Datasets:

wanshenl
/

pgsql-hackers

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 58.4k rows

threads listlengths 1 2.99k
[ { "msg_contents": "I just installed the latest snapshot. And now I cannot access the DB\nanymore:\n\npsql mm\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: mm\n\nmm=> \\z\nERROR: SearchSysCache: Called while cache disabled\n\n\nMichael\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Tue, 10 Feb 1998 12:49:27 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "Access to latest snapshot" }, { "msg_contents": "> \n> I just installed the latest snapshot. And now I cannot access the DB\n> anymore:\n> \n> psql mm\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: mm\n> \n> mm=> \\z\n> ERROR: SearchSysCache: Called while cache disabled\n> \n\ninitdb. Things have changed.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Tue, 10 Feb 1998 09:08:33 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Access to latest snapshot" } ]
[ { "msg_contents": "\nSolaris x86 v2.6 is broken...has been for weeks now, but to date, nobody\nseems able to come up with a solution, or a suggestion on further\ndebugging, so once more I bring it up :)\n\nThe first question is, of course, is anyone else seeing this using Solaris\nx86?\n\nIf not...where should I try to debug this? It just hangs there once I\nhit this, process CPU time remains static...\n\n\n> ./initdb --pglib=/loc/pgsql/lib --pgdata=/loc/pgsql/data\ninitdb: using /loc/pgsql/lib/local1_template1.bki.source as input to\n\tcreate the template database. \ninitdb: using /loc/pgsql/lib/global1.bki.source as input to create the\n\tglobal classes. \ninitdb: using /loc/pgsql/lib/pg_hba.conf.sample as the host-based\n\tauthentication control file. \n\nWe are initializing the database system with username marc (uid=100).\nThis user will own all the files and must also own the server process.\n\ninitdb: creating template database in /loc/pgsql/data/base/template1\nRunning: postgres -boot -C -F -D/loc/pgsql/data -Q template1\nERROR: heap_modifytuple: repl is \\-62\nERROR: heap_modifytuple: repl is \\-62\n\n\n\nfor (attoff = 0;\n attoff < numberOfAttributes;\n attoff += 1)\n{\n\tif (repl[attoff] == ' ')\n\t{\n\t\tvalue[attoff] =\n\t\t\theap_getattr(tuple,\n\t\t\t\t\t AttrOffsetGetAttrNumber(attoff),\n\t\t\t\t\t RelationGetTupleDescriptor(relation),\n\t\t\t\t\t &isNull);\n\t\tnulls[attoff] = (isNull) ? 'n' : ' ';\n\t}\n\telse if (repl[attoff] != 'r')\n\t{\n\t\telog(ERROR, \"heap_modifytuple: repl is \\\\%3d\", repl[attoff]);\n\t}\n\telse\n\t{\t\t\t\t\t\t/* == 'r' */\n\t\tvalue[attoff] = replValue[attoff];\n\t\tnulls[attoff] = replNull[attoff];\n\t}\n}\n\n", "msg_date": "Tue, 10 Feb 1998 08:03:46 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "heap_modifytuple: repl is \\-62" } ]
[ { "msg_contents": "> > I have noticed with delight, that the defined(aix) directives have gone\n> > away (since it is not allways defined),\n> > now there is another one in s_lock.h __AIX with two underscores,\n> > wich also is not defined by gcc.\n> > Please use the singel underscore _AIX since it seems to be defined by\n> > all compilers.\n> \n> \tDarren? can you confirm before we make any changes?\n> \n\nI agree with Andreas. Aix systems here seem to prefer _AIX (with only\none underscore).\n\ndarrenk\n", "msg_date": "Tue, 10 Feb 1998 09:33:47 -0500", "msg_from": "[email protected] (Darren King)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] AIX port s_lock.h __AIX --> _AIX" }, { "msg_contents": "On Tue, 10 Feb 1998, Darren King wrote:\n\n> > > I have noticed with delight, that the defined(aix) directives have gone\n> > > away (since it is not allways defined),\n> > > now there is another one in s_lock.h __AIX with two underscores,\n> > > wich also is not defined by gcc.\n> > > Please use the singel underscore _AIX since it seems to be defined by\n> > > all compilers.\n> > \n> > \tDarren? can you confirm before we make any changes?\n> > \n> \n> I agree with Andreas. Aix systems here seem to prefer _AIX (with only\n> one underscore).\n\n\tSomeone want to send me a patch to fix this? or a list of files\nthat need to be changed?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 13 Feb 1998 01:09:58 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] AIX port s_lock.h __AIX --> _AIX" } ]
[ { "msg_contents": "The numbers are from 6.3.\n\nMichael\n\n--\nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n\n> -----Original Message-----\n> From:\tBruce Momjian [SMTP:[email protected]]\n> Sent:\tTuesday, February 10, 1998 4:41 PM\n> To:\[email protected]\n> Cc:\[email protected]; [email protected]\n> Subject:\tRe: [HACKERS] Profile of current backend\n> \n> > \n> > The Hermit Hacker writes:\n> > > \tJust curious, but do you have -F set to disable fsync()? We\n> > > really really should disable that by default :(\n> > \n> > I tried with -F and it runs nicely. No difference to see between\n> PostgreSQL\n> > and Oracle. I just ran another test which includes table creation,\n> inserts\n> > and drop (source follows). Here's the result:\n> > \n> > Oracle 7.3.3.4.0:\n> > I needed 21 seconds and -345682 microseconds for this test\n> > \n> > PostgreSQL without -F:\n> > I needed 152 seconds and -623545 microseconds for this test\n> > \n> > PostgreSQL with -F:\n> > I needed 5 seconds and 84411 microseconds for this test\n> > \n> > Whow!\n> > \n> \n> \n> This is good news. Is this with 6.3 or 6.2.1?\n> \n> -- \n> Bruce Momjian\n> [email protected]\n", "msg_date": "Tue, 10 Feb 1998 16:53:22 +0100", "msg_from": "\"Meskes, Michael\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] Profile of current backend" } ]
[ { "msg_contents": "\ndon't want to bring up a touchy subject, BUT... does the rule system\nactually work, and if not, what are our plans? It would extend the\nfunctionality of postgresql quite a bit and make it much more\nattractive.. also I'm working on a modified version of pg with perl\nlanguage support. that's right, perl. so far I've got it to the\npoint where you can create perl functions (using anon sub refs) and\naccess your arguments and perform operations (from pg_operator) on\nthem. I'll get the patches together soon, once I add operator\noverloading :)\n\none more thing -- what about making the listen/notify interface\nsynchronous? what must be done? and... and... select foo[5:] for\nelements 5 and onward in array foo..\n", "msg_date": "Tue, 10 Feb 1998 22:41:17 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "rule system, perl and other good stuff" }, { "msg_contents": "\nDOH -- that's another question I meant to ask, as I noticed some\nthings in the new (6.3) code about it.. no, it isn't, it is hacked in\njust like fmgr_dynamic is.. what do I need to know about the create\nlanguage interface.. any docs?\n\nOn Wed, 11 February 1998, at 14:26:23, Vadim B. Mikheev wrote:\n\n> One question: is your perl language support compatible with\n> new dynamic language interface (CREATE LANGUAGE etc) ?\n> \n> Vadim\n", "msg_date": "Tue, 10 Feb 1998 23:26:21 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "Brett McCormick wrote:\n> \n> don't want to bring up a touchy subject, BUT... does the rule system\n> actually work, and if not, what are our plans? It would extend the\n> functionality of postgresql quite a bit and make it much more\n> attractive.. also I'm working on a modified version of pg with perl\n> language support. that's right, perl. so far I've got it to the\n> point where you can create perl functions (using anon sub refs) and\n> access your arguments and perform operations (from pg_operator) on\n> them. I'll get the patches together soon, once I add operator\n> overloading :)\n\nOne question: is your perl language support compatible with\nnew dynamic language interface (CREATE LANGUAGE etc) ?\n\nVadim\n", "msg_date": "Wed, 11 Feb 1998 14:26:23 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "\nI don't see CREATE LANGUAGE in the grammar file... are you asking if\nit is strictly compatible or if it uses the dynamic language interface?\nthere's no reason it shouldn't be compatible..\n\nOn Wed, 11 February 1998, at 14:26:23, Vadim B. Mikheev wrote:\n\n> One question: is your perl language support compatible with\n> new dynamic language interface (CREATE LANGUAGE etc) ?\n> \n> Vadim\n", "msg_date": "Tue, 10 Feb 1998 23:29:39 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "\nokay... yes it can co-exist with CREATE LANGUAGE as long as it is\ncoded into the server like fmgr_dynamic is. it will not function as a\nCREATE LANGUAGE function. this is because all that this functionality\nappears to do is associate a single function with a language. but it\ndoes not pass the prosrc attribute (or probin for that matter) to the\nfunction, so no matter what you say for as 'insert code here', it\nnever gets to the function, so the function has no idea what to do!\n\nI must be missing something.\n\nOn Tue, 10 February 1998, at 23:29:39, Brett McCormick wrote:\n\n> I don't see CREATE LANGUAGE in the grammar file... are you asking if\n> it is strictly compatible or if it uses the dynamic language interface?\n> there's no reason it shouldn't be compatible..\n> \n> On Wed, 11 February 1998, at 14:26:23, Vadim B. Mikheev wrote:\n> \n> > One question: is your perl language support compatible with\n> > new dynamic language interface (CREATE LANGUAGE etc) ?\n> > \n> > Vadim\n", "msg_date": "Tue, 10 Feb 1998 23:40:41 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "Brett McCormick wrote:\n> \n> I don't see CREATE LANGUAGE in the grammar file... are you asking if\n> it is strictly compatible or if it uses the dynamic language interface?\n> there's no reason it shouldn't be compatible..\n...\n> DOH -- that's another question I meant to ask, as I noticed some\n> things in the new (6.3) code about it.. no, it isn't, it is hacked in\n> just like fmgr_dynamic is.. what do I need to know about the create\n> language interface.. any docs?\n...\n> okay... yes it can co-exist with CREATE LANGUAGE as long as it is\n> coded into the server like fmgr_dynamic is. it will not function as a\n> CREATE LANGUAGE function. this is because all that this functionality\n> appears to do is associate a single function with a language. but it\n> does not pass the prosrc attribute (or probin for that matter) to the\n> function, so no matter what you say for as 'insert code here', it\n> never gets to the function, so the function has no idea what to do!\n\nSorry: CREATE PLangTrusted PROCEDURAL LANGUAGE...\n\nPL handler function has to be created (via CREATE FUNCTION)\nbefore language creation..\n\nAfter that you are able to do something like this (having PL/tcl):\n\n create function overpaid_2 (EMP)\n returns bool as ' \n if {200000.0 < $EMP(salary)} {\n return 't'\n } \n if {$EMP(age) < 30 && 100000.0 < $EMP(salary)} { \n return 't' \n } \n return 'f'\n ' language 'pltcl'; \n ^^^^^^^^^^^^^^^^\n ^^^^^^^ in addition to built-in languages!\n\nDynamic procedural language interface and support for TCL are\nimplemented by Jan ([email protected]) - please contact to him\nto get more info (btw, there is create_language.l)...\n\nIt would be nice to have support for perl...\n\nBTW, Mark, I don't see PLtcl in current sources.\nIt would be nice to have it under contrib/pl or, even better, under\nsrc/pl (may be with flag USE_PLTCL in Makefile - like USE_TCL :), -\nwith automatical installation into template database by initdb).\n\nVadim\n", "msg_date": "Wed, 11 Feb 1998 15:19:48 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "Hi,\n\n>\n>\n> don't want to bring up a touchy subject, BUT... does the rule system\n> actually work, and if not, what are our plans? It would extend the\n> functionality of postgresql quite a bit and make it much more\n> attractive.. also I'm working on a modified version of pg with perl\n> language support. that's right, perl. so far I've got it to the\n> point where you can create perl functions (using anon sub refs) and\n> access your arguments and perform operations (from pg_operator) on\n> them. I'll get the patches together soon, once I add operator\n> overloading :)\n\n The rule system is wired up in some places. Especially it's\n impossible to have the values of the OLD and NEW tuples in\n UPDATE rules. But we have triggers that can do all the things\n rules might do.\n\n The function and trigger manager in 6.3 are prepared for\n things like functions in perl. There is a new command CREATE\n PROCEDURAL LANGUAGE. Look at the create_language manpage for\n details. Except for user defined type input-/output-\n functions anything can be done in a procedural language\n (functions, triggers, operators, aggregates). It would be\n very nice if your perl support makes use of the API defined\n for procedural languages.\n\n I've already written a call handler for the Tcl language that\n supports functions and trigger procedures written in Tcl\n (it's not in the contrib up to now, mail me if you want a\n developers copy). The procedural language support of the\n backend is a result from that work. As long as your perl\n stuff isn't able to handle things that cannot be done in C\n right now (like returning sets), there is absolutely no need\n to patch the backend again.\n\n And I currently work on a pure PL/pgSQL handler independent\n of other things like perl/Tcl. This one will also be\n implemented as a handler for the procedural language support.\n\n>\n> one more thing -- what about making the listen/notify interface\n> synchronous? what must be done? and... and... select foo[5:] for\n> elements 5 and onward in array foo..\n>\n>\n\n In contrast I would vote for adding a really async\n functionality of the whole frontend/backend protocol. That\n would fit much better in the event driven world of graphical\n programs.\n\n\nUntil later, 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", "msg_date": "Wed, 11 Feb 1998 09:20:59 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "Vadim wrote:\n\n>\n>\n> It would be nice to have support for perl...\n\n Asked for implementers sometimes :-)\n\n>\n> BTW, Mark, I don't see PLtcl in current sources.\n> It would be nice to have it under contrib/pl or, even better, under\n> src/pl (may be with flag USE_PLTCL in Makefile - like USE_TCL :), -\n> with automatical installation into template database by initdb).\n>\n> Vadim\n>\n\n Sure. Let me do some final tests (checking that PL/Tcl in the\n current version works with Tcl7.5 and Tcl8.0 with the same\n sources). Then I'll upload a tar for putting it into\n src/pl/pltcl or contrib/pl/pltcl (expecting the source\n directory in ../../src).\n\n There is a little test suite included into the directory. It\n checks triggers, functions, operators and aggregates. Since\n building is optional, I don't think that a real regression\n test is a good idea.\n\n Time is limited. So the USE_PLTCL part is up to you - sorry.\n\n\nUntil later, 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", "msg_date": "Wed, 11 Feb 1998 09:45:21 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "> Sure. Let me do some final tests (checking that PL/Tcl in the\n> current version works with Tcl7.5 and Tcl8.0 with the same\n> sources). Then I'll upload a tar for putting it into\n> src/pl/pltcl or contrib/pl/pltcl (expecting the source\n> directory in ../../src).\n\n Ment ../../../src :-)\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, 11 Feb 1998 09:47:32 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "Jan Wieck wrote:\n> \n> Sure. Let me do some final tests (checking that PL/Tcl in the\n> current version works with Tcl7.5 and Tcl8.0 with the same\n> sources). Then I'll upload a tar for putting it into\n> src/pl/pltcl or contrib/pl/pltcl (expecting the source\n> directory in ../../src).\n\nNice!\n\n> \n> There is a little test suite included into the directory. It\n> checks triggers, functions, operators and aggregates. Since\n> building is optional, I don't think that a real regression\n> test is a good idea.\n\nNice2!!\n\n> \n> Time is limited. So the USE_PLTCL part is up to you - sorry.\n\n...Up to Mark (?) - sorry -:)\n\nVadim\n", "msg_date": "Wed, 11 Feb 1998 15:57:13 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": ">\n>\n> okay... yes it can co-exist with CREATE LANGUAGE as long as it is\n> coded into the server like fmgr_dynamic is. it will not function as a\n> CREATE LANGUAGE function. this is because all that this functionality\n> appears to do is associate a single function with a language. but it\n> does not pass the prosrc attribute (or probin for that matter) to the\n> function, so no matter what you say for as 'insert code here', it\n> never gets to the function, so the function has no idea what to do!\n>\n> I must be missing something.\n\n Think so. Using the dynamic language interface, the handler\n is called by fmgr_pl() and one of the arguments is the Oid of\n the called PL function. So the handler has to do a system\n cache lookup on pg_proc (at least the first time this\n function is called) to get the prosrc attribute. The AS '...'\n text on CREATE FUNCTION will be found there for dynamic\n languages. It's handler specific what it expects in this\n attribute. For PL/Tcl it's the procedures body and it builds\n a Tcl proc around it after analyzing pg_proc and some other\n system catalogs. The Tcl proc's name contains the Oid, so\n overloading functions with different parameter types isn't a\n problem.\n\n A few minutes ago I sent down the PL/Tcl directory to this\n list. Look at it and reuse anything that might help to build\n PL/perl. I really hope that PL/perl and PL/Tcl appear in the\n 6.3 distribution. I'll do whatever I can to make this happen.\n\n>\n> On Tue, 10 February 1998, at 23:29:39, Brett McCormick wrote:\n>\n> > I don't see CREATE LANGUAGE in the grammar file... are you asking if\n> > it is strictly compatible or if it uses the dynamic language interface?\n> > there's no reason it shouldn't be compatible..\n> >\n> > On Wed, 11 February 1998, at 14:26:23, Vadim B. Mikheev wrote:\n> >\n> > > One question: is your perl language support compatible with\n> > > new dynamic language interface (CREATE LANGUAGE etc) ?\n> > >\n> > > Vadim\n>\n>\n\n\nUntil later, 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", "msg_date": "Wed, 11 Feb 1998 10:25:44 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "\nI certainly was. It should be easy to switch over to using the\ndynamic PL mechanism. I hope to investing some time into it into the\nnext few weeks to get it up to snuff. It will accept a package\nfunction or an anonymous sub ref as the AS clause src. Strings and\nintegers are passed as scalars, and everything else is passed as a\nPostgres::Type. I was debating whether to make them\nPostgres::Type::datetime (to add type-specific methods in .pm files?)\nA scalar, array or Postgres::Type can be returned and will be cast\nautomatically (if need be).\n\nOn Wed, 11 February 1998, at 10:25:44, Jan Wieck wrote:\n\n> \n> Think so. Using the dynamic language interface, the handler\n> is called by fmgr_pl() and one of the arguments is the Oid of\n> the called PL function. So the handler has to do a system\n> cache lookup on pg_proc (at least the first time this\n> function is called) to get the prosrc attribute. The AS '...'\n> text on CREATE FUNCTION will be found there for dynamic\n> languages. It's handler specific what it expects in this\n> attribute. For PL/Tcl it's the procedures body and it builds\n> a Tcl proc around it after analyzing pg_proc and some other\n> system catalogs. The Tcl proc's name contains the Oid, so\n> overloading functions with different parameter types isn't a\n> problem.\n> \n> A few minutes ago I sent down the PL/Tcl directory to this\n> list. Look at it and reuse anything that might help to build\n> PL/perl. I really hope that PL/perl and PL/Tcl appear in the\n> 6.3 distribution. I'll do whatever I can to make this happen.\n> \n> >\n> > On Tue, 10 February 1998, at 23:29:39, Brett McCormick wrote:\n> >\n> > > I don't see CREATE LANGUAGE in the grammar file... are you asking if\n> > > it is strictly compatible or if it uses the dynamic language interface?\n> > > there's no reason it shouldn't be compatible..\n> > >\n> > > On Wed, 11 February 1998, at 14:26:23, Vadim B. Mikheev wrote:\n> > >\n> > > > One question: is your perl language support compatible with\n> > > > new dynamic language interface (CREATE LANGUAGE etc) ?\n> > > >\n> > > > Vadim\n> >\n> >\n> \n> \n> Until later, 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", "msg_date": "Wed, 11 Feb 1998 01:38:08 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "BTW,\n\n recently I hacked around on the SETUID stuff and it wasn't\n that much to do.\n\n I renamed the obsolete and unsupported proistrusted attribute\n in pg_proc to proissetuid and made it default to false. Then\n I hacked some code into ExecMakeFunctionResult(),\n ExecCallTriggerFunc() and utils/init/miscinit.c and voila -\n setting proissetuid to true works for 'sql', 'C', and any PL\n functions called via a func node by the executor or triggerd.\n It does not work for input/output functions and I haven't\n checked about operators and aggregates. I don't think that\n types input/output functions need it and for the\n operators/aggregates it must be that easy too.\n\n What should the syntax for setting/unsetting proissetuid?\n\n ALTER FUNCTION funcname (args) (NO)SETUID\n\n looks good for me.\n\n But before doing anything here I think we should also be able\n to make a view setuid. I haven't thought much about that up\n to now. Any ideas how and where this could be done?\n\n\nUntil later, 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", "msg_date": "Wed, 11 Feb 1998 10:41:27 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "On Wed, 11 Feb 1998, Jan Wieck wrote:\n \n> A few minutes ago I sent down the PL/Tcl directory to this\n> list. Look at it and reuse anything that might help to build\n> PL/perl. I really hope that PL/perl and PL/Tcl appear in the\n> 6.3 distribution. I'll do whatever I can to make this happen.\n\n\tHrmmmm...I always love these \"blanket\" invitations :)\n\n\tI've installed it as src/pl/tcl...\n\n", "msg_date": "Wed, 11 Feb 1998 09:08:58 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "> \n> On Wed, 11 Feb 1998, Jan Wieck wrote:\n> \n> > A few minutes ago I sent down the PL/Tcl directory to this\n> > list. Look at it and reuse anything that might help to build\n> > PL/perl. I really hope that PL/perl and PL/Tcl appear in the\n> > 6.3 distribution. I'll do whatever I can to make this happen.\n> \n> \tHrmmmm...I always love these \"blanket\" invitations :)\n> \n> \tI've installed it as src/pl/tcl...\n> \n> \n\n\n CVSup'd that - but where's the test subdirectory gone?\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, 11 Feb 1998 15:28:08 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" }, { "msg_contents": "On Wed, 11 Feb 1998, Jan Wieck wrote:\n\n> > \n> > On Wed, 11 Feb 1998, Jan Wieck wrote:\n> > \n> > > A few minutes ago I sent down the PL/Tcl directory to this\n> > > list. Look at it and reuse anything that might help to build\n> > > PL/perl. I really hope that PL/perl and PL/Tcl appear in the\n> > > 6.3 distribution. I'll do whatever I can to make this happen.\n> > \n> > \tHrmmmm...I always love these \"blanket\" invitations :)\n> > \n> > \tI've installed it as src/pl/tcl...\n> > \n> > \n> \n> \n> CVSup'd that - but where's the test subdirectory gone?\n\n\tCheck it again? I just checked here, and its been\ncommitted...yup, just checked the physical CVSROOT directories, and they\nare there too...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Wed, 11 Feb 1998 21:28:16 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] rule system, perl and other good stuff" } ]
