971U<FX@\nM<'-T<F1U<\"AB=68I.PH@(`D)\"7T*(\"`)\"0EB<F5A:SL*(\"`*+2TM(#$S-C,L\nM,3,V.2`M+2TM\"B`@\"B`@\"0EC87-E(%1?4W5B3&EN:SH*(\"`)\"0E[\"B$@\"0D)\nM\"7)E='5R;B!G971?<W5B;&EN:U]E>'!R*'%H+\"!R=%]I;F1E>\"P@;F]D92P@\nM=F%R<')E9FEX*3L*(\"`)\"0E]\"B`@\"0D)8G)E86L[\"B`@\"BHJ*BHJ*BHJ*BHJ\nM*BHJ*@HJ*BH@,3$Q-BPQ,3(R(\"HJ*BH*(\"`@*B`M+2TM+2TM+2TM\"B`@(\"HO\nM\"B`@<W1A=&EC(&-H87(@*@HA(&=E=%]F=6YC7V5X<'(H3&ES=\"`J<G1A8FQE\nM+\"!I;G0@<G1?:6YD97@L($5X<'(@*F5X<'(L(&)O;VP@=F%R<')E9FEX*0H@\nM('L*(\"`)8VAA<@D)[email protected])=.PH@(`E(96%P5'5P;&4)<')O8W1U<#L*\nM+2TM(#$S.#<L,3,Y,R`M+2TM\"B`@(\"H@+2TM+2TM+2TM+0H@(\"`J+PH@('-T\nM871I8R!C:&%R(\"H*(2!G971?9G5N8U]E>'!R*%%R>4AI97(@*G%H+\"!I;G0@\nM<G1?:6YD97@L($5X<'(@*F5X<'(L(&)O;VP@=F%R<')E9FEX*0H@('L*(\"`)\nM8VAA<@D)[email protected])=.PH@(`E(96%P5'5P;&4)<')O8W1U<#L**BHJ*BHJ\nM*BHJ*BHJ*BHJ\"BHJ*B`Q,30S+#$Q-#D@*BHJ*@H@(`D):68@*\"%S=')C;7`H\nM<')O;F%M92P@(FYU;&QV86QU92(I*0H@(`D)>PH@(`D)\"7-T<F-P>2AB=68L\nM(\"(H(BD[\"B$@\"0D)<W1R8V%T*&)U9BP@9V5T7W)U;&5?97AP<BAR=&%B;&4L\nM(')T7VEN9&5X+\"!L9FER<W0H97AP<BT^87)G<RDL\"B`@\"0D)\"0D)\"0D)(\"!V\nM87)P<F5F:7@I*3L*(\"`)\"0ES=')C870H8G5F+\"`B*2!)4TY53$PB*3L*(\"`)\nM\"0ER971U<FX@<'-T<F1U<\"AB=68I.PHM+2T@,30Q-\"PQ-#(P(\"TM+2T*(\"`)\nM\"6EF(\"@A<W1R8VUP*'!R;VYA;64L(\")N=6QL=F%L=64B*2D*(\"`)\"7L*(\"`)\nM\"0ES=')C<'DH8G5F+\"`B*\"(I.PHA(`D)\"7-T<F-A=\"AB=68L(&=E=%]R=6QE\nM7V5X<'(H<6@L(')T7VEN9&5X+\"!L9FER<W0H97AP<BT^87)G<RDL\"B`@\"0D)\nM\"0D)\"0D)(\"!V87)P<F5F:7@I*3L*(\"`)\"0ES=')C870H8G5F+\"`B*2!)4TY5\nM3$PB*3L*(\"`)\"0ER971U<FX@<'-T<F1U<\"AB=68I.PHJ*BHJ*BHJ*BHJ*BHJ\nM*BH**BHJ(#$Q-3$L,3$U-R`J*BHJ\"B`@\"0EI9B`H(7-T<F-M<\"AP<F]N86UE\nM+\"`B;F]N;G5L;'9A;'5E(BDI\"B`@\"0E[\"B`@\"0D)<W1R8W!Y*&)U9BP@(B@B\nM*3L*(2`)\"0ES=')C870H8G5F+\"!G971?<G5L95]E>'!R*')T86)L92P@<G1?\nM:6YD97@L(&QF:7)S=\"AE>'!R+3YA<F=S*2P*(\"`)\"0D)\"0D)\"0D@('9A<G!R\nM969I>\"DI.PH@(`D)\"7-T<F-A=\"AB=68L(\"(I($Y/5$Y53$PB*3L*(\"`)\"0ER\nM971U<FX@<'-T<F1U<\"AB=68I.PHM+2T@,30R,BPQ-#(X(\"TM+2T*(\"`)\"6EF\nM(\"@A<W1R8VUP*'!R;VYA;64L(\")N;VYN=6QL=F%L=64B*2D*(\"`)\"7L*(\"`)\nM\"0ES=')C<'DH8G5F+\"`B*\"(I.PHA(`D)\"7-T<F-A=\"AB=68L(&=E=%]R=6QE\nM7V5X<'(H<6@L(')T7VEN9&5X+\"!L9FER<W0H97AP<BT^87)G<RDL\"B`@\"0D)\nM\"0D)\"0D)(\"!V87)P<F5F:7@I*3L*(\"`)\"0ES=')C870H8G5F+\"`B*2!.3U1.\nM54Q,(BD[\"B`@\"0D)<F5T=7)N('!S=')D=7`H8G5F*3L**BHJ*BHJ*BHJ*BHJ\nM*BHJ\"BHJ*B`Q,38Y+#$Q-S4@*BHJ*@H@(`E[\"B`@\"0ES=')C870H8G5F+\"!S\nM97`I.PH@(`D)<V5P(#T@(BP@(CL*(2`)\"7-T<F-A=\"AB=68L(&=E=%]R=6QE\nM7V5X<'(H<G1A8FQE+\"!R=%]I;F1E>\"P@;&9I<G-T*&PI+\"!V87)P<F5F:7@I\nM*3L*(\"`)?0H@(`ES=')C870H8G5F+\"`B*2(I.PH@(`HM+2T@,30T,\"PQ-#0V\nM(\"TM+2T*(\"`)>PH@(`D)<W1R8V%T*&)U9BP@<V5P*3L*(\"`)\"7-E<\"`](\"(L\nM(\"([\"B$@\"0ES=')C870H8G5F+\"!G971?<G5L95]E>'!R*'%H+\"!R=%]I;F1E\nM>\"P@;&9I<G-T*&PI+\"!V87)P<F5F:7@I*3L*(\"`)?0H@(`ES=')C870H8G5F\nM+\"`B*2(I.PH@(`HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#$Q.30L,3(P,\"`J*BHJ\nM\"B`@(\"H@+2TM+2TM+2TM+0H@(\"`J+PH@('-T871I8R!C:&%R(\"H*(2!G971?\nM=&QE7V5X<'(H3&ES=\"`J<G1A8FQE+\"!I;G0@<G1?:6YD97@L(%1A<F=E=$5N\nM=')Y(\"IT;&4L(&)O;VP@=F%R<')E9FEX*0H@('L*(\"`)2&5A<%1U<&QE\"7!R\nM;V-T=7`[\"B`@\"49O<FU?<&=?<')O8R!P<F]C4W1R=6-T.PHM+2T@,30V-2PQ\nM-#<Q(\"TM+2T*(\"`@*B`M+2TM+2TM+2TM\"B`@(\"HO\"B`@<W1A=&EC(&-H87(@\nM*@HA(&=E=%]T;&5?97AP<BA1<GE(:65R(\"IQ:\"P@:6YT(')T7VEN9&5X+\"!4\nM87)G971%;G1R>2`J=&QE+\"!B;V]L('9A<G!R969I>\"D*(\"![\"B`@\"4AE87!4\nM=7!L90EP<F]C='5P.PH@(`E&;W)M7W!G7W!R;V,@<')O8U-T<G5C=#L**BHJ\nM*BHJ*BHJ*BHJ*BHJ\"BHJ*B`Q,C`X+#$R,3D@*BHJ*@H@(`D@*B`M+2TM+2TM\nM+2TM\"B`@\"2`J+PH@(`EI9B`H=&QE+3YR97-D;VTM/G)E<W1Y<&UO9\"`\\(#`I\nM\"B$@\"0ER971U<FX@9V5T7W)U;&5?97AP<BAR=&%B;&4L(')T7VEN9&5X+\"!T\nM;&4M/F5X<'(L('9A<G!R969I>\"D[\"B`@\"6EF(\"AN;V1E5&%G*'1L92T^97AP\nM<BD@(3T@5%]%>'!R*0HA(`D)<F5T=7)N(&=E=%]R=6QE7V5X<'(H<G1A8FQE\nM+\"!R=%]I;F1E>\"P@=&QE+3YE>'!R+\"!V87)P<F5F:[email protected]@(`EE>'!R(#T@\nM*$5X<'(@*BD@*'1L92T^97AP<BD[\"B`@\"6EF(\"AE>'!R+3YO<%1Y<&4@(3T@\nM1E5.0U]%6%!2*0HA(`D)<F5T=7)N(&=E=%]R=6QE7V5X<'(H<G1A8FQE+\"!R\nM=%]I;F1E>\"P@=&QE+3YE>'!R+\"!V87)P<F5F:[email protected]@(`H@(`EF=6YC(#T@\nM*$9U;F,@*BD@*&5X<'(M/F]P97(I.PH@(`HM+2T@,30W.2PQ-#DP(\"TM+2T*\nM(\"`)(\"H@+2TM+2TM+2TM+0H@(`D@*B\\*(\"`):68@*'1L92T^<F5S9&]M+3YR\nM97-T>7!M;V0@/\"`P*0HA(`D)<F5T=7)N(&=E=%]R=6QE7V5X<'(H<6@L(')T\nM7VEN9&5X+\"!T;&4M/F5X<'(L('9A<G!R969I>\"D[\"B`@\"6EF(\"AN;V1E5&%G\nM*'1L92T^97AP<BD@(3T@5%]%>'!R*0HA(`D)<F5T=7)N(&=E=%]R=6QE7V5X\nM<'(H<6@L(')T7VEN9&5X+\"!T;&4M/F5X<'(L('9A<G!R969I>\"D[\"B`@\"65X\nM<'(@/2`H17AP<B`J*2`H=&QE+3YE>'!R*3L*(\"`):68@*&5X<'(M/F]P5'EP\nM92`A/2!&54Y#7T584%(I\"B$@\"0ER971U<FX@9V5T7W)U;&5?97AP<BAQ:\"P@\nM<G1?:6YD97@L('1L92T^97AP<BP@=F%R<')E9FEX*3L*(\"`*(\"`)9G5N8R`]\nM(\"A&=6YC(\"HI(\"AE>'!R+3YO<&5R*3L*(\"`**BHJ*BHJ*BHJ*BHJ*BHJ\"BHJ\nM*B`Q,C,U+#$R-#4@*BHJ*@H@(`D@*B`M+2TM+2TM+2TM\"B`@\"2`J+PH@(`EI\nM9B`H<')O8U-T<G5C=\"T^<')O;F%R9W,@(3T@,BD*(2`)\"7)E='5R;B!G971?\nM<G5L95]E>'!R*')T86)L92P@<G1?:6YD97@L('1L92T^97AP<BP@=F%R<')E\nM9FEX*3L*(\"`):68@*'!R;V-3=')U8W0M/G!R;W)E='1Y<&4@(3T@<')O8U-T\nM<G5C=\"T^<')O87)G='EP97-;,%TI\"B$@\"0ER971U<FX@9V5T7W)U;&5?97AP\nM<BAR=&%B;&4L(')T7VEN9&5X+\"!T;&4M/F5X<'(L('9A<G!R969I>\"D[\"B`@\nM\"6EF(\"AP<F]C4W1R=6-T+3YP<F]A<F=T>7!E<ULQ72`A/2!)3E0T3TE$*0HA\nM(`D)<F5T=7)N(&=E=%]R=6QE7V5X<'(H<G1A8FQE+\"!R=%]I;F1E>\"P@=&QE\nM+3YE>'!R+\"!V87)P<F5F:[email protected]@(`H@(`DO*B`M+2TM+2TM+2TM\"B`@\"2`J\nM($9I;F%L;'D@*'1O(&)E('1O=&%L;'D@<V%F92D@=&AE('-E8V]N9\"!A<F=U\nM;65N=\"!M=7-T(&)E(&$*+2TM(#$U,#8L,34Q-B`M+2TM\"B`@\"2`J(\"TM+2TM\nM+2TM+2T*(\"`)(\"HO\"B`@\"6EF(\"AP<F]C4W1R=6-T+3YP<F]N87)G<R`A/2`R\nM*0HA(`D)<F5T=7)N(&=E=%]R=6QE7V5X<'(H<6@L(')T7VEN9&5X+\"!T;&4M\nM/F5X<'(L('9A<G!R969I>\"D[\"B`@\"6EF(\"AP<F]C4W1R=6-T+3YP<F]R971T\nM>7!E(\"$]('!R;V-3=')U8W0M/G!R;V%R9W1Y<&5S6S!=*0HA(`D)<F5T=7)N\nM(&=E=%]R=6QE7V5X<'(H<6@L(')T7VEN9&5X+\"!T;&4M/F5X<'(L('9A<G!R\nM969I>\"D[\"B`@\"6EF(\"AP<F]C4W1R=6-T+3YP<F]A<F=T>7!E<ULQ72`A/2!)\nM3E0T3TE$*0HA(`D)<F5T=7)N(&=E=%]R=6QE7V5X<'(H<6@L(')T7VEN9&5X\nM+\"!T;&4M/F5X<'(L('9A<G!R969I>\"D[\"B`@\"B`@\"2\\J(\"TM+2TM+2TM+2T*\nM(\"`)(\"H@1FEN86QL>2`H=&\\@8F4@=&]T86QL>2!S869E*2!T:&4@<V5C;VYD\nM(&%R9W5M96YT(&UU<W0@8F4@80HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#$R-#@L\nM,3(V,B`J*BHJ\"B`@\"2`J+PH@(`ES96-O;F1?87)G(#T@*$-O;G-T(\"HI(&YT\nM:\"@Q+\"!E>'!R+3YA<F=S*3L*(\"`):68@*&YO9&5486<H*$YO9&4@*BD@<V5C\nM;VYD7V%R9RD@(3T@5%]#;VYS=\"D*(2`)\"7)E='5R;B!G971?<G5L95]E>'!R\nM*')T86)L92P@<G1?:6YD97@L('1L92T^97AP<BP@=F%R<')E9FEX*3L*(\"`)\nM:68@*\"AI;G0T*2`H<V5C;VYD7V%R9RT^8V]N<W1V86QU92D@(3T@=&QE+3YR\nM97-D;VTM/G)E<W1Y<&UO9\"D*(2`)\"7)E='5R;B!G971?<G5L95]E>'!R*')T\nM86)L92P@<G1?:6YD97@L('1L92T^97AP<BP@=F%R<')E9FEX*3L*(\"`*(\"`)\nM+RH@+2TM+2TM+2TM+0H@(`D@*B!7:&]W(\"T@9V]T(&ET+B!.;W<@9V5T(')I\nM9\"!O9B!T:&4@<&%D9&EN9R!F=6YC=&EO;@H@(`D@*B`M+2TM+2TM+2TM\"B`@\nM\"2`J+PHA(`ER971U<FX@9V5T7W)U;&5?97AP<BAR=&%B;&4L(')T7VEN9&5X\nM+\"!L9FER<W0H97AP<BT^87)G<RDL('9A<G!R969I>\"D[\"B`@?0H@(`H@(`HM\nM+2T@,34Q.2PQ-3,S(\"TM+2T*(\"`)(\"HO\"B`@\"7-E8V]N9%]A<F<@/2`H0V]N\nM<W0@*BD@;G1H*#$L(&5X<'(M/F%R9W,I.PH@(`EI9B`H;F]D951A9R@H3F]D\nM92`J*2!S96-O;F1?87)G*2`A/2!47T-O;G-T*0HA(`D)<F5T=7)N(&=E=%]R\nM=6QE7V5X<'(H<6@L(')T7VEN9&5X+\"!T;&4M/F5X<'(L('9A<G!R969I>\"D[\nM\"B`@\"6EF(\"@H:6YT-\"D@*'-E8V]N9%]A<F<M/F-O;G-T=F%L=64I(\"$]('1L\nM92T^<F5S9&]M+3YR97-T>7!M;V0I\"B$@\"0ER971U<FX@9V5T7W)U;&5?97AP\nM<BAQ:\"P@<G1?:6YD97@L('1L92T^97AP<BP@=F%R<')E9FEX*3L*(\"`*(\"`)\nM+RH@+2TM+2TM+2TM+0H@(`D@*B!7:&]W(\"T@9V]T(&ET+B!.;W<@9V5T(')I\nM9\"!O9B!T:&4@<&%D9&EN9R!F=6YC=&EO;@H@(`D@*B`M+2TM+2TM+2TM\"B`@\nM\"2`J+PHA(`ER971U<FX@9V5T7W)U;&5?97AP<BAQ:\"P@<G1?:6YD97@L(&QF\nM:7)S=\"AE>'!R+3YA<F=S*2P@=F%R<')E9FEX*3L*(\"!]\"B`@\"B`@\"BHJ*BHJ\nM*BHJ*BHJ*BHJ*@HJ*BH@,3(W-\"PQ,C<Y(\"HJ*BH*+2TM(#$U-#4L,34U,2`M\nM+2TM\"B`@\"6-H87()(\"`@*F5X='9A;#L*(\"`)8F]O;`D):7-N=6QL(#T@1D%,\nM4T4[\"B`@\"6-H87()\"6)U9ELX,3DR73L**R`)8VAA<@D);F%M96)U9ELV-%T[\nM\"B`@\"B`@\"6EF(\"AC;VYS='9A;\"T^8V]N<W1I<VYU;&PI\"B`@\"0ER971U<FX@\nM(DY53$PB.PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#$R.#DL,3(Y-2`J*BHJ\"B`@\nM\"65X='9A;\"`](\"AC:&%R(\"HI(\"@J9FUG<E]F861D<B@F9FEN9F]?;W5T<'5T\nM*2D@*&-O;G-T=F%L+3YC;VYS='9A;'5E+`H@(`D)\"0D)\"0D)\"0D)\"0DF:7-N\nM=6QL+\"`M,2D[\"B`@\"B$@\"7-P<FEN=&8H8G5F+\"`B)R5S)SHZ)7,B+\"!E>'1V\nM86PL(&YA;65O=70H)BAT>7!E4W1R=6-T+3YT>7!N86UE*2DI.PH@(`ER971U\nM<FX@<'-T<F1U<\"AB=68I.PH@('T*(\"`*+2TM(#$U-C$L,38T,R`M+2TM\"B`@\nM\"65X='9A;\"`](\"AC:&%R(\"HI(\"@J9FUG<E]F861D<B@F9FEN9F]?;W5T<'5T\nM*2D@*&-O;G-T=F%L+3YC;VYS='9A;'5E+`H@(`D)\"0D)\"0D)\"0D)\"0DF:7-N\nM=6QL+\"`M,2D[\"B`@\"B$@\"7-P<FEN=&8H;F%M96)U9BP@(CHZ)7,B+\"!N86UE\nM;W5T*\"8H='EP95-T<G5C=\"T^='EP;F%M92DI*3L*(2`):68@*'-T<F-M<\"AN\nM86UE8G5F+\"`B.CIU;FMN;W=N(BD@/3T@,\"D*(2`)\"6YA;65B=69;,%T@/2`G\nM7#`G.PHA(`ES<')I;G1F*&)U9BP@(B<E<R<E<R(L(&5X='9A;\"P@;F%M96)U\nM9BD[\"B$@\"7)E='5R;B!P<W1R9'5P*&)U9BD[\"B$@?0HA(`HA(`HA(\"\\J(\"TM\nM+2TM+2TM+2T*(2`@*B!G971?<W5B;&EN:U]E>'!R\"0D)+2!087)S92!B86-K\nM(&$@<W5B;&EN:PHA(\"`J(\"TM+2TM+2TM+2T*(2`@*B\\*(2!S=&%T:6,@8VAA\nM<B`J\"B$@9V5T7W-U8FQI;FM?97AP<BA1<GE(:65R(\"IQ:\"P@:6YT(')T7VEN\nM9&5X+\"!.;V1E(\"IN;V1E+\"!B;V]L('9A<G!R969I>\"D*(2![\"B$@\"5-U8DQI\nM;FL@(\"`@*G-U8FQI;FL@/2`H4W5B3&EN:R`J*2!N;V1E.PHA(`E1=65R>0D@\nM(\"`J<75E<GD@/2`H475E<GD@*BD@*'-U8FQI;FLM/G-U8G-E;&5C=\"D[\"B$@\nM\"45X<'(@(\"`@(\"`@*F5X<'([\"B$@\"4QI<W0)(\"`@*FP[\"B$@\"6-H87()(\"`@\nM*G-E<#L*(2`)8VAA<B`@(\"`@(\"!B=69;.#$Y,ET[\"B$@\"B$@\"6)U9ELP72`]\nM(\"=<,\"<[\"B$@\"B$@\"6EF(\"AS=6)L:6YK+3YL969T:&%N9\"`A/2!.54Q,*0HA\nM(`E[\"B$@\"0EI9B`H;&5N9W1H*'-U8FQI;FLM/FQE9G1H86YD*2`^(#$I\"B$@\nM\"0D)<W1R8V%T*&)U9BP@(B@B*3L*(2`*(2`)\"7-E<\"`](\"(B.PHA(`D)9F]R\nM96%C:\"AL+\"!S=6)L:6YK+3YL969T:&%N9\"D*(2`)\"7L*(2`)\"0ES=')C870H\nM8G5F+\"!S97`I.PHA(`D)\"7-E<\"`](\"(L(\"([\"B$@\"0D)<W1R8V%T*&)U9BP@\nM9V5T7W)U;&5?97AP<BAQ:\"P@<G1?:6YD97@L\"B$@\"0D)\"0D)\"0D)(\"!L9FER\nM<W0H;\"DL('9A<G!R969I>\"DI.PHA(`D)?0HA(`HA(`D):68@*&QE;F=T:\"AS\nM=6)L:6YK+3YL969T:&%N9\"D@/B`Q*0HA(`D)\"7-T<F-A=\"AB=68L(\"(I(\"(I\nM.PHA(`D)96QS90HA(`D)\"7-T<F-A=\"AB=68L(\"(@(BD[\"B$@\"7T*(2`*(2`)\nM<W=I=&-H(\"AS=6)L:6YK+3YS=6),:6YK5'EP92D@>PHA(`D)8V%S92!%6$E3\nM5%-?4U5\"3$E.2SH*(2`)\"0ES=')C870H8G5F+\"`B15A)4U13(\"(I.PHA(`D)\nM\"6)R96%K.PHA(`HA(`D)8V%S92!!3EE?4U5\"3$E.2SH*(2`)\"0EE>'!R(#T@\nM*$5X<'(@*BEL9FER<W0H<W5B;&EN:RT^;W!E<BD[\"B$@\"0D)<W1R8V%T*&)U\nM9BP@9V5T7V]P;F%M92@H*$]P97(@*BD@*&5X<'(M/F]P97(I*2T^;W!N;RDI\nM.PHA(`D)\"7-T<F-A=\"AB=68L(\"(@04Y9(\"(I.PHA(`D)\"6)R96%K.PHA(`HA\nM(`D)8V%S92!!3$Q?4U5\"3$E.2SH*(2`)\"0EE>'!R(#T@*$5X<'(@*BEL9FER\nM<W0H<W5B;&EN:RT^;W!E<BD[\"B$@\"0D)<W1R8V%T*&)U9BP@9V5T7V]P;F%M\nM92@H*$]P97(@*BD@*&5X<'(M/F]P97(I*2T^;W!N;RDI.PHA(`D)\"7-T<F-A\nM=\"AB=68L(\"(@04Q,(\"(I.PHA(`D)\"6)R96%K.PHA(`HA(`D)8V%S92!%6%!2\nM7U-50DQ)3DLZ\"B$@\"0D)97AP<B`](\"A%>'!R(\"HI;&9I<G-T*'-U8FQI;FLM\nM/F]P97(I.PHA(`D)\"7-T<F-A=\"AB=68L(&=E=%]O<&YA;64H*\"A/<&5R(\"HI\nM(\"AE>'!R+3YO<&5R*2DM/F]P;F\\I*3L*(2`)\"0ES=')C870H8G5F+\"`B(\"(I\nM.PHA(`D)\"6)R96%K.PHA(`HA(`D)9&5F875L=#H*(2`)\"0EE;&]G*$524D]2\nM+\"`B=6YU<'!O<G1E9\"!S=6)L:6YK('1Y<&4@)60B+`HA(`D)\"0D)<W5B;&EN\nM:RT^<W5B3&EN:U1Y<&4I.PHA(`D)\"6)R96%K.PHA(`E]\"B$@\"B$@\"7-T<F-A\nM=\"AB=68L(\"(H(BD[\"B$@\"7-T<F-A=\"AB=68L(&=E=%]Q=65R>5]D968H<75E\nM<GDL('%H*2D[\"B$@\"7-T<F-A=\"AB=68L(\"(I(BD[\"B$@\"B`@\"7)E='5R;B!P\nM<W1R9'5P*&)U9BD[\"B`@?0H@(`ID:69F(\"UC<B!S<F,N;W)I9R]B:6XO:6YI\nM=&1B+VEN:71D8BYS:\"!S<F,O8FEN+VEN:71D8B]I;FET9&(N<V@**BHJ('-R\nM8RYO<FEG+V)I;B]I;FET9&(O:6YI=&1B+G-H\"4UO;B!397`@,C@@,3,Z-#4Z\nM-38@,3DY.`HM+2T@<W)C+V)I;B]I;FET9&(O:6YI=&1B+G-H\"4UO;B!397`@\nM,C@@,C,Z,#DZ,S,@,3DY.`HJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#0S-2PT-#`@\nM*BHJ*@HM+2T@-#,U+#0T,2`M+2TM\"B`@\"B`@96-H;R`B0W)E871I;F<@=FEE\nM=R!P9U]R=6QE<R(*(\"!E8VAO(\")#4D5!5$4@5$%\"3$4@>'!G7W)U;&5S(\"@)\nM\"5P**R`)(\"`@('1A8FQE;F%M90EN86UE+`D)7`H@(`D@(\"`@<G5L96YA;64)\nM;F%M92P)\"5P*(\"`)(\"`@(&1E9FEN:71I;VX)=&5X=\"D[(B!\\('!O<W1G<F5S\nM(\"101U-13%]/4%0@=&5M<&QA=&4Q(#X@+V1E=B]N=6QL\"B`@(VUO=F4@:70@\nM:6YT;R!P9U]R=6QE<PHJ*BHJ*BHJ*BHJ*BHJ*BH**BHJ(#0T-2PT-38@*BHJ\nM*@H@(&UV(\"101T1!5$$O8F%S92]T96UP;&%T93$O>'!G7W)U;&5S(\"101T1!\nM5$$O8F%S92]T96UP;&%T93$O<&=?<G5L97,*(\"`*(\"!E8VAO(\")#4D5!5$4@\nM4E5,12!<(E]2151P9U]R=6QE<UPB($%3($].(%-%3$5#5\"!43R!P9U]R=6QE\nM<R!$3R!)3E-414%$\"5P*(2`)(\"`@(%-%3$5#5\"!R=6QE;F%M92P@<&=?9V5T\nM7W)U;&5D968H<G5L96YA;64I($%3(&1E9FEN:71I;VX)7`HA(`D@(\"`@(\"!&\nM4D]-('!G7W)E=W)I=&4[(B!\\('!O<W1G<F5S(\"101U-13%]/4%0@=&5M<&QA\nM=&4Q(#X@+V1E=B]N=6QL\"B`@\"B`@96-H;R`B0W)E871I;F<@=FEE=R!P9U]V\nM:65W<R(*(\"!E8VAO(\")#4D5!5$4@5$%\"3$4@>'!G7W9I97=S(\"@)\"5P*(\"`)\nM(\"`@('9I97=N86UE\"6YA;64L\"0E<\"B`@\"2`@(\"!D969I;FET:6]N\"71E>'0I\nM.R(@?\"!P;W-T9W)E<R`D4$=344Q?3U!4('1E;7!L871E,2`^(\"]D978O;G5L\nM;`H@(\"-M;W9E(&ET(&EN=&\\@<&=?=FEE=W,*(\"!E8VAO(\")54$1!5$4@<&=?\nM8VQA<W,@4T54(')E;&YA;64@/2`G<&=?=FEE=W,G(%=(15)%(')E;&YA;64@\nM/2`G>'!G7W9I97=S)SLB('Q<\"BTM+2`T-#8L-#8S(\"TM+2T*(\"!M=B`D4$=$\nM051!+V)A<V4O=&5M<&QA=&4Q+WAP9U]R=6QE<R`D4$=$051!+V)A<V4O=&5M\nM<&QA=&4Q+W!G7W)U;&5S\"B`@\"B`@96-H;R`B0U)%051%(%)53$4@7\")?4D54\nM<&=?<G5L97-<(B!!4R!/3B!314Q%0U0@5$\\@<&=?<G5L97,@1$\\@24Y35$5!\nM1`E<\"B$@\"2`@(\"!314Q%0U0@0RYR96QN86UE($%3('1A8FQE;F%M92P)\"0D)\nM7`HA(`D@(\"`@(\"`@(\"`@(%(N<G5L96YA;64@05,@<G5L96YA;64L\"0D)\"5P*\nM(2`)\"2`@('!G7V=E=%]R=6QE9&5F*%(N<G5L96YA;64I($%3(&1E9FEN:71I\nM;VX)\"5P*(2`)(\"`@(\"`@1E)/32!P9U]R97=R:71E(%(L('!G7V-L87-S($,@\nM\"0D)\"5P*(2`)(\"`@(\"`@(\"`@(\"!72$5212!2+G)U;&5N86UE(\"%^(\"=>7U)%\nM5\"<)\"0D)7`HA(`D)(\"`@04Y$($,N;VED(#T@4BYE=E]C;&%S<SLB('P@7`HA\nM(`EP;W-T9W)E<R`D4$=344Q?3U!4('1E;7!L871E,2`^(\"]D978O;G5L;`H@\nM(`H@(&5C:&\\@(D-R96%T:6YG('9I97<@<&=?=FEE=W,B\"B`@96-H;R`B0U)%\nM051%(%1!0DQ%('AP9U]V:65W<R`H\"0E<\"B`@\"2`@(\"!V:65W;F%M90EN86UE\nM+`D)7`HK(`D@(\"`@=FEE=V]W;F5R\"6YA;64L\"0E<\"B`@\"2`@(\"!D969I;FET\nM:6]N\"71E>'0I.R(@?\"!P;W-T9W)E<R`D4$=344Q?3U!4('1E;7!L871E,2`^\nM(\"]D978O;G5L;`H@(\"-M;W9E(&ET(&EN=&\\@<&=?=FEE=W,*(\"!E8VAO(\")5\nM4$1!5$4@<&=?8VQA<W,@4T54(')E;&YA;64@/2`G<&=?=FEE=W,G(%=(15)%\nM(')E;&YA;64@/2`G>'!G7W9I97=S)SLB('Q<\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ\nM*BH@-#8Q+#0V.2`J*BHJ\"BTM+2`T-C@L-3(R(\"TM+2T*(\"`*(\"!E8VAO(\")#\nM4D5!5$4@4E5,12!<(E]2151P9U]V:65W<UPB($%3($].(%-%3$5#5\"!43R!P\nM9U]V:65W<R!$3R!)3E-414%$\"5P*(\"`)(\"`@(%-%3$5#5\"!R96QN86UE($%3\nM('9I97=N86UE+\"`)\"0D)7`HK(`D)(\"`@<&=?9V5T7W5S97)B>6ED*')E;&]W\nM;F5R*2!!4R!V:65W;W=N97(L\"0E<\"B`@\"2`@(\"`@(\"`@(\"`@<&=?9V5T7W9I\nM97=D968H<F5L;F%M92D@05,@9&5F:6YI=&EO;@D)7`H@(`D@(\"`@(\"!&4D]-\nM('!G7V-L87-S(%=(15)%(')E;&AA<W)U;&5S($%.1`D)\"5P*(\"`)(\"`@(\"`@\nM(\"`@(\"!P9U]G971?=FEE=V1E9BAR96QN86UE*2`A/2`G3F]T(&$@=FEE=R<[\nM(B!\\(%P**R`)<&]S=&=R97,@)%!'4U%,7T]05\"!T96UP;&%T93$@/B`O9&5V\nM+VYU;&P**R`**R!E8VAO(\")#<F5A=&EN9R!V:65W('!G7W1A8FQE<R(**R!E\nM8VAO(\")#4D5!5$4@5$%\"3$4@>'!G7W1A8FQE<R`H\"0E<\"BL@\"2`@(\"!T86)L\nM96YA;64);F%M92P)\"5P**R`)(\"`@('1A8FQE;W=N97();F%M92P)\"5P**R`)\nM(\"`@(&AA<VEN9&5X97,)8F]O;\"P)\"5P**R`)(\"`@(&AA<W)U;&5S\"6)O;VPL\nM\"0E<\"BL@\"2`@(\"!H87-T<FEG9V5R<PEB;V]L*3LB('P@<&]S=&=R97,@)%!'\nM4U%,7T]05\"!T96UP;&%T93$@/B`O9&5V+VYU;&P**R`C;6]V92!I=\"!I;G1O\nM('!G7W1A8FQE<PHK(&5C:&\\@(E501$%412!P9U]C;&%S<R!3150@<F5L;F%M\nM92`](\"=P9U]T86)L97,G(%=(15)%(')E;&YA;64@/2`G>'!G7W1A8FQE<R<[\nM(B!\\7`HK(`EP;W-T9W)E<R`D4$=344Q?3U!4('1E;7!L871E,2`^(\"]D978O\nM;G5L;`HK(&5C:&\\@(E501$%412!P9U]T>7!E(%-%5\"!T>7!N86UE(#T@)W!G\nM7W1A8FQE<R<@5TA%4D4@='EP;F%M92`](\"=X<&=?=&%B;&5S)SLB('Q<\"BL@\nM\"7!O<W1G<F5S(\"101U-13%]/4%0@=&5M<&QA=&4Q(#X@+V1E=B]N=6QL\"BL@\nM;78@)%!'1$%402]B87-E+W1E;7!L871E,2]X<&=?=&%B;&5S(\"101T1!5$$O\nM8F%S92]T96UP;&%T93$O<&=?=&%B;&5S\"BL@\"BL@96-H;R`B0U)%051%(%)5\nM3$4@7\")?4D54<&=?=&%B;&5S7\"(@05,@3TX@4T5,14-4(%1/('!G7W1A8FQE\nM<R!$3R!)3E-414%$\"5P**R`)(\"`@(%-%3$5#5\"!R96QN86UE($%3('1A8FQE\nM;F%M92P@\"0D)\"5P**R`)\"2`@('!G7V=E=%]U<V5R8GEI9\"AR96QO=VYE<BD@\nM05,@=&%B;&5O=VYE<BP)\"5P**R`)\"2`@(')E;&AA<VEN9&5X($%3(&AA<VEN\nM9&5X97,L\"0D)\"5P**R`)\"2`@(')E;&AA<W)U;&5S($%3(&AA<W)U;&5S+`D)\nM\"0E<\"BL@\"0D@(\"`H<F5L=')I9V=E<G,@/B`P*2!!4R!H87-T<FEG9V5R<PD)\nM\"5P**R`)(\"`@(\"`@1E)/32!P9U]C;&%S<R!72$5212!R96QK:6YD($E.(\"@G\nM<B<L(\"=S)RD)\"0E<\"BL@\"2`@(\"`@(\"`@(\"`@04Y$('!G7V=E=%]V:65W9&5F\nM*')E;&YA;64I(#T@)TYO=\"!A('9I97<G.R(@?\"!<\"BL@\"7!O<W1G<F5S(\"10\nM1U-13%]/4%0@=&5M<&QA=&4Q(#X@+V1E=B]N=6QL\"BL@\"BL@96-H;R`B0W)E\nM871I;F<@=FEE=R!P9U]I;F1E>&5S(@HK(&5C:&\\@(D-214%412!404),12!X\nM<&=?:6YD97AE<R`H\"5P**R`)(\"`@('1A8FQE;F%M90EN86UE+`D)7`HK(`D@\nM(\"`@:6YD97AN86UE\"6YA;64L\"0E<\"BL@\"2`@(\"!I;F1E>&1E9@ET97AT*3LB\nM('P@<&]S=&=R97,@)%!'4U%,7T]05\"!T96UP;&%T93$@/B`O9&5V+VYU;&P*\nM*R`C;6]V92!I=\"!I;G1O('!G7VEN9&5X97,**R!E8VAO(\")54$1!5$4@<&=?\nM8VQA<W,@4T54(')E;&YA;64@/2`G<&=?:6YD97AE<R<@5TA%4D4@<F5L;F%M\nM92`](\"=X<&=?:6YD97AE<R<[(B!\\7`HK(`EP;W-T9W)E<R`D4$=344Q?3U!4\nM('1E;7!L871E,2`^(\"]D978O;G5L;`HK(&5C:&\\@(E501$%412!P9U]T>7!E\nM(%-%5\"!T>7!N86UE(#T@)W!G7VEN9&5X97,G(%=(15)%('1Y<&YA;64@/2`G\nM>'!G7VEN9&5X97,G.R(@?%P**R`)<&]S=&=R97,@)%!'4U%,7T]05\"!T96UP\nM;&%T93$@/B`O9&5V+VYU;&P**R!M=B`D4$=$051!+V)A<V4O=&5M<&QA=&4Q\nM+WAP9U]I;F1E>&5S(\"101T1!5$$O8F%S92]T96UP;&%T93$O<&=?:6YD97AE\nM<PHK(`HK(&5C:&\\@(D-214%412!254Q%(%PB7U)%5'!G7VEN9&5X97-<(B!!\nM4R!/3B!314Q%0U0@5$\\@<&=?:6YD97AE<R!$3R!)3E-414%$\"5P**R`)(\"`@\nM(%-%3$5#5\"!#+G)E;&YA;64@05,@=&%B;&5N86UE+\"`)\"0D)7`HK(`D)(\"`@\nM22YR96QN86UE($%3(&EN9&5X;F%M92P)\"0D)7`HK(`D)(\"`@<&=?9V5T7VEN\nM9&5X9&5F*%@N:6YD97AR96QI9\"D@05,@:6YD97AD968)\"5P**R`)(\"`@(\"`@\nM1E)/32!P9U]I;F1E>\"!8+\"!P9U]C;&%S<R!#+\"!P9U]C;&%S<R!)\"0D)7`HK\nM(`D@(\"`@(\"`)(\"`@5TA%4D4@0RYO:60@/2!8+FEN9')E;&ED\"0D)\"5P**R`)\nM(\"`@(\"`@(\"`@(\"!!3D0@22YO:60@/2!8+FEN9&5X<F5L:60[(B!\\(%P*(\"`)\nM<&]S=&=R97,@)%!'4U%,7T]05\"!T96UP;&%T93$@/B`O9&5V+VYU;&P*(\"`*\nM(\"!E8VAO(\"),;V%D:6YG('!G7V1E<V-R:7!T:6]N(@ID:69F(\"UC<B!S<F,N\nM;W)I9R]I;F-L=61E+V-A=&%L;V<O<&=?<')O8RYH('-R8R]I;F-L=61E+V-A\nM=&%L;V<O<&=?<')O8RYH\"BHJ*B!S<F,N;W)I9R]I;F-L=61E+V-A=&%L;V<O\nM<&=?<')O8RYH\"4UO;B!397`@,C@@,3,Z-#8Z,#`@,3DY.`HM+2T@<W)C+VEN\nM8VQU9&4O8V%T86QO9R]P9U]P<F]C+F@)36]N(%-E<\"`R.\"`R,#HR,SHR-\"`Q\nM.3DX\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@,C`T,\"PR,#0U(\"HJ*BH*+2TM(#(P\nM-#`L,C`T.2`M+2TM\"B`@1$530U(H(G-O=7)C92!T97AT(&]F(&$@<G5L92(I\nM.PH@($1!5$$H:6YS97)T($])1\"`](#$V-#$@*\"`@<&=?9V5T7W9I97=D968)\nM(\"`@4$=5240@,3$@9B!T(&8@,2!F(#(U(\"(Q.2(@,3`P(#`@,\"`Q,#`@(&9O\nM;R!B87(@*2D[\"B`@1$530U(H(G-E;&5C=\"!S=&%T96UE;G0@;V8@82!V:65W\nM(BD[\"BL@1$%402AI;G-E<G0@3TE$(#T@,38T,B`H(\"!P9U]G971?=7-E<F)Y\nM:60)(\"`@4$=5240@,3$@9B!T(&8@,2!F(#$Y(\"(R,R(@,3`P(#`@,\"`Q,#`@\nM(&9O;R!B87(@*2D[\"BL@1$530U(H(G5S97(@;F%M92!B>2!5240@*'=I=&@@\nM9F%L;&)A8VLI(BD[\"BL@1$%402AI;G-E<G0@3TE$(#T@,38T,R`H(\"!P9U]G\nM971?:6YD97AD968)(\"`@4$=5240@,3$@9B!T(&8@,2!F(#(U(\"(R-B(@,3`P\nM(#`@,\"`Q,#`@(&9O;R!B87(@*2D[\"BL@1$530U(H(FEN9&5X(&1E<V-R:7!T\nM:6]N(BD[\"B`@\"B`@+RH*(\"`@*B!P<F]T;W1Y<&5S(&9O<B!F=6YC=&EO;G,@\n*<&=?<')O8RYC\"@``\n`\nend\n",
"msg_date": "Tue, 29 Sep 1998 00:07:32 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Fix for pg_views/pg_rules and new system views"
}
] |
[
{
"msg_contents": "Gentlemen:\n\tI have tested today's snapshot (retrieved this morning)\non an SGI Irix 6.2 system with the following results:\n\n\t1) The int8 test still fails because the SGI C compiler does\nnot handle a variable argument declaration of \"long long\" correctly in\nthe definition of the function dopr in snprintf.c. According to the\nSGI documentation, the handling of \"long long\" is undefined. I think\nthe best solution is to use sprintf in int8.c, because there is no\nadditional value to using snprintf in this context. There is no risk\nover buffer overflow in int8.c because the maximum length of a\npossible result is known in advance (21 characters + one for the null\ntermination). I've modified int8.c on my machine accordingly.\n\t2) The definition of the slock_t type is incorrect in irix5.h.\n\t3) The new locking mechanism using test_and_set subroutine\ncalls was tested by running 5 parallel processes, two creating indices\nand three doing sequential scans. All queries completed normally. \n\t4) The comment in s_lock.h on possible future SGI locking\nmechanisms should be deleted, since this mechanism is fast, works, and\nis supported by SGI.\n\tPatches to make these corrections (except 4) are included below.\n\n\tAlso, please include the following Makefile.custom into\nthe SGI Irix FAQ:\n\nCUSTOM_CC = cc -32 -g\nLD += -32 -g\n\nThe \"-32\" switch is essential, but I haven't tested the need for \"-g\".\nIf I have time before the release, I will report the results. At least,\nI know that this version passes the regression tests. \n\n\tWould you please put my name on the list of people who\ncontributed to the 6.4 release? Thanks.\n\nHere are the patches:\n\n*** backend/utils/adt/int8.c.orig Tue Sep 1 03:01:35 1998\n--- backend/utils/adt/int8.c Mon Sep 28 21:32:38 1998\n***************\n*** 66,74 ****\n if (!PointerIsValid(val))\n return NULL;\n \n! if ((len = snprintf(buf, MAXINT8LEN, INT64_FORMAT, *val)) < 0)\n elog(ERROR, \"Unable to format int8\", NULL);\n- \n result = palloc(len + 1);\n \n strcpy(result, buf);\n--- 66,73 ----\n if (!PointerIsValid(val))\n return NULL;\n \n! if ((len = sprintf(buf, INT64_FORMAT, *val)) < 0)\n elog(ERROR, \"Unable to format int8\", NULL);\n result = palloc(len + 1);\n \n strcpy(result, buf);\n*** include/port/irix5.h.orig Sun Sep 7 00:59:54 1997\n--- include/port/irix5.h Mon Sep 28 19:37:12 1998\n***************\n*** 3,7 ****\n #define NO_EMPTY_STMTS\n #define SYSV_DIRENT\n #define HAS_TEST_AND_SET\n! #include <abi_mutex.h>\n! typedef abilock_t slock_t;\n--- 3,6 ----\n #define NO_EMPTY_STMTS\n #define SYSV_DIRENT\n #define HAS_TEST_AND_SET\n! typedef unsigned long slock_t;\n*** include/storage/s_lock.h.orig Mon Sep 21 03:00:17 1998\n--- include/storage/s_lock.h Mon Sep 28 19:37:12 1998\n***************\n*** 259,266 ****\n #if defined(__sgi)\n /*\n * SGI IRIX 5\n! * slock_t is defined as a struct abilock_t, which has a single unsigned long\n! * member.\n *\n * This stuff may be supplemented in the future with Masato Kataoka's MIPS-II\n * assembly from his NECEWS SVR4 port, but we probably ought to retain this\n--- 259,269 ----\n #if defined(__sgi)\n /*\n * SGI IRIX 5\n! * slock_t is defined as a unsigned long. We use the standard SGI\n! * mutex API. \n! *\n! * The following comment is left for historical reasons, but is probably\n! * not a good idea since the mutex ABI is supported.\n *\n * This stuff may be supplemented in the future with Masato Kataoka's MIPS-II\n * assembly from his NECEWS SVR4 port, but we probably ought to retain this\n\t\n+------------------------------------------+------------------------------+\n| Robert E. Bruccoleri, Ph.D. | Associate Research Professor |\n| phone: 732 235 5796 | Center for Advanced |\n| Fax: 732 235 4850 | Biotechnology and Medicine |\n| email: [email protected] | Rutgers University |\n| URL: http://www.cabm.rutgers.edu/~bruc | 679 Hoes Lane |\n| | Piscataway, NJ 08854-5638 |\n+------------------------------------------+------------------------------+\n",
"msg_date": "Mon, 28 Sep 1998 23:32:23 -0400 (EDT)",
"msg_from": "Robert Bruccoleri <[email protected]>",
"msg_from_op": true,
"msg_subject": "SGI Port of Postgresql 6.4 snapshot of 09/28/98"
},
{
"msg_contents": "I have applied patches 2, 3, and 4. For patch 1, I changed the 'long\nlong' in va_arg() to a typedef for 'long long'. Let me know if that\nfixes it.\n\n> Gentlemen:\n> \tI have tested today's snapshot (retrieved this morning)\n> on an SGI Irix 6.2 system with the following results:\n> \n> \t1) The int8 test still fails because the SGI C compiler does\n> not handle a variable argument declaration of \"long long\" correctly in\n> the definition of the function dopr in snprintf.c. According to the\n> SGI documentation, the handling of \"long long\" is undefined. I think\n> the best solution is to use sprintf in int8.c, because there is no\n> additional value to using snprintf in this context. There is no risk\n> over buffer overflow in int8.c because the maximum length of a\n> possible result is known in advance (21 characters + one for the null\n> termination). I've modified int8.c on my machine accordingly.\n> \t2) The definition of the slock_t type is incorrect in irix5.h.\n> \t3) The new locking mechanism using test_and_set subroutine\n> calls was tested by running 5 parallel processes, two creating indices\n> and three doing sequential scans. All queries completed normally. \n> \t4) The comment in s_lock.h on possible future SGI locking\n> mechanisms should be deleted, since this mechanism is fast, works, and\n> is supported by SGI.\n> \tPatches to make these corrections (except 4) are included below.\n\n> \n> \tAlso, please include the following Makefile.custom into\n> the SGI Irix FAQ:\n> \n> CUSTOM_CC = cc -32 -g\n> LD += -32 -g\n> \n> The \"-32\" switch is essential, but I haven't tested the need for \"-g\".\n> If I have time before the release, I will report the results. At least,\n> I know that this version passes the regression tests. \n\nPlease contact the IRIX FAQ maintainer directly.\n\n> \n> \tWould you please put my name on the list of people who\n> contributed to the 6.4 release? Thanks.\n\nDone.\n\n> \n> Here are the patches:\n> \n> *** backend/utils/adt/int8.c.orig Tue Sep 1 03:01:35 1998\n> --- backend/utils/adt/int8.c Mon Sep 28 21:32:38 1998\n> ***************\n> *** 66,74 ****\n> if (!PointerIsValid(val))\n> return NULL;\n> \n> ! if ((len = snprintf(buf, MAXINT8LEN, INT64_FORMAT, *val)) < 0)\n> elog(ERROR, \"Unable to format int8\", NULL);\n> - \n> result = palloc(len + 1);\n> \n> strcpy(result, buf);\n> --- 66,73 ----\n> if (!PointerIsValid(val))\n> return NULL;\n> \n> ! if ((len = sprintf(buf, INT64_FORMAT, *val)) < 0)\n> elog(ERROR, \"Unable to format int8\", NULL);\n> result = palloc(len + 1);\n> \n> strcpy(result, buf);\n> *** include/port/irix5.h.orig Sun Sep 7 00:59:54 1997\n> --- include/port/irix5.h Mon Sep 28 19:37:12 1998\n> ***************\n> *** 3,7 ****\n> #define NO_EMPTY_STMTS\n> #define SYSV_DIRENT\n> #define HAS_TEST_AND_SET\n> ! #include <abi_mutex.h>\n> ! typedef abilock_t slock_t;\n> --- 3,6 ----\n> #define NO_EMPTY_STMTS\n> #define SYSV_DIRENT\n> #define HAS_TEST_AND_SET\n> ! typedef unsigned long slock_t;\n> *** include/storage/s_lock.h.orig Mon Sep 21 03:00:17 1998\n> --- include/storage/s_lock.h Mon Sep 28 19:37:12 1998\n> ***************\n> *** 259,266 ****\n> #if defined(__sgi)\n> /*\n> * SGI IRIX 5\n> ! * slock_t is defined as a struct abilock_t, which has a single unsigned long\n> ! * member.\n> *\n> * This stuff may be supplemented in the future with Masato Kataoka's MIPS-II\n> * assembly from his NECEWS SVR4 port, but we probably ought to retain this\n> --- 259,269 ----\n> #if defined(__sgi)\n> /*\n> * SGI IRIX 5\n> ! * slock_t is defined as a unsigned long. We use the standard SGI\n> ! * mutex API. \n> ! *\n> ! * The following comment is left for historical reasons, but is probably\n> ! * not a good idea since the mutex ABI is supported.\n> *\n> * This stuff may be supplemented in the future with Masato Kataoka's MIPS-II\n> * assembly from his NECEWS SVR4 port, but we probably ought to retain this\n> \t\n> +------------------------------------------+------------------------------+\n> | Robert E. Bruccoleri, Ph.D. | Associate Research Professor |\n> | phone: 732 235 5796 | Center for Advanced |\n> | Fax: 732 235 4850 | Biotechnology and Medicine |\n> | email: [email protected] | Rutgers University |\n> | URL: http://www.cabm.rutgers.edu/~bruc | 679 Hoes Lane |\n> | | Piscataway, NJ 08854-5638 |\n> +------------------------------------------+------------------------------+\n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 12:17:05 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: SGI Port of Postgresql 6.4 snapshot of 09/28/98"
}
] |
[
{
"msg_contents": "We need to think about whether to surround all identifiers with double\nquotes all the time in pg_dump output. \n\nThe reason is that Postgres allows reserved keywords to be specified as\ntable and column names if they are surrounded by the double quotes, but\npg_dump doesn't know whether an identifier also happens to be a reserved\nkeyword. afaik it's now only using DQs if there is upper case or funny\ncharacters in the identifier. Instead it should probably surround the\nfields with DQs all the time. It could perhaps have a command-line\nswitch to turn off that feature if necessary.\n\nThe alternative to always using the DQs is to have pg_dump use the\nkeywords.c routine available in the backend. But I don't think that is\nas reliable since it isn't guaranteed to be in sync with the backend\nversion and since there are non-reserved keywords in that file which\ntest the same as the reserved ones.\n\nThis should probably be a \"must do\" for v6.4...\n\nComments?\n\n - Tom\n",
"msg_date": "Tue, 29 Sep 1998 04:39:04 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "pg_dump"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> We need to think about whether to surround all identifiers with double\n> quotes all the time in pg_dump output. \n\nIs there any downside to that? Offhand I don't see one.\nUnless you do, let's just do it...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Tue, 29 Sep 1998 08:55:21 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_dump "
}
] |
[
{
"msg_contents": "Hi. I've been writing documentation on SQL syntax for v6.4, and while\ncompiling lists of parser keywords I've made minor repairs and\nimprovements to gram.y.\n\npreproc.y looks like it has similar sections of code. Can you upgrade\necpg if I send you patches from gram.y? Or do you track changes by\nlooking at the cvs tree? Or??\n\nLet me know if you have time to look at this and I'll send patches or\nwhatever you prefer...\n\nRegards.\n\n - Tom\n",
"msg_date": "Tue, 29 Sep 1998 05:28:58 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "ecpg parser"
}
] |
[
{
"msg_contents": "Hi,\n\n I wonder if anyone ever played around with PL/pgSQL. So far\n (except for the missing Makefile) I got no response if it\n compiles/works on other platforms than my Linux.\n\n Since there are only a few details left in the rewrite code\n for 6.4 I would like to get back on it. I think 6.4 BETA is\n the right time to check if it works on all the platforms (6.4\n is the first time we ship it).\n\n I would do it by adding a build/install with (errors ignored\n first) to the Makefiles. Then another regression test could\n check if it works.\n\n Comments?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Tue, 29 Sep 1998 09:30:57 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "PL/pgSQL in template1?"
},
{
"msg_contents": "Jan,\n\nI have been hacking around trying to get a clean build without much luck. I\nhave adjusted the compile options and fixed one casting problem to get all C\ncode compiled. I also extracted all the external symbols successfully.\n(AIX requirement) However, I get the following message at link time:\n\n ld: 0711-362 ERROR: Glink code cannot be added for function: _ptrgl\n This function is used by compilers and must be called directly.\n\nShared libraries are a bit more tricky on AIX, but I thought I had the\nroutine pretty well figured out. I am at a complete loss here.\n\nAndreas, could you give it a try?\n\nJan Wieck wrote:\n\n> Hi,\n>\n> I wonder if anyone ever played around with PL/pgSQL. So far\n> (except for the missing Makefile) I got no response if it\n> compiles/works on other platforms than my Linux.\n>\n> Since there are only a few details left in the rewrite code\n> for 6.4 I would like to get back on it. I think 6.4 BETA is\n> the right time to check if it works on all the platforms (6.4\n> is the first time we ship it).\n>\n> I would do it by adding a build/install with (errors ignored\n> first) to the Makefiles. Then another regression test could\n> check if it works.\n>\n> Comments?\n\n\n\n",
"msg_date": "Tue, 29 Sep 1998 12:27:34 -0400",
"msg_from": "David Hartwig <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] PL/pgSQL in template1?"
},
{
"msg_contents": ">\n> Jan,\n>\n> I have been hacking around trying to get a clean build without much luck. I\n> have adjusted the compile options and fixed one casting problem to get all C\n> code compiled. I also extracted all the external symbols successfully.\n> (AIX requirement) However, I get the following message at link time:\n>\n> ld: 0711-362 ERROR: Glink code cannot be added for function: _ptrgl\n> This function is used by compilers and must be called directly.\n>\n> Shared libraries are a bit more tricky on AIX, but I thought I had the\n> routine pretty well figured out. I am at a complete loss here.\n>\n> Andreas, could you give it a try?\n\n That's really interesting. The only difference (of interest)\n I can see between my Makefile and that of contrib/spi (where\n refint and autoinc are located which afaik are used in the\n regression test) is that I missed to add CFLAGS_SL to the\n CFLAGS. Seems that one only contains -fpic.\n\n Could you try to use\n\n CFLAGS+= $(CFLAGS_SL) -I$(LIBPQDIR) -I$(SRCDIR)/include\n\n in line 22 of my Makefile?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Tue, 29 Sep 1998 19:19:38 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] PL/pgSQL in template1?"
}
] |
[
{
"msg_contents": "Bruce,\n\n here is the patch that includes PL/pgSQL into the build\n (currently with make errors ignored) and adds a regression\n test for it. A clean build and regression ran fine here.\n\n Can you please apply it?\n\n The tar should be extracted in /usr/local/src/pgsql and\n creates the following files:\n\n src/pl/Makefile\n called by toplevel GNUmakefile and for now only calls\n src/pl/plpgsql/Makefile\n\n src/pl/plpgsql/Makefile\n calls src/pl/plpgsql/src/Makefile (here the call to\n make ignores build errors - this must be changed\n later for the final release).\n\n src/test/regress/input/install_plpgsql.source\n SQL script installing PL/pgSQL language in regression\n database. Will be modified by .../input/Makefile to\n point to correct PGLIB directory where plpgsql.so\n gets installed.\n\n src/test/regress/output/install_plpgsql.source\n expected output for installation script.\n\n src/test/regress/sql/plpgsql.sql\n the main regression test. It tests functions and\n triggers written in PL/pgSQL including views that use\n supportfunctions in this language.\n\n src/test/regress/expected/plpgsql.out\n the expected output for the above regression test.\n\n make_plpgsql.diff\n patch that adds some lines to\n\n src/GNUmakefile.in\n src/test/regress/expected/Makefile\n src/test/regress/input/Makefile\n src/test/regress/output/Makefile\n src/test/regress/sql/Makefile\n src/test/regress/sql/tests\n\n so the plpgsql shared object will be built and\n installed and the regression test for install_plpgsql\n gets patched for LIBDIR to succeed.\n\n\n Anyone please report problems with building PL/pgSQL or\n running the regression tests.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\nbegin 644 regress_plpgsql.tar.gz\nM'XL(`':>$#8\"`^P]:W/;1I+^2OZ*26(718>B\"(\"D8BEVG:-'J#O%UHK2,5NW\nM5UH0'(G8@``#@)95E;K??M,S`V#P&.)!@E:R9I4M$ICI[NGW-!I`]V\"A_X;O\nMEM;RP?O=ZL[,^_L7V_XHO=ZPWT<O$/EH/?I74=E?^-I3API\"PV&OKRC#07\\(\nM9X?JX0O4>[&#S\\KS=1>A%W3]:\\;-IOJ+O^`'!([V#1=YKM%U7//AX.</MZ`2\nM]Z:%NZ8-QQ.'FJ]?OY8-;]RL,!KC)5+?H-Z;(TT](E^4-V]^:.[O[V?`2HX?\nM'/4/V?C7\\0]%JAUV^BJB/P$<_-00^;;?1*CQ<N^7]_]UUD;[)VBJ&[]A>X9T\nMRTJ<,6T?N_>Z@;V,DU.R7#CZ?>SHTN)#S7L;_X[V7NX]FM;,T-T96NAVNX/:\nM\"3#D*)]!:##O,U?2?],9]*.5#'J=P3!K)2O?M#Q\"-5'2-+E\\E9*S9#'!F=2\"\nMHBEE%A7-DB]LJ'2&PVAA0ZTS?+-.1(:%=5LNI*S3L#)V/+6NK.$^]OS,$]WN\nM@>'8OFM.#[RE&0Z1<R1KO6]^Z\"@]+5JPTE/(@<-U2W[Y'_+UIL[!8LG!U$KI\nMP#+\"HQ.8W-(V#SPZ</&#BSWO`']>8L/'LX-?N)U2NUT_).X3UH]-^XB!&O<1\nMQ><K/9C?[\\E]AD+D<<CD@]!W\\&]_6Q\\\"K/D-.KD\\O[@\\&[]M&\"[6?7QWO[(-\nMWW3L.Z7KK(CF)8ZJ[*BS?.)?B%&Y.E$!C_Y>F)X!7P`V8CIY!-)T%VC_'KW<\nMX]C:3-GHVC@I?YZU!8XDS#J*K#=':4U[N?+7:&S\\_!IUC0_,T-7!&EV53R:*\nMJ@Z/E*%<456UH_X0*&KX23.>,`S]8]T0-3V$<IX=_08U8H(A1QGOB4\".\",,O\nM/G\"&DX.P1B`K].`[)8L?32K+.H(SDX9^1SL4DH9!1XLBTL>?_O/M/Y>/LW\\>\nM4\\H:'IZA?8R^]8[NR*G)Z=W1RY?DR\\$_NO_H'CU\\RP8!Z7S0Z>7X]OS\\XE<R\nM;B_XWF8#OX\\-O+SXZ?3B&H:Q;^TTM(.[V_'9]=W!RY?P]X\"<_Q&]_!&]@T!%\nM%GU`/'B.$1`[6F\\%B0%KS\"`QLJ0=K)D-AO##T>!P<T,`3Y*C<:DAH0M*:USH\nMA#8TA.V3)3&$=03+#.%-S!#ZRI_)$/Y(4L:UC)+$/&]`';69,'&@J4^>[8\"%\nMR0U'/+O&:L1A&28S7&,RLJDLPU'ZSS/#`3>=[>AI%@!?0B>?&76><8:SP=KD\nM0:MZA@,*`@>\\;-VAIW)4DXZIH)>I>:\"4;XZ4-W*E'+SI#!5A@TM^AEMUW2+[\nMG3M?GQ);0FCIN(15WMU2)3_<E84]XC$2_\"-'@F\\OOGZV^2&I!9'X,O(]->!8\nM7_]#Z)\"<9/6_WJ&B:7!6409?ZW^[^&S1GW[7_`Z%,?([$&P83^\\=ZM6(GJ&]\nMI>L8>+9R=0M9NOVPTA^PUZ:33XA;)7YK[J,]HPVNI=]!U_@!$Z>*G'ODSS&Z\nMM<U/V/5,_PF.G.B624#;I@[S\"82+T[,/-Q?G%R?O;RX^?F!$O!QA?89=DB'!\nMD.TMMSF^/B'YRUN2:C1-V[!6,TR\\.3O8#LVI^V`Y4]UJ-IO=T[/S][>7-Y\"N\nM!?Z-A0,T(UZ5_(/BT,ST?!XCXA4?JI]0Q3'O6<V'Y$9W-R>7[0[RW15NQX83\nM$#\"4E7N:A>R?8ZC##^3:OZ*%]J]I*CFK#0[[7^W_KVC_@9ZAO2O'\\TF*,?[;\nM9<M#Y'^4X1C^Y'[AH#;7L\"\\8.V$L&'MS@_B?4<1*9;#.RC5P+?:O':K!];_^\nM<#A$2.VIA^I7^]_%AR@L2<^SA=VD)YO-D^NS]S=GZ/SVPPF83Q\",[@R8,M?M\nMF87=/2@[H.NSF]OK#V/D+/7?5Y#<H_=CU`KV_('M1\\6\"%@RY?/_AY]OW/Y^A\nMEM$Z#K'=7-^.;\\Y.T=7UQY.ST]OK]Y?\"0`Z(3A^]_W!Z>7:=214'?_+QEZL+\nM&-.ZNCQ8/H#+.?ZZD9#9/R^K;,\\!Y-I_/XC_RD#I#\\#^^\\.O]K^3S]]NSZ[_\nM?H1V:N)QG%\\-_9G9/^P\"A+I5S?F_HO64?MC_HVH]8O_*8-#_NO_?4?Q'KS?X\nM0/+P&K'_;Z\"@Y]%<'U+RI>X;\\WL36S/6@D'R\\L`HQ5EA&GZP\\MP#RR%6?6!\\\nM\\ES'\\0_8CC2Q0:6Z2HN'M-3:^824KD(+D0>]'PY4*)0?*<,C]1`MG,6_3))#\nMGWU>DB0_PKG)>INL(HPH`>C:<19HCU[)(00O;*=AS'5W[X=VAQXSG,6\"[%(:\nM/O[L-]O'X>25;1+?2;*N&?Y,8=RYMH,<F\\%;>:;]@*:^BS':8V#1=`F`[YRE\nM!V#B1$S&EN-S*CSRU=87F-&A]C@A6<3!4,NT?TL,A2Z5V&$IX13O'6`#TAD5\nM,=H#8M92?W5.58213TD';I7@'P,0TL'AQ0BAYP\"(A(1\\!B[O$Z1MS+VK&/>N\nMJG+OTK1Q#NESQ\\;V:C'%;N*,R-[2Y`/BB'Q*1GGR1ZNI*'F*5>E+R;,!J->`\nM?JD'$O-EM!&H(66`(:T+:VG*UX;Y:II!+QWL!-1)U$1.=$P?1D7U@8,287#J\nM)&\"\"L^M8,'[R?+S8P\"89@'`U'%X9F[PXUXT<Q?:>O`1M9JZ52BFF^$*\"&?8*\nMYCAR\\LPQK=7%B:3@(Z.CR`H1N<TH[YH/1+W#:AUTTMHS%%PAS8K_;&(-20`C\nMY8ND`=MAZ?OS&[)9N;TZA8T0C__T3&,?7=S3L(U(]\">RM!\\(HRWLHT>H#5(_\nM2#AM6<[CQI1P30LDB/R'.T!\\IZ_VVLC%_LHE4F7;/:1[J-6<X@?3Y@:';/S8\nMY4G*-V^18\\V\"7T0'[+#'9;6<`0Z6)'AD&7S06W'^XQR[.#H3P3JF<&BO\\SW[\nMSLB\"R<=PE>.XV0IKQM&F,;(BKBG\"TI!^[Y,#C\"X*,DB_0'^Q;LP)(8\\(?\\;&\nMB@`(M3W&G>/MZL'IV>596@]FF(@=BW(WB8SFID<Y59_T9[G2YY3=NX1K3+(Y\nM$N12(X>K26W&I<805Y+:;(M2^^GL_./U&;KX,#Z[OD&$@,B0*3L\"\"9[,L?$;\nM$9G.U)Z09T*O1@V2>P0%N9N:A2S7<%:VO_>Z30358T*DG$S(4+#/F$6[NNEA\nMLA0#+RGV5HO.?H5F#O%4-E$&NLQ6JR.`V*8AAVM%4TRD#S'2PZX/<HA;-5/-\nM?`41F'=<EX-GNY-`,2Z)(Z29/[^.16R:;9CK\\^Q+BJ\"$;Z?I!/?L+&?+\\.M7\nMH5]G&R6N-_0KTZ?P>`!GFZH0+BK3J_,=8;X\"\"+RIS;/'%>`:+YQ/&$6N/5\"#\nM1/)4DQ*4<_%7@HN7B+.Z@P])RG3QI46X(S=_);KY,]NC5(\"?=U:N6`\"C/I'Z\nMPUK$F>/U9]BP=!?S0H:+C0;YY[BS0&94E,3LD.^@I7<LB-\\C4C#@?@TX`S,)\nM.Y@N,($P9>\"JL/2Z3\"V.0_]!_>Z*V'A.Z+B*6/7MJV\\S0D@\"=H;;`,)+JES1\nM`')5,(`L=Q!`V%9:V\"-X;&]-92#N$X2;RNJ+)@SYMJ,)VWU#-.%;_'0XB4[4\nM$4_\"967&$U[-R-<'@3N[<$:4;8%F`/=@%QY6!(@%7YQW.=^ZK%)2ATJ8H',%\nMG5%4OSD.*CJB=\\KV0VQ0X(BX^L<<$>@*7VAY5\\0!9KNA%&11WYA&'KTE0`BG\nM6RWTQQ_I\">$@]H4,:;7(6-E)F&_&G:J%[0=_OD>'M-$[I/;REL3M*=`$6)KO\nM.,AR[`>T1Z9#J0@M],]M6*)D<70A`82`PDWM+-25/`_,5I!O<8+R;=D#IPW.\nM%;*ZT6H:&!Y;P@'+7PYX<&7)'=%_%_^^,EU<2U8W7TWO]`)6-T]8W6RU6#P%\nM9>OCN,\\F8)TEHBI-&4\"TFBI;@TX\"10@0S_ZU(IM:6.=>X*L[J,=,AM7MB4@:\nMHJXD52R&CG$Y1+<V?#2$V#$*=R)!H3L5.Z(34>QH!(047!B=20]67B'3GD(,\nM#>B,X^T)Z&C&+:(K9X@4(0]V21N$;_%L'*ZIY-LB5\\>MV>%XM81;1025=Y`^\nMFQ$3HWNH</\\4V&(=UB7(A%78.RBXW!-\\:3<#^^,'9`:H6Z;.JN4O%28[(D1^\nMA4LXIX;^-WU..TYE6&P0TVW^/1X=F+[T,I4SFO^C;+ZX(1P)&\\+(J.CJZ#4!\nM=D&,A*@0;J2P&03`DDPH;D:HOT<*ZG8%NBS'68:T<$VE60&_:A>$J$Y`4(=3\nMT0DOL[2;C<8GW5J1V+[7:HW&76IY$/KF;(+9(=;9:K4C\"@%K+-;UUAM8[:G>\nM2-QW+EW\\\"=L^D\\E\"MU>Z!7_,Y<K207O7I81D^9Q1++H[M9A.B?UI,B7\\3'_3\nM]8;YQZN;OU^='0=78O,S1C8HR!C!>Z721<Z$\\NDBT>\\,EE,_!#0'F5W:T#B_\nMB:4I(%OAR#M.<#?3?K=$05ZD#>R9!]O@9Q1OQ0+RNZ\"`G.)M>FZCT@H:XA(:\nMXEHD&3A1;):!DRBWV!-6U*Z<AS/Q'!U%REDI&^<9RO/+QN=%ZR&C@O60>0WU\nM$.X4A=2[M\"NLQ[O-BF3>6_!6@D7E>BM);LAG`:1\"+F<-F$K&7$DM9X%6)G+1\nM<KI88UW8CLKX\\0M_5'!$,^_-SW7HGS'_+7`%1:IPWFKJ^>Z>Z$$Z2.D@M0V^\nMGL#3W8=/_]/[W]PZ2>!^%B0E1A0%>C3].7H%?BN\"L\\VZG+#4N(\\J6ZD5>=:Z\nM&K?:U=!-JJ\";5$;'NNS*KNZR*KJB19<8NHOSRJNC[4QE5S=J[:2V&EVAHSH>\nM[\"4\"0Q]C/SP&:35>+/TG1.P,>K.(T7VXO;Q$=-.!'DP2H6KT`T!\"T6I\\2++I\nMV2NRPIC)QP;0G\"HKE=S8F\"GX+5V`B;%@4Z23+X&TI,UM!^GH2ZRTI+F'2'=M\nMZD%+M&CJP;$O:.H!\"45-/2199NKA@'I,/02_/5./6+`ITLF70%HBFL>0;M4`\nM)*T,,X>GV0=AC<WSL3Z#-)Y7A4&OA&O.-?8UK,J$,[%H$16Z9!5,L:4E'!V?\nM?-QLB'7&X%Z11D/L\\>[PWZQ'H2.<!;$%OT-7TFBT45\"\"Y.=B^;AP+`Y1C,GB\nM,1'R<7+C!O:^U6XK+I9`P3=JEEC5J=.3YZ?3C[7K]*2T3D_6ZC3KW-RF3L<A\nM/@N=?ERKTT7CPV0'.DW#QC/STQ;<\"U:KGZ:QLI2?YC?&2?QT=$]<<(C?E[.1\nMKTY#A<,)R+OWUUP\\$G]=,`V)A%R?;L<ZF)Z';O..DAIUFVWYRNAV<&]<MF[S\nM=J/@IYF9DI11ZP3`J\"4IRX/O1*,#H61K=,EFH5HU>O3\\,I#YMC(0?M5RW97\"\nMO*Z!8OH^6IN?!!?ZQ=/.)OJ>`!A=^OMB^CY?FYV4O`12;W8\"59UGEIU`5*XW\nM.Z&EK%+9\";]/.%NC$PE#%1W.R#EVJK$!TR4Y1\\':7R2ZG79S>M1!917_Q%L\\\nMH)7&62XAY8-77;@N-GSK\"2T=>`@TG0GEP54M.AW05:C?LTI_9]ABE:@>?L.*\nM@T(?B-BW&)+E87]/G->)*6I;[+>LV!&ZAD\"PNZA\"*S:LQ,ZDER)=SLJ&!8FS\nM.S'K;B<;5(JQ;@/VQ=MA+`_'<$I=&'3EE20J<!NEB2HMX'A#K\"#@'+%5%EF2\nMQ%)-M!E.+[+*PJVT1:ME@L'+KU=7P#^I@'^R1?Q%=Y^Q]5^VZNTJ#AHE`^_N\nM11=\\A.=9$,4T_>@>$=TB_)AQ_P]0``*T^V.2\\[!>V%HC`:A[T)/,_A;H1%X\\\nM25J1IU;R!.]#7CSY3\\O@F28\\CX@>7M)O!XF#T<AH<V*3:<,:ZT)A^'G_200,\nM1O\"Q?_P1C&4DQ<9\"_QT,!Y=P=4GTHK6^:X6D/JV6<'W&?\\28</'1804A`$:D\nM&8H4WE-(E($HL:2)DEU0LZ&/^.J2F$4+.H<G8R\"D78\"2F0DY`X-\"4N*(!E\"?\nM\\'D/E)0E=A>F3Q1)0@GG%F7$.,D&L;M,O(<S>R/&9'+,.D'CC66-[%:@Y#W_\nM'C`B!!-Z60(/6LQ$3\\YDFKZ+([J?/!P=#I83G/3GO3Q638JR:O)\\6379#:O2\nMQB73JNQ2[;/0*DI:#:Q*$V]:%G[0+20X&^($;>@1B)KD&'MW<2M!5I2#1XF[\nM\"%XO!9$J'N'(RGB0(ZKED\\4$VQJ(CDMLF/>F47](8SG<%PIJ&X2P'?OC4\"WE\nM1E'2T]($^\\_D9S=BP607+*C=?VZF!5F><3,6U._4RI5RDBV;SZ64$^ZO=U3*\nM\"?OSUF^D4]O^\\!)]W:4<D<!@QYY=REFS%.ERHKI`M*\"BI9PJ^/+9MV$I9Q=$\nM;;>44U8#BXALVZ6<R\"JW7LH1#%Y>RJF`?U(!_V2+^(M>=A7Q2^\\6J(!_5&']\nMHRVNO^A%C9C\\1SLO9876]^Q*63'W]&]=RAH1O2A2RA)Z[1.EK*I5K-&(5[%&\nM8Z\"A2!4K7KYB-S!6Q4[\\`<5^<?X%L(^%\"MZX&/;(F\"!UIP]\\HQDGK]_]%<MW\nM8O3.WU2&H_\\=RW=%6#79#:O`L`JQ2M*-]AQ8)3P6KDY6C8IJU>CY:M5H-UJ5\nM#E,R7Y7=(?,L?!4EK096R8O\"0MA^WD7A>*[X18O\"B=W@UZ)PY7)@6LV+Q>^_\nM4%&X``LFNV!![5%Y(Q9DQMMMLZ#V:+L1\"T:[T(+:H^AFOB`K/C[W2P.GV#-<\nM<XK#2YN>.<.P3=3%IW+7$JOX?8W!-<Q/)G[<2T8I>((4\"U$__DAB+7;?O8L%\nMJUB884$J*Q;YGW2+/RHM^U%\"^7$DB(*Y#Q'*OB4\\?84I*NBF`.QG06`K$T*G\nM\"\"\\90/G@[.MI`0,;*0XV&'<:Y:^Z4>%T19*.0YLBS&?WRE]%?3O\"(\\=@DG!#\nM6#NZE`!G>'^VI&@?0>??Z./)H$AQ#`]0RS@KP)0-:;7:Z2>I\"0)B(YGU9DI:\nMX/ZFL3O!T8C<D'5A<9^OW;39NU/H$K*6%\\YD=Y>V)<,(J/UW(91HW7\".W00:\nM9IK4=&.T2#ODX^O9L7\\CC+;]73JW.(?DSBWVW@$O^=X!KYI/6WJ[\\&H42S<9\nM\\4KX-2_IU^(0DY[-LS*V!DF[B(-HYZES5B1*@,A<NT!+.CN+2`I<6/B8NE&'\nMYTNC,7%!#2:5T;@K\"\":.'D8A>B%SU.4#HN>1'C=SO\"$MKDK=@?BP1B&!C`B3\nM4I5DY9HW3L2G?3FSI[VBAF/;Q$K(#K86TW^L;/JI;,;;.)N9U)G-;-/JB]K\\\nMAEN`M/XFLI-$8B)>.L_-2-9F([(!\\81$#H;E),D-@\\SRXER+ML^-6/+G/;G.\nM8X.]IN/5M?-X0PMJ//V\"4W1G^^KZXR1ZD##-1;/31#HG9R>>$$#ZG14\"B''&\nM*RLHCNA-$0F&T>G=Z$&UA`JTEOL,7/#BB\"#Q9&\"VG7K&H-:3?`I95D%WNQU_\nM^]_$R:42JQESPO2MJ\\'.$F(8';*]6UC!P:(K]CJK3PIX5:Y05^?!0X'@JW#S\nM9U:\\%T:T`49`<3@\\[M\"3P^F:V,/9HW3KZIP%-%CIN>EZQ!&33?;*M*BO17-G\nMY6&TCSP'C@>M%7/V2D%@3U.\\$Y:^-#!\\''NOI\\`5V#/;=W7;P-`<L6ZP\"H,_\nMWM^;^4.UXD/[,/0&&W/;-'1K_6B%$5P$L,+(/7'L>V+X=OYP2O(U)F&$G,@;\nM7([H07&BA_&A7.YC$LZ)SC_2QYOY.GN@7\"(/2<F:1>T0]F3<)?+N*CI@X)*G\nM_Q(42:9-*TU3JV%3JV'3JF'3LK#ESE-#5JIET*DA*TM.4ZMA4ZMATZIAT[*P\nMY<[30E9J9=!I(2M+3E.K85.K8=.J8=.RL.7,4T(#5\\H8@1(:>.EI:C5L:C5L\nM6C5L6A:VW'EJR$JU##HU9&7):6HU;&HU;%HU;%H6MMQY@Y\"5@S+H!B$K2TY3\nMJV%3JV'3JF'3LK#ESAN&K!R603<,65ERFEH-FUH-FU8-FY:%C>=.'\\A64$B+\nM:8[-.BP\\NI$@!TTW?)%`#!E_[6J([>J\\=T=]R(1W&T*&[V'8B0D(;P@60Y]:\nMF#4X$/@D07^$!HXI1@N=[%\\L0HS;@92=G'Q\".CRQQ>:)'&;OW;7I&T2;Z6>A\nM1L2,NX\"\\JU.*0M*RF\":;J5:>J56>V:\\\\<U!YYC!S9B0N(A^00=\"(_0BO#`P:\nML7.2;`G*:5HH4=)8C.JI*@4Q+0I\"DX%0\"U/1EX(H3,5`!D(K3,50\"F*:\\IH2\nM&(9$)%IQD1BJ%$119AB:#$1AD1A]*8C\"5`QD(`J+Q!A*04QCUA5[O7KH!]F;\nM.F?,%0;>MG>GLDJ&R5[JPERLB_D.'OK-@E>A>W-G15]1SM^UG.NW?Z7D0G66\nM/78BYK@+K->/_.RO)?V/'SG:\\E.UZE/[U:<.JD\\=9DXM-'=:G<73ZBR>5F?Q\nMM#J+I]59/\"W\"XI0)*,G41?%\\=&\\YCEM`/F1LF&@H,8OG6\\Y<TBD$50IA6A\"\"\nM)H.@%J6A+X50E(:!#()6E(:A%,*TH#\"F$F&HA84Q5:40\"C)BJLD@%!7&M\"^%\nM4)2&@0Q\"46%,AU((185A2(0Q*\"P,0Y5\"*,@(0Y-!*\"H,HR^%4)2&@0Q\"46$8\nM0RF$HL*8280Q+\"R,F2J%4)`1,TT&H:@P9GTIA*(T#&00B@IC-I1\"F\"8VV(^8\nM[IJR=DQ(T?=5/>PHZ_44OK6\"N5%FV$47-KL+@EWKX/VPCYC=5:C;\"+NN0W;-\nM9,\\.,_ES(^#N\"'K+,!\"S]B$KT?6/]/5D<3-]7`S,=\"T8C8#A%Q`S>BA8SQQ,\nM@(L.R'%G)`>>/J'H\\;7QN9E]5_\\GH,N$4609JOZ76$:.--1GL(S@2AY[.V5P\nM@SJ4>M3IOC9%4WH3$-DL^>:\"/\\`7KGF#13!B^=.<`8AS3X]30KKH=(6#8@6_\nMI=[CMY![*XO,8[\\\\@O@()O]_>U?;VS:NI>=K\\BN$`H/&]R8>FY349CJWP-YB\nM\"@]08+KM7;B+Q2(P;:?QCFME_3+-`,7^]N6[*)&4^\"([N7.E+XEM\\>'#<PX/\nMWWD8&IGYHE5H=X]K&\\F)%4L>0J>GZ@N<<CM4JI;Y\\AYE<L>@AG)-[Y4C3-X`\nM@YW7D[')F&*@?YIB9$]=&ZT#'U`;^_N.?/;ET`>X#]982A\"<$@:G3(-39L$I\nM<V-*EZ0H6+@H6+@H6+@H6+@H6+C((ERQ*>B!];NV)*8YK3*D65$FW>C,FMI!\nM8G6$U'C95I)7GE=WR?$I!IR1J9:V5N:I[37.^Q=E*\\M\\2?8@K?F]$CNV98^V\nM4<K*#?V-G>_8#?G4/5L]^;JD4_;%1EP<P]O#4@:[8?)OGV>KC>AXRI9R=M@7\nM7W!;.\\=4_DCV2TZ(EI&>M>4RH\\M$)&7UZL:A-O=(C[:4*GPWY!M+KD9T/Q(O\nM*,F.*E697JP;@P&)ZO]*6<E6YQ@=TD.>'NCIH4OZE*>'>OK,)7W&TZ=Z^MPE\nM?<[3YUIZY%3^%SS]2SV]4_E?\\O37>OK4)?TU2P]T_2$7^8U'/+VN/^0BOS&7\nM/U3E+UHXA^1<_/\"%EMQ%>F,N?7BM)7<1WI@)'^:@GAPYD;_FR:&6W,5TP(@G\nMUT3G9#E,XSAYIB5W43Q3.$ZN2=Y)[X\")+F-^X^WL(5F*/9_>/@1<<RP@L,KN\nME&90W-7__8#]?_&%^_#=97*_/M\"QS1>6C9PVD#M[Y02!<E,X?[$Z;*F=P=KI\nM#IGNX2^+,!G>S;E+GJSFN)W\"K=!F,=LNY!*6F-!N/)ZO.NZV\"85V.J\"!#O\"C\nMDS7106YT8\",=X$$'@48ZL)7.[>R!:^O-;(,;>/Q9490/E1EHG>NH]4QF]'JK\nM&>[$;)9BT*UT*;@9S\\C9K'MR=)4N6B[)F'FSI`5:;DG0,ZN-KV1'!>>C&2XY\nMY\"4E03GBMUA?N=JU&Y/D^/MQ7JN]_.B#!\"FVJV)#`=[4*SH[:\"'?_.4M>4V^\nMC\\LTTC8@-$I]PL5**`\\;:P@\"S^M3B7LZXY=@E_);PN^++<5.7JWU1]^*;?LL\nM%[Y$_#>Q/.TP[FQ$`$T3.F]7&]IS)#.@L\\V>S5(Z[\"3BDZIB'7NV6B<+-I%S\nMV*S(;;2KS6+Y8)1S;0)L4V!W2VC'36Q^^C3D-^;XX]3J6P0C%<F#D1Z_3HIZ\nM\\I'(6:T_8\\/(JS41&/%46G`]AXJ$!<+DH=0E[QI(O<J-]\"HW1!`W^Z*X6>-Q\nMGT#\\KG^>\\C/\\8;>=_[!?[O8_;)>?<3=F]\\/RX9YN^_N!'Z@:%H=]5!YC/$Q*\nMT^0[<HX(CNC?,6!_\\7_YF/R8Y/DH'>-Q1)J3[[(7+[Y+1J<0P`&WK-LD^8X6\nMM>&]!9K]&?7_[__Q\\X?__%'.:I!=HNP(T,5Y&?R;'=5].;@\\5\\(3TK.ZY[B&\nM5R'4QH(BW6PW!9D(H:B''3TWM]\\NE\\D%GXQEIXAOBON=!L;XB/CD:JQ$?GIX\nMQ#F9>`K/77M5-#'RZY8RT-QO6,13<<MRI1BR#6@M\"&]T64EH*8@,O:7*.P>\"\nM$D>M<&)W^6\"H)C;M8F7=#H5E1S)]7Y'I^RB9\\D#?3:4HKV&I_:(*/;`D)/NR\nM))1,8$E(]UHQ#4I@G%J9;NCH4@0#:*&)L25)DH]N+.WTVLV%7]U0H\\ZN8A!$\nM+7;4QK]B,),P@U&A1)AB,YKXM54H?$`36Z,9C\"P?1_6NT2(P>(.*^!%SA>:J\nMM8ZWD*>Y2NZ,0VAEYG%Q&PJ@5P1?OC23LLK2++WYJG=3DL;G9N8<8)@W>SR@\nM`_]4N?Q!6P:5RY9*^MJ\"9HD5',>WII'R-GQ>0'X5?C66+VW6VV^]ES)J%^2B\nM59#JB,>XNEL7AAH5(EP`\"RZ`ZKW_7@)8-`N`7;*\"5DZV-\"\\.F_W%7P:XO\",F\nM\"TJE)@K%8AKO,W_^G*:N7[!+IG9+B.Y-2Y981(JN!UKP#6^AB+!)TO?L9@7W\nM6JL&8='#@6O3!G+\"1J:M3>0(G.Y%*HMFK*^\\M^@0?+N4D(L@_6JMNJ_\"(I+8\nM.BN)&6NMMQA::NY]2\\VMW(5V6[\\+C1T^(2I/Z!5GULO/;M6[@%@1*BOD]SM^\nM.4CK14A:]5?F<)]]_\\S@!FK8QANQ`E7EZ@1<8^S<NSD!MJS?M1-0[LUEG2W=\nM\"Y0_',\\-R,(9W0#O8CJ$JREEU\"3*%9T!=+/_L@?Z2O1)U0IA-GWVDK#]G>$.\nM)1K#2MR@Y&O]'-!L^1IRY<XOFOV/[$*J(;MH2TL@7]J5MS@-Q0UB^H\\D_:I:\nMC]?+S>?]W<6.W<WS.@&CMB)Q.Q2=65*T?5$D9'(VP=WDA/1MDR^SAP$IHJ5P\nME1!D@F$W]BDMIJW2NP:64DRPR5+QV,XIUMY=S4I#8N^=J0'-:,:+_SGL]G3D\nM?B%\\PF4R8B;&1O2#\\$AZ9XUNRGRQM1CJ:CZJ_*'T4>5%;6X%HRGIE\\$EK(:2\nM:\\Q7\\*SF.QJ$A86S&B[-UCTLV0$YV\"XWRE:[54HK0B\"(>17QCT,PA#M++`0L\nM'CZKI(=#P)K2?X.OM#:2O<2LAO]OND#1?\"MZF?XG6WIMM4LSUSL9'Y'-/&%G\nM*7$;KV4G15J1<R%EUG]-QLEPJ/!:%\\5]&=!06Z^3<P>7@M`E9W$IYRD&YV=G\nM<E6-K$M3FR9.^(XE6%W2F_F4VUE)KA6O._(PW8H1>713Z\\WT`_U,RRG;A._+\nM&Q1Q>=M;<?:2O-,?UPVM\":_<_^K3A&--?YEM#K,U^;.Z/ZQG](?BENG&<N.G\nM;.$*;'-C4G^5;UYSPD.C)7?$H,V;\"\\OF#EU\\K(1#E7,!K\\5<@\"9;/>U94`G,\nML4$;>D78PI7K1Y42#8+[1NKMOA$])-X*/M4>TIWKL,@U].6=V[\"(O[9PZ21U\nM4.D5PW2\\OE=KQFO195MK;@-,4)V(T.Y\"*#<LFFFIJR:-SN]^$S;J,M3EEQ>K\nMILTO+R9.\"./-MI]__Z_1?[<.JD2]^()[+>QZ727258G3_>!7*7\"U\\OC.)*B2\nMX^%ZPS.=AF0ZC<R4K4GZEO1=7*:N8[9*IBP:<$1)'2/P5DLZ>>Y2=4C'S766\nM2&X96^TV!W9\\HZPEE1=^_)OE*O*.[)]'F>MD3JTBB&ZRGCY>UIX&VF76D\\<K\nMM6<-,6==JQUBQX1K[9!;/&VU0[YPS-HA,^FN=I2\"Z\";KZ>-E[=%FV+,V+(\\<\nM?%RH.NB1GZUS`990,&IB$NQ`.]MX<7YVIFXWN.2?^1WKRJ\\TA@/_+$J-/P[D\nM%EG^6Z7;I'Q7153;`?4[%?E5O<=**LP1%@RY<H151*VY')Q6N(]H\"%-O0Y@V\nM&@);^>[2$*J(3\\@0OC8:@JL_FCH9PCTY0G5<CV\")Z=;@$?A>0HM'*+<1BJ_X\nM9J0HKZ\"CDJ]KR(_E&;B2+)[!L94H5>VP@GA$@[#$;FDP\"+&ASFP0?(5/?%P9\nM6PP?6Z@!EJN`)E]Q0C,0JC&;@>?ZG-OD4[P9\\)G<IMG3MC4%-R.9-#8?8AE`\nM_;F(,9(:8#D=^LA&<M?8>'C.9[4U'L1I'K?Q,`?::FH\\^-Y5LQG4_'F(X@U-\nMPB.H68C>TB0X#C!+!3:I6<8R<EFO#UF?EPMYM:&I'AM+77>6M$@D>C7=945W\nM`T-@,]\\5_0:\"Q!3EY\\IB4.47<Y@O8W$.&U(@-?5EQ>`'I@!>[:*+$%]UJ6F]\nM6U;RM-9JLO;K24K&P?(EY:W@ZH8&1<$M:@M669UB)YL@RJKIO!/\"=2\"IU/JV\nM&?<`%M,`%M/.6;AVG2NR>/?<S5L2DQ\"[0^HA.JU[0ECP=L.F$!88WK`CA(6P\nMKX;Q+,]AI#*PIQ+]LW3*++$2*9/E7XF1*28!^;LDUB![EU&JQ].4,3O?O].C\nM@]<6IW\"#^?RY,ANV_[I<8B%^+2KW>\\B=@/3&S]T2J]6RAL\\F-#=D0\\?[=R0Z\nM]B6-DDUX#-J)+%;TYE`*4MPJ%,AN8'I%`PUV2YC<+[=?5OO]<F$APF55#23<\nM%E34W.5E&F$10FOKL6?F];[ZX8$=D8.$D7ZH'C4>^SJF47V?FO%F`/ZRG7#=\nMXXW:1!4;S_PIB&IZ&E'I5<MF5>;IER=A593:$42ED^?W122*K\\$N<$-.FY4K\nMX4R\\@5NZ:GV!1W+\\$6[^Q%Y+*L]N.I[^B';4_IF\\490(IJ<0P=&]3)P5F/Q'\nMG`B\"JKX<M)QH?\"R7F9M')]I8JHR`?N3QL4I0#(/,X^.&HEB+4PZVR@*YCH]#\nM\\FL77^3X^!2DNAT?^UJ@B\\J.,CXNJV;GXV.EUK>-CP-83`-83#MGX;JFH+)H\nMV3X6P&(2((M)Y[)PG5.MV$7+KK9*=?Z7GBN88%$YS!4H.ZQJ<P6ATP23\"9\\F\nMF'PD%!RF\":KS`VQC=6CFN++0S']Y>_K,/RH3)!^=,I?2IWT^>O4BG1'ATR-_\nMQMD1M:5K'XTHMQ3^Z\\V.N(AJ>AI1D6KE)\"K+7H2G(\"KE_/HQ135QM:K)T[6J\nMR6FL2F^B;+[*O&K])'P5I78$4=GGW)0VN^,YM]KXHI]S\"YYMT8W!K97[$\\VY\nM.8A@>@H1'+WMBA*!L57J6@1';Y.B1#`YA14<O:V)\\P6F5N34,Z]\\I[M8=_E]\nMM?QZ47?]Y&0R\\_L__520J_A?OZZT`!7?S3R_R<'O?Y^M^1%\\\\]G:=N<LFI;6\nM4[7FLS;ZK'@Y[Z8!7)D06,F4]DC%J[=*_&7S&H`0X)DFP3,FG3/_E0*JG*%*\nMZ94T5\"Q\\=@CI?;DBKQQE)XF4[=N#<L:7_,*WZUGF5DMT_A\\]]DZ&QZ_(P7S#\nMKPJF[14\\EM9/Z\"L*8F^R*F'4M\"+]V`:Q)M&2KA2=G(/E91<!66D13,63*=D!\nMBH'E-0QU]5JBE.4FO[$3#K+[1JMNA8MUPV2U/*%.HYJSW6E4KHO;U:^+VX7Y\nMBOO=*;P%S658=\\\\>_F)7]Q=5Q+K'V*T-_=BZO54A!FUF8O+P-0ACV14N>E>B\nMI\"1<@[P/87+)&_?)1URUSYA6)A^'BF*JV9.W$KJ.,QGR%_#KE5L4&KP,G2^S\nM5C/U<@VEMU,2L[*JB[+AHL!J,O_J]#6X.FDM[RZZY9T>L^7MLB:YUJ/(/J!N\nM$[66M-:(JJMQK:UG8\\MI>Z':>-IA6/M9[S':K+DJM7+\\=%;IJ.S^V!9?S]B%\nMB]]_*+[^@\\X[\\*X\"^8D.;;[_\\.NTO$R)]IO,71J:IF4H5E.`?I>B`O'1<)4B\nMS:.\\P;`F,)I\\6%[6@UDDC=)G<.)\"0]%)8C!==Y,JJ,?I*\"D]@@`71EQ5\\OY6\nMQ'?\"_HFKYOU;<428_*N<`#&U1LH;`X)!?MVM%N7K5==8?YTNH;!+S\\K.P/NW\nMDJMZL(7>\"RTO+..QSWX6X?H&3FEH.+E?;V]7SBF@=PH:!_$?R_G=AL1.=4K$\nMXY9Z9,,CE;XI-K>XOFR<4]'B?%AB)XQ_<$P35*#,NT!Y:XII-195&9/P,A'V\nM(,(]N:=&,:E!5-X@*F\\8E3>,R!M(F8.PU\"@F-8C*&T3E#:/RAA%Y0RES&)8:\nMQ:0&47F#J+QA5-XP..^Q]\"WC@%HREKXE-#6(RAM$Y0VC\\H81>0,I<Q\"6&L6D\nM!E%Y@ZB\\853>,\"+O3,H\\\"TN-8E*#J+Q!5-XP*F\\8D7<N99Z'I48QJ4%4WB`J\nM;QB5-W3,6P\\L.V*!9:=\\HQ89J^R69'1F0ZCFKX15O4Q*N\"8*%@`0\"P!C`=)8\nM@\"P6((\\$0+H6*C'!/9\"`%0EY(D$;$O#EE%J1?#EE-B3HRRFW(GERFEMT![UU\nM-P=6)%].T(;DJ[MY:D7RY939D'QU-\\^M2,C#?WZB*&3FE)U+#G.@^]*#?@JK\nM_/O2A08CP&B$-!HABT;(8Q%0M\"Y0M\"Y0M\"Y0M\"Y0M\"Z0NRYL,>_+KLEXMT]N\nMUT6Q=2:!4\\BNR;A2S?DXU@L(6(&0'Q\"T`0%/1JD5R)-19@.\"GHQR*Y`?(V31\nM&O#5&@)6($]&T`;DJ3646H$\\&64V($^MH=P*Y,=H;M%:YJNU.;`\">3*\"-B!/\nMK<U3*Y`GH\\P&Y*FU>6X%\\F.TL&@M]]7:`EB!/!E!&Y\"GUA:I%<B3468#\\M3:\nM(K<\"J8P:KS0HUWCTU65U2.Z'AAK18(G&ET3_TA#CEZZA),5VL=PFZ(^DO)5/\nM8BO/-W&/[S?U6[%&^DUR59[S*\\/S5^T?XZ_J<RY%J?+!WPAB^O.M%+#\"1PHQ\nM``?J.,\"#C^E+B8.ZP8$=\\8&Q?\"YR<L9V-S@W&Z-QX]S_*6;L;I1LN;\\;D_0P\nM2)-U?:.#5(N83!9<H@!'%*.P31;JQ`494=).N&2=H.0Q*)H1NOA7,.O]ZTG]\nM:Z-?G'7DIWO_VOO7D_M7,.O]JV__%?3^]5C]5]#[U]Z_/K)_;>YY]O[5R;^:\nM;P]4%L@-#K;<M?G*#RUO0(.SWEO_R;VUJ5:%>&M3C>B]=>^MG;VUR78?RUN;\nM[#_46S?Z5]3[U]Z_GK0W;/33O7_M_>MC^%?4@7_-^OYK[U\\[[+]FO7_M_>NC\nM^5=CS_-(_K5MVQ^H[:@-V_>W+S?^`>_=BPP`Q`+`6(`T%B\"+!<@C`5\"L%E\"L\nM%E\"L%E\"L%E\"L%E\"C%L3J\"*M*I,,B5S]P707/J]='\\-VXKYJ:A*?O]=G.WDH#\nM2+94V?W2V.3UV;9>7Q1D0H&>*,#()?5%,7+)/%&@D4ONBV+B@OQT!,PZ0L`7\nMQ<@%>J(8=8127Q0CE\\P3Q:@CE/NBF+C,_724F74T![XH1B[0$\\6HHWGJBV+D\nMDGFB&'4TSWU13%P6?CK*S3I:`%\\4(Q?HB6+4T2+U13%RR3Q1C#I:Y+XH)B[[\nMVB@$M]:>LQF\\V]D)\"NP$)>T$)>L$)>\\\"!76B(]2)CE`G.D*=Z`AUHB,4JZ-^\nM-/_$1_/E67,?%$0*>*&8NT`_%K\".4>J*8N61^*&8=H=P3Q<AE[J4C:-'1\nM''BBF+E`/Q2SCN:I)XJ92^:'8M;1//=$,7)1.PT4)<ACJIV&&!38\"4K:\"4K6\nM\"4K>!0KJ1$>H$QVA3G2$.M$1ZD1'*$9'%WGS&L?4:T*K7UOKU]8<U]9,_1J,\nM`QQPT-B`@P)P@(X#0OA``TX(GU3'@2%\\,@-.\"!]-7U#3%W3`F8\\-.\"@`!^@X\nM((0/-.\"$\\$EU'!C\")S/@A/\"IZ4N;L_\\V;JNGVKK`N6G6WA$'&'!``!]HP@G@\nMDQIP8`\"?S(03P\"?7<(\"F+]\".@\\8F'.2/`PPX((`/-.$$\\$D-.#\"`3V;\"\">\"C\nMZRO3])6UX\\S')ASDCP,,.\"\"`#S3A!/!)#3@P@$]FP@G@H^LKU_25M^,LQB8<\nMY(\\###@@@`\\TX03P20TX,(!/9L()X%/1UT4*FO:]T(!7Y4+_NR&_=?QJ1*^P\nM7Y(K_-?)?+9>TZ5_Y9HXR\\X!`R#=+'\"E7#FLWA7G#@,Y#-!AH`=,RF&@#I-Y\nMP&0<)M5A<@^8G,/D&@SRD<T+#O-2A_&1S4L.<ZW#I!XPUPP&Z`I''B(>CSB,\nMKG#D(>(QUQ14-25V9KFC<$7!%QJ*AX#'7$_P6D/QD.^8J0GFH(Z\"?$ITS5&@\nMAN)A>6#$433I^A@>LQ2,DFDH'@;##`6C:#KRL1?`I)LQ=_5V]I`L11R34-<%\nMKCDD$)#E7D(G>Z0AE$K(R?!NSOWT9#4OOB2[_6RSF&T7\\O+0VJV\"C5%R5:?N\nM>)%3.SG00`X$D<N:R\"$O<K\"1'/`GAT`C.>A*[G;VP/7Z9K8I-@G^K*@T@-@,\nMN-[VHA(CX>8D+?KFW8%OH\"0Q@6A<HH0M->`?\\/?CW&RX/&Z4Q\"JVJV)#<=Y8\nM3)T%JY()?GE+WI;)EON[D>T2Z$:!3'B)23F&C6:.K#<TL(V-)!X32\\UV'8JM\nMC.:3$WQGHG&25X1C2CR?;S0PD^O;Y[XSP\"Z/7\\)SOBIE7%<OYYM%;$\\R\"7KU\nM.F%1X(2O2RZJ]77P35VYDJ_3\\*M7H^1\"[;D.R\"Q_C0(P+KS7*%RU*.+*1VT:\nM!6A<M;=)@;D&7##A&`:5.2NR\\E:30D9>5]NO@48A-2[YGU0*F7&_P$DIY,;-\nM!J>D@,;&G0HE!:#5\"-!4([*Z+8Q'L(T\",&YSJ%&@GIC$\\B.>.+EX_V90&X(:\nMV@<62G3L(`5HW\"/1)`78(`4$=\"GD;112XP:+&H7CVD)FW)UQ4@JY<6O'*2G,\nMQ\\9](24%>'P*P+BIY*04H'%'RDDII,;M+\">ED!GWPIR40F[<2'-4\"A?CE\\W'\nM)AO[HZ#C_JA?K_,8W='@WJJA-PJ2;^:=3XV]2+4'B]4=VG75&0`S`[T'I_8>\nM2P;>W4:=`30S$`LZH4;<7,$J#%+7+([&(+/)`#;LH!!:\".BPZ0QR&X/T1#)0\nM=Y@]CA80L,D@;]@W4FK!N\\.H,[#6A9>GDD%J8W!]*@:VN@#&IV)@JPO@6![I\nM8@P\\3XC3*;'=77'`/]S.5NMD<5@F^-W#9O6_^+_59K%\\(#-4/W_X\\.N''Y,$\nM.^%-L1>P,_SZ_7HU)[-6ORW_8-G,*HE=+Q[%J,,]Z02T3N`**FHB*8)%@8M&\nM&\"X?5KO]SBGW3Y^&J_5Z^7FV=LZ;OU_\"T\"CVF]7F<_)UM;]+/GURG>3T+_?[\nMF'*KN4>46\\)8RUV9C*V>Q9Y\\)$:G3LJ.^3'L:\"OSRA6,:MENBN3+;'/`Y<-_\nM5O>']6Q/Y@J*6X8E,EG@/BRF0GNP]`=OA-;I8EQZ9A3*C+'*=,=FI9^)]Y[5\nM#\"!H7GJYW2SW-_C=Y?86OWI#+.%F7Q0WZV+SN<Z`H4F#>?;+VR'%&K8!/</N\nMI4C(?\\D%&\"7SN]D6\"^MA</Y=__1/__1/__1/__1/__1/__1/__1/__1/__1/\nC__1/__1/__1/__1/__1/__1/__1/__2/Y?E_M_=+T@#@`0``\n`\nend\n",
"msg_date": "Tue, 29 Sep 1998 12:03:19 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Patch - please apply"
},
{
"msg_contents": "\nDone...\n\n\nOn Tue, 29 Sep 1998, Jan Wieck wrote:\n\n> Bruce,\n> \n> here is the patch that includes PL/pgSQL into the build\n> (currently with make errors ignored) and adds a regression\n> test for it. A clean build and regression ran fine here.\n> \n> Can you please apply it?\n> \n> The tar should be extracted in /usr/local/src/pgsql and\n> creates the following files:\n> \n> src/pl/Makefile\n> called by toplevel GNUmakefile and for now only calls\n> src/pl/plpgsql/Makefile\n> \n> src/pl/plpgsql/Makefile\n> calls src/pl/plpgsql/src/Makefile (here the call to\n> make ignores build errors - this must be changed\n> later for the final release).\n> \n> src/test/regress/input/install_plpgsql.source\n> SQL script installing PL/pgSQL language in regression\n> database. Will be modified by .../input/Makefile to\n> point to correct PGLIB directory where plpgsql.so\n> gets installed.\n> \n> src/test/regress/output/install_plpgsql.source\n> expected output for installation script.\n> \n> src/test/regress/sql/plpgsql.sql\n> the main regression test. It tests functions and\n> triggers written in PL/pgSQL including views that use\n> supportfunctions in this language.\n> \n> src/test/regress/expected/plpgsql.out\n> the expected output for the above regression test.\n> \n> make_plpgsql.diff\n> patch that adds some lines to\n> \n> src/GNUmakefile.in\n> src/test/regress/expected/Makefile\n> src/test/regress/input/Makefile\n> src/test/regress/output/Makefile\n> src/test/regress/sql/Makefile\n> src/test/regress/sql/tests\n> \n> so the plpgsql shared object will be built and\n> installed and the regression test for install_plpgsql\n> gets patched for LIBDIR to succeed.\n> \n> \n> Anyone please report problems with building PL/pgSQL or\n> running the regression tests.\n> \n> \n> Jan\n> \n> --\n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #======================================== [email protected] (Jan Wieck) #\n> \n> \n\n",
"msg_date": "Tue, 29 Sep 1998 08:40:46 -0400 (EDT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Patch - please apply"
},
{
"msg_contents": "> \n> \n> Done...\n> \n> \n> On Tue, 29 Sep 1998, Jan Wieck wrote:\n> \n> > Bruce,\n> > \n> > here is the patch that includes PL/pgSQL into the build\n> > (currently with make errors ignored) and adds a regression\n> > test for it. A clean build and regression ran fine here.\n> > \n> > Can you please apply it?\n> > \n> > The tar should be extracted in /usr/local/src/pgsql and\n> > creates the following files:\n> > \n> > src/pl/Makefile\n> > src/pl/plpgsql/Makefile\n> > src/test/regress/input/install_plpgsql.source\n> > src/test/regress/output/install_plpgsql.source\n> > src/test/regress/sql/plpgsql.sql\n> > src/test/regress/expected/plpgsql.out\n\n Tnx - but these files are missing on cvsup. They are new!\n They are inside the tar too.\n \n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n",
"msg_date": "Tue, 29 Sep 1998 15:07:00 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Patch - please apply"
},
{
"msg_contents": "\nOops, sorry...commiting those one now :(\n\nOn Tue, 29 Sep 1998, Jan Wieck wrote:\n\n> > \n> > \n> > Done...\n> > \n> > \n> > On Tue, 29 Sep 1998, Jan Wieck wrote:\n> > \n> > > Bruce,\n> > > \n> > > here is the patch that includes PL/pgSQL into the build\n> > > (currently with make errors ignored) and adds a regression\n> > > test for it. A clean build and regression ran fine here.\n> > > \n> > > Can you please apply it?\n> > > \n> > > The tar should be extracted in /usr/local/src/pgsql and\n> > > creates the following files:\n> > > \n> > > src/pl/Makefile\n> > > src/pl/plpgsql/Makefile\n> > > src/test/regress/input/install_plpgsql.source\n> > > src/test/regress/output/install_plpgsql.source\n> > > src/test/regress/sql/plpgsql.sql\n> > > src/test/regress/expected/plpgsql.out\n> \n> Tnx - but these files are missing on cvsup. They are new!\n> They are inside the tar too.\n> \n> \n> Jan\n> \n> -- \n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #======================================== [email protected] (Jan Wieck) #\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 1 Oct 1998 00:36:01 -0300 (ADT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Patch - please apply"
}
] |
[
{
"msg_contents": "> > With the current implementation, I don't beleive it will make a\n> difference -\n> > I don't think that any part of the Winsock system is \n> actually hidden if\n> you\n> > ask for a lower version. But it is _permitted_ by the \n> specification that\n> the\n> > DLL can hide parts that belong to a higher version than the \n> one requested.\n> >\n> > So I'm not 100% sure... Does anybody have access to a \n> Winsock?@that\n> actually\n> > hides some details when you ask for the wrong version?\n> >\n> \n> By the specs there may be the DLL that doesn't support lower versions.\n> In that case my code doesn't work well.\n> But we can delay to call WSAStartup() after the first socket \n> call in LIBPQ.\n> My example code in fe-connect.c is such as follows.\nThis looks like a great way to do it :-)\n\n\n> \thp = gethostbyname(conn->pghost);\n> #ifdef\tWIN32\n> \tif ((hp == NULL) && (GetLastError() == WSANOTINITIALISED))\n> \t{\n> \t\tWSADATA\twsaData;\n> \t\tif (WSAStartup(MAKEWORD(1,1),&wsaData))\n> \t\t{\n> \t\t\tfprintf(stderr, \"Failed to start \n> winsock: %i\\n\", WSAGetLastError());\n> \t\t\texit(1);\n\nThis is not the way to do it, though - what if it's a gui program.\nShould instead be, like the other error handling:\n\n\t\t\tsprintf(conn->errorMessage,\n\t\t\t\t\t\"connectDB() -- Failed to start\nwinsock: %i\\n\",\n\t\t\t\t\tWSAGetLstError());\n\t\t\tgoto connect_errReturn;\n\n> \t\t}\n> \t\t....\n> \t\t???? for WSACleanup() ????\n> \t\t....\nYes, this is the problem. Perhaps set a global flag in the DLL that is set\nif we have ever called WSAStartup() from the DLL. Then we can check in\nDllMain() at process detachment if we have to close the winsock?\n\nWill you write up a patch or should I?\n\n//Magnus\n",
"msg_date": "Tue, 29 Sep 1998 14:06:25 +0200",
"msg_from": "Magnus Hagander <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
},
{
"msg_contents": "OK... Whether or not Winsock enables a 2.0 or 1.1 interface is pretty\nirrelevant, unless you're using new features like\nsetsockopt(somethingorother) [I can't remember]. Otherwise, the standard BSD\nsocket functions (socket, bind, connect, listen, etc.) are unchanged.\n\nA much better way to do this is to do a WSAStartup() in DllMain on\nPROCESS_ATTACH requesting any interface from 1.1 to 2.2, and FAIL TO LOAD if\nthe WSAStartup() fails to provide a valid interface. That way, we just\nalways call WSACleanup() in DllMain on PROCESS_DETACH.\n\n(Note that delaying WSAStartup() until the first socket call makes your code\nugly, and can delay the first socket call unnecessarily. Since we're almost\nguaranteed to be calling sockets, we should initialize winsock when we\ninitialize.)\n\nAs for any applications that might not use libpq... They should delay-load\nlibpq.dll. I saw a reference to it in the MSDN library... dunno if win32\nsupports it.\n\nTaral\n\n> -----Original Message-----\n> From: [email protected]\n> [mailto:[email protected]]On Behalf Of Magnus Hagander\n> Sent: Tuesday, September 29, 1998 7:06 AM\n> To: 'Hiroshi Inoue'; pgsql-hackers\n> Subject: RE: [HACKERS] LIBPQ for WIN32\n>\n>\n> > > With the current implementation, I don't beleive it will make a\n> > difference -\n> > > I don't think that any part of the Winsock system is\n> > actually hidden if\n> > you\n> > > ask for a lower version. But it is _permitted_ by the\n> > specification that\n> > the\n> > > DLL can hide parts that belong to a higher version than the\n> > one requested.\n> > >\n> > > So I'm not 100% sure... Does anybody have access to a\n> > Winsock?@that\n> > actually\n> > > hides some details when you ask for the wrong version?\n> > >\n> >\n> > By the specs there may be the DLL that doesn't support lower versions.\n> > In that case my code doesn't work well.\n> > But we can delay to call WSAStartup() after the first socket\n> > call in LIBPQ.\n> > My example code in fe-connect.c is such as follows.\n> This looks like a great way to do it :-)\n>\n>\n> > \thp = gethostbyname(conn->pghost);\n> > #ifdef\tWIN32\n> > \tif ((hp == NULL) && (GetLastError() == WSANOTINITIALISED))\n> > \t{\n> > \t\tWSADATA\twsaData;\n> > \t\tif (WSAStartup(MAKEWORD(1,1),&wsaData))\n> > \t\t{\n> > \t\t\tfprintf(stderr, \"Failed to start\n> > winsock: %i\\n\", WSAGetLastError());\n> > \t\t\texit(1);\n>\n> This is not the way to do it, though - what if it's a gui program.\n> Should instead be, like the other error handling:\n>\n> \t\t\tsprintf(conn->errorMessage,\n> \t\t\t\t\t\"connectDB() -- Failed to start\n> winsock: %i\\n\",\n> \t\t\t\t\tWSAGetLstError());\n> \t\t\tgoto connect_errReturn;\n>\n> > \t\t}\n> > \t\t....\n> > \t\t???? for WSACleanup() ????\n> > \t\t....\n> Yes, this is the problem. Perhaps set a global flag in the DLL that is set\n> if we have ever called WSAStartup() from the DLL. Then we can check in\n> DllMain() at process detachment if we have to close the winsock?\n>\n> Will you write up a patch or should I?\n>\n> //Magnus\n>\n\n",
"msg_date": "Tue, 29 Sep 1998 09:32:18 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
},
{
"msg_contents": ">\n> > > With the current implementation, I don't beleive it will make a\n> > difference -\n> > > I don't think that any part of the Winsock system is\n> > actually hidden if\n> > you\n> > > ask for a lower version. But it is _permitted_ by the\n> > specification that\n> > the\n> > > DLL can hide parts that belong to a higher version than the\n> > one requested.\n> > >\n> > > So I'm not 100% sure... Does anybody have access to a\n> > Winsock?@that\n> > actually\n> > > hides some details when you ask for the wrong version?\n> > >\n> >\n> > By the specs there may be the DLL that doesn't support lower versions.\n> > In that case my code doesn't work well.\n> > But we can delay to call WSAStartup() after the first socket\n> > call in LIBPQ.\n> > My example code in fe-connect.c is such as follows.\n> This looks like a great way to do it :-)\n>\n>\n> > \thp = gethostbyname(conn->pghost);\n> > #ifdef\tWIN32\n> > \tif ((hp == NULL) && (GetLastError() == WSANOTINITIALISED))\n> > \t{\n> > \t\tWSADATA\twsaData;\n> > \t\tif (WSAStartup(MAKEWORD(1,1),&wsaData))\n> > \t\t{\n> > \t\t\tfprintf(stderr, \"Failed to start\n> > winsock: %i\\n\", WSAGetLastError());\n> > \t\t\texit(1);\n>\n> This is not the way to do it, though - what if it's a gui program.\n> Should instead be, like the other error handling:\n>\n> \t\t\tsprintf(conn->errorMessage,\n> \t\t\t\t\t\"connectDB() -- Failed to start\n> winsock: %i\\n\",\n> \t\t\t\t\tWSAGetLstError());\n> \t\t\tgoto connect_errReturn;\n>\n> > \t\t}\n> > \t\t....\n> > \t\t???? for WSACleanup() ????\n> > \t\t....\n> Yes, this is the problem. Perhaps set a global flag in the DLL that is set\n> if we have ever called WSAStartup() from the DLL. Then we can check in\n> DllMain() at process detachment if we have to close the winsock?\n>\n> Will you write up a patch or should I?\n>\n\nJudging from your reply to Vince,it is preferable to call WSAStartup() and\ncorrspoiding WSACleanup() in DllMain.\nIf so,my original code is OK except error handling and the value of version\nnumber ?\n\nHiroshi Inoue\[email protected]\n\n",
"msg_date": "Thu, 1 Oct 1998 19:44:25 +0900",
"msg_from": "\"Hiroshi Inoue\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
}
] |
[
{
"msg_contents": "> > > On 28-Sep-98 Hiroshi Inoue wrote:\n> > > > I think that current DLL supports version \n> 1.0,1.1,2.0,2.1 and 2.2 and we\n> > > > can\n> > > > specify any version from 1.0 to 2.2 .\n> > > > For example,if we call \n> WSAStartup(MAKEWORD(1.0),&wsaData) from libpq\n> > > > and call WSAStartup(MAKEWORD(2,2),&wsaData) from \n> psql,both return OK\n> > > > and wsaData.wVersion is 1.0 for libpq and 2.2 for psql.\n> > >\n> > > Whichever is done, try not to ever request a version \n> higher than what is\n> > > required. It forces upgrades that may not even be \n> necessary. It's been\n> > > awhile since I've done windows code, but I've been bitten \n> by requiring a\n> > > higher version of a library than necessary - more than \n> once without even\n> > > realising it. It's not very fun when it comes time to fix it.\n> > >\n> > \n> > How about my sample code at the end of my previous post.\n> > In that code,libpq does nothing if main application calls \n> WSAStartup().\n> \n> According to the winsock 1.1 spec an intermediate dll should \n> handle it's\n> own startup and cleanup functions. Thinking about it, you \n> really have \n> no way of knowing if the app is going to stop using and unload winsock\n> if you're not done with it. It may have loaded it for any number of \n> reasons not necessarily related to the database access.\n\nYes, that is true. The problem lies in how Winsock really will handle\nrequests for different versions. If the DLL requires 1.1 and the main\nprogram requires 2.0. Will it then open up 2.0 interfaces for both, just the\nmain app, or none of them? It can't very well do it for just the main app,\nsince they are both in the same process & thread.\nOn the other hand, thinking a bit more of it. If we load version 1.1 in\nDllMain(), we are safe in most cases. Since if the app loads winsock later\nwith 2.0, it will probably switch to 2.0, and 2.0 is backwards compatible\nwith 1.1. \nThe problem with that scenario is if the DLL is loaded later on in the\nprogram using LoadLibrary(). In that case, the application will have loaded\nv2.0, and the DLL will then open up and load v1.1. Will the application then\nstill be able to use v2.0? With the current Microsoft implementation of\nWinsock, it will. But reading the specs, I can't find out if we are allowed\nto do this or not.\nOn the other hand, the specs say:\n\"An application or DLL can call WSAStartup more than once if it needs to\nobtain the WSAData structure information more than once. On each such call\nthe application can specify any version number supported by the DLL.\"\nAnd what would the point of this be, if a later call might break an earlier\ncode. So it will probably \"unlock\" parts of the Winsock to support all the\nrequested versions.\n\nIf that's the way it works, then we should just make sure we call\nWSAStartup() once _and_ WSACleanup() once. And again, if this is how it\nworks, then the original change with putting it all in DllMain() seems like\nthe smartest move again. We should probably go with that one, unless\nsomebody knows another good reason not to :-)\n\n\n//Magnus\n",
"msg_date": "Tue, 29 Sep 1998 14:23:17 +0200",
"msg_from": "Magnus Hagander <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
}
] |
[
{
"msg_contents": "> Right now preproc.y and gram.y are in\n> sync in the cvs tree. If you change anything please send me the diff \n> or tell me it has changed so I don't miss it. I have an old gram.y \n> version around so I can diff it locally.\n\nI haven't yet committed the changes to the source tree, but may have a\nchance to regression test and commit today. I'll send patches to you at\nthe same time. btw, there was a missing comma between ADD and AS in the\ntoken list (I see the problem also in preproc.y) and I found one or two\nomissions in parser/keywords.c. Most of my changes were to fix omissions\nand to allow tokens to be column identifiers or column labels as much as\npossible.\n\nThe other thing I'm doing is documenting the reserved and unreserved\nkeywords (I'm most of the way there, and will commit a new file\nsyntax.sgml to the docs tree soon). The only thing I've been looking at\nso far is gram.y. Are there any other keywords added to preproc.y? I see\nthat you did things so that EXEC and SQL are not required to be reserved\nwords (I hadn't thought of how to do that; nice trick!), but perhaps\nthere are others which are. Shall we put all of the keywords used by\neither gram.y or preproc.y into gram.y and keywords.c to make it more\nconsistant and to enforce consistancy in database design? SQL92 doesn't\nseem to make a big distinction between embedded and interactive SQL wrt\nkeyword restrictions...\n\n> Please send mail to me at [email protected]. That\n> way it's guaranteed to reach m.\n\nOK, sorry. I'd put your address into my mailer, and hadn't notice the\nchange.\n\n - Tom\n",
"msg_date": "Tue, 29 Sep 1998 12:55:57 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Antwort: [HACKERS] ecpg parser"
},
{
"msg_contents": "On Tue, Sep 29, 1998 at 12:55:57PM +0000, Thomas G. Lockhart wrote:\n> I haven't yet committed the changes to the source tree, but may have a\n> chance to regression test and commit today. I'll send patches to you at\n> the same time. btw, there was a missing comma between ADD and AS in the\n> token list (I see the problem also in preproc.y) and I found one or two\n> omissions in parser/keywords.c. Most of my changes were to fix omissions\n> and to allow tokens to be column identifiers or column labels as much as\n> possible.\n\nI see. I will add your changes as soon as I find time.\n\n> The other thing I'm doing is documenting the reserved and unreserved\n> keywords (I'm most of the way there, and will commit a new file\n> syntax.sgml to the docs tree soon). The only thing I've been looking at\n> so far is gram.y. Are there any other keywords added to preproc.y? I see\n\nYes, there are some special embedded sql keywords like break, call,\nwhenever, etc. See ecpg/preproc/ecpg_keywords.c for details.\n\n> that you did things so that EXEC and SQL are not required to be reserved\n> words (I hadn't thought of how to do that; nice trick!), but perhaps\n\nThanks. :-)\n\n> there are others which are. Shall we put all of the keywords used by\n> either gram.y or preproc.y into gram.y and keywords.c to make it more\n> consistant and to enforce consistancy in database design? SQL92 doesn't\n\nAt first I thought why making it a keyword when it doesn't have to. But the\nmore I think about it the more I like it. It just doesn't make sense to use\nthese keywords because that would force you to not use ecpg.\n\n> OK, sorry. I'd put your address into my mailer, and hadn't notice the\n> change.\n\nNo problem. But that way it will reach me no matter if I'm at home or in the\noffice.\n\nMichael\n-- \nDr. Michael Meskes | Th.-Heuss-Str. 61, D-41812 Erkelenz | Go SF49ers!\nSenior-Consultant | business: [email protected] | Go Rhein Fire!\nMummert+Partner | private: [email protected] | Use Debian\nUnternehmensberatung AG | [email protected] | GNU/Linux!\n",
"msg_date": "Wed, 30 Sep 1998 07:47:44 +0200",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Antwort: [HACKERS] ecpg parser"
}
] |
[
{
"msg_contents": "Hi,\n\n I'm currently thinking about multiple action and non-INSTEAD\n rules ON SELECT. I'm not sure what users might expect when\n they get fired.\n\n Well if a user types SELECT ... FROM tab and there are rules\n ON SELECT TO tab, then of course. But what about if the user\n issues an INSERT INTO x SELECT ... FROM tab or an UPDATE x\n SET col = tab.attr? In fact tab is scanned and returns data.\n Should the rule ON SELECT then be fired too?\n\n And what the hell is all that good for? Do we need other\n rules ON SELECT than those that build views (which we have\n now)? Tracing which data one user uses? Cannot be what rules\n are made for.\n\n If nobody votes against, I would only add some code checking\n that there is at max one INSTEAD SELECT rule that returns\n exactly the relations tuple structure ON SELECT (currently\n with CREATE TABLE, CREATE RULE someone can setup a situation\n that crashes the backend on SELECT). So SELECT rules are\n totally restricted to view building and nothing else.\n\n After that I'll tidy up the rewrite code (the work I've done\n screwed it up a little with nearly duplicate functions).\n Anything except for bug fixing is then delayed for 6.5.\n\n I still have in mind that we wanted to have views of UNIONS,\n DISTINCT views and some more. But since they require totally\n different semantics (the resulting plan must have something\n like a subselect of union in the case of an UPDATE...) this\n is far too much and has bad bad traps deep inside. We all\n don't want to fall into one during BETA.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Tue, 29 Sep 1998 14:58:40 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "rules ON SELECT"
},
{
"msg_contents": "> Hi,\n> \n> I'm currently thinking about multiple action and non-INSTEAD\n> rules ON SELECT. I'm not sure what users might expect when\n> they get fired.\n> \n> Well if a user types SELECT ... FROM tab and there are rules\n> ON SELECT TO tab, then of course. But what about if the user\n> issues an INSERT INTO x SELECT ... FROM tab or an UPDATE x\n> SET col = tab.attr? In fact tab is scanned and returns data.\n> Should the rule ON SELECT then be fired too?\n> \n> And what the hell is all that good for? Do we need other\n> rules ON SELECT than those that build views (which we have\n> now)? Tracing which data one user uses? Cannot be what rules\n> are made for.\n\nI would guess that SELECT ... FROM tab and INSERT INTO x SELECT ... FROM\ntab would use the rules, but I can see where and UPDATE is using the\ndata from the table just like a SELECT, so I can see it would be\nconfusing for some if it DID do the rule in that case, and for others if\nit DID NOT use the rule. That is a no-win situation, and usually means\nthat there is a bad interface design. However, in our case, it is just\na \"much too powerful\" interface design.\n\nPlease use your judgement. I can see that you are saying that there\nreally is no reason for SELECT rules except to create views. However,\nif someone wants a computed column as part of a table, AND they want the\ntable name to be the same, isn't a SELECT rule the only way to do that. \nThey could of course rename the table, and create a view with the old\nname. This may be the clearest way to do it.\n\nI think if you make it consistent and/or print something to the user\nwhen they try and do something strange, that usually keeps people happy.\nIf they understand what it is doing, the will usually accept it,\nespecially if it is a SUPERSET of SQL.\n\n> \n> If nobody votes against, I would only add some code checking\n> that there is at max one INSTEAD SELECT rule that returns\n> exactly the relations tuple structure ON SELECT (currently\n> with CREATE TABLE, CREATE RULE someone can setup a situation\n> that crashes the backend on SELECT). So SELECT rules are\n> totally restricted to view building and nothing else.\n> \n> After that I'll tidy up the rewrite code (the work I've done\n> screwed it up a little with nearly duplicate functions).\n> Anything except for bug fixing is then delayed for 6.5.\n> \n> I still have in mind that we wanted to have views of UNIONS,\n> DISTINCT views and some more. But since they require totally\n> different semantics (the resulting plan must have something\n> like a subselect of union in the case of an UPDATE...) this\n> is far too much and has bad bad traps deep inside. We all\n> don't want to fall into one during BETA.\n\nSounds good to me.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 21:41:55 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] rules ON SELECT"
}
] |
[
{
"msg_contents": "Does the SQL92 standard have to be purchased from the ANSI Board?\nWell, my actual question has to do with LIKE. Could anyone who has a\ncopy of a standard send me the PATTERN syntax/meanings for LIKE in\nSQL92?\n\tThanks,\n\t-DEJ\n",
"msg_date": "Tue, 29 Sep 1998 09:53:45 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "SQL92"
},
{
"msg_contents": "> Does the SQL92 standard have to be purchased from the ANSI Board?\n\nafaik, yes.\n\n> Well, my actual question has to do with LIKE. Could anyone who has a\n> copy of a standard send me the PATTERN syntax/meanings for LIKE in\n> SQL92?\n\nWell, you may be hoping for more than is in any of your reference books,\nbut there isn't any more to tell :)\n\n % ==> match any (sub)string\n _ ==> match any single character\n everything else matches itself\n\nIn SQL92 (but not yet in Postgres; can't remember if there is a\nworkaround):\n\n LIKE 'pattern' ESCAPE 'char'\n\nallows a single character \"char\" if it precedes the two pattern matching\ncharacters to demote the \"%\" or \"_\" to act like a normal single\ncharacter. The Postgres regex stuff is much more powerful.\n\nSorry, I think that's it :(\n\n - Tom\n",
"msg_date": "Tue, 29 Sep 1998 16:46:06 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
}
] |
[
{
"msg_contents": "> > Does the SQL92 standard have to be purchased from the ANSI Board?\n> \n> afaik, yes.\n> \n> > Well, my actual question has to do with LIKE. Could anyone who has\n> a\n> > copy of a standard send me the PATTERN syntax/meanings for LIKE in\n> > SQL92?\n> \n> Well, you may be hoping for more than is in any of your reference\n> books,\n> but there isn't any more to tell :)\n> \n> % ==> match any (sub)string\n> _ ==> match any single character\n> everything else matches itself\n> \n> In SQL92 (but not yet in Postgres; can't remember if there is a\n> workaround):\n> \n> LIKE 'pattern' ESCAPE 'char'\n> \n> allows a single character \"char\" if it precedes the two pattern\n> matching\n> characters to demote the \"%\" or \"_\" to act like a normal single\n> character. The Postgres regex stuff is much more powerful.\n> \n> Sorry, I think that's it :(\n> \n> - Tom\nWell, In all of the major Databases that I have worked with there is\nalso the \"[character_set]\" matching operator. \nSo a pattern of '[A-D]%' would match all word beginning with \"A\", \"B\",\n\"C\", or \"D\", and \"[^character_set]\" matches everything but the set.\nLooking at the current PgSQL like code this is ignored as well as the\nESCAPE syntax. I'm trying to resolve the problem we have of LIKE not\nmatching the pattern \"%%\" to anything even \"%\".\nBut, looking at the code I see that \"%%\" or \"%%%%%%%%\" should be equal\nto \"%\" so it's not the LIKE matching code that's causing the problem\n(also, read I'm looking in the wrong spot).\n\nDoes anybody know if the \"[character_set]\" stuff is part of the\nstandard?\nCan anyone point me to another place where the pattern could/would get\nmunged?\nAlso, wouldn't doing a memmove from the DATA portion of a varlena struct\ncause it problems if it was accessed again?\n\t\tThanks again,\n\t\t-DEJ\n",
"msg_date": "Tue, 29 Sep 1998 12:03:06 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] SQL92"
},
{
"msg_contents": "> Well, In all of the major Databases that I have worked with there is\n> also the \"[character_set]\" matching operator.\n> Does anybody know if the \"[character_set]\" stuff is part of the\n> standard?\n\nHoping against hope, eh? afaik those are all extensions (though SQL3 has\nsome enhanced functionality with the SIMILAR operator). From my copy of\nthe second draft standard published in 1992:\n\n<like predicate> uses the triadic operator LIKE (or the inverse,\nNOT LIKE), operating on three character strings and returning\na Boolean. LIKE determines whether or not a character string\n\"matches\" a given \"pattern\" (also a character string). The char-\nacters '%' (percent) and '_' (underscore) have special meaning when\nthey occur in the pattern. The optional third argument is a charac-\nter string containing exactly one character, known as the \"escape\ncharacter\", for use when a percent or underscore is required in the\npattern without its special meaning.\n\nWhat version of Postgres are you running? I vaguely recall some\ncomplaints and fixes in the sort-of-recent past. Don't know if it\naffected simple patterns or only more complicated stuff.\n\nCan you send a simple\n\n SELECT text 'your string here' LIKE 'your pattern here';\n\nwhich doesn't behave as you would expect? That would make this a bit\nless theoretical...\n\n - Tom\n",
"msg_date": "Wed, 30 Sep 1998 01:47:15 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
},
{
"msg_contents": "Jackson, DeJuan wrote:\n> \n> > > Does the SQL92 standard have to be purchased from the ANSI Board?\n> >\n> > afaik, yes.\n> >\n> > > Well, my actual question has to do with LIKE. Could anyone who has\n> > a\n> > > copy of a standard send me the PATTERN syntax/meanings for LIKE in\n> > > SQL92?\n> >\n> > Well, you may be hoping for more than is in any of your reference\n> > books,\n> > but there isn't any more to tell :)\n> >\n> > % ==> match any (sub)string\n> > _ ==> match any single character\n> > everything else matches itself\n> >\n> > In SQL92 (but not yet in Postgres; can't remember if there is a\n> > workaround):\n> >\n> > LIKE 'pattern' ESCAPE 'char'\n> >\n> > allows a single character \"char\" if it precedes the two pattern\n> > matching\n> > characters to demote the \"%\" or \"_\" to act like a normal single\n> > character. The Postgres regex stuff is much more powerful.\n> >\n> > Sorry, I think that's it :(\n> >\n> > - Tom\n> Well, In all of the major Databases that I have worked with there is\n> also the \"[character_set]\" matching operator.\n> So a pattern of '[A-D]%' would match all word beginning with \"A\", \"B\",\n> \"C\", or \"D\", and \"[^character_set]\" matches everything but the set.\n> Looking at the current PgSQL like code this is ignored as well as the\n> ESCAPE syntax. I'm trying to resolve the problem we have of LIKE not\n> matching the pattern \"%%\" to anything even \"%\".\n> But, looking at the code I see that \"%%\" or \"%%%%%%%%\" should be equal\n> to \"%\" so it's not the LIKE matching code that's causing the problem\n> (also, read I'm looking in the wrong spot).\n> \n> Does anybody know if the \"[character_set]\" stuff is part of the\n> standard?\nThe SQL92 LIKE:\n\n character-string-expression [NOT] LIKE pattern [ESCAPE\n'escape-char']\n \nallows only the special characters % and _\n\nYou need The SQL3 SIMILAR\n\n character-string-expression [NOT] SIMILAR TO pattern [ESCAPE\n'escape-char']\n\nin this case pattern can involve additional special characters, not just\n% and _ as in LIKE, but every regular expression or \"[character_set]\"\nlike you said. (Refer to \"A Guide to SQL Standard 4th edition\nDate-Rarwen, page 505).\n\n Jose'\n",
"msg_date": "Wed, 30 Sep 1998 12:14:12 +0200",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
},
{
"msg_contents": "> > > Does the SQL92 standard have to be purchased from the ANSI Board?\n> > \n> > afaik, yes.\n> > \n> > > Well, my actual question has to do with LIKE. Could anyone who has\n> > a\n> > > copy of a standard send me the PATTERN syntax/meanings for LIKE in\n> > > SQL92?\n> > \n> > Well, you may be hoping for more than is in any of your reference\n> > books,\n> > but there isn't any more to tell :)\n> > \n> > % ==> match any (sub)string\n> > _ ==> match any single character\n> > everything else matches itself\n> > \n> > In SQL92 (but not yet in Postgres; can't remember if there is a\n> > workaround):\n> > \n> > LIKE 'pattern' ESCAPE 'char'\n> > \n> > allows a single character \"char\" if it precedes the two pattern\n> > matching\n> > characters to demote the \"%\" or \"_\" to act like a normal single\n> > character. The Postgres regex stuff is much more powerful.\n> > \n> > Sorry, I think that's it :(\n> > \n> > - Tom\n> Well, In all of the major Databases that I have worked with there is\n> also the \"[character_set]\" matching operator. \n> So a pattern of '[A-D]%' would match all word beginning with \"A\", \"B\",\n> \"C\", or \"D\", and \"[^character_set]\" matches everything but the set.\n> Looking at the current PgSQL like code this is ignored as well as the\n> ESCAPE syntax. I'm trying to resolve the problem we have of LIKE not\n> matching the pattern \"%%\" to anything even \"%\".\n> But, looking at the code I see that \"%%\" or \"%%%%%%%%\" should be equal\n> to \"%\" so it's not the LIKE matching code that's causing the problem\n> (also, read I'm looking in the wrong spot).\n> \n> Does anybody know if the \"[character_set]\" stuff is part of the\n> standard?\n> Can anyone point me to another place where the pattern could/would get\n> munged?\n> Also, wouldn't doing a memmove from the DATA portion of a varlena struct\n> cause it problems if it was accessed again?\n> \t\tThanks again,\n\nAdded to TODO list:\n\n\t* have LIKE support character classes, 'pg_[a-c]%'\n\nI did not realize we didn't have that.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 21:48:59 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
},
{
"msg_contents": "> The SQL92 LIKE:\n> \n> character-string-expression [NOT] LIKE pattern [ESCAPE\n> 'escape-char']\n> \n> allows only the special characters % and _\n> \n> You need The SQL3 SIMILAR\n> \n> character-string-expression [NOT] SIMILAR TO pattern [ESCAPE\n> 'escape-char']\n> \n> in this case pattern can involve additional special characters, not just\n> % and _ as in LIKE, but every regular expression or \"[character_set]\"\n> like you said. (Refer to \"A Guide to SQL Standard 4th edition\n> Date-Rarwen, page 505).\n> \n> Jose'\n> \n> \n\nTODO item changed to:\n\n\t* add SIMILAR to allow character classes, 'pg_[a-c]%'\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 21:58:09 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
},
{
"msg_contents": "> > You need The SQL3 SIMILAR\n> > in this case pattern can involve additional special characters, not \n> > just % and _ as in LIKE,\n> TODO item changed to:\n> * add SIMILAR to allow character classes, 'pg_[a-c]%'\n\nThe annoying thing is that our regex \"~\" operators do all of this and\nmore, but presumably don't have the right behavior for underscore and\nfor percent. Should we look at the regex code and try to get it to be\ncompatible (for v6.5...)?\n\n - Tom\n",
"msg_date": "Fri, 02 Oct 1998 05:03:32 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
},
{
"msg_contents": "> > > You need The SQL3 SIMILAR\n> > > in this case pattern can involve additional special characters, not \n> > > just % and _ as in LIKE,\n> > TODO item changed to:\n> > * add SIMILAR to allow character classes, 'pg_[a-c]%'\n> \n> The annoying thing is that our regex \"~\" operators do all of this and\n> more, but presumably don't have the right behavior for underscore and\n> for percent. Should we look at the regex code and try to get it to be\n> compatible (for v6.5...)?\n\nNot sure.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 01:10:31 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
}
] |
[
{
"msg_contents": "Jan,\n\nIt builds and runs fine on S/Linux (Linux on a Sun SPARC)\n\nI must admit I was mighty impressed with it's capabilities,\ncourtesy of the demo/test database.\n\nOne day I'll find a problem that I can solve using PL/pgSQL,\nuntil then It's comforting to know it's there.\n\nFor anyone who hasn't taken a look yet, it's well worth it.\n\nKeith.\n\n\[email protected] (Jan Wieck)\n> \n> Hi,\n> \n> I wonder if anyone ever played around with PL/pgSQL. So far\n> (except for the missing Makefile) I got no response if it\n> compiles/works on other platforms than my Linux.\n> \n> Since there are only a few details left in the rewrite code\n> for 6.4 I would like to get back on it. I think 6.4 BETA is\n> the right time to check if it works on all the platforms (6.4\n> is the first time we ship it).\n> \n> I would do it by adding a build/install with (errors ignored\n> first) to the Makefiles. Then another regression test could\n> check if it works.\n> \n> Comments?\n\n\n",
"msg_date": "Tue, 29 Sep 1998 22:22:20 +0100 (BST)",
"msg_from": "Keith Parks <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] PL/pgSQL in template1?"