[ { "msg_contents": "Hi,\n\n hmmm - no access to /pub/incoming on ftp.postgresql.org?\n\n So I put it here.\n\n\nUntil later, 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\nbegin 644 pltcl.tar.gz\nM'XL(`$)GX30\"`^P]^W?:1M;Y%?T54\\)^V%N#02;V)HV[Q8`3MAA<P,GVI#X^\nM`@36%R$1/>P;?_WO??.C%Y(@!]QN[O6\\0-)<^_<][TS&@T+TQN;^\\^^ZL%J\nME:,7+]@S!L=!A?Y75?Y?'!7&#FM'%?7PH'9X!'<KU:/,_;BV2,<ONMI#F//\nM%C/WD[FFW62D/?LO/!:D?T]WO?T_A_YK\\+E:K1X</NG_D?6/?Z[F'TW-FI77\nMR^61[52.:S5LO5_4`WTK[XXJ,+=FHKZKSSI_ZL?RMC1-4]G4]\\:>X9M,3(\nMJ[%FFE?7FC4Q=6=GESFZYSN6R^R%]LG7E9SFLN^[SK[I@TM]TEV^Z8QVB?H\nMLFL7E1S:D:_-=%9L%+]39#^>`P+7)VSAV&-]XCN:R<&!`V@HN,46@CMV)XO\nM#+Q?/._L#P$\"T#][.A[$_UW0].)AW7^#_ZOJB\\,7_Y?>?'D_X]QE$IJ<0:\nMPCFOX5<;F;K+?!><U+#X)3`,!G[O&+JK<(#AU>COJPRPZ46\"\\>8:\\Z2P36.\nM@$UM!QI-/U:9BW+0U`V=>RYA1>-<NR/3;2V1C<>P8=`M!$-W4,$,84,2^9\nMYN@0?J:ZHUL0,\"0.0ARC1=V\"%K7,VH26;G[63!^H,JPH$NQ.4+-'+8.^L:5`\nM`QA-T[XA;#IG#G$@K\"`_#FM8,]DNPH`:!6\">;9>)$(`3E\"'=4T,W)XSS9\\6H\nM=\"$$6IZY9&/;^JP['`=B\\!<+B(UCS=49Q'/#<N'FOK^8@(9)9#(2DVRD(G<4\nM]$7\\F#,LK[8G3]4<2,/942N[_)+WQ<OQ2[7KK(;\">P1=.K#H4-%\"VP>P/.V\nM50D.`HYV`=A!QG8U'M@$P9WG.E9X.MZ../9$[Z%P-\"\"1AJAJ1-Q!@B4(\nM?ON8-P=ZI.-%MA9F&Q'MFTRS+1$#?C#K]]\\,RIS7MRV9OFY>_L5T5641\"Y\nMX=+.`A+D+OO@+HRKA:,OP%)R</SRBY+#1GD7S&WLL6K,&:,M\\+BY!K,E[;%C\nM]LLOA2J#[,N-D,[5/+7'I@3SS(@Z/_]DJ[_3G\\529C%=_Z.,%XUMWT(:\nM\"B&])K#$H.\">GFI!`P7+/8]JT2YY,)A>2\\?Z49>G.:5XDHY$>IMZ!BS&;A&\nMH`KXT<%UEL+C49]K%`?^.Q-Z&^G@F_I811JZSG_P.KT`24/4\"=B_UB>)@:\nM#C\"-Y4T$)E//N>?PP\\X!06AW&\"$F_L(TQM@+6F'$]G+/TVR#W]QL(2EFBI:P\nM1_K?14Z!]WP`EC\"!+&(CT302R;-(Q7NWL&5LGF[Z-E14T;7CIERMAG'U-FV\nM#,_03.-?H#Q3F[F1FT@WEYEM3JZ@`U9)7+?TFRM0%ESG-VX,;WS-\"L,W5_8B\nM8N'M[J#5'Z+\".7+^CYWYKI>I=3`H#A8#6NT:)/V[DKLX;]:'K5OW(,!L\"W.\nM,950,IMR%W(A)Z)Q:QY/0#<ZF\\?02K\"((8\"!]#I-F?0H#T8)0X?XX(+S01+%\nMRA]S7Z';>K_#[;@`L/SC)?OFF`>+7-28XDH!\">12Y<UO;%'->Q1?9@>X:?9\nMZK2R%$+B3G4@(3?0#(?/T'V$CM^C-DWQ-494&&4:Z580\\]/L<!Z$&![4([H\nMA`@1@OJ'&,.CS&<!U+437O&WM?[71;UXWS<'(O%$&VQN!<Y5^%<5'Q$1CZ;\nM;2&9-+97Q)P,4>L3FHQ<7`\"!A49-1TH@R'#KF!<L!X@$RP$V8#E2(X^6(A@F\nM>9=AK&XMO6NT;BB9IX:ELQ#(A22F.YH?!)`<[&8=FQ_=AV!%'FV]V-&F@W&\nM^#S/1G,E$_]X.8J%/J](PY(?#3E(7:@#78/PZ-@W#(7L4Q;E<P:ZDDO)Q6OR\nM?':^=\"-5-97P%E;1V&ZL\\X('SF\"`$JNIP2$A=E_;O@=BTR8HR)MK`T9)\"VW,\nM1TB9I8.Z?>EPSC,>E0PHF]M5\"UF97[UE;:C>L3;<,J$V^-B%PU,&234,.S\nMB.S97!G!.8SWHI'E,M%#1BB+=[/!G:7DU@2TRY2B-;?JSG>-8VH^37Q][I!S\nM>V+`0!%,%Y#3^`.0K/KM!\\UQM\"6;<;%>;N_$ZD8G#IUH>]=5MW1=[HRIE:,<\nMB<OJ`U%$ALE`R2@^PH;A-<0\\DGPC@VQKWXZ;/6IC$E'0H-T;.C:?#9$D@C8\nM<:Z!%S]H9+:<'T`,XXXH0&NZTLQN1`V3!/LC<^T0-:-ST/(+LL;0H@V]5)F\nM9=='$/?V(21>PPM-IZMA30&O7@'D2M`)W$&@%=I$V+!6/]B+%.JUAZGKV;IQ\nMBDA$=^`1(+-Y%&BC9ON'#E-$W7,E?>R<D,0G#%0R1H_]\"QI)++J-SG.F$\"I\nMSY52-3O,1VO&AQD0)-A_V-H_$WEVF0]R$P7<ENU0XA=S)+:\"NUOH^3$WIU\nM)TLS9'<8NUBW-VPW6FRNNRYF\"!SQT6@/;G%K=>VYOK8L+JR&`6G#:RKCS/R:\nM62\\+2A%W3@24\"8.&<LHU.F_K!&$WC+?Y@(5L`7(AAP)<5=%7$QS%EBT$M[6(\nM9$,B\"@2FSD<8?:7V0JQO]FU_NWQI$]@MKAC>ZI3O&.$S,V5_RAP^#[!!\nMNLZ:I>7Y?8M<>XH,T%RB(&/?T4TYF#(<=&W>A6BO,5?[+\"8$B91K=$>PDCF2\nM^:9\"-G8#(V6JS^/IWFR)8VU^D#=&=,+F,`<$D\"P$=LF8\\VC='?^N(Y&A3_\nM]`0%>PK(1[@$-7#?F'QA'_0O\"ZC#35.W9AXD\",V9N9=LGZF7T99$11#4(<1@\nM[6X:UD3_PD%8@>.+3OPF\"Y=@I#`;J0T>2_\"I4BY(%/T;6Q89B>^`?+T%W`1\nMGUPM&:<`P=#6D@^2-UEPBO()KE],!V:&.9T<GX%M>Q;5KT$_Y&\"E!SSWA)C\nMJ070YYC_^1K)LOS8\"`N><L%5Y0XQP[X,)B\\`DDL2)^#5,`F41\"MMCN,I%B\nMZA.R+=XSR,I#W^`7HK)P&B0?,$J!3<%`D3,O!PI`'WYD`S!$)AQ7DGTW+'M\nMCZ`%/MVMFQ,76:4ARF)VI7G0\\0A<?26W1<)A(,MA&4#1&A.U$TD8L!'BX5X,\nM,:O3^=C$\"8T&@>5X$&Z4A4\\P8E8_/#)MZ&Y4-1EL<C[@$[0.1MEVYCPWHY9\nMO8SHT8=RL49T/15C$AVU!'(!YS&P!;CSEV'Y`6$92HMJK+0(!Y)Y_BQ+$`HC\nM$]VUBD6/74,X`6/GSPN1B@FUR<L\")&D17+])O0BS`'/WYY`;^$6TNABX:Q#\nM6FU!$;P8(T=-&\"L1V2!+3R4W`[\\,W\"TJ&?'ZF=6X97$'U<2E3\"H'AY/-?G\nM(]T)GUL3X^3@$+S@ANXX\\%ED8I?WFC0K:0;2C,A6I&:YK=BD!B;R^!9++YZ\nMA5J+,^%@HBX(?&EU.0],0&3_AKO`SE#F;`Z.@,A2DB3JD:Q%6`E_&\"7$#$\nM\";B<B%F5(+I&)H6D_B06K.P_6[J'$TP@Y(FEO/H80J+H@N8[,EKX_CS_@B\nMQHX1!WLNQ&2,JP8@@@11=8LISKO/8FQL-7\\!W^2M:F2B1<611SRX;P/#O7P\nMOY@!Q8^1_VIQ-YTR=5OU`>B3%U+65AZC<TK'9JDPE_NHU_PBJ1K!!#26G\nMIWAZK4)F+:@9LU-9_4R,S[?O9__V_8`[[MR.E>KZ'K39S-%GI%7H0OL\\8SOX\nMO`@[QJF`B!SWX/I(`T?#D'S,^#H%:(KGU>@%A%4CL$!ST%\"--`2WT$QL'&D+\nM<L1;A@5#?MN:8/-BI:CD=K,H=OWYO2F6O553>[email protected]7IUK00(_S7H-V7M\nM@+>Z<F\"Z5D?\\N0AXSP^OB5=3GWH8:4,.P\"&O$Y<\"QXK(Q/0X;W_.]7\\.!&SX\nM__!]5\"MKUW]7JT<'B?5_ZN'AT=/ZO\\<XGG^S/SL??=:49HGW?I9ZY@OND5+\nM4\"\"V9!-^`U%T<?7-LO_%0[6A-N.O:0Y),@S&KHY_\"&#%ODE0EO,AG)ZS$$\nM<L4A+BW,0,\"];SU\\VC)B8D#@6(!26>D3UD!\\M>)=>ZI9-$BR+T@^O#9'9%W\nMW4VX@S6T,=1]W[)HG!993;FJ:0GD(F6B(Y]3\\!E?/\"H?O]_546!0#<QIE-^\nM&;0%]2P^Z)\"-OM^?Z)_W+=\\TJ3W[#BLXBP(?)PP_#74<8N,37P#ME7\\LYQ7=\nM=/745E,-\"]@2,Z$(9YI'G4/PU;T;7;?R\"9`X421`05A>F1I/2[;_E.N_(^;V\nM6.]_U`Z/JI'XK^+Z[Q>5ZE/\\?Y3W/\\2(`<J38/&C?\"Y4Q6U?HFJ?0I<<(H5\nM_U90ZIV@#FX#I8I1R.TH##4.T&M4)@\"IMY)B.J=AC>28CJG82HWDF(ZC9\"\nM%/,Q?XTM3?X.9_E7UK^[U[8/B8CFO!)+<U+1JDZBAO?CZV?A=P$]AO;^4B$\nM`0%-/Y@;B_3G.9KEBFERPPK&Y5ZPQ@#SY4:%K72,4\"EF14\\M)&#1<_0Q0X5%\nM[&M%]BD@:A1$W0KD(`IR($5R)AED$RX<W_IHV3=6<AGR1O2U/I:.N-J!N,D\nMNW3&U0S&\"43=\"N0@\"G)+QM7-Z&M1]+54'T!QI5]&DPX63W!K%D\\$:-!./O:R\nM&%^NI4;7:DF7_VX#FNHFFH23970Q)<Y<%R;NE>W@4OI#^'B=LQND&DF]BG\nM)BOLKZ#9Q+ZPP#0VU(W=J]O`I?1'[&?$RXSXEVY9&88E+HLI?&D.;4;,0>T\nMJ4DJ%G4S%I6['+X4Q=<GX`AA$DQ=].]Z'1PWM/3M0FS<=G]1NKC\\GWY<@MN\nM5D&4X!'?#Z]9R`H4NWBAEJ4@9CL3R!&C)<?EN\\C1#Z\\S]);96ODZ]7\\P7GO@\nM^9\\U]7_U1:UZFC_#XX.G][_?)3CIXM6_^=7[&ZC@-L`J_<!/K@#<'95>QM@\nM]3[`VY.MWD?:ZGVDK=Y'VNI]IW>1]KJ]M+.R)/X]S?\"E#A^\\[YX;)L#,E.I\nM]&TIY4B]F';@E!E1`?I.4B$XV4A%B$]/XJ#>Z!0?^,6<'\\4ZOU1W)F1G4-\\\nM0.KNNOL1_U3V0^O0>^E?%Y6WA_%_>R'U^3W1Z'>'\\7#V<^=)QZVC-F9\\P>M\nM?K_7?\\42;W(64(+H.@;)NG];37KL\"WDZN3#MI\"K<Q#;0JY.+TBYR,4E'@<3\nMJTH\\,6^1A7_S5,2VDLS$MM\"KDY,;`NY.N>0@TU+@TUQ;R4:07^,5ON&KD\nM-WIO(2W81:);:HA3N$0)4V9FY(W41\",UK=%!HM%!TE</-L1Z/A[^>ORE1>X5\nM_M)B\\PI_:=%WE;\\'FC':\"MO6$T=1\"RP&2Y^28DE_Q562<M?)IVW!,^:@,DA7\nM;T/Z`TUC;85MZ]DL_L(`<\";?IZAN_3I%MDRWFQ#;%CQC7BR@7+[F4$U[R^%I\nM</#'#`X.'K[$C%1-VU+Q\\G]M<'#P!Q;$(8J7]U?^W357VYSJGTJ9V_WJ0!)\nMK7=>)DE7TTE?]_A!W%,BIHX=#M5Q)6NLQA;B70/6W=:BR'F<\\%%5J%E6'\nMMZ8JY<E%.I7L]K)+1[U=24+M7QV\\O>Q;?ISZ^]:%T;B`H?LZS\"3C?`;OE`\nMYJD&>.@:X.$2^`-4(E^Y!OA/,M[EP;W3<I_CNHB+`W^I]=_]EOUYEGK_2Q\nM:?]O53U//]5T=/SW\\?9_TG+>[&!6OPHS%WKIDF7Z+N^H;'WZ7DNVR7^0MV\nM_\\#53X[/=[!TQXZQH%,\"P?&XW#2#OUEJZGSSD<1:\\)F&\"S;\"A>KT3C]TP'O0\nMK>UX8_K_V?:1QV5]77ZV.#_!^I1+?C^AX,C^OZ'([7VY/^/\\OY/Z:$.Y;GR\nMG$E#@H]PR+/PY6%Z-<>]IKUP[='_0S@@L':SU1VV3]N-^K#=ZW+@PEM=@PKJ\nM%2M@DX<CDSH<ZA#FYD\">&,(0<;;KS1P=7[1Q;=_!W4),XS.PH@SZC6:[?\\S\nMY7W^XSICQ;#&IC_166&'W]X-O@\\,^V19BJ\\J[9H-X%[N\"\\3ODT4O/`-XFC8\nMUM28E=UKQ@47(`[002,`=@6^\"_S^`_YR?FQ57+!3&D1QB+/&U!BSS[KCTD;'\nMM.LY\\AVWS1&#K[(+38X=/5)&?$/&YVK3OOD:G#>:AR7.OQ;'HP1Z'J_E:N\nM\"\"H:?,?D6/?7]LT<MH)UK8\\\"/@`-;8=A[]S']U>$QJC6=\":=R<;N0;)FY\\\nMRI:X=YJ[=#U]#K2Q@8&[:P,C;ER:(`\\]0Z0\\0>TIS_EF\\M&1!M]@7A(\\HE?S\nM44Y`R8WF0OH;`WGNU#<!F#9LA@09T+\\I;\"#DAJ\\)5E=G)ZV_[G[O>7LJWD\nMHG<ZS5U6LEGA!U9XS>(W4,J#77%12CQR/MC=K'#=HET/:%FZ6(2,;3$/F[1/\nM@J1=;C8-#&\"QRW@7+F@'4^8^'Z-,AN(W>DF]AB`Q3<`S.V)C]\\,#%0?38H\nM5AMAWI[HH%C3)<-IG';J;P;?'K-2\\[R#/%QT?^SVWG=!.N?GO?X0N&DTV+%@\nML-'854(=F$'.#[_\"=V(SJ1+A4X6BHZ#@73XKCO3^<PI7\\>12?\"1-OZH6Q,E\nMO,L1-5NGB#3K#>;\"-)]6\\N/@7O#),&FIW>R3\\&\\@L&)DM+FQLHM67)M\"%&\nMR5!&#HRRX.V/>;@K[#0[PD`4I=T];7=:`Z2#-]IE7,D+TW<9?]_2%7431((I\nMF#EK_?,<X27LMR'PJ[)]7-@1]W=W%6#5F')\\X583H4?\"#2`:`NJ.0`4$)8/,\nMR57WI]_7(T]Y/2FKEDHDC%^@;.G)6F47SBRBHXUGO4.6X^F$.(P;#>Z>PR\nM_I%[0.]\\2!XA10.?6H,AH7T0RH-[CZ]0/AP]9]IC'7+!77ISMS]&O5?;5W]\nM5XU]_Q.-_XYJ3_7?HQS!\\,^UI]X-#MP,3B+)6UIP!]7PX\",O3=P$Y@2>ZO-\nM1[XS$X-!7)<?R4B,[(<VCSIL,#,0L&,::YKCPUZDHQ[)U,Y0AO'R&Y]\"YI1\nM\"`1C-/![32:&.S8U8\\Z_B,:`V/;9F.!.4X0Q0H'F>]>0\"+\"0`VIGCF;A\"\\U`\nMBN%2]8.48(&SQX@S^$<;YB[YUWQ`-VSJ#V^.Q,3WB!2OB00;QF`&!(C;=U#\nMKWW)W7?@[I(M?`=J2$\"S<&P@%=_$QO'-C'C^TDL>+B&H/V#'1TN0T%%$@`\nM@)R2U+`Q/K+F%!$>3@\\'12V)W#@A#%#IC8`BVCT.:0FXP@JIS+HVNW$,S]/A\nM+M2X.CP)QGE<@`FF&,O-=-;`DM0!6`?CO[)IXUL)8^\\DA0B-_X%=T\"PH`PJ\nM]VD;XK'&JSJ+J(\"G&M+K#1CUH7FA0@-AS`(37F+1T;X#HDN3<E=<V)YF_2\nMH[5-=\"SR1D(.:`1)!<C-=GE[)(5A`G/`QLBLQF263&@S+)86N)L`,$R;YM`X\nM)A@L++F-\"R-L=UFWQUKO8!##!F\\AK;'AVQ:K7PS?]OH,?IKMP;#?/KD8]OH#\nM=M)BG7;]I--BPQZK=W\\F#.?U_O!GQDZA,;3NMQK#/<(;?L;JL%WO\\,L-&C#Q\nMLUZ?,#1ZW4'KIPL<1]4[<-ZLG]7?M`:,U?OM0;O[!AI>#.'/2/BV,6@)<_:\nM`\\(PZ)T.W]?[K3W6'@Y8L]>X.,-.<$2VAUP`K:S9ZK??P:5W@!G0]%N]TSUB\nMG,M!(A>LL[?U=RU@N-5E]>:[]J#5Q!ZQQ7EO,&B?M#MMX!HN#2X:;PD#IUKZ\nM=A15O=MD<4%RD>\"@L=/YF='-1J?>/F,L$\"NPTZ^#1%J#/1);YZ()HH#/)RB+\nM;@_^=-IG[2$0!MK8$UVVS\\X[[58S@8%+ZZS5;[R%\"W5./(\"<MH?=UF`@E%?G\nMNFPW+CIUKICSBSYP\"U(E%KJ];@FJI3Z0T4+QEKD\"`N$S^'S>[[T#!8.PN%CK\nM79:O#^!.GIW4!VW@!3%%3\"Q-.\"1ZZ(XP]$XZ[3>D211K#ZGB?:#7UMO=8:M;\nM[S9:9&>\\6H>/?)O?`8\\.K>Y;;($D#\\@:SGK-8,`^#_58O_1]1_?\\V7F0\\PL\nM6\\[CUG_J427V_7_T_E>U]K3_PZ,<Y84)&BA7QDH9JHXJ_[2PF4H?K\"G\\7BME\nM&(&_;2IE=\\'4N5(NTX7387#!,]D5?TCP[_:^M:N-8TMTOJI_15DA0<)\"!O(X\nM<^P3>V20;:T0\\`!.;J[-9352`YU(:D4MV68R_N]WO^K5W1(\"&N32\"LQ4G<]\nM=^W:M6L_3UKJ-8O1#O][5[W>/7F-9^B75.?#A=J@9N#+^M>Z_K=0'_@9P)&-\nM@-L(6H5A?42&&`2V@\\!$#F+!#+6T89H]BX,FL$,/OY&&6<4AUW>%5W>2GD0?\nM)\\A08'Q\\$5,Z4\\CVP!S=D),D8EELUXS1RM(H^+`?PHR\"\"`42)&#=Q`?#E$DH\nMF#G#R$#A(4T+)<9A:=Q/YX@'QB:9@)DY\")@=\\-4BV>VS0@:,)MN-)K05##.\nM-'!PR`CR`.!<17[Y/$EZ[G.`2T-858Q@;.(F2HQ`8K\\Q7O%9%,!LU^,A=#\\\"\nMEA7%11UXU>O%.MXR,VB#&&MI213%C^C^@BM!H=I@`D!2R9EF\"L+WX=QGP5:\nM<'4@\\1L+7/0:)#`\\./4[.DW+.(S3-#A/=7[.38D;4ZYGF(^\"XS!,XWO'1:\nMI.$/PWZJI:F>9$6(2>,\"BJ$PO[A>A^`TM>5I\\.>Q!NU]PO3D(,V@`)CU=D!\nM6/Q&P6ZW<413`#/RT-/A:`QWF7YTSNSS6%CC@-J<<X?,$,M%Y;6;T$0XDW\nMS#34>?P^&C:UQ.1TR(:8!0FW`SR!H/1]/)^J-D.L&_%F&U@2^\"B&+!`@HU\nM3T>PXV.U_DVPHQ]+#DJ\"AKM!\"4E#!>.ZT&A`PEV`B-.D:@'0_6YPJ,&IB'FC\nM'JTGQM>`.ZB9G+M+0HNRI51J=0$JH&4PR(BF+`-?+6_;0=D6[[U?M,5;\nM,-3R:LGQD4X2CJ\\,RR^%&1W#'VDO,4LO!\"BI$!-=%9.'1R+<`8I#8XAD)#-\nMXPBV4:KODHS5SE#]@0S\"B;?=``4F`8_BPP4\">,8P6!UKFDH1O3'71<##T(@\nM)Q>#\"&Z5.C'A0C'&&^MB&D]N,+\"<$F]&_:[4PYY0N0$$$P$S(27)$<FD\"&8\nM8`@:8JD$F18:A+C](:'X8:@((!T!TBV@@!%&+X_&9U$7A0`XD@^XVA\\BNA`\"\nMLC@+'YCPD()H'!`.J7A8NCI1/W3Q-M-&9,<1,-0N.$X$$JJPT@SM8?93D(@\nM`VY^'W`FR$O4CQFQ#>JEG%TCCMM-0U@ZB.`3JQ3S_J;R^X44ED\\@N%9R+2X\nM/5QZ)$B4YA7`P4!LR#IW,6(\\C1K!:F9&),X`!.%DJB(=)$H48E_3[F1J,8(6\nM%G'%1((.:(-P/&4\\8R:K<!3T(P%?$&P[F,84A0/Q0'>IY(4TBP`5#A,?L\\%\nM>R(4B(>$.E&,2(.?Z$HSI[V8^]YE(S\"/`.A@=U@^\"Y$ID]E(;.QN2PT`@W\nM3?HP$Z%%/J<K`%2\"$.&39SBE^-(($`-FYA%D_5-I11(+?2#U4C?I]X&39D%7\nMP<\"A<,[email protected]#'J/3GEPIF&%I^.2@H\"F]BKP#6:$QXYP$`[I!\\3&7Q3[\nM-TPY!CR7@IE@4_J`,E`!PG!.X)NVZ50%NP4)PIR0(5/0N[?4[HZ!X:I+D#\nM&!>NM9D['W96MABF2R`IOW$I3%39Y@O%%QRR$B4B&Y#GL#Z`I0GJWRO#-\nM<:_ED+8%0A,)_VHT85IUH<,\\\\0[2=)80&B@S#-><II97TPQ:9J@-.=PCQ3J[\nM]!(6\\>/C@!AN$N8<M#'#SHLW>]MT:<?V:'\"UM[@D]/5ML]D\\/JYK%A>K';2/\nMWASL'2I3IG6H5MT\",@Z+1=)[Y(K'.!5;7;VGOYIO6R;</@RFMDIX.?+^`H\nMP[.%3T0,KNU,57(M-+P0Y\\S;$9<\"BPA'Z#AF[GEE4\\$4,`N),([^L(ARA$\"W\nM)I,^\"@\\_6@Z88P<RNHB/K_@38I'\"L5CEE.'M^-I9/`B3!T@9R,,8Q>2BIF\"\nM\"[N`RT=[=EYN.&G'&Y8R-_IA=A>&7+!W,^VK`+YD,.\\9Z#7PY0,!1<4$KL\nM)F>;D&JAD8`G/G-VF7I)`%V4T<32F>N,PCH>G!6A)BR;!)<\"$VD>#L$8.1`\nM.49DG8YDMP.']\"`D&L?A7'O9`M)QZ@/Q$[KZK-XGD:2F)D'!5<#[-TLHW\"#\nMY&N&LTYG+Z;3::W]X^O\"Q;1!HS.+N;6!G^8&1I\"&RK4T[(?CRSK&DLXUHU8G\nMJ^[2?K(H@55AB>O0S-<;ZJNOZ))[NW;UN[/5:Z%3@OP=KG8M$@8#<UCB7F:[\nM9`0VWJLL'392=<ZD4$=F$W8[TVS3&6?.0C!\\G6+L!9A)-IJP5TF@H?A'1\nMP@D'_\"$)C.%!:)&8CD/X-4$5>I9\"B:U,BN=I=(9,8=H(D`$BA,8X>V$O<W,D\nM<;W=(G@KY+A]0'H\"N7#R5I&1$ZDDE(/V-'W2VGYG)/H@#JB\"YC3Z25>W$6.J\nMFN+#SN1%<INS1UW^+1VF%M^=PS),G76212!0RTPL8.)=8X'.$R-TQ;Q32M-\nM-\"&X],^G('/3Q!0\\>.W(F\"S)6`&,(?Z=0YWH=)$N[$EN3@\"DQ8SV`BV=T`@@\nM.81V)5Q\\BYFB,?10>?ER_8!X1]1OL!MCM)UF/;VV4[#<D@3L9^A?-5\\7:Z\nMLC,AP0YA+YI)93H()Y.4NFBQ-\"BV,AP.).T17+CRG<4?M1X44#D:C(#IHUI\"\nMNYNP8\\F`!\\%%R4X\\JJY7C:I8+B8%\\MMEAE.P/-\"4SK!MHR2%R^-[P]FB\\)-T\nMDP353<WW2DH87%HN&/6<&Y+-KN-?DWS`('DQ@)<+[_/VB_V#-K+0E'43N,T(\nM$Y'@[E%1N]QEIDE1[C[C9-0)MOZP?[/V/3A$:`%BINWGPRRK;-Z=4;)@GZ\nM6))N+]I'D.D$$[02RFA2YES'L)YP,-(:2\\)59%_-]G>'_%H@I\")00.'Q=AD\nMDH([76(6Z\\6[I'1D;I?<UR/)-5[8)8_Z?+ECH%CP2DA3(E[V&AB8>B;QQ'$\nM<-T2[;O3\">:YF;']O,MQCG-BJL5=!K.H\"9\\4/C,6]E,M3HSEEFR86;-/?&:6\nM#P4^D[%!3U)!FR\\L(#MQ2I=:S:.QG$#M_T!H_D/G=2-`M3Q/V=R\\S?U'K9J$\nMOZLN62!&;Y(9!)[N^S\\PZSY$LM]W\\H'7/'P3'@SW>1$MW1VR@0\"X@DG:%T[\nMZH==.';V3]_'R33MP]4`AZTF$;$!!:<)BD/B/D`:(_E.1R@?XX)F,+P@-F^2\nM.TEN5TXS%!#S!$.;(1F1>9(4=TUB)ASZJ4ZD0QA)D24#O2V&D<4;W,49V)!P\nMN]Y4>U'4(X,6G%`HAK!H^S817PM(SZ(0MQ^R%(1-^8.\"\\H(N<03'97+[=;V\nMR1R@/NO@\"2OIO9Z%'TDSKKHY(U2L:<HI)X&\"P@:SXTDAX8YV]1-,8QIC\nMF&?[!7[1R8MY\"F,2?2-CP.F-<>\\`4)MDBXET\"GZLI@9^#4M:V&9%9Z8E\\?.&\nMYJ)PVW3'M.&X6<ZEJ),8Z;67VYMP#44IDO''\":PVI9^:D[<QF]16VWOR8CH/\nMW`RUYMUR99K.7+\\F(2XLRI1DD[XXA6@U%Y2HJ!18/[\":7^2:=4FK\\IFZ9R3\nMZ)LHY6:@YB6Y82PXVE0UCN:!UV\"$XW`B/Y!(PP;[.'&R\\YB(UYB>\\VC3`'Y6\nM.BHV-!IB-U=DHR)+G-QZKO(,&R1<[]Q^.TBS9VH;HA8%,N6KJ>%PH9(I8\"$\nMG;4)V5M+CK.EE&:L21JDYW1:O8@I926^$`U6$[E\\/E&H(^'XYDT\"%:0U'`\nMF887K3(_&>G_?S-2]I$>_N=O9V`I+Q:C<FZ,&7EM?;0Y-R$DO<11[.C8X2Q\nMV#[I=J=BH(^J.RC5QV216`D[0TU9V@_3\"\\R9CGE#HW':##H3;6\"OQ:Q$US$\nMA^S\\4GO,N]DGF:HBT389^E`)X60$K&40<^-0J?]4-39024%]8V'OXF!NI^\nM]D\\2@C,,8&G$;7=U!=I:E>,S\"P_\"ZHN0VR(1OF94547I5/5DOW!>P:8U\nMEQY,L>4/(=>]-...($21^X=Q,CPG[90`5;M1%#>`XH`E\"]F\"8S%7YJKHY-3\nMPM2/;<\\`?'2-L&ABEN9Y/,;/H.O3%WP)O),YO8,+BNC==P/SPB;MA&'8-%@\nM]S_\"K'U#^)N,)O$`\"/_X$>_VA-5KU`\"M&M)YQ%`F/3&B\\R'NAD\"X8^QH/I\nMP)'/HP-OP.IWVH64S4;8(.XB\"(3KX2$#+QFJ0V!DMH\\LLV<N]68D9FJ%IH-\nM[_!2::!5O')PFB:J^.1)_K9F`0B$4Z)1)&&_N'I2[DDB;G-$+ABJALC-,D\nM)B4&#(B'T^C173T&S-\"S+4C>V+:<&YOHZD$@[.;D'.#RD6/8R=HM007`55\nMMH)<HDPFP6MUK(O'@DZ8B?\"N!LST42Y;&I7LFQ4L0R7OUK7BA),R$&%OHN\nMJ7E47@FU;@0^`D]G]%K'B-HRU#R:NZHV.L8+KT7:GEK%/ZNLV+1HXL`B<.)\nM.X;'!HA-H>];Q(X82,D@MPW(UORLHE4YS.DT8-)=)3F:Y'3=KDGRT'HU^&3L\nM5]:W!+@C(\"!(TKBF<W`P,A''1)>U,],CXXC'YR.FZ`'':='\"A\"1G8\\#P9D&1\nM$ZP^$FXZ^>0R9U;Y\\'7GQ&P[GWQHZLT-H3RTLDBE9\"LL\"F9+QZVT94TVM\nMB^WD<260<Q_9\\)F/?QQ-%56:&0G+(\\(NZ:L$&41M0P-@`$\"$5;72'JHY@Y\nM4F1G)\\E(CNIH@+Y.N.V`[]$:LHDV+8$C<DPW%6UD@H\"G/DEA%5!?(J##=-]D\nM5C/)$B(WA;R^93)ZD_\".A4;D&W=JU2@``!+8`2CS2RPB4P?X!#'NKK/C6BG!\nM>J;3TS2\"=ZB$U])9<YZR%LRF=24S<!'CYX^)T>QSXMDZIB$CP`Y-.>>L8[\nM]YEW?@1M809#G=A;LJ3G9D6B+X.#.(])/!$T6?@\\&=WB0/'Z$-`#@&2J[%#/\nMISYMNK`KMB.KPU6'<0IP0)37P`R)%?UR\"7/(&^I>5E.BC?980=P(B(S`+9\",\nM)D@<R$G9M0#)S$O4\\CQXO1!\"!P`WLB($%]XBPF+L1?@3'NHM;87`3IT&(ZH1\nMRA@94^\".E423,F`?TUU<Y\"@!S@(`&`S3UQ6T\"`(#)\"$4PV63YBLR[J3A,VM\nM\\KR!RO,&#3M43V1U)FPJCA&6`3,KIX9-M>R#YBSL0:S+R'&MCV_G=C\\T]WK>\nM=W+-,!=>5MYD]HES]X\"S:<[M#M8B>Y\\S+<-=CH_[><EN4<GUX\"UJ\"=C'!;4?\nMQV[\"U2_TR\"SI];Q1$:$\"][[\\<J>&I3IP-_!Q[D$S^5'CIR@5)2K_MV[E<UJ\nMM@F<Q[%S;YZ1MUK&PCFR-X\\]?2[T./=>;4[O;X)@#WEM-K!$CL_>GHR1GC#M\nM1!;C9N]9`IK9HY<QDQ:)RVPU:O5P`L:>T:[>B+FOPWU2@/3+[E=^U89/XJ'\nMKT;3C#MQH$UHV<Y6RFAA`XSA#(JDW`R.9N06A%-M3IHD==-SH?`]J>D_QN$\nMYW'7I-=DL_.!T[5Y@G/#^69^'L3@C4<9V2UD-<)>LF]L,E5M?8S+\"-:L`N\nM5TP\\&'YX0`E`[),N,OT9'W;\"#SA2>'3\"%R($+-JWI&-'25(CS,V-UC&4>`E\nMPTC+$&7\"PNR.'`R1GVZ,ST[=@RO8CDMF@88+RSK1-:UDL;3D+KJ(U+PV[:\nMQ!<;C1X:76D$W4,1S/#7Y-&>OGU`SK\\_5<O.HW3)B\"!JADOP/K2V>.OYO\\A\nMWMAWU,=5_K\\;&]G\\GYM;__ANZ?]Q'Y_.3.,V&_DI\"#:;RHDHX$5QD4@?UAMW\nM?M`7'1A$S/F8I^)H\\:7UK5UX.L;1R%!0PL,O?+(1FI9_ZZYU=S$ATW5[L43\nMS9<,DTFS8$BLJ]5VSOGHWP@HRJ)-!@3(O'4C+;E%&&?G-=LO\\KZ(F5\"I9#-\nMKSMF>,@>LD#P60Y$\\B)KTEJLH:>LC.YX(O8(,4JR9D9N5R<)KTT8,%Y9QP\nM_D`E9'GTA)$;YN+#;`0(`KV,W41:@7I;L+YX\"&ANL<BZTK5!X9Y,/!84-YB\"\nMODT`3A:]':A9L<W1DO2&\\;=%O9%I0.QTM<T><K8CMH(8G0+J7%=EB,0>AZA\nMR&#3%\"&;>;;RG$PD<:=(X[3+L2,(L,)_7\"&T;(L^]\"_-6$R_#G(9Y8%UI\\A-\nM[email protected]^L\\PV,STX@!,)PZ)RFBNJ7D@[ST)Q?KDV2=JJWSXJ[#XJY&<N+\nM;J-^IV)T-4#J\\(Y0X%LE7&M0J>@!Y,<44,OH3Q!C@55&$.Q!0T+W`SWAV3,\nM$LO9:=I%5VOMMP4;6$]?C$IR_P)`R87B;;9!:XS&D^9MQ4_FKIQI?=\\:'(>\nM6%0_:^+:U`'P&\"%.TZ0_A=%=DLM2`W\"13(.LW`?&=,$7+U&4TU@0;7N))1;&\nMA#NAK1:.28`F\"B77PLV[Z,4[5A#1NO82N^28U)%MOUHBC2]='$2$8.\"D\nMX2IRP2IM<@O+.)MI+=HLR[/F,HCDO\\#_D]N#H_N,;(-QO_^/;;V?$_E<K&\nM?]G\\[NN-_U#?+OF_>UY_IC$H[X`G9?+_\\];_FV^VLNO_S;?+^(_W%/_QP:/3\nM>`CG#\\;70V-09(J!&WX7L+5?E17-JIJY;\\D\"%C20SU;W&=A$(O(M3P\"N.4O\nMH!A+EP&F(\"+;##^/SJ'5IMI$HS\"[F_(\"4B0&?4:7YCX@BA30;8/77A0-&^4\nM(Q083?V!=G]=]2^UJ26.H^D$]0N]:#Q6U6K^T11598^5A^.J=TH6M>_>LAP_\nM?7<,WWD[R(MFL_GN^-UQ=:$^I)7'!:_PP?H%^G?C/]CXK$(D4L-_6&H_L^?H\nM([\"QF\\$GAMA!-(A(R&\\,58R9NK6Z$G4%PA#EK3+[MWU2A2H$Z7NU<4SOI\"1V\nMAC^AWP08L@W]3C]'3)\\I>\\QC.;#!=ZC_EB)T<D!ET@O#Z[96U&':]4_9N[\nM]X>Q$A_#4%1UO7JL'GRO-ES!,NF_GX@`-]#28QQ5KHW`F%K%UC,C5D^_]T>7\nM`_$P$8T.R(0$T45O0)_JT&%VF2H`(\\MFGE7IBS`F3LBUQ9M?9V:=@;$Z/)'\nM4\"&^%#4+/;LL_/)=%6TXWE45C.A34&'<F56#7J[0BE%I:V)6R>\\6$7?I%=Q\nM:M5+NA-7=N?\"1J^(),7]FACR:C>KX)!3^U2_HUW+$F194HH[OT#,&B($ESL\nM1L,'4>@C36I%WH@XCR=GJIUO\";]\\>[=)[7\"A>C[Q^)BJY^4;:ZJHN97EB[\nMS[:-<)D7#<L#2IKS.`)[5G9(J/#/%S<R6I5L;[EQ_T=!Z%M5YM!);^)2:&T\nM0@5R^[+&@H#\"<768&R4X`\\L)2KP!>^1;N=%+?&DOK4N25IC)@!#>.2Z0.%\nM]!8P=]HGLX!C<T/H5JKD`Y>;,\\5!QC<U7;!>=4B,GN4G)K\\TOK=H8D3V'_\"+\nMABT;9D4P_3@(A`9^GZ&!LZ;:H_AU\\$7/)(\\'E7<B?#`SD[)5\\\\;'!`?UN;,5\nM>%^32G5-4E34AW.,:\\T\\6`J;RU/=BG,($+6\"Z6(00TF_/+WNP[[_+\\0_28]\nM+9/_GQ?__9MOO\\WQ__!CR?_?\"_]?4NQW=>BH@K/XY(ZX4-3-_X'_&(.=WR>\nM:EI`3I[JBXWR<GK78Q'):5/1%-]@BK]$R$SE1%^NUZ]=-R%:AC_G>K=9F;<\nM@-H?.M80.8.)M)QN\\J8:-'?/6@./$'JJWKK6<Q7\\P(%6T<=!\\2E`!Q\\7E'//\nMG)#?[+!P._'F=X`D%YO19VET>_DBH!E\"T\\>TZ,]::%#W9)RCR1HRQW(I])6\nM<9?\"K8M7U>_3B`3A;FP$'U#E=(LP'#+\\LM8KLI!DNJ-O67#'!\\?VYO0$.]6\nM=(6NW&HXE2_4GF'(6%(NRK%UA=&8(J,A&>2O77+'C,>,3!U=!''.X:]O.@0\nM9MZ[JO5DM=].^ESDCRCQU2C..IJOW4/5<M;=-P<J6PB[\\BV\\E;<]D2Q\\?.\nM1I=[(19>X7U5^F8(>RY$\"LUR2J-K;[LA7I]]/ZFIF/!^L:&JS##JO670@\\<\nM(GP5LL-F_QI5936U8,YCP21F.E&<A6*@7]IJ$H+Z)D##]/SDL'H:IK95HG-\nM\"\\L''1T,>(X:(O').QH]IA<<>9L-55C2%;7LV`8;WL#4GR!]3=Q0.,4Q8)\nMWI$`S#'.,J:'M^Y2^]QIC-LT,3+!]RW8'+!OP?_?S<IX[@_S>__38;_W7K\nMN\\VOE_S_?7R,\\E[VI;;HI%\";17EE1%@O@;B,!8>8B[BY9#B3`)D+2+\"+&9EA\nMFEK+?0X[EH.&<=X>CM4XQ,W\\/Y&JT9C^\\P?U>+VNC40<NT\\S%+9U]L54-LR-\nM\"!.A'C)(W`^[4/!WCB:$<LO4,ZJT'B29\"&R]!\"-/XMA(`Z_E?!SR/T&9DCC#\nM]A(.HL@!5JG/2PXET#7#LD<11']C:&XB5XUH?DJ#.4C,&OUO5@YCKVI8THJ\nM&1C&D6_LPM8.O9YQ?,E\"CFUP^1B:Z)@X4^L'P3<GQRS6Q$KIX0D_@'@>,8\nMV&HDXM6C^K\"T\"(Q\"VQ7P0,7TQVS9(Z68Z&'S.PNQZ)F$#+-/APH!15_L9\nMDV/9.#['LX(77`;&@8N&[Y/?,#2K2$,DR#\\%:V-/`HHTVAM$CK/E`)<8LU>$\nMPY`,;]?5Z72BURK`I(;!9T__O<UQ3_I?(/H;6SG][^;6DOXO];^+Z'^W;J9-\nMD@.@0)M$AJ\"B15IJ?9=:WZ76]S9:7PMT];__JVZM!/X+;%L68VM!L\\SHA(!V\nMPFP6X/A.TT2F`KF-.$+Q1QVU<,8BRU5;^'5&'!.?\"A7=)5CX$+$;\\WL9M'R\nMNEK?N=J_[:8XI3=4JPG=N5^G@X8Z:L@:!&U><YO:OMAA\\TKM4P8;:/6$3+\nMTEO3$6E9?:K1HQ+-(VZI9A^4&B[F<2]@O<P+$2PBO_J\"$M#+2H\"O=NJ1D#,\nM=:MJ,L600++9\"4A!A2G=\"KRJ<;$Z%A`,!P3'=?A4N>.V+E?+R<>FM''R5A'\nM6ETU;:(X)+_9^+<A-;#S>#`??2)NL(1TG52`0.P6X(CG6(\"'GB7,WDC$7,\"2\nMX?S,6F9-`0AP5?/F2/%+E,C)\"O?.>3+)SMZ&NS\"&!'EX?)HS>L%%'#W.]I:C\nMQR:N'KWNTQ\\]UBT>?<;B@0T>4!N_42!SVYA1_%\\NSF41%3-3P,J/4>O@QI<\nM%8Z'XXRA009'=3X%\\BHG-VE`%)LUCL#<V-C\"O.T<\"9;;FEYMAA!3)DBSE3R\nMZ2H<7+#\"?X(<V;,Z$X+&'W]\\,AHC5@V)ELF>Y+QAL#$W9QLD<OT)^\\(?SN/\nMEC,Q]:41[(<U+PKF=_0B$I\"?II&C6__W)INH8[V,[`--A<NOT'EM[R-4&TC\nM8'!EXV&1U.0JY/5!:=1]VHAI%A3YY4T!R`8VMX(@-7$=$$J%Z\\0JBU$`>;W\nM^]57ZL'O-6=R<X>%0S(;&X24I&0PL=.?]-J'1#;7%T2S#8+&])[:<N4S0\nM\"6N8#/F?`(Q%TGG391AL>A,S41F3-1\"XLYF2D]GCC?;;>L46$A4#&;F#.\nM6W3YLSL)Y50G$JA&;)P,XUY,=[T:UZ[FO#.+2LV%!5'\\3BS+X2:C,K5F<^\nMLRQ3JDL?LD2_]2E_C`WJ$!:TFKUMH9O535@;I'/73'61Y`[B>6\",UN9HY\nM`U@9&'--66X=1$R:<LI'P`U/(+TV]KL>ERWYEXD!SVQY'K$TF9(Z^]IB&\nML2)SA=]N%$MAT;1JMC(+/T.X-U[^#UF98,>5ZK;&U0&;#\\IS_FE8^LJ[G\\8\nM?3AAJ7EWXT#'/?)>LS<_=4.E->-QE_C3H#1D<;Y)2&??9?I!$L0RODE9$C%\nM5\"&=0Q8'Q=2AYD[6,/T#XQZBYI=O<X:9N\\-Q-[;J?B[?+.R\\MZ#ZWGVI?4T\nM)(C,WYUAO\\\\&0OK+E<4)IA7]9=Y^3OJP^Q4;-\\F(C[V79_(2@>6\\HEEID.L'\nM9R36RID65(IL\"A\"6178%/FN](JOPF$P$,BD##<%NWWV5/!E9BH>Q<`1#\"@\nM\"3L(FH:.#;].49?4\"H-L!6L<HWS&WI#<>R9O8==H3\"]'C:K6CXLO5G;&+SE.\nM')7&XU.B0)L@H-H8(--NH<!)VP(;9BU3R=K(X]`UK>5%IA)!Q5G`Z1\";X.=6\nM>J=!B2CH#Q;!)1P=\\0:UP6E$!;\\<S4H7VA;,BKO`M-1^CGP=!M?&)YNI=GP\nME.FB<3T?4C2B/SXM0.SRAB0.]+_@#4.+2N&U90\"/A:16W5+,P>5+G57G7\"D=\nM[ND6=\\HK&=E/6:3)'8,(;U38OH6CFQ0M='!;@95252]RE\\_8UFPQ_&AS5I+,\nMF1.2C^XWEG]A\\!]<=BOZK&<D=8BH>`=+UQPD[-O\"7FY\"-\"[1S99^53:5!A\nMF98WCI-8^6!)AH5&`>3<10).&K&F03^.4E&Z8V`4CB8:X&EXDIA_$&+\\A_O\nM\"^<G'%U8K2\"0UBF'_?6NDG>(F+P?,YAI9`(YU\"2$5<P@WQX/W9OVGXJ(-)\"Y\nMF,AP8E2\\\"-,+C8D>J,I\"11E.\";C(PUX,&>=?+3$PRK_MM?)NF=RK9D@#J<B7\nM>2PQ'H29.ZZY5\\):QF`==%^LU07Q+A<9),G%^8A0YJHB_CH]GW?'HTA3W\nMU-L$6`937E7'\"R`TF='FN16/G\\I<WO2MYBVB%S`C<0^ZZ?8Q7`S^\"`J8+W/?\nMDSNZ<TDL4BQDBYL+XK&1VE?F.RL\"-ER@,+ZGUE5\\/L0@U;\")K`Y(O/),Z'Z\nM]JF[=KIB'C;#MA4VQ5<#?4V=UY3+47M-<8G,:-Q7F=8M7SD3>ES$Y2RM6,2P\nMGCDV]7;-75LP4=V+/CSB[!4]:_FMA1RO'.E$?,E$4:?EUWH%:=^`7SO3<P0\nM.(!G:#J@G4WK\\:]A8A:C]W=%PQ>EG`Z)-$[7KN1--[9YC.>%H$KEI,STN<C\nM5/4T.H^'<+[K<],]3S-EC4-4+^I'D^A:/MBF_9LVSZ?^U>T'E86(%@&12LP\"\nM(EVCY52P0ZY4M.#/`:T+3\\DS0WXFF<%+Q.D:[[email protected]>;9A)UM.TQ9@G[3&\nM?\\,_9')G05`1VZ?<J6$%M1DM-(&!,C#K.JHU34VU-?__.<_CW.E9P\"-WCJE\nMS8%8W/HW&QL;K`.^`M\"51:!,&)@#,H&VH59B`C2.CP%=`.<%M@M:Y\\83?[_0\nM,7LXI7!J9U/T^&!+8D.K<\\XZ5S\"G#,P_#-',1UE(\\2]^2V?$5T@YMD+J!8I.\nMJ4M<EQETR2KJQ5:,%V=&#42^WG0PNQK6[Q9I/&8IJSO:882^N+,8'E\"T6\nM,%\\P=PXSRF2F,-49@8S2F4F,+N4.WY$$@FN4#C=A4,M<&^Q-EC35GER\\#A\nM'[01!'.-'U:>L_7,;>K!<O[TK#B2G-/]1G]9.(JS\"BDH;L,0/C9V?\\#U\"F\nM^?^5\\=\\VO_LZ:___]=;FTOY_:?]_=_'?&,>+[(CSH:263@!+)X\"E$\\#GXP3P\nME]F[=`_AV2QCP-UK##ASU;19@;7SAO)6^!^ZZA\";MS-E7(7V\\^<F';C:)\"\\\nM1T'S=>/#S23D%E=G15ES\"-PRQ-J_#_^?RSE1<OR'>?S_YL8_LO'?OO[N'TO^\nM_Y[Y?R\"\\;]7ZF6H^\\MB'#_A_.RX4LW,RP!U,YIJ4V6=`Z6XC4K4O4CP,\"<5\nMOI=\\!<F?K@PDNUG<DF1/A4XK?H\\ZS<J,?JG:`GWK9H39:<YK'8/_I)';MM>J\nM=S2YJ6$>^(&.FH8830&#+B`F_&!>@TT/M7I7/97TQ)KQG`DUW.:YW'L;#\nMIJ_G-P8:S!G#'_B/#HG_5>?C:16_]_1]NZJ>IJE\"P$UL+$DXG\\I^D__-KMW\nMU,?<_%];6T#^O\\WE?]A8QO^YE\\^CM5(^`=P,!(DJE74W@Y>;(A(C^X0IEV@\nM;5114J#:Z]TZO,?F.COMO:/.B\\YVZZBSOX=/T&+I513\"7>IQ@7\")OQ,Z'VP?\nM:I1NO%>;S<UOU>8___F?CS:V'@&J;7[W>&OK\\<9_JJHVA]':D7ZI%A(&'0M\nM.9M\\D,3)W61T22DD.4J-28.HUM6K<'`Z'9\\WW=H49.<\")HO,.Y0_'X=##+>)\nM\\7#2%._32+C1`KJAJ.T&&7#&9Y<-:0.]3M@QM<&>RT#NN]&0(N\"X8R/QSB3U\nM<T%&Y0O=G@)%X/Q,&N`-CO8PKU0VE(=7PB,SD\\KV+.01U)&V/,&HF&Q7B\"\nM44XFSI.=RJBH)1X35^;$9MW^M(=)TZB-]]'X%,8UH'1:,!XS-[QI--4>IRW\"\nMC$AX(ODV97I:GC`5-0XN0S[DTN86!0I[$=DACTS4\\FMR<\"/_P?#]U\">+F[C\nM1X1PW`U9')/3@/07B)4DQOK7&QL<EX+&W(NO+DZ:\"F0QY&T.