}
] |
[
{
"msg_contents": "Hi Michael. Here are the patches I just applied to the main parser. I'm\nnot sure how ecpg builds, but I also updated gram.c and parse.h to get\nthings properly self-consistant.\n\nThanks for looking at this...\n\n - Tom",
"msg_date": "Wed, 30 Sep 1998 05:54:07 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Patches for gram.y and keywords.c"
}
] |
[
{
"msg_contents": "Hi,\n\nPostgres 6.4.B1 compiles, links and runs with minimal changes in source code\n(mainly adding some includes and 1 or 2 source code changes) under Windows\nNT.\n\nAt this moment I can run initdb . Allthought it crashes after saying \"Adding\ntemplate1 database to pg_database...\" it does create some files. Can anyone\ntell me what files there should be (My Linux PC is broken so I cannot test\npostgres at Linux)?\n\nIn addition to the previous question, could anyone be so kind to send me the\noutput of a correctly working \"initdb --debug --username=Administrator\" (or\nwhatever your account is)?\nJoost Kraaijeveld\n\nAskesis B.V.\nGroenewoudseweg 46\n6524VB Nijmegen\ntel: 024-3888063\nfax: 024-3608416\nweb: http:/www.askesis.nl\n\n",
"msg_date": "Wed, 30 Sep 1998 10:02:00 +0200",
"msg_from": "\"Joost Kraaijeveld\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Windows NT port: what file are created by initdb?"
}
] |
[
{
"msg_contents": "Hi,\n\nCan anyone tell where the file is actually read that is given as input in\nthe initdb script where \"Adding template1 database to pg_database...\" is\nechoed?\n\nCan anyone tell me how I can run that command line through gdb (I know that\nthis not a postgres question but if anyone could helpme I would ben\ngratefull)?\n\nthe commandline is\n\npostgres -boot -C -F -D/usr/local/pgsql/data -d template1 < file.txt\n\nI know how to set the args but not how to get the redirection.\n\n\n\nJoost Kraaijeveld\n\nAskesis B.V.\nGroenewoudseweg 46\n6524VB Nijmegen\ntel: 024-3888063\nfax: 024-3608416\nweb: http:/www.askesis.nl\n\n",
"msg_date": "Wed, 30 Sep 1998 15:47:13 +0200",
"msg_from": "\"Joost Kraaijeveld\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "initdb question"
},
{
"msg_contents": "[Charset iso-8859-1 unsupported, filtering to ASCII...]\n> Hi,\n> \n> Can anyone tell where the file is actually read that is given as input in\n> the initdb script where \"Adding template1 database to pg_database...\" is\n> echoed?\n\nSee pgsql/lib for the *.source files.\n\n> \n> Can anyone tell me how I can run that command line through gdb (I know that\n> this not a postgres question but if anyone could helpme I would ben\n> gratefull)?\n> \n> the commandline is\n> \n> postgres -boot -C -F -D/usr/local/pgsql/data -d template1 < file.txt\n> \n> I know how to set the args but not how to get the redirection.\n\nSure. I usually make a copy of initdb, and call it inittest. I then\nedit the file, and put an 'exit' right before the postgres -boot command\nis executed. I then start the debugger on the postgres binary, and type\nthe command as though it is being run from initdb, because the initdb\nstuff is at the right point to do that. I can then watch if fail, and\nlook at the backtrace and output, because it is not being sent to\n/dev/null.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Thu, 1 Oct 1998 22:01:41 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] initdb question"
}
] |
[
{
"msg_contents": "> Okay, here's the patch for ecpg...\n> It contains other changes I made that haven't made it into the archive \n> yet. Could you please put all this into cvs?\n\nHi Michael. I got the patches and they all applied cleanly. I notice\nthat there might be several keywords missing from your version of\nkeywords.c (I had done some updates recently that I forgot to let you\nknow about):\n\n> {\"absolute\", ABSOLUTE},\n> {\"encoding\", ENCODING},\n> {\"names\", NAMES},\n> {\"next\", NEXT},\n> {\"old\", CURRENT},\n> {\"only\", ONLY},\n> {\"prior\", PRIOR},\n> {\"read\", READ},\n> {\"relative\", RELATIVE},\n> {\"scroll\", SCROLL},\n> {\"serial\", SERIAL},\n\nIt seems simple to make the changes to make this compatible with the\nnewest version of parser/keywords.c; should I go ahead and do that?\n\nSince the only difference between the files is in which files are\nmentioned in the #includes at the top, I was going to suggest changing\nparser/keywords.c to have something like:\n\n #ifdef ECPG\n #include \"type.h\"\n #include \"y.tab.h\"\n #else\n #include \"parse.h\"\n #endif\n\nBut I think that is not the best idea since it might make preproc.y and\nyour keywords.c inconsistant at times. OK, never mind: it's a bad idea\n:)\n\nSo, should I update your keywords.c to add the missing entries? It looks\nlike at least one entry (NEXT) does not appear in the keywords files but\ndoes appear in preproc.y already. Probably doesn't get recognized\nthough...\n\n - Tom\n",
"msg_date": "Wed, 30 Sep 1998 15:28:50 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Antwort: Patches for gram.y and keywords.c"
},
{
"msg_contents": "> will you apply my patches to the cvs tree, too? Since we will release \n> 6.4 pretty soon I think the changes should make it into the tree asap.\n\nApplied. \n\nThese patches from Michael bring the ecpg parser up to date with the\nmain Postgres parser, as well as solidifying some strdup() usage.\n\n - Tom\n",
"msg_date": "Sat, 03 Oct 1998 05:23:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Antwort: Patches for gram.y and keywords.c"
}
] |
[
{
"msg_contents": "> > Well, In all of the major Databases that I have worked with there is\n> > also the \"[character_set]\" matching operator.\n> > Does anybody know if the \"[character_set]\" stuff is part of the\n> > standard?\n> \n> Hoping against hope, eh? afaik those are all extensions (though SQL3\n> has\n> some enhanced functionality with the SIMILAR operator). From my copy\n> of\n> the second draft standard published in 1992:\n> \n> <like predicate> uses the triadic operator LIKE (or the inverse,\n> NOT LIKE), operating on three character strings and returning\n> a Boolean. LIKE determines whether or not a character string\n> \"matches\" a given \"pattern\" (also a character string). The char-\n> acters '%' (percent) and '_' (underscore) have special meaning when\n> they occur in the pattern. The optional third argument is a charac-\n> ter string containing exactly one character, known as the \"escape\n> character\", for use when a percent or underscore is required in the\n> pattern without its special meaning.\n> \n> What version of Postgres are you running? I vaguely recall some\n> complaints and fixes in the sort-of-recent past. Don't know if it\n> affected simple patterns or only more complicated stuff.\n> \n> Can you send a simple\n> \n> SELECT text 'your string here' LIKE 'your pattern here';\n> \n> which doesn't behave as you would expect? That would make this a bit\n> less theoretical...\n> \n> - Tom\n> \nWell, this all started with this message:\n> \nSince I didn't get any feedback from the list about it I decided to jump\ninto the code myself.\nUpon finding the like.c I looked through the code and found that\n\"[character_set]\"'s were ignored and that the problem expressed in the\nabove message was not evident in the code. (btw this is the CVS version\nof 6.4; I cvs update daily.) \nSo, I asked weather the \"[character_set]\" stuff was in the standard,\nwhich it is not. \nAnd I asked if I should be looking somewhere else for munging of \"%%\".\n(I'm currently looking at src/backend/utils/adt/like.c)\n\nSo, that's where I stand.\n\t-DEJ\n\nCC'd to the PostgreSQL Hackers list.\n\nI performed the same test as Brian on the cvs version of 6.4 and it\nexhibits the same behavior. could we get a fix in for the release.\n\t\t-DEJ\n\n> Hehehe... \n> \n> Try using the \"*\", as I posted to your early send... see if it makes\n> any\n> difference.\nDoing a like '*' will search for the character '*';\n\n> > er sorry about that last \"early send\" message...\n> > \n> > well, at first I though you were correct, but it turns out that\n> > postgresql\n> > is also inconsistent. Consider a table with a field username. 3\n> records:\n> > \n> > \"Brian % Schaffner\"\n> > \"Brian T Schaffner\"\n> > \"%\"\n> > \n> > select * from table where username='%' gets all rows (expected)\n> > select * from table where username='%%' gets the row with \"%\"\n> (expected)\n> > select * from table where username='%%%' gets the row with \"%\"\n> > (expected)\n> > select * from table where username='%%%%' gets no rows (expected)\n> > select * from table where username='% %' gets the 2 name rows\n> (expected)\n> > select * from table where username='% %%' gets the 2 name rows (NOT\n> > expected)\n> > select * from table where username='%% %' gets no rows (NOT\n> expected)\n> > select * from table where username='% % %' gets the 2 name rows\n> > (expected)\n> > select * from table where username='% %% %' gets the 2 name rows\n> (NOT\n> > expected)\n> > select * from table where username='% % % %' gets no rows (expected)\n> > \n> > so, if %% is the LIKE representation for a literal %, then why does\n> '%\n> > %%' return\n> > the 2 name rows, and '%% %' return no rows, and '% %% %' not return\n> the\n> > row with the\n> > embedded literal %?\n> > \n> > I could not get postgres to single out the row with the embedded %\n> using\n> > LIKE.\n> > \n> > why is this not getting any easier to define?\n> > \n> > -----Original Message-----\n> > From: Daniel J. Lashua [mailto:[email protected]]\n> > Sent: Friday, September 25, 1998 10:58 AM\n> > To: Brian Schaffner\n> > Cc: 'Rasmus Lerdorf'; '[email protected]';\n> > [email protected]\n> > Subject: RE: [PHP3] ARGH!! strstr() changed?\n> > \n> > \n> > On Fri, 25 Sep 1998, Brian Schaffner wrote:\n> > \n> > > in PostgreSQL (6.3.2 on FreeBSD 2.2.6):\n> > > \n> > > select * from table where field like '%%';\n> > > \n> > > returns NO rows;\n> > > \n> > > select * from table where field like '%';\n> > > \n> > > returns ALL rows;\n> > > \n> > > -brian-\n> > > \n> > \n> > \n> > I am not in any way doubting what you say... but that doesn't seem\n> > right.\n> > They should both reuturn all rows. Maybe in Postgres %% is the way\n> to\n> > state you actually want to search for ONE \"%\"?\n> > \n> > Out of curiosity, if you have time, could you test that. Make a\n> table\n> > with\n> > a field and in one row of the table in the field insert \"%\". Then do\n> > your\n> > select * from table where somefield LIKE '%%' and see if it returns\n> the\n> > one row?\n> > \n> > Am I offbase, or does that sound like incorrect behavior to you too?\n> > \n> > Daniel\n> > \n> > \n> \n> \n> --\n> PHP 3 Mailing List http://www.php.net/\n> To unsubscribe send an empty message to [email protected]\n> To subscribe to the digest list: [email protected]\n> For help: [email protected] Archive:\n> http://www.php.net/mailsearch.php3\n>",
"msg_date": "Wed, 30 Sep 1998 10:30:08 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] SQL92"
}
] |
[
{
"msg_contents": "> Jackson, DeJuan wrote:\n> > > > Does the SQL92 standard have to be purchased from the ANSI\n> Board?\n> > > afaik, yes.\n> > >\n> > > > Well, my actual question has to do with LIKE. Could anyone who\n> has\n> > > a\n> > > > copy of a standard send me the PATTERN syntax/meanings for LIKE\n> in\n> > > > SQL92?\n> > >\n> > > Well, you may be hoping for more than is in any of your reference\n> > > books,\n> > > but there isn't any more to tell :)\n> > >\n> > > % ==> match any (sub)string\n> > > _ ==> match any single character\n> > > everything else matches itself\n> > >\n> > > In SQL92 (but not yet in Postgres; can't remember if there is a\n> > > workaround):\n> > >\n> > > LIKE 'pattern' ESCAPE 'char'\n> > >\n> > > allows a single character \"char\" if it precedes the two pattern\n> > > matching\n> > > characters to demote the \"%\" or \"_\" to act like a normal single\n> > > character. The Postgres regex stuff is much more powerful.\n> > >\n> > > Sorry, I think that's it :(\n> > >\n> > > - Tom\n> > Well, In all of the major Databases that I have worked with there is\n> > also the \"[character_set]\" matching operator.\n> > So a pattern of '[A-D]%' would match all word beginning with \"A\",\n> \"B\",\n> > \"C\", or \"D\", and \"[^character_set]\" matches everything but the set.\n> > Looking at the current PgSQL like code this is ignored as well as\n> the\n> > ESCAPE syntax. I'm trying to resolve the problem we have of LIKE\n> not\n> > matching the pattern \"%%\" to anything even \"%\".\n> > But, looking at the code I see that \"%%\" or \"%%%%%%%%\" should be\n> equal\n> > to \"%\" so it's not the LIKE matching code that's causing the problem\n> > (also, read I'm looking in the wrong spot).\n> > \n> > Does anybody know if the \"[character_set]\" stuff is part of the\n> > standard?\n> The SQL92 LIKE:\n> \n> character-string-expression [NOT] LIKE pattern [ESCAPE\n> 'escape-char']\n> \n> allows only the special characters % and _\n> \n> You need The SQL3 SIMILAR\n> \n> character-string-expression [NOT] SIMILAR TO pattern [ESCAPE\n> 'escape-char']\n> \n> in this case pattern can involve additional special characters, not\n> just\n> % and _ as in LIKE, but every regular expression or \"[character_set]\"\n> like you said. (Refer to \"A Guide to SQL Standard 4th edition\n> Date-Rarwen, page 505).\n> \n> Jose'\nWell, I'm not looking for a new operator just trying to clarify what the\nscope of the current one should be. But, thanks for the information,\nJose'.\nIf you have any clue where in the source-tree LIKE is implemented could\nyou let me know?\n\t-DEJ\n",
"msg_date": "Wed, 30 Sep 1998 10:47:00 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] SQL92"
}
] |
[
{
"msg_contents": "I'm getting:\nmake -C pl clean\nmake[1]: Entering directory `/home/djackson/pgsql/src/pl'\nmake[1]: *** No rule to make target `clean'. Stop.\nmake[1]: Leaving directory `/home/djackson/pgsql/src/pl'\nmake: *** [clean] Error 2\non a 'make distclean', same error on a 'make all'.\nDid a cvs update at 9/30/98 11:00am CDT\n\t-DEJ\n",
"msg_date": "Wed, 30 Sep 1998 11:09:53 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Make error in current cvs."
},
{
"msg_contents": "> \n> I'm getting:\n> make -C pl clean\n> make[1]: Entering directory `/home/djackson/pgsql/src/pl'\n> make[1]: *** No rule to make target `clean'. Stop.\n> make[1]: Leaving directory `/home/djackson/pgsql/src/pl'\n> make: *** [clean] Error 2\n> on a 'make distclean', same error on a 'make all'.\n> Did a cvs update at 9/30/98 11:00am CDT\n> \t-DEJ\n> \n> \n\nThis is due to some missing files that Marc forgot to install.\nBut the patch for GNUmakefile.in went in. Just put a NIL Makefile\nin place until it all is there. BTW - the regression tests\nfor plpgsql will fail too.\n\n\nJan\n\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n",
"msg_date": "Wed, 30 Sep 1998 18:48:26 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Make error in current cvs."
},
{
"msg_contents": "On Wed, 30 Sep 1998, Jan Wieck wrote:\n\n> > \n> > I'm getting:\n> > make -C pl clean\n> > make[1]: Entering directory `/home/djackson/pgsql/src/pl'\n> > make[1]: *** No rule to make target `clean'. Stop.\n> > make[1]: Leaving directory `/home/djackson/pgsql/src/pl'\n> > make: *** [clean] Error 2\n> > on a 'make distclean', same error on a 'make all'.\n> > Did a cvs update at 9/30/98 11:00am CDT\n> > \t-DEJ\n> > \n> > \n> \n> This is due to some missing files that Marc forgot to install.\n\n\tBut...I did install these, after you mentioned that I had missed\nit...are they *still* missing?\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Mon, 5 Oct 1998 01:02:00 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Make error in current cvs."
},
{
"msg_contents": "\"Marc G. Fournier\" <[email protected]> wrote:\n> On Wed, 30 Sep 1998, Jan Wieck wrote:\n> \n> > > \n> > > I'm getting:\n> > > make -C pl clean\n> > > make[1]: Entering directory `/home/djackson/pgsql/src/pl'\n> > > make[1]: *** No rule to make target `clean'. Stop.\n> > > make[1]: Leaving directory `/home/djackson/pgsql/src/pl'\n> > > make: *** [clean] Error 2\n> > > on a 'make distclean', same error on a 'make all'.\n> > > Did a cvs update at 9/30/98 11:00am CDT\n> > > \t-DEJ\n> > > \n> > > \n> > \n> > This is due to some missing files that Marc forgot to install.\n> \n> \tBut...I did install these, after you mentioned that I had missed\n> it...are they *still* missing?\n> \n> Marc G. Fournier [email protected]\n> Systems Administrator @ hub.org \n> scrappy@{postgresql|isc}.org ICQ#7615664\n\nIt worked when I did a 'gmake clean' from the pl directory. (I just retrieved \nthe CVS tree tonight).\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Mon, 05 Oct 1998 01:55:30 -0400",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Make error in current cvs. "
}
] |
[
{
"msg_contents": "unsubscribe [email protected]\n\n",
"msg_date": "Wed, 30 Sep 1998 17:18:40 +0100",
"msg_from": "Sergio Brandano <[email protected]>",
"msg_from_op": true,
"msg_subject": "None"
}
] |
[
{
"msg_contents": "> > I'm getting:\n> > make -C pl clean\n> > make[1]: Entering directory `/home/djackson/pgsql/src/pl'\n> > make[1]: *** No rule to make target `clean'. Stop.\n> > make[1]: Leaving directory `/home/djackson/pgsql/src/pl'\n> > make: *** [clean] Error 2\n> > on a 'make distclean', same error on a 'make all'.\n> > Did a cvs update at 9/30/98 11:00am CDT\n> > \t-DEJ\n> \n> This is due to some missing files that Marc forgot to install.\n> But the patch for GNUmakefile.in went in. Just put a NIL Makefile\n> in place until it all is there. BTW - the regression tests\n> for plpgsql will fail too.\n> \n> \n> Jan\n> \nCould you forward me the files (and where they go). Part of what I want\nto do is test pl/pgsql.\n\t-DEJ\n> -- \n> \n> #=====================================================================\n> =#\n> # It's easier to get forgiveness for being wrong than for being right.\n> #\n> # Let's break this rule - forgive me.\n> #\n> #======================================== [email protected] (Jan Wieck)\n> #\n",
"msg_date": "Wed, 30 Sep 1998 11:48:45 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] Make error in current cvs."
}
] |
[
{
"msg_contents": "unsubscribe\n\n\n",
"msg_date": "Fri, 8 Feb 2036 08:20:54 +0200 (EET)",
"msg_from": "Alexzander Blashko <[email protected]>",
"msg_from_op": true,
"msg_subject": "None"
}
] |
[
{
"msg_contents": "> > > > With the current implementation, I don't beleive it will make a\n> > > difference -\n> > > > I don't think that any part of the Winsock system is\n> > > actually hidden if\n> > > you\n> > > > ask for a lower version. But it is _permitted_ by the\n> > > specification that\n> > > the\n> > > > DLL can hide parts that belong to a higher version than the\n> > > one requested.\n> > > >\n> > > > So I'm not 100% sure... Does anybody have access to a\n> > > Winsock?@that\n> > > actually\n> > > > hides some details when you ask for the wrong version?\n> > > >\n> > >\n> > > By the specs there may be the DLL that doesn't support \n> lower versions.\n> > > In that case my code doesn't work well.\n> > > But we can delay to call WSAStartup() after the first socket\n> > > call in LIBPQ.\n> > > My example code in fe-connect.c is such as follows.\n> > This looks like a great way to do it :-)\n> >\n> >\n> > > \thp = gethostbyname(conn->pghost);\n> > > #ifdef\tWIN32\n> > > \tif ((hp == NULL) && (GetLastError() == WSANOTINITIALISED))\n> > > \t{\n> > > \t\tWSADATA\twsaData;\n> > > \t\tif (WSAStartup(MAKEWORD(1,1),&wsaData))\n> > > \t\t{\n> > > \t\t\tfprintf(stderr, \"Failed to start\n> > > winsock: %i\\n\", WSAGetLastError());\n> > > \t\t\texit(1);\n> >\n> > This is not the way to do it, though - what if it's a gui program.\n> > Should instead be, like the other error handling:\n> >\n> > \t\t\tsprintf(conn->errorMessage,\n> > \t\t\t\t\t\"connectDB() -- Failed to start\n> > winsock: %i\\n\",\n> > \t\t\t\t\tWSAGetLstError());\n> > \t\t\tgoto connect_errReturn;\n> >\n> > > \t\t}\n> > > \t\t....\n> > > \t\t???? for WSACleanup() ????\n> > > \t\t....\n> > Yes, this is the problem. Perhaps set a global flag in the \n> DLL that is set\n> > if we have ever called WSAStartup() from the DLL. Then we \n> can check in\n> > DllMain() at process detachment if we have to close the winsock?\n> >\n> > Will you write up a patch or should I?\n> >\n> \n> Judging from your reply to Vince,it is preferable to call \n> WSAStartup() and\n> corrspoiding WSACleanup() in DllMain.\n> If so,my original code is OK except error handling and the \n> value of version\n> number ?\n\nYes, I think that is where we ended up :-)\n\n\nBTW, I came to think of another thing - the Win32 library does _not_ support\ncrypt-password authentication as it is now. Because there is no crypt()\nfunction.\nAny chance to have a file with crypt() included in the source as we did with\ngetopt? If that can be done, and one of you FreeBSD (or someone else who has\nit under BSD license) could mail it to me/put it in the tree, I can take a\nlook at implementing it - shouldn't bve hard at all.\n\n//Magnus\n",
"msg_date": "Thu, 1 Oct 1998 12:53:15 +0200 ",
"msg_from": "Magnus Hagander <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] LIBPQ for WIN32"
}
] |
[
{
"msg_contents": "Hi,\n\n found a little bug in libpgtcl/pgtclCmds.c causing a\n segmentation fault. Please apply.\n\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n*** pgtclCmds.c.orig\tThu Oct 1 09:15:11 1998\n--- pgtclCmds.c\tThu Oct 1 12:59:59 1998\n***************\n*** 241,258 ****\n {\n \tPQconninfoOption *option;\n \tchar\t\tbuf[8192];\n \n \tTcl_ResetResult(interp);\n \tfor (option = PQconndefaults(); option->keyword != NULL; option++)\n \t{\n! \t\tif (option->val == NULL)\n! \t\t\toption->val = \"\";\n \t\tsprintf(buf, \"{%s} {%s} {%s} %d {%s}\",\n \t\t\t\toption->keyword,\n \t\t\t\toption->label,\n \t\t\t\toption->dispchar,\n \t\t\t\toption->dispsize,\n! \t\t\t\toption->val);\n \t\tTcl_AppendElement(interp, buf);\n \t}\n \n--- 241,258 ----\n {\n \tPQconninfoOption *option;\n \tchar\t\tbuf[8192];\n+ \tchar\t\t*val;\n \n \tTcl_ResetResult(interp);\n \tfor (option = PQconndefaults(); option->keyword != NULL; option++)\n \t{\n! \t\tval = (option->val == NULL) ? \"\" : option->val;\n \t\tsprintf(buf, \"{%s} {%s} {%s} %d {%s}\",\n \t\t\t\toption->keyword,\n \t\t\t\toption->label,\n \t\t\t\toption->dispchar,\n \t\t\t\toption->dispsize,\n! \t\t\t\tval);\n \t\tTcl_AppendElement(interp, buf);\n \t}\n \n",
"msg_date": "Thu, 1 Oct 1998 13:06:42 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Patch for libpgtcl"
},
{
"msg_contents": "[email protected] (Jan Wieck) writes:\n> found a little bug in libpgtcl/pgtclCmds.c causing a\n> segmentation fault. Please apply.\n\nPatch applied. That routine had other problems too: it would fail if\nany of the values contained characters that are special in a Tcl list.\nAll better now...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Thu, 01 Oct 1998 21:42:01 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Patch for libpgtcl "
}
] |
[
{
"msg_contents": "Hi postgres gurus,\n\nI have problems with running initdb. As I mentioned in a precious message I\ncannot run postgres on a Linux machine at this moment so I have to ask\ninstead of trying myself. And in making the situation worse, for some reason\nI cannot debug the program if input is redirected to a file (see question 3)\nbut I have to use printf statements instead.\n\nI have three main questions.\n\nRunning \"initdb --username=Administrator -d\" seems to do:\n1. checking the environment setting (goes OK)\n2. checking/setting file templates \t (goes OK)\n3. checking user settings (goes OK)\n4. creating directories if needed (goes OK)\n5. creating template1 database (question 1)\n6. creating global classes in two steps (question 2 and 3)\n7. do some database filling (goes ????)\n\nQuestion 1.\n\nPostgres is run after creating the template1 directory saying:\n\"Creating template database in /usr/local/pgsql/data/base/template1\"\n\"Running: postgres -boot -C -F -D/usr/local/pgsql/data -d template1\"\n\nWhat should be the effect of running postgres at this point? At my machine\nit does nothing, other than run (that is , no files are created/changed\netc).\n\nN.B.\n/usr/local/pgsql/data/base/template1/PG_VERSION is created, not by running\npostgres but by running pg_version after that\n\nQuestion 2.\n\nPostgres is run again after saying:\n\"Creating global classes in /usr/local/pgsql/data/base\"\n\"Running: postgres -boot -C -F -D/usr/local/pgsql/data -d template1\"\n\nWhat should be the effect of running postgres at this point? At my machine\nit does nothing, other than run (that is , no files are created/changed\netc).\n\nN.B.\n/usr/local/pgsql/data/PG_VERSION is created, not by running postgres but by\nrunning pg_version after that\n/usr/local/pgsql/data/pg_geqo.sample, created by a simple copy\n/usr/local/pgsql/data/pg_hba.conf, created by a simple copy\n\nQuestion 3.\n\nPostgres is run again after saying:\n\"Adding template1 database to pg_database...\"\n\"Running: postgres -boot -C -F -D/usr/local/pgsql/data -d template1 <\n/tmp/create.8796\" (number varies)\n\nThe program crashes while running. Two empty files are created: pg_class and\npg_type. The program crashes because for some unknown reason no valid\nHeapTuple is created for pg_database. Is there anyone with a briljant idea?\nWhat could be the reason? Is it related with the fact that prevoius runs of\npostgres did't do anything?\n\nTIA\n\nJoost Kraaijeveld\n\nAskesis B.V.\nGroenewoudseweg 46\n6524VB Nijmegen\ntel: 024-3888063\nfax: 024-3608416\nweb: http:/www.askesis.nl\n\n",
"msg_date": "Thu, 1 Oct 1998 14:36:16 +0200",
"msg_from": "\"Joost Kraaijeveld\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Initdb questions Windows NT port"
}
] |
[
{
"msg_contents": "\n\n Hi everyone.\n\n I am new to postgres. I need a relational db that can be used with\na web client\nfrom any plataform, and this seems to be it, BUT.....\n\n Our lab has ONLY sgi machines. Indigo 2s running Irix 6.2, an\nOrigin 200 running\nIRIX 6.4, and a lonely Indigo running IRIX 5.3 !!!\n\n I have read about all the problems with the ld on IRIX 6.2 and\n6.4, and try to follow the\ninstallation instructions, and did not have any luck on all 3\nsystesms. Here is a little summary\nof what I did !!! BTW, these systems are patched up with the latest\npatches, so, it has the\nlatest compilers, on the 6.2 and 6.4, and 5.3 has the latest of what it\ncan have ( I think it is 7.1 for\nthe c compiler, but I am not sure there ).\n\n ANyways, I installed gmake and GNU readline, but could not find\nGNU install !!! I looked on the\nGNU page, and it is not there !!! ANybody knows ?????\n\n Thenm I created the postgres account, logged in, and did a configure\nwith the --prefix and where\nthe location of the src directory is. Then, issue the gmake all and\nI get a bunch of warnings and\nit compiles for a while, and then, I get these errors !!!!!\n\nWarning: parse_target.o: fmgr_pl_finfo: multiply defined\n previous (used) definition from 'analyze.o';\n new (ignored) definition from 'parse_target.o'\nWarning: scan.o: fmgr_pl_finfo: multiply defined\n previous (used) definition from 'analyze.o';\n new (ignored) definition from 'scan.o'\ngmake[2]: Leaving directory `/purple/disk3/pgsql/src/backend/parser'\ngmake -C port all\ngmake[2]: Entering directory `/purple/disk3/pgsql/src/backend/port'\ncc -I../../include -I../../backend -I.. -c dynloader.c -o\ndynloader.o\ncc -I../../include -I../../backend -I.. -c isinf.c -o isinf.o\ncc -I../../include -I../../backend -I.. -c snprintf.c -o\nsnprintf.o\ncfe: Error: snprintf.c, line 115: Syntax Error\n static void fmtstr __P((char *value, int ljust, int len, int zpad, int\nmaxwidth));\n -------------------^\ncfe: Error: snprintf.c, line 115: Syntax Error\n static void fmtstr __P((char *value, int ljust, int len, int zpad, int\nmaxwidth));\n -------------------------------------^\ncfe: Error: snprintf.c, line 115: Syntax Error\n static void fmtstr __P((char *value, int ljust, int len, int zpad, int\nmaxwidth));\n ------------------------------------------------^\ncfe: Error: snprintf.c, line 115: Syntax Error\n static void fmtstr __P((char *value, int ljust, int len, int zpad, int\nmaxwidth));\n ---------------------------------------------------------^\ncfe: Error: snprintf.c, line 115: Syntax Error\n static void fmtstr __P((char *value, int ljust, int len, int zpad, int\nmaxwidth));\n -------------------------------------------------------------------^\ncfe: Warning 625: snprintf.c, line 115: Empty declaration\n static void fmtstr __P((char *value, int ljust, int len, int zpad, int\nmaxwidth));\n ---------------------------------------------------------------------------------^\n\ncfe: Error: snprintf.c, line 116: Syntax Error\n static void fmtnum __P((long value, int base, int dosign, int ljust,\nint len, int zpad));\n -------------------^\ncfe: Error: snprintf.c, line 116: Syntax Error\n static void fmtnum __P((long value, int base, int dosign, int ljust,\nint len, int zpad));\n ------------------------------------^\ncfe: Error: snprintf.c, line 116: Syntax Error\n static void fmtnum __P((long value, int base, int dosign, int ljust,\nint len, int zpad));\n ----------------------------------------------^\ncfe: Error: snprintf.c, line 116: Syntax Error\n static void fmtnum __P((long value, int base, int dosign, int ljust,\nint len, int zpad));\n ----------------------------------------------------------^\ncfe: Error: snprintf.c, line 116: Syntax Error\n static void fmtnum __P((long value, int base, int dosign, int ljust,\nint len, int zpad));\n ---------------------------------------------------------------------^\ncfe: Error: snprintf.c, line 116: Syntax Error\n static void fmtnum __P((long value, int base, int dosign, int ljust,\nint len, int zpad));\n ------------------------------------------------------------------------------^\n\ncfe: Warning 625: snprintf.c, line 116: Empty declaration\n static void fmtnum __P((long value, int base, int dosign, int ljust,\nint len, int zpad));\n ----------------------------------------------------------------------------------------^\n\ncfe: Error: snprintf.c, line 117: Syntax Error\n static void dostr __P(( char * , int ));\n ------------------^\ncfe: Warning 625: snprintf.c, line 117: Empty declaration\n static void dostr __P(( char * , int ));\n ---------------------------------------^\ncfe: Error: snprintf.c, line 119: Syntax Error\n static void dopr_outch __P(( int c ));\n -----------------------^\ncfe: Warning 625: snprintf.c, line 119: Empty declaration\n static void dopr_outch __P(( int c ));\n -------------------------------------^\ngmake[2]: *** [snprintf.o] Error 1\ngmake[2]: Leaving directory `/purple/disk3/pgsql/src/backend/port'\ngmake[1]: *** [port.dir] Error 2\ngmake[1]: Leaving directory `/purple/disk3/pgsql/src/backend'\ngmake: *** [all] Error 2\n\n\n SO !!!!! Any help ? Did anybody had some luck with SGIs ?\nHow do I get rid of this errors,\nand or warnings ? Since I do not have much experience with unix\ncompilation, I am in a loss\nhere !! My boss wants this up and running ASAP, or else, he will buy a\nwindoze app, argh !!!!!!\n\n If anybody has binaries, it would be nice to have them uploaded to\nthe postgres site !!!!!\n\n--\n==============================================================================\n Henrique Moreira da Silva Jr |\n X-ray Crystallography Lab | voice:(617) 632-3981\n |\n Dana-Farber Cancer Institute | fax: (617) 632-4393\n 44 Binney Street, SM 1036 |\n Boston, MA 02115 | [email protected]\n==============================================================================\n\n\n\n",
"msg_date": "Thu, 01 Oct 1998 11:17:09 -0400",
"msg_from": "Henrique Silva <[email protected]>",
"msg_from_op": true,
"msg_subject": "Having problems compiling postgres on IRIX"
}
] |
[
{
"msg_contents": "Well 1st problem --prefix is not where the source code is at but where\nyou want the binaries to go once the compile is finished.\nThe syntax errors that you are getting on snprintf.c I'm not\nexperiencing here. I'm using gcc v2.7.2.3 (you can do gcc --version to\ncheck yours).\nYou might want to try editing out the extra parens and see what you get.\n\n\nAnd I have no idea where you get the install.sh script, but one does\ncome with the postgres distribution.\n\t\t-DEJ\n\n> -----Original Message-----\n> From:\tHenrique Silva [SMTP:[email protected]]\n> Sent:\tThursday, October 01, 1998 10:17 AM\n> To:\[email protected]; Henrique Silva;\n> [email protected]; [email protected]\n> Subject:\t[ADMIN] Having problems compiling postgres on IRIX\n> \n> \n> \n> Hi everyone.\n> \n> I am new to postgres. I need a relational db that can be used\n> with\n> a web client\n> from any plataform, and this seems to be it, BUT.....\n> \n> Our lab has ONLY sgi machines. Indigo 2s running Irix 6.2, an\n> Origin 200 running\n> IRIX 6.4, and a lonely Indigo running IRIX 5.3 !!!\n> \n> I have read about all the problems with the ld on IRIX 6.2 and\n> 6.4, and try to follow the\n> installation instructions, and did not have any luck on all 3\n> systesms. Here is a little summary\n> of what I did !!! BTW, these systems are patched up with the latest\n> patches, so, it has the\n> latest compilers, on the 6.2 and 6.4, and 5.3 has the latest of what\n> it\n> can have ( I think it is 7.1 for\n> the c compiler, but I am not sure there ).\n> \n> ANyways, I installed gmake and GNU readline, but could not find\n> GNU install !!! I looked on the\n> GNU page, and it is not there !!! ANybody knows ?????\n> \n> Thenm I created the postgres account, logged in, and did a\n> configure\n> with the --prefix and where\n> the location of the src directory is. Then, issue the gmake all\n> and\n> I get a bunch of warnings and\n> it compiles for a while, and then, I get these errors !!!!!\n> \n> Warning: parse_target.o: fmgr_pl_finfo: multiply defined\n> previous (used) definition from 'analyze.o';\n> new (ignored) definition from 'parse_target.o'\n> Warning: scan.o: fmgr_pl_finfo: multiply defined\n> previous (used) definition from 'analyze.o';\n> new (ignored) definition from 'scan.o'\n> gmake[2]: Leaving directory `/purple/disk3/pgsql/src/backend/parser'\n> gmake -C port all\n> gmake[2]: Entering directory `/purple/disk3/pgsql/src/backend/port'\n> cc -I../../include -I../../backend -I.. -c dynloader.c -o\n> dynloader.o\n> cc -I../../include -I../../backend -I.. -c isinf.c -o isinf.o\n> cc -I../../include -I../../backend -I.. -c snprintf.c -o\n> snprintf.o\n> cfe: Error: snprintf.c, line 115: Syntax Error\n> static void fmtstr __P((char *value, int ljust, int len, int zpad,\n> int\n> maxwidth));\n> -------------------^\n> cfe: Error: snprintf.c, line 115: Syntax Error\n> static void fmtstr __P((char *value, int ljust, int len, int zpad,\n> int\n> maxwidth));\n> -------------------------------------^\n> cfe: Error: snprintf.c, line 115: Syntax Error\n> static void fmtstr __P((char *value, int ljust, int len, int zpad,\n> int\n> maxwidth));\n> ------------------------------------------------^\n> cfe: Error: snprintf.c, line 115: Syntax Error\n> static void fmtstr __P((char *value, int ljust, int len, int zpad,\n> int\n> maxwidth));\n> ---------------------------------------------------------^\n> cfe: Error: snprintf.c, line 115: Syntax Error\n> static void fmtstr __P((char *value, int ljust, int len, int zpad,\n> int\n> maxwidth));\n> -------------------------------------------------------------------^\n> cfe: Warning 625: snprintf.c, line 115: Empty declaration\n> static void fmtstr __P((char *value, int ljust, int len, int zpad,\n> int\n> maxwidth));\n> \n> ----------------------------------------------------------------------\n> -----------^\n> \n> cfe: Error: snprintf.c, line 116: Syntax Error\n> static void fmtnum __P((long value, int base, int dosign, int ljust,\n> int len, int zpad));\n> -------------------^\n> cfe: Error: snprintf.c, line 116: Syntax Error\n> static void fmtnum __P((long value, int base, int dosign, int ljust,\n> int len, int zpad));\n> ------------------------------------^\n> cfe: Error: snprintf.c, line 116: Syntax Error\n> static void fmtnum __P((long value, int base, int dosign, int ljust,\n> int len, int zpad));\n> ----------------------------------------------^\n> cfe: Error: snprintf.c, line 116: Syntax Error\n> static void fmtnum __P((long value, int base, int dosign, int ljust,\n> int len, int zpad));\n> ----------------------------------------------------------^\n> cfe: Error: snprintf.c, line 116: Syntax Error\n> static void fmtnum __P((long value, int base, int dosign, int ljust,\n> int len, int zpad));\n> \n> ---------------------------------------------------------------------^\n> cfe: Error: snprintf.c, line 116: Syntax Error\n> static void fmtnum __P((long value, int base, int dosign, int ljust,\n> int len, int zpad));\n> \n> ----------------------------------------------------------------------\n> --------^\n> \n> cfe: Warning 625: snprintf.c, line 116: Empty declaration\n> static void fmtnum __P((long value, int base, int dosign, int ljust,\n> int len, int zpad));\n> \n> ----------------------------------------------------------------------\n> ------------------^\n> \n> cfe: Error: snprintf.c, line 117: Syntax Error\n> static void dostr __P(( char * , int ));\n> ------------------^\n> cfe: Warning 625: snprintf.c, line 117: Empty declaration\n> static void dostr __P(( char * , int ));\n> ---------------------------------------^\n> cfe: Error: snprintf.c, line 119: Syntax Error\n> static void dopr_outch __P(( int c ));\n> -----------------------^\n> cfe: Warning 625: snprintf.c, line 119: Empty declaration\n> static void dopr_outch __P(( int c ));\n> -------------------------------------^\n> gmake[2]: *** [snprintf.o] Error 1\n> gmake[2]: Leaving directory `/purple/disk3/pgsql/src/backend/port'\n> gmake[1]: *** [port.dir] Error 2\n> gmake[1]: Leaving directory `/purple/disk3/pgsql/src/backend'\n> gmake: *** [all] Error 2\n> \n> \n> SO !!!!! Any help ? Did anybody had some luck with SGIs ?\n> How do I get rid of this errors,\n> and or warnings ? Since I do not have much experience with unix\n> compilation, I am in a loss\n> here !! My boss wants this up and running ASAP, or else, he will buy\n> a\n> windoze app, argh !!!!!!\n> \n> If anybody has binaries, it would be nice to have them uploaded to\n> the postgres site !!!!!\n> \n> --\n> ======================================================================\n> ========\n> Henrique Moreira da Silva Jr |\n> X-ray Crystallography Lab | voice:(617) 632-3981\n> |\n> Dana-Farber Cancer Institute | fax: (617) 632-4393\n> 44 Binney Street, SM 1036 |\n> Boston, MA 02115 |\n> [email protected]\n> ======================================================================\n> ========\n> \n> \n> \n",
"msg_date": "Thu, 1 Oct 1998 11:56:42 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [ADMIN] Having problems compiling postgres on IRIX"
}
] |
[
{
"msg_contents": "Well,\n\nAfter looking at the code for pg_dump, it appears that all identifiers\ngo through fmtId which does as you thougt, it looks to see if a character\nis non-lower case, a digit, or an '_' then it double quotes it.\nI am going to change this behavior to always double quote it and see \nif it breaks anything.\n\nI also noted a possible memory leak that I introduced when I added\nthe code to dump the ACLs for tables and such. I will take care of\nthis as well.\n\nMy concern is that I do not have a database with enough stuff to\nreally test this. I tried to dump the regression and template1 \ndatabases, but template1 had nothing in it and regression failed\nnot being able to find a trigger.\n\nDoes anyone have a good test database that they could send me the\npg_dump of? If it is really big, I can either ftp it from you, or\nyou could ftp it to me.\n\nThanks,\n\nMatt\n> We need to think about whether to surround all identifiers with double\n> quotes all the time in pg_dump output. \n> \n> The reason is that Postgres allows reserved keywords to be specified as\n> table and column names if they are surrounded by the double quotes, but\n> pg_dump doesn't know whether an identifier also happens to be a reserved\n> keyword. afaik it's now only using DQs if there is upper case or funny\n> characters in the identifier. Instead it should probably surround the\n> fields with DQs all the time. It could perhaps have a command-line\n> switch to turn off that feature if necessary.\n> \n> The alternative to always using the DQs is to have pg_dump use the\n> keywords.c routine available in the backend. But I don't think that is\n> as reliable since it isn't guaranteed to be in sync with the backend\n> version and since there are non-reserved keywords in that file which\n> test the same as the reserved ones.\n> \n> This should probably be a \"must do\" for v6.4...\n> \n> Comments?\n> \n> - Tom\n> \n\n----------\nMatthew C. Aycock\nOperating Systems Analyst/Admin, Senior\nDept Math/CS\nEmory University, Atlanta, GA \nInternet: [email protected] \t\t\n\n\n",
"msg_date": "Thu, 1 Oct 1998 15:01:24 -0400 (EDT)",
"msg_from": "\"Matthew C. Aycock\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] pg_dump"
},
{
"msg_contents": "> After looking at the code for pg_dump, it appears that all identifiers\n> go through fmtId which does as you thougt, it looks to see if a character\n> is non-lower case, a digit, or an '_' then it double quotes it.\n> I am going to change this behavior to always double quote it and see\n> if it breaks anything.\n\nI've already got patches to do the double-quote thing, with a command\nline option (-n) to recover the old behavior. Will commit that, so you\nwill only need to do the memory leak stuff. Or would you prefer that I\nsend you my patches and you can integrate them?\n\n - Tom\n",
"msg_date": "Fri, 02 Oct 1998 05:00:04 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] pg_dump"
}
] |
[
{
"msg_contents": "Here is a report regarding the backend-crash from a user in Japan.\nIncluded shell script should reproduce the phenomenon.\nNote that\n\tselect * from getting; vacuum;\ndoes cause a crash, while\n\tselect * from getting;\n\tvacuum;\nnot.\n--\nTatsuo Ishii\[email protected]\n\n========================================================================\n#!/bin/sh\n\nDBNAME=ptest\n\ndestroydb $DBNAME\ncreatedb $DBNAME\npsql -e $DBNAME <<EOF\ncreate table header\n(\n\thost\ttext\tnot null,\n\tport\tint\tnot null,\n\tpath\ttext\tnot null,\n\tfile\ttext\tnot null,\n\textra\ttext\tnot null,\n\tname\ttext\tnot null,\n\tvalue\ttext\tnot null\n);\ncreate index header_url_idx on header (host, port, path, file, extra);\ncreate unique index header_uniq_idx on header (host, port, path, file, extra, name);\n\ncreate table reference\n(\n\tf_url\ttext\tnot null,\n\tt_url\ttext\tnot null\n);\ncreate index reference_from_idx on reference (f_url);\ncreate index reference_to_idx on reference (t_url);\ncreate unique index reference_uniq_idx on reference (f_url, t_url);\n\ncreate table extension\n(\n\text\ttext\tnot null,\n\tnote\ttext\n);\ncreate unique index extension_ext_idx on extension (ext);\n\ncreate table getting\n(\n\thost\ttext\tnot null,\n\tport\tint\tnot null,\n\tip\ttext\tnot null,\n\twhen\tdatetime\tnot null\n);\ncreate unique index getting_ip_idx on getting (ip);\nEOF\n#psql -c \"delete from getting; vacuum;\" $DBNAME\npsql -c \"select * from getting; vacuum;\" $DBNAME\n#psql -c \"delete from getting;\" $DBNAME\n#psql -c \"select * from getting;\" $DBNAME\n#psql -c \"vacuum;\" $DBNAME\n#psql -c \"vacuum; vacuum;\" $DBNAME\n",
"msg_date": "Fri, 02 Oct 1998 12:44:54 +0900",
"msg_from": "Tatsuo Ishii <[email protected]>",
"msg_from_op": true,
"msg_subject": "delete from ..;vacuum crashes"
},
{
"msg_contents": "Tatsuo Ishii <[email protected]> writes:\n> Here is a report regarding the backend-crash from a user in Japan.\n> Included shell script should reproduce the phenomenon.\n\nOn which postgres version(s)?\n\n> Note that\n> \tselect * from getting; vacuum;\n> does cause a crash, while\n> \tselect * from getting;\n> \tvacuum;\n> not.\n\nSpecifically I see you are using\n\n> psql -c \"select * from getting; vacuum;\" $DBNAME\n\nrather than entering the commands at the psql prompt. The -c option\nworks differently from entering multiple commands at psql's prompt.\nIn ordinary interactive use, psql will break what you type at\nsemicolon boundaries and send each SQL command to the backend\nseparately, even if you typed several commands on one line.\n*But* the -c option doesn't work that way --- it just sends the\nwhole given string to the backend as one query.\n\nThe implication of this is that psql -c \"select * from getting; vacuum;\"\nexecutes the select and the vacuum as part of a single transaction,\nwhereas any other way of doing it with psql will make the commands be\ntwo separate transactions. I speculate that this has something to do\nwith the different behavior you see.\n\nExactly what the bug is is beyond my abilities, but perhaps that tidbit\nwill help someone more competent to find it.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Fri, 02 Oct 1998 02:14:22 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] delete from ..;vacuum crashes "
},
{
"msg_contents": "At 2:14 AM 98.10.2 -0400, Tom Lane wrote:\n>Tatsuo Ishii <[email protected]> writes:\n>> Here is a report regarding the backend-crash from a user in Japan.\n>> Included shell script should reproduce the phenomenon.\n>\n>On which postgres version(s)?\n\nThe latest cvs source tree (I have not tried this for 6.3.2).\n\n>The implication of this is that psql -c \"select * from getting; vacuum;\"\n>executes the select and the vacuum as part of a single transaction,\n>whereas any other way of doing it with psql will make the commands be\n>two separate transactions. I speculate that this has something to do\n>with the different behavior you see.\n\nSo basically the backend treats multiple SQL statements conjucted with \";\"\nas a single transaction?\n\nIf this is the cause of the problem, following SQLs should produce\nthe backend death too. I will try this.\n\nbegin;\nselect * from getting;\nvacuum;\nend;\n--\nTatsuo Ishii\[email protected]\n\n",
"msg_date": "Fri, 2 Oct 1998 23:16:10 +0900",
"msg_from": "[email protected] (Tatsuo Ishii)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] delete from ..;vacuum crashes"
},
{
"msg_contents": "[email protected] (Tatsuo Ishii) writes:\n> So basically the backend treats multiple SQL statements conjucted with \";\"\n> as a single transaction?\n\nIf they arrive in a single query string, as you can cause with psql -c\nor with a direct PQexec() call.\n\nActually it's even more subtle than that: multiple statements in a\nsingle query string act like a *single statement* as far as the\ntransaction mechanism is concerned. There's one StartTransactionCommand\nin postgres.c before the string starts to execute, and one\nCommitTransactionCommand after it's done.\n\nIt is possible that that is a bug, and that we need to take the start/\ncommit calls out of the postgres.c main loop and put them somewhere\ndown inside the parsing/execution code, at a point where the software\nhas parsed off a single SQL statement. As it stands, if there are any\nstatements in the system that assume there is a StartTransactionCommand\njust before they begin and a CommitTransactionCommand just after they\nfinish, those statements will break when executed as part of a\nmulti-statement query. Question for the gurus: would you consider this\na bug in the particular statement (it shouldn't assume that), or a bug\nin the outer layers (they should make that be true)?\n\nIn particular, since vacuum.c does some peculiar things with transaction\nboundaries, it seems to me that it might be an example of such a\nstatement and that what I just described is the root cause of your bug.\nBut someone who knows the system better than me will have to figure out\njust what's going on.\n\n> If this is the cause of the problem, following SQLs should produce\n> the backend death too. I will try this.\n\n> begin;\n> select * from getting;\n> vacuum;\n> end;\n\nYou should try it and let us know. But that is a different test case,\nbecause there will be CommitTransactionCommand & StartTransactionCommand\nbetween the select and the vacuum.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Fri, 02 Oct 1998 10:40:36 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] delete from ..;vacuum crashes "
},
{
"msg_contents": "(I have changed the subject \"delete from\" to \"select * from\" )\n\nAs I reported,\n\n\tselect * from getting; vacuum;\n\ndoes crash the backend with included test data.\n\nThis time I have tried:\n\nbegin;\nselect * from getting;\nvacuum;\nend;\n\nand have a crash too.\n\n(using current source tree + FreeBSD)\n\nI think this should be added to the Open 6.4 items list.\n\n>Here is a report regarding the backend-crash from a user in Japan.\n>Included shell script should reproduce the phenomenon.\n>Note that\n>\tselect * from getting; vacuum;\n>does cause a crash, while\n>\tselect * from getting;\n>\tvacuum;\n>not.\n>--\n>Tatsuo Ishii\n>[email protected]\n>\n>========================================================================\n>#!/bin/sh\n>\n>DBNAME=ptest\n>\n>destroydb $DBNAME\n>createdb $DBNAME\n>psql -e $DBNAME <<EOF\n>create table header\n>(\n>\thost\ttext\tnot null,\n>\tport\tint\tnot null,\n>\tpath\ttext\tnot null,\n>\tfile\ttext\tnot null,\n>\textra\ttext\tnot null,\n>\tname\ttext\tnot null,\n>\tvalue\ttext\tnot null\n>);\n>create index header_url_idx on header (host, port, path, file, extra);\n>create unique index header_uniq_idx on header (host, port, path, file, extra, name);\n>\n>create table reference\n>(\n>\tf_url\ttext\tnot null,\n>\tt_url\ttext\tnot null\n>);\n>create index reference_from_idx on reference (f_url);\n>create index reference_to_idx on reference (t_url);\n>create unique index reference_uniq_idx on reference (f_url, t_url);\n>\n>create table extension\n>(\n>\text\ttext\tnot null,\n>\tnote\ttext\n>);\n>create unique index extension_ext_idx on extension (ext);\n>\n>create table getting\n>(\n>\thost\ttext\tnot null,\n>\tport\tint\tnot null,\n>\tip\ttext\tnot null,\n>\twhen\tdatetime\tnot null\n>);\n>create unique index getting_ip_idx on getting (ip);\n>EOF\n>#psql -c \"delete from getting; vacuum;\" $DBNAME\n>psql -c \"select * from getting; vacuum;\" $DBNAME\n>#psql -c \"delete from getting;\" $DBNAME\n>#psql -c \"select * from getting;\" $DBNAME\n>#psql -c \"vacuum;\" $DBNAME\n>#psql -c \"vacuum; vacuum;\" $DBNAME\n>\n\n",
"msg_date": "Tue, 06 Oct 1998 11:50:56 +0900",
"msg_from": "Tatsuo Ishii <[email protected]>",
"msg_from_op": true,
"msg_subject": "select * from ..;vacuum crashes "
},
{
"msg_contents": "> (I have changed the subject \"delete from\" to \"select * from\" )\n> \n> As I reported,\n> \n> \tselect * from getting; vacuum;\n> \n> does crash the backend with included test data.\n> \n> This time I have tried:\n> \n> begin;\n> select * from getting;\n> vacuum;\n> end;\n\nI am attaching the original test script that will crash the backend. \nThe backtrace is:\n\n exceptionP=0x8152500, detail=0x0, fileName=0x8113761 \"heapam.c\", \n lineNumber=1055) at assert.c:74\n#6 0x805a3ea in heap_fetch (relation=0x8187310, snapshot=0x0, tid=0x82f1128, \n userbuf=0x8045430) at heapam.c:1055\n#7 0x8081986 in vc_updstats (relid=141974, num_pages=0, num_tuples=0, \n hasindex=1 '\\001', vacrelstats=0x8186890) at vacuum.c:1767\n#8 0x807ef8d in vc_vacone (relid=141974, analyze=0, va_cols=0x0)\n at vacuum.c:579\n#9 0x807e6f1 in vc_vacuum (VacRelP=0x0, analyze=0 '\\000', va_cols=0x0)\n at vacuum.c:257\n#10 0x807e5ce in vacuum (vacrel=0x0, verbose=0, analyze=0 '\\000', va_spec=0x0)\n at vacuum.c:160\n#11 0x80e2d07 in ProcessUtility (parsetree=0x8185950, dest=Debug)\n at utility.c:644\n#12 0x80e0745 in pg_exec_query_dest (query_string=0x80455f8 \"vacuum;\\n\", \n dest=Debug, aclOverride=0) at postgres.c:758\n#13 0x80e0664 in pg_exec_query (query_string=0x80455f8 \"vacuum;\\n\")\n at postgres.c:699\n#14 0x80e1708 in PostgresMain (argc=4, argv=0x8047644, real_argc=4, \n real_argv=0x8047644) at postgres.c:1622\n#15 0x809ae39 in main (argc=4, argv=0x8047644) at main.c:103\n#16 0x804a96c in __start ()\n\nSomething in the heap fetch it does not like. I am kind of lost in this\npart of the code.\n\nThe Assert line is:\n\n Assert(ItemIdIsUsed(lp)); \n\nwhich is checking for:\n\n (bool) (((itemId)->lp_flags & LP_USED) != 0)\n\nwhich is saying the disk identifer should be in use, but is not during\nthe vacuum, for some reason.\n\nYou must enable Assert to see the crash.\n\nThe cause may be because you are doing a vacuum INSIDE a transaction. I\nthink that also explains the psql -e thing, because that does both\ncommands in the same transaction.\n\nPerhaps we need to disable vacuum inside transactions. Vadim?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n\n#!/bin/sh\n\nDBNAME=ptest\n\ndestroydb $DBNAME\ncreatedb $DBNAME\npsql -e $DBNAME <<EOF\ncreate table header\n(\n\thost\ttext\tnot null,\n\tport\tint\tnot null,\n\tpath\ttext\tnot null,\n\tfile\ttext\tnot null,\n\textra\ttext\tnot null,\n\tname\ttext\tnot null,\n\tvalue\ttext\tnot null\n);\ncreate index header_url_idx on header (host, port, path, file, extra);\ncreate unique index header_uniq_idx on header (host, port, path, file, extra, name);\n\ncreate table reference\n(\n\tf_url\ttext\tnot null,\n\tt_url\ttext\tnot null\n);\ncreate index reference_from_idx on reference (f_url);\ncreate index reference_to_idx on reference (t_url);\ncreate unique index reference_uniq_idx on reference (f_url, t_url);\n\ncreate table extension\n(\n\text\ttext\tnot null,\n\tnote\ttext\n);\ncreate unique index extension_ext_idx on extension (ext);\n\ncreate table getting\n(\n\thost\ttext\tnot null,\n\tport\tint\tnot null,\n\tip\ttext\tnot null,\n\twhen\tdatetime\tnot null\n);\ncreate unique index getting_ip_idx on getting (ip);\nEOF\n#psql -c \"delete from getting; vacuum;\" $DBNAME\npsql -c \"select * from getting; vacuum;\" $DBNAME\n#psql -c \"delete from getting;\" $DBNAME\n#psql -c \"select * from getting;\" $DBNAME\n#psql -c \"vacuum;\" $DBNAME\n#psql -c \"vacuum; vacuum;\" $DBNAME",
"msg_date": "Tue, 6 Oct 1998 00:06:45 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "> begin;\n> select * from getting;\n> vacuum;\n> end;\n\nQuestion: Does the following really make sense?\n\nbegin transaction;\nselect * from table;\nvacuum;\nabort transaction;\n\nTaral\n",
"msg_date": "Mon, 5 Oct 1998 23:13:24 -0500",
"msg_from": "\"Taral\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] select * from ..;vacuum crashes "
},
{
"msg_contents": ">You must enable Assert to see the crash.\n\nI saw the crash without assertion enabled? This is FreeBSD 2.2.6.\n\n>The cause may be because you are doing a vacuum INSIDE a transaction. I\n>think that also explains the psql -e thing, because that does both\n>commands in the same transaction.\n>\n>Perhaps we need to disable vacuum inside transactions. Vadim?\n\nFYI, it is reported that 6.3.2 does not have the crash.\n--\nTatsuo Ishii\[email protected]\n",
"msg_date": "Tue, 06 Oct 1998 13:14:05 +0900",
"msg_from": "Tatsuo Ishii <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes "
},
{
"msg_contents": "> >You must enable Assert to see the crash.\n> \n> I saw the crash without assertion enabled? This is FreeBSD 2.2.6.\n> \n> >The cause may be because you are doing a vacuum INSIDE a transaction. I\n> >think that also explains the psql -e thing, because that does both\n> >commands in the same transaction.\n> >\n> >Perhaps we need to disable vacuum inside transactions. Vadim?\n> \n> FYI, it is reported that 6.3.2 does not have the crash.\n\nOf course, you are right. I can reproduce it with the SELECT, then\nVACUUM, with no transactions at all. VACUUM alone works, but not with\nthe SELECT before it.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Tue, 6 Oct 1998 00:35:03 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "> > select * from getting; vacuum;\n> > does crash the backend with included test data.\n> > This time I have tried:\n> > begin;\n> > select * from getting;\n> > vacuum;\n> > end;\n\nOleg, can you try compiling with asserts enabled and see if you can\nlearn anything new about your vacuum problems? It's probably not related\nto this report, but you never know for sure. I can send you this\noriginal message if you did not receive it...\n\n - Tom\n",
"msg_date": "Tue, 06 Oct 1998 04:35:31 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "Taral wrote:\n> \n> > begin;\n> > select * from getting;\n> > vacuum;\n> > end;\n> \n> Question: Does the following really make sense?\n\nI'm glad somebody asked this, since I was beginning to wonder if I was missing\nsomething. My vote would be to disable vacuum in a transaction, since it\ndoesn't make a whole lot of sense anyhow.\n\n--\nNick Bastin\nRBB Systems, Inc.\n",
"msg_date": "Tue, 06 Oct 1998 00:37:54 -0400",
"msg_from": "Nick Bastin <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "> >You must enable Assert to see the crash.\n> \n> I saw the crash without assertion enabled? This is FreeBSD 2.2.6.\n> \n> >The cause may be because you are doing a vacuum INSIDE a transaction. I\n> >think that also explains the psql -e thing, because that does both\n> >commands in the same transaction.\n> >\n> >Perhaps we need to disable vacuum inside transactions. Vadim?\n> \n> FYI, it is reported that 6.3.2 does not have the crash.\n\nI think I will be able to fix this if no one gets to it first. Looks\nlike a problem with the cache lookup and updating class statistics. \nCould take me a few days until I can get to it. If someone else wants\nto debug it, go ahead.\n\nI am on jury duty.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Tue, 6 Oct 1998 01:34:37 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "> >You must enable Assert to see the crash.\n> \n> I saw the crash without assertion enabled? This is FreeBSD 2.2.6.\n> \n> >The cause may be because you are doing a vacuum INSIDE a transaction. I\n> >think that also explains the psql -e thing, because that does both\n> >commands in the same transaction.\n> >\n> >Perhaps we need to disable vacuum inside transactions. Vadim?\n\nTurns out getting is being removed from pg_class. If you run the\nscript, without the final psql command, and go into ptest, and look do:\n\n\tselect oid, relname from pg_class\n\nyou see the table getting. If you then run the 'select * from getting'\nand 'vacuum' you will see from the backtrace it is trying to update\nstatistics on the 'getting' table, but it is gone. If you go back into\nptest after the vacuum crash, 'getting' is gone from pg_class.\n\nLooks like I may need help on this one. How does that happen?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Tue, 6 Oct 1998 01:42:44 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "Taral wrote:\n> \n> > begin;\n> > select * from getting;\n> > vacuum;\n> > end;\n> \n> Question: Does the following really make sense?\n> \n> begin transaction;\n> select * from table;\n> vacuum;\n> abort transaction;\n\nUsing vacuum inside BEGIN/END is Bad Idea!!!\nMUST be disabled: vacuum uses MANY transaction but\nCommitTransactionCommand() used by vacuum does nothing\ninside BEGIN/END (only increments command counter).\n\nVadim\n",
"msg_date": "Tue, 06 Oct 1998 14:51:27 +0800",
"msg_from": "Vadim Mikheev <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "> >You must enable Assert to see the crash.\n> \n> I saw the crash without assertion enabled? This is FreeBSD 2.2.6.\n> \n> >The cause may be because you are doing a vacuum INSIDE a transaction. I\n> >think that also explains the psql -e thing, because that does both\n> >commands in the same transaction.\n> >\n> >Perhaps we need to disable vacuum inside transactions. Vadim?\n> \n> FYI, it is reported that 6.3.2 does not have the crash.\n\nI am still working on it, but it appears that the vacuuming of pg_class\nis causing the failure of the vacuum on 'getting'. Probably the\nvacuuming of pg_class it not invalidating the pg_class cache values. I\nadded additional debugging statements to find out exactly when the cache\nlookups start to fail.\n\n\t\n\tUnUsed 0, MinLen 110, MaxLen 144; Re-using: Free/Avail. Space 1496/1496; EndEmpty/Avail. Pages 0/1. Elapsed 0/0 sec.\n\tDEBUG: Index pg_class_relname_index: Pages 2; Tuples 58: Deleted 12. Elapsed 0/0 sec.\n\tDEBUG: Index pg_class_oid_index: Pages 2; Tuples 58: Deleted 12. Elapsed 0/0 sec.\n\tDEBUG: Rel pg_class: Pages: 2 --> 1; Tuple(s) moved: 2. Elapsed 0/0 sec.\n\tDEBUG: Index pg_class_relname_index: Pages 2; Tuples 58: Deleted 2. Elapsed 0/0 sec.\n\tDEBUG: Index pg_class_oid_index: Pages 2; Tuples 58: Deleted 2. Elapsed 0/0 sec.\n\tTRAP: Failed Assertion(\"!(( (void)((bool) ((! assert_enabled) || (! !((bool)((void*)(lp) != 0))) || (ExceptionalCondition(\"!((bool)((void*)(lp) != 0))\", &( FailedAssertion), (char*) 0, \"heapam.c\", 1057)))), (bool) (((lp)->lp_flags & 0x01) != 0) )):\", File: \"heapam.c\", Line: 1057)\n\t\n\t!(( (void)((bool) ((! assert_enabled) || (! !((bool)((void*)(lp) != 0))) || (ExceptionalCondition(\"!((bool)((void*)(lp) != 0))\", &( FailedAssertion), (char*) 0, \"heapam.c\", 1057)))), (bool) (((lp)->lp_flags & 0x01) != 0) )) (0) [Permission denied]\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sat, 10 Oct 1998 21:51:26 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "> >You must enable Assert to see the crash.\n> \n> I saw the crash without assertion enabled? This is FreeBSD 2.2.6.\n> \n> >The cause may be because you are doing a vacuum INSIDE a transaction. I\n> >think that also explains the psql -e thing, because that does both\n> >commands in the same transaction.\n> >\n> >Perhaps we need to disable vacuum inside transactions. Vadim?\n> \n> FYI, it is reported that 6.3.2 does not have the crash.\n\nI have a fix and will apply tomorrow with a posting. Heading to bed.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 11 Oct 1998 01:31:10 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "> >You must enable Assert to see the crash.\n> \n> I saw the crash without assertion enabled? This is FreeBSD 2.2.6.\n> \n> >The cause may be because you are doing a vacuum INSIDE a transaction. I\n> >think that also explains the psql -e thing, because that does both\n> >commands in the same transaction.\n> >\n> >Perhaps we need to disable vacuum inside transactions. Vadim?\n> \n> FYI, it is reported that 6.3.2 does not have the crash.\n\nOK, this is fixed now. The problem is that my new cache-use code finds\ntuples by looking in the cache, getting the t_ctid value, and using\nheap_fetch to get the tuple, rather than the older sequential\nscan/ScanKey method.\n\nHowever, when you vacuum a table, as the rows are moved, the t_ctid\nchanges, but there was no call to invalidate the system cache for the\nrow, so when you vacuum pg_class, and later use the cache to look up\nsomething, the cache points to an old tuple.\n\nThis is fixed now, along with a little cache code cleanup that removes\nsome unused code that was confusing things.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 11 Oct 1998 21:16:10 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
},
{
"msg_contents": "Bruce,\n\nI was hoping this will fix 'vacuum analyze' problem on\nmy Linux box when I run postmaster with -B 1024 option, but it doesn't :-(\nAfter clean reinstalling I still get error message:\npsg post\n 9080 p1 S 0:00 ssh dv -l postgres \n13916 ? S 0:00 /usr/local/pgsql/bin/postmaster -i -B 1024 -S -D/usr/local/p\ndv:~$ !psq\n\nregression=> vacuum analyze;\nNOTICE: AbortTransaction and not in in-progress state \nNOTICE: AbortTransaction and not in in-progress state \nregression=> \\q\n\n\n\tRegards,\n\n\t\tOleg\n\n\nOn Sun, 11 Oct 1998, Bruce Momjian wrote:\n\n> Date: Sun, 11 Oct 1998 21:16:10 -0400 (EDT)\n> From: Bruce Momjian <[email protected]>\n> To: [email protected]\n> Cc: [email protected], [email protected]\n> Subject: Re: [HACKERS] select * from ..;vacuum crashes\n> \n> > >You must enable Assert to see the crash.\n> > \n> > I saw the crash without assertion enabled? This is FreeBSD 2.2.6.\n> > \n> > >The cause may be because you are doing a vacuum INSIDE a transaction. I\n> > >think that also explains the psql -e thing, because that does both\n> > >commands in the same transaction.\n> > >\n> > >Perhaps we need to disable vacuum inside transactions. Vadim?\n> > \n> > FYI, it is reported that 6.3.2 does not have the crash.\n> \n> OK, this is fixed now. The problem is that my new cache-use code finds\n> tuples by looking in the cache, getting the t_ctid value, and using\n> heap_fetch to get the tuple, rather than the older sequential\n> scan/ScanKey method.\n> \n> However, when you vacuum a table, as the rows are moved, the t_ctid\n> changes, but there was no call to invalidate the system cache for the\n> row, so when you vacuum pg_class, and later use the cache to look up\n> something, the cache points to an old tuple.\n> \n> This is fixed now, along with a little cache code cleanup that removes\n> some unused code that was confusing things.\n> \n> -- \n> Bruce Momjian | http://www.op.net/~candle\n> [email protected] | (610) 853-3000\n> + If your life is a hard drive, | 830 Blythe Avenue\n> + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n> \n> \n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n",
"msg_date": "Mon, 12 Oct 1998 10:28:44 +0300 (MSK)",
"msg_from": "Oleg Bartunov <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] select * from ..;vacuum crashes"
}
] |
[
{
"msg_contents": "I have implemented a regproc fix.\n\nNow, if you supply just the pg_proc.proname, and it is unique, you can\nuse it.\n\nIf there is more than one match for that proname, it prints a message\nsaying you have to supply the oid in quotes:\n\n\ttest=> create table testr(x regproc);\n\tCREATE\n\ttest=> insert into testr values ('int4in');\n\tINSERT 18665 1\n\ttest=> insert into testr values (300);\n\tERROR: parser: attribute 'x' is of type 'regproc' but expression is of type 'int4'\n\t You will need to rewrite or cast the expression\n\ttest=> insert into testr values ('300');\n\tINSERT 18666 1\n\ttest=> insert into testr values ('int4');\n\tERROR: There is more than one procedure named oid4.\n\t\tSupply the desired pg_proc oid inside single quotes.\n\nOne remaining problem is that you have to supply the oid in quotes,\nbecause regproc has to get a string, not an int. Perhaps we need\nanother regprocin that allows int4 or char*, but I don't think you can\nallow two input functions for a type.\n\nPerhaps we can just leave it. We also output the proname, even if they\nused the oid as input.\n\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 01:28:26 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "regproc fix"
},
{
"msg_contents": "> One remaining problem is that you have to supply the oid in quotes,\n> because regproc has to get a string, not an int. Perhaps we need\n> another regprocin that allows int4 or char*, but I don't think you can\n> allow two input functions for a type.\n> \n> Perhaps we can just leave it. We also output the proname, even if \n> they used the oid as input.\n\nThe int4 vs. string issue would be easily solved by having a routine\nregproc(int4), which the new type coersion stuff would use\nautomatically.\n\n - Tom\n",
"msg_date": "Fri, 02 Oct 1998 14:26:46 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] regproc fix"
},
{
"msg_contents": "> > One remaining problem is that you have to supply the oid in quotes,\n> > because regproc has to get a string, not an int. Perhaps we need\n> > another regprocin that allows int4 or char*, but I don't think you can\n> > allow two input functions for a type.\n> > \n> > Perhaps we can just leave it. We also output the proname, even if \n> > they used the oid as input.\n> \n> The int4 vs. string issue would be easily solved by having a routine\n> regproc(int4), which the new type coersion stuff would use\n> automatically.\n\nBut we define a pg_type.typinput. What do we put in there. I assume we\ncan leave it alone, and allow the type coersion stuff to over-ride it.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 10:45:23 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] regproc fix"
},
{
"msg_contents": "> > One remaining problem is that you have to supply the oid in quotes,\n> > because regproc has to get a string, not an int. Perhaps we need\n> > another regprocin that allows int4 or char*, but I don't think you can\n> > allow two input functions for a type.\n> > \n> > Perhaps we can just leave it. We also output the proname, even if \n> > they used the oid as input.\n> \n> The int4 vs. string issue would be easily solved by having a routine\n> regproc(int4), which the new type coersion stuff would use\n> automatically.\n\nI started coding it, but realized that things like CREATE FUNCTION will\nstill be looking for a string for the input function, so we would have\nto change those too. Does not seem worth the confusion.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 11:00:19 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] regproc fix"
},
{
"msg_contents": "> > > One remaining problem is that you have to supply the oid in \n> > > quotes, because regproc has to get a string, not an int. Perhaps \n> > > we need another regprocin that allows int4 or char*, but I don't \n> > > think you can allow two input functions for a type.\n> > > Perhaps we can just leave it. We also output the proname, even if\n> > > they used the oid as input.\n> > The int4 vs. string issue would be easily solved by having a routine\n> > regproc(int4), which the new type coersion stuff would use\n> > automatically.\n> I started coding it, but realized that things like CREATE FUNCTION \n> will still be looking for a string for the input function, so we would \n> have to change those too. Does not seem worth the confusion.\n\nWell, I've been really confused through this whole issue, so I'm used to\nit :)\n\nIf everything works the way you want, but you would like to be able to\nenter OID-style regproc names using integer conventions as well as using\nstring conventions, then defining this extra routine will let you do\nthat. No other changes, no changes to input/output routines, nada.\nCREATE FUNCTION, if it works now, would continue to work; everything\nelse stays the same. The default behavior of handling regproc OID\nidentifiers as strings seems fine if it does what you need. This would\njust give a user additional flexibility in how they specify regprocs for\ninput.\n\n - Tom\n",
"msg_date": "Fri, 02 Oct 1998 15:23:48 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] regproc fix"
},
{
"msg_contents": "> > > > One remaining problem is that you have to supply the oid in \n> > > > quotes, because regproc has to get a string, not an int. Perhaps \n> > > > we need another regprocin that allows int4 or char*, but I don't \n> > > > think you can allow two input functions for a type.\n> > > > Perhaps we can just leave it. We also output the proname, even if\n> > > > they used the oid as input.\n> > > The int4 vs. string issue would be easily solved by having a routine\n> > > regproc(int4), which the new type coersion stuff would use\n> > > automatically.\n> > I started coding it, but realized that things like CREATE FUNCTION \n> > will still be looking for a string for the input function, so we would \n> > have to change those too. Does not seem worth the confusion.\n> \n> Well, I've been really confused through this whole issue, so I'm used to\n> it :)\n> \n> If everything works the way you want, but you would like to be able to\n> enter OID-style regproc names using integer conventions as well as using\n> string conventions, then defining this extra routine will let you do\n> that. No other changes, no changes to input/output routines, nada.\n> CREATE FUNCTION, if it works now, would continue to work; everything\n> else stays the same. The default behavior of handling regproc OID\n> identifiers as strings seems fine if it does what you need. This would\n> just give a user additional flexibility in how they specify regprocs for\n> input.\n\nBut no one really assigns regproc fields. They usually do it through\nCREATE FUNCTION, and that would still require the quotes, so is it worth\nmaking that exception?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 11:31:43 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] regproc fix"
},
{
"msg_contents": "Bruce Momjian wrote:\n> \n> > > > > One remaining problem is that you have to supply the oid in\n> > > > > quotes, because regproc has to get a string, not an int. Perhaps\n> > > > > we need another regprocin that allows int4 or char*, but I don't\n> > > > > think you can allow two input functions for a type.\n> > > > > Perhaps we can just leave it. We also output the proname, even if\n> > > > > they used the oid as input.\n> > > > The int4 vs. string issue would be easily solved by having a routine\n> > > > regproc(int4), which the new type coersion stuff would use\n> > > > automatically.\n> But no one really assigns regproc fields. They usually do it through\n> CREATE FUNCTION, and that would still require the quotes, so is it \n> worth making that exception?\n\nIf you don't think so, no! ;-)\n\n - Tom\n",
"msg_date": "Fri, 02 Oct 1998 16:23:10 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] regproc fix"
}
] |
[
{
"msg_contents": "We still have some release-stoppper items on this list.\n\nI need to integrate the cidr type into the system. \n\n---------------------------------------------------------------------------\n\nPossible additions\n------------------\ncidr/IP address type(Tom Helbekkmo)\nrewrite system(Jan)\nnotify fixes(Tom)\n\nSerious Bugs\n------------\nchange pg args for platforms that don't support argv changes\n\t(setproctitle()?, sendmail hack?)\nhave psql dump out rules text with new function\nman pages/sgml synchronization\ngenerate html/postscript documentation\ngenerate postmaster pid file and remove flock/fcntl lock code\nCREATE TABLE test (x text, s serial) fails if no database creation permission\nSELECT * FROM pg_rules WHERE pg_rules.oid = pg_class.oid crashes\nhandle oid's on views by either disallowing it or meaningful results\nCREATE INDEX i_test ON pg_class (upper(relname)) fails\n\nNew Bugs\n--------\ncnf-ify still can exhaust memory, make SET KSQO more generic\npermissions on indexes: what do they do? should it be prevented?\nlow level locking - work-in-progress for 6.5\nimprove reporting of syntax errors by showing location of error in query\nuse index with constants on functions\nallow chaining of pages to allow >8k tuples\nallow multiple generic operators in expressions without the use of parentheses\ndocument/trigger/rule so changes to pg_shadow create pg_pwd\nlarge objects orphanage\nimprove group handling\nno min/max for oid type\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 02:18:30 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Open 6.4 items"
},
{
"msg_contents": ">\n> We still have some release-stoppper items on this list.\n>\n> I need to integrate the cidr type into the system.\n>\n> ---------------------------------------------------------------------------\n>\n> Possible additions\n> ------------------\n> cidr/IP address type(Tom Helbekkmo)\n> rewrite system(Jan)\n\n I'm still waiting for 2 patches I sent to you to get applied.\n Subjects where:\n\n rewrite rules patch\n\n Fix for pg_views/pg_rules and new system views\n\n Do you need them again?\n\n Still must tidy up the rewrite stuff and add the checks for\n rules on select. But I would like the two patches applied\n first to get in sync with CVS.\n\n> notify fixes(Tom)\n>\n> Serious Bugs\n> ------------\n> change pg args for platforms that don't support argv changes\n> (setproctitle()?, sendmail hack?)\n> have psql dump out rules text with new function\n\n This requires the second patch, because it covers subselect\n backparsing for pg_get_ruledef.\n\n> man pages/sgml synchronization\n> generate html/postscript documentation\n> generate postmaster pid file and remove flock/fcntl lock code\n> CREATE TABLE test (x text, s serial) fails if no database creation permission\n> SELECT * FROM pg_rules WHERE pg_rules.oid = pg_class.oid crashes\n\n Will take a look at it.\n\n> handle oid's on views by either disallowing it or meaningful results\n> CREATE INDEX i_test ON pg_class (upper(relname)) fails\n\n Don't functional indices need an operator class too?\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Fri, 2 Oct 1998 12:21:13 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Thus spake Bruce Momjian\n> Possible additions\n> ------------------\n> cidr/IP address type(Tom Helbekkmo)\n> rewrite system(Jan)\n> notify fixes(Tom)\n\nWhat about the completion of PRIMARY KEY? I know that we are putting\nFOREIGN KEY support off for a later version. Does that mean full PRIMARY\nKEY hs been put off too? All that is left is to actually set the\nrelevant indisprimary bool in pg_index. I guess it also needs to be\nadded to pg_dump output as well as suppressing the related CREATE INDEX.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid|vex}.net> | Democracy is three wolves\nhttp://www.druid.net/darcy/ | and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) | what's for dinner.\n",
"msg_date": "Fri, 2 Oct 1998 07:29:47 -0400 (EDT)",
"msg_from": "[email protected] (D'Arcy J.M. Cain)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> We still have some release-stoppper items on this list.\n\nI'm a little confused by this list. Some of these are things we clearly\nhave gotta do before 6.4 release, but some others are things I thought\nwere being put off for another day. Wouldn't it help to mark the list\nitems with priorities? Perhaps distinguish \"must fix for 6.4\", \"would\nlike to have in 6.4\", \"future work\".\n\n\t\t\tregards, tom lane\n",
"msg_date": "Fri, 02 Oct 1998 10:09:37 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "> Bruce Momjian <[email protected]> writes:\n> > We still have some release-stoppper items on this list.\n> \n> I'm a little confused by this list. Some of these are things we clearly\n> have gotta do before 6.4 release, but some others are things I thought\n> were being put off for another day. Wouldn't it help to mark the list\n> items with priorities? Perhaps distinguish \"must fix for 6.4\", \"would\n> like to have in 6.4\", \"future work\".\n\nOK, but I normally don't make those decisions myself as to which are\nwhich.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 10:43:06 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> >\n> > We still have some release-stoppper items on this list.\n> >\n> > I need to integrate the cidr type into the system.\n> >\n> > ---------------------------------------------------------------------------\n> >\n> > Possible additions\n> > ------------------\n> > cidr/IP address type(Tom Helbekkmo)\n> > rewrite system(Jan)\n> \n> I'm still waiting for 2 patches I sent to you to get applied.\n> Subjects where:\n> \n> rewrite rules patch\n> \n> Fix for pg_views/pg_rules and new system views\n> \n> Do you need them again?\n\nI am afraid I do. I don't have them.\n\n> \n> Still must tidy up the rewrite stuff and add the checks for\n> rules on select. But I would like the two patches applied\n> first to get in sync with CVS.\n> \n> > notify fixes(Tom)\n> >\n> > Serious Bugs\n> > ------------\n> > change pg args for platforms that don't support argv changes\n> > (setproctitle()?, sendmail hack?)\n> > have psql dump out rules text with new function\n> \n> This requires the second patch, because it covers subselect\n> backparsing for pg_get_ruledef.\n> \n> > man pages/sgml synchronization\n> > generate html/postscript documentation\n> > generate postmaster pid file and remove flock/fcntl lock code\n> > CREATE TABLE test (x text, s serial) fails if no database creation permission\n> > SELECT * FROM pg_rules WHERE pg_rules.oid = pg_class.oid crashes\n> \n> Will take a look at it.\n> \n> > handle oid's on views by either disallowing it or meaningful results\n> > CREATE INDEX i_test ON pg_class (upper(relname)) fails\n> \n> Don't functional indices need an operator class too?\n\nyes.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 10:47:16 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": ">\n> > >\n> > > We still have some release-stoppper items on this list.\n> > >\n> > > I need to integrate the cidr type into the system.\n> > >\n> > > ---------------------------------------------------------------------------\n> > >\n> > > Possible additions\n> > > ------------------\n> > > cidr/IP address type(Tom Helbekkmo)\n> > > rewrite system(Jan)\n> >\n> > I'm still waiting for 2 patches I sent to you to get applied.\n> > Subjects where:\n> >\n> > rewrite rules patch\n> >\n> > Fix for pg_views/pg_rules and new system views\n> >\n> > Do you need them again?\n>\n> I am afraid I do. I don't have them.\n>\n\n Here's a combination of all the patches I'm currently waiting\n for against a just updated CVS tree. It contains\n\n Partial new rewrite system that handles subselects, view\n aggregate columns, insert into select from view, updates\n with set col = view-value and select rules restriction to\n view definition.\n\n Updates for rule/view backparsing utility functions to\n handle subselects correct.\n\n Little changes to system views pg_rules and pg_views.\n\n New system views pg_tables and pg_indexes (where you can\n see the complete index definition in the latter one).\n\n Enabling array references on query parameters.\n\n Bugfix for functional index.\n\n The rule system isn't a release-stopper any longer.\n\n But another stopper is that I don't know if the latest\n changes to PL/pgSQL (not already in CVS) made it compile on\n AIX. Still wait for some response from Dave.\n\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\nbegin 644 combined.diff.gz\nM'XL(\"*WQ%#8\"`V-O;6)I;F5D+F1I9F8`U#QK<]O&KI_57['*:1.]9=G.P_:)\nM.Q[':7TGM5/'33]T.AI:7$EL*%+ADG9\\TOSW\"V`?W\"6IARVW:3,9R2*Q6`\"+\nMQ6M!^L%XS+JCA(EDU(N38-*_\\D8?>.3WO=&(\"]$/(I]_DI_>K#=\"N!4@W[1:\nMK771U5XG`3L?I8QML\\'S_>W=_=T7;+\"W]^*;;K>[QES%\\<_W=Y[+\\2WW'Q&U\nM\\VR[L_/L!:,+C-5J-<:NXCAD+2]-S[(P;,+5SWCGE9=FLUHMX6F61.^]\\.\";\nM.JLA:*T6\"(0\\`\"@$#,:L\\38.HI0GIP(``[\\Q/HW&<9,]?LQ>GTYX^C:)1^?Y\nMY?I+=AI=(R!<Q/EJGXE731O\\W=V0-O:2O3YZ\\^[D06FLE.?SIYV=%\\\\M><I_\nMC`'-R5DV$[\\%OW?<.]/+;/Z*BU&G-(\"8/*#+7^C3<`@,-8CK)AO/)LG02Q+O\nM%CXGHE%!OH.Y5B.(Z`B!*^\\W1E,O`1::2`/,XA7N/Y9\"593-QPGG#079)-'7\nM]`K9<I<L\\%#P$BM3[LV'0!/*J)%F\\Y!W<GG!'^?C\\>^=7$Z6:$A/2.9[EIX\\\nM@,S;>!G5)%<VNI0KU.7%+R?_\\K7)-^Y?LCC^0DLZ]Q+!D_XD`8MUZQ@UYTZU\nMW71`RN;RZ=-J<[ETV//]P1(KN?MT[T5G]]F6LZ_EJI-%J;.W'J`]B_<9>WMT\nM<?23`OFLU^;;;T&0,^\\#/XM]WE#`:HWH=O<PRF97/`&P;P<'Y1F0&4V%;1&7\nM4L'B>3H$QQ`D?)0&<;0I5>W\\NH45;VY7D+QR]>EKR#_-DX(3K;B_5!-LP#OK\nMP[+!R[5BL+W=&6S;OK,V1QEV#^DKO9US-`7H,%@:!_Y!!=`P#0.1(M@;_(9M\nM??;+FS=J.1$XX2(+\"0`7\"0%HG,9UE7#O@_HA93_R!&>7PZ/A\";\"T3UP3H<^V\nM;+5Y>$+15LZCV%4-<)MGIV^:\"N2S^JXA\"K\"VK!7XGP!E<1C:*0UZ,PU\"SAH(\nMZ\"++L=6.AJ<P&((AP.@%2&-^!0@-QT$B4D2A+\"#^0VD2\"2$N/8Q!=CKF=HWN\nM97C/IJ:6*>@T\\2(QCI,92KDQ%V`=T2P&W<,,YNFP.;#\"?1Z->/.@UF^Q609\"\nMXY^`;R\"I;]\"AT!#C)2Q!@W#+4./L<O?\\]%4S)X>'\\:1Q<G%Q?M%AC\\BA,(KX\nM&(Z!*!\"D)N0D5QSNI+M/Q\".+6YP'B0NU'('9''LNR5JXDK^PQ%\\^VF$G7,+.\nMIOQ\\,7\\9KEZR4*]6?B>3=\\KKV&^9/UF+17'*V0UG\"<]@\\Z13SDC9XS$2A6K4\nM(3/)1)8@G#W4CZ,G*4,WB^-F=78.7\\E-(+@:X[&(WQ`^>QC`V&@LC9\"L1/Q3\nM06,-RV6+@/.H*\"/AXX:^3ONV4]I=!J7&2%Y_(?*[F9O%1C_A-TF0\\GX8CSZ(\nM@L$OW*LV]@6@HJU^NK^U56WH5PQ\\MK_U8K&1W][M[&QI$_^?(!J%&<CE498&\nMH>A?94&8!I'H31\\Y=T<0;8&\"]^>3H9AZ?GPC`;X!NF`C!2-V#5:6C:9\\].$-\nM$/66)S.A[\"H1V6$_9SRY92UR3BDH5P<W`4O(G?-/397'])$HT*;+:2`0T:^>\nMN`2134\"Y?!(#4+^M`N-*#_9\\JS-X_DRS]T5B!5VPR(0+&Q\"J,[0+'GJH?#5^\nM#3H:RL!]\\&P/IG^J_5(^_5\\X[VH%5=^O^#B(^`)%+<`L5]@\"\\-T5=P4\"U.$E\nM4<K>7F=[:\\]$*>\",NN8?7B#]X6R4)0E'$<K)V-2+_!!\"OT\"P#U%\\$T%DP&[B\nMY`,#MYXHJ8+)O>:A0I)D(?C;.`IO>^PH\\AEX#_;NY,W)\\24#J\"B%10O0!\\[!\nM#\"$<![,;=<'T3H-H@O8'<'@R;$BG7HH3P]9*N>?WV#O86F2A@1*BC,59RJX#\nM?B,8_J*Y%0Z@,^&3+(0T\":QY`M/\".'(EX*]N`?R69F1H]GIRC!IY2IS*V4_E\nM$\"D+:<.+DADG\\8R)4>*EHZG\"@#P_ZSUE(L9)1QZ*:A9?DV\\`.TD*+2GVPA\"F\nM@D]R-T1_3PI2?F+\\\\7]>I&1KK1AM'%K39R:@^Q>NJ5PV2C,%(`\"@$-:&47Z)\nMCA>DHL9+XE.>D]@C@P<_19H$.+$B25(+/)%BK\"?.2IL_>-[9WMZ66X;,Z&\"7\nM+FCQFCR9BS!.AT@&*-3+/\"1NXYK@)TQV%A-A2@Y\"\"F(6@R&Z9>=O7C$0Z=G)\nMKQ*XKXM3GAL_XUP@>^Z-IHVPHU`M*4-M[^QTMB'Q,:G)%YWCZPEH^88R]'_)\nMCG]Z-91\"5$AQ#(*%/)JD4T5.$T&W)#4JB:0X[NS\\\\O3X!`(YO;9*`8SZ,,%#\nMT!\"MC[#2``'1U'P>)RGW'^F<TXD*<?>J+:[P/G)J817D';)!LP+3#$*:8&Z6\nMX*YD?217`R&1\\CDFH_!T/*7)J0=BJ&7PYY^,!G8/1_%L!MJ.03$N:$'6!5IQ\nMKVE^NXH^26].I:$PYT'1JE>XKTV:W\"BH?H&-@:P/C._\";HG-1H(+R@1VV178\nMUYMXQB.]BR,,6(+0!PR,V,_W4U_-*G7*&!)5.HKG/$J4OL57?W0/`2#R9EQ5\nM[U!5=P9632-757L779`,SL_T7O<2;M.M-CV8>/\"1`6VTPHY:H?!8V[CTD@E/\nM3Z(TN:W56FG(9<GC@@L_GL%\"M1+Z2UY]#<G1$`(\\+(4%(\"U>P[_D/:S1U62E\nM#3F5%R$^P?@>?]!OR1KR]BY&<Y>0PT`-A)3G].S=Y<G1*P9;Z\\?3LQ\\ZK-?K\nM*?#^/WUSEO@#VDL,%O:D6(O#A]_?E822(U.X=`[J*<4KT_EO,0]M187<\"!\\S\nM+\\PK`>W21`3%$\"H8!R-/NBY'?!0,%$0MYUD@U92VE\\ZL\\8K>R[FH]1@=$<PP\nMN\")829`Q+O8JX/>=;8\\KCGS@8I%<!TF*4M.P17G<<IO_,FZ8W4=ST3V,X%.@\nM])5^*TV0`L(TIUDUO[/<$E;6*4A'I:1<*0E(Y))L!$$6MRC#=6M@F6QPP`+V\nMWY=L*9D`TVXWF;20-3\"*J.J6I02%CX\"%`'S&H,/*G*BRL32=6%D*.2Y#;DC)\nM;K+%5.!M\\1OA_ST?@:L(@_`+4H'&8SSD2`B8UC?GEA8\"Q#\":S1MR6IH>H>BL\nM@L!Q+;;T\"<]*H7/B^SL?XFG!,*'D%&6/XC\";1422S`V^$X\\@/LYG<6DR!(-/\nM\"GRDP\"(/_=2F])\"O(T*,CS(DK4$3A*AEHN#BIF2)X'_KD;70EV`L`F[>\"Y57\nM(?U7?L5#]P8ICHD4]$`9@X<00_NWKNE8L%,E+]I$Z@U@-L\\6;1ZV:\"`=G6#M\nMHK\"!()J@7`SC&0@H,AEDJ'M$X\\LE..GKMT#O`R(=KW4/Y<8O1#;JP,99J>\\$\nMAH)*$.#6T(F#\\/.-5#4[#W/3V=1;^HN[1F1P1V$L>`'%.J%I'I\"B<\\.T.$A[\nM&T:7:Y=Z?I1I\\?):CP%:J]ACH.]=[5F(X=G^UM[B<L^SSF!;IWW,^F\"GKT[.\nM+D]?GQX?79Z>G\\$:4%+\\[8^@!CS99_U,)%@A]<+^Z%HD<9SVYQ/Q,>RO0V+G\[email protected]^UU]\\:L*W=_9W!_LX.9+JS/P(O8B>?YNS;G)[N0_TC;'V2)O%N\nMKOVMO&\\IWE_TMW?9%C#^?/_IWM_%>V79[T5G>R>O8.#/72V;8AD;MDRQ@AW%\nM/A=8O\\8`8\\'-))C17WB[7;IMBC6R]&V/MT]>L78+(%C\"0M-JKD$\\$P7S3&UV\nM,@+5\"(;V1)4M.8/.TXH*O@J[17\\$EC`-HFF12]5>A5;&FU5P(64G;L7(&TUY\nM<;2\\ZXW\"^YT-',WGX>T%2\"(`TX:^HE$N>%N>A)$GZ0`2>71IU\\)595Q):4CE\nMO1Q2%\\@-@`1O48$JX#YFZ9HR68\"'Y(;F%([email protected][)K(:LLYQ2=W\"I%&8X\nM#KU)ASF5??FCA:G'<)[$/L2OPJ:\"Y$-3*P&\\RZ[>!-$'=6*%VE<FVH8_C\\K2\nM+(_P.9^K`02\\=`C1I*#?)AS(1MU9/4[=TR-CD:JA9W'TCD*JJN'.V<_$PWA'\nM8?B)IQZ3->*=[4Y^M+*AVK=+BATNU_N'V17M:@\",:^5M\\U^)4TD!^Z98JR28\nMLKZJ`(GI=<!--)39:'Y/JA7=0IW,;SC*;JX6E%Y?+JL]A5&*<+J;>-&$7WI7\nM(:?$:I@)[EM*W:G8U2*[HOTLAMF\\A,^$U^M@PL`[7HZ4E#P>CT%XPVLO(3-?\nMPBKOKX%H-$5V%R-RR8OXS3I\\$V996S^:3!(^^65.<*YM6`Q^3$2]=VFB86_3\nMY\"'I>I7$\\Y]!FTI3R-\\86DI1X+$Z:U$KA85/F3QEIA7.G[P/:`M#M(5R&**!\nMH>73TJ6T5=)5B6'-_ZZQ(_]Q>I%4NA!MW588N,J(!V*<W9T=<^8A#Y=@6<;Q\nM@3Y=+MAX^'T^3X,99*1`D71JEB]JPOW/Z+9,OQ)N*>E!&9W,*#]*K4P'#J3K\nM;R4T0B\"0/LF!A-?,4Z.):B6?/D0([`A134Q-C<3D?@0ARR11C!E@=]\"49)5H\nM\"+TYI$M^HQ`,,(.&L*N.D*ZNS92&YQ=+0[_@AY)\\-(JC46FN?#\"-^&(6&Z-R\nM%OC<PWPPC?&0@9N37)F,DPPZF/5'NC0).:,RJA*!!,0*01:50-3=T(.8EAU%\nMMRSDLQG6PO%P),XB5=7LJ;BZ7M\"6.HL32!@J]`2[#>K.ZDNRU;+;M[0+D*0L\nM`<BIKP+2G7=E7:HSU=\"U4)<:SIVFK5NJG:^@6X$XE43)%B[5+U8`HHKR2[>W\nMK,BLUB#GNJM$=>:V)%6)HXC&W\"NCDDU..;Y\\9?+-0%=*0VED#JZU60&7II:#\nM\"JI?`\"9V\"?\"+TV<#K@-^'(6ACOB5=JFS$92FJU]1<?DC?2AL*P`95X*?I[D)\nMB?)E5EUPN.\\B>?`1^FP.42BX`=P1Z(!N5-L\"NYERJJRI\"%(>I6'I1N,`GL4<\nM2_M/&D]:3WK6N4:S)\\\\3^9-K#B[S&K\"8LKD>31V'01H`!!:1,J\"@QW[$#D/5\nM0^'76)=YMVRPU=_;[:EA?:.N\\W3I48I:?%J=L0?J8*\\Q74V3C)>7QFS]U^BV\nMK(1,'*75\\7W>1%J($,WE<LZ5WZM(CPKX4$LI?BQI1:!`J:V52H_X&XG%0B39\nMI71(55E2<V$Z%FPDU`^B+(Y4$Y1NHS2R7)YLNKM?B=5\"I(?240751AOY$G68\nM.T7'BK0*X4&=.E[&@6RCP2X26A5CY:56_'66L;[8--9@$T!6)36I5E/\"U*8&\nM?U;;&41+P%:3\\V<;12E.D3S):2Q6$=RQTNM$%-*<T_FTZZZ=6[HDD#_4HQMJ\nM2X'&PL#DH-)UZ-:8EW:'M^Y*=C:#>>Y&=\\?G.E9G>4]KODHTZ/%C!:_.V\\I3\nM-]GA2[:=`]JVN`+Z[B26D;@4EPL^5EG%#+8W12F(*QD3V,^/K3J.D8RYMCX7\nMKX-/L@T\";:O(:2L<]3DLA^@0;#9P;4WS<#UO]<AML#(6>2<H[/-3%1*&P0?L\nMHPO$/D5Z*MQ3$NOJ,W%C#&3(,?7@3^OLB8-_`+-.75@1R^:^EV+/FD_=@]B*\nMP;'?&:[!/M-!)QY5T:&2Z-7>>D*H'QBCOM:UL)X<?H[QH346XU!5T>E95`L(\nM<QD'HD;I?F[<G@A874S_P;1Q<,2S8#*E,RV9*OO%J)2Z9.OKE`K5NK\"JDF&]\nMLIKA7\"V4#NL+2X?.L+R$J,/C/!ZQ\"BVV$S(/99AJB^N;I'@L+Y>D!\"#MDVIL\nMT50`,W*[6R!7GB\\^ND^#HO+IV=\"MT6F:F3H_8\"N'5U:#B@M\"?2KHHK[__GM8\nMOY&'S3(01\\EG!603IO4L1\"WOM+P)TBGDX@F7^AE')3@YQ_>L\"Y&0/!HC`/T\\\nM`T#\\>G1Q=GKV0[U>9Z=CG%;$,P[.@W8#3KUB+C,))5JRK<H;A;IY56HEGN]>\nM)?$''O7,*(0\\&H7G$-XEJ.!B&F>A3]VC*`6@@F(LUKCB87S39,&8MC/L[7SB\nMDE53^[JC>WEDMTV7FL_4J+X=W$D;Y.K7*)[?GE_]@=54Y?3<]5*&2S=&6>Y-\nM&B>3/>00^N%/4A^Y:9UY+`,I;\\LYM*-.TH[:ZW940O6_JU\"U54!PG'OI)*V*\nMVI7XT!+-$WD\"[/G^<!8(`61B;4T,TWAXY<EC$M%HJH-WWU?]N5+H5_$D$^J\\\nM/J5F-&IXMGI,W'@;2,-'2EX#KF/<-,6`6N\\\\9CI<C!`*(E.)DMS7M@\"M%30B\nME#(LB19W+?WA[&=WK`4#`.=4I'SO)6=4YG/@J<;+#`O-J@'FX1AG9.X)5XW/\nMB\\D.H*1^K;DF29S-CT.T+/<C%APCJ,#/\"V@P==!%`C*6GKV]>+<]//[EXN+D\nM['+X'FS/.6OG&.V`94N25\\\"]4I8/-Y4E]KNAJA3*ZH7YJX5DK>%]I\\HS/-?P\nM6F9)GG2_!^->\"DO5MEVV\"]S<%^AS`U+'OMKS00QUI,\\O[CJO'0#;`;,T^I2.\nM=`P='?981@6%;,R$+_A*B+H\"R:D\\\\GVJDQ>C='N75)ELZGJY\"2!M);](V5T@\nMNR&E:YMYL-+*\".<MS4#&U6UN]B%$C+%C!IV1I7'R;P&@!\".P$\\V<`@E\\TL\"/\nMY7,RZJE'9@8!'\";3]7)=)3>W]DQNH6\\1T,*]<:\"%^*/1X86BS-5<.6IKMJDG\nMCB8385+.?(R\\_N>?K`Q=B48=J52B,O=*Z/2=7'D6E(_>RBQ`Q^DK2D=JG!/O\nM+B\\.J2/$3$84A6ITZH;1ZE1]B`H7.$5H50-06X:ZW$PJ=L?@1=9R\"A&,++ZH\nM1_),O<?JZ3*I@%U+5X^D6I4D0V*[?:!_EW!9_6'ZMMT>IL^9\\L.4(HZ*-D`\"\nM_JR/5R1AJI=B88'06MZN'<SGUK%XV26C<-]=?G=H%G)UA$,',79/7@FM`B1;\nMYYQ+:^YDW8*85!U\\W3PADE#E&E])=[K5BI.?>*VE.-WU%(?`5M<[NXOUR[EE\nM1I7N^-EL=FO:4\\R9G*V6W?NII3/,%%L7J^;!.@MM]K\"KZ'E=5!>!U(3*JFQ8\nMJ-5E'DJ4\";']/-MGG;8I\\>=]18ZK-\\U$.0&.?W<[B11%C]WUL8M?!5UW*U:J\nM3HLGNTKSG8JE)-5]#T79F,J<IG#=PL>J*E\\%<*L*YIZ>PX5CR)*.(E]'(&4G\nMHMZ5Q9Q#I([L`]S=P2<936-0Y2FY<R);U97\"NK*'$A_(Y?+I*U7E@F3RXO*$\nM6OGR*IA\\:&6,CXH&LLX&TU)%;@2#L3QQ0WW<0<0@997E./V(\"`5%I0-9=';P\nM:]..&5V8PL7%`94'&$ZI2-P$*1ZW`?\"E-Z'O9E,BT:]\"L)Z-V'<+K/;S9:WJ\nM)RD`G_/R%VF#J]@TFT#G!@UZLH(Z1?(=4CY]*K2HV#5H\\W8'FQW9$U+@1%ZL\nMM;S)A%ZW0C\\WI1^0Z3#^03GX(0\\[\"VQ8=VJM23)'7JQKFS($&&%!$.)!^9%O\nMVG`8P4LU[!)\"#F3OSV:D`RI(E1*(E1^2\\M.4%RG'2[56D-)+IAKX:U/2$5?W\nM$#_ONA=8F?ZZNQ?P'2L7?%S<#>HR/B`ZIOV@+AA6ZO9;DM;D`W!A7##.P\"<E\nM1/A=N&GF;W0AX=F'.?>G!1_2^*?0\\D^@P1-B$CTX(:[CJ:_2RO=><5/!E5KK\nMFMI6&N^]I%H-X7[W$#XD:=D<G9_K'5U\"-;P\\R]5<'E2^1.A.]-.+Y_:=#@WW\nM?%G!'4,4DZX!AV%/02`4M[=\"2P(J)V'X.@7]2C/RYLR\\#6NU/AB%4*E*:*E!\nMA1XL4(0J3?AR/U50Y8@\"]^IJK07SHTJ8[M$-K!.@ZAZ&?(QG1_[FVG]/`N!#\nMG@NM30%KL\\$J*NXH=`K`\"R*71YVMC^Z3X?<7]\\?\"4[5?0>!$`I4UO];D5B'P\nM:Y%@E3`?C(9UU<T'?P,9Y+YND;1?[I!%\\H4^E,6DW@0?NPT6A$Q-?-S326`.\nMJE`2WX'8EP_GTH#X';WIQAUE$ZFS6BNQ->F3W=0A4TOW`06=5!Y3MR$LC,?$\nMG(_P]0/6D\\+JA:3Q6&6:@<`:MFS^K3FII,D>%R2/#_)XQ+\\D@RSP>M_<44KA\nM:Z:2*QE9-XG<D)6'SRE7<K9F-KDA8QLEERMY6\"^MW)\"%S;+,E3S<(;_<=+_<\nM-]UT7LJ[DJ$[)YK+V%K#WZY-T-K9YM]%T#^&D/7RSKM3P\\R;*!S/9;WNH5I3\nMUT]!V3\\D!56<NBEHFU6P;L.I%'0EG$Q!VU(@;3<%-6_[6):\"MO,4M*@4[27)\nM9[N<?)I+2AO:\"Y+1=E4RVE;)Z)I*X0K!)*..'%8EHT8\\RWA?F(:VV2+^\"^Q7\nM<U]@_LZ4V/GHO4A1Z>E2<NZX#\"H]=19A67IZ-[;+B>G76P([0_W*5#BIZE>F\nMQ<U9_Q)BUE5)D\\*VUTYA\"SQ6):_M>R6O!6_&F'E+42EYK9ODM<WH5520O!8>\nMA(?L%2_7?E5'HOH]5-<\\P9?98\\):Z,MLXA-UD3XXQ7Y<0F\".3>EGWWU.&W[=\nM^P%\\&$O;OYRJMEG>/?TW)JDF]\"FRM%YZJAB]0W1=WS0;O0/%_U_>M[^U<2R)\nM_JS]*QIVCRU9`^C],\"<^GV/CA.\\XX`-XC[,GN7Q\"&D`;H5'TL,,FOG_[[:KJ\nM]_2,9O0`LI<X(,WTH[JZNKM>7>670]>%>7VQ,\\<0O`+GNB/()5_F`-8G6:X+\nM:SY!,@>PB2+DVC2]MH%R.?!+Q,4L0\\C12X(,N.%>MMNZ3UI;=Z:W9?33YB]#\nMV\"I_(XZ5P_A+N#(>>_\\UG]7/6\\`T]WD+K&?GDR-)G,94XYX[?;GG;TU+W7+B\nM2S+0+26\\?#WX+'!I7:\"`LTVCVG*XDVUIF\\.-SUBVX=:]UK`-]^$W=ZV\\=ZU@\nMP7(!W:;Q2GGPRZ_24.7$JI*6*LSC(?A\\Y\\(.)%*+L%3\\PN[F(E])3_\\TZY-$\nM_L.R].X85[4XZ=$_\"I^_?!A9[4T;&,@&F?_EX\\IH;=K`L%:3\"):/()NM:0,#\nM6%%,6#Z\"'):F3:R3M66'Y2/*;6K*.*X<76<V*FV^ZT?H,INA:`/4LR5Y1$=D\nML002P_KCEUB^T>=F@LSB*?$GDEH2*2\"O2Z)GZA]:EEE.S=F=#;-0<KYNL[D8\nM+NEWZU+/\\L'D\\2#<*!:S^0UNOLN,WH*;[SBKC^`ZN^X*DI0+_:-(4I[8O%*:\nM.A[+P\".C>X;1J>%R&,6SQ0MEO3'K846(SR22K:#CX'!,+9#'GY&Q2*=^[`W^\nM&_[@I7KCM`'[$J5@I!;4];/]!-EM26SAIRN0^0!/$\\JV+5EE@\\>4KK8G2V2#\nM1<L36Q0*LH'B\"`8/P]UG`\\W+X2O&-5\\;%JN^8AOKU-7,\\Y8X8)?[E=$,_A0L\nM:\"HNXVSH`S\"2V6;79B9+J]6/.'6O6E=SE%ME\";-!Y+*%I=5:0/YNQ;H&H[9B\nM\"R;'M6FVR0?&H[-.=IX\"?7,\"GDHF1]RC!Q9'L5&2K4I@[email protected]&#NE<CCS\nMYB!(F[RCKYJA^M>+L4CY'+%)!$IK$<ZH'TVG4`U3+<'MC2Q<UD93,DC^#`--\nMOB`.C<G&))4\\;>;-P8>[(OB2>&:R<4]8N9YW5!8S^.?1M><=IJ%Y?YJJ][P#\nM,ACGIZF)SSL@E_U^\\HKYO`/T,_&;T%RO\"HDM\"CPF)(\\/@2&6_/^AU+=#Z_('\nMI<--J?P+<5@A;ID:3D$R!S!@P7/RTG]\"BT$6:C/)30MN3]Z`D'<A.1+@AK3#\nM*T%!<N1C0F!(HT_9P)%W;#&Y]I%P;$C'CPF!*6,_)AR6I/Z$C\",I@WET85_F\nM\"I1B/J5#0HEZLIA*\\9OU1[#$*+Z[+;=GD+)7R4?XD,*TYD0V)X::92*WD!QQ\nM`L$K?.615)72*O)[C8DLCP_`4L3%VWSR(20>N`1P/5*B+@4X=$MY\\`HD8A3]\nMAB$>-`>(HRF\\Z,L$769\\)BC*,7D_\"4^'`XC,^>WIZ?O3X[<ZN!*NTJ.SL],S\nM@T'9U9HO:`'N7T40!-/1=PUGD+`#M%R040SBANQJUA/AP6Q6OX0PL46$TWD/\nM.<GX4ZRO@/,6&840'+;J?<?A6(Q&5F(*M\\CGW@BA*;[MS1=W)3.6E%7NZOXS\nMYCEQ;I_[^%R%<#,@E7G).ON\"4/025P3H12%)(.:UEW5=Y!7_=8'(*;'6ZD]6\nM#^BQ)KKV91WK&DH!NUP4+[C6Z)?I#HSY3G<&-#&Q0O])&H-X_PD>@6OVO[S?\nMS?;GU0K$._4Y!&:EN71!5'^-,`6<$$)-8BJ(7'*CH2-F9XC&EXX1\"R5:=#5E\nM5XF+\\?RV.!H&++)<XRSL<P\"5*3:+C+^J_&J5BF+%UEJ'R<*=)HI$![>L%)%?\nMG#-*1$Z1M8;KE;/T2'U.:&J0JW7F%ZF<+KU.:$NQNZ8`HX!_=-'%2$FNC)1H\nM@C1LDC@0-$@ZW)?!F8'A41DG9UY/,(P(!PE37&E'K2='Y,F3+EV*.G*QXIJ&\nMFH<ZKQT^PX!S.I>=>\":2R5FI#@J0X`L?\"QD$!17^`9^!GI6SV2@AX`/79D@-\nM8DK&`ASV\\`&?(>LM+3B&VE0,KN0*9D@;?$NL!,PPU``;+L6JD@KB;Y:O+BEO\nM,MYL=TJ)!U]\"]F\".UJ&,B5\"+3>0,\\]*H],\"2X=9\\I[]='SU@^AQT_[N\"/#B,\nM8V!^2\\D7!:N_J]`AA&A+3G1'B9#0G!24_KM(-94X1VF@YR&4<');8(M43\"JM\nM]U@U8`G)XO@L%PSQPJ``F6L:9]\\H0O3@C&GO%5\\S(#_,'+:?Z@-7,!M'!2E[\nM&$]!:H%A%$\\GR!63)8]4EO\"7,AI`*4]-OM08I'QU7T&:\"_YB,IM/!XM)<?>O\nMXP@>O=HMN25_\">\\+@DUP'D=*36^^^&^^-18,O@*MX`I)QA)1YO&\"%G;TVD.-\nM!*W:Q)Q^OZID1-Y]5YT5ST1#@>K':82S(<ZZE-NB,:UBUZ$.Q7L\\M^'I!4W2\nMT:</9Y?G'[]]?WSR=[L<G_V([PBF\\*C>225XX1L[]T4\\8Z*QUATX@!X**0W`\nMX.*@$\\M1T!E_%-IQ+S5&C[MK2@,2L^;N:S9FI06'EG4FD4.[W&(^'`WG]^?S\nMNWG!@,PH012O$M%((M0%.'<;%1(J#V<?^);6&\\7F0KW_=CCNT>A][Q=CWF=O\nM-$IYSS^]&_5N$@\"8\\>Z%YT+\"\\'#WB4TFW\\U*3DFM68^5GH]BI8'2$2OF::A?\nM&TIA!.S8A4MS40F`B\\1>\\%9I/JRW,D]7*F[CN!&DY/,!,\\AN)!>G6];1).MM\nMH1A;@:6`62>#X!:KUF-II(QO/1EZPGU[Z[T8TD3>KHRX2P9WE<Y@FGKQ\\[`W\nMY3(DL)))C.2,7-@@27MO.$YR7*.\\S3\"2^7\"^P`2YX1TDU1-@S?;9Q6TX,W6'\nM(JT-;.9#6*/W#//=HC;>`D9``)Y\",<Y'M\"$9V2S><5Z=?2+[^@0T]=O076_$\nM-2M9VO,[9)FIKC;M:.5J#XRMJ&#EG=:G,NB^:PGJDN2A>5VS8B,S!B45(@8H\nM*4+$CA8*#)^/%\"G\"J$!C>=CAD`8F24@$T3\"A5?'8HT07LU<HI.2;-)3U6_9N\nM6T5IG8O0-^NKMKJ6.3[I^=I-TAZOV^ZFV_-J?7/-V.:]BW)K9G/!NT%?H964\nMIU);^6AW.7+J0+T;58[V_&K.;%.VIA[SX768@M$B'/#O[_@6\\`,DP>3M7+Q'\nMW8L@__D(,USW;T$)%%Y\"[%3(WQG+O3N4R)!)4(<@(8%'DLZ=:_G^TWT`L5\"&\nM`FSH1B8KA7X4AK5\"Q=!V*#W+H=+<[<SFT_[=I*@@Q7R4\"'+)RJADW!W0^;!E\nMFE<A%27C\"X3V2]`U%<$<#YHAA1&RXL,8^GZC.13JV_9RI5KJ6T9RCJQ+_HI_\nM*6(75A%I\"/\\`-T_\"Z7?A'\"WB161IK9+*JJ[LW'W'2\"XZPJ]&5Q)5,BE[WXL4\nMP:UST7AT?WEV?'8)//V2^RNQ9+A\\GBQZ@^(O=.KR_RT77!PDK7.I!7\"FOA#6\nMU%>%N$>Y[Y)ED)GON&QWF!N4L;*,.NN5E^T.>K7;,%G&E_$&S':'MZ+XD&5\\\nM>2[$;'E]KBU_9!EO_OLQZX\\Z%V#9K\\L\\,&!/#J\",EVFV2[5/X9Z-L%=S5)AN\nM@;3U^=A@+;XYZ\"@`UR2XS\"+'&R:W'QK8T<!@Q@'KTK=TZE2LB0;0U*YH/E,V\nMI;DS(TV!^O[5Z2&&N5>L8F)+Z:C`>$6V;3X+)C`Q^[JTX0E2$[,&3]W[16[G\nM%I3N2,5\\T&-):;;.Z,]T\"REI:6;0$GCN'EFK,D:'\\77YT!>3LFQ$.2XCK;D)\nMY04JX\\V@U:':^J6A+$/-=5'HX68@X[VA!P8HZS6B!P8K\\ZVB+1WC*^BYG+$]\nMAI9+*\"A>`R1G(6^%C_5L,0J+KOU08H2Q,W*W@E+L!62IT:\\L18;]5+A+7\"(R\nMS<;[email protected])JT)I'-1CO^K)'448,AP#D8=1[Y6F@E''DWR\"ZH0=8`'R2\nMX<R?