\\G'SCV%`^=\nM6`Q8<\"=-L+2!2)%=#)TMC&O@<!1=[?J7BG>=4F=J<S:VORY(]PR,''6O..=\nMC:^,$XHL-1KU+QW4[.RIO7W5_@EVE3I\\U=K=54>OVJKUYNC5_H&\"_W8ZAT<'\nMG>=OCO8/#M7SMMKMM)[OMM71OFKM_2)MO&X='/VBU`LH#N4/VMM'#6K9?C]\\\nMW=[NM';Y\\3;M8?ZU?R!M;._O';;_^PUN[M8N_-YI_=AZV3Y4JG70.>SLO82B\nM;X[@GQ>!JC>'+;UK\\ZAM'&X_^+HY]9!NZ$Z1X=J9W_[S8_8$1*!LX%1JQV\nMV@>=G^`1II6'A@[:^R\\:-'T-#]V!@$\"]:OW4AHFW]U1KYZ?.87L'>\\42K_</\nM#SO/.[L=F#L\\.GRS_4K:X+&[^]]M$-/:^V!E\\\"!-V]W]1='+[=U6YT>,Y&6`\nM#!,[:`%TVH<-`N+NFQT`\"WQ_CG#9VX=_=CL_=HY@@+`Z#>FT\\^/KW4Y[)]<&\nMP^[']L'VWC0XFE`I1>=H[WVX:$L9HO7MK/]9K>E%^KUFP.8.<\"8)KWO[?>\nMV7MQ`$-I([\";O\"!F11\\?WVP_Q,L.0!.`[FUIZJM0WA75<];AQV8$;;E(%X1\nMD&@AH$-I8__Y;N<EK2T\">!]'QOW@OFYU]H[:>ZV][39A'P>4@Z]O7N^TC@\"\"\nMTD9[[Q66P8$?$H;\\N+]CSI9#6<!2/H^\"X`LA@NI?>`1=/'6?I)->G.`S[Q$F\nMB,P^\"^%@\\9Y-AT`\\>_ZSL^YPTL]61;EZYEDT^74P\\H=2Y6!HR?A1.HJ;%U7G\nMC4YI^$ARQ_EO@7CWTT<8^;3H.:64C(&X>^_.!N>95C@1WJ.+!R%`WR5:RF]\nM3+M`V+,V#BO))SZ)\\A`S'R)GN'4;%`BBW-$L41-#F7U`0.)=O'8.4VF$QW+\nMS@;DP^I<#IP_U<H%4DA=C1\"1XQ(I[K7H3CRII$'V6IZ@L`?`=&7&%!Y$D\\\nM).,?KF/,5#LBZ@?#9Z(!&_BONA'0^?Y$.:9EJ'1R<``6K[[8^M__/BQY<'\nMK8.7A\\=.1[H,=E-0AMK69:#'>47B]&0<]3,E/F6A\\N1>%U\\BF/>(/TC#]X@(\nM',#]3G&`\\L+GD.!WQ(\"W6QL\"N/<8&<-AU.\\CK0^:_\";XBEDUA:?P[Q1)%U0\nM&A93'N^$D^F`GK,YF-,5/B0D^I0;]:T7\"9<&I13JYW`\\1'D[23Y(J#,!?LXD\nM\"VYVU?LXC4\\E,_&4>4-<U'XR/`?R6NK081FVG$ZT.PD4KE:^^!@_Z#1;#;K\nM2C??QQ3!6-^):W(N3\"%R\\RE9/MLJL,$T80\"\"$SBWJ6:CC@N)49.7>.PEAQV\nM<IHVL\"0%7,'$I62>/+&#6T5N=1!A.$GD7[&V&)@!4SH]OR#^/!-B6P_H-.S^\nMIGF4#L9AGY[W,3XEQJ14MT/)+^(S[7]8&T8?)^DD&M6#+_A1)8UQZ\">GT[.\nM_-6O%+SB0ZGVL7%9K^CO=;>`3+TB?X,O(N\"&SX(`^HE@XD[K[M(_J3Q:(R!G\nM%A]/YAI@;BVHL:]O!V.GBYQE_<,;Q1B]#%<:Z7RO=I\\4E\"\"<O322D.)C:(2\nM`ABDY$`;$1RI<HNB=5:'T`9I!-:X0]6QB1H=._-[NZ<PI@E=T[A5V%Z0<Y$\nMNCS19W)]6J0T4PJO>)G+^\"(9PPVP!TC<[8=COA^6OIY(?H7PH3T&&AO6\\%G]\nMR:P2#DQU2=QK2/T+0RC/:H?L]42?KQOB;:1@EQ0V11N%:'E!$NB:/A44LAEX\nM&C:\"\"C[[B2V`\\2D>PUUFB1]M1:G>]-^'T<O`W0;QB.EM(9?`2-YA`[>TKBP\nMKK/;=^K&0WV:$KW?[L?1<`(C#547]W?#07KHF?XVT`H$LSJ?=QMTY\"HZ^-X>\nMUY_D6R73XELWFV]7YWXH?\\1.2I:[:9Q25I0&$@?MX20YF2`>G#`74BMHT30S\nM#B_),Y6[@&K#I.%@$K4#8\\(_.\\AAP0/D60KW+5Y\\>MUME6K7=T6](TCW#ED\nM<Z@U.+EAX-!#)2A;-FN)3QV%M!UK&D;)$.Z`Z`5%1\"]@/E53^_9P,KZLK\"&%\nM!X@]\\=X=4JR%\"K[CL`O\\.L/GTY6'>4G[UG8E37Z+LPFEK@EOXF?-;636.,Z\nMOA!0GOIDV!4/Q=M!4;2JM0<9/J`N?%^94]D6CVY@'\\G\\7>'Y[W&2Z/$Q!AX#\nM=@E[5B9GNG(ZZ>'UAY6.9.<$H-@AJQ;>=F5\"(\\>>H`,F<@5UW[.]5E22T(KG\nMRB.KU>MN?5:06HY?<L(\\9DB@;1M'$ES'^/4H^A4?\\%7L956)AK52J3J0J^).\nM]B)CE(-\\P--DTN^HVAA/C(-N10[I0+>84O0,#,+>\")DSLKG:R$T\\DUIQJYL\nM+T81Y')V4#>]APXH'S`&)ZV\"\"Q\"V@0!7^!.-H0O6[^AOK+4#J'%ME@UW<:#\nM#+HY8.V:UF9T%H=NWL946?$1NE6L%W2$@%\\:M+\"^E.1&Q6]U,_\\F>S!1<Q.\nM)#/N3WN$3H8KCX[72S)'>6?9R%[#U@B$APZ+^X47^RMYH8(8QNX+L:8P]#B\nMC\"-#6A!I3\"/K3[74J/A]V9ACYYU!'?]%#LIWA3Q[R0<XD.4@\"$F/=U=D>,;U\nML\"Z<0(9/X!N^=41\\0FZ(=\\-2>L,!SM+A)P@8J/?W`'+'K&;N]BPL9]G,DLLA\nMA2D::.%LTSX\\42`Z'!2S3+)2WH%)7PZY;-.<2AY%DAX&RP8K6A-GT.2!,@\nMEPF2/&\\+LD+8[JK,$!E;[KVD(+3]+3Z?DYI5F:3D93+?2]8(0[O9Q$%&9,\nM<E#I?:B.$I4Q9^,PV[4RVE#OBJ$R]-AS@M71\\5TOM[N[^0YG2G_?S-RY>=\nMO9</'CPH:WT?K9E+T&$T^2D<YQ>[0=:&)S3^;9X?K&9U$_[9$(\"C&<TB#MW\nM1<-/(<>?/2G3<HF=/?SQL/+D@W!2#8#,)=T:)@=+SORW]7K^+BCGP$V:81E\"\nM/)G)QZ`6?GOW!+78>R]/?FC_<GA%3??\\SM<L<^&`1HIJ0MU2?]Q^+IC;+,D\nM5=9='5R6'FWS\"`KQ&6E.M<$DS`KC&D2;^%\\'I%<W1S(WOSUZE&GP&BVFHYC$\nM5GZC6IAUXX%BL\\)YYEN6%[=JG$1MQ8,>9:$Q5\\9<&E(>`<V8))+,SLDQ:'/(\nMS4B>6]+AB=,7IY1:'1ED?+#_P\\GV_MY>>_MHUDE8G<7R/%9.B^HLQ(R']MS+\nM\\&Z>2%XZ4&=08GTPAO3B\\Y>Y_#5#8?$[>5&M(`&`)G%N<AP1ZRD#QW@)7?G\nM(PD;DP+4A@G$R89B[5C6K]SMC.O;!&^$X7+%=RXZZC%\"?57-?N9A$!5TSJ\nM>?U2S#I0MVEKQ\\[WLV%2JCB0?\\!=D4P96\"6M1L9.I[D\\:ZICT--Y6?1:@J\nMS&VERC:.3JX@;X\"KQ-7:;24]V`UUV-Y=<(_[N/A8R<C\\<17M=KV=\"553C'6,\nM8>+H!JXO8V6SQ`<8^TX?]./DPZHD(`?TS[D1RI203=?5.2$F!GV'>PM'CBWS\nM]!?41D:H]I6#W_H`GHT&/):&1'TK\\L;,(D;11L46CX;U=[RRG>))G!PCBR\nM%,BMT>7SG&T2B;9F&:VH>,@:?+I)<4H^U#-([PG&F8L0'E;%@LHD7J-4[%;\nMN?N/_M#G>/MB$W>UM8X(K>+4Y6EZ`HV/2SLO30:(74.7%1\\)):,V-5/#7\nM9$%+Z?J$OKC3<&)JQ7?.B\\-1(=OLC-W4:=TFTQ'1^O6G6BNIJCSBJMZB>&>H\nMQ2154N@7[\"TW/'KXT%UNBE?&_9R30G,:93J\"9^G;^+@!&V+&\"&\"H3O(,DIUB\nMJUEYIH-9+8IZX.%6@X;24.N;6A(Y8J$A/'55,V183P>RB\"C9OJ<-PRQBK)U.\nM64CJX7,]=Z@7X3Q&35C`%-\\1LLUZ$!%,+F\\Q)SNFG28QDY.8LX;)DN$6J]W\nM7=3D[-G6Y<>4UM457$_\"<Y21H:D;VV7H9Z9PEP\\O]&?4U1#EV$B9MMGIJPD\nM&[=VMLWR>#LR4`GF6+Y4'`,5QTE8BQ4X#':J%0JVDA%N$\"VCV3]_I.[D<TX\nMJYB&I;=1+JDWB];A>%HI)U]G[$S)\\:;\\LM\"^Z/,(P:/29W`X?&X%HX3A]\nM/K$4@/P012,29/XFALCL3R3YY\\^0\"JW#\"-;Q;[/9)&PH64QG31`?/BS5Q,)-\nM(X^=H'6U0R\"`/FAR`IM4UR)[email protected][&,0JET^FI;.(RT402%1WQ`,B&R;4E\nMX'T.ATF1L9NF(,I8M1E[-HZ/K8,>VU9J1#[JQ<9MUJ3-#7269OU]5+7QL;U\nM(\";G#C?C#:4BK@@$A=RW6)#K8GRG:C(O)4GRZM7_(,.N'%T/NV'8XO4LC]+\nMEE6XQUJAW>6-CS620&AQ!(DJEI)=\"(F!6^_W=PZ=@NLH?5W'MRI2$8-XZ/\nMAM&'18R_7$$25BZ@U[!&PP;>C(9C-Q>7'D,G]7FL8[JC4WKZ![A[$W+W%3\nM/4<;0@57!ZMG(U6+33Y'SI\\F\\]9^W\"M-$D,K`I0&_H531).4]:=GPQ.V&Y92\nMW=%E+;O`L.5TF$]:FI3/)P4%.?>M)&.'+SJ#QP[B;);].15>_Y\\)2;%B)7\nMJ58>6>N-86^.M4\\6-(Y`PK3CG\"JW`TN%103B9X)2`+E.3W+`(&\\4B8%/]7S$\nM,\\$8B:=GDKIIMD,,!G;2O%P_1$@U+JM+'UITZ73?H\\\"R9&)'T&U613\"C,<\nM%2'XB&#&DR:6H%)P\\2`_+-)L,)ABE)==).FVK=:Y(JQZJU@?_#MB?L,X<#/\nM7B3CP8DXE+10_+<@C='78?81&J85\\Y=(H0Q:!(+SJ;^_XTZ5U\"`2:Q_BND\nMX&^_V?CG=\\>FP!J58#F%>0C438AU&;C60L6J-8RQB\\-N\\1Q>D);_]FLRUQ#.\nM5_%FWM?)8\"AC!0?MZ$J3<3\\:YFA873TD.7-%2&V@<M7090':(FJ,6VY2C8\nMQ/.@4@()$\\M\"1&V4XI!IUN%ENHWNCH2VM=<'^_N='60]D]_!2:PTWL93>C4\nMS)T<6&BCH398W%HAHV:S6SHI\\!EQKR:]U8U8Y]JVB<5WL2ZYT!'$M-R>0&:Y\nM04IY9/8FXH^[;>OJ9?OH\\.C@S?:1&6,YZ_@RXJN:B63A^I82H1NB30LPV>_A\nM-L\\2$)(P5JSO'PG62B!H0K)FK/71+Z_GK+4`CM8(AH5-+;;DTNF=+CDQV.PF\nMB:>3O^J6[))9B$N2W677XT27#)JK;?^%+Z/(Y@-WDTZP_<X,=B$Y4[)F4%\nM2(\"C)`F_MN``[+F,)COH(+>]B>(3MFA$DHUU%=V2+Y;N$\\?<]PA!%-+=LH\nMOIQ1#P@G6NYDI`G\\+UL(;$+LWN(J@$(441@CR,;6KJL8VC]7E7P]Q=:B@>\nMUT5G\\\\8QJS>;:P^H5PJGJ@]WX06MQ/++AY24MVU0;6QKQOX^/L)C8.?,W\nM<F4FRE>R^%Y9:/_JQ<GM8\"<YU;5WLIW.@IO9+JP3?R#&1=O$F6!+,08TYL+.\nMW2>_UC!711-`&E1)1\\#)3<YJP'DYMR@8Y<F7O2HP$9K-J,QI#^I2$1V6@/'\nMRYWE>`OD)[%AX\"FD:18%X1WED1`O]@/@C5&T9`M0X`?J;CXA:R,$2&@[DQJ\nMHM?3KH().'SE0CB8Y\"^'OP17-\":K\\D4TN98&<.VT9[$T6_WCW4DLU##F[\nM/1'+J>QS]Y#ZDRM+X6TD4XSU<^T^Q?DJ:!6I`#RK7KM:`:M]O0;RZUD2EP<$\nMY2S^Z)PU\"!;U(9Y<'JODY:I?[ESH$R@3]7+';X+P\\)3#5>H3179E)/9_Q(/\nML05:GYM8;',XM:P;I@S;Q`Z%K5JP\"9=T/3Z37/2_:J+2$=94AH3XY((=A\nM\"F#^9<\\FMF%:3>,A4I&A&`O.B7K@^<B%1.[F2/&BCQ.4R7V5/:Y%V;PHW)Q6\nM=5\\C?2CK%]?>.D&EDM>&%Y\"\"3,.L$2\\D&2733U<F5&Q;<?O[T^6`\\(704R\nMI+$+0$Q(,3$R/`9@7=\"28/^'LJ]5KIN')TUEV'Z9JG7XIZHYS?P6SP%#WQ3\nM.S5;/2L&S`N9E$@$K<3WU@OORWH9Z69+>UTVTH#E]%NZ-4^+/\"]:R@G5$+=\nM+I?/Y=T>?B]('#P'@CF9>8G2IAOYL99N7C\"3[^IZ$9:0ZY(J!I[<'P>\"U1\nM;>4MC?BUUF_EWV<(NK13+TL#ZH4SLK2E;H\"FH'V\":>L\\#HS7EL\"\\++@-H@&\nM)S]RE78P99SXT4UE(B%W3)UUS#4AKRJVV20MJS<M3=^04SO4MW;IEB4F_7\nM./_.6>\"Y49XVY7187\\^:`UB;9M5G1DIBOAFHA<.4\"4OP^1DN.&8V,.HVT2\nM\"EXREA0PD(X0QW/9/PT@W$4.#ZZ$>451W3#RW/HFH8$0JDH3B&%<M`]IL\nM-E=UQ9L<7%=1#0<I\"@/;B)\"C5B-!\"WFV'?:!05BKU\\1@8/TI1O0A@0!9>S:N\nM664RZ7G.W()6/=1&Z@\"]\"W/,-E&BEH);_)K@28+VVY#1X<0OL&\"2@@-.`!\nMFSY)TK2(IV=J,`&-T5Q2&+-Q!`=0:J\"WQ)3.%X2V4N(P&V`\\L;:&LF)SF!X\nMX]H\\B5`]J%2T/\":WV`WGY4RAT?Q\"+!RZ!C:X2$XLL%Y$QSIO'CK<EMB7>;0\nM6XGIS78+8<PWLHB6A]?LZRA[?%6MNL2ZP1(:4(>X3YH^0#P,JQX9.N@+8V\nMZ2CJQK1G)&NB0(Q\\TY$-NQ,;/[email protected]=2:>31-D$'GDR#H=IZX`FF2.\nM)Z5Z_F?O>9A1;^;I;8ZM!_.+7,4\";#H@'FLQ%41)UU1:]$U$R^05UT4_RWY\nMF6U206EU%>6$XQR3QJIKQM6!#5EB#-OJ9-5`TJE8U^58X)!MC&:@,.L/AS;@\nMFTM\"0OT:BKBFP'.?Q91\"@E:!)8QF3Y@@M'`;9,:#6F%[UWR`6;(4&ML=-62'\nM`)<.\\BQA+[PG1QK.'GN>)+U2=\\9],>J?&;]M;(3%1-@_E6=J>@/G1)ZUX<1I\nM/'_P^JKA1C\"W\")S.FLC?!N[W9:^;-:ZN9RQVM<GYW1CJ&NUJL&@<4QU/T1JB\nM5RIK6`V)Q.=BEJL9^4I%_+OXN;6:0UO;R71T+2M>#9_9]6!XYF7>RM>8+CN_\nMU@9)#[WTO#F]-2-:B[P0E_36\"!UP),Y&0X':\"#FU<80WN>Z$;NS:3XD(BY\nM0[\\U^M(10EF32N,6!>>`%.6=([email protected]:47]5[.BIL+W;Z!\\:QOB\nM(@OGV=;,10;%?S/[86!L@/?4(A\"RFEI:&R^MC3\\;:^.E/=&]VA-AF(^CER>T\nMM^$O6Q7R%V`N4_SZ`7U7X\"]?#N%+,EN2<E>^V?O]_[N#BI\"-;%9VV9!'10\nM^Q[#'!H!MXPI2E'HW4>2A5PRM$X!BY407<QQ1!L_MTLF!:MAEAE54\"((\"L9\nMC'&LHF[2).+82@;I;C9,;\"U6&U!9_WIO?L$EFC)6OE<K9\"3]ASS'\"U07F';S\nM$ZNMQ&KEO3SY)'^G0VX>GKNC6QI-_76-II8&4TN#J:7!5+D&4R)0UHIDIOU\nMIY-SYE%@!H!MO1/@5IXLS:N7HO@<5X1ECZ6HSY6=0Y1M%OQ5BG$::AQHC<\nM8TYQZ-A;+>VIKK\"G\\E>T^+U9SL_/XHHT;O9LH7M''K>O;_M'[6J),<M/79<L\nM?<59O\"N/TDC#^+V@6Q2DBF6@'H$F3;E1\\`5KAA!V-G6+>XL;2\"X@;%4MMM5\"\nM(P$=Z(^@EA8-F\"Z\"\"T%.$!.8\"RT2SEJL\"2E'BS5H-1>.;\\'F32ND`'D;'],W\nMG$`3X>?%KKJ1R96[R^:9A2Q`6VT1%V?H4DUA`G4\"5\\I9A>#25VY'87QTT'GY\nMLGUP\\J)ST-XY>=Y^L7_0]K&%JM<7!R7-N\\HMF5!#G)>!F(Y\\KZT71^V#,CJE\nMAOSX8L+G7<5B4!8+20P3,5F@I=?SX0OP.3D8/_G,N8S2P&7>SS\\AU1)G?\nMR^C9-'9WD&;U%25C(Q6=MO[4A)#EKUX`MD(3S1RIY12$HM(W839G[\"9,;..8\nMVLY>]61$@AUMD4)B\"7R`LH39>^V7D\\[>8?N@E#7AEJK%$HH;D5%&B%7`E$\nM:YUEH0VT%\\%1@,5.>[=]5`KIX986AD6UNN!\\KP3:HJ\"X`A)O7N^TRH$$M\\20\nM6'27,(='>KN<T.236+9^9M#3-JX:ZCKB91'H^YO:Y6':J68'%V/02FX+928\nMJ2X&)KU_B=+V\"=+)X73]D$;D%[)MKHK_XC93+SK+G$-1)O5((7&(,/NTNTQ\nMEV;3?TNSZ>]9VDM2D:7E]!W+)=(/,?K]\">Q=FM`-X2QBL?MC#\"PZCL+?;QW\nM9E%[T5D(0W\\L#&O1I=#&P`)6O.PEV)W2,]WDA3`\"\\)E[<JL%&80GU]0'KED\nMZ\"9,IG0`X7G<%6^>5.W_@&,#W_HO,:_FN4GDVE2S<%FM5.>C+N#D9SK'Y5\nM=?^':CV7P\\$:WQJ;RD_7:10G.+M9FRN]3@:!GY,-/D6UC)#E2$F/B#'G=18U\nMS$$1GUT2XZ$#GI>T43=&/7CR2X@1-'I<:7A]E>.,2HP/=8\"M9[1>Q7FWQ`=\nM5HI#,)V/JBK)G_'=4F2L\\`X76>LZDNU9:3!E>YO)&YS1O]D;MSGF&,WZ2';\nM[:3EB.N<XI$<'?!VH;[`$TJ\\ZM+,E5P;!%,.0DSEM%8?L;&6+W7#K!_,1L9#\nMPST:NV%CGW]U=3;COYFD3P]68B51)!(T3]/#D/A+W$4N):ZV9\\:JQ8-\\:&B\\\nM6+SI@VU.EQM9(9G@4!EWOIO;<BS,ODZ47Z,DI236#A:D%51DOP[.IT4[8'/\nM30\\1VZ3#E(2ZA@CK[&H7&&0KC]+EZ6!&]-S]N%>IZ\"%SF_<[email protected]']\nM7SD&$)WS80+X,TJC:2\\Q]$!,G%0/:\"$PV^O!<=U67/6DCS(8556FZLV3)%Z\nM^+W:$L;/\"4OTJ=30B8Z>@M(!W7YNO:9)$-&D.+.^>'#8VV+H2MQ+=JM)WO[\nMK2.XR#Y_<]0V0)E'[>U45K^D-'!.7^OK[![\\J6RC6-LIA2?4>NK+E(Q!J;26\nMW7(\\5W/7_/.\"MF;4/`+[=;7)^A[<<KT;AG`M6I_M7$1';Z$8<%^B2R);R6J#\nMF,SRW7PJ-MPK+X&)/UC+A(K,1XJLV[\"IV(I0'W7])B2V81G8=YC;NG#?!M:-\nM_!LXCKK#Z\"+E\"<JG#G%'2CC\":XH9JD;,%'F3V3I\"0.^^'7+?18Y%KHGUC+\nM4`M$`8$EV835'OS4.MAM[[5.7G3^3WOGY+#S?]MS4,+X/#Y35^!-/QIT<?J\nM:A2#V045>Q)J1T664&9O\"(7Z(:4S_)/FHW77%W#\\I,-PU5I21#KI'513U%M\nM:F6?>@T(#T[CGI\\3WDVPXEU_9`\\#>`ZV8?(2%@^+%3']L_CZ6T=6R-Q4[\\1-\nME.!!Z5SXF\\FFC22.+&A+]PW%RX?3N;5P4UU$J89D&\\=;EEJ3.V`AH4H)H;#\nME\\-XP/?W;X_=W+3$%>:RG-R91R+`B?C+`?P3GL/52()!R;V<A(F26K#\\M&5%\nM<KRL,)(\",N/@>7.47AJ(MFGT3GP#B&1ZW&\"F:YT/1W]^R9GE^0%#7G@9\\J\nM3G1\\08J!].]H['7532B_LAGA`^X0'-+76=4%'+<'1+MJ-JOZY7`2?E31>)R@\nM+&&58BCPU6J0GJ\\B!U/!GG[:WVT==7;;CKCBRG&('(TVZ>8Q=,9\"F0YF=SE\nM%!=P;5MG-_9SZV!O3E/NDYHB8J6T]2+UE%K=TY3]'ZQIG;:S]^\\G-,4O5^L\nMJ;W]SM[.7+##^XQ%L8LWK.OHLX;R67KH`Q>572WE<8']#I^Z24O;$0ZI6G\nM5./3L$%4#B_3G%HZ/H^'CJX(.3U+WMT0)CK\"&V;[&U,&=SC1^(9]4JS8Z5!\nM$O@%]EO')?`>&2CO_W!'K,?OTV0245P^J;Z,6`70%?K$R87&!8&89]PSE2%\nM6HAI2AD23(YMJ!R-+W6E.V16>+BE<\"OD:;QF$K?QS^YHT_^Y5;YNE%263/?+\nMY$GTJ;-U@U.'UYZ7#TE&YL2IW/U^-_[9Z2CL1D;0,0@_QH.I:%0X8%TW-K=\nM\\S%:EB6TYV.;NG:4T$1+8UXX0MW(\\ZH6EN'46PYDFS`'R@J+_6C+;0NH8!W\nMY=GNAWU4LH^)D1/`D.E<,B5!]7O<CKRJQ!&W=_2?IA>E`42DLFJ&NX73Q)[\nMQL](`/ENU4H@<2<]?CX:2[\"H5_MW:J7<\\&I^F[5J_A/_,&ZVZV-(ULY0R\nM5GN9.&`FG\\$)^ISR7;0LLXQLK`F8_;^!63[=%L_,'[.D#T$<#W5]1EHVW<\nMQ#T<3(A5G<Q;-G=>17IGYX8995D77<RV$IOH/W1:8,[[[XY=N(!<=0SH\\C@\nM,X7T\\!IP6K-J#R=HB=Q[Z/S+':^]Y-DY$<<&G(\"IVP$I#7WL0V8Y,5+FG\\]\nMY[%.\\5(-X\\P=&&9MJR:89O79.L]W^,Q_RH8'..-GPB<\\PXF0D_BS:JF)A!,,\nMH=+#`^%R<H&;-3Z3D')PVQU'8>\\2N19C993THK^25,!87CD<!I&K\\T@+<'%]\nMDE&9!FJ&]?C7(IP'(50A[;KZ1LO7;^?T094@=IT]@=S[5XSW+T9!YP)F3IV'\nM#V/U]'O32N4FHZ\\4L$>NB96[X7E,#Q\\>%RKCYD^\"5N_/FH2V(7D938!8FL;,\nM?.!>0^.K%SA!%R^MVWI&)^G8<`E4#&5$0P0K$K/W?E?:]YWS]J^'D<COD$I\nMN;$ZHCDKDQMKH9P3K3,KF_,M'-FT<2FBRXOH9EQ[CKP^U.N.FTX-Z?\"4_)Y\nM:(UB7;/#TNXOSF(`HFX(-JV9O/`YN6]6Y+1-MJ<K\\Y6TT<Q?X/)V\\`XIVC\nM7Q[;0-L%[91L$(%G&UA!X?MW<[>T?[CPI?LZ%/\\CAU?BM^Q\\CCF1G>L)@H\nM'>&T4\\/SFJ=8O)M;>/9!Q\"+_,%V?JA=?#R#?J9S06U)=6N1HXY>ZNRS=:\nMY9`:\\Z?A\\CH%`WRSM[V_MP?S:>^4.D:GW3&N;W_^I=2QX<-ECP-P>'^P?E\nM#HV:+&-P1P>MO</6]E%G?Z_4$3KMEC',_=?BH53J($VK-QJB,5N?16V$PCID\nMIEA5,&>LQD:F.A5]PL$VAK\"B7CJ@_L1_>\\/^Y>Z1N=CS$:2+./#?-(P$A=\nM1,!0K<-@X+`ZC4B<8&SB.:4Q7/O.DU8(VI#+MIRB!H;UD?74\"G#)6?4@%^\nM4<P62E^\">7MMP!3_M7=%+U7`BD'<,5+;Z`YC>BQYP`57HT/>SV,83],E)&'\nMJ//X?31LJ%_1ICR5:[N)9J=KV\\##\\1CL1<WZK?\"X67=TX!=:&-,UVF$,3.-\nM^E%W$O5%P3`_=3>RU`TKMV?LEO$F@S#3,69EHUWG/NEW)W)?)#KO]TX-L9-\nM-M'#`HI?=9G(D01%U8?F:XN#OP&F+T\"1WL6=?&+F3<(>!&?(@0KN$D<AV'\nMJ`;L`@R@5Q82F(77XOBLSX$66&:=#1;'@]C@0>SA0:#%H^(L26Z2GIS\"7GCJ\nM3WQABZEI91>.2@!/\"51YV]-^VKRAVTC]X<[!ELOQ5\"94Y?;MHZXV5Z?G[0\nM;L$T,NYP96'T'9%%[95<1+YDU<L6=]Q4.3^7UESOTG>?JAIA`%A;\\UPT-:ZY\nMH13H09T0[;]$9W/D)U543ENF#\"S=A_`RI70M3LP0AMJFL;#<U,(V(!>F6/\nMJ1COT67.`3;7A=)6=.2.C@IT2^X2[T7._0%KF8Z2P%U(.DNA?-N]8NZ0!\nM6YZ\"J<(^\\%[>!_M$VT,@2>,L&;];QO%]@@XN:PB!3'J38,XR)S@S0=]Y%R\nMXO/6Z%3N79US%P8C-S%3=#!>RS#'YXA,Z<U-%LNQ\"\"8'4S><!3N%H./FW?G3\nM>FS&%BI#:._\"7_R3T8G<'S\",40T=I-$'7BJZ=QKGX](X/S^$J\"5,,W(>V`+9\nM$_5W]EKX7?8?-G_\"&=K#2V<Z[\\[L5=_=_J\"D2_5P[)_-WD=22\\Y/?H:F,'\nMQ'6-L^Q^;$)EV,KQ$-VLV&/7)DR:T8(ND&]&?$W2&XU!W'Z5Z_A;7-WS][4-\nM].AAI\"X'1=7%V_CI>[KWUCW139`&>2OYY\\+7M>+:R\"J%KRR[L9%YR7S7GJ\nM^!D3\";R7ZX/K3.E2_Q13C+@IAM++=!(-.,N&;SALY8GPX_=I/\"9#5TZ/J.%\nMLNL?>\\:E)9I2Y.[A3&0KO35W&O]V,8S^#6;01H'1QT!R7?'=_!97AEXD6-\nMO#!3XX2I;.PEOQF<)/#\"$A'ZT\\G)Y0VSO$\"S/I%JKF^E3B[K_)X3KF#GV0[\nM1S27,`:+=B(NX]D]41`,H>)M#RF`%';Q&?&VNC.#!'U)(D=/8C=)+HQ,7FGW\nM;.K@>_>^:!TCA$G$RN'`<?:3KXMNBVAI=WFUG\\>2[XHM087@309H@SH5LKK\nM\"FNN:XZ3H<;[^2I59N=P#&@L5Z`D?6(C$Q6VE]>`SFK2U6E>U:JCL)S5'D@\nMKVS'U2_.;(DUAE>UE5<'SFK05?!=U6I6>S>K3:N/R[7H]OR(AY'/68T_38B\nM5L7T!J6=MDUC;-DTR_/5B&6L8HY\"WG\"K=T`##ETI2:EW@/6GSI;7DA[:R`ZS\nMY)5<;N\\%M_>?BJ)F>\\-1SO#-(K@TEYH2(431D#\\HP,T55NH\"S1\\__Z7.3&\nMBV0MX;LQYZ/T;\\-.PA)'6G-%SA+-D,\\12_N]_(GR:31/&+%T6MO)A;Y,^AZ\\\nM!$9E2G%S?@(%C@5N%^=7^3,H<3:%XEF9TK\"!:Y,4?B<OT-Y,4<9X2L]X(\\\nMA,WD\"%D$A%H\\C6])HI!O+>^[,,MY07LO>(WDLSY[KZ^01<OXKG1G&/F>\"PPV\nM]OMYIJ[AWX`P>'9G;@Y'<>]2Z;R0(2IPQX\"NXW)%(GAF^5\\D(_A(9=[783N\nM^T6K_VGIWU'2['6<+?;?2)U;6\\FZ@;^\\H\\7\\.=\"N_[/FH\"GU+<;_=_835'S\nM-HB'Y'I,VC<3&[U,>NMQHT6X^Y%JIKH>\"R824X'I^<ID-S):HH,]_WE:W<9\nM9(-:\\S.U7QU/0P?_H\\;AKPFF02'CE!M+8R;@K]$S<FF5ZIQ8:A<W@5M)\\2P\nMT(KYNLEX''4Q8M/P?')1)EYSUP[/4!%\"A8[S'\"8,W[H/$,#9IE=5ZH\\3I0O\nM>U/K)5&7,&`8B)1\\%VS68M4T3-LKH6JWHF-YX<0K5M$O\"EIO;0IKL1L^X#Z\nM61NH\":^U6H'!49IUK#<>+.>!O6IKZ+T:=Y7CJ#.$W)-5Z/&4=[M>>.!7TI\nML0SSFG\\6S9<38W6.PE^?:_IJ)-\\+5/]W\"P%M,1(R&/R]P!ND+&B8N5YO(YMS\nMONJABM\"@N7N9:!;1TJIMI@C!9EU<I6\"TO^SNR&W>V4M=6BU<6UXD[(?`X\nM.7LX3B.=;LOR,;>/X7D+G7EE085Y.4:SI%?]E?6JO\\)5QLHPU^L6ZU4''HH\nM^KI?CZ'HAOKJSQ]\\OH^>/W`4_AQ.S8,1XU%0O5:054:6J7(R`/;S>H1Y]\\\nM)CB/QBQR+L+D1?)W,)3XFXI.J5)8`\\(/`-9T*%OXZT4$Q,R+[I6.64LJ<\nMD7C/C95<`N(O@$Y7KJP?IC:OK?Z5(LQRQ%;XF,6&%Q38#<5/!:KLPK>,S.6%\nMK,;<06Q%@<<AYDLIX%7HR^W\"7%_M70J&V[/'FQ&\\%DF3WC@&IC`Q2;ZJ\"ES\nM/TQ-J%+):V2X00W>=<_9@Y&L)L<IX<Y@^.C/`[1B`YY,+2)NPSM@E;D\"3-\nM.^06/>`6/=P6.=AL^_3$\"',M7IN;@-(U=OWL@M64-)^I12HSF4X]@#\"/H^\nM1`%`R.E=R=GH<[T69]LY5I4C=W8[@7WEONC6OP?HNR?I_N2+A9F)_/X:A\nM0^%E)!E)!E)!E)!E)!E)!E)#/(Q3('5_`1`#I6N,E4]]C(M6W7K9.\nM7:96,8O^2SBEUB9AZOYNI?8)66D92@K_LG?,W;)YQ2JQ\"#B,E3),E3)WS!4\nM26YW8882S%9HSA)._XBF63IWH0U6@GIC/HEX)4HW&<?X&K-&6F`H;JS#-65@\nMRW&H-DP:UO9!]NPMLZ&9+@&Y;$;3#Z93H\"S2B?C67'H29J$B52EQ&F2]\"MQ\nMBH)\"+]([YB$D65ML;WX(#IK+G?J/V.W2=!G.W(&]?:&O:0>88#\\>S/RV=S\nM=3H0PAA!]9_\"\\19<?2B%LZ0/8<MZ/ML&%@FK12.HEON3$XC,EJWE2-(0B0\nM^<JL3D5`B`\\UZ&9HLIS:NA._OFVT@$#0C!VN@0!6<,15&3;ZEKCANP$MF''\\\nM;I(0EY%>V0\"3886.RGX.SYA3=^++)@JVG]Q-2N64Y>EI%3&Q;^`'7YR'DWP\nM9TW(#B5X-VP'#!,)A`)_KQ,QTQK/L-4Q_'GE>%8<+\"AKC5N;11@B/KULQ%S\nM/2^C\\9^:TMB]F8:\\:[email protected]`<YH..9T//J07KP8#S,J[-_6X\"_4AAM_=\nM<`I4UEB38%HTK(%B>,PO4#\"#IU=JAZYJT0?1VAF\"T^P4;3R5&DRB,A_Q:V;\nM#H!E`_81!]\\LQTKO`>]@5&'!>CAX(MF,%''K:E0PF3M;NHE8N8J3B9F+F,8\nM$%STXMS+;,6\">=G'#:5S+NO'Q=C)<0\\\\Z:MW``ES`5_D/&O8@<@>M,HX\\Z:>\nM2SR%;;_!Y;^J]8VZHXR[$XX7LR;UA)/4RE#B>C%(7P]06>R=IRDYPB)I6&6#\nM!C^JLVM1#YA/N+>)SM<.(VK.=P&,QH[ASS)-;ACP-9=F._]$[C<)6NS9&V6\nMK,V2M?FS61OT]TK(!!TK=5&X&R80$1M'WFDMGY+)D=.019?=MFF4.-#\\2&\nMIM:NGG=N>8_YF<43.7S-?RP_R\\_RL_PL/\\O/\\K/\\+#_+S_S_\"P_R\\_RL_PL\nC/\\O/\\K/\\+#_+S_S_\"P_R\\_RL_PL/W_WS_\\'[P<2WP!8`@``\n`\nend\n", "msg_date": "Wed, 11 Feb 1998 10:02:14 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "PL/Tcl" }, { "msg_contents": "On Wed, 11 Feb 1998, Jan Wieck wrote:\n\n> Hi,\n> \n> hmmm - no access to /pub/incoming on ftp.postgresql.org?\n\n\tSomeone decided that storing WAREZ files in there was a good\nidea... :(\n\n\n", "msg_date": "Wed, 11 Feb 1998 09:06:51 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/Tcl" }, { "msg_contents": "> \n> On Wed, 11 Feb 1998, Jan Wieck wrote:\n> \n> > Hi,\n> > \n> > hmmm - no access to /pub/incoming on ftp.postgresql.org?\n> \n> \tSomeone decided that storing WAREZ files in there was a good\n> idea... :(\n> \n> \n> \n\n What's a WAREZ file :-?\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, 11 Feb 1998 15:22:59 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PL/Tcl" }, { "msg_contents": "Thus spake Jan Wieck\n> > \tSomeone decided that storing WAREZ files in there was a good\n> > idea... :(\n> What's a WAREZ file :-?\n\nBasically it is stolen software on an FTP site. For example, check out\nwarez.druid.net with your ftp client.\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": "Wed, 11 Feb 1998 09:32:09 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/Tcl" }, { "msg_contents": "On Wed, 11 Feb 1998, Jan Wieck wrote:\n\n> > \n> > On Wed, 11 Feb 1998, Jan Wieck wrote:\n> > \n> > > Hi,\n> > > \n> > > hmmm - no access to /pub/incoming on ftp.postgresql.org?\n> > \n> > \tSomeone decided that storing WAREZ files in there was a good\n> > idea... :(\n> > \n> > \n> > \n> \n> What's a WAREZ file :-?\n\n\tPirated, commercial software...:(\n\n\n", "msg_date": "Wed, 11 Feb 1998 09:41:57 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/Tcl" }, { "msg_contents": "On Wed, 11 Feb 1998, D'Arcy J.M. Cain wrote:\n\n> Thus spake Jan Wieck\n> > > \tSomeone decided that storing WAREZ files in there was a good\n> > > idea... :(\n> > What's a WAREZ file :-?\n> \n> Basically it is stolen software on an FTP site. For example, check out\n> warez.druid.net with your ftp client.\n\n\tHey, that's not funny :( There are no warez there...I checked as\nsoon as I found the www.postgresql.org stuff sigh\n\n\n", "msg_date": "Wed, 11 Feb 1998 09:46:50 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/Tcl" } ]
[ { "msg_contents": "Hi,\n\nI'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\nfar no problems, however I noted some cleanups / enhancements which I\nwould like to do. Before I send you a bunch of patches I thought I'll\ntell you what I'm planning to do. Please comment on my list and indicate\nwhether I should go ahead.\n\n- Fix all Makefiles so 'make dep' and 'make depend' work\n- Fix all Makefiles so 'make clean' throws away the depend file\n- Some other Makefile cleanups\n- gcc 2.8.0 issues some additional warnings which are very easy to fix:\n - register i --> register int i\n - Ambiguous else --> add braces:\n if (cond1)\n if (cond2)\n ...\n else\n ...\n - etc.\n- Add a template for linux-elf-586 with (optimized) code for a Pentium\n(gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\nWhy not use template names similar to the output of config.guess (maybe\nwith some symbolic links)?\n- Fix for tools/find_static: add two (in my opinion funny) indices to\nimprove speed tremendously:\n\nNested Loop (cost=6.05 size=1 width=42)\n -> Index Scan on debug (cost=2.05 size=2 width=12)\n -> Index Scan on debug2 (cost=2.00 size=2 width=30)\n\nrather than\n\nHash Join (cost=993.76 size=1 width=42)\n -> Seq Scan on debug (cost=495.81 size=2 width=12)\n -> Hash (cost=0.00 size=0 width=0)\n -> Seq Scan on debug2 (cost=495.81 size=2 width=30)\n\n- Cleanup of some code that uses heap_formtuple to allow a NULL value\nfor the nulls parameter, indicating there are no null columns (comes in\nhandy for catalog/pg_.c), among others.\n- Why is there some code to change the case of the procedural language\nto lower case except for 'C' (in fact it's there twice)? Why not use\nstrcasecmp and remove these pices of code?\n- Add pcalloc(n,s) to allocate ns bytes and set them to zero and use\nthem where appropriate (I won't touch all code :-))\n- Shouldn't same_tuple in executor/nodeGroup.c use the equality operator\nfor the type concerned to test for equality of the attributes rather\nthan print them to a buffer and use strcmp? Shouldn't the pointers for\nthese functions be looked up once in ExecInitGroup and stored somewhere?\nShouldn't this function go to heaptuple.c and be renamed heap_sametuple?\n- Lump heaptuple.c and heapvalid.c together\n- I also saw quite some #ifdef NOT_USED and other similar stuff. I don't\nwant to touch these now, but shouldn't some of these be removed soon?\n- Add a pg_version function that returns a string like 'PostgreSQL 6.3'\nto indicate the version of PostgreSQL a user is using (with 'select\npg_version()'). Might be handy to include in the bug reports.\n\nThese are all the things that I found after browsing through the code\none night (primarily in backend/access, backend/catalog and\nbackend/executor).\n\nLet me know what you think of the above list and I will proceed. If you\nhave any hints on how I might proceed (especially with same_tuple)\nplease don't hesitate. Expect the changes to be available somewhere\nafter the weekend.\n\nCheers,\n\nJeroen van Vianen\n", "msg_date": "Wed, 11 Feb 1998 10:22:05 +0100", "msg_from": "Jeroen van Vianen <[email protected]>", "msg_from_op": true, "msg_subject": "Some cleanups/enhancements" }, { "msg_contents": "These all sound good to me. \n\nThe NOT_USED stuff was left because we thought some day we may need\nthem. If you see stuff that clearly doesn't do anything valuable, get\nrid of it.\n\n> \n> Hi,\n> \n> I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> far no problems, however I noted some cleanups / enhancements which I\n> would like to do. Before I send you a bunch of patches I thought I'll\n> tell you what I'm planning to do. Please comment on my list and indicate\n> whether I should go ahead.\n> \n> - Fix all Makefiles so 'make dep' and 'make depend' work\n> - Fix all Makefiles so 'make clean' throws away the depend file\n> - Some other Makefile cleanups\n> - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> - register i --> register int i\n> - Ambiguous else --> add braces:\n> if (cond1)\n> if (cond2)\n> ...\n> else\n> ...\n> - etc.\n> - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> Why not use template names similar to the output of config.guess (maybe\n> with some symbolic links)?\n> - Fix for tools/find_static: add two (in my opinion funny) indices to\n> improve speed tremendously:\n> \n> Nested Loop (cost=6.05 size=1 width=42)\n> -> Index Scan on debug (cost=2.05 size=2 width=12)\n> -> Index Scan on debug2 (cost=2.00 size=2 width=30)\n> \n> rather than\n> \n> Hash Join (cost=993.76 size=1 width=42)\n> -> Seq Scan on debug (cost=495.81 size=2 width=12)\n> -> Hash (cost=0.00 size=0 width=0)\n> -> Seq Scan on debug2 (cost=495.81 size=2 width=30)\n> \n> - Cleanup of some code that uses heap_formtuple to allow a NULL value\n> for the nulls parameter, indicating there are no null columns (comes in\n> handy for catalog/pg_.c), among others.