`X,CH>\"#(2N;4>2J-YC]:EW$0'V0[(V_@,^T<`^EJH4>$;0S0^UGV/'L\nM(IP;J98*!R_8W_[V-W85]E&P_A*R00374>\"<&(4&-\\6'\\H)2!K$OP_DMNXO0\nM9-T;LV@<*T=]_(WM]>[)MFQJ_XAZ;.3&78%L8`7)\">19=U1PK2C'/5W\"<D+#\nM*4ETM#+]X:0&=3H'EQ1\\KI$9UYQI+>I4*Q@+Z$2`F(`[*1#,!J+43#EE0I;?\nMP>#R;HC)F\"!.XNQR'EU>\\3/HDE,HQ'2GRW6#`21K$JWTV%5TLY@QU%[(0#?C\nM\\(OP<F\"PQ^R+P@>$7F\")A^-WO*TW0#'7/8XA0]<JR0[@%]2AD6\"A;-R/QGWQ\nMUG*Y,F90H9!P&$,MD\"Q^L(C9KFN5B:4.4&I0M0P\"OK34,`)6*66H1ST&\\7JH\nMEW5\"5_JZ5'L'^W!V7KM\\\\_'L[.B$L^VOSTY..:-AM*O5K0*R].8%9*NWK_8!\nM:Q_4M,O<PT\"=;^R9WH.-9JD2/P=]_J6R&+F6BLD7T%MSJWF>@&F-LD!)&DSN\nM?#T<5-:&(G#W/6Z'',[7,E$<GF$6XFB!)/4C9U8/1>QPO$%(^JW!>$;[OP$,\nMS*L6RYX]8SNB\"!/@O1X,T(QEPF/\"H3U)&>,[_@5L2E^&HQ'?NT?W[$LTQ53H\nM*I<YEWR!C:!M!OV;.>\\Q'/.^K^[QV74T&D5?Q`8%1QPSF\"/Z/)-E9V`Y4OGU\nM9BR:\\B,F)`=L51H<IWNCT8[8OAB`8^[09O.F/8'WS?S%OK$08+PX5\"C[7O&9\nMB8@S6%%1S_:O`']`$`V<2O3\\CS\\\\I?WM&'X:L;;4.]Y>0C5!*X44!HP?4D\"_\nMT?A<\"EQ_CI\"R'K@?,Z)L)G#R!I3=H\"$D$WRF)61K)HM,D#Q(N-M,D#Q*M-M,\nMD*4'N\\W5A#_6;:XFUJBZ:J3;)ZUK3,.`1H&A;#S<[DU/K;+*2V?)(6M)_R4:\nM01FPF.ZAE*/;7\\6U\"^T<:;1FN#@R,:)M:?$RH2@MCFWV!F)A;+-738ABF[V!\nM;0:Q]4#Q*'=#34*5;ONH?\\(P_\"KU,0.G?5C;4Q`E2`8E`1NCV2(G[#K'\"^+1\nM_!2MAD1?>%3H2)8_[7JFNNW#Y3>M3N)[4=3_948E>$_O^;<\"ZIW$,ZTIH\\>Q\nMIS@?HWNM79)BA'XRM/K3\\K^0I>3UHR^WPU&H1:R_2KU!7(\\BF<%RV1RY]A)#\nM?%]`440\"I=R.75VS15RQW\\*=BN%8!@R1F@[04L3$OP3]#B\"8E[\\->Y/+:!*.\nMI\\IZB'YJVIF-/^&/!Y=$*ZZE$.OW1]$LA(*2/FWPE#U0-,&L)JD,SB]3UY$<\nM[=$;+F?!-@ARE*9;O-NA<DX`85NJ'[B:4J38!VS(YPDK[;T:+^[>(S&Q8;FL\nMQ@$OA4*1E\\$__QK^+(:C]8JDN=HQK[7)$]`><\\&M2?*M8_K<<?*MQR<_W2)G\nM2<Z*.\"Q0Z)%$K[P[AM])QBLYDV1@_3;L_\\(XLX5:.KIVH]';A[>`R0^\\P$RV\nMZ\"%;GSKP)/I\",X:W?=QI4YP&-NK.D>(A]/#$ZM['Z2E(O3!^.[1+$+)4$?KJ\nME,$+=#'ELM$&*6-U(_1=@Q-7\\L<G])EH33V(S:\\-,JR\\O:HNK77[A6=*=P\\9\nMY8VI]\"U.^=)S2\"D6P_$WCGD)&^JIV#W=@,5]H(4+IWKAOVDV'(0]H2(&X/@N\nM<;,8]:9BL>.D!T`N*FX4!DZ?S</>@!J@@G\";;#&.%1%O1[W9?)^]'M_S;?ON\nM#C3DO,=^M!B+0.VQ:V#$Z_+OT7003L_PF*-G1)PQ'V0)MKY:J=V3\"12YE287\nMT-#[\"GF<G!4L4D'OFHXNH02P?=:;DNWZK/=\\60/NR1X34\"6DK=_M?4T4(G.,\nMJ2J!T#/V8.768ST7:B9H17-$*N:5#QUN,^J=W12M@[*,95`FLXN<&=F(>!*K\nM*ORP97&A^I>%8UU;CLH)A7&XV/I7_C\\'\"/XQEY]B_*B;W+\\>#Z0F,VX=9,*<\nM1D0H]I[@WU[8/_\"=-5JUH-%J,WRPM[?':K5F(ZC56A7&O^U)6)B`Y0#ZAYN9\nML#4(IA$?\\%^@48#+D_-;SCC?W\"+\"B.IH'X>5A$/<]T/2Z03-2H4@X2,X@#2_\nM$=^8J\"&YY-'D`T8<\"\"A!A,/7(B-]G#C#<$4@,1(M?IA&_7`VD]LM(JSH.T#C\nMG)#Y4L[3]:AW$Z#1S.\"!7IAO;1=^ZS:V>U%#T[)TZH?%C`2&>%?GLK&YT%(^\nM!!3QY3WR[%^`$YK-3I7/9J<M9W.3:-TTP#ZJ:%::0;-:550A[,1X;@N#,CQ5\nM.Y\\Z:*VGT\\-_V^/?(6!@`>\\$7YJS!1[YYO=#,9#\"`77)SXUC\"^\"^SNMA!R\"0\nM\"3A0(-IGYV@6$!/1K?&)Z(J)\\`VUU6@&K6;-6(J-.I^\\1D,M1;08[^7^D34/\nMZ,-P?!T1_QHW*XIWG\"7\"[5`@P+^`8NT$JKK8@FC9/(LA7\"T>[`3AH@H':XR.\nMG<_#\"6OX4=NN!.UN6U&1H%R:O)G:<??`XF.TO`<L@S@-/TS#\"6&!]EMZ6Q!Y\nM[`9<*!Z.9C)+B\\'B(I'`2IO=<P3<4:U!\"!L\"B-74A],G7[N626(O`0B/,+W'\nM68T]7.56FX`?RVQUV^.[LD.]<`6IQ]GXFS&1+=:B(FP0W<$%?2X97873F31X\nM*=E$O1B:=87)7=I9]]F[X6\\GX%S#-`+8D`MHP+'QH8UZ?3\"44>47X_`+='\\=\nM3L-Q7R;.050*7PAA-3,O_5-5W,QH=0+:YL@/PHH4[UW4`+X+CAG-6=A<*`/D\nM%A\"['N<&VF\"T:F)`#]0.-!_1`_=J&3TE9Q:Z+`:RQA[#QR2OQ^RW*,$G@PHB\nM/59W;W,:TCVV/D@5[@4,DH.<CRP]@:'9PV*$EX+73\"4YR/E(]*Q&RLL*Q^37\nM\\_G)XJXX'03>BW.R8NQ=Q&0D(<(;L'0\"<E.T&5`#\\$Z\\3U_LT6PN%MH)%W](\nMP2I7/,JD\\@03RE>E)1/*,#Q44=-W=4_5-#E:6T'*TI<,WUX:5,E;@.%DA6,7\nM-&;IJO9,'RZ^WJ\"!!-<?11&FUT^<M%+6R\"5I#&TO('O5B#*7<L=0OF183&CU\nM:$RBI%(*J6)JW):^R'HCSZ+D$O!&`N>(2<;J)%1\"9$KQ/0:]M;SHM;W&C&H2\nMFE@[IB),](W\"E5716H8TUW#N]N\\&XNZFF%XK^I)JK^![R^S03'N&EDCL+W=P\nM!P%5945=,F`V9*9V4G0CEW-\"KP;@>DT7#%9T1[\"BQI`E64(!)>SND6V#@!6Z\nMH03Q5DAV>^HR!REN#-E6;*DN\\9`P9SVE_F5K7YTAV`TL'8I5W!F2/#=,/Z9#\nMZYUQ0Z&B9SL/+NS6I=>%TE:BS@F%5E$^KMD2LVP>2HR1LZKU1&FWK*>.LZI3\nMPUXI]FM3Z>7,1.R`<%O2YP7\\E7([#>003Q%FR.9\"`Z3\\<?W<68`1RVU90UXG\nM?@&RS\"7?W0>+_GP&0M;OP*B^N1-1TK2B4NU1HO`E'#A#2UFDV`G.%IB[445*\nM-J]GLW!J;.Y(A70?FPJ0$OL;<X.WM@TJA7>#J\"1M%1\\OCM\\?7_Q(>B!;/!02\nM3;/#)9I.5TDT\"9RX_/>(^\"TL0[`7OSZS%9&68[M*/.IL^Y7_G-O<+)9!5H#?\nM($#A]BUUI--0>#8/N(`.SB+(I7^!IY]#E:6;:DH,H;[email protected]\\]9\\[#/\nMI8,!%3N`/S&:D2:2LD,R91NXCQ@+<$B0@=NUYNW)(8[SO5]N[U$[##Y[MUQL\nM^)OL-R^M^F3(3J4>=#I*$T'$ZU)`(E>.S>Z88,0M1+_[?*)#7`/A'5098J3=\nMQ630@S@$G,SY]*.3(&<%YZ%E(DG@O@Z%ZC<+%69AN*353C-W_.-U\"$YO#G)2\nMK(SY^::=W'S3SO+]WE1NVVWOF&IKH=YVV4QZ:C!$.,L>7B@1+8:Q*K;SF,X+\nMJ%'S:\"]9K&WC%4%C/+`V2>,Y@6D\\L#?.F\"70%&`$T0[\"<&)NU-J)E384HNKA\nM?\"&3X6\"M+[!B'?F)2_A\"Y,<PTW)@,QE_&BOB?N25QZ0K;A$<<\\GL`UV/AK^$\nM)7D'`)1T?%/KPWX'YF`,+0B@J`*Q`1XH.G'G23)S/CJ9_\"JL%UKBFM_)NZ2J\nM%*19DLAV;-V*+YS\\RJG&[KEDN_'PADW_'<G737Y5WFA4(DVPY\"54:0<JJ=`V\nM'W.Y_<Z]E!4_0,T:KN5?6`#M.D*SO\"?<S(5TB>S;P0O\\X]TE!1V(`JC?,?`>\nM\\0I3N2,;NVV,=16:WADIDTCIBYK23Y\\^L8]C\"DTYA7B1$\\YFJBLH-[WAV+YC\nM@CTF2]:\">284$*_J5;EB,X'/2F'PO3;[@[=_)#>KN\"MAQP$]7@]4C*\"Y'8Y!\nM-S'E)_9TACO3%U1O\\_Z%6P7I^%#'/;R#@U@X_F!+?7XR71D:MWWVEI])_3E8\nM??KW?5B.?&+`9TXN,]`V#GA[/3S5PM_ZX;0_G/%U1YQBMQG4FK6.Y!17/&]I\nM<G.>I)R5(/,R[\"*B/FHJ]^G+@?#Z6?G,6^W(RWWB93KPEIQWM*5X3[O-''8)\nM9UW:49=\\TKD'7<(YYQYS\\5-.VK,,UX3XQF18:I_XDO(RM)UVT.6BF&!HA:@%\nMH@V'\"O<;<3P<CS]'OT@5@MP^4'EIEI+Z4:;4G-%86AL^#WMV6=*`<GJ(9K/A\nMU>B>$9)A9!4(P<ZXD'(OVQ,85_I1(0WN64TF*T(3;\"=QS:;9G/`8]L287U[0\nMN@EC*1#,BJ?CT('AJU1+T\\`]O>`+>HL7DP2FZ7R23LXBB/90>\\0,(D`LLC[1\nMEQ!LO&/Y<I]:^R=(=$`\\]\\1.]2/*R0!UT/`CYW0411-Q11-QR4M.<16&`VKI\nM<V\\ZA(UN1E88],2WY</P-YB_X=@`=#_![.5#M'D51\\QP_\"K.GG$5AV:`KN(P\nMYRZ.5E?[KN/L&8ZT>_%+,0FOZ0Z(I:?#.R!HM[:3[S'RK%9J.=]XE<N5D8K-\nMTF=Y0,`+`0GOR)??`D_O[@8L2@NIO?F9Q'HA\"5CMO+\\,1.6X;T$B:1B`T8D7\nM#/]]A;&\"Q4G3-F.RO/$`Z(?N:*?AW!BO8\"?)]Q`6R0#\\MFF1,/9/N.U-FH59\nM!*;4GEH2O<&`;[>ZL@1671\"72V4XAXV9KWE*80AMP7+:5U4/LI\"!$<*_E)5T\nM8F'[S9H<#1QS[M;TJ\\CRH!7B.[R@FOPX>BTC#/FUX-7$T3C\\;5Z$NLJ#)*T-\nM:1P*YR6S*5G'2C;S%M$+#O;LEO]O[$HRD0#M.'J;C-@=YQN&DU$H]Z3]W612\nM58[]>^2<P\"`?%H6?9\\B*S]2<Z5I?]7X?W]0)S1R_RP]-=TOTGGHP4XD'7]HQ\nM3K-/-(^6^E]`]@;)G<%0YK<H+??450#3?\"EN\"[#>O6$^%R><*_LG'G'(X<0T\nM!<J;S#K9:\"HU0-*G8D:B%HO&JM@PG&F&\"1L3CE#->H/+$_5NLO]-M=*H!/R7\nMZ8'3[%:\"6JM;MZ40H;(6O9#2NBP<]E!AZ9WH<M)$EQGI1NV)+F>8Z#*?Z'+Z\nM1#MJU%4G6BMPLTQTV4*(J4$II[K4E%=RJ2D[+C5EQZ[NX!=YBW)VEQJ!9'<<\nMJ[C4E-=PJ2FOXU)37MVEIKS$I:;LNM38NOXE+C5EX+_*\"2XU9=>EINRZU)2]\nM+C5ECTL-K8-U7&K*Z2XUY>4N-0*&#\"[email protected])17=:DI9W\"I*0LO:?S@\nMNM24&;Y+V/V6;WTH&-*G:0;)4+8CY$,E&,H%[MTYDU>TN,>%_8/KZZ'Q4!@D\nMI0JV;-[Z*NL,@R+!E;/)HB=BE;Z(1Q>HW9[-(9+$+`SO@(V\\\"RD`S]64;]=C\nMBA4D?$HY\\N7Z%PW,(E(>Z-WD3DIP>L%E$'5]H-8L4/7]OW$TWK.,@FI2.\"FB\nM&J+BF0L)C,*KA[M<!E$]`2+[=I>\\E2B/4=FS>3](05&B22_$1`7CLA#9(.7,\nMZRL`^\"2P+S(2@ZP7@7$V*FNV7AJ#X?4UV^M/V6S:WP?_EH.K7I_/^N!`$+[\\\nM^T-O/)SL]Z'<DB+(M&1LKO!N.F2GG,%F-59MOZPU7U8JK-KM=I#-6=Z76[_U\nMLM*E^CY.JET/VAW-1?&OW:;APUPXQ<A#7$@]L</[:`DW8!2=2%S30%V<YK++\nM;IR#,LF091WG`-V\\[%`'0G046UDR%'9X`P.2,D%2]D.B0QQ8T*@0!R1QNE$.\nML@+EBVS@@K:TLAG3('?EE2JI.`;I:&2.ZH(1!ID*O8O8L\\+O\"L1Y&?EF.ZBV\nM##:^VFX$U6[+)$&5.2V-!$<#BD&%?FORHQ.$<&ODF01AC#RS0/D0I)L$<`+I\nM9@9[:<,.66^NX8TW:\"Z'%:=MLTNEWJP']::Q5]<[[:!1,9?*63B+1I]#$#+@\nM>H=,%TCI+_0US^*S8M%2:Y9TA'TFK2S>Q9.Z:O+T;BZDDAW^OIP$0;85D0<*\nM=X64O)'ITP%:OS\\WX/RV^WNH?JSP\\)EF-77!D)0(JMA+$8L=_',YKU@TUTY)\nM1&KWKZ%&LQ$TFEV]AIJ56J`T3<P,]\\D>[/2`KPF1^\"P\\QS)%R.A\\IK@+'9HS\nMI?VY05MQR0656'$=HX_B[\\4*/,(!E14G\"=DD'A,SN6%WUO^?\"O8_(\\SFOO1T\nMEE#NO2_[KM?L-H)FU]CU6IS);EN<PS*SI[SU:2#[/X?AESB2#<MGR<0\"AK6E\nM6*(&;MT\"$M$:G[GVGHULPOYQQ=/T//#8-KVO^H>9E)UG.X/-#$X\"B_18X#P1\nM,\"P&BRV#A>6$1?`[R=JPQ7PXFAWT!O,#T++A-T<9YBWAUX5YBWI4836_*BQ3\nM];;2I'EWR5K0JAJ<82UH-^0.^>_#<7^TX#.P2QV-9O>S?J]_&^[?[EJO^[UY\nM;Q3='$QN+OLC/C\\I[^'*'+PN^U_/;GN#Z$M*`9R_E/?1Q`O!]=W-E)Y)_[\"8\nMS>]]U.^-V(`WA5DD9XE6.G@[X/O,;#Y=<#3_8WK__9#OKK@WV<\\*!5#JX[T=\nMK9(W=/)?9<'#!&,DVJ\"_&T57`C)ZX!0X2`C6$+2,PZ_=#CKJZ!-V1DI%BM!<\nM\\@4CKMSM\"H7Z\"_;N[/0'QG$J+1___/[H[`@UVR*#Z']4=P]U<^A3^&(RZHVA\nM-0PUK:\\#^;L4A7)U\"0I]\\T%M]]`VG6H8>G<*@G(\"!%@DWC]_3%U#>V*@\"9T(\nMBEO:DRX7[TZ^B_7IIU>O2INO7\"-*3J<;:'TBW3\":A[\\)1_\\;@`?=<#D=%T\\X\nM(M\\\"V;^8\"K-;O\"Q,E+]L&<LR^I'%<95\"^5-T,^5?,*,7%E=-Z.*+63B]NA]\"\nM0*1Y@RV&@U*>D7<J0;<AG41IJ-8,8.HO.=COP][D8@'N-?!DOI@$#+^^#6=]\nM>C3OEPZ]34@<9&QBQVX\"QDFT`&T(TX[T8?*4)7^C7%7(2SQ7%7(PSU6%/-!S\nM5<$0+'!D\"S\\]>5#;J6RUWZ2X)\\F9^PD_\\(>_^5N]7HS[RULECIY4FEE:G6<!\nMU;7H>YIF\\:8Q4S`UC@&-V0N9?]A?7MTLQL1XL)3D*O(4UE$8[=((>@U-[HL[\nMLRI\"C`$0+X?7EU,9OC%Y3BAU,#!)L,%TNT&UVK!WF#[email protected]]O5UB)\nM1BM)]5.798;ZJ6LT0_W4!9NAOEZ]1M'-+=Z41E=>NREM;F'IECTT0YZCZZ'M\nMS[0E^`[F&F=\":O)H1MM-I1LT.Q7#5]'/BAL<!^>7]]@/O>%X9&0@H:Q;C`^-\nM?0G)YP8%/,R`(EP<\\,(!\\5G2H;F'^4^2^7J#C^%?E[`RTD<'.:7\"\"UD,.7VU\nM515N>0.#W[P/,3AT[/EX$'\\^FPPOYY#UDS]56QX]G??Q*6K*X,&U\\(AZ%TWO\nM+J6X5,#>^M8+Y#?IQ2CA'8(BWW&.;7%'?8\":ZU_5G_$Q$!L]A=RJLW_5?K;@\nMF1K@C<`157[Y);P7D%(+5XOK?W6JW9K9*B_C>_QB%DZT3T[<Y9&OSS'&68[8\nM^8=C\\/WIW83395Z`O\"@L\\C$&.H?[9?#@].^7;TY/3M3M?\\NWVR2.EW`5\"NX/\nM][V=[Y;2`#XE8D4G'][.:`39Q>!((K=2+F:@[^4HBGY93$@X$37!,=:0>9=Z\nM.AIBD;R'HMP;.7E#`L7I#0J_<GX+(.N0(,`K*Y6:*E;YF4LJI\\=O^3\\J#\\48\nM84^,HFC(6@$$\\Y:5A>9'0F;!Y.#:;*[$KGO#$46Z93_M_F7VT^ZN\\**B3DH:\nM$B7A00.SWF=P-QUC;Z4UQR)P_@`#$CTYH]+2I']HRN/+1W'OX\"XH49?8(M@<\nM.2OA70M;G0@4B6_3\"(LV,X`C[$W[M^?WLS>@(L)]JGA\\\\O;HT]G1^^.W0BM(\nM5XN/!]^%<]Q1BL:.&L#-N@JE[A(4NZ,VP>/9?_9XH2)U5_*L1ZF<D@L%L$EC\nM^\\M`8'@W<(51L36\"_MC:,DO?'5V<7YQ]?',A>SS,BD^:%\\*GB43E$KT<FQ36\nMWH=0CLO3%&3B8\"!7X\"I(A?#:&?&*;=MXC7>M\\3MR4$P$[:*8PGNOB>5PD!G/\nM>+ZMB^><6,8*&\\&RB6-Y4B_#,?6>AF,^1'+ROV.H7Y/7=K-3L.(0:#OUHFZ$\nM@X#6Q&:M&0BL]9P]=QY7\\?%/%?5\\VA=;'T3PF1A'6\\`D!/0)ZP>8-EK,#-76\nM1[P9W\\>:$[$M\\[,7;C205I`HSMA;,,D)1U6TF(/!S!H>JLA*NF?:]='OF;?+\nM(:ANNU?!.@I<\\4^HS@\"[*>):EYGW8V7XIP%G-%69:[3`0IEK\\OPNBIH<Z-X=\nM])K.YYS/@0_7U(3!3X;+Z6G*.<;K(F<%>3]OSHY>7QSQ\\>/)PO^RTQ/X_?'\\\nM^.0[^%#<I45KK)3A;#$>\\N.4_8WM?CPY_L?'([;+7K)=43(=C;$R<JW%R@!B\nM0*7=&RW\"HL`[D2\"B2&\"P).U=R>RK3A,B=S2=<3$54R',\\RYJJB7S7+'6#683\nM0=:;,HK0Q[\\R1.;E#Z\\_7?[]Z,=S\\;Q<-KA##`NH4<H+_`L+_0Q,SO'X,VQP\nMX(1P@H11<LRY^'4VG_9[\\R(!QO$12M.?@)O/KBH;0PPZV0\\&%E:NA^%H@*M`\[email protected]`L%#0%NU(AA.(#O<HYX5X\"NX_([A3SBM[MW:SNI%[92*1T;\nMQ]<8L*C'0),\"BPBCEI/#AD!4-(%X]WS.Z;#,CRG8JXP)/B6Q5YGALVSUU!`\"\nM(`=VJ&M[]OQ\"XJY?2-GW%3N>MOD7EF[_R[9BR773?OP7YTCVC3)MV]]\\;UFV\nM^T*F#;^P=,OGT.D]W[-,6-*;?!OEUZ7\"#\"=P=PVP6>^>/8>GK`C+<%9B`IG/\nMEPG)+OGOQ,B?J%^K9Z\"85,\\H'0H\\Q#?G:!%6:UD43F`[/YR=:K83?])6%;3E\nMLIX)O*?HMN2316-\\)XX[@>W$/JW1T`!-UA/+,,UZRLX/S7,`24&?J[JEO5?\\\nML\\G#6#5V^0GO>4RTY2M?8IHKR;UC5>3*2MJL$O>J=;>JI3O5^EN'&ESJ'K7!\nM?C+M3IDVIYQ[DTD0N78?W'SD(9^T^1C!%G#;@0U';C?L7P';W]__.77;^3V5\nMG%/W/LY`SVY3F4)K+:0SZ&_@NKV9-02N7U-\\'(J3\"E<IP[M(7F7T]S@*06G&\nM^^4?H-\\2*[/_?'WV_=NS\\_^2XBHJ3:5JK3<:1?TB+TWCY67/C__KJ\"B+E>\"B\nMH]`=\\][[D_LB+_+V]<5K701L`7R`T/6>ZBQUK&^',ZFKQ9OC.57%UXAX2U/\\\nM[OCD^/S[!$6QL&:\\9%9EL<$J0(7IPK0EI(8I,'TGT%`\"XGL/0E%.27R_NL>X\nME/>SX0\"4Q8:IY)IC'=S(8\"T_ESE`BQ^/WWXS+CU/-)&8OAO:1!)WWY#&$7U(\nM0AEY2,IS@IR]\\-6EM#/('N0%Y[0Y?`V$`Q0[F_3ZH1%B#\"JFS:*@;DA(J_Q9\nM(&,<$N+)ZQ^.@+K>'YV4)-'-PKFX9+OW\"ORP\\,1U\"R[3J:@1B\\RDJ+T'X38C\nMU`*%\":S#Q_.C\\Q_/DW56\"[^:*L8IB%Y*!JLC)\\@\\XVDDAGY)UJ,=EQ_E)V_X\nM2K61QL]YV=;>*_Z)3GGY-H9X:_<5FZ30!A3)REFB]HTTMDC\"?QE`KMK%<)!R\nM^3EM<:&7G>F_@(N+0PTV7^1V8,:$+N+LX_LC)J)64W`86F\"<&GOL9OB9[TF&\nM0QL>E#$W/N_-MWHUJ#=:9CCG^)$G_!XX`L+/F&UJEWQ8Q3=16IUWRG-\"U8.S\nM[E\"&*!O.@`NA\\$>Z!3.B]Q((*$B'AD$$[5@+\"M6&A$.$J)8O5\"Q(<-T5V;6`\nM4F5(L!GF'KZ(P&ZMJZD0Y.[,P_+[=C$<#8P@/EI3A49B4<I,C`+7\"ZKMH%7K\nMFH%XMC9?#S`9&\\5S>7.H]BZ5;C5H5&HJ>Q9=(`%`1.PUF87-B<>F;G@(W+M#\nM$\"',F\"\\H@PQP(C\"A)L615LBC<U?$'3;?V:XU(@\":=#>K!!CP4^0A*UGIZ^P.\nMWIYBZ_*\"2ZNC`M+F0`-LK=)+A?_\\>KM9S&!4B]M]<K,R'&7Q*=7G3^TI60V-\nMS\\#%)C?RO)<FJ\\V@43/R^LF@.#*5G!R>2\"F*%P>6X$)&GW;AUZY9*F8'<\\OM\nM'K)=E1,TAJ#2H22\"=J4:M\"LZG>(#PAW0!K$J]-Y9X(=@H]XVLBLF+>$L=+@&\nM\\BWHOQ*F^8;?KM6-^V3;ABT5P0Z$_@RFM:\"A?:4?CZ78W?U?RE`@7;0Z0;O=\nM2H[CUZQ4@F9%<7:R>TR#Q3M_7GW._OB#0<N@27CU#:O`]YWA3.:HA*\\P^7<3\nMN1=S;.V6K$PANR=H'0#!<U?&&H<ZG\"E7Y!7S;#57@(^ZXJ<X#IB35=LX=QYM\nM/&G+)=MPO--5[03-1E7?,(A?P'&<+SEL*>[^,K6/#/@KCGXC\\4U)G]YT$5+G\nM87EII>)-F40[=H78B%1C'S]P2>LHWEC,USA+8\\<GYT=G'LABCL]9&GO+A^F#\nM+.;%G*6QD].+[X]/OGM)5V+XLNOH>/WKSJ//?QRG%JYZR?M?P$F1\\&GR/[+\\\nMH7CALD#$+F^5/@+V#/V]-T<D65K,1RE96LQ'+EE:5#3CW0N:[:#9JN??\"]*O\nM\\\\@M(>8,*Y]AN&GP>T5\";C2\"3K.2GY!SW&3(#I'W\\F&]&;0:!O-L[KYD^6>.\nMY5_D.G$YH'FR@(3.^Y`9@C,.X*?.7K%JJ50RDF.S-^#+#D%0OX3L#F*V@H+^\nM]3FHY&6V:1F:'4*N8Z1)O%H,YQ:&)\"C199=:+>CJK6,CX[E]R$%X)ZE3Y>*B\nMR5LG#,L=F%O2\"2*ZHS-4VL,2565X!T<6VU6]^1J_E@E)838X@]75UY2W`+A@\nM1LSW#B2`V\\I&1N2=FBZ$JJA8'*(15MUB47_?L-9!]9)$ETJ6/M`^.=^=G7[\\\nMP+[]D;UY__KC^9$,:MWE)-;MM\"W.<\"OCN-TP\\-X;MI56T*YJ+9.<]@3:,TG/\nM*]RE38,*^9DX#BGG\":&CVZT'W:X1TFESP-WFA\\A_0;G.Q>5N_J,S_5IKOJ.S\nM6JE\"\"/5J*__AF>,:WWJ'9Z<*?*J5QW'U>5SM_/1K:7;U\"B*SC&]1ILF424K@\nM-]'DWHA33@(V7\"Z+6$_8^`;\"R`=&,VDFG8LI[;:#:K5B)>+:!,)NMXBE&%_*\nMJ68[*/)26)U36+V5?RFF7Q?/N12K59@WK4#;$!@;78I-\")M1-_A8AY\\PYS4A\nMAGG\\5/LF\\41+6JWRJQVZWPQN@V3JZKL)S5Q:J%:UN/\"`@[A=&7+O9+2:0:==\nM?6BF:/D(;+[#/PG-&DQ\"YT$XH74A]N=BJP6=3B?_EI$>+B)M>?H2+A`R6Q![\nMH-W,OW'DN`J?%S(OTKK=H%MI/C8;3QDV8^ROH\\6M5CLMCM9N[5&X]>4P^A#<\nMK3:Y7+Z\":G;MB\"0IY$'8K%7XBJ]55E`WKAMO(0TT+Q+KK:#;,.0:)<^ZL1(\\\nM:3Y4H,0=F:I+:N#T,\"1R%?1FS%FF8E`;FHO`B8,0BT%+*.82!?]EQK'=,.2`\nM]PU\"[<5^HQUTFW6;V[<63^_F9N\\5_V4<Q<[**^XJ[47RJLLZ\"]@=A=\"S1N3O\nMNJ25'#(-*7\\_6$R*Y,!*$U7G+$BM4;$Y]&T,,LN$;7:`WCEM=H)NJ^%94=]-\nMH\\7DS:C')0GVXF:*2;\"M9YM=4;P'3IM`\\5F75*/)9ZI9]2RI#8&>98;[email protected]%\nM?[L1=#LM2[,IC\"D4J36:*#.*J;Y#&\\#IA\\NC3Q_.7JH,\\RLN*8PXJ<8$%2$1\nM;C0ITEYA%^5E/828HE\",015-\\)99L7@ZH5\"O<J3AM`1_R*<]N=T-C-8:['1X\nM<YMWM\"D[2*O*Z;)=MW2^VYC21/+\\7S>=J2/=SE1ZUVJ'[Y7=CG/^B7D[$_/F\nM'6=Q0V2[V5DM<:A9\\:FLJ/A11LNI#=M\\I^D<R!O$^@.NHSP8WR[1Q['M3^-9\nMJ8(.NNJE^=<G;_^,1,_!?OI4S[>::JW;\\5+]9O#^H&2?`^=/@^ZK7`RHU\"I>\nMNC\\YO4C#/W_-GBCM*_JJP[JN:]_;+8WO(6DL>2YK-9C+CLECNRXU<NCO/IZ\\\nM<1@P0S30D2\"739R5G<%]&1,3',D&8@]6Z]6FG?1F/8#39F)58/W8KK<A\\;-A\nMB%@MPQ:-)JZ^3,PXRQL1238PQ2Q3R605QO\"^N>TO<4W^$KLO3H[^^4*Z<;JD\nM-0Z_[.N54^,[0[UN>*UO=GS*MX/B5?)Z<F`4^Q\\N)NY(L?7+[7`4%ODS#AFD\nM`&:_WH)KWNW>*QG,78BW.?`&M2TM;<GH$/6LJGD9-HX]>Q:#DE(2\"YK,V'\\Y\nMJ?^-36%\"KG(@V6;=WB\"L^OW%%`:\\;\\A!:#`5GW]7BP\\`5$O&Q`L?WLXW:@DJ\nM*+7XE^IYXRVS:U!E`_;S1MO>,E8?0DZOH@T,PS\\S8#ZI.\"*76&_GBZOWP_$O\nM:LT1^.(I1D85D6V!\\N1C9_DYMR\"L:P]%47WO%?]`K@+JS%-YLT?RB6V@M=:+\nM:@>.N%LP>2OKQ`ZS4)_(QNEX2>*!O\"DSHJP@5O-JU?WN/:')PT\"]<9T,<O(K\nM.\\Y!;?C>6&>V*/?5.]02.SXQ5`%?-?ZRJE4]/AP[\"0K2'4?G9\\O]'K6=-)/4\nM6W6^QEK6M<ET8C1=F\\THR]9A3*?N,MUA@D&D6JV\"):Q6RV]76CLF?9K9T1-R\nMAG#8X5MMO;N\"(\\6Z4;AS0NM'=J,.D>2[^J`PSR,1\"P9.F\\5HA-=]Y'%D>+C)\nM2R`&->=8;W)]&5F\\;,VTAU-VE]HY[#WJ%$A69#9`!FK4#&^,+8_67A8/,U+_\nM/#>KX`+17CK/T?A)3S67U++.-<A(#=-M?/LCWO1T9QJM?[ZEEX9]*WNYYWRZ\nM\\T#BE(X\\QKROL9M?PIPG)JA1X1/4:#DWIM<%T3L'^<#S8[0+9E3M%IY]G]],\nM4A.YVWMV=L:LR&?,C'Q&J&XU.:K;U?QGU$:2.JP(NG<::A50(U;U@<7B_IH'\nMTG_&]:>;WT_NH@'[JY3CLIEWF7(RB/$T:5<_X.3>4#=N@DRC+WV%UC#$`0!*\nMC;(9&/\"Z,(3T@S`T\\!?`,,QBQKIN\\ZVGT=6'[,9GR%[B6YF=K%UL:V8R]I]I\nM5OSKJ`Y&J$8STSJ*Q0?DAQF,I+;!9>3TP=N45YB=-T;*@.UU;Z1BP\"\"4)Q>-\nMT^.WFUM*/B_S=\\-Q;S2ZYS04L:N0S:,Y?IWUKL,2I5^!&$@#R'BPN`-=%=Y'\nMXR5[M/\":%2XX-?7MBHU/:9YUM^'I7*/KW%.98^UM>!K]*[4!)UZK9JQ4G$5J\nM!H)8PNH7B9%(`5D-F,%UQO9\"9:7038A=$5O9X*K\"H'2@>%(=P55JRMVT((V;\nM_P38[DK[YVWTA>VQFVC.AO-]=L*_04C-J<Z\\,>D-!G#'0\\:V35Q..RPKE!YQ\nMP`%7N-Z*U5SEQVBS7F?.XGV,:<^S^K8QY6NNR`>8[J627\\)4^U=\\&T2-MI%`\nMMMH$U\\\"FX1JHM+/A;X!91#^PW5S,Q8`KG#C0J_O0JR%2^9Q`\"(;GK<;/$H,P\nMB7+\"Q-3)&\"Y61!`03).8]$X7;/V:N2`@@6!%Z$!6?`%9@2^O.>*GQ6?7P_%U\nM=!DMYI/%O%2*]8^DXYI0\"\\\\(K(#M54LJ*(F5L.#Y7V;/7[[$?`T$@1GT&`X%\nM>4SPSRKH,6-)>E1]HZ0)'G&MAEZ<#SY`,7%\\D&*`RP>VPTQ#B-F`B-^X6P+S\nME^\"\\)64820R\\^(UAE[\"UDX#%'29TX?R?O59W9%!54[&,L1\\_]*:SD&'`U)XT\nM#E!QI_J!FY=0AEY8/R$@;PEUX#FL,2N98E\"0$'GP`%Q\\Z!AG7,L,/A\"5C$5.\nM9H'8'\"HZ2#7BD,(?RHW\"\\0T_7.)&&0HK$%.)24T9Z6DL,\\]R&\\_O\\?8,ZT[.\nMB]->FTZZ04<9:E8:>4F9?-3=4[]WJ.Q&>M&:U`!TA-ZT1'!DCSGZ='Q^<7YY\nM_O';]\\<G?W_I:9I*:)N3L+*(X0@OJQ^=%ASQ56!&02-D2\"^J8SZRIMQI>,EZ\nMD,#A2(?S_?NG`>?[]ZEP@@S_)`!-`'(07O<6H[D$S0R+O1@O)I-HBO>6Q2Z&\nM\\AD$D5<&31]1NAUI4O;9-3/>O-YA?JOFDJ-X,+R^9GO]*9M-^_O1='AS<#4<\nM'T`$T\\&5^+,_NX6WWA?(JZ16+;R;#MEI?\\Y8C57;+VO-EY4VJW:['>0\"DMIU\nM:[5?5BI4RQ]DL1DT&A7-Z]&#JF&3#?NW$=O%`/'`I$(4-HQJ##%;=_5[BHA\\\nM\\?K;]T?L-_F>%0N%GS\".-?]!F02H#$_W`-XP>@-EO2]T2%A,MEHZW&5_L$DT\nMF]]PQIW]QX?OSO_Q_O+TPP6;AW>346\\>5OGF>#`(/Q\\`Z\\);^?>[Z'/(^6ZZ\nM+B_!\\J.\",[F-IKHY<O<9VH>@U`=7?+D=J!X.].@2\"JAN3/P9$:-_VKT\\.[J0\nMI7[:A8@^IR<R9>W%J8(3@I0>GYQ?'+U^RW&R0S@1Q23.`AF*7>;;EB]*T*R!\nM0-4`8^_.3G\\P(E/GQFHR5<#?-*K`]T058HKAR0/-/7:N8*,09SJAX/G1!1.N\nM.<\"OR/+/Q=U6XY4:R'.`Z2=:-XU6T&@IEORI4\\^;?3D>7ELMS``V7H-.\\.=L\nM7U(4E%5DIXI\"68<$=167\"%.HD)T%>C+>,!\\H-!,&0#O_ESW_/X\".YQ8XX!W]\nM9C_B$O8WO'3XF1I%`L(R3X?6R_H%%X'\"Z9]X%7CW5\"ZI-EJ&*J'1Z@3-6BU^\nMO/@('%OW$SCAUB)P9F^/FKPEU@-!4V6+9E4&\"UX#IZ`DZ^`W<R[$CZ@H\\GH4\nM91\"Q.*4SE]()Q0J3M[T9+51.KX7L_4!<51T1];D@ZW(>LL:XE`EDC;L!I^MR\nM$EU3@65'N_G&(6WQB@^?\\OG-4'?DO$+,^%[,.;]T$T[I7=[U4(ZO!QJ.&N[R\nM!4$5$E:$>$E+(N><N``@3PS]\"SV*MW_CW;K]IQ]:8MJ33RV%2)=T[%5-Q?S+\nM6O1AK>MRTKK6QU:.A:WI,3\"JT&*D_'&\\D*;,P&K96K/GBDKM0M\"?I%%R(J>B\nMBFX-<D[<&G[A_8,39_'Y]'G`GL^>E^QJ^`/'7-(FL=T]0F`G99,0)?@ND6&3\nMP,+);R`+RRJ'GV>Q\"[!RK'91(V&YR[?;7.\\.!,Z\"7PN\"]!4OYS!YR6MTIJ]Y\nM4<Z_Z&4W*:L^@5VUU_VQ64B15.#;&R19%3_M&UG+53U,^^-9I+1!?#+Y5./S\nML;U$\"XIAE:PH=H9=%1)6\\[%94H`EUBY;D6U]'_70N,3!A\"1ZT^$$P^I[U!?#\nM<7^T&(0'_=Z\\-XIN#H17U#[I,!+?VHJ,Q&)Q;4:U:FLS,E=-5VG4*HU*P'\\U\nM-=<I'RE7\\[='YV_.BKNS:#'MAYAM!FQQ/91NR`U/))F#N(_L]/@MGY,J1#4N\nMNFP9S.\"'[S[R$GPXUVS._X>_M2;;K79WX<(O@__@+[N.(G;5FS*R\\T@84/V.\nMAH,0#?((\",:ZQ_0I7D!J!B`J\"YP?DFJ7[=;J*9\"4)216[CAHI_AE.+]5Z>)*\nM*?#4#7C4\"DK&3*V5!1Z9*EP3K;2S'@@72$X>\\PC=*S#GEK2?SI@DG7X*F8.E\nM9'8P`>L.?G3H//[:3^CQ<AY*K_DI?7G=]LMJ)YG46ZUVT&K7-*'3`VT8/\"'C\nM>Z$`2\"(C^7C.]Q[*MW%8P\"#38(*F!]H@/>5D,`=%,%K]0?<*Z?(`Z=`2$,FA\nMJ@M/^CUP!,!<=<I<!/G7IR'.R\"&4[4VGO7M^@%Z'TW#<YW.&C7\\%`UOO[B32\nM<YL\\9T)5<#\"*^K^X$^:\\\\\\^64\\@S50W_5\"VIV'I9Z2;/4[4;U.J&/9U_;9@B\nM\\&\\<U6-&#KIWO7G_]CWT4GQSAR'_6?B9[PL!.^.;$[Q@+V\";0D#(A(BWX<A^\nMR_$NK'Q(5?-I&.*2$EU\\AN.E#T'%H:$/X?1N)OR\"17-N9=M&*=??OX=\\:OD\"\nMYY/Z_O3-W\\\\OOX>YC$_;/)S-.>;@S)H=A+]Q,IF'@P/DF?>CQ1P1O*2,/8U+\nM\"L>GD^],UG3F;\"!]6EOM*JXVO?SX@XYVDZA5*G^<_IU_*-;8-/J\">9?X7/SC\nMX]'9CR]97Z1<!6X&[O7-Z$SA9TP/D-X(V!6>3`'[S$C475JW)NMF*%MW^\\E0\nMIQ&'K9_4'XH*HBJ-JS=CXK#K07W,O6J.^Q/L'E]N^>[`CX6_@@\\/%NY'B_&\\\nM^*+D5JBQ'QF5_G&_ASQ3+P,8M:5@4)N?ES=53VN*^GJ0(34,.'B-@/^Z\"D0?\nMO)T2O.:[+CX`>.*(#VP8--\"?$`S>\"#RZ@8A?P!O(3N*@R1/8:`\\>585;%IG3\nM9D@R'\"IH]/E2E!`D`,=_5#&7\\J@WOEGT;D(N%O\\Z>KX</TUGGH(8=,MQ9/0B\nM>!^4*$TLHA<-%WBK7%[G)\\4=JX+D/G]>RE&W)NO6\\M>MR[KU_'4;LFX#ZE[G\nMJMN4=9OY^VW)NJW\\_;9EW7;^?CNR;B=KW9J>HZT5;N0IW,Q3N/T$\"XL5^2*V\nM;>)BZ_UQA5(Q_[A75OF0:W_0ZN`?&_2QP3\\VZ6.3?VS3QS;_6&R(0[?,,G1:\nM2^ZT2FU6[?[K]+&>V'^'/G8`E%8>4.J9QI\\PZ'J>GAIF3W\\8^Y_LLFSDHI9]\nM_R'T%>8DR&=5`S#SF810/FODGI]F'E#EC,GN*@G@RSDTRV4=4HN>M<RZOF%*\nM,C#+%3N>H?L73SUEG:BC\\:^LO6QA9:8K_[KRTA6%U/9VL^J8]'&_\\PUK8CZ#\nM*[3RUIZO-T)C$1LK-V&Y-AYFL/6-35J.S6#5<312MA$Q:77V2C%1&89F[SY]\nM:RV;\"];::*SE6,TWR,8Z@VPJRGS%A4-*M$%#Y9QI9?WAJKVH8FXN%7,'J:P^\nM7(]`UX_N)JR(%H/>=(X&E@\"^+<;#.?KT<GX$'\\R&_Q,6KD=1#[;;)5(B5+[N\nM\\>$7O4W!JVB:UIC)O\".$)\"8H\\09@#5CQTSY`Q9'\\XSZUB1P\\/+L<CB_[=ZCO\nM-[EX;$L).@K,'ZF@E';@.0H\\\\\"&56E0+DK=Y?L=9R6JELE\\IY:S8QYKYZPW'\nM_5M>L[;?7,(XXMA5M0G*)=!G<UF?3L6:K%C/61$EDOXJ7394S6K>JBB.\"!RU\nM<]9M&74;^XVE;\"M6CZ:#<`KR,=`HU(\"_?U@T:3$JN(E/JGPS$*Q!LP(K?%+3\nM3^KTI*Z?,-SD)PWC216?-.63:GN_#:?8I\"6>5*O[U78K%QN*XZ%%H>$'OW-.\nMWGJ<QJO%#&Q,KS8ZZOB(XJ..C[&IQ[A$(3G[=20T?_Q37#MHO4Y10UKE/!K(\nM5HH&,K5NZV4US7.WT@@:,I,TNE;Q!\\V*TOW[YG4<S6_'M<.TM_5#4E*NI)W,\nMI);,I8]<21&YN@:2;49?MPG=XZ,K'9^BMO%QU8Q;UB_R?ROI%E=2*JZD35Q)\nMC;B2_G`EQ>%*&L.55(6)]5S-WV9*-3*5:F8JU7ZD4OQ?NL8O7367+D2G:[C2\nME4J)(\\BC>DD5WU-$[E4Z7Z(C>4A0ZEOI<`550ZK0G2(>KP+#<DW`JM`@;\\Q9\nMN0O^%#U:@`]?S'ODU#(<,_+=%P4W*]/3O\\V(]/ZS4HCS[#'E^:0I7R;.YY;C\nK5Q'@O=66\">_YI?85Q/55Y/25!/25)/.D%>>7RE-+KR3S_C^-YP;Y!>P!````\n`\nend\n",
"msg_date": "Fri, 2 Oct 1998 17:45:06 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> Here's a combination of all the patches I'm currently waiting\n> for against a just updated CVS tree. It contains\n> \n> Partial new rewrite system that handles subselects, view\n> aggregate columns, insert into select from view, updates\n> with set col = view-value and select rules restriction to\n> view definition.\n> \n> Updates for rule/view backparsing utility functions to\n> handle subselects correct.\n> \n> Little changes to system views pg_rules and pg_views.\n> \n> New system views pg_tables and pg_indexes (where you can\n> see the complete index definition in the latter one).\n> \n> Enabling array references on query parameters.\n> \n> Bugfix for functional index.\n> \n> The rule system isn't a release-stopper any longer.\n> \n> But another stopper is that I don't know if the latest\n> changes to PL/pgSQL (not already in CVS) made it compile on\n> AIX. Still wait for some response from Dave.\n> \n\nApplied.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 12:26:04 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> > > SELECT * FROM pg_rules WHERE pg_rules.oid = pg_class.oid crashes\n> > \n> > Will take a look at it.\n> > \n\n Please apply the patch at the end. Disables use of system\n columns of views at all (not only oid, cmin etc. too).\n \n pgsql=> select cmin from pg_rules;\n ERROR: system column cmin not available - pg_rules is a view\n pgsql=> select * from pg_rules where pg_rules.oid = pg_class.oid;\n ERROR: system column oid not available - pg_rules is a view\n pgsql=> \n\n initdb and regression tests still O.K.\n\n> > > handle oid's on views by either disallowing it or meaningful results\n> > > CREATE INDEX i_test ON pg_class (upper(relname)) fails\n\n Fixed by earlier applied patch.\n\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\ndiff -cr src.orig/backend/rewrite/rewriteHandler.c src/backend/rewrite/rewriteHandler.c\n*** src.orig/backend/rewrite/rewriteHandler.c\tFri Oct 2 19:32:27 1998\n--- src/backend/rewrite/rewriteHandler.c\tFri Oct 2 19:43:15 1998\n***************\n*** 1777,1782 ****\n--- 1777,1784 ----\n \t\t\t\t\t\tvar->varno == rt_index) {\n \t\t\t\t\tNode\t\t*exp;\n \n+ \t\t\t\t\tif (var->varattno < 0)\n+ \t\t\t\t\t\telog(ERROR, \"system column %s not available - %s is a view\", get_attname(rte->relid, var->varattno), rte->relname);\n \t\t\t\t\texp = FindMatchingTLEntry(\n \t\t\t\t\t\t\ttlist,\n \t\t\t\t\t\t\tget_attname(rte->relid,\n",
"msg_date": "Fri, 2 Oct 1998 21:05:55 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "Applied.\n\n\n> > > > SELECT * FROM pg_rules WHERE pg_rules.oid = pg_class.oid crashes\n> > > \n> > > Will take a look at it.\n> > > \n> \n> Please apply the patch at the end. Disables use of system\n> columns of views at all (not only oid, cmin etc. too).\n> \n> pgsql=> select cmin from pg_rules;\n> ERROR: system column cmin not available - pg_rules is a view\n> pgsql=> select * from pg_rules where pg_rules.oid = pg_class.oid;\n> ERROR: system column oid not available - pg_rules is a view\n> pgsql=> \n> \n> initdb and regression tests still O.K.\n> \n> > > > handle oid's on views by either disallowing it or meaningful results\n> > > > CREATE INDEX i_test ON pg_class (upper(relname)) fails\n> \n> Fixed by earlier applied patch.\n> \n> \n> Jan\n> \n> -- \n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #======================================== [email protected] (Jan Wieck) #\n> \n> diff -cr src.orig/backend/rewrite/rewriteHandler.c src/backend/rewrite/rewriteHandler.c\n> *** src.orig/backend/rewrite/rewriteHandler.c\tFri Oct 2 19:32:27 1998\n> --- src/backend/rewrite/rewriteHandler.c\tFri Oct 2 19:43:15 1998\n> ***************\n> *** 1777,1782 ****\n> --- 1777,1784 ----\n> \t\t\t\t\t\tvar->varno == rt_index) {\n> \t\t\t\t\tNode\t\t*exp;\n> \n> + \t\t\t\t\tif (var->varattno < 0)\n> + \t\t\t\t\t\telog(ERROR, \"system column %s not available - %s is a view\", get_attname(rte->relid, var->varattno), rte->relname);\n> \t\t\t\t\texp = FindMatchingTLEntry(\n> \t\t\t\t\t\t\ttlist,\n> \t\t\t\t\t\t\tget_attname(rte->relid,\n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 17:34:05 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "On Fri, 2 Oct 1998, Bruce Momjian wrote:\n\n> Serious Bugs\n> ------------\n> change pg args for platforms that don't support argv changes\n> \t(setproctitle()?, sendmail hack?)\n\n\tI'll try and dive into this on Tuesday night...maybe even\nafternoon...will use sendmail as a basis for it, which *should* hit most\nof the operating systems out there...\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Mon, 5 Oct 1998 01:05:14 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "> On Fri, 2 Oct 1998, Bruce Momjian wrote:\n> \n> > Serious Bugs\n> > ------------\n> > change pg args for platforms that don't support argv changes\n> > \t(setproctitle()?, sendmail hack?)\n> \n> \tI'll try and dive into this on Tuesday night...maybe even\n> afternoon...will use sendmail as a basis for it, which *should* hit most\n> of the operating systems out there...\n\nLet me remind you what has to be done.\n\nWe need to use set_proctitle() for those platforms that have it. I\ndon't have it here, but Linux does.\n\nSecond, we need to determine if set_proctitle is a cheap function by\nlooking at the library source code.\n\nThird, we need a 'configure' test to see if the modification of argv[0]\ndisplays in the ps output.\n\nIf modification of argv[0] works, use it, as the current code does.\n\nIf not, then if set_proctitle() is available, use to change the backend\n'ps' display on backend startup, and if it is cheap, use it to change\nstatus for every command.\n\nIf a platform has neither, use the sendmail code to change the ps\ndisplay on backend startup only. The sendmail code it so slow to use\nfor every command, I think.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Mon, 5 Oct 1998 21:56:50 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
}
] |
[
{
"msg_contents": "Attached is the new IP type that Tom wants installed in the current type\nsystem. I am planning to do it, but if someone else has time to do it,\nthat would be nice.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n\nBruce Momjian wrote:\n\n> When do you think you can you look over both versions, and send me\n> one good version to work with?\n\nI've got something working right now, by taking what I was using, and\nchanging it to work the way Paul's code does, with some enhancements:\nMy code is ready for storing both IPV4 and IPV6 at the same time with\nvariable length storage in the data base, and it's got Aleksei's index\nintegration in place. The type name is still \"ipaddr\" -- but that's\neasy to change, of course.\n\n> Even if you just say, \"Paul's is better, throw out ip_and_mac\", that\n> is all we need.\n\nI am definitely *not* going to say \"I can do this better than Vixie\".\n\nThe way I feel about this right now is that Paul's code is better than\nwhat I originally submitted, and better than Aleksei's improvements on\nthat code. However, what I currently run has certain improvements\nover all of those versions, and I kind of like the way it's going...\n\nI'll append it below. Take a look, and let me know what you think.\n\nOh, and a correction: Paul Vixie wrote:\n\n> the ip_and_mac type is host-only but has an unfortunate bridging\n> between layer 3 and layer 2.\n\nNo, it was (and is) two different types, just packaged in the same\ndirectory because I thought they conceptually belonged together.\n\nAnyway, I'm appending a shar of what I've got right now. It's only\nminimally tested so far, and I'm *not* a professional programmer.\nI'm basically just having a lot of fun with this, while it is at the\nsame time useful for me in my work, and if what I write can be of use\nto others, that's great! :-)\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n\n# This is a shell archive. Save it in a file, remove anything before\n# this line, and then unpack it by entering \"sh file\". Note, it may\n# create directories; files and directories will be owned by you and\n# have default permissions.\n#\n# This archive contains:\n#\n#\tMakefile\n#\tREADME\n#\tinet_net_ntop.c\n#\tinet_net_pton.c\n#\tip.c\n#\tip.sql.in\n#\tmac.c\n#\tmac.h\n#\tmac.sql.in\n#\ttest.sql\n#\necho x - Makefile\nsed 's/^X//' >Makefile << 'END-of-Makefile'\nX#\nX#\tPostgreSQL types for IP and MAC addresses\nX#\nX#\t$Id: Makefile,v 1.4 1998/09/08 12:23:30 tih Exp $\nX\nXPGINST=/usr/local/pgsql\nXTARGET=/u/tih/databases\nX\nXall: ip.so mac.so ip.sql mac.sql\nX\nXip.so: ip.o inet_net_ntop.o inet_net_pton.o\nX\tld -Bshareable -o ip.so ip.o inet_net_ntop.o inet_net_pton.o\nX\nXip.o: ip.c\nX\tcc -g -O -fPIC -I${PGINST}/include -c ip.c\nX\nXinet_net_ntop.o: inet_net_ntop.c\nX\tcc -g -O -fPIC -c inet_net_ntop.c\nX\nXinet_net_pton.o: inet_net_pton.c\nX\tcc -g -O -fPIC -c inet_net_pton.c\nX\nXmac.so: mac.o\nX\tld -Bshareable -o mac.so mac.o\nX\nXmac.o: mac.c mac.h\nX\tcc -g -O -fPIC -I${PGINST}/include -c mac.c\nX\nXip.sql: ip.sql.in\nX\tcat ip.sql.in | sed s@TARGET@${TARGET}/modules@ > ip.sql\nX\nXmac.sql: mac.sql.in\nX\tcat mac.sql.in | sed s@TARGET@${TARGET}/modules@ > mac.sql\nX\nXinstall: ip.so mac.so\nX\tinstall -c ip.sql ip.so mac.sql mac.so ${TARGET}/modules\nX\nXclean:\nX\trm -f *.o *.so ip.sql mac.sql\nX\nXFILES=Makefile README inet_net_ntop.c inet_net_pton.c \\\nX\tip.c ip.sql.in mac.c mac.h mac.sql.in test.sql\nX\nXip+mac.shar: ${FILES}\nX\tshar ${FILES} > ip+mac.shar\nX\nX#\nX#\teof\nX#\nEND-of-Makefile\necho x - README\nsed 's/^X//' >README << 'END-of-README'\nXPostgreSQL type extensions for IP and MAC addresses.\nX---------------------------------------------------\nX\nX$Id: README,v 1.2 1998/09/08 12:10:22 tih Exp $\nX\nXI needed to record IP and MAC level ethernet addresses in a data\nXbase, and I really didn't want to store them as plain strings, with\nXno enforced error checking, so I put together the accompanying code\nXas my first experiment with adding a data type to PostgreSQL. I\nXthen thought that this might be useful to others, both directly and\nXas a very simple example of how to do this sort of thing, so I\nXsubmitted it to the PostgreSQL project for inclusion in the contrib\nXdirectory. Since then, that directory has been modified to contain\nXAleksei Roudnev's implementation, which is based on mine.\nX\nXFor those who have seen my previous contribution of these types, note\nXthat much has changed: I've modified the IP address type to work the\nXway Paul Vixie did with his CIDR type. In fact, I've pretty much just\nXstolen his solution, modifying it into my framework in such a way as\nXto facilitate the addition of IPV6 handling code in the future. I've\nXpretty much ignored Aleksei's C code, but I've added his SQL code to\nXenter the necessary operators into the various system tables needed to\nXmake the types indexable.\nX\nXIP addresses are implemented as a struct of fixed in-memory length,\nXbut variable on-disk storage size. For IPV4, it contains the address\nXfamily (AF_INET), the CIDR prefix length and four byte address. For\nXIPV6, the address family will be different, and the address longer.\nX\nXThe external representation of an IP address generally looks like\nX'158.37.96.15/32'. This address happens to be part of a subnet where\nXI work; '158.37.96.0/24', which itself is a part of the larger subnet\nXallocated to our site, which is '158.37.96.0/21', which again, if you\nXgo by the old book, is part of the class \"B\" net '158.37.0.0/16'.\nX\nXInput and output functions are supplied, along with the \"normal\" <,\nX<=, =, >=, > and <> operators, which all do what you expect. In\nXaddition, there are operators to check for networks or addresses being\nXsubnets of or addresses contained within other networks. << tests\nXwhether the left operand is contained within the right, <<= includes\nXequality, >> and >>= do the same things the opposite way.\nX\nXThe input and output functions use routines from Paul Vixie's BIND,\nXand I've snarfed the source files inet_net_ntop.c and inet_net_pton.c\nXdirectly from a recent distribution of that code. They are included\nXhere to avoid the need to fetch and install the BIND libraries to be\nXable to use this code. IANAL, but it looks from the copyright\nXmessages in the files as if this should be acceptable. Read the\nXdocumentation in inet_net_pton.c to see the legal input formats.\nX\nXMAC level ethernet addresses are implemented as a 6 byte struct that\nXcontains the address as unsigned chars. Several input forms are\nXaccepted; the following are all the same address: '08002b:010203',\nX'08002b-010203', '0800.2b01.0203', '08-00-2b-01-02-03' and\nX'08:00:2b:01:02:03'. Upper and lower case is accepted for the digits\nX'a' through 'f'. Output is always in the latter of the given forms.\nX\nXAs with IP addresses, input and output functions are supplied as well\nXas the \"normal\" operators, which do what you expect. As an extra\nXfeature, a function macaddr_manuf() is defined, which returns the name\nXof the manufacturer as a string. This is currently held in a\nXhard-coded struct internal to the C module -- it might be smarter to\nXput this information into an actual data base table, and look up the\nXmanufacturer there.\nX\nXMany thanks to Aleksei Roudnev and Paul Vixie for their fine work!\nX\nXI don't know what changes are needed to the Makefile for other systems\nXthan the one I'm running (NetBSD 1.3), but anyway: to install on a BSD\nXsystem: fix the path names in the Makefile if you need to, then make,\nXmake install, slurp the SQL files into psql or whatever, and you're\nXoff. Enjoy!\nX\nXBergen, Norway, 1998-08-09, Tom Ivar Helbekkmo ([email protected]).\nEND-of-README\necho x - inet_net_ntop.c\nsed 's/^X//' >inet_net_ntop.c << 'END-of-inet_net_ntop.c'\nX/*\nX * Copyright (c) 1996 by Internet Software Consortium.\nX *\nX * Permission to use, copy, modify, and distribute this software for any\nX * purpose with or without fee is hereby granted, provided that the above\nX * copyright notice and this permission notice appear in all copies.\nX *\nX * THE SOFTWARE IS PROVIDED \"AS IS\" AND INTERNET SOFTWARE CONSORTIUM DISCLAIMS\nX * ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES\nX * OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL INTERNET SOFTWARE\nX * CONSORTIUM BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL\nX * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR\nX * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS\nX * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS\nX * SOFTWARE.\nX */\nX\nX#if defined(LIBC_SCCS) && !defined(lint)\nXstatic const char rcsid[] = \"$Id: inet_net_ntop.c,v 8.2 1996/08/08 06:54:44 vixie Exp $\";\nX#endif\nX\nX#include <sys/types.h>\nX#include <sys/socket.h>\nX#include <netinet/in.h>\nX#include <arpa/inet.h>\nX\nX#include <errno.h>\nX#include <stdio.h>\nX#include <string.h>\nX#include <stdlib.h>\nX\nX#ifdef SPRINTF_CHAR\nX# define SPRINTF(x) strlen(sprintf/**/x)\nX#else\nX# define SPRINTF(x) ((size_t)sprintf x)\nX#endif\nX\nXstatic char *\tinet_net_ntop_ipv4 __P((const u_char *src, int bits,\nX\t\t\t\t\tchar *dst, size_t size));\nX\nX/*\nX * char *\nX * inet_net_ntop(af, src, bits, dst, size)\nX *\tconvert network number from network to presentation format.\nX *\tgenerates CIDR style result always.\nX * return:\nX *\tpointer to dst, or NULL if an error occurred (check errno).\nX * author:\nX *\tPaul Vixie (ISC), July 1996\nX */\nXchar *\nXinet_net_ntop(af, src, bits, dst, size)\nX\tint af;\nX\tconst void *src;\nX\tint bits;\nX\tchar *dst;\nX\tsize_t size;\nX{\nX\tswitch (af) {\nX\tcase AF_INET:\nX\t\treturn (inet_net_ntop_ipv4(src, bits, dst, size));\nX\tdefault:\nX\t\terrno = EAFNOSUPPORT;\nX\t\treturn (NULL);\nX\t}\nX}\nX\nX/*\nX * static char *\nX * inet_net_ntop_ipv4(src, bits, dst, size)\nX *\tconvert IPv4 network number from network to presentation format.\nX *\tgenerates CIDR style result always.\nX * return:\nX *\tpointer to dst, or NULL if an error occurred (check errno).\nX * note:\nX *\tnetwork byte order assumed. this means 192.5.5.240/28 has\nX *\t0x11110000 in its fourth octet.\nX * author:\nX *\tPaul Vixie (ISC), July 1996\nX */\nXstatic char *\nXinet_net_ntop_ipv4(src, bits, dst, size)\nX\tconst u_char *src;\nX\tint bits;\nX\tchar *dst;\nX\tsize_t size;\nX{\nX\tchar *odst = dst;\nX\tchar *t;\nX\tu_int m;\nX\tint b;\nX\nX\tif (bits < 0 || bits > 32) {\nX\t\terrno = EINVAL;\nX\t\treturn (NULL);\nX\t}\nX\tif (bits == 0) {\nX\t\tif (size < sizeof \"0\")\nX\t\t\tgoto emsgsize;\nX\t\t*dst++ = '0';\nX\t\t*dst = '\\0';\nX\t}\nX\nX\t/* Format whole octets. */\nX\tfor (b = bits / 8; b > 0; b--) {\nX\t\tif (size < sizeof \"255.\")\nX\t\t\tgoto emsgsize;\nX\t\tt = dst;\nX\t\tdst += SPRINTF((dst, \"%u\", *src++));\nX\t\tif (b > 1) {\nX\t\t\t*dst++ = '.';\nX\t\t\t*dst = '\\0';\nX\t\t}\nX\t\tsize -= (size_t)(dst - t);\nX\t}\nX\nX\t/* Format partial octet. */\nX\tb = bits % 8;\nX\tif (b > 0) {\nX\t\tif (size < sizeof \".255\")\nX\t\t\tgoto emsgsize;\nX\t\tt = dst;\nX\t\tif (dst != odst)\nX\t\t\t*dst++ = '.';\nX\t\tm = ((1 << b) - 1) << (8 - b);\nX\t\tdst += SPRINTF((dst, \"%u\", *src & m));\nX\t\tsize -= (size_t)(dst - t);\nX\t}\nX\nX\t/* Format CIDR /width. */\nX\tif (size < sizeof \"/32\")\nX\t\tgoto emsgsize;\nX\tdst += SPRINTF((dst, \"/%u\", bits));\nX\treturn (odst);\nX\nX emsgsize:\nX\terrno = EMSGSIZE;\nX\treturn (NULL);\nX}\nEND-of-inet_net_ntop.c\necho x - inet_net_pton.c\nsed 's/^X//' >inet_net_pton.c << 'END-of-inet_net_pton.c'\nX/*\nX * Copyright (c) 1996 by Internet Software Consortium.\nX *\nX * Permission to use, copy, modify, and distribute this software for any\nX * purpose with or without fee is hereby granted, provided that the above\nX * copyright notice and this permission notice appear in all copies.\nX *\nX * THE SOFTWARE IS PROVIDED \"AS IS\" AND INTERNET SOFTWARE CONSORTIUM DISCLAIMS\nX * ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES\nX * OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL INTERNET SOFTWARE\nX * CONSORTIUM BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL\nX * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR\nX * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS\nX * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS\nX * SOFTWARE.\nX */\nX\nX#if defined(LIBC_SCCS) && !defined(lint)\nXstatic const char rcsid[] = \"$Id: inet_net_pton.c,v 8.3 1996/11/11 06:36:52 vixie Exp $\";\nX#endif\nX\nX#include <sys/types.h>\nX#include <sys/socket.h>\nX#include <netinet/in.h>\nX#include <arpa/inet.h>\nX\nX#include <assert.h>\nX#include <ctype.h>\nX#include <errno.h>\nX#include <stdio.h>\nX#include <string.h>\nX#include <stdlib.h>\nX\nX#ifdef SPRINTF_CHAR\nX# define SPRINTF(x) strlen(sprintf/**/x)\nX#else\nX# define SPRINTF(x) ((size_t)sprintf x)\nX#endif\nX\nXstatic int\tinet_net_pton_ipv4 __P((const char *src, u_char *dst,\nX\t\t\t\t\tsize_t size));\nX\nX/*\nX * static int\nX * inet_net_pton(af, src, dst, size)\nX *\tconvert network number from presentation to network format.\nX *\taccepts hex octets, hex strings, decimal octets, and /CIDR.\nX *\t\"size\" is in bytes and describes \"dst\".\nX * return:\nX *\tnumber of bits, either imputed classfully or specified with /CIDR,\nX *\tor -1 if some failure occurred (check errno). ENOENT means it was\nX *\tnot a valid network specification.\nX * author:\nX *\tPaul Vixie (ISC), June 1996\nX */\nXint\nXinet_net_pton(af, src, dst, size)\nX\tint af;\nX\tconst char *src;\nX\tvoid *dst;\nX\tsize_t size;\nX{\nX\tswitch (af) {\nX\tcase AF_INET:\nX\t\treturn (inet_net_pton_ipv4(src, dst, size));\nX\tdefault:\nX\t\terrno = EAFNOSUPPORT;\nX\t\treturn (-1);\nX\t}\nX}\nX\nX/*\nX * static int\nX * inet_net_pton_ipv4(src, dst, size)\nX *\tconvert IPv4 network number from presentation to network format.\nX *\taccepts hex octets, hex strings, decimal octets, and /CIDR.\nX *\t\"size\" is in bytes and describes \"dst\".\nX * return:\nX *\tnumber of bits, either imputed classfully or specified with /CIDR,\nX *\tor -1 if some failure occurred (check errno). ENOENT means it was\nX *\tnot an IPv4 network specification.\nX * note:\nX *\tnetwork byte order assumed. this means 192.5.5.240/28 has\nX *\t0x11110000 in its fourth octet.\nX * author:\nX *\tPaul Vixie (ISC), June 1996\nX */\nXstatic int\nXinet_net_pton_ipv4(src, dst, size)\nX\tconst char *src;\nX\tu_char *dst;\nX\tsize_t size;\nX{\nX\tstatic const char\nX\t\txdigits[] = \"0123456789abcdef\",\nX\t\tdigits[] = \"0123456789\";\nX\tint n, ch, tmp, dirty, bits;\nX\tconst u_char *odst = dst;\nX\nX\tch = *src++;\nX\tif (ch == '0' && (src[0] == 'x' || src[0] == 'X')\nX\t && isascii(src[1]) && isxdigit(src[1])) {\nX\t\t/* Hexadecimal: Eat nybble string. */\nX\t\tif (size <= 0)\nX\t\t\tgoto emsgsize;\nX\t\t*dst = 0, dirty = 0;\nX\t\tsrc++;\t/* skip x or X. */\nX\t\twhile ((ch = *src++) != '\\0' &&\nX\t\t isascii(ch) && isxdigit(ch)) {\nX\t\t\tif (isupper(ch))\nX\t\t\t\tch = tolower(ch);\nX\t\t\tn = strchr(xdigits, ch) - xdigits;\nX\t\t\tassert(n >= 0 && n <= 15);\nX\t\t\t*dst |= n;\nX\t\t\tif (!dirty++)\nX\t\t\t\t*dst <<= 4;\nX\t\t\telse if (size-- > 0)\nX\t\t\t\t*++dst = 0, dirty = 0;\nX\t\t\telse\nX\t\t\t\tgoto emsgsize;\nX\t\t}\nX\t\tif (dirty)\nX\t\t\tsize--;\nX\t} else if (isascii(ch) && isdigit(ch)) {\nX\t\t/* Decimal: eat dotted digit string. */\nX\t\tfor (;;) {\nX\t\t\ttmp = 0;\nX\t\t\tdo {\nX\t\t\t\tn = strchr(digits, ch) - digits;\nX\t\t\t\tassert(n >= 0 && n <= 9);\nX\t\t\t\ttmp *= 10;\nX\t\t\t\ttmp += n;\nX\t\t\t\tif (tmp > 255)\nX\t\t\t\t\tgoto enoent;\nX\t\t\t} while ((ch = *src++) != '\\0' &&\nX\t\t\t\t isascii(ch) && isdigit(ch));\nX\t\t\tif (size-- <= 0)\nX\t\t\t\tgoto emsgsize;\nX\t\t\t*dst++ = (u_char) tmp;\nX\t\t\tif (ch == '\\0' || ch == '/')\nX\t\t\t\tbreak;\nX\t\t\tif (ch != '.')\nX\t\t\t\tgoto enoent;\nX\t\t\tch = *src++;\nX\t\t\tif (!isascii(ch) || !isdigit(ch))\nX\t\t\t\tgoto enoent;\nX\t\t}\nX\t} else\nX\t\tgoto enoent;\nX\nX\tbits = -1;\nX\tif (ch == '/' && isascii(src[0]) && isdigit(src[0]) && dst > odst) {\nX\t\t/* CIDR width specifier. Nothing can follow it. */\nX\t\tch = *src++;\t/* Skip over the /. */\nX\t\tbits = 0;\nX\t\tdo {\nX\t\t\tn = strchr(digits, ch) - digits;\nX\t\t\tassert(n >= 0 && n <= 9);\nX\t\t\tbits *= 10;\nX\t\t\tbits += n;\nX\t\t} while ((ch = *src++) != '\\0' &&\nX\t\t\t isascii(ch) && isdigit(ch));\nX\t\tif (ch != '\\0')\nX\t\t\tgoto enoent;\nX\t\tif (bits > 32)\nX\t\t\tgoto emsgsize;\nX\t}\nX\nX\t/* Firey death and destruction unless we prefetched EOS. */\nX\tif (ch != '\\0')\nX\t\tgoto enoent;\nX\nX\t/* If nothing was written to the destination, we found no address. */\nX\tif (dst == odst)\nX\t\tgoto enoent;\nX\t/* If no CIDR spec was given, infer width from net class. */\nX\tif (bits == -1) {\nX\t\tif (*odst >= 240)\t/* Class E */\nX\t\t\tbits = 32;\nX\t\telse if (*odst >= 224)\t/* Class D */\nX\t\t\tbits = 4;\nX\t\telse if (*odst >= 192)\t/* Class C */\nX\t\t\tbits = 24;\nX\t\telse if (*odst >= 128)\t/* Class B */\nX\t\t\tbits = 16;\nX\t\telse\t\t\t/* Class A */\nX\t\t\tbits = 8;\nX\t\t/* If imputed mask is narrower than specified octets, widen. */\nX\t\tif (bits >= 8 && bits < ((dst - odst) * 8))\nX\t\t\tbits = (dst - odst) * 8;\nX\t}\nX\t/* Extend network to cover the actual mask. */\nX\twhile (bits > ((dst - odst) * 8)) {\nX\t\tif (size-- <= 0)\nX\t\t\tgoto emsgsize;\nX\t\t*dst++ = '\\0';\nX\t}\nX\treturn (bits);\nX\nX enoent:\nX\terrno = ENOENT;\nX\treturn (-1);\nX\nX emsgsize:\nX\terrno = EMSGSIZE;\nX\treturn (-1);\nX}\nEND-of-inet_net_pton.c\necho x - ip.c\nsed 's/^X//' >ip.c << 'END-of-ip.c'\nX/*\nX *\tPostgreSQL type definitions for IP addresses. This\nX *\tis for IP V4 CIDR notation, but prepared for V6: just\nX *\tadd the necessary bits where the comments indicate.\nX *\nX *\t$Id: ip.c,v 1.2 1998/09/08 12:10:36 tih Exp $\nX */\nX\nX#include <sys/types.h>\nX#include <sys/socket.h>\nX\nX#include <stdio.h>\nX#include <errno.h>\nX\nX#include <netinet/in.h>\nX#include <arpa/inet.h>\nX\nX#include <postgres.h>\nX#include <utils/palloc.h>\nX\nX/*\nX *\tThis is the internal storage format for IP addresses:\nX */\nX\nXtypedef struct {\nX unsigned char family;\nX unsigned char bits;\nX union {\nX u_int32_t ipv4_addr;\t/* network byte order */\nX\t\t\t\t/* add IPV6 address type here */\nX } addr;\nX} ipaddr_struct;\nX\nXtypedef struct varlena ipaddr;\nX\nX/*\nX *\tAccess macros. Add IPV6 support.\nX */\nX\nX#define ip_addrsize(ipaddrptr) \\\nX\t(((ipaddr_struct *)VARDATA(ipaddrptr))->family == AF_INET ? 4 : -1)\nX\nX#define ip_family(ipaddrptr) \\\nX\t(((ipaddr_struct *)VARDATA(ipaddrptr))->family)\nX\nX#define ip_bits(ipaddrptr) \\\nX\t(((ipaddr_struct *)VARDATA(ipaddrptr))->bits)\nX\nX#define ip_v4addr(ipaddrptr) \\\nX\t(((ipaddr_struct *)VARDATA(ipaddrptr))->addr.ipv4_addr)\nX\nX/*\nX *\tVarious forward declarations:\nX */\nX\nXipaddr *ipaddr_in(char *str);\nXchar *ipaddr_out(ipaddr *addr);\nX\nXbool ipaddr_lt(ipaddr *a1, ipaddr *a2);\nXbool ipaddr_le(ipaddr *a1, ipaddr *a2);\nXbool ipaddr_eq(ipaddr *a1, ipaddr *a2);\nXbool ipaddr_ge(ipaddr *a1, ipaddr *a2);\nXbool ipaddr_gt(ipaddr *a1, ipaddr *a2);\nX\nXbool ipaddr_ne(ipaddr *a1, ipaddr *a2);\nX\nXbool ipaddr_sub(ipaddr *a1, ipaddr *a2);\nXbool ipaddr_subeq(ipaddr *a1, ipaddr *a2);\nXbool ipaddr_sup(ipaddr *a1, ipaddr *a2);\nXbool ipaddr_supeq(ipaddr *a1, ipaddr *a2);\nX\nXint4 ipaddr_cmp(ipaddr *a1, ipaddr *a2);\nX\nXint v4bitncmp(u_int32_t a1, u_int32_t a2, int bits);\nX\nX/*\nX *\tIP address reader.\nX */\nX\nXipaddr *ipaddr_in(char *src) {\nX int bits;\nX ipaddr *dst;\nX\nX dst = palloc(VARHDRSZ + sizeof(ipaddr_struct));\nX if (dst == NULL) {\nX elog(ERROR, \"unable to allocate memory in ipaddr_in()\");\nX return(NULL);\nX }\nX /* First, try for an IP V4 address: */\nX ip_family(dst) = AF_INET;\nX bits = inet_net_pton(ip_family(dst), src, &ip_v4addr(dst), ip_addrsize(dst));\nX if ((bits < 0) || (bits > 32)) {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"could not parse \\\"%s\\\"\", src);\nX pfree(dst);\nX return(NULL);\nX }\nX VARSIZE(dst) = VARHDRSZ\nX + ((char *)&ip_v4addr(dst) - (char *)VARDATA(dst))\nX + ip_addrsize(dst);\nX ip_bits(dst) = bits;\nX return(dst);\nX}\nX\nX/*\nX *\tIP address output function.\nX */\nX\nXchar *ipaddr_out(ipaddr *src) {\nX char *dst, tmp[sizeof(\"255.255.255.255/32\")];\nX\nX if (ip_family(src) == AF_INET) {\nX /* It's an IP V4 address: */\nX if (inet_net_ntop(AF_INET, &ip_v4addr(src), ip_bits(src),\nX\t\t tmp, sizeof(tmp)) < 0) {\nX elog(ERROR, \"unable to print address (%s)\", strerror(errno));\nX return(NULL);\nX }\nX } else {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"unknown address family (%d)\", ip_family(src));\nX return(NULL);\nX }\nX dst = palloc(strlen(tmp) + 1);\nX if (dst == NULL) {\nX elog(ERROR, \"unable to allocate memory in ipaddr_out()\");\nX return(NULL);\nX }\nX strcpy(dst, tmp);\nX return(dst);\nX}\nX\nX/*\nX *\tBoolean tests for magnitude. Add V4/V6 testing!