\n> - Why is there some code to change the case of the procedural language\n> to lower case except for 'C' (in fact it's there twice)? Why not use\n> strcasecmp and remove these pices of code?\n> - Add pcalloc(n,s) to allocate ns bytes and set them to zero and use\n> them where appropriate (I won't touch all code :-))\n> - Shouldn't same_tuple in executor/nodeGroup.c use the equality operator\n> for the type concerned to test for equality of the attributes rather\n> than print them to a buffer and use strcmp? Shouldn't the pointers for\n> these functions be looked up once in ExecInitGroup and stored somewhere?\n> Shouldn't this function go to heaptuple.c and be renamed heap_sametuple?\n> - Lump heaptuple.c and heapvalid.c together\n> - I also saw quite some #ifdef NOT_USED and other similar stuff. I don't\n> want to touch these now, but shouldn't some of these be removed soon?\n> - Add a pg_version function that returns a string like 'PostgreSQL 6.3'\n> to indicate the version of PostgreSQL a user is using (with 'select\n> pg_version()'). Might be handy to include in the bug reports.\n> \n> These are all the things that I found after browsing through the code\n> one night (primarily in backend/access, backend/catalog and\n> backend/executor).\n> \n> Let me know what you think of the above list and I will proceed. If you\n> have any hints on how I might proceed (especially with same_tuple)\n> please don't hesitate. Expect the changes to be available somewhere\n> after the weekend.\n> \n> Cheers,\n> \n> Jeroen van Vianen\n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 10:27:59 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "On Wed, 11 Feb 1998, Jeroen van Vianen wrote:\n\n> Hi,\n> \n> I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> far no problems, however I noted some cleanups / enhancements which I\n> would like to do. Before I send you a bunch of patches I thought I'll\n> tell you what I'm planning to do. Please comment on my list and indicate\n> whether I should go ahead.\n> \n> - Fix all Makefiles so 'make dep' and 'make depend' work\n\n\tCan someone explain what, exactly, 'make depend' accomplishes? We\ndon't use it right now, so I'm wondering why (if?) we need it now?\n\n> - Some other Makefile cleanups\n> - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> - register i --> register int i\n> - Ambiguous else --> add braces:\n> if (cond1)\n> if (cond2)\n> ...\n> else\n> ...\n> - etc.\n\n\tSounds great...\n\n> - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> Why not use template names similar to the output of config.guess (maybe\n> with some symbolic links)?\n\n\tErk...I think 'templates' are getting a little out of hand here,\nno? \n\n> - Why is there some code to change the case of the procedural language\n> to lower case except for 'C' (in fact it's there twice)? Why not use\n> strcasecmp and remove these pices of code?\n\n\tI question this in alot of places...like why pg_dlopen is\ndefined as just 'dlopen()' in some ports shrug Why not just call it\ndirectly? raised eyebrow\n\n> These are all the things that I found after browsing through the code\n> one night (primarily in backend/access, backend/catalog and\n> backend/executor).\n> \n> Let me know what you think of the above list and I will proceed. If you\n> have any hints on how I might proceed (especially with same_tuple)\n> please don't hesitate. Expect the changes to be available somewhere\n> after the weekend.\n\n\tThe only thing I ask is that you submit these in such a way that\nthey can be easily reviewed before committing them...we are in a beta mode\nright now, and altho some of this makes for nice cleanups, some of this\nshould most likely be gingerly added...\n\n\tIf at all possible, a seperate patch for each point above would be\nreally good, with an explanation of each. If it weren't for beta-status,\nI wouldn't care, since we could debug after, but with only 2/2.5 weeks\ntill release, we are getting tight for debugging...:(\n\n\n\n", "msg_date": "Wed, 11 Feb 1998 10:43:00 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "I recommend running the regression test before/after the changes, to\nmake sure something didn't get broken.\n\n> \n> On Wed, 11 Feb 1998, Jeroen van Vianen wrote:\n> \n> > Hi,\n> > \n> > I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> > far no problems, however I noted some cleanups / enhancements which I\n> > would like to do. Before I send you a bunch of patches I thought I'll\n> > tell you what I'm planning to do. Please comment on my list and indicate\n> > whether I should go ahead.\n> > \n> > - Fix all Makefiles so 'make dep' and 'make depend' work\n> \n> \tCan someone explain what, exactly, 'make depend' accomplishes? We\n> don't use it right now, so I'm wondering why (if?) we need it now?\n> \n> > - Some other Makefile cleanups\n> > - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> > - register i --> register int i\n> > - Ambiguous else --> add braces:\n> > if (cond1)\n> > if (cond2)\n> > ...\n> > else\n> > ...\n> > - etc.\n> \n> \tSounds great...\n> \n> > - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> > Why not use template names similar to the output of config.guess (maybe\n> > with some symbolic links)?\n> \n> \tErk...I think 'templates' are getting a little out of hand here,\n> no? \n> \n> > - Why is there some code to change the case of the procedural language\n> > to lower case except for 'C' (in fact it's there twice)? Why not use\n> > strcasecmp and remove these pices of code?\n> \n> \tI question this in alot of places...like why pg_dlopen is\n> defined as just 'dlopen()' in some ports shrug Why not just call it\n> directly? raised eyebrow\n> \n> > These are all the things that I found after browsing through the code\n> > one night (primarily in backend/access, backend/catalog and\n> > backend/executor).\n> > \n> > Let me know what you think of the above list and I will proceed. If you\n> > have any hints on how I might proceed (especially with same_tuple)\n> > please don't hesitate. Expect the changes to be available somewhere\n> > after the weekend.\n> \n> \tThe only thing I ask is that you submit these in such a way that\n> they can be easily reviewed before committing them...we are in a beta mode\n> right now, and altho some of this makes for nice cleanups, some of this\n> should most likely be gingerly added...\n> \n> \tIf at all possible, a seperate patch for each point above would be\n> really good, with an explanation of each. If it weren't for beta-status,\n> I wouldn't care, since we could debug after, but with only 2/2.5 weeks\n> till release, we are getting tight for debugging...:(\n> \n> \n> \n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 10:52:56 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "> I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> far no problems, however I noted some cleanups / enhancements which I\n> would like to do. Before I send you a bunch of patches I thought I'll\n> tell you what I'm planning to do. Please comment on my list and indicate\n> whether I should go ahead.\n>\n> - Fix all Makefiles so 'make dep' and 'make depend' work\n> - Fix all Makefiles so 'make clean' throws away the depend file\n> - Some other Makefile cleanups\n\nThese all sound good. If there is a possibility of large breakage, wait\nuntil after v6.3.\n\n> - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> - register i --> register int i\n\nI'm sure there will be differing opinions :-/, but imho all register\ndeclarations should be removed. Modern compilers do a good job of\noptimization, and register declarations can get in the way by forcing the\ncompiler to burn a register to accomodate the declared item.\n\n> - Ambiguous else --> add braces:\n> if (cond1)\n> if (cond2)\n> ...\n> else\n> ...\n\nSure.\n\n> - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> Why not use template names similar to the output of config.guess (maybe\n> with some symbolic links)?\n\nDoes gcc 2.7.x support the -mpentium and -mpentiumpro switches? If not,\nthen the template should be more explicit in name (e.g.\n\"linux-elf-586-gcc2.8\") or we should update the FAQ or include comments in\nlinux-elf with some suggestions. It was only in the last release or two\nthat the -m486 was added, and I worried about causing trouble for _all_ of\nthose 386 users :)\n\n> - Shouldn't same_tuple in executor/nodeGroup.c use the equality operator\n> for the type concerned to test for equality of the attributes rather\n> than print them to a buffer and use strcmp? Shouldn't the pointers for\n> these functions be looked up once in ExecInitGroup and stored somewhere?\n> Shouldn't this function go to heaptuple.c and be renamed heap_sametuple?\n\nYes, I would think so. The only downside to this is that, since two items\nwhich fail the equality test may look identical when formatted (e.g.\nfloating point numbers with the lsb differing) the results may look a bit\nfunny and be difficult to track down. Still, I think this is the right way\nto go...\n\n> - Lump heaptuple.c and heapvalid.c together\n> - I also saw quite some #ifdef NOT_USED and other similar stuff. I don't\n> want to touch these now, but shouldn't some of these be removed soon?\n\nOnly when the module is completely understood. So, don't remove blindly,\nbut if it is clear that it is obsolete code which is not providing hints on\nwhat should be done in the future, then it is OK to remove.\n\n> - Add a pg_version function that returns a string like 'PostgreSQL 6.3'\n> to indicate the version of PostgreSQL a user is using (with 'select\n> pg_version()'). Might be handy to include in the bug reports.\n\nGood idea.\n\nSome or all of these changes might not be appropriate for v6.3, since we\nare in beta testing and since they do not affect the current functionality.\nFor those cases, how about submitting patches based on the final v6.3\nrelease?\n\n - Tom\n\n", "msg_date": "Wed, 11 Feb 1998 16:03:51 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "> \n> > I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> > far no problems, however I noted some cleanups / enhancements which I\n> > would like to do. Before I send you a bunch of patches I thought I'll\n> > tell you what I'm planning to do. Please comment on my list and indicate\n> > whether I should go ahead.\n> >\n> > - Fix all Makefiles so 'make dep' and 'make depend' work\n> > - Fix all Makefiles so 'make clean' throws away the depend file\n> > - Some other Makefile cleanups\n> \n> These all sound good. If there is a possibility of large breakage, wait\n> until after v6.3.\n> \n> > - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> > - register i --> register int i\n> \n> I'm sure there will be differing opinions :-/, but imho all register\n> declarations should be removed. Modern compilers do a good job of\n> optimization, and register declarations can get in the way by forcing the\n> compiler to burn a register to accomodate the declared item.\n\nTotally agree. Get rid of all register's competely. I don't think this\nwill affect Vadim, as most of them are in utility directories.\n\nI agree with your other points too.\n\n\n> \n> > - Ambiguous else --> add braces:\n> > if (cond1)\n> > if (cond2)\n> > ...\n> > else\n> > ...\n> \n> Sure.\n> \n> > - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> > Why not use template names similar to the output of config.guess (maybe\n> > with some symbolic links)?\n> \n> Does gcc 2.7.x support the -mpentium and -mpentiumpro switches? If not,\n> then the template should be more explicit in name (e.g.\n> \"linux-elf-586-gcc2.8\") or we should update the FAQ or include comments in\n> linux-elf with some suggestions. It was only in the last release or two\n> that the -m486 was added, and I worried about causing trouble for _all_ of\n> those 386 users :)\n> \n> > - Shouldn't same_tuple in executor/nodeGroup.c use the equality operator\n> > for the type concerned to test for equality of the attributes rather\n> > than print them to a buffer and use strcmp? Shouldn't the pointers for\n> > these functions be looked up once in ExecInitGroup and stored somewhere?\n> > Shouldn't this function go to heaptuple.c and be renamed heap_sametuple?\n> \n> Yes, I would think so. The only downside to this is that, since two items\n> which fail the equality test may look identical when formatted (e.g.\n> floating point numbers with the lsb differing) the results may look a bit\n> funny and be difficult to track down. Still, I think this is the right way\n> to go...\n> \n> > - Lump heaptuple.c and heapvalid.c together\n> > - I also saw quite some #ifdef NOT_USED and other similar stuff. I don't\n> > want to touch these now, but shouldn't some of these be removed soon?\n> \n> Only when the module is completely understood. So, don't remove blindly,\n> but if it is clear that it is obsolete code which is not providing hints on\n> what should be done in the future, then it is OK to remove.\n> \n> > - Add a pg_version function that returns a string like 'PostgreSQL 6.3'\n> > to indicate the version of PostgreSQL a user is using (with 'select\n> > pg_version()'). Might be handy to include in the bug reports.\n> \n> Good idea.\n> \n> Some or all of these changes might not be appropriate for v6.3, since we\n> are in beta testing and since they do not affect the current functionality.\n> For those cases, how about submitting patches based on the final v6.3\n> release?\n> \n> - Tom\n> \n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 11:18:11 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nThen <[email protected]> spoke up and said:\n> > - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> > Why not use template names similar to the output of config.guess (maybe\n> > with some symbolic links)?\n\nIMHO, none of the gcc-specific optimization switches belong in any\ntemplates. If you're using gcc on a pentium, and your version\nsupportes -mpentium, then it should be in the SPECS file in your gcc\ninstallation. Admittedly, this presumes a certain clue level on the\npart of the gcc installer/maintainer, but it reduces the clutter level\nwith templates somewhat.\n\n- -- \n\n\n=====================================================================\n\| \"If you're all through trying to burn the field down, will you \|\n\| kindly get up and tell me why you're sitting in a fruit field, \|\n\| stark naked, frying peaches?\" \|\n=====================================================================\n\| Finger [email protected] for my public key. \|\n=====================================================================\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQBVAwUBNOHZkIdzVnzma+gdAQEg7QH+MOviZ+pcq5wdpE1d7tK3yB2Ai/03qGti\n7cBxzMQLq82g1/5wT+lXm9Rh3plbyvTBCUpU48kpXTYAYjeAvVIzzQ==\n=C0DN\n-----END PGP SIGNATURE-----\n\n", "msg_date": "11 Feb 1998 12:02:09 -0500", "msg_from": "Brian E Gallew <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "Thomas G. Lockhart wrote:\n> \n> > I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> > far no problems, however I noted some cleanups / enhancements which I\n> > would like to do. Before I send you a bunch of patches I thought I'll\n> > tell you what I'm planning to do. Please comment on my list and indicate\n> > whether I should go ahead.\n> >\n> > - Fix all Makefiles so 'make dep' and 'make depend' work\n> > - Fix all Makefiles so 'make clean' throws away the depend file\n> > - Some other Makefile cleanups\n> \n> These all sound good. If there is a possibility of large breakage, wait\n> until after v6.3.\n\n> Some or all of these changes might not be appropriate for v6.3, since we\n> are in beta testing and since they do not affect the current functionality.\n> For those cases, how about submitting patches based on the final v6.3\n> release?\n\nAfter the messages I've read so far, I'll wait until after the final\nrelease of 6.3 and try to do the patches one at a time, so there'll be\nplenty of time :-) to review them.\n\n> [snip]\n\n> > - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> > Why not use template names similar to the output of config.guess (maybe\n> > with some symbolic links)?\n> \n> Does gcc 2.7.x support the -mpentium and -mpentiumpro switches? If not,\n> then the template should be more explicit in name (e.g.\n> \"linux-elf-586-gcc2.8\") or we should update the FAQ or include comments in\n> linux-elf with some suggestions. It was only in the last release or two\n> that the -m486 was added, and I worried about causing trouble for _all_ of\n> those 386 users :)\n\nNo, it doesn't. linux-elf-586-gcc2.8 sounds OK to me.\n\n> [snip]\n\n\nThe Hermit Hacker wrote:\n> > - Fix all Makefiles so 'make dep' and 'make depend' work\n> \n> Can someone explain what, exactly, 'make depend' accomplishes? We\n> don't use it right now, so I'm wondering why (if?) we need it now?\n\nIt makes sure that your C files get compiled if you change a header\nfile. Your C compiler should be able to find out which files are\nincluded and create lines which can be included in the Makefile (man gcc\n:-) )\n\n> > - Some other Makefile cleanups\n> > - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> > - register i --> register int i\n> > - Ambiguous else --> add braces:\n> > if (cond1)\n> > if (cond2)\n> > ...\n> > else\n> > ...\n> > - etc.\n> \n> Sounds great...\n\nIf I find something like this, I'll remove the register as well.\n\n> > - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> > Why not use template names similar to the output of config.guess (maybe\n> > with some symbolic links)?\n> \n> Erk...I think 'templates' are getting a little out of hand here,\n> no?\n> \n> > - Why is there some code to change the case of the procedural language\n> > to lower case except for 'C' (in fact it's there twice)? Why not use\n> > strcasecmp and remove these pices of code?\n> \n> I question this in alot of places...like why pg_dlopen is\n> defined as just 'dlopen()' in some ports shrug Why not just call it\n> directly? raised eyebrow\n> \n\n[snip]\n\nBruce Momjian wrote:\n> \n> I recommend running the regression test before/after the changes, to\n> make sure something didn't get broken.\n\nSure.\n\n> [snip]\n\nSee you in a 2-2.5 weeks :-)\n\nJeroen van Vianen\n", "msg_date": "Wed, 11 Feb 1998 18:10:20 +0100", "msg_from": "Jeroen van Vianen <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "> > > - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> > > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> > > Why not use template names similar to the output of config.guess (maybe\n> > > with some symbolic links)?\n>\n> IMHO, none of the gcc-specific optimization switches belong in any\n> templates. If you're using gcc on a pentium, and your version\n> supportes -mpentium, then it should be in the SPECS file in your gcc\n> installation. Admittedly, this presumes a certain clue level on the\n> part of the gcc installer/maintainer, but it reduces the clutter level\n> with templates somewhat.\n\nGreat! Want to write it up for the docs? Need a cookbook and an explanation;\nI'll add it in...\n\nIf it requires a new install of gcc, then the other option might be to include\nit in Makefile.custom as\n\n CFLAGS+= -mpentium\n\n - Tom\n\n", "msg_date": "Wed, 11 Feb 1998 17:19:08 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "> \n> Thomas G. Lockhart wrote:\n> > \n> > > I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> > > far no problems, however I noted some cleanups / enhancements which I\n> > > would like to do. Before I send you a bunch of patches I thought I'll\n> > > tell you what I'm planning to do. Please comment on my list and indicate\n> > > whether I should go ahead.\n> > >\n> > > - Fix all Makefiles so 'make dep' and 'make depend' work\n> > > - Fix all Makefiles so 'make clean' throws away the depend file\n> > > - Some other Makefile cleanups\n> > \n> > These all sound good. If there is a possibility of large breakage, wait\n> > until after v6.3.\n> \n> > Some or all of these changes might not be appropriate for v6.3, since we\n> > are in beta testing and since they do not affect the current functionality.\n> > For those cases, how about submitting patches based on the final v6.3\n> > release?\n> \n> After the messages I've read so far, I'll wait until after the final\n> release of 6.3 and try to do the patches one at a time, so there'll be\n> plenty of time :-) to review them.\n> \n> > [snip]\n> \n> > > - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> > > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> > > Why not use template names similar to the output of config.guess (maybe\n> > > with some symbolic links)?\n> > \n> > Does gcc 2.7.x support the -mpentium and -mpentiumpro switches? If not,\n> > then the template should be more explicit in name (e.g.\n> > \"linux-elf-586-gcc2.8\") or we should update the FAQ or include comments in\n> > linux-elf with some suggestions. It was only in the last release or two\n> > that the -m486 was added, and I worried about causing trouble for _all_ of\n> > those 386 users :)\n> \n> No, it doesn't. linux-elf-586-gcc2.8 sounds OK to me.\n> > > - Some other Makefile cleanups\n> > > - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> > > - register i --> register int i\n> > > - Ambiguous else --> add braces:\n> > > if (cond1)\n> > > if (cond2)\n> > > ...\n> > > else\n> > > ...\n> > > - etc.\n> > \n> > Sounds great...\n> \n> If I find something like this, I'll remove the register as well.\n\nRegister is gone. Just removed them.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 14:12:14 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "[email protected], can you send in patches for these?\n\n> \n> > I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> > far no problems, however I noted some cleanups / enhancements which I\n> > would like to do. Before I send you a bunch of patches I thought I'll\n> > tell you what I'm planning to do. Please comment on my list and indicate\n> > whether I should go ahead.\n> >\n> > - Fix all Makefiles so 'make dep' and 'make depend' work\n> > - Fix all Makefiles so 'make clean' throws away the depend file\n> > - Some other Makefile cleanups\n> \n> These all sound good. If there is a possibility of large breakage, wait\n> until after v6.3.\n> \n> > - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> > - register i --> register int i\n> \n> I'm sure there will be differing opinions :-/, but imho all register\n> declarations should be removed. Modern compilers do a good job of\n> optimization, and register declarations can get in the way by forcing the\n> compiler to burn a register to accomodate the declared item.