\nX */\nX\nXbool ipaddr_lt(ipaddr *a1, ipaddr *a2) {\nX if ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET)) {\nX int order = v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a2));\nX return((order < 0) || ((order == 0) && (ip_bits(a1) < ip_bits(a2))));\nX } else {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"cannot compare address families %d and %d\",\nX\t ip_family(a1), ip_family(a2));\nX return(FALSE);\nX }\nX}\nX\nXbool ipaddr_le(ipaddr *a1, ipaddr *a2) {\nX return(ipaddr_lt(a1, a2) || ipaddr_eq(a1, a2));\nX}\nX\nXbool ipaddr_eq(ipaddr *a1, ipaddr *a2) {\nX if ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET)) {\nX return((ip_bits(a1) == ip_bits(a2))\nX\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a1)) == 0));\nX } else {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"cannot compare address families %d and %d\",\nX\t ip_family(a1), ip_family(a2));\nX return(FALSE);\nX }\nX}\nX\nXbool ipaddr_ge(ipaddr *a1, ipaddr *a2) {\nX return(ipaddr_gt(a1, a2) || ipaddr_eq(a1, a2));\nX}\nX\nXbool ipaddr_gt(ipaddr *a1, ipaddr *a2) {\nX if ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET)) {\nX int order = v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a2));\nX return((order > 0) || ((order == 0) && (ip_bits(a1) > ip_bits(a2))));\nX } else {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"cannot compare address families %d and %d\",\nX\t ip_family(a1), ip_family(a2));\nX return(FALSE);\nX }\nX}\nX\nXbool ipaddr_ne(ipaddr *a1, ipaddr *a2) {\nX return(!ipaddr_eq(a1, a2));\nX}\nX\nXbool ipaddr_sub(ipaddr *a1, ipaddr *a2) {\nX if ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET)) {\nX return((ip_bits(a1) > ip_bits(a2))\nX\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a2)) == 0));\nX } else {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"cannot compare address families %d and %d\",\nX\t ip_family(a1), ip_family(a2));\nX return(FALSE);\nX }\nX}\nX\nXbool ipaddr_subeq(ipaddr *a1, ipaddr *a2) {\nX if ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET)) {\nX return((ip_bits(a1) >= ip_bits(a2))\nX\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a2)) == 0));\nX } else {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"cannot compare address families %d and %d\",\nX\t ip_family(a1), ip_family(a2));\nX return(FALSE);\nX }\nX}\nX\nXbool ipaddr_sup(ipaddr *a1, ipaddr *a2) {\nX if ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET)) {\nX return((ip_bits(a1) < ip_bits(a2))\nX\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a1)) == 0));\nX } else {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"cannot compare address families %d and %d\",\nX\t ip_family(a1), ip_family(a2));\nX return(FALSE);\nX }\nX}\nX\nXbool ipaddr_supeq(ipaddr *a1, ipaddr *a2) {\nX if ((ip_family(a1) == AF_INET) && (ip_family(a2) == AF_INET)) {\nX return((ip_bits(a1) <= ip_bits(a2))\nX\t && (v4bitncmp(ip_v4addr(a1), ip_v4addr(a2), ip_bits(a1)) == 0));\nX } else {\nX /* Go for an IPV6 address here, before faulting out: */\nX elog(ERROR, \"cannot compare address families %d and %d\",\nX\t ip_family(a1), ip_family(a2));\nX return(FALSE);\nX }\nX}\nX\nX/*\nX *\tComparison function for sorting. Add V4/V6 testing!\nX */\nX\nXint4 ipaddr_cmp(ipaddr *a1, ipaddr *a2) {\nX if (ntohl(ip_v4addr(a1)) < ntohl(ip_v4addr(a2))) {\nX return(-1);\nX } else if (ntohl(ip_v4addr(a1)) > ntohl(ip_v4addr(a2))) {\nX return(1);\nX }\nX return 0;\nX}\nX\nX/*\nX *\tBitwise comparison for V4 addresses. Add V6 implementation!\nX */\nX\nXint v4bitncmp(u_int32_t a1, u_int32_t a2, int bits) {\nX unsigned long mask = 0;\nX int i;\nX for (i = 0; i < bits; i++) {\nX mask = (mask >> 1) | 0x80000000;\nX }\nX a1 = ntohl(a1);\nX a2 = ntohl(a2);\nX if ((a1 & mask) < (a2 & mask)) {\nX return(-1);\nX } else if ((a1 & mask) > (a2 & mask)) {\nX return(1);\nX }\nX return(0);\nX}\nX\nX/*\nX *\teof\nX */\nEND-of-ip.c\necho x - ip.sql.in\nsed 's/^X//' >ip.sql.in << 'END-of-ip.sql.in'\nX--\nX--\tPostgreSQL code for IP addresses.\nX--\nX--\t$Id: ip.sql.in,v 1.2 1998/09/08 12:10:45 tih Exp $\nX--\nX\nXload 'TARGET/ip.so';\nX\nX--\nX--\tInput and output functions and the type itself:\nX--\nX\nXcreate function ipaddr_in(opaque)\nX\treturns opaque\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_out(opaque)\nX\treturns opaque\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate type ipaddr (\nX\tinternallength = variable,\nX\texternallength = variable,\nX\tinput = ipaddr_in,\nX\toutput = ipaddr_out\nX);\nX\nX--\nX--\tThe various boolean tests:\nX--\nX\nXcreate function ipaddr_lt(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_le(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_eq(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_ge(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_gt(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_ne(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_sub(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_subeq(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_sup(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_supeq(ipaddr, ipaddr)\nX\treturns bool\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nXcreate function ipaddr_cmp(ipaddr, ipaddr)\nX\treturns int4\nX\tas 'TARGET/ip.so'\nX\tlanguage 'c';\nX\nX--\nX--\tNow the operators. Note how some of the parameters to some\nX--\tof the 'create operator' commands are commented out. This\nX--\tis because they reference as yet undefined operators, and\nX--\twill be implicitly defined when those are, further down.\nX--\nX\nXcreate operator < (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX--\tnegator = >=,\nX\trestrict = intltsel,\nX\tjoin = intltjoinsel,\nX\tprocedure = ipaddr_lt\nX);\nX\nXcreate operator <= (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX--\tnegator = >,\nX\trestrict = intltsel,\nX\tjoin = intltjoinsel,\nX\tprocedure = ipaddr_le\nX);\nX\nXcreate operator = (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX\tcommutator = =,\nX--\tnegator = <>,\nX\trestrict = eqsel,\nX\tjoin = eqjoinsel,\nX\tprocedure = ipaddr_eq\nX);\nX\nXcreate operator >= (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX\tnegator = <,\nX\trestrict = intgtsel,\nX\tjoin = intgtjoinsel,\nX\tprocedure = ipaddr_ge\nX);\nX\nXcreate operator > (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX\tnegator = <=,\nX\trestrict = intgtsel,\nX\tjoin = intgtjoinsel,\nX\tprocedure = ipaddr_gt\nX);\nX\nXcreate operator <> (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX\tnegator = =,\nX\trestrict = neqsel,\nX\tjoin = neqjoinsel,\nX\tprocedure = ipaddr_ne\nX);\nX\nXcreate operator << (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX\tprocedure = ipaddr_sub\nX);\nX\nXcreate operator <<= (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX\tprocedure = ipaddr_subeq\nX);\nX\nXcreate operator >> (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX\tprocedure = ipaddr_sup\nX);\nX\nXcreate operator >>= (\nX\tleftarg = ipaddr,\nX\trightarg = ipaddr,\nX\tprocedure = ipaddr_supeq\nX);\nX\nX--\nX--\tFinally, make it possible to do btree indexing on IP addresses.\nX--\nX\nXbegin;\nX\nXcreate table tmp_op (oprname name, opi int2);\nX\nXinsert into tmp_op values ('<', 1);\nXinsert into tmp_op values ('<=', 2);\nXinsert into tmp_op values ('=', 3);\nXinsert into tmp_op values ('>=', 4);\nXinsert into tmp_op values ('>', 5);\nX\nXdelete from pg_opclass\nX\twhere opcname = 'ipaddr_ops';\nX\nXinsert into pg_opclass (opcname, opcdeftype)\nX\tselect 'ipaddr_ops', oid from pg_type\nX\t\twhere typname = 'ipaddr';\nX\nXselect o.oid as opoid, o.oprname\nX\tinto table ipaddr_tmp\nX\tfrom pg_operator o, pg_type t\nX\twhere o.oprleft = t.oid and\nX\t o.oprright = t.oid and\nX\t t.typname = 'ipaddr';\nX\nXinsert into pg_amop (amopid, amopclaid, amopopr, amopstrategy,\nX\t\t amopselect, amopnpages)\nX\tselect am.oid, opcl.oid, c.opoid, t.opi,\nX\t\t\t'btreesel'::regproc, 'btreenpage'::regproc\nX\t\tfrom pg_am am, pg_opclass opcl, ipaddr_tmp c, tmp_op t\nX\t\twhere t.oprname = c.oprname and\nX\t\t amname = 'btree' and opcname = 'ipaddr_ops';\nX\nXinsert into pg_amproc (amid, amopclaid, amproc, amprocnum)\nX\tselect pgam.oid, opc.oid, prc.oid, '1'::int2\nX\t\tfrom pg_am pgam, pg_opclass opc, pg_proc prc\nX\t\twhere prc.proname = 'ipaddr_cmp' and\nX\t\t pgam.amname = 'btree' and\nX\t\t opc.opcname = 'ipaddr_ops';\nX\nXdrop table tmp_op;\nXdrop table ipaddr_tmp;\nX\nXcommit;\nX\nX--\nX--\teof\nX--\nEND-of-ip.sql.in\necho x - mac.c\nsed 's/^X//' >mac.c << 'END-of-mac.c'\nX/*\nX *\tPostgreSQL type definitions for MAC addresses.\nX *\nX *\t$Id: mac.c,v 1.1 1998/09/07 12:31:18 tih Exp $\nX */\nX\nX#include <stdio.h>\nX\nX#include <postgres.h>\nX#include <utils/palloc.h>\nX\nX#include \"mac.h\"\nX\nX/*\nX *\tThis is the internal storage format for MAC addresses:\nX */\nX\nXtypedef struct macaddr {\nX unsigned char a;\nX unsigned char b;\nX unsigned char c;\nX unsigned char d;\nX unsigned char e;\nX unsigned char f;\nX} macaddr;\nX\nX/*\nX *\tVarious forward declarations:\nX */\nX\nXmacaddr *macaddr_in(char *str);\nXchar *macaddr_out(macaddr *addr);\nX\nXbool macaddr_lt(macaddr *a1, macaddr *a2);\nXbool macaddr_le(macaddr *a1, macaddr *a2);\nXbool macaddr_eq(macaddr *a1, macaddr *a2);\nXbool macaddr_ge(macaddr *a1, macaddr *a2);\nXbool macaddr_gt(macaddr *a1, macaddr *a2);\nX\nXbool macaddr_ne(macaddr *a1, macaddr *a2);\nX\nXint4 macaddr_cmp(macaddr *a1, macaddr *a2);\nX\nXtext *macaddr_manuf(macaddr *addr);\nX\nX/*\nX *\tUtility macros used for sorting and comparing:\nX */\nX\nX#define hibits(addr) \\\nX ((unsigned long)((addr->a<<16)|(addr->b<<8)|(addr->c)))\nX\nX#define lobits(addr) \\\nX ((unsigned long)((addr->c<<16)|(addr->e<<8)|(addr->f)))\nX\nX/*\nX *\tMAC address reader. Accepts several common notations.\nX */\nX\nXmacaddr *macaddr_in(char *str) {\nX int a, b, c, d, e, f;\nX macaddr *result;\nX int count;\nX\nX if (strlen(str) > 0) {\nX\nX count = sscanf(str, \"%x:%x:%x:%x:%x:%x\", &a, &b, &c, &d, &e, &f);\nX if (count != 6)\nX count = sscanf(str, \"%x-%x-%x-%x-%x-%x\", &a, &b, &c, &d, &e, &f);\nX if (count != 6)\nX count = sscanf(str, \"%2x%2x%2x:%2x%2x%2x\", &a, &b, &c, &d, &e, &f);\nX if (count != 6)\nX count = sscanf(str, \"%2x%2x%2x-%2x%2x%2x\", &a, &b, &c, &d, &e, &f);\nX if (count != 6)\nX count = sscanf(str, \"%2x%2x.%2x%2x.%2x%2x\", &a, &b, &c, &d, &e, &f);\nX \nX if (count != 6) {\nX elog(ERROR, \"macaddr_in: error in parsing \\\"%s\\\"\", str);\nX return(NULL);\nX }\nX \nX if ((a < 0) || (a > 255) || (b < 0) || (b > 255) ||\nX\t(c < 0) || (c > 255) || (d < 0) || (d > 255) ||\nX\t(e < 0) || (e > 255) || (f < 0) || (f > 255)) {\nX elog(ERROR, \"macaddr_in: illegal address \\\"%s\\\"\", str);\nX return(NULL);\nX }\nX } else {\nX a = b = c = d = e = f = 0;\t/* special case for missing address */\nX }\nX\nX result = (macaddr *)palloc(sizeof(macaddr));\nX\nX result->a = a;\nX result->b = b;\nX result->c = c;\nX result->d = d;\nX result->e = e;\nX result->f = f;\nX\nX return(result);\nX}\nX\nX/*\nX *\tMAC address output function. Fixed format.\nX */\nX\nXchar *macaddr_out(macaddr *addr) {\nX char *result;\nX\nX if (addr == NULL)\nX return(NULL);\nX\nX result = (char *)palloc(32);\nX\nX if ((hibits(addr) > 0) || (lobits(addr) > 0)) {\nX sprintf(result, \"%02x:%02x:%02x:%02x:%02x:%02x\",\nX\t addr->a, addr->b, addr->c, addr->d, addr->e, addr->f);\nX } else {\nX result[0] = 0;\t\t/* special case for missing address */\nX }\nX return(result);\nX}\nX\nX/*\nX *\tBoolean tests.\nX */\nX\nXbool macaddr_lt(macaddr *a1, macaddr *a2) {\nX return((hibits(a1) < hibits(a2)) ||\nX\t ((hibits(a1) == hibits(a2)) && lobits(a1) < lobits(a2)));\nX};\nX\nXbool macaddr_le(macaddr *a1, macaddr *a2) {\nX return((hibits(a1) < hibits(a2)) ||\nX\t ((hibits(a1) == hibits(a2)) && lobits(a1) <= lobits(a2)));\nX};\nX\nXbool macaddr_eq(macaddr *a1, macaddr *a2) {\nX return ((hibits(a1) == hibits(a2)) && (lobits(a1) == lobits(a2)));\nX};\nX\nXbool macaddr_ge(macaddr *a1, macaddr *a2) {\nX return((hibits(a1) > hibits(a2)) ||\nX\t ((hibits(a1) == hibits(a2)) && lobits(a1) >= lobits(a2)));\nX};\nX\nXbool macaddr_gt(macaddr *a1, macaddr *a2) {\nX return((hibits(a1) > hibits(a2)) ||\nX\t ((hibits(a1) == hibits(a2)) && lobits(a1) > lobits(a2)));\nX};\nX\nXbool macaddr_ne(macaddr *a1, macaddr *a2) {\nX return ((hibits(a1) != hibits(a2)) || (lobits(a1) != lobits(a2)));\nX};\nX\nX/*\nX *\tComparison function for sorting:\nX */\nX\nXint4 macaddr_cmp(macaddr *a1, macaddr *a2) {\nX if (hibits(a1) < hibits(a2))\nX return -1;\nX else if (hibits(a1) > hibits(a2))\nX return 1;\nX else if (lobits(a1) < lobits(a2))\nX return -1;\nX else if (lobits(a1) > lobits(a2))\nX return 1;\nX else\nX return 0;\nX}\nX\nX/*\nX *\tThe special manufacturer fetching function. See \"mac.h\".\nX */\nX\nXtext *macaddr_manuf(macaddr *addr) {\nX manufacturer *manuf;\nX int length;\nX text *result;\nX\nX for (manuf = manufacturers; manuf->name != NULL; manuf++) {\nX if ((manuf->a == addr->a) &&\nX\t(manuf->b == addr->b) &&\nX\t(manuf->c == addr->c))\nX break;\nX }\nX if (manuf->name == NULL) {\nX result = palloc(VARHDRSZ + 1);\nX memset(result, 0, VARHDRSZ + 1);\nX VARSIZE(result) = VARHDRSZ + 1;\nX } else {\nX length = strlen(manuf->name) + 1;\nX result = palloc(length + VARHDRSZ);\nX memset(result, 0, length + VARHDRSZ);\nX VARSIZE(result) = length + VARHDRSZ;\nX memcpy(VARDATA(result), manuf->name, length);\nX }\nX return result;\nX}\nX\nX/*\nX *\teof\nX */\nEND-of-mac.c\necho x - mac.h\nsed 's/^X//' >mac.h << 'END-of-mac.h'\nX/*\nX *\tPostgreSQL type definitions for MAC addresses.\nX *\nX *\t$Id: mac.h,v 1.1 1998/09/07 12:31:22 tih Exp $\nX */\nX\nXtypedef struct manufacturer {\nX unsigned char a;\nX unsigned char b;\nX unsigned char c;\nX char *name;\nX} manufacturer;\nX\nXmanufacturer manufacturers[] = {\nX {0x00, 0x00, 0x0C, \"Cisco\"},\nX {0x00, 0x00, 0x0E, \"Fujitsu\"},\nX {0x00, 0x00, 0x0F, \"NeXT\"},\nX {0x00, 0x00, 0x10, \"Sytek\"},\nX {0x00, 0x00, 0x1D, \"Cabletron\"},\nX {0x00, 0x00, 0x20, \"DIAB\"},\nX {0x00, 0x00, 0x22, \"Visual Technology\"},\nX {0x00, 0x00, 0x2A, \"TRW\"},\nX {0x00, 0x00, 0x32, \"GPT Limited\"},\nX {0x00, 0x00, 0x5A, \"S & Koch\"},\nX {0x00, 0x00, 0x5E, \"IANA\"},\nX {0x00, 0x00, 0x65, \"Network General\"},\nX {0x00, 0x00, 0x6B, \"MIPS\"},\nX {0x00, 0x00, 0x77, \"MIPS\"},\nX {0x00, 0x00, 0x7A, \"Ardent\"},\nX {0x00, 0x00, 0x89, \"Cayman Systems\"},\nX {0x00, 0x00, 0x93, \"Proteon\"},\nX {0x00, 0x00, 0x9F, \"Ameristar Technology\"},\nX {0x00, 0x00, 0xA2, \"Wellfleet\"},\nX {0x00, 0x00, 0xA3, \"Network Application Technology\"},\nX {0x00, 0x00, 0xA6, \"Network General\"},\nX {0x00, 0x00, 0xA7, \"NCD\"},\nX {0x00, 0x00, 0xA9, \"Network Systems\"},\nX {0x00, 0x00, 0xAA, \"Xerox\"},\nX {0x00, 0x00, 0xB3, \"CIMLinc\"},\nX {0x00, 0x00, 0xB7, \"Dove Fastnet\"},\nX {0x00, 0x00, 0xBC, \"Allen-Bradley\"},\nX {0x00, 0x00, 0xC0, \"Western Digital\"},\nX {0x00, 0x00, 0xC5, \"Farallon\"},\nX {0x00, 0x00, 0xC6, \"Hewlett-Packard\"},\nX {0x00, 0x00, 0xC8, \"Altos\"},\nX {0x00, 0x00, 0xC9, \"Emulex\"},\nX {0x00, 0x00, 0xD7, \"Dartmouth College\"},\nX {0x00, 0x00, 0xD8, \"3Com (?)\"},\nX {0x00, 0x00, 0xDD, \"Gould\"},\nX {0x00, 0x00, 0xDE, \"Unigraph\"},\nX {0x00, 0x00, 0xE2, \"Acer Counterpoint\"},\nX {0x00, 0x00, 0xEF, \"Alantec\"},\nX {0x00, 0x00, 0xFD, \"High Level Hardware\"},\nX {0x00, 0x01, 0x02, \"BBN internal usage\"},\nX {0x00, 0x20, 0xAF, \"3Com\"},\nX {0x00, 0x17, 0x00, \"Kabel\"},\nX {0x00, 0x80, 0x64, \"Wyse Technology\"},\nX {0x00, 0x80, 0x2B, \"IMAC (?)\"},\nX {0x00, 0x80, 0x2D, \"Xylogics, Inc.\"},\nX {0x00, 0x80, 0x8C, \"Frontier Software Development\"},\nX {0x00, 0x80, 0xC2, \"IEEE 802.1 Committee\"},\nX {0x00, 0x80, 0xD3, \"Shiva\"},\nX {0x00, 0xAA, 0x00, \"Intel\"},\nX {0x00, 0xDD, 0x00, \"Ungermann-Bass\"},\nX {0x00, 0xDD, 0x01, \"Ungermann-Bass\"},\nX {0x02, 0x07, 0x01, \"Racal InterLan\"},\nX {0x02, 0x04, 0x06, \"BBN internal usage\"},\nX {0x02, 0x60, 0x86, \"Satelcom MegaPac\"},\nX {0x02, 0x60, 0x8C, \"3Com\"},\nX {0x02, 0xCF, 0x1F, \"CMC\"},\nX {0x08, 0x00, 0x02, \"3Com\"},\nX {0x08, 0x00, 0x03, \"ACC\"},\nX {0x08, 0x00, 0x05, \"Symbolics\"},\nX {0x08, 0x00, 0x08, \"BBN\"},\nX {0x08, 0x00, 0x09, \"Hewlett-Packard\"},\nX {0x08, 0x00, 0x0A, \"Nestar Systems\"},\nX {0x08, 0x00, 0x0B, \"Unisys\"},\nX {0x08, 0x00, 0x11, \"Tektronix\"},\nX {0x08, 0x00, 0x14, \"Excelan\"},\nX {0x08, 0x00, 0x17, \"NSC\"},\nX {0x08, 0x00, 0x1A, \"Data General\"},\nX {0x08, 0x00, 0x1B, \"Data General\"},\nX {0x08, 0x00, 0x1E, \"Apollo\"},\nX {0x08, 0x00, 0x20, \"Sun\"},\nX {0x08, 0x00, 0x22, \"NBI\"},\nX {0x08, 0x00, 0x25, \"CDC\"},\nX {0x08, 0x00, 0x26, \"Norsk Data\"},\nX {0x08, 0x00, 0x27, \"PCS Computer Systems GmbH\"},\nX {0x08, 0x00, 0x28, \"Texas Instruments\"},\nX {0x08, 0x00, 0x2B, \"DEC\"},\nX {0x08, 0x00, 0x2E, \"Metaphor\"},\nX {0x08, 0x00, 0x2F, \"Prime Computer\"},\nX {0x08, 0x00, 0x36, \"Intergraph\"},\nX {0x08, 0x00, 0x37, \"Fujitsu-Xerox\"},\nX {0x08, 0x00, 0x38, \"Bull\"},\nX {0x08, 0x00, 0x39, \"Spider Systems\"},\nX {0x08, 0x00, 0x41, \"DCA Digital Comm. Assoc.\"},\nX {0x08, 0x00, 0x45, \"Xylogics (?)\"},\nX {0x08, 0x00, 0x46, \"Sony\"},\nX {0x08, 0x00, 0x47, \"Sequent\"},\nX {0x08, 0x00, 0x49, \"Univation\"},\nX {0x08, 0x00, 0x4C, \"Encore\"},\nX {0x08, 0x00, 0x4E, \"BICC\"},\nX {0x08, 0x00, 0x56, \"Stanford University\"},\nX {0x08, 0x00, 0x58, \"DECsystem 20 (?)\"},\nX {0x08, 0x00, 0x5A, \"IBM\"},\nX {0x08, 0x00, 0x67, \"Comdesign\"},\nX {0x08, 0x00, 0x68, \"Ridge\"},\nX {0x08, 0x00, 0x69, \"Silicon Graphics\"},\nX {0x08, 0x00, 0x6E, \"Concurrent\"},\nX {0x08, 0x00, 0x75, \"DDE\"},\nX {0x08, 0x00, 0x7C, \"Vitalink\"},\nX {0x08, 0x00, 0x80, \"XIOS\"},\nX {0x08, 0x00, 0x86, \"Imagen/QMS\"},\nX {0x08, 0x00, 0x87, \"Xyplex\"},\nX {0x08, 0x00, 0x89, \"Kinetics\"},\nX {0x08, 0x00, 0x8B, \"Pyramid\"},\nX {0x08, 0x00, 0x8D, \"XyVision\"},\nX {0x08, 0x00, 0x90, \"Retix Inc\"},\nX {0x48, 0x44, 0x53, \"HDS (?)\"},\nX {0x80, 0x00, 0x10, \"AT&T\"},\nX {0xAA, 0x00, 0x00, \"DEC\"},\nX {0xAA, 0x00, 0x01, \"DEC\"},\nX {0xAA, 0x00, 0x02, \"DEC\"},\nX {0xAA, 0x00, 0x03, \"DEC\"},\nX {0xAA, 0x00, 0x04, \"DEC\"},\nX {0x00, 0x00, 0x00, NULL}\nX};\nX\nX/*\nX *\teof\nX */\nEND-of-mac.h\necho x - mac.sql.in\nsed 's/^X//' >mac.sql.in << 'END-of-mac.sql.in'\nX--\nX--\tPostgreSQL code for MAC addresses.\nX--\nX--\t$Id: mac.sql.in,v 1.2 1998/09/08 12:11:18 tih Exp $\nX--\nX\nXload 'TARGET/mac.so';\nX\nX--\nX--\tInput and output functions and the type itself:\nX--\nX\nXcreate function macaddr_in(opaque)\nX\treturns opaque\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nXcreate function macaddr_out(opaque)\nX\treturns opaque\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nXcreate type macaddr (\nX\tinternallength = 6,\nX\texternallength = variable,\nX\tinput = macaddr_in,\nX\toutput = macaddr_out\nX);\nX\nX--\nX--\tThe boolean tests:\nX--\nX\nXcreate function macaddr_lt(macaddr, macaddr)\nX\treturns bool\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nXcreate function macaddr_le(macaddr, macaddr)\nX\treturns bool\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nXcreate function macaddr_eq(macaddr, macaddr)\nX\treturns bool\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nXcreate function macaddr_ge(macaddr, macaddr)\nX\treturns bool\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nXcreate function macaddr_gt(macaddr, macaddr)\nX\treturns bool\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nXcreate function macaddr_ne(macaddr, macaddr)\nX\treturns bool\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nXcreate function macaddr_cmp(macaddr, macaddr)\nX\treturns int4\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nX--\nX--\tNow the operators. Note how some of the parameters to some\nX--\tof the 'create operator' commands are commented out. This\nX--\tis because they reference as yet undefined operators, and\nX--\twill be implicitly defined when those are, further down.\nX--\nX\nXcreate operator < (\nX\tleftarg = macaddr,\nX\trightarg = macaddr,\nX--\tnegator = >=,\nX\tprocedure = macaddr_lt\nX);\nX\nXcreate operator <= (\nX\tleftarg = macaddr,\nX\trightarg = macaddr,\nX--\tnegator = >,\nX\tprocedure = macaddr_le\nX);\nX\nXcreate operator = (\nX\tleftarg = macaddr,\nX\trightarg = macaddr,\nX\tcommutator = =,\nX--\tnegator = <>,\nX\tprocedure = macaddr_eq\nX);\nX\nXcreate operator >= (\nX\tleftarg = macaddr,\nX\trightarg = macaddr,\nX\tnegator = <,\nX\tprocedure = macaddr_ge\nX);\nX\nXcreate operator > (\nX\tleftarg = macaddr,\nX\trightarg = macaddr,\nX\tnegator = <=,\nX\tprocedure = macaddr_gt\nX);\nX\nXcreate operator <> (\nX\tleftarg = macaddr,\nX\trightarg = macaddr,\nX\tnegator = =,\nX\tprocedure = macaddr_ne\nX);\nX\nX--\nX--\tFinally, the special manufacurer matching function:\nX--\nX\nXcreate function macaddr_manuf(macaddr)\nX\treturns text\nX\tas 'TARGET/mac.so'\nX\tlanguage 'c';\nX\nX--\nX--\tFinally, make it possible to do btree indexing on MAC addresses.\nX--\nX\nXbegin;\nX\nXcreate table tmp_op (oprname name, opi int2);\nX\nXinsert into tmp_op values ('<', 1);\nXinsert into tmp_op values ('<=', 2);\nXinsert into tmp_op values ('=', 3);\nXinsert into tmp_op values ('>=', 4);\nXinsert into tmp_op values ('>', 5);\nX\nXdelete from pg_opclass\nX\twhere opcname = 'macaddr_ops';\nX\nXinsert into pg_opclass (opcname, opcdeftype)\nX\tselect 'macaddr_ops', oid from pg_type\nX\t\twhere typname = 'macaddr';\nX\nXselect o.oid as opoid, o.oprname\nX\tinto table macaddr_tmp\nX\tfrom pg_operator o, pg_type t\nX\twhere o.oprleft = t.oid and\nX\t o.oprright = t.oid and\nX\t t.typname = 'macaddr';\nX\nXinsert into pg_amop (amopid, amopclaid, amopopr, amopstrategy,\nX\t\t amopselect, amopnpages)\nX\tselect am.oid, opcl.oid, c.opoid, t.opi,\nX\t\t\t'btreesel'::regproc, 'btreenpage'::regproc\nX\t\tfrom pg_am am, pg_opclass opcl, macaddr_tmp c, tmp_op t\nX\t\twhere t.oprname = c.oprname and\nX\t\t amname = 'btree' and opcname = 'macaddr_ops';\nX\nXinsert into pg_amproc (amid, amopclaid, amproc, amprocnum)\nX\tselect pgam.oid, opc.oid, prc.oid, '1'::int2\nX\t\tfrom pg_am pgam, pg_opclass opc, pg_proc prc\nX\t\twhere prc.proname = 'macaddr_cmp' and\nX\t\t pgam.amname = 'btree' and\nX\t\t opc.opcname = 'macaddr_ops';\nX\nXdrop table tmp_op;\nXdrop table macaddr_tmp;\nX\nXcommit;\nX\nX--\nX--\teof\nX--\nEND-of-mac.sql.in\necho x - test.sql\nsed 's/^X//' >test.sql << 'END-of-test.sql'\nX--\nX--\tA quick test of the IP address code\nX--\nX--\t$Id: test.sql,v 1.2 1998/09/08 12:11:34 tih Exp $\nX--\nX\nX-- temporary table:\nXcreate table addresses (address ipaddr);\nX\nX-- sample data from two subnets:\nXinsert into addresses values ('158.37.96.15');\nXinsert into addresses values ('158.37.96.16');\nXinsert into addresses values ('158.37.96.17');\nXinsert into addresses values ('158.37.97.15');\nXinsert into addresses values ('158.37.97.16');\nXinsert into addresses values ('158.37.97.17');\nXinsert into addresses values ('158.37.98.15');\nXinsert into addresses values ('158.37.98.16');\nXinsert into addresses values ('158.37.98.17');\nXinsert into addresses values ('158.37.96.150');\nXinsert into addresses values ('158.37.96.160');\nXinsert into addresses values ('158.37.96.170');\nXinsert into addresses values ('158.37.97.150');\nXinsert into addresses values ('158.37.97.160');\nXinsert into addresses values ('158.37.97.170');\nXinsert into addresses values ('158.37.98.150');\nXinsert into addresses values ('158.37.98.160');\nXinsert into addresses values ('158.37.98.170');\nX\nX-- show them all:\nXselect * from addresses;\nX\nX-- select the ones in subnet 96:\nXselect * from addresses where address << '158.37.96.0/24';\nX\nX-- select the ones not in subnet 96:\nXselect * from addresses where not address << '158.37.96.0/24';\nX\nX-- select the ones in subnet 97:\nXselect * from addresses where address << '158.37.97.0/24';\nX\nX-- select the ones not in subnet 97:\nXselect * from addresses where not address << '158.37.97.0/24';\nX\nX-- select the ones in subnet 96 or 97, sorted:\nXselect * from addresses where address << '158.37.96.0/23'\nX\torder by address;\nX\nX-- now some networks:\nXcreate table networks (network ipaddr);\nX\nX-- now the subnets mentioned above:\nXinsert into networks values ('158.37.96.0/24');\nXinsert into networks values ('158.37.97.0/24');\nXinsert into networks values ('158.37.98.0/24');\nX\nX-- select matching pairs of addresses and containing nets:\nXselect address, network from addresses, networks\nX\twhere address << network;\nX\nX-- tidy up:\nXdrop table addresses;\nXdrop table networks;\nX\nX--\nX--\teof\nX--\nEND-of-test.sql\nexit",
"msg_date": "Fri, 2 Oct 1998 02:23:17 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: CIDR/IP types. Was: [GENERAL] big numbers (fwd)"
}
] |
[
{
"msg_contents": "Hi all,\n\nI supposed v6.4 were released on October 1, but today is already Oct 2,\nand I see only postgresql.snapshot.tar.gz.\n\nIs there any delay ?\n\nJose'\n",
"msg_date": "Fri, 02 Oct 1998 09:53:15 +0200",
"msg_from": "\"Jose' Soares\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "I can't find v6.4"
},
{
"msg_contents": "On Fri, 2 Oct 1998, Jose' Soares wrote:\n\n> Hi all,\n> \n> I supposed v6.4 were released on October 1, but today is already Oct 2,\n> and I see only postgresql.snapshot.tar.gz.\n> \n> Is there any delay ?\n\n\tYes...unfortunately, a new release date is currently being\ndiscusssed, and isn't available as of yet. Ppl appear to want to keep\nthrowing in new features *glare* even after the beta freeze...\n\n\tHope to have something in stone by Tuesday evening...this past\nweekend I've been away on a business trip :(\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Mon, 5 Oct 1998 00:58:17 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] I can't find v6.4"
}
] |
[
{
"msg_contents": "Hi,\n\n the following little patch adds array references to query\n parameters. With it applied a function like\n\n CREATE FUNCTION getname(oid8, int4) RETURNS name AS\n 'SELECT typname FROM pg_type WHERE oid = $1[$2]'\n LANGUAGE 'sql';\n\n is possible. Mainly I need this to enable array references in\n expressions for PL/pgSQL. Complete regression test ran O.K.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\nbegin 644 paramarray.diff\nM9&EF9B`M8W(@<W)C+F]R:6<O8F%C:V5N9\"]P87)S97(O9W)A;2YY('-R8R]B\nM86-K96YD+W!A<G-E<B]G<F%M+GD**BHJ('-R8RYO<FEG+V)A8VME;F0O<&%R\nM<V5R+V=R86TN>0E&<FD@3V-T(\"`R(#$R.C4X.C`V(#$Y.3@*+2TM('-R8R]B\nM86-K96YD+W!A<G-E<B]G<F%M+GD)1G)I($]C=\"`@,B`Q,CHU.3HP.2`Q.3DX\nM\"BHJ*BHJ*BHJ*BHJ*BHJ*@HJ*BH@-#4Y.\"PT-C`W(\"HJ*BH*(\"`)\"0D)?0H@\nM(`D).PH@(`HA(%!A<F%M3F\\Z(\"!005)!30H@(`D)\"0E[\"B`@\"0D)\"0DD)\"`]\nM(&UA:V5.;V1E*%!A<F%M3F\\I.PH@(`D)\"0D))\"0M/FYU;6)E<B`](\"0Q.PH@\nM(`D)\"0E]\"B`@\"0D[\"B`@\"BTM+2`T-3DX+#0V,#@@+2TM+0H@(`D)\"0E]\"B`@\nM\"0D[\"B`@\"B$@4&%R86U.;SH@(%!!4D%-(&]P=%]I;F1I<F5C=&EO;@H@(`D)\nM\"0E[\"B`@\"0D)\"0DD)\"`](&UA:V5.;V1E*%!A<F%M3F\\I.PH@(`D)\"0D))\"0M\nM/FYU;6)E<B`](\"0Q.PHK(`D)\"0D))\"0M/FEN9&ER96-T:6]N(#T@)#([\"B`@\nM\"0D)\"7T*(\"`)\"3L*(\"`*9&EF9B`M8W(@<W)C+F]R:6<O8F%C:V5N9\"]P87)S\nM97(O<&%R<V5?97AP<BYC('-R8R]B86-K96YD+W!A<G-E<B]P87)S95]E>'!R\nM+F,**BHJ('-R8RYO<FEG+V)A8VME;F0O<&%R<V5R+W!A<G-E7V5X<'(N8PE&\nM<FD@3V-T(\"`R(#$R.C4X.C`V(#$Y.3@*+2TM('-R8R]B86-K96YD+W!A<G-E\nM<B]P87)S95]E>'!R+F,)1G)I($]C=\"`@,B`Q,CHU.3HP.2`Q.3DX\"BHJ*BHJ\nM*BHJ*BHJ*BHJ*@HJ*BH@,3(R+#$R.\"`J*BHJ\"B`@\"0D)\"7!A<F%M+3YP87)A\nM;71Y<&4@/2`H3VED*2!T;VED.PH@(`D)\"0EP87)A;2T^<&%R86U?=&QI<W0@\nM/2`H3&ES=\"`J*2!.54Q,.PH@(`HA(`D)\"0ER97-U;'0@/2`H3F]D92`J*2!P\nM87)A;3L*(\"`)\"0D)8G)E86L[\"B`@\"0D)?0H@(`D)8V%S92!47T%?17AP<CH*\nM+2TM(#$R,BPQ-C`@+2TM+0H@(`D)\"0EP87)A;2T^<&%R86UT>7!E(#T@*$]I\nM9\"D@=&]I9#L*(\"`)\"0D)<&%R86TM/G!A<F%M7W1L:7-T(#T@*$QI<W0@*BD@\nM3E5,3#L*(\"`*(2`)\"0D):68@*'!N;RT^:6YD:7)E8W1I;VX@(3T@3DE,*0HA\nM(`D)\"0E[\"B$@\"0D)\"0E,:7-T\"2`@(\"II9'@@/2!P;F\\M/FEN9&ER96-T:6]N\nM.PHA(`HA(`D)\"0D)=VAI;&4@*&ED>\"`A/2!.24PI\"B$@\"0D)\"0E[\"B$@\"0D)\nM\"0D)05]);F1I8V5S(\"`J86D@/2`H05]);F1I8V5S(\"HI(&QF:7)S=\"AI9'@I\nM.PHA(`D)\"0D)\"4YO9&4)(\"`@*FQE>'!R(#T@3E5,3\"P*(2`)\"0D)\"0D)\"2`@\nM(\"IU97AP<CL*(2`*(2`)\"0D)\"0EU97AP<B`]('1R86YS9F]R;45X<'(H<'-T\nM871E+\"!A:2T^=6ED>\"P@<')E8V5D96YC92D[\"2\\J(&UU<W0@97AI<W1S(\"HO\nM\"B$@\"0D)\"0D):68@*&5X<')4>7!E*'5E>'!R*2`A/2!)3E0T3TE$*0HA(`D)\nM\"0D)\"0EE;&]G*$524D]2+\"`B87)R87D@:6YD97@@97AP<F5S<VEO;G,@;75S\nM=\"!B92!I;G0T)W,B*3L*(2`)\"0D)\"0EI9B`H86DM/FQI9'@@(3T@3E5,3\"D*\nM(2`)\"0D)\"0E[\"B$@\"0D)\"0D)\"6QE>'!R(#T@=')A;G-F;W)M17AP<BAP<W1A\nM=&4L(&%I+3YL:61X+\"!P<F5C961E;F-E*3L*(2`)\"0D)\"0D):68@*&5X<')4\nM>7!E*&QE>'!R*2`A/2!)3E0T3TE$*0HA(`D)\"0D)\"0D)96QO9RA%4E)/4BP@\nM(F%R<F%Y(&EN9&5X(&5X<')E<W-I;VYS(&UU<W0@8F4@:6YT-\"=S(BD[\"B$@\nM\"0D)\"0D)?0HA(`D)\"0D)\"6%I+3YL:61X(#T@;&5X<'([\"B$@\"0D)\"0D)86DM\nM/G5I9'@@/2!U97AP<CL*(2`*(2`)\"0D)\"0DO*@HA(`D)\"0D)\"2`J(&YO=&4@\nM=V4@<F5U<V4@=&AE(&QI<W0@;V8@:6YD:6-E<RP@;6%K92!S=7)E('=E\"B$@\nM\"0D)\"0D)(\"H@9&]N)W0@9G)E92!T:&5M(2!/=&AE<G=I<V4L(&UA:V4@82!N\nM97<@;&ES=`HA(`D)\"0D)\"2`J(&AE<F4*(2`)\"0D)\"0D@*B\\*(2`)\"0D)\"0EI\nM9'@@/2!L;F5X=\"AI9'@I.PHA(`D)\"0D)?0HA(`D)\"0D)<F5S=6QT(#T@*$YO\nM9&4@*BD@;6%K95]A<G)A>5]R968H*$YO9&4@*BEP87)A;2P@<&YO+3YI;F1I\nM<F5C=&EO;BD[\"B$@\"0D)\"7T*(2`)\"0D)96QS90HA(`D)\"0D)<F5S=6QT(#T@\nM*$YO9&4@*BD@<&%R86T[\"B`@\"0D)\"6)R96%K.PH@(`D)\"7T*(\"`)\"6-A<V4@\nM5%]!7T5X<'(Z\"F1I9F8@+6-R('-R8RYO<FEG+VEN8VQU9&4O;F]D97,O<&%R\nM<V5N;V1E<RYH('-R8R]I;F-L=61E+VYO9&5S+W!A<G-E;F]D97,N:`HJ*BH@\nM<W)C+F]R:6<O:6YC;'5D92]N;V1E<R]P87)S96YO9&5S+F@)1G)I($]C=\"`@\nM,B`Q,CHU.#HR-\"`Q.3DX\"BTM+2!S<F,O:6YC;'5D92]N;V1E<R]P87)S96YO\nM9&5S+F@)1G)I($]C=\"`@,B`Q,CHU.3HU-B`Q.3DX\"BHJ*BHJ*BHJ*BHJ*BHJ\nM*@HJ*BH@-C8W+#8W,B`J*BHJ\"BTM+2`V-C<L-C<S(\"TM+2T*(\"`)3F]D951A\nM9PD)='EP93L*(\"`):6YT\"0D);G5M8F5R.PD)\"2\\J('1H92!N=6UB97(@;V8@\nM=&AE('!A<F%M971E<B`J+PH@(`E4>7!E3F%M92`@(\"IT>7!E;F%M93L)\"2\\J\nM('1H92!T>7!E8V%S=\"`J+PHK(`E,:7-T\"2`@(\"II;F1I<F5C=&EO;CL)+RH@\nI87)R87D@<F5F97)E;F-E<R`J+PH@('T@4&%R86U.;SL*(\"`*(\"`O*@H`\n`\nend\n",
"msg_date": "Fri, 2 Oct 1998 14:20:07 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": true,
"msg_subject": "Patch - please apply"
},
{
"msg_contents": "Applied.\n\n> Hi,\n> \n> the following little patch adds array references to query\n> parameters. With it applied a function like\n> \n> CREATE FUNCTION getname(oid8, int4) RETURNS name AS\n> 'SELECT typname FROM pg_type WHERE oid = $1[$2]'\n> LANGUAGE 'sql';\n> \n> is possible. Mainly I need this to enable array references in\n> expressions for PL/pgSQL. Complete regression test ran O.K.\n> \n> \n> Jan\n> \n> --\n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #======================================== [email protected] (Jan Wieck) #\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 12:21:36 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Patch - please apply"
}
] |
[
{
"msg_contents": "Dear Henrique,\n\tI have been using PostgreSQL on Irix for several years, and it\nworks very well. The Perl interface (src/interfaces/perl5) works\nnicely too, especially for CGI scripts. The problem you have run into\nstems from the use of the current developmental snapshot instead of\nthe released version 6.3.2. Download the released version, and unpack\nit. Put the following into src/Makefile.custom:\n\nCUSTOM_CC = cc -32\nLD += -32\nMK_NO_LORDER = 1\nLIBS = $(filter-out -nsl, $(LIBS))\n\nUse a initialization script like the following to setup the\nenvironment for running or compiling PostgreSQL (obviously,\nyou have to replace the paths, machine names, and port number\nwith appropriate values for your installation.\n\n# Setup to run postgresql 6.3.2 under csh.\n\nset path = ( /u/stone/pg/pg6.3.2/bin `printenv PATH | tr ':' '\\012' | grep -v postgres/bin` )\nif ($?MANPATH) then\n setenv MANPATH ${MANPATH}:/u/stone/pg/pg6.3.2/man\nelse\n setenv MANPATH /usr/share/catman:/usr/share/man:/usr/catman:/usr/man:/u/stone/pg/pg6.3.2/man\nendif\nsetenv PGHOST stone\nsetenv PGPORT 6546\nsetenv POSTGRES_HOME /u/stone/pg/pg6.3.2\nsetenv PGDATA /u/stone/pg/pg6.3.2/data\nsetenv PGLIB /u/stone/pg/pg6.3.2/lib\nif ($?POSTGRESDIR) then\n unsetenv POSTGRESDIR\nendif\n\n========================================================================\n\nNext, apply the locking patch I posted to pgsql-patches (volume 1 # 107)\n\nThen, study the script below, edit it to suit your installation,\nand then run it to build PostgreSQL\n\n#!/bin/ksh -x\n\ncd src\ncat >config.cache <<EOF\nac_cv_lib_nsl_main=${ac_cv_lib_nsl_main='no'}\nEOF\ngmake clean\nINSTALL=/usr/local/bin/install ./configure --prefix=/pg/pg6.3.2 --enable-hba --with-pgport=6546 --disable-locale --enable-cassert --with-template=irix5 --with-tcl --with-includes=\"/usr/local/include /usr/local/include/readline\" --with-libs=/usr/local/lib\ngmake -j 2\ngmake install\nexport PATH=/pg/pg6.3.2/bin:$PATH\ninitdb --pgdata=/pg/pg6.3.2/data --pglib=/pg/pg6.3.2/lib\ncd ../data\nif [[ ! -r pg_hba.conf.bak ]]\nthen\n chmod 644 pg_hba.conf\n cp pg_hba.conf pg_hba.conf.bak\n cat >>pg_hba.conf <<EOF\nhost all 0.0.0.0 0.0.0.0 trust\n\n# The above would allow anyone anywhere to connect to any database under\n# any username.\nEOF\n chmod 444 pg_hba.conf\nfi\ncd ../src\nexport TZ=PST8PDT7\npostmaster -S -i\ncd test/regress\ngmake clean\ngmake all runtest\n\n========================================================================\n\nThe Perl5 interface can be built using the following set of commands\nwhen your default directory is src/interfaces/perl5. Note that you\nmust have built perl using the -32 compilation switch and\nit be able to dynamically load objects.\n\n#!/bin/csh -x\n\nperl Makefile.PL\nsmake\nsmake test\nsu root -c \"smake install\"\n\n========================================================================\n\nWith regard to the install script, you can use /usr/bin/X11/bsdinst\nand that should work OK.\n\nIf you need more help, send me email. We certainly don't you\nto use Windoze!\n\n+------------------------------------------+------------------------------+\n| Robert E. Bruccoleri, Ph.D. | Associate Research Professor |\n| phone: 732 235 5796 | Center for Advanced |\n| Fax: 732 235 4850 | Biotechnology and Medicine |\n| email: [email protected] | Rutgers University |\n| URL: http://www.cabm.rutgers.edu/~bruc | 679 Hoes Lane |\n| | Piscataway, NJ 08854-5638 |\n+------------------------------------------+------------------------------+\n",
"msg_date": "Fri, 2 Oct 1998 08:46:26 -0400 (EDT)",
"msg_from": "Robert Bruccoleri <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Having problems compiling postgres on IRIX"
}
] |
[
{
"msg_contents": "\nThomas,\n\nGo ahead and commit your patches for the double quote stuff.\nI need to resync with CVS. I inadverently deleted the original\nfiles that I was using. I will create new patches after the\ncommits from you and Jan.\n\nMatt\n> > After looking at the code for pg_dump, it appears that all identifiers\n> > go through fmtId which does as you thougt, it looks to see if a character\n> > is non-lower case, a digit, or an '_' then it double quotes it.\n> > I am going to change this behavior to always double quote it and see\n> > if it breaks anything.\n> \n> I've already got patches to do the double-quote thing, with a command\n> line option (-n) to recover the old behavior. Will commit that, so you\n> will only need to do the memory leak stuff. Or would you prefer that I\n> send you my patches and you can integrate them?\n> \n> - Tom\n\n----------\nMatthew C. Aycock\nOperating Systems Analyst/Admin, Senior\nDept Math/CS\nEmory University, Atlanta, GA \nInternet: [email protected] \t\t\n\n",
"msg_date": "Fri, 2 Oct 1998 09:16:58 -0400 (EDT)",
"msg_from": "\"Matthew C. Aycock\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] pg_dump"
}
] |
[
{
"msg_contents": "Is it too late to consider distributing a patching mechanism with the\nupcoming release? \n\nProbably everyone has some sort of utility for doing this; I've got\nGenPatches and ApplyPatches which traverse the source tree looking for\n\".orig\" files and running diff on any it finds, saving the results to\n.patch files. If we packaged these routines (perhaps slightly modified\nto encapsulate any patches into a specific area; or we could use someone\nelse's routines) into a new directory\n\n $(POSTGRESDIR)/patch\n\nthen we could generate standard patch tarballs which people could put\ninto that directory and then automatically expand and apply.\n\nDoes anyone want to work on it?\n\n - Tom\n",
"msg_date": "Fri, 02 Oct 1998 15:40:12 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "patching utilities?"
},
{
"msg_contents": "> Is it too late to consider distributing a patching mechanism with the\n> upcoming release? \n> \n> Probably everyone has some sort of utility for doing this; I've got\n> GenPatches and ApplyPatches which traverse the source tree looking for\n> \".orig\" files and running diff on any it finds, saving the results to\n> .patch files. If we packaged these routines (perhaps slightly modified\n> to encapsulate any patches into a specific area; or we could use someone\n> else's routines) into a new directory\n> \n> $(POSTGRESDIR)/patch\n> \n> then we could generate standard patch tarballs which people could put\n> into that directory and then automatically expand and apply.\n> \n> Does anyone want to work on it?\n\nThe Developer's FAQ mentions our difforig, cporig, rmorig tools. To\napply the patch, just use any standard patch program, no?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 12:23:23 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patching utilities?"
},
{
"msg_contents": "> > Is it too late to consider distributing a patching mechanism with \n> > the upcoming release?\n> The Developer's FAQ mentions our difforig, cporig, rmorig tools. To\n> apply the patch, just use any standard patch program, no?\n\nOK, I'll look at those...\n\nI'm also suggesting that we have a convention for and place to put the\npatch collections, say a directory \"patch\" just under the main tree.\nThen we can distribute tarballs and patch files which just need to be\ngenerated and unpacked in a single location to succeed.\n\n - Tom\n",
"msg_date": "Fri, 02 Oct 1998 16:52:29 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] patching utilities?"
}
] |
[
{
"msg_contents": "> > > > You need The SQL3 SIMILAR\n> > > > in this case pattern can involve additional special characters,\n> not \n> > > > just % and _ as in LIKE,\n> > > TODO item changed to:\n> > > * add SIMILAR to allow character classes, 'pg_[a-c]%'\n> > \n> > The annoying thing is that our regex \"~\" operators do all of this\n> and\n> > more, but presumably don't have the right behavior for underscore\n> and\n> > for percent. Should we look at the regex code and try to get it to\n> be\n> > compatible (for v6.5...)?\n> \n> Not sure.\nI thought that the regex code was already being revamped by the\noriginator.\n\t-DEJ\n",
"msg_date": "Fri, 2 Oct 1998 10:41:09 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] SQL92"
},
{
"msg_contents": "> > > > > You need The SQL3 SIMILAR\n> > > > > in this case pattern can involve additional special characters,\n> > not \n> > > > > just % and _ as in LIKE,\n> > > > TODO item changed to:\n> > > > * add SIMILAR to allow character classes, 'pg_[a-c]%'\n> > > \n> > > The annoying thing is that our regex \"~\" operators do all of this\n> > and\n> > > more, but presumably don't have the right behavior for underscore\n> > and\n> > > for percent. Should we look at the regex code and try to get it to\n> > be\n> > > compatible (for v6.5...)?\n> > \n> > Not sure.\n> I thought that the regex code was already being revamped by the\n> originator.\n\nHe is speeding it up, but had no release date yet.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 12:05:14 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
}
] |
[
{
"msg_contents": "> > The SQL92 LIKE:\n> > \n> > character-string-expression [NOT] LIKE pattern [ESCAPE\n> > 'escape-char']\n> > \n> > allows only the special characters % and _\n> > \n> > You need The SQL3 SIMILAR\n> > \n> > character-string-expression [NOT] SIMILAR TO pattern [ESCAPE\n> > 'escape-char']\n> > \n> > in this case pattern can involve additional special characters, not\n> just\n> > % and _ as in LIKE, but every regular expression or\n> \"[character_set]\"\n> > like you said. (Refer to \"A Guide to SQL Standard 4th edition\n> > Date-Rarwen, page 505).\n> > \n> > Jose'\n> > \n> > \n> \n> TODO item changed to:\n> \n> \t* add SIMILAR to allow character classes, 'pg_[a-c]%'\nSo are we going to add the character class support to LIKE or SIMILAR or\nboth? Just looking for clarification. Or are we going to use the regex\ncode for both?\n\t-DEJ \n",
"msg_date": "Fri, 2 Oct 1998 10:44:24 -0500",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] SQL92"
},
{
"msg_contents": "> > > The SQL92 LIKE:\n> > > \n> > > character-string-expression [NOT] LIKE pattern [ESCAPE\n> > > 'escape-char']\n> > > \n> > > allows only the special characters % and _\n> > > \n> > > You need The SQL3 SIMILAR\n> > > \n> > > character-string-expression [NOT] SIMILAR TO pattern [ESCAPE\n> > > 'escape-char']\n> > > \n> > > in this case pattern can involve additional special characters, not\n> > just\n> > > % and _ as in LIKE, but every regular expression or\n> > \"[character_set]\"\n> > > like you said. (Refer to \"A Guide to SQL Standard 4th edition\n> > > Date-Rarwen, page 505).\n> > > \n> > > Jose'\n> > > \n> > > \n> > \n> > TODO item changed to:\n> > \n> > \t* add SIMILAR to allow character classes, 'pg_[a-c]%'\n> So are we going to add the character class support to LIKE or SIMILAR or\n> both? Just looking for clarification. Or are we going to use the regex\n> code for both?\n\nNo idea. It is just a valid TODO item.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 12:09:06 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
},
{
"msg_contents": "> > * add SIMILAR to allow character classes, 'pg_[a-c]%'\n> So are we going to add the character class support to LIKE or SIMILAR \n> or both? Just looking for clarification. Or are we going to use the \n> regex code for both?\n\nI would expect us to stay compatible with the standard for LIKE, and\nperhaps have a superset of the SQL3 (draft) standard for SIMILAR. I'd\nexpect it to be a candidate for v6.5, or perhaps for v6.4.x.\n\nBut I'm not doing the work, and whoever does has a much stronger vote :)\n\n - Tom\n",
"msg_date": "Fri, 02 Oct 1998 16:21:33 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] SQL92"
}
] |
[
{
"msg_contents": "hullo,\n\nive two serious problems with postgresql (latest version):\n\n1.) i seem to be unable (well, i _am_ ;-) to create an index\n on a text field and then have that index being used if\n i perform a search using ~*, ~~, and related operators.\n i found out that these operators can only successfully\n be implemented if i use box* field types etc. this is a real\n pain, since i have that database with more than 70,000\n entries and it always does a sequential scan.\n a query might look like \n \"select * from table where field ~* 'string'\" or\n \"select * from table where field ~~ '%string%' etc.\n there are indices (hash, btree, (field text_ops)), but\n they wont be used.\n can anybody of you possibly help me any further on this issue?\n\n2.) another thing i noticed is that while \"~*\" is supposed to\n perform a case insensitive search, it does not. i havent\n checked too much into the bug behind it (like \"all ascii\n values > 127 ...\"), but the typical austrian and german \n characters like \"ae\" with two dots (forgot the terminologically\n exact name, forgive me) etc. will not be searched \n properly, thus, if i have a field entry with\n 'AEyadayada' \n and perform a search with \n ... ~* 'aeyadayada', \n it wont find anything, i have to use ~* 'AEyadayada'.\n opinions on that one woul dbe appreciated as well!\n\n \n\nbest regards,\n\n+bl.\n\n--\nBernhard Lorenz Managing Partner\n!C Internet Consult http://www.iconsult.at/\nPacassistrasse 32, A-1130 Wien +43/1/319 09 90\nAichholzgasse 6/5, A-1120 Wien +43/1/817 39 23\n",
"msg_date": "Fri, 2 Oct 1998 18:01:19 +0200 (CEST)",
"msg_from": "Bernhard Lorenz <[email protected]>",
"msg_from_op": true,
"msg_subject": "indices: ~* / text_ops"
},
{
"msg_contents": "[Charset ISO-8859-1 unsupported, filtering to ASCII...]\n> hullo,\n> \n> ive two serious problems with postgresql (latest version):\n> \n> 1.) i seem to be unable (well, i _am_ ;-) to create an index\n> on a text field and then have that index being used if\n> i perform a search using ~*, ~~, and related operators.\n> i found out that these operators can only successfully\n> be implemented if i use box* field types etc. this is a real\n> pain, since i have that database with more than 70,000\n> entries and it always does a sequential scan.\n> a query might look like \n> \"select * from table where field ~* 'string'\" or\n> \"select * from table where field ~~ '%string%' etc.\n> there are indices (hash, btree, (field text_ops)), but\n> they wont be used.\n> can anybody of you possibly help me any further on this issue?\n\nIndexes on strings can only be used if the start of the string is\nanchored with ^ because the index only sorts the strings starting with\nthe first character.\n\nI have added this to the FAQ, which is on the web site.\n\n> \n> 2.) another thing i noticed is that while \"~*\" is supposed to\n> perform a case insensitive search, it does not. i havent\n> checked too much into the bug behind it (like \"all ascii\n> values > 127 ...\"), but the typical austrian and german \n> characters like \"ae\" with two dots (forgot the terminologically\n> exact name, forgive me) etc. will not be searched \n> properly, thus, if i have a field entry with\n> 'AEyadayada' \n> and perform a search with \n> ... ~* 'aeyadayada', \n> it wont find anything, i have to use ~* 'AEyadayada'.\n> opinions on that one woul dbe appreciated as well!\n> \n\nNot sure how we handle this. I don't think we handle this, but we\nshould, but it is hard.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 12:37:23 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] indices: ~* / text_ops"
}
] |
[
{
"msg_contents": "Where to start ... base system:\nRed Hat 4.2, flex 2.5.4, 2.0.35 kernel\n(3 differant boxes, all Pentium+, all 90MB+ RAM)\n\npg version:\n6.4 snap shot, several over the last few weeks.\n(the last install went perfect, passing all regresion tests)\n\nProblem:\nA script that does an insert, then about a dozen updates on tiny tables\ngets slower and slower each time it runs, till it finally just will not\nreturn.\n\nWhat I have found out:\nIt has some thing to do with a unique index the never exited in the first\nplace.\n\nThis may be related to some of the posts that I have been seeing here\nlately, but most of them go over my head, so I'm not sure.\n\nBasically, a 2 field table (char() and bool) has a unique index on the\nchar() field, but there is NO index on the bool field, but if one\nrepeatedly runs some things like:\nupdate table set boolfield = 'f';\n(once each on several similar tables)\nthen\nupdate table set boolfield = 't' where charfield = 'something';\n(once each on several similar tables)\n\nafter this is done a few times, some where along the line postgres starts\nto thing there is a unique index on the boolean field of one or more of\nthe tables???\n\nI did have it once telling me that it could not do the secound update,\nit said that it can not insert a duplicate value in a unique field!\n\nIf any one is interested, I do have some screen snips of psql.\nThe whole database is very small, I can even email a dump file and\nassociated PHP 3 scripts to any one that may be that interested.\n\nI have reinstalled several times, with several differant snap shots,\nhoping the problem whould go away, but it is still here.\nAt this point I am going to get the latest snap shoot and try it all from\nscratch, trying to make a 'test case'.\n\nI will try to make a simple case that can show the problem.\n\nThanks all, have a great day\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1..3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n",
"msg_date": "Fri, 2 Oct 1998 16:00:39 -0400 (EDT)",
"msg_from": "Terry Mackintosh <[email protected]>",
"msg_from_op": true,
"msg_subject": "Bug, index problem??"
}
] |
[
{
"msg_contents": "Hiya. Compiled current CVS of postgres. Haven't dl'd a CVS in a while -\nsince about july I guess - and that one worked right up 'till recently...\n\npostgres server seems to work fine.\ninitdb also works fine.\ncreatedb fails, I don't (yet) know how to convince it to tell me -WHY- it\nfails.....\npsql segfaults in pqWait()\n\t[again haven't traced this down - figured I'd try the JDBC driver\nfirst but it doesn't know how to make databases afaik...]\n\nAny ideas where to begin? Or are there any unknowns in my current\ntoollist.. I think I tried gcc-2.7.2.1 as well - I'll try it again. But\nIIRC that didn't work either....\n\nG'day, eh? :)\n\t- Teunis\n\n",
"msg_date": "Fri, 2 Oct 1998 13:08:23 -0700 (MST)",
"msg_from": "teunis <[email protected]>",
"msg_from_op": true,
"msg_subject": "egcs, glibc-2.0.7, linux - any problems known?"