\n> \n> > - Ambiguous else --> add braces:\n> > if (cond1)\n> > if (cond2)\n> > ...\n> > else\n> > ...\n> \n> Sure.\n> \n> > - Add a template for linux-elf-586 with (optimized) code for a Pentium\n> > (gcc 2.8.0 not only supports -m486 but also -mpentium and -mpentiumpro).\n> > Why not use template names similar to the output of config.guess (maybe\n> > with some symbolic links)?\n> \n> Does gcc 2.7.x support the -mpentium and -mpentiumpro switches? If not,\n> then the template should be more explicit in name (e.g.\n> \"linux-elf-586-gcc2.8\") or we should update the FAQ or include comments in\n> linux-elf with some suggestions. It was only in the last release or two\n> that the -m486 was added, and I worried about causing trouble for _all_ of\n> those 386 users :)\n> \n> > - Shouldn't same_tuple in executor/nodeGroup.c use the equality operator\n> > for the type concerned to test for equality of the attributes rather\n> > than print them to a buffer and use strcmp? Shouldn't the pointers for\n> > these functions be looked up once in ExecInitGroup and stored somewhere?\n> > Shouldn't this function go to heaptuple.c and be renamed heap_sametuple?\n> \n> Yes, I would think so. The only downside to this is that, since two items\n> which fail the equality test may look identical when formatted (e.g.\n> floating point numbers with the lsb differing) the results may look a bit\n> funny and be difficult to track down. Still, I think this is the right way\n> to go...\n> \n> > - Lump heaptuple.c and heapvalid.c together\n> > - I also saw quite some #ifdef NOT_USED and other similar stuff. I don't\n> > want to touch these now, but shouldn't some of these be removed soon?\n> \n> Only when the module is completely understood. So, don't remove blindly,\n> but if it is clear that it is obsolete code which is not providing hints on\n> what should be done in the future, then it is OK to remove.\n> \n> > - Add a pg_version function that returns a string like 'PostgreSQL 6.3'\n> > to indicate the version of PostgreSQL a user is using (with 'select\n> > pg_version()'). Might be handy to include in the bug reports.\n> \n> Good idea.\n> \n> Some or all of these changes might not be appropriate for v6.3, since we\n> are in beta testing and since they do not affect the current functionality.\n> For those cases, how about submitting patches based on the final v6.3\n> release?\n> \n> - Tom\n> \n> \n> \n\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Sun, 15 Mar 1998 23:59:32 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> [email protected], can you send in patches for these?\n> \n> >\n> > > I'm running PostgreSQL 6.3 on Linux 2.1.85 with gcc 2.8.0 and libc5. So\n> > > far no problems, however I noted some cleanups / enhancements which I\n> > > would like to do. Before I send you a bunch of patches I thought I'll\n> > > tell you what I'm planning to do. Please comment on my list and indicate\n> > > whether I should go ahead.\n> > >\n> > > - Fix all Makefiles so 'make dep' and 'make depend' work\n> > > - Fix all Makefiles so 'make clean' throws away the depend file\n> > > - Some other Makefile cleanups\n> >\n> > These all sound good. If there is a possibility of large breakage, wait\n> > until after v6.3.\n> >\n> > > - gcc 2.8.0 issues some additional warnings which are very easy to fix:\n> > > - register i --> register int i\n\nThere's a patch attached to fix gcc 2.8.x warnings, except for the\nyyerror ones from bison. It also includes a few 'enhancements' to the C\nprogramming style (which are, of course, personal).\n\nThe other patch removes the compilation of backend/lib/qsort.c, as\nqsort() is a standard function in stdlib.h and can be used any where\nelse (and it is). It was only used in\nbackend/optimizer/geqo/geqo_pool.c, backend/optimizer/path/predmig.c,\nand backend/storage/page/bufpage.c\n\n> > Some or all of these changes might not be appropriate for v6.3, since we\n> > are in beta testing and since they do not affect the current functionality.\n> > For those cases, how about submitting patches based on the final v6.3\n> > release?\n\nThere's more to come. Please review these patches. I ran the regression\ntests and they only failed where this was expected (random, geo, etc).\n\nCheers,\n\nJeroen", "msg_date": "Tue, 17 Mar 1998 19:42:11 +0100", "msg_from": "Jeroen van Vianen <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "> There's a patch attached to fix gcc 2.8.x warnings, except for the\n> yyerror ones from bison. It also includes a few 'enhancements' to the C\n> programming style (which are, of course, personal).\n> \n> The other patch removes the compilation of backend/lib/qsort.c, as\n> qsort() is a standard function in stdlib.h and can be used any where\n> else (and it is). It was only used in\n> backend/optimizer/geqo/geqo_pool.c, backend/optimizer/path/predmig.c,\n> and backend/storage/page/bufpage.c\n> \n\nWoh, this is some pretty serious code. Nice.\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Tue, 17 Mar 1998 14:07:53 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" }, { "msg_contents": "Applied. Looked good, and qsort.c removed as you suggested.\n\n\n> There's a patch attached to fix gcc 2.8.x warnings, except for the\n> yyerror ones from bison. It also includes a few 'enhancements' to the C\n> programming style (which are, of course, personal).\n> \n> The other patch removes the compilation of backend/lib/qsort.c, as\n> qsort() is a standard function in stdlib.h and can be used any where\n> else (and it is). It was only used in\n> backend/optimizer/geqo/geqo_pool.c, backend/optimizer/path/predmig.c,\n> and backend/storage/page/bufpage.c\n> \n> > > Some or all of these changes might not be appropriate for v6.3, since we\n> > > are in beta testing and since they do not affect the current functionality.\n> > > For those cases, how about submitting patches based on the final v6.3\n> > > release?\n> \n> There's more to come. Please review these patches. I ran the regression\n> tests and they only failed where this was expected (random, geo, etc).\n> \n> Cheers,\n> \n> Jeroen\n\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Mon, 30 Mar 1998 11:45:01 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Some cleanups/enhancements" } ]
[ { "msg_contents": "I ran my performance tests some more times and it seems the numbers are not\nreally comparable. When I run PostgreSQL without -F I get a sync after every\ninsert. With -F I get no sync at all as all inserts fit well into the\nbuffer. However, Oracle in comparison does sync. Simply hearing the disk\naccess it seems as if they sync every two or three seconds.\n\nDoes anyone know a way to really check both DBMSs?\n\nMichael\n\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Wed, 11 Feb 1998 10:36:03 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "Problem with the numbers I reported yesterday" }, { "msg_contents": "> \n> I ran my performance tests some more times and it seems the numbers are not\n> really comparable. When I run PostgreSQL without -F I get a sync after every\n> insert. With -F I get no sync at all as all inserts fit well into the\n> buffer. However, Oracle in comparison does sync. Simply hearing the disk\n> access it seems as if they sync every two or three seconds.\n> \n> Does anyone know a way to really check both DBMSs?\n\nMany dbms's do buffered logging, that is they sync after the buffer gets\nfull or after a minute or so. We have the logic to add buffered logging\nto PostgreSQL and will be doing it later. Right now, we only have\nnon-buffered logging, and no logging.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 10:31:20 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Problem with the numbers I reported yesterday" }, { "msg_contents": "> \n> Hi,\n> \n> perhaps a stupid question:\n> \n> What will happen, if you start PostgreSQL with -F and write a script \n> which is a loop that fsyncs every 2 seconds? Does this lead to a \n> database which is \"almost\" correct? Does this seem like a good \n> compromise?\n\nWell, you really have to sync the data pages BEFORE sync'ing pg_log. \nOur buffered logging idea for post-6.3 will do exactly that. I don't\nthink it is a workaround. You could get pg_log to disk saying a\ntransaction is complete without getting all the data pages to disk if\nthe crash happened during the sync.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 09:47:43 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Problem with the numbers I reported yesterday" }, { "msg_contents": "Hi,\n\nperhaps a stupid question:\n\nWhat will happen, if you start PostgreSQL with -F and write a script \nwhich is a loop that fsyncs every 2 seconds? Does this lead to a \ndatabase which is \"almost\" correct? Does this seem like a good \ncompromise?\n\n> > I ran my performance tests some more times and it seems the numbers are not\n> > really comparable. When I run PostgreSQL without -F I get a sync after every\n> > insert. With -F I get no sync at all as all inserts fit well into the\n> > buffer. However, Oracle in comparison does sync. Simply hearing the disk\n> > access it seems as if they sync every two or three seconds.\n> > \n> > Does anyone know a way to really check both DBMSs?\n> \n> Many dbms's do buffered logging, that is they sync after the buffer gets\n> full or after a minute or so. We have the logic to add buffered logging\n> to PostgreSQL and will be doing it later. Right now, we only have\n> non-buffered logging, and no logging.\n> \n> -- \n> Bruce Momjian\n> [email protected]\n> \n> \n\nCiao\n\nDas Boersenspielteam.\n\n---------------------------------------------------------------------------\n http://www.boersenspiel.de\n \t Das Boersenspiel im Internet\n Realitaetsnah Kostenlos Ueber 6000 Spieler\n---------------------------------------------------------------------------\n", "msg_date": "Thu, 12 Feb 1998 15:41:45 +0000", "msg_from": "\"Boersenspielteam\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Problem with the numbers I reported yesterday" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> > \n> > Hi,\n> > \n> > perhaps a stupid question:\n> > \n> > What will happen, if you start PostgreSQL with -F and write a script \n> > which is a loop that fsyncs every 2 seconds? Does this lead to a \n> > database which is \"almost\" correct? Does this seem like a good \n> > compromise?\n> \n> Well, you really have to sync the data pages BEFORE sync'ing pg_log. \n\nWhy should this be necessary? If the transaction is considered\ncommitted once the log has be written, and the system crashes before\nthe data are written, then postgres can look at the data and logs when\nit is next started up and apply all the transactions that were logged\nbut not committed to the data pages.\n\nAm I missing something? It seems to me if you sync the data pages\nfirst, then what good is the log? (other than being able to retrace\nyour steps)\n\nOcie Mitchell\n", "msg_date": "Thu, 12 Feb 1998 10:22:47 -0800 (PST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Problem with the numbers I reported yesterday" }, { "msg_contents": "> \n> Bruce Momjian wrote:\n> > \n> > > \n> > > Hi,\n> > > \n> > > perhaps a stupid question:\n> > > \n> > > What will happen, if you start PostgreSQL with -F and write a script \n> > > which is a loop that fsyncs every 2 seconds? Does this lead to a \n> > > database which is \"almost\" correct? Does this seem like a good \n> > > compromise?\n> > \n> > Well, you really have to sync the data pages BEFORE sync'ing pg_log. \n> \n> Why should this be necessary? If the transaction is considered\n> committed once the log has be written, and the system crashes before\n> the data are written, then postgres can look at the data and logs when\n> it is next started up and apply all the transactions that were logged\n> but not committed to the data pages.\n\nNo, on restart, you can't identify the old/new data. Remember, pg_log\nis just the transaction id and a flag. The superceeded/new rows are\nmixed on the data pages, with transaction id's as markers.\n\n\n> \n> Am I missing something? It seems to me if you sync the data pages\n> first, then what good is the log? (other than being able to retrace\n> your steps)\n\nAgain, the log is just a list of transaction ids, and their statuses.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 14:16:29 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Problem with the numbers I reported yesterday" }, { "msg_contents": ">>>>> \"ocie\" == ocie <[email protected]> writes:\n\n > Bruce Momjian wrote:\n >> > > Hi, > > perhaps a stupid question: > > What will happen,\n >> if you start PostgreSQL with -F and write a script > which is a\n >> loop that fsyncs every 2 seconds? Does this lead to a >\n >> database which is \"almost\" correct? Does this seem like a good\n >> > compromise?\n >> \n >> Well, you really have to sync the data pages BEFORE sync'ing\n >> pg_log.\n\n > Why should this be necessary? If the transaction is considered\n > committed once the log has be written, and the system crashes\n > before the data are written, then postgres can look at the data\n > and logs when it is next started up and apply all the\n > transactions that were logged but not committed to the data\n > pages.\n\n > Am I missing something? It seems to me if you sync the data\n > pages first, then what good is the log? (other than being able\n > to retrace your steps)\n\nI do not think that pg_log is used like a normal 'log' device in other\ndatabases. My quick look at the code looks like pg_log only has a\nlist of transactions and not the actual data blocks. Notice that\nTransRecover is commented out in backent/access/transam/transam.c.\n\nMost database log has the before images and after images of any page\nthat has been modified in a transaction followed by commit/abort\nrecord. This allows for only this file to have to be synced. The\nrest of the database can float (generally checkpoints are done every\nso often to reduce recover time). The method of recovering from a\ncrash is to replay the log from the last checkpoint until the end of\nthe log by applying the before/after images (as needed based on\nweather the transaction commited) to the actual database relations.\n\nI would appreciate anyone correcting any mistakes in my understanding\nof how postgres works.\n\n > Ocie Mitchell\n\nKent S. Gordon\nArchitect\niNetSpace Co.\nvoice: (972)851-3494 fax:(972)702-0384 e-mail:[email protected]\n", "msg_date": "Fri, 13 Feb 1998 11:33:22 -0600 (CST)", "msg_from": "\"Kent S. Gordon\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Problem with the numbers I reported yesterday" }, { "msg_contents": "This is 100% correct. See my other posting describing the issues.\n\n> \n> >>>>> \"ocie\" == ocie <[email protected]> writes:\n> \n> > Bruce Momjian wrote:\n> >> > > Hi, > > perhaps a stupid question: > > What will happen,\n> >> if you start PostgreSQL with -F and write a script > which is a\n> >> loop that fsyncs every 2 seconds? Does this lead to a >\n> >> database which is \"almost\" correct? Does this seem like a good\n> >> > compromise?\n> >> \n> >> Well, you really have to sync the data pages BEFORE sync'ing\n> >> pg_log.\n> \n> > Why should this be necessary? If the transaction is considered\n> > committed once the log has be written, and the system crashes\n> > before the data are written, then postgres can look at the data\n> > and logs when it is next started up and apply all the\n> > transactions that were logged but not committed to the data\n> > pages.\n> \n> > Am I missing something? It seems to me if you sync the data\n> > pages first, then what good is the log? (other than being able\n> > to retrace your steps)\n> \n> I do not think that pg_log is used like a normal 'log' device in other\n> databases. My quick look at the code looks like pg_log only has a\n> list of transactions and not the actual data blocks. Notice that\n> TransRecover is commented out in backent/access/transam/transam.c.\n> \n> Most database log has the before images and after images of any page\n> that has been modified in a transaction followed by commit/abort\n> record. This allows for only this file to have to be synced. The\n> rest of the database can float (generally checkpoints are done every\n> so often to reduce recover time). The method of recovering from a\n> crash is to replay the log from the last checkpoint until the end of\n> the log by applying the before/after images (as needed based on\n> weather the transaction commited) to the actual database relations.\n> \n> I would appreciate anyone correcting any mistakes in my understanding\n> of how postgres works.\n> \n> > Ocie Mitchell\n> \n> Kent S. Gordon\n> Architect\n> iNetSpace Co.\n> voice: (972)851-3494 fax:(972)702-0384 e-mail:[email protected]\n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 14:34:09 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Problem with the numbers I reported yesterday" }, { "msg_contents": "Kent wrote:\n>\n> I do not think that pg_log is used like a normal 'log' device in other\n> databases. My quick look at the code looks like pg_log only has a\n> list of transactions and not the actual data blocks. Notice that\n> TransRecover is commented out in backent/access/transam/transam.c.\n>\n> Most database log has the before images and after images of any page\n> that has been modified in a transaction followed by commit/abort\n> record. This allows for only this file to have to be synced. The\n> rest of the database can float (generally checkpoints are done every\n> so often to reduce recover time). The method of recovering from a\n> crash is to replay the log from the last checkpoint until the end of\n> the log by applying the before/after images (as needed based on\n> weather the transaction commited) to the actual database relations.\n>\n> I would appreciate anyone correcting any mistakes in my understanding\n> of how postgres works.\n>\n> > Ocie Mitchell\n>\n> Kent S. Gordon\n> Architect\n> iNetSpace Co.\n> voice: (972)851-3494 fax:(972)702-0384 e-mail:[email protected]\n>\n>\n\n Totally right, PostgreSQL doesn't have a log mechanism that\n collects all the information to recover a corrupted database\n from a backup.\n\n I hacked around on that a little bit.\n\n When doing a complete after image logging, that is taking all\n the tuples that are stored on insert/update, the tuple id's\n of deletes plus the information about transaction id's that\n commit, the regression tests produce log data that is more\n than the size of the final regression database. The\n performance increase when only syncing the log- and\n controlfiles (2 control files on different devices and the\n logfile on a different device from the database files) and\n running the backends with -F is about 15-20% for the\n regression test.\n\n I thought this is far too much logging data and so I didn't\n spent much time trying to implement a recovery. But as far as\n I got it I can tell that the updates to system catalogs and\n keeping the indices up to date will be really tricky.\n\n Another possible log mechanism I'll try sometimes after v6.3\n release is to log the queries and data from copy commands\n along with informations about Oid and Tid allocations.\n\n\nUntil later, 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", "msg_date": "Sat, 14 Feb 1998 14:58:05 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "DB logging (was: Problem with the numbers I reported yesterday)" }, { "msg_contents": ">>>>> \"jwieck\" == Jan Wieck <[email protected]> writes:\n\n > Kent wrote:\ndescription of current logging method deleted.\n\n > Totally right, PostgreSQL doesn't have a log mechanism that\n > collects all the information to recover a corrupted database\n > from a backup.\n\n > I hacked around on that a little bit.\n\n > When doing a complete after image logging, that is taking\n > all the tuples that are stored on insert/update, the tuple id's\n > of deletes plus the information about transaction id's that\n > commit, the regression tests produce log data that is more than\n > the size of the final regression database. The performance\n > increase when only syncing the log- and controlfiles (2 control\n > files on different devices and the logfile on a different device\n > from the database files) and running the backends with -F is\n > about 15-20% for the regression test.\n\nLog files do get very big with image logging. I would not expect a\nhuge win in performance unless the log device is a raw device. On a\ncooked device (file system) buffer cache effects are very large (all\ndisk data is being buffered both by postgresql and the OS buffer\ncache. The buffer cache is actual harmfully in this case, since data\nis not reused, and the writes are synced. The number of writes to the\nlog also flush out other buffer from the cache leading to even more\nio.). If a system does not have raw devices (linux, NT), it would be\nvery useful if a flag exists to tell the OS that the file will be read\nsequential like in the madvise() call for mmap. Is your code\navailable anywhere?\n\n > I thought this is far too much logging data and so I didn't\n > spent much time trying to implement a recovery. But as far as I\n > got it I can tell that the updates to system catalogs and\n > keeping the indices up to date will be really tricky.\n\nI have not looked at this area of the code. Do the system catalogs\nhave a separate storage manager? I do not see why the could not be\nhandled like any other data except for keeping the buffer in the cache.\n\nKent S. Gordon\nArchitect\niNetSpace Co.\nvoice: (972)851-3494 fax:(972)702-0384 e-mail:[email protected]\n\n\n", "msg_date": "Mon, 16 Feb 1998 08:18:31 -0600 (CST)", "msg_from": "\"Kent S. Gordon\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: DB logging (was: Problem with the numbers I reported yesterday)" }, { "msg_contents": "Kent S. Gordon wrote:\n\n[SNIP]\n\n> Log files do get very big with image logging. I would not expect a\n> huge win in performance unless the log device is a raw device. On a\n> cooked device (file system) buffer cache effects are very large (all\n> disk data is being buffered both by postgresql and the OS buffer\n> cache. The buffer cache is actual harmfully in this case, since data\n> is not reused, and the writes are synced. The number of writes to the\n> log also flush out other buffer from the cache leading to even more\n> io.). If a system does not have raw devices (linux, NT), it would be\n ^^^^^ \nWhat exactly do you mean by \"raw devices\" that it is not offered by\nLinux? If I take a partition of one of my hard drives and I don't\nmake a filesystem on it, I can perform reads and writes on the \"raw\ndevice\" /dev/hd?? or /dev/sd?? I didn't think these writes were\nbuffered (if that's what you're referring to), but I could be wrong.\n\nOcie Mitchell.\n", "msg_date": "Mon, 16 Feb 1998 17:44:46 -0800 (PST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: DB logging (was: Problem with the numbers I reported yesterday)" }, { "msg_contents": "> What exactly do you mean by \"raw devices\" that it is not offered by\n> Linux? If I take a partition of one of my hard drives and I don't\n> make a filesystem on it, I can perform reads and writes on the \"raw\n> device\" /dev/hd?? or /dev/sd?? I didn't think these writes were\n> buffered (if that's what you're referring to), but I could be wrong.\n\nYour /dev/hd* goes through the the buffer cache, the raw versions\n/dev/rhd* does not.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Mon, 16 Feb 1998 22:15:58 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: DB logging (was: Problem with the numbers I reported yesterday)" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> > What exactly do you mean by \"raw devices\" that it is not offered by\n> > Linux? If I take a partition of one of my hard drives and I don't\n> > make a filesystem on it, I can perform reads and writes on the \"raw\n> > device\" /dev/hd?? or /dev/sd?? I didn't think these writes were\n> > buffered (if that's what you're referring to), but I could be wrong.\n> \n> Your /dev/hd* goes through the the buffer cache, the raw versions\n> /dev/rhd* does not.\n\nActually on Linux, there is no raw/cooked drive interface as in\nSolaris. In Solaris, the /dev/dsk/ devices are buffered by the OS,\nwhile their counterparts in /dev/rdsk are not. Linux only has the one\ninterface to the partition, which is raw. Code which uses these raw\ndevices (the filesystem code) must supply its own buffering.\n\nAnyway, I don't want to prolong this tangential topic. Linux should\nprovide raw devices, but does not (as Solaris does) provide buffered\nor cooked access to disks.\n\nOcie Mitchell\n\n", "msg_date": "Mon, 16 Feb 1998 20:21:25 -0800 (PST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: DB logging (was: Problem with the numbers I reported yesterday)" }, { "msg_contents": "\nKent wrote:\n>\n> >>>>> \"jwieck\" == Jan Wieck <[email protected]> writes:\n> > When doing a complete after image logging, that is taking\n> > all the tuples that are stored on insert/update, the tuple id's\n> > of deletes plus the information about transaction id's that\n> > commit, the regression tests produce log data that is more than\n> > the size of the final regression database. The performance\n> > increase when only syncing the log- and controlfiles (2 control\n> > files on different devices and the logfile on a different device\n> > from the database files) and running the backends with -F is\n> > about 15-20% for the regression test.\n>\n> Log files do get very big with image logging. I would not expect a\n> huge win in performance unless the log device is a raw device. On a\n> cooked device (file system) buffer cache effects are very large (all\n> disk data is being buffered both by postgresql and the OS buffer\n> cache. The buffer cache is actual harmfully in this case, since data\n> is not reused, and the writes are synced. The number of writes to the\n> log also flush out other buffer from the cache leading to even more\n> io.). If a system does not have raw devices (linux, NT), it would be\n> very useful if a flag exists to tell the OS that the file will be read\n> sequential like in the madvise() call for mmap. Is your code\n> available anywhere?\n\n I don't have that code any more. It wasn't that much so I can\n redo it if at least you would like to help on that topic. But\n since this will be a feature we should wait for the 6.3\n release before touching anything.\n\n>\n> > I thought this is far too much logging data and so I didn't\n> > spent much time trying to implement a recovery. But as far as I\n> > got it I can tell that the updates to system catalogs and\n> > keeping the indices up to date will be really tricky.\n>\n> I have not looked at this area of the code. Do the system catalogs\n> have a separate storage manager? I do not see why the could not be\n> handled like any other data except for keeping the buffer in the cache.\n\n I just had some problems on the system catalogs (maybe due to\n the system caching). I think that it can be handled somehow.\n\n There are other details in the logging we should care about\n when we implement it.\n\n The logging should be configurable per database. Some\n databases have logging enabled while others are unprotected.\n\n It must be able to do point in time recovery (restore the\n database from a backup and recover until an absolute time or\n transaction ID).\n\n The previous two produce a problem for shared system\n relations. If a backend running on an unlogged database\n updates pg_user for example, this time it must go into the\n log!\n\n We should give query logging instead of image logging a try.\n\n\nUntil later, 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", "msg_date": "Thu, 19 Feb 1998 13:09:02 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: DB logging (was: Problem with the numbers I\n\treported yesterday)" } ]
[ { "msg_contents": "subscribe\n\n", "msg_date": "Wed, 11 Feb 1998 20:09:50 +0900", "msg_from": "Yu HyungSic <[email protected]>", "msg_from_op": true, "msg_subject": "(no subject)" } ]
[ { "msg_contents": "I wrote a message about 4 months ago asking about an Alpha Linux version\nof PostGreSQL. Last I heard, they were working on it and it would be\nreleased in early January with a 64-bit clean version of the server.\n\nThis is my third request.\n\nIt's now almost the middle of February, and I still don't see a fix for\nthe problem I saw 4 months ago. Here's the problem in full detail:\n\n[postgres@nimue pgsql]$ initdb\ninitdb: using /usr/local/pgsql/lib/local1_template1.bki.source as input to\ncreate the template database.\ninitdb: using /usr/local/pgsql/lib/global1.bki.source as input to create\nthe global classes.\ninitdb: using /usr/local/pgsql/lib/pg_hba.conf.sample as the host-based\nauthentication control file.\n\nWe are initializing the database system with username postgres (uid=510).\nThis user will own all the files and must also own the server process.\n\nCreating Postgres database system directory /usr/local/pgsql/data\n\nCreating Postgres database system directory /usr/local/pgsql/data/base\n\ninitdb: creating template database in /usr/local/pgsql/data/base/template1\nRunning: postgres -boot -C -F -D/usr/local/pgsql/data -Q template1\nERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\nERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\ninitdb: could not create template database\ninitdb: cleaning up by wiping out /usr/local/pgsql/data/base/template1\n\nAnd it dumps core. Any ideas on when this will be fixed? Anytime soon??\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Wed, 11 Feb 1998 11:57:13 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> As v6.2.1 is >4mos old, have you tried using the v6.3beta? \n\nNot quite sure where I can find 6.3beta ... if you're talking about the\nsnapshot, that's what I tried.\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Wed, 11 Feb 1998 13:35:47 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> \tAh, okay, then your subject was a little misleading :(\n\nOops - sorry about that - I should have put snapshot ;)\n\n> \tHave you tried running gdb against your core dump?\n\nNot yet - I haven't had the time ... but I do get a large number of\npointer-to-integer mismatches. I think it is in the lookup function for\nthe called SQL routine, meaning, when it tries to look-up a function from\nthe internal table, it either runs out of memory, or out of bounds. I can\nlook into this a little later tonight and give a little more definite bug\nreport.\n\nBut, the \"MemSet\" function returns \"cast from pointer to integer of\ndifferent size\" and there's places in the code that specify uint32 when\nuint32 on the Alpha is a different size. :/\n\nSo, if I can get a version of PostgreSQL working, I'll post a patch for\nAlpha, but I can't promise anything. I'll keep everyone informed in what\nI find.\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Wed, 11 Feb 1998 13:57:42 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL Snapshot (and 6.2.x) for Linux Alpha" }, { "msg_contents": "\n\nAs v6.2.1 is >4mos old, have you tried using the v6.3beta? \n\n\nOn Wed, 11 Feb 1998, Kenji T. Hollis wrote:\n\n> I wrote a message about 4 months ago asking about an Alpha Linux version\n> of PostGreSQL. Last I heard, they were working on it and it would be\n> released in early January with a 64-bit clean version of the server.\n> \n> This is my third request.\n> \n> It's now almost the middle of February, and I still don't see a fix for\n> the problem I saw 4 months ago. Here's the problem in full detail:\n> \n> [postgres@nimue pgsql]$ initdb\n> initdb: using /usr/local/pgsql/lib/local1_template1.bki.source as input to\n> create the template database.\n> initdb: using /usr/local/pgsql/lib/global1.bki.source as input to create\n> the global classes.\n> initdb: using /usr/local/pgsql/lib/pg_hba.conf.sample as the host-based\n> authentication control file.\n> \n> We are initializing the database system with username postgres (uid=510).\n> This user will own all the files and must also own the server process.\n> \n> Creating Postgres database system directory /usr/local/pgsql/data\n> \n> Creating Postgres database system directory /usr/local/pgsql/data/base\n> \n> initdb: creating template database in /usr/local/pgsql/data/base/template1\n> Running: postgres -boot -C -F -D/usr/local/pgsql/data -Q template1\n> ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n> ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n> initdb: could not create template database\n> initdb: cleaning up by wiping out /usr/local/pgsql/data/base/template1\n> \n> And it dumps core. Any ideas on when this will be fixed? Anytime soon??\n> \n> -- Ken\n> ------\n> =========================================================================\n> Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n> Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n> C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n> =========================================================================\n> Home page URL: http://www.houston-interweb.com/\n> \n> \n\n", "msg_date": "Wed, 11 Feb 1998 13:10:41 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "On Wed, 11 Feb 1998, Kenji T. Hollis wrote:\n\n> > As v6.2.1 is >4mos old, have you tried using the v6.3beta? \n> \n> Not quite sure where I can find 6.3beta ... if you're talking about the\n> snapshot, that's what I tried.\n\n\tAh, okay, then your subject was a little misleading :(\n\n\tHave you tried running gdb against your core dump?\n\n\n", "msg_date": "Wed, 11 Feb 1998 14:50:59 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" } ]
[ { "msg_contents": "I execute PQexec(simple_connection, copiedquery); with copiedquery being\n\"declare cur cursor for select name , born from meskes \" followed by another\ncall to PQexec with the command \"fetch in cur \". I get an OK status back,\nbut no tuples. However, just using \"select name , born from meskes;\" in psql\ngives me a correct list with five rows. \n\nTwo days ago when I tried this the last time it worked without a problem.\nThat is to say, I'm using the current snapshot I got from cvs this morning.\n\nShould a problem with the development version be reported here, or on the\nbugs list?\n\nMichael\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Wed, 11 Feb 1998 13:03:48 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "cursor access doesn't work anmore" }, { "msg_contents": "On Wed, 11 Feb 1998, Michael Meskes wrote:\n\n> I execute PQexec(simple_connection, copiedquery); with copiedquery being\n> \"declare cur cursor for select name , born from meskes \" followed by another\n> call to PQexec with the command \"fetch in cur \". I get an OK status back,\n> but no tuples. However, just using \"select name , born from meskes;\" in psql\n> gives me a correct list with five rows. \n> \n> Two days ago when I tried this the last time it worked without a problem.\n> That is to say, I'm using the current snapshot I got from cvs this morning.\n> \n> Should a problem with the development version be reported here, or on the\n> bugs list?\n\n\tHere works...is anyone even using bugs?\n\n\n", "msg_date": "Wed, 11 Feb 1998 09:11:54 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] cursor access doesn't work anmore" }, { "msg_contents": "> \n> I execute PQexec(simple_connection, copiedquery); with copiedquery being\n> \"declare cur cursor for select name , born from meskes \" followed by another\n> call to PQexec with the command \"fetch in cur \". I get an OK status back,\n> but no tuples. However, just using \"select name , born from meskes;\" in psql\n> gives me a correct list with five rows. \n> \n> Two days ago when I tried this the last time it worked without a problem.\n> That is to say, I'm using the current snapshot I got from cvs this morning.\n> \n> Should a problem with the development version be reported here, or on the\n> bugs list?\n> \n\nHave you done initdb. We changed things recently.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 10:33:48 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] cursor access doesn't work anmore" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> Have you done initdb. We changed things recently.\n\nShouldn't PG_VERSION be changed every time something changes in the\ndatabase file formats?\n\nIt is now set to 6.3, but should probably be 6.3.<something>\n\nAs I understand it, this is the place where the backend can look and\ndecide if it can access the files or not.\n\n/* m */\n", "msg_date": "Fri, 13 Feb 1998 14:00:00 +0100", "msg_from": "Mattias Kregert <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] cursor access doesn't work anmore" }, { "msg_contents": "On Fri, 13 Feb 1998, Mattias Kregert wrote:\n\n> Shouldn't PG_VERSION be changed every time something changes in the\n> database file formats?\n> \n> It is now set to 6.3, but should probably be 6.3.<something>\n> \n> As I understand it, this is the place where the backend can look and\n> decide if it can access the files or not.\n\nI would have thought so, yes. However, not until 6.3 is released. Until\nthen, the files can change often.\n\nRemember, were in beta. If we changed it before release, then we would get\nquestions like \"Why does PG_VERSION show 6.3.9 when I'm running 6.3?\"\n\nI think the only time this should change is between public\nreleases/subreleases.\n\n-- \nPeter T Mount [email protected] or [email protected]\nMain Homepage: http://www.demon.co.uk/finder\nWork Homepage: http://www.maidstone.gov.uk Work EMail: [email protected]\n\n", "msg_date": "Sat, 14 Feb 1998 09:56:36 +0000 (GMT)", "msg_from": "Peter T Mount <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] cursor access doesn't work anmore" } ]
[ { "msg_contents": "hi,\n\nI am going to do my first hack:\n\n\tusing indexes in aggregates (whenever possible)\n\nIf someone claimed it already please let me know.\nTIA\n\n.co.\n\nCostin Oproiu ---------------- http://www2.deuroconsult.ro/~co\nCiprian Porumbescu 10, ap.13 tel/fax: +(40)-68-183528\nBrasov 2200, ROMANIA email: [email protected]\n--------------------------------------------------------------\n\n", "msg_date": "Wed, 11 Feb 1998 14:53:52 +0200 (EET)", "msg_from": "Costin Oproiu <[email protected]>", "msg_from_op": true, "msg_subject": "Working on some performance optimization" }, { "msg_contents": "> \n> hi,\n> \n> I am going to do my first hack:\n> \n> \tusing indexes in aggregates (whenever possible)\n> \n> If someone claimed it already please let me know.\n> TIA\n> \n\nAdded to TODO.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 10:39:37 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Working on some performance optimization" } ]
[ { "msg_contents": "\tIn working with testing the ODBC driver with 6.3 I've noticed\nsome odd behavior. I establish a connection and get authenticated.\nThen I send a query. This creates yet another postgres process... Each\nquery does this... And they all seem to hang around... And take up CPU\nand Memory. And quckly system performance goes way out the window...\n\njulie\n\n-- \n[ Julia Anne Case ] [ Ships are safe inside the harbor, ]\n[Programmer at large] [ but is that what ships are really for. ] \n[ Admining Linux ] [ To thine own self be true. ]\n[ Windows/WindowsNT ] [ Fair is where you take your cows to be judged. ]\n", "msg_date": "Wed, 11 Feb 1998 07:59:47 -0500", "msg_from": "Julia Anne Case <[email protected]>", "msg_from_op": true, "msg_subject": "Frontend woes..." } ]
[ { "msg_contents": "Brian:\n\n> > I wrote a message about 4 months ago asking about an Alpha Linux version\n> > of PostGreSQL. Last I heard, they were working on it and it would be\n> > released in early January with a 64-bit clean version of the server.\n> > \n> > This is my third request.\n> \n> So, fix it your own self and stop whining. Can't fix it? Too bad, I\n> guess you'll just have to wait until they get around to it.\n\nIs this how you guys normally handle inquiries? Idiot replies like this?\nOh by the way, I tried fixing it. Next question?\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Wed, 11 Feb 1998 13:34:57 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> \tNo, this isn't...generally we tend to try and point you in the\n> right direction towards determining the problem...\n\nInstead of bitching out at the person who found the problem. I see.\nMaybe I should have said \"pretty please\". ;)\n\n> \t...first question is, how many ppl out there are running under a\n> Linux/Alpha platform, or at least are trying? I don't have one myself, so\n> can only be of limited help...\n\nI'm not really sure - I've had this Alpha here for about 6 months and I\nhave resulted to using msql since that's about the only thing that will\nwork efficiently.\n\n> \t...next, do you actually get a core file that you can analyze\n> using gdb? If so, what do the results show?\n\nI get a core file, but it doesn't tell me squat,\n\n> \tWe can't debug this for you (well, I imagine Thomas would love it\n> if you sent him a Linux/Alpha machine, then he could help you grin), but\n> if you are willing, we can try and steer you towards a solution...\n\nI've actually zeroed in on the problem. It lies somewhere in the\nSearchSysCache routine. I'm attempting to debug it now... So far, what I\nget is: (Note \"[KTH]\" is a debug message added by yours truly)\n\n--- SearchSysCache starts here ---\nSearchSysCache: [KTH] Hash: 433\nSearchSysCache: [KTH] Tuple not found in cache, attempting to find.\nSearchSysCache: [KTH] RelationGetRelationName (pg_proc)\nSearchSysCache: performing scan (override==0)\nSearchSysCache: [KTH] IsBootstrapProcessingMode() is true\nSearchSysCache: [KTH] relation check skipped.\nSearchSysCache: [KTH] heap_beginscan is okay.\nheap_getnext([pg_proc,nkeys=3],backw=0,0x1ffff040) called\nheap_getnext returning EOS\nSearchSysCache: [KTH] heap_getnext returns null\nSearchSysCache: [KTH] tuple not found.\nSearchSysCache: [KTH] Heap scan ends.\nSearchSysCache: Heap tuple (ntp) is invalid.\nERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\nERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n--- End of debug ---\n\nLooks like it lies somewhere in heap_getnext. Heap_getnext is a HUMONGOUS\ncommand, and I'm not about to spend another 2 hours debugging that. ;)\n\nAnyone have any suggestions of a patch for this?\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Wed, 11 Feb 1998 15:57:30 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "On Wed, 11 Feb 1998, Kenji T. Hollis wrote:\n\n> Brian:\n> \n> > > I wrote a message about 4 months ago asking about an Alpha Linux version\n> > > of PostGreSQL. Last I heard, they were working on it and it would be\n> > > released in early January with a 64-bit clean version of the server.\n> > > \n> > > This is my third request.\n> > \n> > So, fix it your own self and stop whining. Can't fix it? Too bad, I\n> > guess you'll just have to wait until they get around to it.\n> \n> Is this how you guys normally handle inquiries? Idiot replies like this?\n> Oh by the way, I tried fixing it. Next question?\n\n\tNo, this isn't...generally we tend to try and point you in the\nright direction towards determining the problem...\n\n\t...first question is, how many ppl out there are running under a\nLinux/Alpha platform, or at least are trying? I don't have one myself, so\ncan only be of limited help...\n\n\t...next, do you actually get a core file that you can analyze\nusing gdb? If so, what do the results show?\n\n\tWe can't debug this for you (well, I imagine Thomas would love it\nif you sent him a Linux/Alpha machine, then he could help you grin), but\nif you are willing, we can try and steer you towards a solution...\n\n\n\n", "msg_date": "Wed, 11 Feb 1998 15:30:04 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "Bruce:\n\n> \tbackend/catalog/index.c:298: func_error(\"BuildFuncTupleDesc\",\n> \tfuncname, nargs, argtypes);\n> \n> The function he is having trouble with is one that gets created by\n> initdb for use in an index. Must be failing there somehow, but without\n> initdb completing, you can't easily debug to see what is in the pg_proc\n> table.\n\nThat's the line that's the problem. It seems to find the other routines\nthat it needs from the hash table, but this seems to be the culprit in\nboth v6.2.1, and the 6.3beta that I'm trying.\n\nI spent a good day working on finding the problem, and found that this was\nwhere it lied. Further study showed it was in hash_getnext, but I didn't\nhave time to debug hash_getnext.\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Thu, 12 Feb 1998 01:05:07 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "Bruce:\n\n> The problem here is that it can't find the function to make/use the\n> index. Try using initdb --debug to get more output, and see what is\n> says about the mkoidname function creation. Looks like pg_proc is not\n> working, because a scan is returning nothing. mkoidname is function\n> used to index pg_attribute. If you do initdb with --noclean, is\n> data/template1/pg_proc indeed zero bytes. Try adding a define to the\n> postgres.h\n>\n> #define long int\n> \n> and see if it works. Maybe the 64-bit longs are causing problems, and\n> we have to fix them or change to ints.