}
] |
[
{
"msg_contents": "There is a bug in check_foreign_key of refint.c which is bundled with\nthe standard distribution. It occurs when a trigger calling this\nfunction recursively fires another trigger which calls the same\nfunction. The calling check_foreign_key loses its plan informantion and\nwhen it tries to use it the backend closes its channel. You can check it\nwith the sql script I am attaching below.\nThe solution to this is to do a find_plan again before executing it at\nline 483 of refint.c.\nTherefore two more lines should be added before line 483:\n\nsprintf(ident, \"%s$%u\", trigger->tgname, rel->rd_id);\nplan = find_plan(ident, &FPlans, &nFPlans);\n\nbefore line 483 which is:\n ret = SPI_execp(plan->splan[r], kvals, NULL, tcount);\n\nHere is the script to check it:\ncreate table a (\n a int4 primary key,\n b int4,\n c int4,\n d int4\n );\n\ncreate table b (\n a int4 primary key,\n b int4,-- foreign key references a\n c int4,\n d int4\n );\n\ncreate table ab (\n a int4 primary key,\n b int4,\n c int4,\n d int4\n );\n\ncreate table bc (\n a int4 primary key,\n b int4,\n c int4,\n d int4\n );\n\n\ncreate function check_primary_key()\n returns opaque\n as '/home/bye2data/postgresql-6.3.2/contrib/spi/refint.so'\n language 'c';\n\ncreate function check_foreign_key()\n returns opaque\n as '/home/bye2data/postgresql-6.3.2/contrib/spi/refint.so'\n language 'c';\n\ncreate trigger t1 before delete or update\n on a for each row execute procedure\n check_foreign_key(1,'cascade','a','b','b');\n\ncreate trigger t2 before insert or update\n on b for each row execute procedure\n check_primary_key('b','a','a');\n\ncreate trigger t3 before delete or update\n on ab for each row execute procedure\n check_foreign_key(2,'cascade','a','a','a','bc','a');\n\ncreate trigger t4 before insert or update\n on a for each row execute procedure\n check_primary_key('a','ab','a');\n\ncreate trigger t5 before insert or update\n on bc for each row execute procedure\n check_primary_key('a','ab','a');\n\ninsert into ab values(1,1,1,1);\ninsert into bc values (1,1,1,1);\ninsert into a values (1,1,1,1);\ndelete from ab where a=1;\n\nSorry for the confusing names of tables and attributes.\n\n- Anand.\n\n",
"msg_date": "Fri, 02 Oct 1998 14:57:23 -0700",
"msg_from": "Anand Surelia <[email protected]>",
"msg_from_op": true,
"msg_subject": "Bug in contrib/spi/refint.c"
},
{
"msg_contents": "> There is a bug in check_foreign_key of refint.c which is bundled with\n> the standard distribution. It occurs when a trigger calling this\n> function recursively fires another trigger which calls the same\n> function. The calling check_foreign_key loses its plan informantion and\n> when it tries to use it the backend closes its channel. You can check it\n> with the sql script I am attaching below.\n> The solution to this is to do a find_plan again before executing it at\n> line 483 of refint.c.\n> Therefore two more lines should be added before line 483:\n> \n> sprintf(ident, \"%s$%u\", trigger->tgname, rel->rd_id);\n> plan = find_plan(ident, &FPlans, &nFPlans);\n> \n> before line 483 which is:\n> ret = SPI_execp(plan->splan[r], kvals, NULL, tcount);\n\nPlease send a context diff to the patches list.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 18:36:07 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [GENERAL] Bug in contrib/spi/refint.c"
}
] |
[
{
"msg_contents": "Possible additions\n------------------\ncidr/IP address type(Tom Helbekkmo, Bruce Momjian)\nrewrite system(Jan)\nnotify fixes(Tom)\n\nSerious Items\n------------\nchange pg args for platforms that don't support argv changes\n\t(setproctitle()?, sendmail hack?)\nhave psql dump out rules text with new function\nman pages/sgml synchronization\ngenerate html/postscript documentation\ngenerate postmaster pid file and remove flock/fcntl lock code\nCREATE TABLE test (x text, s serial) fails if no database creation permission\nregression test all platforms\n\nMinor items\n-----------\ncnf-ify still can exhaust memory, make SET KSQO more generic\npermissions on indexes: what do they do? should it be prevented?\nmulti-verion concurrency control - work-in-progress for 6.5\nimprove reporting of syntax errors by showing location of error in query\nuse index with constants on functions\nallow chaining of pages to allow >8k tuples\nallow multiple generic operators in expressions without the use of parentheses\ndocument/trigger/rule so changes to pg_shadow create pg_pwd\nlarge objects orphanage\nimprove group handling\nno min/max for oid type\nimprove PRIMARY KEY handling\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Fri, 2 Oct 1998 17:58:00 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Open 6.4 items"
}
] |
[
{
"msg_contents": "Merged into the current source tree. Please test.\n\n\n> Bruce Momjian <[email protected]> writes:\n> \n> > Is this the version you want me to install in the current source tree?\n> \n> Yup. It seems to work right, so it's what I'm using right now. As\n> soon as I've got an IPV6 kit installed here, I'll expand it with code\n> to handle that data type -- as you can see, I've left comments all\n> over the code showing what needs to be done for that. I reckon the\n> thing to do is for you to get it properly into the tree, and then I'll\n> do further maintenance on that code, and send patches.\n> \n> The SQL code to set up indexing capability doesn't work after your\n> changes to add OIDs to various names, but I guess that's irrelevant\n> when you integrate it into the base system, right? As for the MAC\n> level ethernet address stuff, if you want to add that as well, I guess\n> we should consider replacing the hard-coded table of manufacturers\n> with a table in template1 -- that is the SQL way, after all. Also,\n> the names of these types should maybe be changed? I use \"ipaddr\" and\n> \"macaddr\", but I can't remember if consensus was reached on whether\n> the IP address data type should be \"cidr\", or \"inet\", or what... If\n> you want me to go through the code and change that before you do the\n> integration, just let me know.\n> \n> -tih\n> -- \n> Popularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sat, 3 Oct 1998 01:46:33 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: CIDR/IP types. Was: [GENERAL] big numbers"
},
{
"msg_contents": "On Sat, 3 Oct 1998, Bruce Momjian wrote:\n\n> Merged into the current source tree. Please test.\n> \n> \n> > Bruce Momjian <[email protected]> writes:\n> > \n> > > Is this the version you want me to install in the current source tree?\n> > \n> > Yup. It seems to work right, so it's what I'm using right now. As\n> > soon as I've got an IPV6 kit installed here, I'll expand it with code\n> > to handle that data type -- as you can see, I've left comments all\n> > over the code showing what needs to be done for that. I reckon the\n> > thing to do is for you to get it properly into the tree, and then I'll\n> > do further maintenance on that code, and send patches.\n> > \n> > The SQL code to set up indexing capability doesn't work after your\n> > changes to add OIDs to various names, but I guess that's irrelevant\n> > when you integrate it into the base system, right? As for the MAC\n> > level ethernet address stuff, if you want to add that as well, I guess\n> > we should consider replacing the hard-coded table of manufacturers\n> > with a table in template1 -- that is the SQL way, after all. Also,\n> > the names of these types should maybe be changed? I use \"ipaddr\" and\n> > \"macaddr\", but I can't remember if consensus was reached on whether\n> > the IP address data type should be \"cidr\", or \"inet\", or what... If\n\n\t*Please* make the appropriate changes and submit a patch...it\nshould be listed as CIDR, not as IPADDR. MACADDR, IMHO, is okay, as its\nthe only way I've ever heard of it being referred...but I do not consider\nIPADDR to be correct terminology...\n\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Sun, 4 Oct 1998 19:52:38 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: CIDR/IP types. Was: [GENERAL] big numbers"
},
{
"msg_contents": "> \t*Please* make the appropriate changes and submit a patch...it\n> should be listed as CIDR, not as IPADDR. MACADDR, IMHO, is okay, as its\n> the only way I've ever heard of it being referred...but I do not consider\n> IPADDR to be correct terminology...\n\nThat is OK, but just let me that CIDR is for networks addresses, and\nthis type supports both host and network addresses.\n\nAttached is a message from Paul Vixie mentioning this issue. It argues\nfor the use of INET, rather than CIDR or IPADDR.\n\n---------------------------------------------------------------------------\n\n\nReplies to three messages here.\n\n> From: Bruce Momjian <[email protected]>\n> Date: Tue, 21 Jul 1998 01:13:34 -0400 (EDT)\n>\n> The only problem is that if we assume /32, how do we auto-netmask class\n> A/B/C addresses? I guess we don't. If they want a netmask, they are\n> going to have to specify it in cidr format.\n\nRight. But read on -- what you're calling a netmask is really a \nprefix length, and I think there's some confusion as to what it is.\n\n> I will be honest. I always found the network/host IP address\n> distinction to be very unclearly outlined in old/non-cidr address\n> displays, and this causes major confusion for me when trying to figure\n> out how things are configured.\n\nMe too.\n\n> I like INET too. It is up to you.\n\nHow do folks feel about polymorphism between IPv4 and IPv6? Should we (a)\nmake it work (either by making internal_length=10 or going variable length)\nor (b) just make this thing IPv4 only and take care of IPv6 separately/later?\n\nI've started to wonder if we ought to call the type INET and limit it to V4.\n(In the C socket bindings, IPv6 addresses are in_addr6 / sockaddr_in6, and\nthe address family is AF_INET6 -- I don't know whether to plan on reflecting\nthis in the postgres types, i.e., use a separate one for IPv6, or not.)\n\n> From: Bruce Momjian <[email protected]>\n> Date: Tue, 21 Jul 1998 01:30:05 -0400 (EDT)\n>\n> > ... but why would you want to know the mantissa without the scale?\n> \n> I guess I thought someone might want to have ipaddr() and netmask()\n> functions so they can do:\n> \n> \tx = 192.7.34.21/24\n> \tipaddr(x) -> 192.7.34.21\n> \tnetmask(x) -> 255.255.255.0\n\nThis is the downreference from above. It does not work that way. /24 is\nnot a shorthand for specifying a netmask -- in CIDR, it's a \"prefix length\".\nThat means \"192.7.34.21/24\" is either (a) a syntax error or (b) equivilent\nto \"192.7.34/24\".\n\nBtw, it appears from my research that the BIND functions *do* impute a \"class\"\nif (a) no \"/width\" is specified and (b) the classful interpretation would be\nlonger than the classless interpretation. No big deal but it qualifies \nsomething I said earlier so I thought I'd mention it.\n\n> \tx = 192.7.0.0/16\n> \tipaddr(x) -> 192.7.0.0\n> \tnetmask(x) -> 255.255.0.0\n> \n> These function are defined on the cidr type, and can be called if\n> someone wants the old output format.\n\nCan we wait and see if someone misses / asks for these before we make them?\n\n> ..., the 127.1 ambiguity was very strange. netstat -rn is very hard to\n> understand using the old format.\n\nI was amazed at the number of people who had hardwired \"127.1\" though :-(.\n\n> From: Bruce Momjian <[email protected]>\n> Date: Tue, 21 Jul 1998 01:33:41 -0400 (EDT)\n>\n> Doing complex stuff like indexing with contrib stuff is tricky, and one\n> reason we want to move stuff out of there as it becomes popular. It is\n> just too hard for someone not experienced with the code to implement. \n> Add to this the fact that the oid at the time of contrib installation\n> will change every time you install it, so it is even harder/impossible\n> to automate.\n\nPerhaps we ought to make new type insertion easier since it's so cool?\n\n\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 4 Oct 1998 20:57:19 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: CIDR/IP types. Was: [GENERAL] big numbers"
},
{
"msg_contents": "\"Marc G. Fournier\" <[email protected]> writes:\n\n> \t*Please* make the appropriate changes and submit a patch...it\n> should be listed as CIDR, not as IPADDR. MACADDR, IMHO, is okay, as its\n> the only way I've ever heard of it being referred...but I do not consider\n> IPADDR to be correct terminology...\n\nI agree. I think, though, that the best argument presented in the\ndebate was from Paul Vixie, who wanted INET to be the name covering\nboth IPV4 and IPV6. The following kit makes the needed changes:\n\nWARNING! This posting contains two separate patch kits, to be applied\nat different times, renaming a couple of files in between them! The\nsteps to be taken are marked with \"[***]\".\n\n[***] First apply these patches to change ipaddr to inet. Affected files are:\n\n- src/backend/utils/adt/ip.c\n- src/include/catalog/pg_opclass.h\n- src/include/catalog/pg_operator.h\n- src/include/catalog/pg_proc.h\n- src/include/catalog/pg_type.h\n- src/include/utils/builtins.h\n- src/include/utils/mac.h\n- src/tools/pgindent/pgindent\n\nIndex: src/backend/utils/adt/ip.c\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/backend/utils/adt/ip.c,v\nretrieving revision 1.4\ndiff -r1.4 ip.c\n2c2\n< *\tPostgreSQL type definitions for IP addresses. This\n---\n> *\tPostgreSQL type definitions for the INET type. This\n23a24,25\n> static int v4bitncmp(unsigned int a1, unsigned int a2, int bits);\n> \n28,29c30,31\n< #define ip_addrsize(ipaddrptr) \\\n< \t(((ipaddr_struct *)VARDATA(ipaddrptr))->family == AF_INET ? 4 : -1)\n---\n> #define ip_addrsize(inetptr) \\\n> \t(((inet_struct *)VARDATA(inetptr))->family == AF_INET ? 4 : -1)\n31,32c33,34\n< #define ip_family(ipaddrptr) \\\n< \t(((ipaddr_struct *)VARDATA(ipaddrptr))->family)\n---\n> #define ip_family(inetptr) \\\n> \t(((inet_struct *)VARDATA(inetptr))->family)\n34,35c36,37\n< #define ip_bits(ipaddrptr) \\\n< \t(((ipaddr_struct *)VARDATA(ipaddrptr))->bits)\n---\n> #define ip_bits(inetptr) \\\n> \t(((inet_struct *)VARDATA(inetptr))->bits)\n37,38c39,40\n< #define ip_v4addr(ipaddrptr) \\\n< \t(((ipaddr_struct *)VARDATA(ipaddrptr))->addr.ipv4_addr)\n---\n> #define ip_v4addr(inetptr) \\\n> \t(((inet_struct *)VARDATA(inetptr))->addr.ipv4_addr)\n44,45c46,47\n< ipaddr *\n< ipaddr_in(char *src)\n---\n> inet *\n> inet_in(char *src)\n48c50\n< \tipaddr\t *dst;\n---\n> \tinet\t *dst;\n50c52\n< \tdst = palloc(VARHDRSZ + sizeof(ipaddr_struct));\n---\n> \tdst = palloc(VARHDRSZ + sizeof(inet_struct));\n53c55\n< \t\telog(ERROR, \"unable to allocate memory in ipaddr_in()\");\n---\n> \t\telog(ERROR, \"unable to allocate memory in inet_in()\");\n78c80\n< ipaddr_out(ipaddr *src)\n---\n> inet_out(inet *src)\n102c104\n< \t\telog(ERROR, \"unable to allocate memory in ipaddr_out()\");\n---\n> \t\telog(ERROR, \"unable to allocate memory in inet_out()\");\n114c116\n< ipaddr_lt(ipaddr *a1, ipaddr *a2)\n---\n> inet_lt(inet *a1, inet *a2)\n132c134\n< ipaddr_le(ipaddr *a1, ipaddr *a2)\n---\n> inet_le(inet *a1, inet *a2)\n134c136\n< \treturn (ipaddr_lt(a1, a2) || ipaddr_eq(a1, a2));\n---\n> \treturn (inet_lt(a1, a2) || inet_eq(a1, a2));\n138c140\n< ipaddr_eq(ipaddr *a1, ipaddr *a2)\n---\n> inet_eq(inet *a1, inet *a2)\n155c157\n< ipaddr_ge(ipaddr *a1, ipaddr *a2)\n---\n> inet_ge(inet *a1, inet *a2)\n157c159\n< \treturn (ipaddr_gt(a1, a2) || ipaddr_eq(a1, a2));\n---\n> \treturn (inet_gt(a1, a2) || inet_eq(a1, a2));\n161c163\n< ipaddr_gt(ipaddr *a1, ipaddr *a2)\n---\n> inet_gt(inet *a1, inet *a2)\n179c181\n< ipaddr_ne(ipaddr *a1, ipaddr *a2)\n---\n> inet_ne(inet *a1, inet *a2)\n181c183\n< \treturn (!ipaddr_eq(a1, a2));\n---\n> \treturn (!inet_eq(a1, a2));\n185c187\n< ipaddr_sub(ipaddr *a1, ipaddr *a2)\n---\n> inet_sub(inet *a1, inet *a2)\n202c204\n< ipaddr_subeq(ipaddr *a1, ipaddr *a2)\n---\n> inet_subeq(inet *a1, inet *a2)\n219c221\n< ipaddr_sup(ipaddr *a1, ipaddr *a2)\n---\n> inet_sup(inet *a1, inet *a2)\n236c238\n< ipaddr_supeq(ipaddr *a1, ipaddr *a2)\n---\n> inet_supeq(inet *a1, inet *a2)\n257c259\n< ipaddr_cmp(ipaddr *a1, ipaddr *a2)\n---\n> inet_cmp(inet *a1, inet *a2)\n270c272\n< int\n---\n> static int\nIndex: src/include/catalog/pg_opclass.h\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/include/catalog/pg_opclass.h,v\nretrieving revision 1.12\ndiff -r1.12 pg_opclass.h\n112c112\n< DATA(insert OID = 935 (\tipaddr_ops 869 ));\n---\n> DATA(insert OID = 935 (\tinet_ops 869 ));\nIndex: src/include/catalog/pg_operator.h\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/include/catalog/pg_operator.h,v\nretrieving revision 1.39\ndiff -r1.39 pg_operator.h\n647,656c647,656\n< DATA(insert OID = 1201 ( \"=\"\t PGUID 0 b t t 869 869\t 16 1201 1202 0 0 ipaddr_eq eqsel eqjoinsel ));\n< DATA(insert OID = 1202 ( \"<>\"\t PGUID 0 b t f 869 869\t 16 1202 1201 0 0 ipaddr_ne neqsel neqjoinsel ));\n< DATA(insert OID = 1203 ( \"<\"\t PGUID 0 b t f 869 869\t 16 1205 1206 0 0 ipaddr_lt intltsel intltjoinsel ));\n< DATA(insert OID = 1204 ( \"<=\"\t PGUID 0 b t f 869 869\t 16 1206 1205 0 0 ipaddr_le intltsel intltjoinsel ));\n< DATA(insert OID = 1205 ( \">\"\t PGUID 0 b t f 869 869\t 16 1203 1204 0 0 ipaddr_gt intltsel intltjoinsel ));\n< DATA(insert OID = 1206 ( \">=\"\t PGUID 0 b t f 869 869\t 16 1204 1203 0 0 ipaddr_ge intltsel intltjoinsel ));\n< DATA(insert OID = 931 ( \"<<\"\t PGUID 0 b t f 869 869 16 933 934 0 0 ipaddr_sub intltsel intltjoinsel ));\n< DATA(insert OID = 932 ( \"<<=\"\t PGUID 0 b t f 869 869 16 934 933 0 0 ipaddr_subeq intltsel intltjoinsel ));\n< DATA(insert OID = 933 ( \">>\"\t PGUID 0 b t f 869 869 16 931 932 0 0 ipaddr_sup intltsel intltjoinsel ));\n< DATA(insert OID = 934 ( \">>=\"\t PGUID 0 b t f 869 869 16 932 931 0 0 ipaddr_supeq intltsel intltjoinsel ));\n---\n> DATA(insert OID = 1201 ( \"=\"\t PGUID 0 b t t 869 869\t 16 1201 1202 0 0 inet_eq eqsel eqjoinsel ));\n> DATA(insert OID = 1202 ( \"<>\"\t PGUID 0 b t f 869 869\t 16 1202 1201 0 0 inet_ne neqsel neqjoinsel ));\n> DATA(insert OID = 1203 ( \"<\"\t PGUID 0 b t f 869 869\t 16 1205 1206 0 0 inet_lt intltsel intltjoinsel ));\n> DATA(insert OID = 1204 ( \"<=\"\t PGUID 0 b t f 869 869\t 16 1206 1205 0 0 inet_le intltsel intltjoinsel ));\n> DATA(insert OID = 1205 ( \">\"\t PGUID 0 b t f 869 869\t 16 1203 1204 0 0 inet_gt intltsel intltjoinsel ));\n> DATA(insert OID = 1206 ( \">=\"\t PGUID 0 b t f 869 869\t 16 1204 1203 0 0 inet_ge intltsel intltjoinsel ));\n> DATA(insert OID = 931 ( \"<<\"\t PGUID 0 b t f 869 869 16 933 934 0 0 inet_sub intltsel intltjoinsel ));\n> DATA(insert OID = 932 ( \"<<=\"\t PGUID 0 b t f 869 869 16 934 933 0 0 inet_subeq intltsel intltjoinsel ));\n> DATA(insert OID = 933 ( \">>\"\t PGUID 0 b t f 869 869 16 931 932 0 0 inet_sup intltsel intltjoinsel ));\n> DATA(insert OID = 934 ( \">>=\"\t PGUID 0 b t f 869 869 16 932 931 0 0 inet_supeq intltsel intltjoinsel ));\nIndex: src/include/catalog/pg_proc.h\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/include/catalog/pg_proc.h,v\nretrieving revision 1.71\ndiff -r1.71 pg_proc.h\n2072c2072\n< DATA(insert OID = 910 ( ipaddr_in\t\t\tPGUID 11 f t f 1 f 869 \"0\" 100 0 0 100\tfoo bar ));\n---\n> DATA(insert OID = 910 ( inet_in\t\t\tPGUID 11 f t f 1 f 869 \"0\" 100 0 0 100\tfoo bar ));\n2074c2074\n< DATA(insert OID = 911 ( ipaddr_out\t\t\tPGUID 11 f t f 1 f 23 \"0\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 911 ( inet_out\t\t\tPGUID 11 f t f 1 f 23 \"0\" 100 0 0 100 foo bar ));\n2077c2077\n< DATA(insert OID = 920 ( ipaddr_eq\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 920 ( inet_eq\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2079c2079\n< DATA(insert OID = 921 ( ipaddr_lt\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 921 ( inet_lt\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2081c2081\n< DATA(insert OID = 922 ( ipaddr_le\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 922 ( inet_le\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2083c2083\n< DATA(insert OID = 923 ( ipaddr_gt\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 923 ( inet_gt\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2085c2085\n< DATA(insert OID = 924 ( ipaddr_ge\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 924 ( inet_ge\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2087c2087\n< DATA(insert OID = 925 ( ipaddr_ne\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 925 ( inet_ne\t\t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2089c2089\n< DATA(insert OID = 926 ( ipaddr_cmp\t \t PGUID 11 f t f 2 f 23 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 926 ( inet_cmp\t \t PGUID 11 f t f 2 f 23 \"869 869\" 100 0 0 100 foo bar ));\n2091c2091\n< DATA(insert OID = 927 ( ipaddr_sub\t \t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 927 ( inet_sub\t \t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2093c2093\n< DATA(insert OID = 928 ( ipaddr_subeq \t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 928 ( inet_subeq \t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2095c2095\n< DATA(insert OID = 929 ( ipaddr_sup\t \t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 929 ( inet_sup\t \t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n2097c2097\n< DATA(insert OID = 930 ( ipaddr_supeq \t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\n---\n> DATA(insert OID = 930 ( inet_supeq \t PGUID 11 f t f 2 f 16 \"869 869\" 100 0 0 100 foo bar ));\nIndex: src/include/catalog/pg_type.h\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/include/catalog/pg_type.h,v\nretrieving revision 1.48\ndiff -r1.48 pg_type.h\n303c303\n< DATA(insert OID = 869 ( ipaddr\t PGUID -1 -1 f b t \\054 0 0 ipaddr_in ipaddr_out ipaddr_in ipaddr_out i _null_ ));\n---\n> DATA(insert OID = 869 ( inet\t PGUID -1 -1 f b t \\054 0 0 inet_in inet_out inet_in inet_out i _null_ ));\n342c342\n< DATA(insert OID = 1041 ( _ipaddr PGUID -1 -1 f b t \\054 0 869 array_in array_out array_in array_out i _null_ ));\n---\n> DATA(insert OID = 1041 ( _inet PGUID -1 -1 f b t \\054 0 869 array_in array_out array_in array_out i _null_ ));\nIndex: src/include/utils/builtins.h\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/include/utils/builtins.h,v\nretrieving revision 1.57\ndiff -r1.57 builtins.h\n522,535c522,534\n< ipaddr\t *ipaddr_in(char *str);\n< char\t *ipaddr_out(ipaddr * addr);\n< bool\t\tipaddr_lt(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_le(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_eq(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_ge(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_gt(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_ne(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_sub(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_subeq(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_sup(ipaddr * a1, ipaddr * a2);\n< bool\t\tipaddr_supeq(ipaddr * a1, ipaddr * a2);\n< int4\t\tipaddr_cmp(ipaddr * a1, ipaddr * a2);\n< int\t\t\tv4bitncmp(unsigned int a1, unsigned int a2, int bits);\n---\n> inet\t *inet_in(char *str);\n> char\t *inet_out(inet * addr);\n> bool\t\tinet_lt(inet * a1, inet * a2);\n> bool\t\tinet_le(inet * a1, inet * a2);\n> bool\t\tinet_eq(inet * a1, inet * a2);\n> bool\t\tinet_ge(inet * a1, inet * a2);\n> bool\t\tinet_gt(inet * a1, inet * a2);\n> bool\t\tinet_ne(inet * a1, inet * a2);\n> bool\t\tinet_sub(inet * a1, inet * a2);\n> bool\t\tinet_subeq(inet * a1, inet * a2);\n> bool\t\tinet_sup(inet * a1, inet * a2);\n> bool\t\tinet_supeq(inet * a1, inet * a2);\n> int4\t\tinet_cmp(inet * a1, inet * a2);\nIndex: src/include/utils/mac.h\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/include/utils/mac.h,v\nretrieving revision 1.2\ndiff -r1.2 mac.h\n29c29\n< }\t\t\tipaddr_struct;\n---\n> }\t\t\tinet_struct;\n31c31\n< typedef struct varlena ipaddr;\n---\n> typedef struct varlena inet;\nIndex: src/tools/pgindent/pgindent\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/tools/pgindent/pgindent,v\nretrieving revision 1.11\ndiff -r1.11 pgindent\n789c789\n< -Tipaddr \\\n---\n> -Tinet \\\n\n[***] Then, some renaming is in order:\n\n- rename doc/README.ipaddr to doc/README.inet\n- rename src/backend/utils/adt/ip.c to src/backend/utils/adt/inet.c\n- rename src/include/utils/mac.h to src/include/utils/network.h\n\n[***] Finally, apply these patches to reflect the renaming. Files are:\n\n- src/backend/utils/adt/Makefile\n- src/backend/utils/adt/inet.c\n- src/backend/utils/adt/mac.c\n- src/include/utils/builtins.h\n\n*** src/backend/utils/adt/Makefile.old\tWed Oct 7 10:45:34 1998\n--- src/backend/utils/adt/Makefile\tWed Oct 7 10:46:38 1998\n***************\n*** 24,30 ****\n \toid.o oracle_compat.o \\\n \tregexp.o regproc.o ruleutils.o selfuncs.o sets.o \\\n \ttid.o timestamp.o varchar.o varlena.o version.o \\\n! \tip.o mac.o inet_net_ntop.o inet_net_pton.o\n \n all: SUBSYS.o\n \n--- 24,30 ----\n \toid.o oracle_compat.o \\\n \tregexp.o regproc.o ruleutils.o selfuncs.o sets.o \\\n \ttid.o timestamp.o varchar.o varlena.o version.o \\\n! \tinet.o mac.o inet_net_ntop.o inet_net_pton.o\n \n all: SUBSYS.o\n \n*** src/backend/utils/adt/inet.c.old\tWed Oct 7 10:45:38 1998\n--- src/backend/utils/adt/inet.c\tWed Oct 7 10:46:52 1998\n***************\n*** 19,25 ****\n #include <postgres.h>\n #include <utils/palloc.h>\n #include <utils/builtins.h>\n! #include <utils/mac.h>\n \n static int v4bitncmp(unsigned int a1, unsigned int a2, int bits);\n \n--- 19,25 ----\n #include <postgres.h>\n #include <utils/palloc.h>\n #include <utils/builtins.h>\n! #include <utils/network.h>\n \n static int v4bitncmp(unsigned int a1, unsigned int a2, int bits);\n \n*** src/backend/utils/adt/mac.c.old\tWed Oct 7 10:45:43 1998\n--- src/backend/utils/adt/mac.c\tWed Oct 7 10:47:11 1998\n***************\n*** 10,16 ****\n #include <postgres.h>\n #include <utils/palloc.h>\n #include <utils/builtins.h>\n! #include <utils/mac.h>\n \n manufacturer manufacturers[] = {\n {0x00, 0x00, 0x0C, \"Cisco\"},\n--- 10,16 ----\n #include <postgres.h>\n #include <utils/palloc.h>\n #include <utils/builtins.h>\n! #include <utils/network.h>\n \n manufacturer manufacturers[] = {\n {0x00, 0x00, 0x0C, \"Cisco\"},\n*** src/include/utils/builtins.h.old\tWed Oct 7 10:45:54 1998\n--- src/include/utils/builtins.h\tWed Oct 7 10:47:28 1998\n***************\n*** 28,34 ****\n #include <utils/nabstime.h>\n #include <utils/int8.h>\n #include <utils/cash.h>\n! #include <utils/mac.h>\n #include <utils/rel.h>\n \n /*\n--- 28,34 ----\n #include <utils/nabstime.h>\n #include <utils/int8.h>\n #include <utils/cash.h>\n! #include <utils/network.h>\n #include <utils/rel.h>\n \n /*\n***************\n*** 518,524 ****\n /* inet_net_pton.c */\n int inet_net_pton(int af, const char *src, void *dst, size_t size);\n \n! /* ip.c */\n inet\t *inet_in(char *str);\n char\t *inet_out(inet * addr);\n bool\t\tinet_lt(inet * a1, inet * a2);\n--- 518,524 ----\n /* inet_net_pton.c */\n int inet_net_pton(int af, const char *src, void *dst, size_t size);\n \n! /* inet.c */\n inet\t *inet_in(char *str);\n char\t *inet_out(inet * addr);\n bool\t\tinet_lt(inet * a1, inet * a2);\n\n[***] Hope that I got it right! \"It works for me.\" :-)\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n",
"msg_date": "07 Oct 1998 11:28:08 +0200",
"msg_from": "Tom Ivar Helbekkmo <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: CIDR/IP types. Was: [GENERAL] big numbers"
},
{
"msg_contents": "Applied.\n\nThomas, did remove the README.ip, or did you leave it when you decided\nto merge it into the docs. I still show it here, so I am renaming it.\n\n\n\n> \"Marc G. Fournier\" <[email protected]> writes:\n> \n> > \t*Please* make the appropriate changes and submit a patch...it\n> > should be listed as CIDR, not as IPADDR. MACADDR, IMHO, is okay, as its\n> > the only way I've ever heard of it being referred...but I do not consider\n> > IPADDR to be correct terminology...\n> \n> I agree. I think, though, that the best argument presented in the\n> debate was from Paul Vixie, who wanted INET to be the name covering\n> both IPV4 and IPV6. The following kit makes the needed changes:\n> \n> WARNING! This posting contains two separate patch kits, to be applied\n> at different times, renaming a couple of files in between them! The\n> steps to be taken are marked with \"[***]\".\n> \n> [***] First apply these patches to change ipaddr to inet. Affected files are:\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Wed, 7 Oct 1998 20:19:51 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: CIDR/IP types. Was: [GENERAL] big numbers"
},
{
"msg_contents": "> Thomas, did remove the README.ip, or did you leave it when you decided\n> to merge it into the docs. I still show it here, so I am renaming it.\n\nI have not yet merged it; it is on my ToDo list for documenting along\nwith the other new datatypes INT8 and SERIAL. We should leave it in the\nREADME until I do the merge, just in case...\n\n - Tom\n",
"msg_date": "Thu, 08 Oct 1998 07:16:24 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: CIDR/IP types. Was: [GENERAL] big numbers"
}
] |
[
{
"msg_contents": "> Bruce Momjian <[email protected]> writes:\n> \n> > Possible additions\n> > ------------------\n> > cidr/IP address type(Tom Helbekkmo)\n> \n> It looks like the code I sent you is working fine -- if you want to\n> integrate it into the backend, I'd say go ahead. Also, if you at the\n> same time could make notes explaining how this is done, that would\n> make it possible for the rest of us to do such things ourselves...\n\nIt is very hard to do this. If you look at how I integrated this type,\nyou can see what is involved. There are things pointing everywhere, so\nit is really hard to describe without understanding all the tables and\ntheir interaction.\n\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sat, 3 Oct 1998 01:50:24 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 Items"
}
] |
[
{
"msg_contents": "Additions\n------------------\ntest new cidr/IP address type(Tom Helbekkmo)\ncomplete rewrite system changes(Jan)\nnotify fixes(Tom)\n\nSerious Items\n------------\nchange pg args for platforms that don't support argv changes\n\t(setproctitle()?, sendmail hack?)\nhave psql dump out rules text with new function\nman pages/sgml synchronization\ngenerate html/postscript documentation\ngenerate postmaster pid file and remove flock/fcntl lock code\nCREATE TABLE test (x text, s serial) fails if no database creation permission\nregression test all platforms\nmake sure all changes are documented properly\n\nMinor items\n-----------\ncnf-ify still can exhaust memory, make SET KSQO more generic\npermissions on indexes: what do they do? should it be prevented?\nmulti-verion concurrency control - work-in-progress for 6.5\nimprove reporting of syntax errors by showing location of error in query\nuse index with constants on functions\nallow chaining of pages to allow >8k tuples\nallow multiple generic operators in expressions without the use of parentheses\ndocument/trigger/rule so changes to pg_shadow create pg_pwd\nlarge objects orphanage\nimprove group handling\nno min/max for oid type\nimprove PRIMARY KEY handling\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sat, 3 Oct 1998 02:05:22 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Open 6.4 items"
},
{
"msg_contents": "I'm working on the multi-byte support for win32 env. with help from\nHiroshi Inoue who appears in the discussion on some win32 issues.\nIn one or two days, I will post new patches that should make\nthe multi-byte support usable on win32.\nNote that only Japanese(Shift-JIS) will be supported in the first\nversion.\n\nI hope this is not too late for the 6.4 release.\n--\nTatsuo Ishii\[email protected]\n",
"msg_date": "Mon, 05 Oct 1998 11:04:12 +0900",
"msg_from": "Tatsuo Ishii <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "On Sat, 3 Oct 1998, Bruce Momjian wrote:\n\n> Additions\n\n\tRename as 'release stopper'...\n\n> ------------------\n> test new cidr/IP address type(Tom Helbekkmo)\n> complete rewrite system changes(Jan)\n> notify fixes(Tom)\n\n\tThese have obviously become show stoppers, since they are now half\nimplemented, and have to be completed before release. Do we have ETAs on\nthis stuff? As things stand right now, we are looking at November 1st\nfor a release date on v6.4...a month late, but not much worse then other\nreleases :) Unfortunately, because of the lateness for some of this, we\nhaven't done a solid freeze yet, which means that the testing cycle is\ngoing to be a little bit shorter then we like :( \n\n\tAt this time, I'd like to say that the source tree is now\n*absolutely* frozen...no more \"just one more thing\"...the above three\nissues will still need work, as they haven't been fully tested, nor do I\nbelieve them to be fully implemented...but, as a result, they can still be\nconsidered to be \"bugs\", and, as such, code changes can still be\nmade/submitted to fix those bugs...\n\n> Serious Items\n> ------------\n> change pg args for platforms that don't support argv changes\n> \t(setproctitle()?, sendmail hack?)\n> have psql dump out rules text with new function\n> man pages/sgml synchronization\n> generate html/postscript documentation\n> generate postmaster pid file and remove flock/fcntl lock code\n\n\tNone of these, IMHO, are release stoppers...documentation related\nissues don't truly apply to the beta freeze anyway, but unless there is a\nreal reason that the above coding issues interfere with the ability for\nthe backend to run on a particular platform, then let's leave them as they\nare. I'm going to look at the setproctitle stuff this week, but will not\nbe committing any changes until *after* the release, at which time I will\ncreate a seperate patch...\n\n\tAs for the flock/fcntl lock code...my understanding of it is that\nit is only 'startup' related, and doesn't really affect anything...leave\nit there for now, create a patch shortly after release that can be used to\nremove it if it is causing problems for some...\n\n> CREATE TABLE test (x text, s serial) fails if no database creation permission\n\n\trelease stopper...please move it up with the other three...\n\n> regression test all platforms\n\n\tsame as the CREATE TABLE...\n\n> make sure all changes are documented properly\n\n\tnot included as part of beta freeze, changes here can be made at\nany time up till the release date, since it does not affect the running of\nthe backend...\n\n> Minor items\n> -----------\n> cnf-ify still can exhaust memory, make SET KSQO more generic\n> permissions on indexes: what do they do? should it be prevented?\n> multi-verion concurrency control - work-in-progress for 6.5\n> improve reporting of syntax errors by showing location of error in query\n> use index with constants on functions\n> allow chaining of pages to allow >8k tuples\n> allow multiple generic operators in expressions without the use of parentheses\n> document/trigger/rule so changes to pg_shadow create pg_pwd\n> large objects orphanage\n> improve group handling\n> no min/max for oid type\n> improve PRIMARY KEY handling\n\n\tactually, I'd pretty much say that none of the 'minor' ones are\nrelease stoppers...and, as such, should be moved over to \"todo for\nv6.5\"...\n\t\n\tFor v6.4, can we just concentrate on, and remove from the list\neverything else, the following:\n\n> test new cidr/IP address type(Tom Helbekkmo)\n> complete rewrite system changes(Jan)\n> notify fixes(Tom)\n> CREATE TABLE test (x text, s serial) fails if no database creation permission\n> regression test all platforms\n\n\tEverything else appears to be a \"ya, be nice if we could, but it can\nwait\" sort of issue...\n\n\tIf we can get those 5 cleaned up/out, then I think we have a good\nrelease candidate to snapshot and test...\n\t\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Mon, 5 Oct 1998 01:28:18 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
},
{
"msg_contents": "\"Marc G. Fournier\" <[email protected]> writes:\n> \tRename as 'release stopper'...\n>> notify fixes(Tom)\n>> [other items snipped]\n> \tThese have obviously become show stoppers, since they are now half\n> implemented, and have to be completed before release. Do we have ETAs on\n> this stuff?\n\nMy notify rewrite is not in the tree at all right now. I thought we had\nagreed (off-list) not to put that change into 6.4, but to postpone it\nto the next release.\n\n> As things stand right now, we are looking at November 1st\n> for a release date on v6.4...a month late, but not much worse then other\n> releases :)\n\nOn the other hand, if release is going to be 11/1 not 10/1, my personal\nvote is to put the notify changes in. If there are any bugs, that ought\nto be time enough to shake them out.\n\nI can commit those changes tonight if I have the go-ahead. Or I can\nwait till post-6.4. Your call.\n\n>> Serious Items\n>> ------------\n>> [snippage]\n>> generate postmaster pid file and remove flock/fcntl lock code\n\n> \tNone of these, IMHO, are release stoppers...\n\nflock is a release stopper as far as I'm concerned, because the backend\n*does not compile* on my platform without diking out that code. I agree\nthat it is too late to try to rewrite the feature correctly for 6.4.\nWhat say we put in an autoconf test for whether flock exists, and make\nthe new code conditional on that? People without flock would see the\nsame old behavior, which is good enough for now. I will volunteer to\nmake the necessary changes if that strategy is agreed on.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 05 Oct 1998 10:36:06 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "On Mon, 5 Oct 1998, Tom Lane wrote:\n\n> \"Marc G. Fournier\" <[email protected]> writes:\n> > \tRename as 'release stopper'...\n> >> notify fixes(Tom)\n> >> [other items snipped]\n> > \tThese have obviously become show stoppers, since they are now half\n> > implemented, and have to be completed before release. Do we have ETAs on\n> > this stuff?\n> \n> My notify rewrite is not in the tree at all right now. I thought we had\n> agreed (off-list) not to put that change into 6.4, but to postpone it\n> to the next release.\n\n\t*slap forehead*\n\n> > As things stand right now, we are looking at November 1st\n> > for a release date on v6.4...a month late, but not much worse then other\n> > releases :)\n> \n> On the other hand, if release is going to be 11/1 not 10/1, my personal\n> vote is to put the notify changes in. If there are any bugs, that ought\n> to be time enough to shake them out.\n> \n> I can commit those changes tonight if I have the go-ahead. Or I can\n> wait till post-6.4. Your call.\n\n\tGo for it...that will at least get them off the list...\n\n> flock is a release stopper as far as I'm concerned, because the backend\n> *does not compile* on my platform without diking out that code. I agree\n> that it is too late to try to rewrite the feature correctly for 6.4.\n> What say we put in an autoconf test for whether flock exists, and make\n> the new code conditional on that? People without flock would see the\n> same old behavior, which is good enough for now. I will volunteer to\n> make the necessary changes if that strategy is agreed on.\n\n\tMake it so...\n\n\n",
"msg_date": "Mon, 5 Oct 1998 14:29:41 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": ">I'm working on the multi-byte support for win32 env. with help from\n>Hiroshi Inoue who appears in the discussion on some win32 issues.\n>In one or two days, I will post new patches that should make\n>the multi-byte support usable on win32.\n>Note that only Japanese(Shift-JIS) will be supported in the first\n>version.\n\nHere are the patches against the current source tree. I have run the\nregression test on a FreeBSD box with both non-MULTIBYTE and\nMULTIBYTE-enabled, and confirmed that the results are same.\n\nHowever I do not tested on PCs(I don't have access to win). Please let \nme know if the patches break anything on PCs.\n\nAlso please note that the patch for varchar.c is a fix for a nasty bug\nof char(n) types that I introduced and I believe at least this should\nbe applied.\n\nAffected files are:\nsrc/win32.mak\nsrc/bin/psql/win32.mak\nsrc/interfaces/libpq/win32.mak\nsrc/interfaces/libpq/libpqdll.def\nsrc/interfaces/libpq/fe-print.c\nsrc/backend/utils/mb/common.c\nsrc/backend/utils/adt/varchar.c\n\n\nIndex: src/win32.mak\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/win32.mak,v\nretrieving revision 1.2\ndiff -c -r1.2 win32.mak\n*** win32.mak\t1998/08/27 13:25:11\t1.2\n--- win32.mak\t1998/10/05 06:33:48\n***************\n*** 10,19 ****\n NULL=nul\n !ENDIF \n \n ALL: \n cd interfaces\\libpq\n! nmake /f win32.mak\n cd ..\\..\\bin\\psql\n! nmake /f win32.mak\n cd ..\\..\n echo All Win32 parts have been built!\n--- 10,23 ----\n NULL=nul\n !ENDIF \n \n+ !IFDEF\tMULTIBYTE\n+ MAKEMACRO = \"MULTIBYTE=$(MULTIBYTE)\"\n+ !ENDIF\n+ \n ALL: \n cd interfaces\\libpq\n! nmake /f win32.mak $(MAKEMACRO)\n cd ..\\..\\bin\\psql\n! nmake /f win32.mak $(MAKEMACRO)\n cd ..\\..\n echo All Win32 parts have been built!\nIndex: src/bin/psql/win32.mak\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/bin/psql/win32.mak,v\nretrieving revision 1.2\ndiff -c -r1.2 win32.mak\n*** win32.mak\t1998/09/03 02:16:27\t1.2\n--- win32.mak\t1998/10/05 06:33:48\n***************\n*** 29,34 ****\n--- 29,42 ----\n CPP_PROJ=/nologo /ML /W3 /GX /O2 /D \"WIN32\" /D \"NDEBUG\" /D \"_CONSOLE\" /D\\\n \"_MBCS\" /Fp\"$(INTDIR)\\psql.pch\" /YX /Fo\"$(INTDIR)\\\\\" /Fd\"$(INTDIR)\\\\\" /FD /c \\\n /I ..\\..\\include /I ..\\..\\interfaces\\libpq\n+ \n+ !IFDEF\tMULTIBYTE\n+ !IFNDEF MBFLAGS\n+ MBFLAGS=\"-DMULTIBYTE=$(MULTIBYTE)\"\n+ !ENDIF\n+ CPP_PROJ=$(MBFLAGS) $(CPP_PROJ)\n+ !ENDIF\n+ \n CPP_OBJS=.\\Release/\n CPP_SBRS=.\n \nIndex: src/interfaces/libpq/win32.mak\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/interfaces/libpq/win32.mak,v\nretrieving revision 1.3\ndiff -c -r1.3 win32.mak\n*** win32.mak\t1998/09/18 16:46:07\t1.3\n--- win32.mak\t1998/10/05 06:33:48\n***************\n*** 33,39 ****\n \t-@erase \"$(OUTDIR)\\libpq.lib\"\n \t-@erase \"$(OUTDIR)\\libpq.dll\"\n \t-@erase \"$(OUTDIR)\\libpq.res\"\n! \t-@erase \"$(OUTDIR)\\vc*.*\"\n \t-@erase \"$(OUTDIR)\\libpq.pch\"\n \t-@erase \"$(OUTDIR)\\libpqdll.exp\"\n \t-@erase \"$(OUTDIR)\\libpqdll.lib\"\n--- 33,39 ----\n \t-@erase \"$(OUTDIR)\\libpq.lib\"\n \t-@erase \"$(OUTDIR)\\libpq.dll\"\n \t-@erase \"$(OUTDIR)\\libpq.res\"\n! \t-@erase \"vc50.pch\"\n \t-@erase \"$(OUTDIR)\\libpq.pch\"\n \t-@erase \"$(OUTDIR)\\libpqdll.exp\"\n \t-@erase \"$(OUTDIR)\\libpqdll.lib\"\n***************\n*** 44,49 ****\n--- 44,57 ----\n CPP_PROJ=/nologo /ML /W3 /GX /O2 /I \"..\\..\\include\" /D \"NDEBUG\" /D\\\n \"WIN32\" /D \"_WINDOWS\" /Fp\"$(INTDIR)\\libpq.pch\" /YX\\\n /Fo\"$(INTDIR)\\\\\" /Fd\"$(INTDIR)\\\\\" /FD /c \n+ \n+ !IFDEF MULTIBYTE\n+ !IFNDEF\tMBFLAGS\n+ MBFLAGS=\"-DMULTIBYTE=$(MULTIBYTE)\"\n+ !ENDIF\n+ CPP_PROJ = $(CPP_PROJ) $(MBFLAGS)\n+ !ENDIF\n+ \n CPP_OBJS=.\\Release/\n CPP_SBRS=.\n \t\n***************\n*** 57,62 ****\n--- 65,74 ----\n \t\"$(INTDIR)\\fe-lobj.obj\" \\\n \t\"$(INTDIR)\\fe-misc.obj\" \\\n \t\"$(INTDIR)\\fe-print.obj\"\n+ \n+ !IFDEF MULTIBYTE\n+ LIB32_OBJS = $(LIB32_OBJS) $(INTDIR)\\common.obj $(INTDIR)\\wchar.obj $(INTDIR)\\conv.obj\n+ !ENDIF\n \n RSC_PROJ=/l 0x409 /fo\"$(INTDIR)\\libpq.res\"\n \nIndex: src/interfaces/libpq/libpqdll.def\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/interfaces/libpq/libpqdll.def,v\nretrieving revision 1.4\ndiff -c -r1.4 libpqdll.def\n*** libpqdll.def\t1998/10/01 01:40:26\t1.4\n--- libpqdll.def\t1998/10/05 06:33:48\n***************\n*** 63,66 ****\n \tlo_unlink\t\t@ 60\n \tlo_import\t\t@ 61\n \tlo_export\t\t@ 62\n! \tPQresultErrorMessage\t@ 63\n--- 63,67 ----\n \tlo_unlink\t\t@ 60\n \tlo_import\t\t@ 61\n \tlo_export\t\t@ 62\n! \tpgresStatus\t\t@ 63\n! \tPQmblen\t\t\t@ 64\nIndex: src/interfaces/libpq/fe-print.c\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/interfaces/libpq/fe-print.c,v\nretrieving revision 1.13\ndiff -c -r1.13 fe-print.c\n*** fe-print.c\t1998/10/04 20:46:39\t1.13\n--- fe-print.c\t1998/10/05 06:33:49\n***************\n*** 513,519 ****\n \treturn (pg_encoding_mblen(encoding, s));\n }\n \n! #endif\n \n static void\n do_field(PQprintOpt *po, PGresult *res,\n--- 513,529 ----\n \treturn (pg_encoding_mblen(encoding, s));\n }\n \n! #else\n! \n! #ifdef WIN32\n! int\n! PQmblen(unsigned char *s)\n! {\n! }\n! #endif\t/* WIN32 */\n! \n! \n! #endif\t/* MULTIBYTE */\n \n static void\n do_field(PQprintOpt *po, PGresult *res,\nIndex: src/backend/utils/mb/common.c\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/backend/utils/mb/common.c,v\nretrieving revision 1.2\ndiff -c -r1.2 common.c\n*** common.c\t1998/09/01 04:33:19\t1.2\n--- common.c\t1998/10/05 06:33:49\n***************\n*** 4,10 ****\n * Tatsuo Ishii\n * $Id: common.c,v 1.2 1998/09/01 04:33:19 momjian Exp $ */\n \n! #include <stdio.h>\n #include <string.h>\n \n #include \"mb/pg_wchar.h\"\n--- 4,19 ----\n * Tatsuo Ishii\n * $Id: common.c,v 1.2 1998/09/01 04:33:19 momjian Exp $ */\n \n! #include <stdlib.h>\n! \n! #ifdef WIN32\n! #include \"win32.h\"\n! #else\n! #if !defined(NO_UNISTD_H)\n! #include <unistd.h>\n! #endif\n! #endif\n! \n #include <string.h>\n \n #include \"mb/pg_wchar.h\"\nIndex: src/backend/utils/adt/varchar.c\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/backend/utils/adt/varchar.c,v\nretrieving revision 1.41\ndiff -c -r1.41 varchar.c\n*** varchar.c\t1998/09/25 15:51:02\t1.41\n--- varchar.c\t1998/10/05 06:33:50\n***************\n*** 163,169 ****\n #ifdef MULTIBYTE\n \t/* truncate multi-byte string in a way not to break\n \t multi-byte boundary */\n! \tslen = pg_mbcliplen(VARDATA(s), rlen, rlen);\n #else\n \tslen = VARSIZE(s) - VARHDRSZ;\n #endif\n--- 163,173 ----\n #ifdef MULTIBYTE\n \t/* truncate multi-byte string in a way not to break\n \t multi-byte boundary */\n! \tif (VARSIZE(s) > len) {\n! \t\tslen = pg_mbcliplen(VARDATA(s), VARSIZE(s)-VARHDRSZ, rlen);\n! } else {\n! \t\tslen = VARSIZE(s) - VARHDRSZ;\n! }\n #else\n \tslen = VARSIZE(s) - VARHDRSZ;\n #endif\n",
"msg_date": "Tue, 06 Oct 1998 11:05:56 +0900",
"msg_from": "Tatsuo Ishii <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "\"Marc G. Fournier\" <[email protected]> writes:\n>> I can commit those changes tonight if I have the go-ahead. Or I can\n>> wait till post-6.4. Your call.\n\n> \tGo for it...that will at least get them off the list...\n\nOK, the NOTIFY rewrite is checked in. We'll see what breaks, if\nanything.\n\n>> flock is a release stopper as far as I'm concerned, because the backend\n>> *does not compile* on my platform without diking out that code.\n\nActually, it looks like Vadim replaced the flock() call with fcntl() a\nfew weeks ago, and I'd not noticed because I had a locally modified copy\nof pqcomm.c. I don't know if fcntl(F_SETLK) is any more portable than\nflock() --- it compiles on my platform, where flock() didn't, but that\nproves little. So I went ahead and put in an autoconf test, only\nchecking for fcntl(F_SETLK) rather than flock(). I still think the\nprocess-pid-in-a-textfile approach to locking is safer, but we can\nleave that for the next release.\n\nThat's two items off the must-fix list and onto the are-there-bugs?\nlist...\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 05 Oct 1998 22:49:36 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Open 6.4 items "
},
{
"msg_contents": "Applied.\n\n> >I'm working on the multi-byte support for win32 env. with help from\n> >Hiroshi Inoue who appears in the discussion on some win32 issues.\n> >In one or two days, I will post new patches that should make\n> >the multi-byte support usable on win32.\n> >Note that only Japanese(Shift-JIS) will be supported in the first\n> >version.\n> \n> Here are the patches against the current source tree. I have run the\n> regression test on a FreeBSD box with both non-MULTIBYTE and\n> MULTIBYTE-enabled, and confirmed that the results are same.\n> \n> However I do not tested on PCs(I don't have access to win). Please let \n> me know if the patches break anything on PCs.\n> \n> Also please note that the patch for varchar.c is a fix for a nasty bug\n> of char(n) types that I introduced and I believe at least this should\n> be applied.\n> \n> Affected files are:\n> src/win32.mak\n> src/bin/psql/win32.mak\n> src/interfaces/libpq/win32.mak\n> src/interfaces/libpq/libpqdll.def\n> src/interfaces/libpq/fe-print.c\n> src/backend/utils/mb/common.c\n> src/backend/utils/adt/varchar.c\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Mon, 5 Oct 1998 23:03:44 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Open 6.4 items"
}
] |
[
{
"msg_contents": "Hi, there's a small bug in inv_fetchup().\nIn the following do-while loop, res is freed at the very\nend of the loop. However, after getting out of the loop,\nit gets freed again. Not good. Possible fix would be\nchanging the pfree(res) in the loop to:\n\n if (tuple == (HeapTuple)NULL) pfree(res);\n\n------------------- inv_api.c ----------------------\n\t\tdo\n\t\t{\n\t\t\tres = index_getnext(obj_desc->iscan, ForwardScanDirection);\n\n\t\t\tif (res == (RetrieveIndexResult) NULL)\n\t\t\t{\n\t\t\t\tItemPointerSetInvalid(&(obj_desc->htid));\n\t\t\t\treturn (HeapTuple) NULL;\n\t\t\t}\n\n\t\t\t/*\n\t\t\t * For time travel, we need to use the actual time qual here,\n\t\t\t * rather that NowTimeQual. We currently have no way to pass\n\t\t\t * a time qual in.\n\t\t\t *\n\t\t\t * This is now valid for snapshot !!! And should be fixed in some\n\t\t\t * way...\t- vadim 07/28/98\n\t\t\t *\n\t\t\t */\n\n\t\t\ttuple = heap_fetch(obj_desc->heap_r, SnapshotNow,\n\t\t\t\t\t\t\t &res->heap_iptr, buffer);\n\t\t\tpfree(res);\n\t\t} while (tuple == (HeapTuple) NULL);\n\n\t\t/* remember this tid -- we may need it for later reads/writes */\n\t\tItemPointerCopy(&(res->heap_iptr), &(obj_desc->htid));\n\t\tpfree(res);\n\t}\n------------------- inv_api.c -----------------\n--\nTatsuo Ishii\[email protected]\n\n",
"msg_date": "Sat, 3 Oct 1998 23:30:45 +0900",
"msg_from": "[email protected] (Tatsuo Ishii)",
"msg_from_op": true,
"msg_subject": "too many pfree in large object"
},
{
"msg_contents": "> Hi, there's a small bug in inv_fetchup().\n> In the following do-while loop, res is freed at the very\n> end of the loop. However, after getting out of the loop,\n> it gets freed again. Not good. Possible fix would be\n> changing the pfree(res) in the loop to:\n> \n> if (tuple == (HeapTuple)NULL) pfree(res);\n> \n> ------------------- inv_api.c ----------------------\n> \t\t\ttuple = heap_fetch(obj_desc->heap_r, SnapshotNow,\n> \t\t\t\t\t\t\t &res->heap_iptr, buffer);\n> \t\t\tpfree(res);\n> \t\t} while (tuple == (HeapTuple) NULL);\n> \n> \t\t/* remember this tid -- we may need it for later reads/writes */\n> \t\tItemPointerCopy(&(res->heap_iptr), &(obj_desc->htid));\n> \t\tpfree(res);\n> \t}\n\nThanks. Fixed. I changed the line to read:\n\n ItemPointerCopy(&tuple->t_ctid, &obj_desc->htid);\n\nThis way, I am getting it from the heap, not from the index tuple, and\nremoved the duplicate pfree().\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Mon, 5 Oct 1998 23:58:05 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] too many pfree in large object"
}
] |
[
{
"msg_contents": "> > Added Files:\n> > README.ipaddr\n> > Log Message:\n> > Integrate new IP type from Tom Ivar Helbekkmo.\n> \n> I'm going to take this and move it to the datatypes docs, OK?\n> \n> - Tom\n> \n\nGood.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n",
"msg_date": "Sat, 3 Oct 1998 10:54:15 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [COMMITTERS] 'pgsql/doc README.ipaddr'"
}
] |
[
{
"msg_contents": "I'm trying to fix up initlocation to accept an environment variable as\nan input parameter (in addition to the absolute path name it already\naccepts). \n\nI'd like to be able to say:\n\n setenv PGDATA2 /home/postgres/data\n initlocation PGDATA2\n\nHowever, initlocation sets a local variable PGALTDIR to the input\nargument as the program starts, and I'm fumbling around trying to\nevaluate the contents of PGALTDIR as an environment variable.\n\n(So, I need PGALTDIR -> PGDATA2 -> /home/postgres/data, and to assign\nthe path name back to a local variable.)\n\nAnyone know how to do this? Please??\n\nTIA\n\n - Tom\n",
"msg_date": "Sat, 03 Oct 1998 18:58:04 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "need help with csh"
}
] |
[
{
"msg_contents": "Hi all\n\nIs it too late to add a feature to pg_dump for 6.4??\n\nI just spent most of the day learning pg_dump and modifing it so it would\ndump views also.\n\nThis is the first time I have ever contributed any code changes, so I'm\nnot sure of how to submit it.\n\nThe diff's and a readme as a tgz file are attached.\n\nThanks\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!",
"msg_date": "Sat, 3 Oct 1998 16:25:30 -0400 (EDT)",
"msg_from": "Terry Mackintosh <[email protected]>",
"msg_from_op": true,
"msg_subject": "Dumping of views -- done!"
},
{
"msg_contents": "> Hi all\n> \n> Is it too late to add a feature to pg_dump for 6.4??\n> \n> I just spent most of the day learning pg_dump and modifing it so it would\n> dump views also.\n> \n> This is the first time I have ever contributed any code changes, so I'm\n> not sure of how to submit it.\n> \n> The diff's and a readme as a tgz file are attached.\n> \n> Thanks\n> Terry Mackintosh <[email protected]> http://www.terrym.com\n> sysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nI though we dumped views already. If not, I am inclinded to apply it,\nbut want to see what others say. I thought Jan was working on this, or\nwas that just rewrite rules? Are you using his code? I know there is\nan item on the Open Items list that talks about this:\n\n\thave psql dump out rules text with new function\n\nIs that what this is?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sat, 3 Oct 1998 21:06:44 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "Hi Bruce\n\nOn Sat, 3 Oct 1998, Bruce Momjian wrote:\n\n> > Hi all\n> > \n> > Is it too late to add a feature to pg_dump for 6.4??\n\n> I though we dumped views already. If not, I am inclinded to apply it,\n\nOH, ... well, if it does, I sure would like to know how, I missed it.\n\n> but want to see what others say. I thought Jan was working on this, or\n> was that just rewrite rules? Are you using his code? I know there is\n\nI tryed to stay as much in style with the code that was there as possible.\n\n> an item on the Open Items list that talks about this:\n> \n> \thave psql dump out rules text with new function\n> Is that what this is?\n\nI'm not sure that is the same, what my change does is add:\nCREATE VIEW viewname AS select clause......;\nto the dump file.\n\nI did notice that if used with '-z' it also puts the revoke grant\nstatments there too. Not my doing, just nice.\n\nAt this time there are no provisions (by me any way) to dump any\nassociated update/delete rules on the view. Unless, like -z above, it\nalready works?\n\nTo the best of my knowlage this feature does not yet exist, if any one\nelse is working on it, I don't know.\n\n> -- \n> Bruce Momjian | http://www.op.net/~candle\n> [email protected] | (610) 853-3000\n> + If your life is a hard drive, | 830 Blythe Avenue\n> + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n> \n\nHave a great day\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n",
"msg_date": "Sat, 3 Oct 1998 22:49:14 -0400 (EDT)",
"msg_from": "Terry Mackintosh <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "Hi Bruce\n \nOn Sat, 3 Oct 1998, Bruce Momjian wrote:\n\n> I though we dumped views already. If not, I am inclinded to apply it,\n\nYour comment gave me cause to go read the pg_dump man page again, here is\nwhat it has to say:\n\n rules and views\n pg_dump does not understand user-defined rules and\n views and will fail to dump them properly. (This\n is due to the fact that rules are stored as plans\n in the catalogs and not textually)\n\n\nNow, I did not (yet) do any thing about rules, only views.\n\nHave a great day\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n",
"msg_date": "Sat, 3 Oct 1998 23:11:04 -0400 (EDT)",
"msg_from": "Terry Mackintosh <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "On Sat, 3 Oct 1998, Bruce Momjian wrote:\n\n> > Hi all\n> > \n> > Is it too late to add a feature to pg_dump for 6.4??\n> > \n> > I just spent most of the day learning pg_dump and modifing it so it would\n> > dump views also.\n> > \n> > This is the first time I have ever contributed any code changes, so I'm\n> > not sure of how to submit it.\n> > \n> > The diff's and a readme as a tgz file are attached.\n> > \n> > Thanks\n> > Terry Mackintosh <[email protected]> http://www.terrym.com\n> > sysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n> \n> I though we dumped views already. If not, I am inclinded to apply it,\n> but want to see what others say. \n\nAs this does not affect, in any way, shape or form, the backend...I'd be\ninclined to say go for it also...\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Mon, 5 Oct 1998 00:56:08 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "Terry Mackintosh wrote:\n>\n> Hi Bruce\n>\n> On Sat, 3 Oct 1998, Bruce Momjian wrote:\n>\n> > > Hi all\n> > >\n> > > Is it too late to add a feature to pg_dump for 6.4??\n>\n> > I though we dumped views already. If not, I am inclinded to apply it,\n>\n> OH, ... well, if it does, I sure would like to know how, I missed it.\n>\n\n No we didn't. I just mentioned that it would be possible with\n the new rule utility functions.\n\n> > but want to see what others say. I thought Jan was working on this, or\n> > was that just rewrite rules? Are you using his code? I know there is\n>\n> I tryed to stay as much in style with the code that was there as possible.\n>\n\n Took a look at the patch. Terry's using the new system view\n pg_views for it. But...\n\n> > an item on the Open Items list that talks about this:\n> >\n> > have psql dump out rules text with new function\n> > Is that what this is?\n>\n> I'm not sure that is the same, what my change does is add:\n> CREATE VIEW viewname AS select clause......;\n> to the dump file.\n>\n> I did notice that if used with '-z' it also puts the revoke grant\n> statments there too. Not my doing, just nice.\n>\n> At this time there are no provisions (by me any way) to dump any\n> associated update/delete rules on the view. Unless, like -z above, it\n> already works?\n>\n> To the best of my knowlage this feature does not yet exist, if any one\n> else is working on it, I don't know.\n\n Not me. But I think I should assist since I'm one of the\n rulegurus here on the list.\n\n IMHO the better way would be to dump views as regular CREATE\n TABLE statements and just omit the data (views have none).\n After all tables are restored and all functions, operators\n and operator classes are created, we should dump rules using\n the pg_rules view.\n\n The checks in pg_dump if a table is a view rely on the\n existence of a rule named _RETtablename. I'll add another\n check to rewriteDefine.c making sure that a rewrite rule ON\n SELECT follows this convention (rules ON SELECT are now\n restricted to view rules at all so it would make sense for\n me).\n\n I'll add double quotes around identifiers in the output of\n pg_get_ruledef() and pg_get_viewdef().\n\n Terry, would you mind to implement dumping of views again in\n the above mentioned way? This would cover rewrite rules at\n all. A good place to dump rules would be where triggers get\n dumped (if they are dumped at all). Dumping all rules would\n automagically turn the formerly created tables into views.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 5 Oct 1998 10:35:27 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "Hi Jan\n\nOn Mon, 5 Oct 1998, Jan Wieck wrote:\n\n> Took a look at the patch. Terry's using the new system view\n> pg_views for it. But...\n> \n> IMHO the better way would be to dump views as regular CREATE\n> TABLE statements and just omit the data (views have none).\n> After all tables are restored and all functions, operators\n> and operator classes are created, we should dump rules using\n> the pg_rules view.\n> \nWell, if you see need to do it differant, you most certainly know more\nabout it then me, so sure.\n\nI am confused though ....\nHow is \"CREATE TABLE ....\" going to become \"CREATE VIEW .....\"????\n\nNote also, that the way I did it kept in mind that some day we will have\nouter joins, at which time most of the code I put in getTables() can come\nout, and just make a very minor change to the main query to get the view\ninfo all at one time.\n\nIf there is a cleaner way to do it though, then that is the way to do it,\nI just do not understand what it is you said.\n\nI am lacking time today to really think about it, but will take time\ntonight maybe to look at pg_rules and such.\n\nHave a great day\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n",
"msg_date": "Mon, 5 Oct 1998 09:46:18 -0400 (EDT)",
"msg_from": "Terry Mackintosh <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "> Hi all\n> \n> Is it too late to add a feature to pg_dump for 6.4??\n> \n> I just spent most of the day learning pg_dump and modifing it so it would\n> dump views also.\n> \n> This is the first time I have ever contributed any code changes, so I'm\n> not sure of how to submit it.\n> \n> The diff's and a readme as a tgz file are attached.\n\nApplied, and I have updated the manual pages, and the sgml docs. Yoo,\nwhooo.\n\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Mon, 5 Oct 1998 23:10:53 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "> Applied, and I have updated the manual pages, and the sgml docs. Yoo,\n> whooo.\n\nThat sgml is pretty exciting, eh? Now calm down and get back to work :)\n\n - Tom\n",
"msg_date": "Tue, 06 Oct 1998 04:40:06 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "> > Applied, and I have updated the manual pages, and the sgml docs. Yoo,\n> > whooo.\n> \n> That sgml is pretty exciting, eh? Now calm down and get back to work :)\n\nI actually have a hardcover SGML book, called \"Readme.1st: SGML for\nWriters and Editors\".\n\nI have not really looked through it, but it was only $7. It is a\nPrentice Hall book, 1996.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Wed, 7 Oct 1998 20:08:37 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "> I actually have a hardcover SGML book, called \"Readme.1st: SGML for\n> Writers and Editors\".\n\nWithout having seen the book, I will make the comment that the SGML\nbooks I've run into (not very many, but that won't stop me from\ngeneralizing :) tend to tell you what you could do with SGML, rather\nthan telling you how to do something with the particular implementation\nwe are using (DocBook).\n\nThe fellow who wrote and maintains the \"Modular Style Sheets\", Norm\nWalsh, is in the middle of writing a book (Nutshell? can't remember) on\nDocBook, and presumably his style sheets.\n\nThe most useful single information source I've found so far, other than\nstealing markup from other parts of the documentation, is the DocBook\nelement dictionary at the O'Reilly site:\n\n http://www.ora.com/homepages/dtdparse/docbook/3.0/\n\nI hope this is mentioned in our docguide...\n\n - Tom\n",
"msg_date": "Thu, 08 Oct 1998 07:26:18 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "\nI haven't seen any comments on this yet...am tempted to add it in,\nconsidering we have another 2 weeks before release, and it doesn't affect\nthe backend...\n\nAnyone want to comment? If I hear nothing by, say, Monday morning, I'll\napply these...\n\n\n\n\n\nOn Sat, 3 Oct 1998, Terry Mackintosh wrote:\n\n> Hi all\n> \n> Is it too late to add a feature to pg_dump for 6.4??\n> \n> I just spent most of the day learning pg_dump and modifing it so it would\n> dump views also.\n> \n> This is the first time I have ever contributed any code changes, so I'm\n> not sure of how to submit it.\n> \n> The diff's and a readme as a tgz file are attached.\n> \n> Thanks\n> Terry Mackintosh <[email protected]> http://www.terrym.com\n> sysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n> \n> Proudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n> -------------------------------------------------------------------\n> Success Is A Choice ... book by Rick Patino, get it, read it!\n> \n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Mon, 12 Oct 1998 15:53:59 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "Hi Marc\n\nOn Mon, 12 Oct 1998, Marc G. Fournier wrote:\n\n> \n> I haven't seen any comments on this yet...am tempted to add it in,\n> considering we have another 2 weeks before release, and it doesn't affect\n> the backend...\n\nHugh?-)\nThere were lots, any way it prompted Jan to do some major work that did\nthis and much more, so it is all done as far as I know.\n\nThanks for asking though, have a great night.\n\n> Anyone want to comment? If I hear nothing by, say, Monday morning, I'll\n> apply these...\n> \n> On Sat, 3 Oct 1998, Terry Mackintosh wrote:\n> \n> > Hi all\n> > \n> > Is it too late to add a feature to pg_dump for 6.4??\n\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n",
"msg_date": "Mon, 12 Oct 1998 21:26:07 -0400 (EDT)",
"msg_from": "Terry Mackintosh <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": ">\n>\n> I haven't seen any comments on this yet...am tempted to add it in,\n> considering we have another 2 weeks before release, and it doesn't affect\n> the backend...\n>\n> Anyone want to comment? If I hear nothing by, say, Monday morning, I'll\n> apply these...\n>\n\n Stop - don't apply.\n\n Dumping of views is already done in the current CVS. Pg_dump\n now outputs CREATE TABLE for view relations too, but no data.\n At the end of the schema (after data for real tables has been\n filled in) it outputs commands to recreate ALL rewrite rules,\n silently turning the tables into the views they where at dump\n time.\n\n>\n>\n>\n>\n> On Sat, 3 Oct 1998, Terry Mackintosh wrote:\n>\n> > Hi all\n> >\n> > Is it too late to add a feature to pg_dump for 6.4??\n> >\n> > I just spent most of the day learning pg_dump and modifing it so it would\n> > dump views also.\n> >\n> > This is the first time I have ever contributed any code changes, so I'm\n> > not sure of how to submit it.\n> >\n> > The diff's and a readme as a tgz file are attached.\n> >\n> > Thanks\n> > Terry Mackintosh <[email protected]> http://www.terrym.com\n> > sysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n> >\n> > Proudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n> > -------------------------------------------------------------------\n> > Success Is A Choice ... book by Rick Patino, get it, read it!\n> >\n>\n> Marc G. Fournier [email protected]\n> Systems Administrator @ hub.org\n> scrappy@{postgresql|isc}.org ICQ#7615664\n>\n>\n>\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Tue, 13 Oct 1998 09:10:31 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
},
{
"msg_contents": "On Tue, 13 Oct 1998, Jan Wieck wrote:\n\n> >\n> >\n> > I haven't seen any comments on this yet...am tempted to add it in,\n> > considering we have another 2 weeks before release, and it doesn't affect\n> > the backend...\n> >\n> > Anyone want to comment? If I hear nothing by, say, Monday morning, I'll\n> > apply these...\n> >\n> \n> Stop - don't apply.\n> \n> Dumping of views is already done in the current CVS. Pg_dump\n> now outputs CREATE TABLE for view relations too, but no data.\n> At the end of the schema (after data for real tables has been\n> filled in) it outputs commands to recreate ALL rewrite rules,\n> silently turning the tables into the views they where at dump\n> time.\n\n\tD'oh...I hate losing a hard drive :( catching up takes soooo long\nto do :(\n\n> \n> >\n> >\n> >\n> >\n> > On Sat, 3 Oct 1998, Terry Mackintosh wrote:\n> >\n> > > Hi all\n> > >\n> > > Is it too late to add a feature to pg_dump for 6.4??\n> > >\n> > > I just spent most of the day learning pg_dump and modifing it so it would\n> > > dump views also.\n> > >\n> > > This is the first time I have ever contributed any code changes, so I'm\n> > > not sure of how to submit it.\n> > >\n> > > The diff's and a readme as a tgz file are attached.\n> > >\n> > > Thanks\n> > > Terry Mackintosh <[email protected]> http://www.terrym.com\n> > > sysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n> > >\n> > > Proudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n> > > -------------------------------------------------------------------\n> > > Success Is A Choice ... book by Rick Patino, get it, read it!\n> > >\n> >\n> > Marc G. Fournier [email protected]\n> > Systems Administrator @ hub.org\n> > scrappy@{postgresql|isc}.org ICQ#7615664\n> >\n> >\n> >\n> \n> \n> Jan\n> \n> --\n> \n> #======================================================================#\n> # It's easier to get forgiveness for being wrong than for being right. #\n> # Let's break this rule - forgive me. #\n> #======================================== [email protected] (Jan Wieck) #\n> \n> \n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Tue, 13 Oct 1998 10:33:15 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Dumping of views -- done!"