\n\nOkay, I can try this. In the current version of Postgres, when I run\n\"initdb --debug\", I get the following output:\n\n> CREATED relation pg_description with OID 17847\n> Commit End\nAmopen: relation pg_description. attrsize 63\ncreate attribute 0 name objoid len 4 num 1 type 26\ncreate attribute 1 name description len -1 num 2 type 25\n> Amclose: relation (null).\n> initdb: could not create template database\ninitdb: cleaning up by wiping out /usr/local/pgsql/data/base/template1\n\nInstalling the \"#define long int\" gives about 40 pages of errors.\n\n> Make sure you turn on Assert checking in configure so it may give you an\n> earlier error.\n\nI have no idea of how to do set this in configure. Configure has no\noption to do this.\n\n> These are very hard to debug because there is no running system to run\n> tests on, and it is all very inter-related.\n\nI can give access to my Alpha, but I have to talk to the person I am\ndealing with. I will not give access to this machine with a complete\nstranger via E-Mail.\n\n> My lib/local1_template1.bki.source has the following two lines for this\n> function:\n> \n> insert OID = 949 ( mkoidname PGUID 11 f t f 2 f 911 \"26 19\" 100 0 0 100 foo bar)\n> declare index pg_attribute_relid_attnam_index on pg_attribute using btree(mkoidname(attrelid, attname) oidname_ops)\n\nExact carbon-copy of mine.\n\n> That is all the ideas I have for now. Would like to get it working.\n\nSo would I. Bruce, If you would like access to the Postgres machine,\nplease feel free to give me a call tomorrow. \n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Thu, 12 Feb 1998 01:15:42 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "On Wed, 11 Feb 1998, Kenji T. Hollis wrote:\n\n> > \tNo, this isn't...generally we tend to try and point you in the\n> > right direction towards determining the problem...\n> \n> Instead of bitching out at the person who found the problem. I see.\n> Maybe I should have said \"pretty please\". ;)\n\n\tNope, that wouldn't have helped either...its difficult for us to\nattempt to debug a problem with a platform that we don't have access to,\nso we tend to rely on the person reporting the problem to provide us as\nmuch detail as possible concerning the problem. At times, you have to\nnudge the developers a little more then others, but its a volunteer\ndevelopment environment, and everyone has their own priorities...:(\n\n> > \t...next, do you actually get a core file that you can analyze\n> > using gdb? If so, what do the results show?\n> \n> I get a core file, but it doesn't tell me squat,\n\n\tIt can generally tell you alot...do you have gdb on your system?\nhave you compiled with -g (debugging symbols) compiled in? Using gdb and\nthe core file, you should be able to get pretty close to the exact line\n(and values) where the error is occuring...combine that with 'script', and\nyou can pretty much give us a screen capture of what you are doing...\n\n> I've actually zeroed in on the problem. It lies somewhere in the\n> SearchSysCache routine. I'm attempting to debug it now... So far, what I\n> get is: (Note \"[KTH]\" is a debug message added by yours truly)\n> \n> --- SearchSysCache starts here ---\n> SearchSysCache: [KTH] Hash: 433\n> SearchSysCache: [KTH] Tuple not found in cache, attempting to find.\n> SearchSysCache: [KTH] RelationGetRelationName (pg_proc)\n> SearchSysCache: performing scan (override==0)\n> SearchSysCache: [KTH] IsBootstrapProcessingMode() is true\n> SearchSysCache: [KTH] relation check skipped.\n> SearchSysCache: [KTH] heap_beginscan is okay.\n> heap_getnext([pg_proc,nkeys=3],backw=0,0x1ffff040) called\n> heap_getnext returning EOS\n> SearchSysCache: [KTH] heap_getnext returns null\n> SearchSysCache: [KTH] tuple not found.\n> SearchSysCache: [KTH] Heap scan ends.\n> SearchSysCache: Heap tuple (ntp) is invalid.\n> ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n> ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n> --- End of debug ---\n> \n> Looks like it lies somewhere in heap_getnext. Heap_getnext is a HUMONGOUS\n> command, and I'm not about to spend another 2 hours debugging that. ;)\n> \n> Anyone have any suggestions of a patch for this?\n\n\tBruce...I just checked backend/catalog/index.c, where\nBuildFuncTupleDesc() exists, and there is no error message that matches\nhis above ERROR...\n\n\tKen...look at the top of backend/catalog/index.c and tell me what\nversion it stays it is? On the $Header: line?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Wed, 11 Feb 1998 21:26:19 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> > > \t...next, do you actually get a core file that you can analyze\n> > > using gdb? If so, what do the results show?\n> > \n> > I get a core file, but it doesn't tell me squat,\n> \n> \tIt can generally tell you alot...do you have gdb on your system?\n> have you compiled with -g (debugging symbols) compiled in? Using gdb and\n> the core file, you should be able to get pretty close to the exact line\n> (and values) where the error is occuring...combine that with 'script', and\n> you can pretty much give us a screen capture of what you are doing...\n> \n> > I've actually zeroed in on the problem. It lies somewhere in the\n> > SearchSysCache routine. I'm attempting to debug it now... So far, what I\n> > get is: (Note \"[KTH]\" is a debug message added by yours truly)\n> > \n> > --- SearchSysCache starts here ---\n> > SearchSysCache: [KTH] Hash: 433\n> > SearchSysCache: [KTH] Tuple not found in cache, attempting to find.\n> > SearchSysCache: [KTH] RelationGetRelationName (pg_proc)\n> > SearchSysCache: performing scan (override==0)\n> > SearchSysCache: [KTH] IsBootstrapProcessingMode() is true\n> > SearchSysCache: [KTH] relation check skipped.\n> > SearchSysCache: [KTH] heap_beginscan is okay.\n> > heap_getnext([pg_proc,nkeys=3],backw=0,0x1ffff040) called\n> > heap_getnext returning EOS\n> > SearchSysCache: [KTH] heap_getnext returns null\n> > SearchSysCache: [KTH] tuple not found.\n> > SearchSysCache: [KTH] Heap scan ends.\n> > SearchSysCache: Heap tuple (ntp) is invalid.\n> > ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n> > ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n> > --- End of debug ---\n> > \n> > Looks like it lies somewhere in heap_getnext. Heap_getnext is a HUMONGOUS\n> > command, and I'm not about to spend another 2 hours debugging that. ;)\n> > \n> > Anyone have any suggestions of a patch for this?\n> \n> \tBruce...I just checked backend/catalog/index.c, where\n> BuildFuncTupleDesc() exists, and there is no error message that matches\n> his above ERROR...\n\nI think it is at:\n\n\tbackend/catalog/index.c:298: func_error(\"BuildFuncTupleDesc\",\n\tfuncname, nargs, argtypes);\n\nThe function he is having trouble with is one that gets created by\ninitdb for use in an index. Must be failing there somehow, but without\ninitdb completing, you can't easily debug to see what is in the pg_proc\ntable.\n\n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 20:37:42 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> --- SearchSysCache starts here ---\n> SearchSysCache: [KTH] Hash: 433\n> SearchSysCache: [KTH] Tuple not found in cache, attempting to find.\n> SearchSysCache: [KTH] RelationGetRelationName (pg_proc)\n> SearchSysCache: performing scan (override==0)\n> SearchSysCache: [KTH] IsBootstrapProcessingMode() is true\n> SearchSysCache: [KTH] relation check skipped.\n> SearchSysCache: [KTH] heap_beginscan is okay.\n> heap_getnext([pg_proc,nkeys=3],backw=0,0x1ffff040) called\n> heap_getnext returning EOS\n> SearchSysCache: [KTH] heap_getnext returns null\n> SearchSysCache: [KTH] tuple not found.\n> SearchSysCache: [KTH] Heap scan ends.\n> SearchSysCache: Heap tuple (ntp) is invalid.\n> ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n> ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n> --- End of debug ---\n\nThe problem here is that it can't find the function to make/use the\nindex. Try using initdb --debug to get more output, and see what is\nsays about the mkoidname function creation. Looks like pg_proc is not\nworking, because a scan is returning nothing. mkoidname is function\nused to index pg_attribute. If you do initdb with --noclean, is\ndata/template1/pg_proc indeed zero bytes. Try adding a define to the\npostgres.h\n\n#define long int\n\nand see if it works. Maybe the 64-bit longs are causing problems, and\nwe have to fix them or change to ints.\n\nMake sure you turn on Assert checking in configure so it may give you an\nearlier error.\n\nThese are very hard to debug because there is no running system to run\ntests on, and it is all very inter-related.\n\nMy lib/local1_template1.bki.source has the following two lines for this\nfunction:\n\ninsert OID = 949 ( mkoidname PGUID 11 f t f 2 f 911 \"26 19\" 100 0 0 100 foo bar)\ndeclare index pg_attribute_relid_attnam_index on pg_attribute using btree(mkoidname(attrelid, attname) oidname_ops)\n\nThat is all the ideas I have for now. Would like to get it working.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 23:36:24 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "Bruce:\n\n> Take a look at utils/hash/hashfn.c:tag_hash. Is there a problem in that\n> code for your platform. Is the hash getting set, or is it falling\n> through the case statements? This code is clearly broken for\n> sizeof(int) > 4, but I think your ints are 4, and longs are 8. I bet\n> somewhere we are using a long where we should be using an int, and that\n> is why only your platform is seeing it. Is this true about long vs.\n> int. I can review our use of longs to see if there are problems.\n\nI created a small program to return the size of values. They are:\n\nSize of short: 2\nSize of char: 1\nSize of int: 4\nSize of long: 8\n\nDoes this help?\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Thu, 12 Feb 1998 10:06:38 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "Bruce:\n\n> OK, I have a new idea. See in utils/hash/hashfn.c:tag_hash, there is\n> the line:\n> \n> for (; keysize > (sizeof(int) - 1); keysize -= sizeof(int),key++)\n> h = h * PRIME1 ^ (key);\n> \n> Now, since h is a long, shouldn't the for loop be comparing\n> sizeof(long)? However, key is an int. \n\nHow is this a problem? key is getting the value of the current pointer\nof key. This means, if key contains a string: \"Ooga\" and key++, then the\nvalue of key would be \"o\" in decimal. This is a standard hashing\nroutine, and the problem does not lie here as far as I can see.\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Thu, 12 Feb 1998 10:09:30 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "Bruce:\n\n> Use this:\n> \n> \tconfigure --enable-cassert\n\nDone, installed, and no difference in output in any way, shape, or form.\n\n> I am not sure how I would debug this even if I had access. I guess I\n> would try the assert, and then start reviewing the long vs. int issues.\n\nI would probably want a PostGreSQL hacker working on this if that's the\ncase, then. ;)\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Thu, 12 Feb 1998 10:10:15 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "On Thu, 12 Feb 1998, Kenji T. Hollis wrote:\n\n> > Make sure you turn on Assert checking in configure so it may give you an\n> > earlier error.\n> \n> I have no idea of how to do set this in configure. Configure has no\n> option to do this.\n\n\t--enable-cassert\n\n\n", "msg_date": "Thu, 12 Feb 1998 06:48:03 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> \n> Bruce:\n> \n> > \tbackend/catalog/index.c:298: func_error(\"BuildFuncTupleDesc\",\n> > \tfuncname, nargs, argtypes);\n> > \n> > The function he is having trouble with is one that gets created by\n> > initdb for use in an index. Must be failing there somehow, but without\n> > initdb completing, you can't easily debug to see what is in the pg_proc\n> > table.\n> \n> That's the line that's the problem. It seems to find the other routines\n> that it needs from the hash table, but this seems to be the culprit in\n> both v6.2.1, and the 6.3beta that I'm trying.\n> \n> I spent a good day working on finding the problem, and found that this was\n> where it lied. Further study showed it was in hash_getnext, but I didn't\n> have time to debug hash_getnext.\n\nAgain, I will say that the problems with initdb are usually very\ncomplicated to debug. It seems like you have gotten pretty far. For\nme, it is just a challenge to get initdb running inside a debugger\nbecause there is so much shell script startup before the postgres\nprocess runs.\n\nTake a look at utils/hash/hashfn.c:tag_hash. Is there a problem in that\ncode for your platform. Is the hash getting set, or is it falling\nthrough the case statements? This code is clearly broken for\nsizeof(int) > 4, but I think your ints are 4, and longs are 8. I bet\nsomewhere we are using a long where we should be using an int, and that\nis why only your platform is seeing it. Is this true about long vs.\nint. I can review our use of longs to see if there are problems.\n\n\n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 09:21:06 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> > CREATED relation pg_description with OID 17847\n> > Commit End\n> Amopen: relation pg_description. attrsize 63\n> create attribute 0 name objoid len 4 num 1 type 26\n> create attribute 1 name description len -1 num 2 type 25\n> > Amclose: relation (null).\n> > initdb: could not create template database\n> initdb: cleaning up by wiping out /usr/local/pgsql/data/base/template1\n> \n> Installing the \"#define long int\" gives about 40 pages of errors.\n> \n> > Make sure you turn on Assert checking in configure so it may give you an\n> > earlier error.\n> \n> I have no idea of how to do set this in configure. Configure has no\n> option to do this.\n\nUse this:\n\n\tconfigure --enable-cassert\n\n> \n> > These are very hard to debug because there is no running system to run\n> > tests on, and it is all very inter-related.\n> \n> I can give access to my Alpha, but I have to talk to the person I am\n> dealing with. I will not give access to this machine with a complete\n> stranger via E-Mail.\n\n610-853-3000. That is me.\n\n> \n> > My lib/local1_template1.bki.source has the following two lines for this\n> > function:\n> > \n> > insert OID = 949 ( mkoidname PGUID 11 f t f 2 f 911 \"26 19\" 100 0 0 100 foo bar)\n> > declare index pg_attribute_relid_attnam_index on pg_attribute using btree(mkoidname(attrelid, attname) oidname_ops)\n> \n> Exact carbon-copy of mine.\n\nGood.\n\n> \n> > That is all the ideas I have for now. Would like to get it working.\n> \n> So would I. Bruce, If you would like access to the Postgres machine,\n> please feel free to give me a call tomorrow. \n\nI am not sure how I would debug this even if I had access. I guess I\nwould try the assert, and then start reviewing the long vs. int issues.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 09:24:08 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> Take a look at utils/hash/hashfn.c:tag_hash. Is there a problem in that\n> code for your platform. Is the hash getting set, or is it falling\n> through the case statements? This code is clearly broken for\n> sizeof(int) > 4, but I think your ints are 4, and longs are 8. I bet\n> somewhere we are using a long where we should be using an int, and that\n> is why only your platform is seeing it. Is this true about long vs.\n> int. I can review our use of longs to see if there are problems.\n\nOK, I have a new idea. See in utils/hash/hashfn.c:tag_hash, there is\nthe line:\n\n for (; keysize > (sizeof(int) - 1); keysize -= sizeof(int),key++)\n h = h * PRIME1 ^ (key);\n\nNow, since h is a long, shouldn't the for loop be comparing\nsizeof(long)? However, key is an int. \n\nComments? \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 09:46:01 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> \n> > Take a look at utils/hash/hashfn.c:tag_hash. Is there a problem in that\n> > code for your platform. Is the hash getting set, or is it falling\n> > through the case statements? This code is clearly broken for\n> > sizeof(int) > 4, but I think your ints are 4, and longs are 8. I bet\n> > somewhere we are using a long where we should be using an int, and that\n> > is why only your platform is seeing it. Is this true about long vs.\n> > int. I can review our use of longs to see if there are problems.\n> \n> OK, I have a new idea. See in utils/hash/hashfn.c:tag_hash, there is\n> the line:\n> \n> for (; keysize > (sizeof(int) - 1); keysize -= sizeof(int),key++)\n> h = h * PRIME1 ^ (key);\n> \n> Now, since h is a long, shouldn't the for loop be comparing\n> sizeof(long)? However, key is an int. \n> \n> Comments? \n> \n\nNever mind. I think the code is good, but if someone sees a problem\nwith longs/ints anywhere in the code, please let's discuss it.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 09:51:52 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "Bruce Momjian wrote:\n\n> >\n> > Bruce:\n> >\n> > > backend/catalog/index.c:298: func_error(\"BuildFuncTupleDesc\",\n> > > funcname, nargs, argtypes);\n> > >\n> > > The function he is having trouble with is one that gets created by\n> > > initdb for use in an index. Must be failing there somehow, but without\n> > > initdb completing, you can't easily debug to see what is in the pg_proc\n> > > table.\n> >\n> > That's the line that's the problem. It seems to find the other routines\n> > that it needs from the hash table, but this seems to be the culprit in\n> > both v6.2.1, and the 6.3beta that I'm trying.\n> >\n> > I spent a good day working on finding the problem, and found that this was\n> > where it lied. Further study showed it was in hash_getnext, but I didn't\n> > have time to debug hash_getnext.\n>\n> Again, I will say that the problems with initdb are usually very\n> complicated to debug. It seems like you have gotten pretty far. For\n> me, it is just a challenge to get initdb running inside a debugger\n> because there is so much shell script startup before the postgres\n> process runs.\n>\n> Take a look at utils/hash/hashfn.c:tag_hash. Is there a problem in that\n> code for your platform. Is the hash getting set, or is it falling\n> through the case statements? This code is clearly broken for\n> sizeof(int) > 4, but I think your ints are 4, and longs are 8. I bet\n> somewhere we are using a long where we should be using an int, and that\n> is why only your platform is seeing it. Is this true about long vs.\n> int. I can review our use of longs to see if there are problems.\n\nI recall seeing the \"Datum\" type in the code as an int/pointer coersed type. That\nis, there are lots of places where it is forced from an int to a pointer and back\nagain. Might look at that, and make sure it is a long/pointer rather than an\nint/pointer. Of course, that may have already been fixed up...\n\n - Tom\n\n", "msg_date": "Thu, 12 Feb 1998 15:02:26 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> \n> Bruce:\n> \n> > Take a look at utils/hash/hashfn.c:tag_hash. Is there a problem in that\n> > code for your platform. Is the hash getting set, or is it falling\n> > through the case statements? This code is clearly broken for\n> > sizeof(int) > 4, but I think your ints are 4, and longs are 8. I bet\n> > somewhere we are using a long where we should be using an int, and that\n> > is why only your platform is seeing it. Is this true about long vs.\n> > int. I can review our use of longs to see if there are problems.\n> \n> I created a small program to return the size of values. They are:\n> \n> Size of short: 2\n> Size of char: 1\n> Size of int: 4\n> Size of long: 8\n> \n> Does this help?\n\nYep. Does anyone else run OS's with the size of long greater than int? \nDoes Dec Unix Alpha? Are people running that?\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 11:50:51 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> \n> Bruce:\n> \n> > OK, I have a new idea. See in utils/hash/hashfn.c:tag_hash, there is\n> > the line:\n> > \n> > for (; keysize > (sizeof(int) - 1); keysize -= sizeof(int),key++)\n> > h = h * PRIME1 ^ (key);\n> > \n> > Now, since h is a long, shouldn't the for loop be comparing\n> > sizeof(long)? However, key is an int. \n> \n> How is this a problem? key is getting the value of the current pointer\n> of key. This means, if key contains a string: \"Ooga\" and key++, then the\n> value of key would be \"o\" in decimal. This is a standard hashing\n> routine, and the problem does not lie here as far as I can see.\n\nYou are right.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 11:51:21 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> \n> Bruce:\n> \n> > Use this:\n> > \n> > \tconfigure --enable-cassert\n> \n> Done, installed, and no difference in output in any way, shape, or form.\n> \n> > I am not sure how I would debug this even if I had access. I guess I\n> > would try the assert, and then start reviewing the long vs. int issues.\n> \n> I would probably want a PostGreSQL hacker working on this if that's the\n> case, then. ;)\n\nOK, it is true if I am not sure how to debug it, it is tough to see how\nyou will. I will check on the long/int issue in the source. Can you\ntry to see if the hash keys set/lookup match for mkoidname?\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 11:53:13 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "On Thu, 12 Feb 1998, Kenji T. Hollis wrote:\n\n> > I am not sure how I would debug this even if I had access. I guess I\n> > would try the assert, and then start reviewing the long vs. int issues.\n> \n> I would probably want a PostGreSQL hacker working on this if that's the\n> case, then. ;)\n\n\tYou might want to be slightly careful here rofl There are three\nppl that I consider to be the ones that \"know it all\", and I don't think\nany of them have the same opinion of themselves:\n\t\n\tBruce, Vadim and Thomas\n\n\tWe've just been blessed with a fourth that is working on\nobseleting Oracle for himself by bringing into PostgreSQL what he\nrequires.\n\n\tAnd then there are about a dozen (or more) of us that have\nknowledge of various aspects...\n\n\tI think the question that should be asked is if anyone out there\nis currently working on/with PostgreSQL and a 64bit architecture, namely\nthe Alpha, since the problem appears to be 64bit related, which, I don't \nbelieve, any of the core currently uses, or has much knowledge of.\n\n\tAs has been shown over the past few hours, even, there is alot of\nknowledge and ideas that you can tap...but each of us works hard at\nmaintaining our individual platforms (Vadim and I use FreeBSD, Thomas uses\nLinux, Bruce - BSDi, Darren - Aix, Andrew - Irix5/Linux, etc)...