}
] |
[
{
"msg_contents": "Attached is the IP type test program supplied by Tom Ivar Helbekkmo. \nThis can be used for testing, and for integration into the regression\ntests.\n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n\n--\n--\tA quick test of the IP address code\n--\n--\t$Id: test.sql,v 1.2 1998/09/08 12:11:34 tih Exp $\n--\n\n-- temporary table:\ncreate table addresses (address ipaddr);\n\n-- sample data from two subnets:\ninsert into addresses values ('158.37.96.15');\ninsert into addresses values ('158.37.96.16');\ninsert into addresses values ('158.37.96.17');\ninsert into addresses values ('158.37.97.15');\ninsert into addresses values ('158.37.97.16');\ninsert into addresses values ('158.37.97.17');\ninsert into addresses values ('158.37.98.15');\ninsert into addresses values ('158.37.98.16');\ninsert into addresses values ('158.37.98.17');\ninsert into addresses values ('158.37.96.150');\ninsert into addresses values ('158.37.96.160');\ninsert into addresses values ('158.37.96.170');\ninsert into addresses values ('158.37.97.150');\ninsert into addresses values ('158.37.97.160');\ninsert into addresses values ('158.37.97.170');\ninsert into addresses values ('158.37.98.150');\ninsert into addresses values ('158.37.98.160');\ninsert into addresses values ('158.37.98.170');\n\n-- show them all:\nselect * from addresses;\n\n-- select the ones in subnet 96:\nselect * from addresses where address << '158.37.96.0/24';\n\n-- select the ones not in subnet 96:\nselect * from addresses where not address << '158.37.96.0/24';\n\n-- select the ones in subnet 97:\nselect * from addresses where address << '158.37.97.0/24';\n\n-- select the ones not in subnet 97:\nselect * from addresses where not address << '158.37.97.0/24';\n\n-- select the ones in subnet 96 or 97, sorted:\nselect * from addresses where address << '158.37.96.0/23'\n\torder by address;\n\n-- now some networks:\ncreate table networks (network ipaddr);\n\n-- now the subnets mentioned above:\ninsert into networks values ('158.37.96.0/24');\ninsert into networks values ('158.37.97.0/24');\ninsert into networks values ('158.37.98.0/24');\n\n-- select matching pairs of addresses and containing nets:\nselect address, network from addresses, networks\n\twhere address << network;\n\n-- tidy up:\ndrop table addresses;\ndrop table networks;\n\n--\n--\teof\n--",
"msg_date": "Sat, 3 Oct 1998 21:02:21 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "IP test program"
},
{
"msg_contents": "> Attached is the IP type test program supplied by Tom Ivar Helbekkmo.\n> This can be used for testing, and for integration into the regression\n> tests.\n\nI'll look at it if no one else picks it up. Just remember to cut me some\nslack when I'm running up against the release deadline :)\n\n - Tom\n",
"msg_date": "Sun, 04 Oct 1998 04:18:25 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: IP test program"
}
] |
[
{
"msg_contents": "Recent additions to the PostgreSQL 6.4 tree broke the UnixWare 7.x port. Here \nare the patches to fix them:\n\n1. In 'src/backend/utils/adt/ip.c', the functions 'ipaddr_eq()' and\n 'ipaddr_ge()' were used before being defined. I reordered the functions so\n that they were defined before being used.\n\n2. In 'src/include/utils/mac.h', I included the file 'sys/bitypes.h' if the\n code is being compiled by a USL compiler (i.e. UnixWare). This is needed\n to define the 'u_int32_t' type.\n\n3. In 'src/include/c.h', I add a check to see if 'bool' had been defined as a\n macro before attempting to typeset it. In some instances, 'bool' is\n defined as a macro that expands to 'char'. The UnixWare compiler complains\n if you try to compile \"typeset char char;\".\n\n4. In 'src/pl/plpgsql/src/pl_exec.c', there is a construct that is incorrect.\n This construct is '((int4)(var->value))--' and '((int4)(var->value))++'.\n What this is saying is to take result of the expression '(var->value)',\n cast it as an 'int4', and then increment it. This can not work as the\n result of the expression is not an 'LVALUE' and thus can not be the target\n of the '--' or the '++' operator. What is needed is '(int4)var->value--'\n or '(int4)var->value++' [assuming the desired result is to decrement(in-\n crement) 'var->value' as if it were an 'int4'].\n\n5. In 'src/test/regress/checkresults', the use of the '-u' option is assuming\n the use of GNU diff.\n\nThere were some additional portability problems with the make file for plpgsql \nthat I do not address with these patches. I will supply a Makefile.in for \nplpgsql that will address these problems shortly.\n\nThe UnixWare 7.x port now passes the regression tests except for differences \nin the some error messages and floating point precision.\n\nBTW: The plpgsql language seems to working well, at least for the simple\n things that I am currently using them for.\n\n Thanks Jan for the good work on PL/pgsql!\n\n\n\n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 |",
"msg_date": "Sun, 04 Oct 1998 05:30:36 -0400",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "UnixWare 7.x port updates."
},
{
"msg_contents": "> The UnixWare 7.x port now passes the regression tests except for \n> differences in the some error messages and floating point precision.\n\nGot the report. Will assume that your patches will make it into the\ndistribution, and will update the porting information for your platform.\nCould you please let us know if anything gets broken, or if your patches\ndon't get in completely so you continue to have a broken port?\n\nTIA\n\n - Tom\n",
"msg_date": "Sun, 04 Oct 1998 14:58:32 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] UnixWare 7.x port updates."
},
{
"msg_contents": "Hi all,\nif I ./configure with tcl then make reports:\n\nmake[2]: Entering directory `/usr/local/pgsql/src/pl/tcl'\n'cc -shared' -o pltcl.so pltcl.o '-ltcl8.0' '-L/usr/lib -ltcl8.0' '-ldl\n-lieee -lm'\nmake[2]: cc -shared: Command not found\nmake[2]: *** [pltcl.so] Error 127\n\n-Egon\n\n",
"msg_date": "Sun, 4 Oct 1998 17:35:34 +0200 (MET DST)",
"msg_from": "Egon Schmid <[email protected]>",
"msg_from_op": false,
"msg_subject": "tcl support broken"
},
{
"msg_contents": "> Recent additions to the PostgreSQL 6.4 tree broke the UnixWare 7.x port. Here \n> are the patches to fix them:\n> \n> 1. In 'src/backend/utils/adt/ip.c', the functions 'ipaddr_eq()' and\n> 'ipaddr_ge()' were used before being defined. I reordered the functions so\n> that they were defined before being used.\n\nI was not including builtin.h as I should have. Fixed now.\n\n> \n> 2. In 'src/include/utils/mac.h', I included the file 'sys/bitypes.h' if the\n> code is being compiled by a USL compiler (i.e. UnixWare). This is needed\n> to define the 'u_int32_t' type.\n\nI have changed u_int32_t to unsigned int. No reason to use u_int32_t.\n\n> \n> 3. In 'src/include/c.h', I add a check to see if 'bool' had been defined as a\n> macro before attempting to typeset it. In some instances, 'bool' is\n> defined as a macro that expands to 'char'. The UnixWare compiler complains\n> if you try to compile \"typeset char char;\".\n> \n> 4. In 'src/pl/plpgsql/src/pl_exec.c', there is a construct that is incorrect.\n> This construct is '((int4)(var->value))--' and '((int4)(var->value))++'.\n> What this is saying is to take result of the expression '(var->value)',\n> cast it as an 'int4', and then increment it. This can not work as the\n> result of the expression is not an 'LVALUE' and thus can not be the target\n> of the '--' or the '++' operator. What is needed is '(int4)var->value--'\n> or '(int4)var->value++' [assuming the desired result is to decrement(in-\n> crement) 'var->value' as if it were an 'int4'].\n> \n> 5. In 'src/test/regress/checkresults', the use of the '-u' option is assuming\n> the use of GNU diff.\n\nI will apply the rest.\n\n> \n> There were some additional portability problems with the make file for plpgsql \n> that I do not address with these patches. I will supply a Makefile.in for \n> plpgsql that will address these problems shortly.\n> \n> The UnixWare 7.x port now passes the regression tests except for differences \n> in the some error messages and floating point precision.\n> \n> BTW: The plpgsql language seems to working well, at least for the simple\n> things that I am currently using them for.\n> \n> Thanks Jan for the good work on PL/pgsql!\n> \nContent-Description: uw7.patch\n\n[Attachment, skipping...]\n\n> ____ | Billy G. Allie | Domain....: [email protected]\n> | /| | 7436 Hartwell | Compuserve: 76337,2061\n> |-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n> |/ |LLIE | (313) 582-1540 | \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 4 Oct 1998 11:35:52 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] UnixWare 7.x port updates."
},
{
"msg_contents": "> Recent additions to the PostgreSQL 6.4 tree broke the UnixWare 7.x port. Here \n> are the patches to fix them:\n> \n> 1. In 'src/backend/utils/adt/ip.c', the functions 'ipaddr_eq()' and\n> 'ipaddr_ge()' were used before being defined. I reordered the functions so\n> that they were defined before being used.\n> \n> 2. In 'src/include/utils/mac.h', I included the file 'sys/bitypes.h' if the\n> code is being compiled by a USL compiler (i.e. UnixWare). This is needed\n> to define the 'u_int32_t' type.\n\nFixed.\n\n> \n> 3. In 'src/include/c.h', I add a check to see if 'bool' had been defined as a\n> macro before attempting to typeset it. In some instances, 'bool' is\n> defined as a macro that expands to 'char'. The UnixWare compiler complains\n> if you try to compile \"typeset char char;\".\n> \n> 4. In 'src/pl/plpgsql/src/pl_exec.c', there is a construct that is incorrect.\n> This construct is '((int4)(var->value))--' and '((int4)(var->value))++'.\n> What this is saying is to take result of the expression '(var->value)',\n> cast it as an 'int4', and then increment it. This can not work as the\n> result of the expression is not an 'LVALUE' and thus can not be the target\n> of the '--' or the '++' operator. What is needed is '(int4)var->value--'\n> or '(int4)var->value++' [assuming the desired result is to decrement(in-\n> crement) 'var->value' as if it were an 'int4'].\n> \n> 5. In 'src/test/regress/checkresults', the use of the '-u' option is assuming\n> the use of GNU diff.\n\nThe rest are applied.\n\n> \n> There were some additional portability problems with the make file for plpgsql \n> that I do not address with these patches. I will supply a Makefile.in for \n> plpgsql that will address these problems shortly.\n> \n> The UnixWare 7.x port now passes the regression tests except for differences \n> in the some error messages and floating point precision.\n> \n> BTW: The plpgsql language seems to working well, at least for the simple\n> things that I am currently using them for.\n> \n> Thanks Jan for the good work on PL/pgsql!\n> \nContent-Description: uw7.patch\n\n[Attachment, skipping...]\n\n> ____ | Billy G. Allie | Domain....: [email protected]\n> | /| | 7436 Hartwell | Compuserve: 76337,2061\n> |-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n> |/ |LLIE | (313) 582-1540 | \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 4 Oct 1998 11:39:03 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] UnixWare 7.x port updates."
}
] |
[
{
"msg_contents": "unsubscribr [email protected]\n\n",
"msg_date": "Sun, 04 Oct 1998 15:13:57 +0100",
"msg_from": "Sergio Brandano <[email protected]>",
"msg_from_op": true,
"msg_subject": "None"
}
] |
[
{
"msg_contents": "\nunsubscribe [email protected]\n\n\n",
"msg_date": "Sun, 04 Oct 1998 15:15:12 +0100",
"msg_from": "Sergio Brandano <[email protected]>",
"msg_from_op": true,
"msg_subject": "None"
}
] |
[
{
"msg_contents": "... it does not build on HPUX, nor on any other platform that requires\nspecialized switches to build shared libraries. I recommend the author\nconsult the makefile for libpq, and transpose the necessary fooling\naround into plpgsql. (On HPUX it fails for lack of -fPIC in the compile\ncommands for the pl/plpgsql/src subdirectory; there might also be\nproblems in the link command, but the compile commands are certainly\nwrong.)\n\nWe have enough different shared libraries now that it'd be a good idea\nto try to eliminate most of that cruft from the individual module\nmakefiles, replacing it with definitions made in the port makefiles.\nBut it's a bit too late in the 6.4 beta cycle to try to do that for 6.4,\nI think --- the testing problem looms too large. Something for next time.\n\nBTW there is also a configuration check that needs to be applied: the\npl/tcl subdirectory should be built only if a *shared library* version\nof Tcl is available. I happen to have only a static version of libtcl.a\ninstalled, which is fine for pgtclsh, but pl/tcl tries to link with it\nand falls over...\n\nIf this stuff can't be fixed properly before 6.4 release, I suggest\nmaking plpgsql an optional component that is only built if suitable\nswitches are provided to configure.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sun, 04 Oct 1998 16:00:10 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "plpgsql is not ready for prime time"
},
{
"msg_contents": "> We have enough different shared libraries now that it'd be a good idea\n> to try to eliminate most of that cruft from the individual module\n> makefiles, replacing it with definitions made in the port makefiles.\n> But it's a bit too late in the 6.4 beta cycle to try to do that for 6.4,\n> I think --- the testing problem looms too large. Something for next time.\n\nI had a patch to do just this earlier this year for 6.4, but I changed\njobs before I could submit it, then other patches were applied that made\nmy patch fail. I just finally got access to a machine to redo my patch,\nbut it wasn't in time for 6.4 beta. I will have it in the first month or\nso after 6.4 is released...\n\ndarrenk\n\n",
"msg_date": "Sun, 4 Oct 1998 16:12:37 -0400",
"msg_from": "\"Stupor Genius\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] plpgsql is not ready for prime time"
},
{
"msg_contents": "Tom Lane <[email protected]> wrote:\n> ... it does not build on HPUX, nor on any other platform that requires\n> specialized switches to build shared libraries. I recommend the author\n> consult the makefile for libpq, and transpose the necessary fooling\n> around into plpgsql. (On HPUX it fails for lack of -fPIC in the compile\n> commands for the pl/plpgsql/src subdirectory; there might also be\n> problems in the link command, but the compile commands are certainly\n> wrong.)\n\nI have made change to configure/configure.in and added a Makefile.in to help \nwith the problem (I had problems compiling for UnixWare). I will be testing \n[the UnixWare part of] the patches over the new few days. I will submit the \nafter I complete prlimiary testing. I will also mail the patch to anyone else \nwho would like to test for a specific port.\n\n> We have enough different shared libraries now that it'd be a good idea\n> to try to eliminate most of that cruft from the individual module\n> makefiles, replacing it with definitions made in the port makefiles.\n> But it's a bit too late in the 6.4 beta cycle to try to do that for 6.4,\n> I think --- the testing problem looms too large. Something for next time.\n\nI aggree that the shared library handling needs to be standardized and the \nport specific information moved to the port specific Makefile.\n\n> \n> BTW there is also a configuration check that needs to be applied: the\n> pl/tcl subdirectory should be built only if a *shared library* version\n> of Tcl is available. I happen to have only a static version of libtcl.a\n> installed, which is fine for pgtclsh, but pl/tcl tries to link with it\n> and falls over...\n\nHmmmm.... I don't now autoconf well enough (yet) to handle this change.\n\n> [...]\n> \t\t\tregards, tom lane\n> \n\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Mon, 05 Oct 1998 01:31:49 -0400",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] plpgsql is not ready for prime time "
},
{
"msg_contents": "Billy G. Allie wrote:\n>\n> Tom Lane <[email protected]> wrote:\n> > ... it does not build on HPUX, nor on any other platform that requires\n> > specialized switches to build shared libraries. I recommend the author\n> > consult the makefile for libpq, and transpose the necessary fooling\n> > around into plpgsql. (On HPUX it fails for lack of -fPIC in the compile\n> > commands for the pl/plpgsql/src subdirectory; there might also be\n> > problems in the link command, but the compile commands are certainly\n> > wrong.)\n>\n> I have made change to configure/configure.in and added a Makefile.in to help\n> with the problem (I had problems compiling for UnixWare). I will be testing\n> [the UnixWare part of] the patches over the new few days. I will submit the\n> after I complete prlimiary testing. I will also mail the patch to anyone else\n> who would like to test for a specific port.\n>\n> > We have enough different shared libraries now that it'd be a good idea\n> > to try to eliminate most of that cruft from the individual module\n> > makefiles, replacing it with definitions made in the port makefiles.\n> > But it's a bit too late in the 6.4 beta cycle to try to do that for 6.4,\n> > I think --- the testing problem looms too large. Something for next time.\n>\n> I aggree that the shared library handling needs to be standardized and the\n> port specific information moved to the port specific Makefile.\n>\n> >\n> > BTW there is also a configuration check that needs to be applied: the\n> > pl/tcl subdirectory should be built only if a *shared library* version\n> > of Tcl is available. I happen to have only a static version of libtcl.a\n> > installed, which is fine for pgtclsh, but pl/tcl tries to link with it\n> > and falls over...\n>\n> Hmmmm.... I don't now autoconf well enough (yet) to handle this change.\n>\n\n I'd really appreciated to have PL/pgSQL installed in\n template1 with the 6.4 release. But I see that the problems\n we have with the make process of shared libraries is too much\n trouble that close to release.\n\n Since PL/Tcl also causes trouble let's put both into 6.5 and\n don't delay 6.4 longer than required for the other fixes.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 5 Oct 1998 09:28:34 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] plpgsql is not ready for prime time"
}
] |
[
{
"msg_contents": "There is some code in libpgtcl that purports to convert Postgres array\ndata values into Tcl lists. Is anyone prepared to argue that that code\ndoes something useful in its present state? I can name half a dozen\nbugs in it without breathing hard:\n\n1. Blithely assumes that any data value beginning with '{' and ending\nwith '}' must represent an array value. Should have some more robust\nway of discovering whether a column is array type. (In fairness, this\nmight require a FE/BE protocol change, unless arrayness can be\ndetermined from the tuple descriptors provided by the backend, ie,\nfield type OID, size, and attmod. Anybody know a way to do that?)\n\n2. Applies a translation that converts all backslash escape sequences\ndefined for C string constants into their equivalent single characters.\nSince neither the backend nor Tcl generate anything close to C-string\nescapes, the point of this is difficult to determine. It does however\nresult in unexpected output, eg disappearing backslashes.\n\n3. Applies said translation even when processing a non-array data value.\n\n4. Doesn't actually manage to produce a valid Tcl list, if the data\ncontains anything Tcl considers a special character. What it *should*\nbe doing is quoting, not de-quoting.\n\n5. Fails to modify \\\\, \\{, and \\} (thus quite unintentionally doing\nalmost the right thing...) when these sequences appear inside an array\nvalue, because \"they will be unescaped by Tcl in Tcl_AppendElement\".\nBut in fact Tcl_AppendElement is not invoked on the results of this\ncode.\n\n6. Modifies the string returned by libpq *in place*. This would be a\nconst-ness violation if we had been more careful about declaring things\nconst. More importantly, it means that re-examining the same tuple of\nthe PGresult will yield a different result. Not cool.\n\n7. The TCL_ARRAYS code is only invoked in the \"-assign\" variant of\nthe pgtcl pg_result statement, not in any of the other paths that allow\ntuple values to be examined. This is presumably an oversight, not\nthe intended behavior.\n\n8. Does not cope with MULTIBYTE strings. (But I don't think Tcl does\neither, so it's not clear that this can be called a bug.)\n\n\nI am strongly inclined to rip this code out, because it is responsible\nfor several behaviors that were correctly called bugs when backslash-\nhandling was discussed on pgsql-interfaces back in August. If we don't\nrip it out, it needs a complete rewrite.\n\nUnless there is a bulletproof solution to problem #1 (how to tell\nwhether a field's data type is array), I do not think it is appropriate\nfor the basic pg_result code to be applying any such transform. Perhaps\nit would be reasonable to invent a separate string-formatting function,\nsay \"pg_arraytolist\", that would perform the conversion. It would then\nbe the application writer's responsibility to know which fields were\narrays and apply the conversion if he wanted it.\n\nComments?\n\n\t\t\tregards, tom lane\n",
"msg_date": "Sun, 04 Oct 1998 18:24:16 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "TCL_ARRAYS code in libpgtcl is pretty seriously broken"
},
{
"msg_contents": "> There is some code in libpgtcl that purports to convert Postgres array\n> data values into Tcl lists. Is anyone prepared to argue that that code\n> does something useful in its present state? I can name half a dozen\n> bugs in it without breathing hard:\n> \n> 1. Blithely assumes that any data value beginning with '{' and ending\n> with '}' must represent an array value. Should have some more robust\n> way of discovering whether a column is array type. (In fairness, this\n> might require a FE/BE protocol change, unless arrayness can be\n> determined from the tuple descriptors provided by the backend, ie,\n> field type OID, size, and attmod. Anybody know a way to do that?)\n> \n> 2. Applies a translation that converts all backslash escape sequences\n> defined for C string constants into their equivalent single characters.\n> Since neither the backend nor Tcl generate anything close to C-string\n> escapes, the point of this is difficult to determine. It does however\n> result in unexpected output, eg disappearing backslashes.\n> \n> 3. Applies said translation even when processing a non-array data value.\n> \n> 4. Doesn't actually manage to produce a valid Tcl list, if the data\n> contains anything Tcl considers a special character. What it *should*\n> be doing is quoting, not de-quoting.\n> \n> 5. Fails to modify \\\\, \\{, and \\} (thus quite unintentionally doing\n> almost the right thing...) when these sequences appear inside an array\n> value, because \"they will be unescaped by Tcl in Tcl_AppendElement\".\n> But in fact Tcl_AppendElement is not invoked on the results of this\n> code.\n> \n> 6. Modifies the string returned by libpq *in place*. This would be a\n> const-ness violation if we had been more careful about declaring things\n> const. More importantly, it means that re-examining the same tuple of\n> the PGresult will yield a different result. Not cool.\n> \n> 7. The TCL_ARRAYS code is only invoked in the \"-assign\" variant of\n> the pgtcl pg_result statement, not in any of the other paths that allow\n> tuple values to be examined. This is presumably an oversight, not\n> the intended behavior.\n> \n> 8. Does not cope with MULTIBYTE strings. (But I don't think Tcl does\n> either, so it's not clear that this can be called a bug.)\n> \n> \n> I am strongly inclined to rip this code out, because it is responsible\n> for several behaviors that were correctly called bugs when backslash-\n> handling was discussed on pgsql-interfaces back in August. If we don't\n> rip it out, it needs a complete rewrite.\n> \n> Unless there is a bulletproof solution to problem #1 (how to tell\n> whether a field's data type is array), I do not think it is appropriate\n> for the basic pg_result code to be applying any such transform. Perhaps\n> it would be reasonable to invent a separate string-formatting function,\n> say \"pg_arraytolist\", that would perform the conversion. It would then\n> be the application writer's responsibility to know which fields were\n> arrays and apply the conversion if he wanted it.\n\nI have just started to learn TCL, and have Practical Tcl and TK by Brent\nWelch on my desk.\n\nSounds like our interface needs fixing. I am sure we currently do not\nhandle full protocol correctly.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 4 Oct 1998 20:32:58 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] TCL_ARRAYS code in libpgtcl is pretty seriously broken"
},
{
"msg_contents": "> 1. Blithely assumes that any data value beginning with '{' and ending\n> with '}' must represent an array value. Should have some more robust\n> way of discovering whether a column is array type. (In fairness, this\n> might require a FE/BE protocol change, unless arrayness can be\n> determined from the tuple descriptors provided by the backend, ie,\n> field type OID, size, and attmod. Anybody know a way to do that?)\n> Comments?\n\nPostgres seems to use a convention that a type name which starts with an\nunderscore is the array type for the corresponding non-underscore,\nnon-array type. Also, the typelem field in pg_type is non-zero for array\ntypes. \n\nThis isn't a definitive answer and there may be another way to discover\narray-ness but it's where I would look. Not sure if you'd be happy\nhaving to do a select on pg_type for every query unless you're doing it\nalready...\n\n - Tom\n",
"msg_date": "Mon, 05 Oct 1998 00:57:02 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] TCL_ARRAYS code in libpgtcl is pretty seriously broken"
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> This isn't a definitive answer and there may be another way to discover\n> array-ness but it's where I would look. Not sure if you'd be happy\n> having to do a select on pg_type for every query unless you're doing it\n> already...\n\nI think that won't fly for performance reasons. I know I wouldn't want\nmy Tcl applications paying an additional frontend-to-backend round trip\nfor every SELECT result...\n\nI was actually a tad surprised to realize that the column-type info sent\nby the backend couldn't answer such a basic question as \"is this an\narray?\". Something to fix if we ever rev the FE/BE protocol again.\nDon't think I'd propose a protocol rev just for this, though.\n\nIn the meantime, I'm inclined to take the fallback approach I suggested\nyesterday: provide the array-to-list reformatting function as a separate\nTcl statement that the application programmer can decide to invoke.\nThe app writer is likely to know perfectly well where he needs that\nfeature anyway.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 05 Oct 1998 10:11:51 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] TCL_ARRAYS code in libpgtcl is pretty seriously broken "
},
{
"msg_contents": "> \n> There is some code in libpgtcl that purports to convert Postgres array\n> data values into Tcl lists. Is anyone prepared to argue that that code\n> does something useful in its present state? I can name half a dozen\n> bugs in it without breathing hard:\n> \n> 1. Blithely assumes that any data value beginning with '{' and ending\n> with '}' must represent an array value. Should have some more robust\n> way of discovering whether a column is array type. (In fairness, this\n> might require a FE/BE protocol change, unless arrayness can be\n> determined from the tuple descriptors provided by the backend, ie,\n> field type OID, size, and attmod. Anybody know a way to do that?)\n> \n> 2. Applies a translation that converts all backslash escape sequences\n> defined for C string constants into their equivalent single characters.\n> Since neither the backend nor Tcl generate anything close to C-string\n> escapes, the point of this is difficult to determine. It does however\n> result in unexpected output, eg disappearing backslashes.\n> \n> 3. Applies said translation even when processing a non-array data value.\n> \n> 4. Doesn't actually manage to produce a valid Tcl list, if the data\n> contains anything Tcl considers a special character. What it *should*\n> be doing is quoting, not de-quoting.\n> \n> 5. Fails to modify \\\\, \\{, and \\} (thus quite unintentionally doing\n> almost the right thing...) when these sequences appear inside an array\n> value, because \"they will be unescaped by Tcl in Tcl_AppendElement\".\n> But in fact Tcl_AppendElement is not invoked on the results of this\n> code.\n> \n> 6. Modifies the string returned by libpq *in place*. This would be a\n> const-ness violation if we had been more careful about declaring things\n> const. More importantly, it means that re-examining the same tuple of\n> the PGresult will yield a different result. Not cool.\n> \n> 7. The TCL_ARRAYS code is only invoked in the \"-assign\" variant of\n> the pgtcl pg_result statement, not in any of the other paths that allow\n> tuple values to be examined. This is presumably an oversight, not\n> the intended behavior.\n> \n> 8. Does not cope with MULTIBYTE strings. (But I don't think Tcl does\n> either, so it's not clear that this can be called a bug.)\n> \n> \n> I am strongly inclined to rip this code out, because it is responsible\n> for several behaviors that were correctly called bugs when backslash-\n> handling was discussed on pgsql-interfaces back in August. If we don't\n> rip it out, it needs a complete rewrite.\n> \n> Unless there is a bulletproof solution to problem #1 (how to tell\n> whether a field's data type is array), I do not think it is appropriate\n> for the basic pg_result code to be applying any such transform. Perhaps\n> it would be reasonable to invent a separate string-formatting function,\n> say \"pg_arraytolist\", that would perform the conversion. It would then\n> be the application writer's responsibility to know which fields were\n> arrays and apply the conversion if he wanted it.\n> \n> Comments?\n> \n> \t\t\tregards, tom lane\n\nI wrote this code and used it for two years without any problem. All\nthe bugs you mentioned disappear if you use the proper string output\nfunctions which C-like escapes (code in contrib/string-io). This makes\nalso possible to distinguish between array and normal attibutes.\nThe code works fine in this case. I did a lot of testing at the time.\nHowever it is ok to move it into a separate tcl command.\n\n-- \nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n| Massimo Dal Zotto email: [email protected] |\n| Via Marconi, 141 phone: ++39-461-534251 |\n| 38057 Pergine Valsugana (TN) www: http://www.cs.unitn.it/~dz/ |\n| Italy pgp: finger [email protected] |\n+----------------------------------------------------------------------+\n",
"msg_date": "Mon, 5 Oct 1998 21:59:33 +0200 (MET DST)",
"msg_from": "Massimo Dal Zotto <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] TCL_ARRAYS code in libpgtcl is pretty seriously broken"
},
{
"msg_contents": "Massimo Dal Zotto <[email protected]> writes:\n>> [ my gripes about TCL_ARRAYS code snipped ]\n\n> I wrote this code and used it for two years without any problem. All\n> the bugs you mentioned disappear if you use the proper string output\n> functions which C-like escapes (code in contrib/string-io).\n\nWell, not the bug involving overwriting libpq's PGresult storage.\nBut still, this explains a good deal.\n\nI guess my gripe here is that the TCL_ARRAYS option is enabled *by\ndefault* in the Postgres distribution, but the code does not work\nproperly unless one plugs in a contrib function in the backend\n(and if the connection between the two is documented anywhere, I sure\ndidn't see it). This is not a reasonable default setup.\n\nAs a short-term fix I suggest that we just turn off TCL_ARRAYS in\nthe distributed config.h.in --- does that sound reasonable to you?\n\nIn the longer term, there needs to be some way of applying the\nTCL_ARRAYS conversion code only when your contrib stuff is being used.\nBacking the conversion code out of pg_result and making it a separate\nfunction might do, but that is a low-tech solution that puts the\nresponsibility on the application programmer to combine the right bits\nof frontend and backend functionality. Perhaps someone can think of a\nbetter way?\n\nUltimately I would like the text I/O functions to be completely 8-bit\nclean and able to deal with arbitrary byte strings. That is not\nsomething we can hope to shoehorn into 6.4, though.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 05 Oct 1998 19:39:39 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] TCL_ARRAYS code in libpgtcl is pretty seriously broken "
},
{
"msg_contents": "> \n> Massimo Dal Zotto <[email protected]> writes:\n> >> [ my gripes about TCL_ARRAYS code snipped ]\n> \n> > I wrote this code and used it for two years without any problem. All\n> > the bugs you mentioned disappear if you use the proper string output\n> > functions which C-like escapes (code in contrib/string-io).\n> \n> Well, not the bug involving overwriting libpq's PGresult storage.\n> But still, this explains a good deal.\n\nI didn't realize it could be called more than once for the same tuple.\nWe must allocate and free a new string for each value. I will write a patch.\n\n> I guess my gripe here is that the TCL_ARRAYS option is enabled *by\n> default* in the Postgres distribution, but the code does not work\n> properly unless one plugs in a contrib function in the backend\n> (and if the connection between the two is documented anywhere, I sure\n> didn't see it). This is not a reasonable default setup.\n\nI submitted the two patches at the same time but one of them was put in the\ncontribs.\n\n> As a short-term fix I suggest that we just turn off TCL_ARRAYS in\n> the distributed config.h.in --- does that sound reasonable to you?\n\nGiven the situation I agree that TCL_ARRAYS should be disabled by default.\n\n> In the longer term, there needs to be some way of applying the\n> TCL_ARRAYS conversion code only when your contrib stuff is being used.\n> Backing the conversion code out of pg_result and making it a separate\n> function might do, but that is a low-tech solution that puts the\n> responsibility on the application programmer to combine the right bits\n> of frontend and backend functionality. Perhaps someone can think of a\n> better way?\n> \n> Ultimately I would like the text I/O functions to be completely 8-bit\n> clean and able to deal with arbitrary byte strings. That is not\n> something we can hope to shoehorn into 6.4, though.\n\nI completely agree on the last thing. It is something I have been asking\nfrom a long time but apparently nobody cared bout it.\n\n-- \nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n| Massimo Dal Zotto email: [email protected] |\n| Via Marconi, 141 phone: ++39-461-534251 |\n| 38057 Pergine Valsugana (TN) www: http://www.cs.unitn.it/~dz/ |\n| Italy pgp: finger [email protected] |\n+----------------------------------------------------------------------+\n",
"msg_date": "Sun, 11 Oct 1998 14:04:47 +0200 (MET DST)",
"msg_from": "Massimo Dal Zotto <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] TCL_ARRAYS code in libpgtcl is pretty seriously broken"
}
] |
[
{
"msg_contents": "> Did you update the man pages or docs? I don't see the man page \n> changed. Do you want me to do it?\n\nFunny you should ask...\n\nI'm just in the process of converting all of the \"utilities\" man pages\nto sgml. For this release, we should keep the man pages as-is, but we\nwill have equivalent info in hardcopy and html.\n\nSo, I haven't updated the man page yet, but I'll do that. I did update\nthe usage printout from pg_dump itself.\n\nAnd, can we be sure to update both sgml and man pages from here to\nrelease to keep them in sync? I should be able to commit most sgml\nsources for the utilities this evening; on my list of \"must-haves\" just\npg_dump, pg_dumpall, and vacuum are left to do.\n\nbtw, is \"-n\" an acceptable choice for the flag? I'd be happy with\nanything, but decided that \"-q\" probably shouldn't be used since many\nprograms use it to mean \"quiet\". So -n for \"No quotes\" is what I chose\ninstead...\n\n - Tom\n",
"msg_date": "Sun, 04 Oct 1998 22:58:41 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: pg_dump new -n flag"
},
{
"msg_contents": "> > Did you update the man pages or docs? I don't see the man page \n> > changed. Do you want me to do it?\n> \n> Funny you should ask...\n> \n> I'm just in the process of converting all of the \"utilities\" man pages\n> to sgml. For this release, we should keep the man pages as-is, but we\n> will have equivalent info in hardcopy and html.\n\nGood.\n\n> So, I haven't updated the man page yet, but I'll do that. I did update\n> the usage printout from pg_dump itself.\n\nI saw that.\n\n> And, can we be sure to update both sgml and man pages from here to\n> release to keep them in sync? I should be able to commit most sgml\n> sources for the utilities this evening; on my list of \"must-haves\" just\n> pg_dump, pg_dumpall, and vacuum are left to do.\n\nYes. Good idea. Just one question. Are we sure we have all the man\npage sync with the sgml? I remember Tom Lane saying there was some\nstuff missing from the sgml.\n\n\n> btw, is \"-n\" an acceptable choice for the flag? I'd be happy with\n> anything, but decided that \"-q\" probably shouldn't be used since many\n> programs use it to mean \"quiet\". So -n for \"No quotes\" is what I chose\n> instead...\n\nSeems good.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 4 Oct 1998 20:39:27 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: pg_dump new -n flag"
},
{
"msg_contents": "> > And, can we be sure to update both sgml and man pages from here to\n> > release to keep them in sync?\n> Yes. Good idea. Just one question. Are we sure we have all the man\n> page sync with the sgml? I remember Tom Lane saying there was some\n> stuff missing from the sgml.\n\nI just committed a bunch of new sgml sources covering the essential\nutilities. They are based on the equivalent man pages, so we should try\nto keep them in sync until after the upcoming release.\n\nI think that Tom Lane is working on the sgml docs right now, and I saw a\ncommit of at least one already. I'll go ahead and remove one or two more\nof the man page docs which are obsolete.\n\nI still have to consolidate or write new info for the SERIAL, INT8, and\nCIDR data types (I've got Tom H's info on CIDR, but expect to have to\nmove things around a bit).\n\nbtw, I'm thinking of changing the src/ and doc/ Makefiles to have the\nman pages installed from the doc directory, not the src directory. So,\none would get the man pages installed by doing\n\n % cd doc\n % make install\n\nrather than having them installed every time you install a new\nexecutable, as currently happens. I won't move the location of the man\nsources, just change the makefiles.\n\nComments?\n\n - Tom\n",
"msg_date": "Mon, 05 Oct 1998 03:15:34 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: pg_dump new -n flag"
},
{
"msg_contents": "Bruce Momjian <[email protected]> writes:\n> I remember Tom Lane saying there was some\n> stuff missing from the sgml.\n\nLast night I committed a ton of revisions to libpq.sgml; it's in good\nshape now (except possibly for some bugs in sgml markup, which I don't\nhave an easy way to check here). I still need to work on libpgtcl.sgml.\n\nI've recommended to Tom Lockhart that we drop libpq.3 (for now), because\nI don't want to transpose the results of that mammoth editing session\nback into nroff format. Eventually it ought to be possible to generate\nman format from the sgml, if you prefer man pages.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 05 Oct 1998 10:01:28 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: pg_dump new -n flag "
},
{
"msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> btw, I'm thinking of changing the src/ and doc/ Makefiles to have the\n> man pages installed from the doc directory, not the src directory. So,\n> one would get the man pages installed by doing\n> % cd doc\n> % make install\n> rather than having them installed every time you install a new\n> executable, as currently happens.\n\nThat sounds good...\n\n> I won't move the location of the man\n> sources, just change the makefiles.\n\n... but as long as the man pages live under src/man, I think they ought\nto be installed by the src makefile. Cross-subtree installs are confusing.\nI'd vote for going all the way and moving the src/man subdirectory into\nthe doc tree.\n\n\t\t\tregards, tom lane\n",
"msg_date": "Mon, 05 Oct 1998 10:18:47 -0400",
"msg_from": "Tom Lane <[email protected]>",
"msg_from_op": false,
"msg_subject": "man-page install (was Re: pg_dump new -n flag)"
},
{
"msg_contents": "> > I won't move the location of the man\n> > sources, just change the makefiles.\n> ... but as long as the man pages live under src/man, I think they \n> ought to be installed by the src makefile. Cross-subtree installs are \n> confusing.\n> I'd vote for going all the way and moving the src/man subdirectory \n> into the doc tree.\n\nI agree, but would like to keep it this way for this release for three\nreasons:\n\n1) the src/ makefile is actually doing the install. So from src/ you can\ndo a\n\n make install-man\n\nbut it is no longer part of \"src/make install\". The doc/ makefile simply\ndoes the above make.\n\n2) moving the files will (unfortunately) eliminate the cvs log\ninformation (unless I do bad stuff with the RCS tree behind cvs, and I'm\nnot touching that :)\n\n3) the next release should have man pages derived from sgml so it won't\nbe an issue.\n\n - Tom\n",
"msg_date": "Mon, 05 Oct 1998 14:44:01 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: man-page install (was Re: pg_dump new -n flag)"
}
] |
[
{
"msg_contents": "> Modified Files:\n> protocol.sgml\n> Log Message:\n> Add note explaining that NoticeResponse messages can\n> arrive even while idle.\n\nOh boy! Another unsolicited patch for sgml docs. I'm so excited!\n\nIf anyone feels inspired, there is probably another one week window when\nwe can make updates to the sgml docs which would appear in new hardcopy.\nIf you've been thinking about writing anything remember contribs are\nwelcome. I'll post new sources and html output on the web site in the\nnext couple of days.\n\nOne thing that would be helpful would be for someone to verify that the\ntutorial docs and tutorial files actually work as advertised. Or if\nsomeone has run through them recently (the last 4 months?) please report\nin. I wasn't planning any updates to them since I haven't heard of any\nproblems but I'm not sure if any of us old timers have run through those\nin a long time...\n\nDeJuan, could you ask on the more general lists whether anyone has tried\nthe tutorials with success, and solicit comments or corrections? TIA\n\n - Tom\n",
"msg_date": "Mon, 05 Oct 1998 01:19:32 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [COMMITTERS] 'pgsql/doc/src/sgml protocol.sgml'"
}
] |
[
{
"msg_contents": "Hi,\n\ncvs still has problem with vacuum analyze on my Linux machine:\nafter completion of regression test which passed fine I tried\nvacuum analyze on regression db and got \n\nNOTICE: AbortTransaction and not in in-progress state \nNOTICE: AbortTransaction and not in in-progress state \n\nin regress.log there were following messages:\n......skipped....\nages 0/0. Elapsed 0/0 sec.\nDEBUG: --Relation lseg_tbl--\nDEBUG: Pages 1: Changed 0, Reapped 0, Empty 0, New 0; Tup 5: Vac 0, Crash 0, UnUsed 0, MinLen 72, MaxLen 72; Re-using: Free/Avail. Space 0/0; EndEmpty/Avail. Pages 0/0. Elapsed 0/0 sec.\nERROR: cannot write block 0 of pkeys [regression] blind\nNOTICE: AbortTransaction and not in in-progress state \nERROR: cannot write block 0 of pkeys [regression] blind\nNOTICE: AbortTransaction and not in in-progress state \nERROR: cannot write block 0 of pkeys [regression] blind\nNOTICE: AbortTransaction and not in in-progress state \n\nEarlier this happened only if I run postmaster with number of buffers > 128\nand never happens when I run regression test following procedure\ndescribed in INSTALL. \n\n\tOleg\n\n\n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n",
"msg_date": "Mon, 5 Oct 1998 05:05:34 +0300 (MSK)",
"msg_from": "Oleg Bartunov <[email protected]>",
"msg_from_op": true,
"msg_subject": "vacuum analyze problem"
},
{
"msg_contents": "> Hi,\n> \n> cvs still has problem with vacuum analyze on my Linux machine:\n> after completion of regression test which passed fine I tried\n> vacuum analyze on regression db and got \n> \n> NOTICE: AbortTransaction and not in in-progress state \n> NOTICE: AbortTransaction and not in in-progress state \n> \n> in regress.log there were following messages:\n> ......skipped....\n> ages 0/0. Elapsed 0/0 sec.\n> DEBUG: --Relation lseg_tbl--\n> DEBUG: Pages 1: Changed 0, Reapped 0, Empty 0, New 0; Tup 5: Vac 0, Crash 0, UnUsed 0, MinLen 72, MaxLen 72; Re-using: Free/Avail. Space 0/0; EndEmpty/Avail. Pages 0/0. Elapsed 0/0 sec.\n> ERROR: cannot write block 0 of pkeys [regression] blind\n> NOTICE: AbortTransaction and not in in-progress state \n> ERROR: cannot write block 0 of pkeys [regression] blind\n> NOTICE: AbortTransaction and not in in-progress state \n> ERROR: cannot write block 0 of pkeys [regression] blind\n> NOTICE: AbortTransaction and not in in-progress state \n> \n> Earlier this happened only if I run postmaster with number of buffers > 128\n> and never happens when I run regression test following procedure\n> described in INSTALL. \n\nJust tested on BSDI, and I can't reproduce the problem here.\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 4 Oct 1998 22:54:14 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] vacuum analyze problem"
},
{
"msg_contents": "> cvs still has problem with vacuum analyze on my Linux machine:\n> after completion of regression test which passed fine I tried\n> vacuum analyze on regression db and got\n> NOTICE: AbortTransaction and not in in-progress state\n> NOTICE: AbortTransaction and not in in-progress state\n*snip*\n> Earlier this happened only if I run postmaster with number of buffers > 128 and never happens when I run regression test following procedure\n> described in INSTALL.\n\nI'm not seeing this problem on my Linux machine (2.0.30 kernel) so it\nprobably isn't a general problem with Linux. I haven't done a completely\nnew refresh of my sources for at least a few days, though I don't think\nthis is the problem because you've been seeing something like this on\nyour machine for a while.\n\nYour problem would be easier to diagnose if someone could reproduce it\non another machine, but if the test is as simple as trying a vacuum on\nthe regression database then it's easy to test and we aren't finding it\nhappening everywhere :(\n\nKeep plugging away at it though because there's always a chance that\nyour particular machine is uncovering an obscure feature of the code...\n\n - Tom\n",
"msg_date": "Mon, 05 Oct 1998 03:28:31 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] vacuum analyze problem"
}
] |
[
{
"msg_contents": "\nThanks Daniele, thanks to everyone who replied, and to those who considered...\n\nNothing helps. Disabling fsync does speed up COPY, but not with the table in question. It does not even seem to be the question of speed anymore. Even if COPY terminates, the table and the database containnig it become unusable: queries run awfully long and return nothing. Looks like I managed to make a killer table, because even a fraction of that table kills the database it is being copied to.\n\nJust in case it matters, the table I am trying to create is this:\n\nCREATE TABLE key(\n tag char(3), \n id char(12), \n rel char(70), \n pos char(16), \n item int2, \n unit char(48),\n data text\n );\n\nHere is the fragment of the data (can provide the whole table for testing):\n\nKM APS94172-01 CO:KM:KT:VM:MA:KC:CEF:ACC:1 66:15:66:19 1 mmol/l 0.03\nKT APS94172-01 CO:KM:KT:VM:MA:KC:CEF:ACC:1 66:23:66:24 1 NULL H\nVM APS94172-01 CO:KM:KT:VM:MA:KC:CEF:ACC:1 66:29:66:36 1 umol/min/mg 86\nMA APS94172-01 CO:KM:KT:VM:MA:KC:CEF:ACC:1 66:40:66:45 1 1/min 1.9E3\nOR APS94172-01 NULL 16:4:16:17 1 NULL `Glicine max`\nOCN APS94172-01 NULL 17:4:17:11 1 NULL soybean\nPS APS94172-01 NULL 32:5:32:1208 1 NULL T=4(C)\nPS APS94172-01 NULL 32:5:32:1208 10 NULL Blue Sepharose CL-6B dye affinity chromatography, elution with 50(mmol/l) imidazole buffer, pH = 7.2, 10(mmol/l) 2-mercaptoethanol, 25(%) glycerol, 500(mmol/l) KCl, 2.0(mmol/l) folic acid, peak fraction\n\n--Gene\n\n\nDaniele Orlandi wrote:\n>\n> Gene Selkov Jr. wrote:\n> > \n> > Hi,\n> > \n> > I am watching the progress of a COPY from a 2 megarow table. It seems to slow \n> > down almost exponentially. The database file grew at about 2 megabytes a \n> > minute in the beginning and now, 20 hours since started, it grows at 80 \n> > kilobytes a minute. The table has 7 columns of char and int types, the last \n> > column is text. The scene is a dual 400MHz i586 running RedHat 5.1 and \n> > postgres 6.3.2, as originally distributed, no patches. As far as I can tell \n> > by 'tail | strings | tail' on the database file, it is more than halfway \n> > there, but such exponential slowdown makes me anxious. Other users will have \n> > to take vacation:\n> \n> I had the same behaviour trying to copy about 250K records. Just follow the\n> suggestions in the documentation (remove indexes, disable fsync etc...).\n> \n> It would be nice to understand why this happens, it could be a bug or something\n> that could be improved.\n",
"msg_date": "Mon, 05 Oct 1998 01:21:27 -0400",
"msg_from": "\"Gene Selkov Jr.\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [ADMIN] COPY slows down; is it normal? "
},
{
"msg_contents": "What version is this? I would be interested to see if you see the same\nthing in 6.4 beta.\n\n\n> \n> Thanks Daniele, thanks to everyone who replied, and to those who considered...\n> \n> Nothing helps. Disabling fsync does speed up COPY, but not with the table in question. It does not even seem to be the question of speed anymore. Even if COPY terminates, the table and the database containnig it become unusable: queries run awfully long and return nothing. Looks like I managed to make a killer table, because even a fraction of that table kills the database it is being copied to.\n> \n> Just in case it matters, the table I am trying to create is this:\n> \n> CREATE TABLE key(\n> tag char(3), \n> id char(12), \n> rel char(70), \n> pos char(16), \n> item int2, \n> unit char(48),\n> data text\n> );\n> \n> Here is the fragment of the data (can provide the whole table for testing):\n> \n> KM APS94172-01 CO:KM:KT:VM:MA:KC:CEF:ACC:1 66:15:66:19 1 mmol/l 0.03\n> KT APS94172-01 CO:KM:KT:VM:MA:KC:CEF:ACC:1 66:23:66:24 1 NULL H\n> VM APS94172-01 CO:KM:KT:VM:MA:KC:CEF:ACC:1 66:29:66:36 1 umol/min/mg 86\n> MA APS94172-01 CO:KM:KT:VM:MA:KC:CEF:ACC:1 66:40:66:45 1 1/min 1.9E3\n> OR APS94172-01 NULL 16:4:16:17 1 NULL `Glicine max`\n> OCN APS94172-01 NULL 17:4:17:11 1 NULL soybean\n> PS APS94172-01 NULL 32:5:32:1208 1 NULL T=4(C)\n> PS APS94172-01 NULL 32:5:32:1208 10 NULL Blue Sepharose CL-6B dye affinity chromatography, elution with 50(mmol/l) imidazole buffer, pH = 7.2, 10(mmol/l) 2-mercaptoethanol, 25(%) glycerol, 500(mmol/l) KCl, 2.0(mmol/l) folic acid, peak fraction\n> \n> --Gene\n> \n> \n> Daniele Orlandi wrote:\n> >\n> > Gene Selkov Jr. wrote:\n> > > \n> > > Hi,\n> > > \n> > > I am watching the progress of a COPY from a 2 megarow table. It seems to slow \n> > > down almost exponentially. The database file grew at about 2 megabytes a \n> > > minute in the beginning and now, 20 hours since started, it grows at 80 \n> > > kilobytes a minute. The table has 7 columns of char and int types, the last \n> > > column is text. The scene is a dual 400MHz i586 running RedHat 5.1 and \n> > > postgres 6.3.2, as originally distributed, no patches. As far as I can tell \n> > > by 'tail | strings | tail' on the database file, it is more than halfway \n> > > there, but such exponential slowdown makes me anxious. Other users will have \n> > > to take vacation:\n> > \n> > I had the same behaviour trying to copy about 250K records. Just follow the\n> > suggestions in the documentation (remove indexes, disable fsync etc...).\n> > \n> > It would be nice to understand why this happens, it could be a bug or something\n> > that could be improved.\n> \n> \n\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Mon, 5 Oct 1998 22:00:48 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [ADMIN] COPY slows down; is it normal?"
},
{
"msg_contents": "Bruce Momjian wrote:\n\n> What version is this? I would be interested to see if you see the same\n> thing in 6.4 beta.\n>\n\nIt is 6.3.2, no patches applied. I will check it in 6.4, the latest versions I can see are\n\n postgresql.snapshot.tar.gz 4091 Kb Mon Oct 5 11:01:00 1998\n postgresql.v6.4-BETA1.tar.gz 4057 Kb Mon Sep 14 08:40:00 1998\n\nWhich of these should I try?\n\n--Gene\n",
"msg_date": "Tue, 06 Oct 1998 00:11:30 -0400",
"msg_from": "\"Gene Selkov Jr.\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [ADMIN] COPY slows down; is it normal? "
},
{
"msg_contents": "On Tue, 6 Oct 1998, Gene Selkov Jr. wrote:\n\n> Bruce Momjian wrote:\n> \n> > What version is this? I would be interested to see if you see the same\n> > thing in 6.4 beta.\n> >\n> \n> It is 6.3.2, no patches applied. I will check it in 6.4, the latest versions I can see are\n> \n> postgresql.snapshot.tar.gz 4091 Kb Mon Oct 5 11:01:00 1998\n> postgresql.v6.4-BETA1.tar.gz 4057 Kb Mon Sep 14 08:40:00 1998\n> \n> Which of these should I try?\n\n\tWait a day or so...I'm just testing out a BETA2 right now on a\ncouple of different platforms, and hope to have a new tar file up\ntomorrow...\n\nMarc G. Fournier [email protected]\nSystems Administrator @ hub.org \nscrappy@{postgresql|isc}.org ICQ#7615664\n\n",
"msg_date": "Tue, 13 Oct 1998 17:44:56 -0400 (EDT)",
"msg_from": "\"Marc G. Fournier\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [ADMIN] COPY slows down; is it normal? "
},
{
"msg_contents": "Bruce Momjian wrote:\n> \n> What version is this? I would be interested to see if you see the same\n> thing in 6.4 beta.\n\nPostgresql 6.4, 13-Oct snapshot, importing about80MB of data *without* indexes\nfrom a pg_dump. This is the size of the database file minute per minute starting\nfrom about 13MB already imported:\n\n-rw------- 1 postgres postgres 13434880 Oct 14 00:49 log2\n-rw------- 1 postgres postgres 16015360 Oct 14 00:50 log2\n-rw------- 1 postgres postgres 22765568 Oct 14 00:51 log2\n-rw------- 1 postgres postgres 25034752 Oct 14 00:52 log2\n-rw------- 1 postgres postgres 29212672 Oct 14 00:53 log2\n-rw------- 1 postgres postgres 32923648 Oct 14 00:54 log2\n-rw------- 1 postgres postgres 35717120 Oct 14 00:55 log2\n-rw------- 1 postgres postgres 38363136 Oct 14 00:56 log2\n-rw------- 1 postgres postgres 41394176 Oct 14 00:57 log2\n-rw------- 1 postgres postgres 44310528 Oct 14 00:58 log2\n-rw------- 1 postgres postgres 47063040 Oct 14 00:59 log2\n-rw------- 1 postgres postgres 49176576 Oct 14 01:00 log2\n-rw------- 1 postgres postgres 51142656 Oct 14 01:01 log2\n-rw------- 1 postgres postgres 53305344 Oct 14 01:02 log2\n-rw------- 1 postgres postgres 55394304 Oct 14 01:03 log2\n-rw------- 1 postgres postgres 57581568 Oct 14 01:04 log2\n-rw------- 1 postgres postgres 59236352 Oct 14 01:05 log2\n-rw------- 1 postgres postgres 60784640 Oct 14 01:06 log2\n-rw------- 1 postgres postgres 62578688 Oct 14 01:07 log2\n-rw------- 1 postgres postgres 64389120 Oct 14 01:08 log2\n-rw------- 1 postgres postgres 66199552 Oct 14 01:09 log2\n-rw------- 1 postgres postgres 67723264 Oct 14 01:10 log2\n-rw------- 1 postgres postgres 69140480 Oct 14 01:11 log2\n-rw------- 1 postgres postgres 70656000 Oct 14 01:12 log2\n-rw------- 1 postgres postgres 72122368 Oct 14 01:13 log2\n-rw------- 1 postgres postgres 73596928 Oct 14 01:14 log2\n-rw------- 1 postgres postgres 74883072 Oct 14 01:15 log2\n-rw------- 1 postgres postgres 76201984 Oct 14 01:16 log2\n-rw------- 1 postgres postgres 77692928 Oct 14 01:17 log2\n-rw------- 1 postgres postgres 79175680 Oct 14 01:18 log2\n-rw------- 1 postgres postgres 80650240 Oct 14 01:19 log2\n-rw------- 1 postgres postgres 81821696 Oct 14 01:20 log2\n-rw------- 1 postgres postgres 83042304 Oct 14 01:21 log2\n-rw------- 1 postgres postgres 83574784 Oct 14 01:22 log2\n\nAs you may see, the speed decreases as the table size increases.\n\nBye!\n\n-- \n Daniele\n\n-------------------------------------------------------------------------------\n\"It's God. No, not Richard Stallman, or Linus Torvalds, but God.\"\n(By Matt Welsh)\n-------------------------------------------------------------------------------\n Dal 28-09-1998, con un solo POP abbiamo fatto guadagnare a telecom italia\n 1769237 Lire solo in scatti alla risposta. Non male eh ?\n-------------------------------------------------------------------------------\n Daniele Orlandi - Utility Line Italia - http://www.orlandi.com\n Via Mezzera 29/A - 20030 - Seveso (MI) - Italy\n-------------------------------------------------------------------------------\n",
"msg_date": "Wed, 14 Oct 1998 01:33:01 +0200",
"msg_from": "Daniele Orlandi <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [ADMIN] COPY slows down; is it normal?"
}
] |
[
{
"msg_contents": "When I dumped (via pg_dump) a database that had PL/pgsql functions defined, \nthe following happened:\n\n CREATE FUNCTION val0_id (int4 ) RETURNS text AS 'unknown' LANGUAGE 'unknown';\n CREATE FUNCTION val1_id (int4 ) RETURNS text AS 'unknown' LANGUAGE 'unknown';\n CREATE FUNCTION val0_sect (text,char ) RETURNS text AS 'unknown' LANGUAGE\n 'unknown';\n CREATE FUNCTION val1_sect (text,char ) RETURNS text AS 'unknown' LANGUAGE\n 'unknown';\n\nApparently, pg_dump does not recognize/handle PL/pgsql functions yet.\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Mon, 05 Oct 1998 01:22:38 -0400",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Problem dumping PL/pgsql functions."
},
{
"msg_contents": ">\n> When I dumped (via pg_dump) a database that had PL/pgsql functions defined,\n> the following happened:\n>\n> CREATE FUNCTION val0_id (int4 ) RETURNS text AS 'unknown' LANGUAGE 'unknown';\n> CREATE FUNCTION val1_id (int4 ) RETURNS text AS 'unknown' LANGUAGE 'unknown';\n> CREATE FUNCTION val0_sect (text,char ) RETURNS text AS 'unknown' LANGUAGE\n> 'unknown';\n> CREATE FUNCTION val1_sect (text,char ) RETURNS text AS 'unknown' LANGUAGE\n> 'unknown';\n>\n> Apparently, pg_dump does not recognize/handle PL/pgsql functions yet.\n\n Seems that pg_dump doesn't lookup pg_language but has the\n languages hardcoded compiled in. That's not right. I'll take\n a look at it.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Mon, 5 Oct 1998 09:17:57 +0200 (MET DST)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Problem dumping PL/pgsql functions."
},
{
"msg_contents": "> When I dumped (via pg_dump) a database that had PL/pgsql functions defined, \n> the following happened:\n> \n> CREATE FUNCTION val0_id (int4 ) RETURNS text AS 'unknown' LANGUAGE 'unknown';\n> CREATE FUNCTION val1_id (int4 ) RETURNS text AS 'unknown' LANGUAGE 'unknown';\n> CREATE FUNCTION val0_sect (text,char ) RETURNS text AS 'unknown' LANGUAGE\n> 'unknown';\n> CREATE FUNCTION val1_sect (text,char ) RETURNS text AS 'unknown' LANGUAGE\n> 'unknown';\n> \n> Apparently, pg_dump does not recognize/handle PL/pgsql functions yet.\n\nIs this still an open issue?\n\n-- \n Bruce Momjian | http://www.op.net/~candle\n [email protected] | (610) 853-3000\n + If your life is a hard drive, | 830 Blythe Avenue\n + Christ can be your backup. | Drexel Hill, Pennsylvania 19026\n\n",
"msg_date": "Sun, 11 Oct 1998 21:19:44 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Problem dumping PL/pgsql functions."
},
{
"msg_contents": ">> Apparently, pg_dump does not recognize/handle PL/pgsql functions yet.\n> \n> Is this still an open issue?\n> \n\nNo - it is now working.\n\nBruce,\n\nDo not apply the patches I sent recently (PostgreSQL 6.4 patches - portability \nrelated). I will get a fresh copy of the CVS tree and recreate the patches.\n\nThanks.\n-- \n____ | Billy G. Allie | Domain....: [email protected]\n| /| | 7436 Hartwell | Compuserve: 76337,2061\n|-/-|----- | Dearborn, MI 48126| MSN.......: [email protected]\n|/ |LLIE | (313) 582-1540 | \n\n\n",
"msg_date": "Sun, 11 Oct 1998 21:28:15 -0400",
"msg_from": "\"Billy G. Allie\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Problem dumping PL/pgsql functions. "
}
] |
[
{
"msg_contents": "Hi all.\nI'm looking forward to the appearance of LLL in PostgreSQL 6.5 and have a\nquestion about the sensitivity of cursors in LLL.\n\nIn LLL cursors are INSENSITIVE as Oracle ?\n\nCurrently cursors are indeterminate and in some cases they are strangely\nsensitive(for me).\nIn LLL the behavior of cursors will be more complicated, if changes by other\ntransactions can be seen by fetch statements(especially for read committed\nisolation level).\n\nI hope INSENSITIVE cursors to be implemented whose behavior we can predict\nand I think that they can be realized according to proposals for LLL by\nVadim.\n\nIn LLL access methods return snapshot of data as they were in _some_ point\nin time.\nFor read committed mode this moment is the time when statement began.\nFor serializable mode this is the time when current transaction began.\n\nFor a INSENSITIVE cursor this is the time when it was opened(declared),\nnot the time when the fetch statements for it began ?\n\nThanks.\n\nHiroshi Inoue\[email protected]\n\n\n",
"msg_date": "Mon, 5 Oct 1998 19:29:10 +0900",
"msg_from": "\"Hiroshi Inoue\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "=?iso-2022-jp?B?Y3Vyc28bJEIjchsoQnMgaW4gTExM?="
},
{
"msg_contents": "Hiroshi Inoue wrote:\n> \n> Hi all.\n> I'm looking forward to the appearance of LLL in PostgreSQL 6.5 and have a\n> question about the sensitivity of cursors in LLL.\n> \n> In LLL cursors are INSENSITIVE as Oracle ?\n> \n> Currently cursors are indeterminate and in some cases they are strangely\n> sensitive(for me).\n\nDo you mean seeing row inserted between fetches ?\nShould this be changed ?\nHow is this in Oracle, Informix, Sybase, standards ?\n\n> In LLL the behavior of cursors will be more complicated, if changes by other\n> transactions can be seen by fetch statements(especially for read committed\n> isolation level).\n> \n> I hope INSENSITIVE cursors to be implemented whose behavior we can predict\n> and I think that they can be realized according to proposals for LLL by\n> Vadim.\n> \n> In LLL access methods return snapshot of data as they were in _some_ point\n> in time.\n> For read committed mode this moment is the time when statement began.\n> For serializable mode this is the time when current transaction began.\n> \n> For a INSENSITIVE cursor this is the time when it was opened(declared),\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nThis is easy to implement.\nBut I'd like to know what standards say about cursor sensitivness...\n\n> not the time when the fetch statements for it began ?\n\nThanks.\n\nVadim\n",
"msg_date": "Tue, 06 Oct 1998 10:07:18 +0800",
"msg_from": "Vadim Mikheev <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] curso\u001b$B#r\u001b(Bs in LLL"
},
{
"msg_contents": ">\n> Hiroshi Inoue wrote:\n> >\n> > Hi all.\n> > I'm looking forward to the appearance of LLL in PostgreSQL 6.5\n> and have a\n> > question about the sensitivity of cursors in LLL.\n> >\n> > In LLL cursors are INSENSITIVE as Oracle ?\n> >\n> > Currently cursors are indeterminate and in some cases they are strangely\n> > sensitive(for me).\n>\n> Do you mean seeing row inserted between fetches ?\n\nYes and in some cases updated rows can be seen many times.\n\n> Should this be changed ?\n\nYes,at least the option is necessary that changes can't be seen.\n\n> How is this in Oracle, Informix, Sybase, standards ?\n>\n\nSo far as I know,in Oracle cursors are fixed(logically) when\nthey are opened.\n\n> > In LLL the behavior of cursors will be more complicated, if\n> changes by other\n> > transactions can be seen by fetch statements(especially for\n> read committed\n> > isolation level).\n> >\n> > I hope INSENSITIVE cursors to be implemented whose behavior we\n> can predict\n> > and I think that they can be realized according to proposals for LLL by\n> > Vadim.\n> >\n> > In LLL access methods return snapshot of data as they were in\n> _some_ point\n> > in time.\n> > For read committed mode this moment is the time when statement began.\n> > For serializable mode this is the time when current transaction began.\n> >\n> > For a INSENSITIVE cursor this is the time when it was opened(declared),\n> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n> This is easy to implement.\n> But I'd like to know what standards say about cursor sensitivness...\n>\n\nIn SQL92,cursors are not necessarily INSENSITIVE and there's the option\nINSENSITIVE for DECLARE CURSOR statements.\n\nThanks.\n\nHiroshi Inoue\[email protected]\n\n",
"msg_date": "Tue, 6 Oct 1998 12:01:50 +0900",
"msg_from": "\"Hiroshi Inoue\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] cursors in LLL"
}
] |
[
{
"msg_contents": "\nI am re-posting this query to the Interfaces and hackers mailing lists\nwith a little more information...\n\nOS: HP_UX 10.2\nVersion of JDK : 1.1.5\nVersion of Postgres: 6.3.2\nDate Style : ISO\n\nI have a default record with a 'timestamp' field. Postgres saves 'epoch'\nas the default value for this field but the JDBC driver/JDK on HP-UX 10.2\ndo not recognise this and I get a\n'java.lang.StringIndexOutOfBoundsException' and\n'java.sql.SQLException: Bad Date Format: at 0 in epoch'\n\nBasically it is choking on 'epoch'.\n- Is there a way to have the date returned in its numeric format\n'1997-01-01.....' instead of epoch??\nor\n- Is there a patch for the JDBC driver with a fix for this?\n\nI appreciate any help.\nThank You\n\n-SR\n\n---------------------- Forwarded by Sangeeta Rao/MW/US/3Com on 10/05/98\n11:52 AM ---------------------------\n\nSangeeta Rao/MW/US/3Com\n09/29/98 04:58 PM\n\n\nTo: [email protected]@3COM-MWGATE\ncc:\nSubject: Default 'timestamp' value on HP-UX running Postgresql6.3.2\n\nI am running postgres 6.3.2 on HP_UX 10.2. I have a field defined as\ntype 'timestamp' and am using ISO datestyle. The default value shows\nup as the string constant 'epoch'. Is there any way of having the numeric\nvalue of '1970-01-01 00:00:00+00' returned instead??\n\nI appreciate any help! Thanks.\n-sr\n\n\n\n\n\n",
"msg_date": "Mon, 5 Oct 1998 12:03:41 -0500",
"msg_from": "\"Sangeeta Rao\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Default 'timestamp' value on HP-UX running Postgresql6.3.2"
}
] |
[
{
"msg_contents": "> Just did full and clean cvs'ing, cleaning, recompiling on 2 Linux \n> machines.\n> The problem with vacuum analyze on regression db dissapears, but it's\n> still exists when I run postmaster with -B 1024 option.\n> usually I start postmaster as\n> \n> /usr/local/pgsql/bin/postmaster -i -B 1024\n> -S -D/usr/local/pgsql/data/ -o '-Fe'\n> \n> and 6.3.2 is rock solid, no problem at all with vacuuming.\n> > Your problem would be easier to diagnose if someone could reproduce \n> > it on another machine, but if the test is as simple as trying a \n> > vacuum on the regression database then it's easy to test and we \n> > aren't finding it happening everywhere :(\n> \n> I could reproduce the problem myself :-) on 3 Linux machines, running \n> different kernels, different compilers etc. This is an excerption from \n> debug messages:\n> il. Pages 0/0. Elapsed 0/0 sec.\n> DEBUG: Index bt_txt_index: Pages 27; Tuples 10000. Elapsed 0/0 sec.\n> ERROR: cannot write block -1 of [] blind\n> AbortCurrentTransaction\n> NOTICE: AbortTransaction and not in in-progress state\n> > Keep plugging away at it though because there's always a chance that\n> > your particular machine is uncovering an obscure feature of the \n> > code...\n> 3 particular machines and all I managed to install software, so \n> problem could be in me :=)\n> What could you suggest I do to track the problem ?\n\nWell there is something about your installations which is different. The\nonly one I know about is that at least one of your machines is running\nthe 2.1.xxx kernels, which I don't have any experience with. If we can\nconfirm that this is then only difference between your machines and\nmine, then it is something which we should understand and report as a\nnew kernel problem. If that isn't the common element, then I'm at a loss\nas to why your machines and mine differ in behavior.\n\nI'll interpret the lack of \"hey, I'm having trouble too!\" messages from\nother people running Linux/iX86 as meaning that others aren't seeing\ntrouble, but it could just be that others aren't trying it yet. Does\nanyone else on Linux see this problem? Does anyone on Linux _not_ see\nthis problem?\n\n - Tom\n",
"msg_date": "Mon, 05 Oct 1998 17:15:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] vacuum analyze problem"
}
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.