\n\n\tActually, Andrew...what is the Irix5? 64bit or 32bit still?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 12 Feb 1998 12:59:19 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "Good grief! I don't check my email for a few days a fire fight\nover getting pgsql on Linux/Alpha ensues! :)\n\tJeff Strum has been doing some extensive work (quietly) on getting\nLinux/Alpha to run. He has gotten initdb to finish successfully and the\nregression tests are partially successful. I haven't had time to test\nthings for myself, but things look promsing. Apparently there was a\ncompiler define that should have been enabled, but was not due to\ndifferent naming conventions for Linux/Alpha. Attached is the patch he\nsent me, that I have not had a chance to test yet.\n\tFor the future, he plans to track down and clean up the warnings\nabout casting from different pointer sizes and such. I plan to help\n(mainly testing) as much as I can, but I am quite busy, and I am pretty\nsure he is too.\n\tAs for you, Kenji, (not to be a jerk, but instead a reality\ncheck) just calm down and be patient, or jump in and\nhelp us (i.e. email Jeff and ask him what you can do to help, or get a\nnew version of pgsql, apply the patch included in this email, and report\nback to us what happens). Nothing gets done by people continously asking\nfor something to be done. And anyway, porting to the alpha, a 64bit\narchitecture is not simple to say the very least. Add to that, a very\nlarge (I think 100k+ lines of code has been stated before) and powerful\ndatabase to port, and things reach the near impossible point. We will get\nit figured out in due time, but we don't have 50 programmers working away\nat this all day long. Then again, you are not paying anything for the\ndatabase. So I think things are fair. :)\n\tI will try and keep everyone updated a bit more regularly to head\noff fire fights like these. Thanks, and talk to you all later.\n\nPS. Kenji, where did you get that pgsql would be running on Linux/Alpha in\nJan '98? I have never seen any such projected date!\n\n----------------------------------------------------------------------------\n\| \"For to me to live is Christ, and to die is gain.\" \|\n\| --- Philippians 1:21 (KJV) \|\n----------------------------------------------------------------------------\n\| Ryan Kirkpatrick \| Boulder, Colorado \| [email protected] \|\n----------------------------------------------------------------------------\n\| http://www-ugrad.cs.colorado.edu/~rkirkpat/ \|\n----------------------------------------------------------------------------", "msg_date": "Fri, 13 Feb 1998 19:27:25 -0600 (CST)", "msg_from": "Ryan Kirkpatrick <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "Ryan:\n\n> \tAs for you, Kenji, (not to be a jerk, but instead a reality\n> check) just calm down and be patient, or jump in and\n> help us (i.e. email Jeff and ask him what you can do to help, or get a\n> new version of pgsql, apply the patch included in this email, and report\n> back to us what happens). Nothing gets done by people continously asking\n> for something to be done. And anyway, porting to the alpha, a 64bit\n> architecture is not simple to say the very least. Add to that, a very\n> large (I think 100k+ lines of code has been stated before) and powerful\n> database to port, and things reach the near impossible point. We will get\n> it figured out in due time, but we don't have 50 programmers working away\n> at this all day long. Then again, you are not paying anything for the\n> database. So I think things are fair. :)\n\nI totally understand where you're coming from, however, I wrote over 4\nmonths ago reporting the problem, and was kind of brushed off like it was\na non-issue, and I reported it twice. Same thing twice. Finally, I raise\nhell about it and I get a response. I'm not impatient. I am currently\nusing msql for what I need and it seems to be getting the job done. I\njust need somethimg much more powerful. PostGres is much more powerful.\n\nI'm trying to help, but the first message I get was from some idiot on\nthis list who told me to fix the problem myself and quit whining. I don't\ntake kindly to people who write messages like those to me, especially when\nI'm trying to help resolve an issue with the coders. I don't deserve to\nbe treated with this disrespect, and I never shoot back at any question I\nget - the most I'll do is delete the message and bite my damn tongue.\n\nSo, now that you understand my position, I just wanted an update, and\nwanted to help fix the problem.\n\n> PS. Kenji, where did you get that pgsql would be running on Linux/Alpha in\n> Jan '98? I have never seen any such projected date!\n\nOne of the coders told me this privately. I wish I had saved that\nmessage!\n\n-- Ken\n------\n =========================================================================\n Houston InterWeb Design, Inc. \|\| Office: +1 (713) 627-9494\n Lead Programmer/Designer \|\| Fax: +1 (713) 627-2744\n C++, Windows, and Web Programming \|\| Pager: +1 (713) 727-2529\n =========================================================================\n Home page URL: http://www.houston-interweb.com/\n\n", "msg_date": "Sun, 15 Feb 1998 02:15:59 +0000 ()", "msg_from": "\"Kenji T. Hollis\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "> I totally understand where you're coming from, however, I wrote over 4\n> months ago reporting the problem, and was kind of brushed off like it was\n> a non-issue, and I reported it twice. Same thing twice. Finally, I raise\n> hell about it and I get a response. I'm not impatient. I am currently\n> using msql for what I need and it seems to be getting the job done. I\n> just need somethimg much more powerful. PostGres is much more powerful.\n> \n> I'm trying to help, but the first message I get was from some idiot on\n> this list who told me to fix the problem myself and quit whining. I don't\n> take kindly to people who write messages like those to me, especially when\n> I'm trying to help resolve an issue with the coders. I don't deserve to\n> be treated with this disrespect, and I never shoot back at any question I\n> get - the most I'll do is delete the message and bite my damn tongue.\n> \n> So, now that you understand my position, I just wanted an update, and\n> wanted to help fix the problem.\n> \n> > PS. Kenji, where did you get that pgsql would be running on Linux/Alpha in\n> > Jan '98? I have never seen any such projected date!\n> \n> One of the coders told me this privately. I wish I had saved that\n> message!\n\nAsk and you shall receive. Here is the message. At the time of 6.2,\nAlpha/Linux was broken because the inlinine of asm() would not work\nbecause alpha/linux does not support local labels in asm statements, so\nif multiple lock requests were in the same C file, you had multiple\nlabels with the same name, and it would not compile.\n\nThis was worked on and a new fix was made, but that did not work because\nof the labels problem, so the attached fix was also broken. The final\nsolution was to move the alpha/linux lock code to a separate C file, so\nthere was every feeling it would now work in 6.3.\n\nTurns out there were other problems related to compiler flags, and that\nhas only recently been added to the source. Good thing Kenji spoke up\nso we could get all the alpha/linux people together on this.\n\n---------------------------------------------------------------------------\n", "msg_date": "Sun, 15 Feb 1998 12:47:57 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" }, { "msg_contents": "On Sun, 15 Feb 1998, Kenji T. Hollis wrote:\n\n> Ryan:\n\n> I totally understand where you're coming from, however, I wrote over 4\n> months ago reporting the problem, and was kind of brushed off like it was\n> a non-issue, and I reported it twice. Same thing twice. Finally, I raise\n> hell about it and I get a response. I'm not impatient. I am currently\n> using msql for what I need and it seems to be getting the job done. I\n> just need somethimg much more powerful. PostGres is much more powerful.\n\n\tI remember the two - three times you have posted to the list\nrequesting information about running pgsql on Linux/Alpha. Each time I\nresponded (politely) that it was in the works, but it was slow going due\nto lack of pgsql hackers with Alphas and Alpha experience. If you had\nwanted more information then, or felt you were brushed off, why did you\nrarely ever respond to my responses with more than a few words of\nacknowledgement? Not to pick bones here or anything, but I am just trying\nto put the record straight and defend those who have been trying to get\nthe alpha port working.\n\tAlso, I am also currently using msql on my UDB until pgsql gets\nfinished being ported. That is one of the initial reasons I got myself\ninvolved in this, I wanted a real (free) RDBS database for my UDB. So,\nsince the pgsql hackers appeared to be \"Alpha\"less I jumped in and tried\nto help as best I could. \n\n> I'm trying to help, but the first message I get was from some idiot on\n> this list who told me to fix the problem myself and quit whining. I don't\n> take kindly to people who write messages like those to me, especially when\n> I'm trying to help resolve an issue with the coders. I don't deserve to\n> be treated with this disrespect, and I never shoot back at any question I\n> get - the most I'll do is delete the message and bite my damn tongue.\n\n\tFirst, let me say in the \"Open Source Software\" community (yes,\nnotice the new name for free software), there is only a gray fuzzy line\nbetween the coders and the users. Here we don't ask coders to do something\nfor us. Instead we ask if they have any plans/ideas on how to do something\nwe want. If they don't, then it is our responsiblity to pitch in and do\nthe modifications ourself. I would much rather have a piece of software\nthat requires some coding on my part if they source is open and free to\nall.\n\tAnd as for the disrespectful messages, give them a break, they had\nprobably had a bad day (or maybe you did :). If it happens in a regular\npattern, then there is a problem. A one time occurance is not worth\nworrying over.\n \n> So, now that you understand my position, I just wanted an update, and\n> wanted to help fix the problem.\n\n\tGood, now that we got this mess aired out, we can get back to the\nimportant things, that is the actual port. Though in the future, kindly\npost a query for an update and in a little while (day or two maybe, we are\nall busy I am sure) you should recieve a response. Also, this project is\nso huge, that often the left hand does not know what the foot is doing.\nAnd to make matters worse, when the right hand asks about something that\nthe foot is doing, the left hand jumps in to try and save the day,\nthinking that this is something totally new and undealth with. I think you\nget the picture. Just be patient I think is the best advice here.\n \n> > PS. Kenji, where did you get that pgsql would be running on Linux/Alpha in\n> > Jan '98? I have never seen any such projected date!\n> \n> One of the coders told me this privately. I wish I had saved that\n> message!\n\n\tSo do I! Though there is/was a rumor that old versions (pre 6.0)\nversions of pgsql ran fine on Linux/Alpha. But then again Linux/Alpha has\nonly been around for about 1.5 years, and so the entire platform is rather\nyoung and still a bit unstable and changing.\n\n\tAnyway... back to hacking! :)\n\n----------------------------------------------------------------------------\n\| \"For to me to live is Christ, and to die is gain.\" \|\n\| --- Philippians 1:21 (KJV) \|\n----------------------------------------------------------------------------\n\| Ryan Kirkpatrick \| Boulder, Colorado \| [email protected] \|\n----------------------------------------------------------------------------\n\| http://www-ugrad.cs.colorado.edu/~rkirkpat/ \|\n----------------------------------------------------------------------------\n\n\n", "msg_date": "Sun, 15 Feb 1998 12:11:40 -0600 (CST)", "msg_from": "Ryan Kirkpatrick <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha" } ]
[ { "msg_contents": "There is an info file describing ecpg. But it seems we do not have any other\ninfo file. Shall I install it regardless?\n\nMichael\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Wed, 11 Feb 1998 14:56:13 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "info files" }, { "msg_contents": "On Wed, 11 Feb 1998, Michael Meskes wrote:\n\n> There is an info file describing ecpg. But it seems we do not have any other\n> info file. Shall I install it regardless?\n\n\tI don't see why not...have you submitted docs to Thomas yet?\n\n\n", "msg_date": "Wed, 11 Feb 1998 09:42:43 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] info files" }, { "msg_contents": "> \n> There is an info file describing ecpg. But it seems we do not have any other\n> info file. Shall I install it regardless?\n\nSure, or convert it to a text file.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 10:44:16 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] info files" } ]
[ { "msg_contents": "> PostgreSQL SUMs population column given -1523690296 (overflow)\n> While SOLID and MySQL gives 2771277000.\n>\n> Who are right PostgreSQL or SOLID and MySQL ?\n\nDuh.\n\n> Is it correct to have an overflow with SUM() function ?\n\nDo you know what technique Solid and/or MySQL use to allow an integer summation to\nexceed the range of a signed 32-bit integer? Do they do summations using floating\npoint? Let us know...\n\n - Tom\n\n> name \| population\n> ----------+-----------\n> RUSSIA \| 281170000\n> INDIA \| 766140000\n> CHINA \| 1072220000\n> JAPAN \| 129947000\n> CANADA \| 25610000\n> U.S.A. \| 242080000\n> MEXICO \| 81160000\n> BRAZIL \| 141450000\n> ARGENTINA \| 31500000\n>\n> postgres=> select sum(populazione) from nations;\n> sum\n> -----------\n> -1523690296\n>\n> mysql> select sum( populazione) from nations;\n> 1 row in set (0.05 sec)\n> +-------------------+\n> \| sum( population) \|\n> +-------------------+\n> \| 2771277000 \|\n> +-------------------+\n>\n> SOLID SQL Editor (teletype) v.02.20.0007\n> (C) Copyright Solid Information Technology Ltd 1993-1997\n> Execute SQL statements terminated by a semicolon.\n> Exit by giving command: exit;\n> Connected to default server.\n>\n> select sum(population) from nations;\n> SUM(POPULATION)\n> ----------------\n> 2771277000\n> 1 rows fetched.\n\n", "msg_date": "Wed, 11 Feb 1998 14:28:41 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [QUESTIONS] Error on PostgreSQL agregate SUM() function??" }, { "msg_contents": "On Wed, 11 Feb 1998, Thomas G. Lockhart wrote:\n\n> > PostgreSQL SUMs population column given -1523690296 (overflow)\n> > While SOLID and MySQL gives 2771277000.\n> >\n> > Who are right PostgreSQL or SOLID and MySQL ?\n> \n> Duh.\n\nJust to add in here...Oracle:\n\nSQL> select sum(population) from nations;\n\nSUM(POPULATION)\n---------------\n 2771277000\n\nSQL> \n\n\n\n", "msg_date": "Wed, 11 Feb 1998 10:14:15 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [QUESTIONS] Error on PostgreSQL agregate SUM()\n\tfunction??" }, { "msg_contents": ">>>>> \"scrappy\" == The Hermit Hacker <[email protected]> writes:\n\n > On Wed, 11 Feb 1998, Thomas G. Lockhart wrote:\n >> > PostgreSQL SUMs population column given -1523690296\n >> (overflow) > While SOLID and MySQL gives 2771277000. > > Who\n >> are right PostgreSQL or SOLID and MySQL ?\n >> \n >> Duh.\n\n > Just to add in here...Oracle:\n\n SQL> select sum(population) from nations;\n\n > SUM(POPULATION) --------------- 2771277000\n\nOracle does all math in base 100 so it is not the same as using\ninteger.\n\n SQL>\n\nKent S. Gordon\nArchitect\niNetSpace Co.\nvoice: (972)851-3494 fax:(972)702-0384 e-mail:[email protected]\n\n\n\n", "msg_date": "Wed, 11 Feb 1998 13:27:54 -0600 (CST)", "msg_from": "\"Kent S. Gordon\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [QUESTIONS] Error on PostgreSQL agregate SUM()\n\tfunction??" }, { "msg_contents": "On Wed, 11 Feb 1998, Thomas G. Lockhart wrote:\n\n> > PostgreSQL SUMs population column given -1523690296 (overflow)\n> > While SOLID and MySQL gives 2771277000.\n> >\n> > Who are right PostgreSQL or SOLID and MySQL ?\n> \n> Duh.\n> \n> > Is it correct to have an overflow with SUM() function ?\n> \n> Do you know what technique Solid and/or MySQL use to allow an integer summation to\n> exceed the range of a signed 32-bit integer? Do they do summations using floating\n> point? Let us know...\n\nTo have overflows isn't a merit but a lack.\nMySQL and SOLID don't use overflow even on SUM(float):\n\nmysql> select * from t;\n2 rows in set (0.01 sec)\n+--------------------------------------------+------------+---------+\n\| myfloat \| myint \| mysmall \|\n+--------------------------------------------+------------+---------+\n\| 340282346638528859811704183484516925440.00 \| 2147483647 \| 32767 \|\n\| 340282346638528859811704183484516925440.00 \| 2147483647 \| 32767 \|\n+--------------------------------------------+------------+---------+\n\nmysql> select sum(myfloat),sum(myint),sum(mysmall) from t;\n1 row in set (0.00 sec)\n+--------------------------------------------+------------+--------------+\n\| sum(myfloat) \| sum(myint) \| sum(mysmall) \|\n+--------------------------------------------+------------+--------------+\n\| 680564693277057719623408366969033850880.00 \| 4294967294 \| 65534 \|\n+--------------------------------------------+------------+--------------+\n\n-----------------------------------------------------------\n(C) Copyright Solid Information Technology Ltd 1993-1997\nExecute SQL statements terminated by a semicolon.\nExit by giving command: exit;\nConnected to default server.\n\nmysql> select * from t;\n\n MYFLOAT MYINT MYSMALL\n ------- ----- -------\n3.40282347e+38 2147483647 32767\n3.40282347e+38 2147483647 32767\n2 rows fetched.\n\nselect sum(myfloat),sum(myint),sum(mysmall) from t;\n\n SUM(MYFLOAT) SUM(MYINT) SUM(MYSMALL)\n ------------ ---------- ------------\n6.80564694e+38 4294967294 65534\n1 rows fetched.\n-----------------------------------------------------------\n\n... PostgreSQL isn't coherent; \nit gives an overflow message on sum(float)\nand nothing when overflow on sum(int) or sum(smallint).\n\npostgres=> select * from t;\nmyfloat \| myint\|mysmall\n---------------+----------+-------\n9.99999999e+307\|2147483647\| 32767\n9.99999999e+307\|2147483647\| 32767\n(2 rows)\n\npostgres=> select sum(myfloat) from t;\nERROR: Bad float8 input format -- overflow\npostgres=> select sum(myint),sum(mysmall) from t;\nsum\|sum\n---+---\n -2\| -2\n(1 row)\n Ciao, Jose'\n\n", "msg_date": "Thu, 12 Feb 1998 10:23:10 +0000 (GMT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: [QUESTIONS] Error on PostgreSQL agregate SUM() function??" }, { "msg_contents": "> > > PostgreSQL SUMs population column given -1523690296 (overflow)\n> > > While SOLID and MySQL gives 2771277000.\n> > > Is it correct to have an overflow with SUM() function ?\n> > Do you know what technique Solid and/or MySQL use to allow an integer summation to\n> > exceed the range of a signed 32-bit integer? Do they do summations using floating\n> > point? Let us know...\n>\n> To have overflows isn't a merit but a lack.\n> MySQL and SOLID don't use overflow even on SUM(float):\n\nYes, I understood your point. However, we need a description of an alternative\nimplementation to evaluate; can you describe the implementation of Solid or MySQL for\nthe sum() operator wrt integers?? Does it use a float8 as the accumulator?\n\n> ... PostgreSQL isn't coherent;\n> it gives an overflow message on sum(float)\n> and nothing when overflow on sum(int) or sum(smallint).\n\nYes, on some or all platforms Postgres allows silent overflows on integer types. I'm\nnot certain about the behavior for all platforms. Is the \"float\" type on MySQL and\nSolid 4 bytes or eight? If eight, how do they \"allow\" overflows??\n\n - Tom\n\n", "msg_date": "Thu, 12 Feb 1998 16:57:12 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [QUESTIONS] Error on PostgreSQL agregate SUM() function??" }, { "msg_contents": "On Thu, 12 Feb 1998, Thomas G. Lockhart wrote:\n\n> > > > PostgreSQL SUMs population column given -1523690296 (overflow)\n> > > > While SOLID and MySQL gives 2771277000.\n> > > > Is it correct to have an overflow with SUM() function ?\n> > > Do you know what technique Solid and/or MySQL use to allow an integer summation to\n> > > exceed the range of a signed 32-bit integer? Do they do summations using floating\n> > > point? Let us know...\n> >\n> > To have overflows isn't a merit but a lack.\n> > MySQL and SOLID don't use overflow even on SUM(float):\n> \n> Yes, I understood your point. However, we need a description of an alternative\n> implementation to evaluate; can you describe the implementation of Solid or MySQL for\n> the sum() operator wrt integers?? Does it use a float8 as the accumulator?\n> \n> > ... PostgreSQL isn't coherent;\n> > it gives an overflow message on sum(float)\n> > and nothing when overflow on sum(int) or sum(smallint).\n> \n> Yes, on some or all platforms Postgres allows silent overflows on integer types. I'm\n> not certain about the behavior for all platforms. Is the \"float\" type on MySQL and\n> Solid 4 bytes or eight? If eight, how do they \"allow\" overflows??\n------------------------------------------------------------------\nBoth MySQL and Solid they have types like:\n\n float (4 byte)\n double precision (8 byte)\n\nand they use 8 bytes as the accumulator\nlook at this:\n\n--SOLID-------------------------------------------\ndrop table t;\ncreate table t ( mydouble double precision);\ninsert into t values(8.0e+307);\ninsert into t values(8.0e+307);\nselect * from t;\n\n MYDOUBLE\n --------\n 8e+307\n 8e+307\n2 rows fetched.\n\nselect sum(mydouble) from t;\n\nSOLID Table Error 13072: Numerical value out of range\n\n-MySQL--------------------------------------------\nmysql> update t set mydouble=mydouble*1.1;\nQuery OK, 2 rows affected (0.01 sec)\n\nmysql> select sum(mydouble) from t;\n1 row in set (0.01 sec)\n\n+---------------+\n\| sum(mydouble) \|\n+---------------+\n\| Inf \| <-- (seems that \"Inf\" meaning overflow)\n+---------------+\n Ciao, Jose'\n\n", "msg_date": "Fri, 13 Feb 1998 13:36:10 +0000 (GMT)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: [QUESTIONS] Error on PostgreSQL agregate SUM() function??" } ]
[ { "msg_contents": "No, I've just added a small man page. Where do I have to submit what in\nterms of doc? How is documentation handled anyway?\n\nMichael\n\n--\nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n\n> -----Original Message-----\n> From:\tThe Hermit Hacker [SMTP:[email protected]]\n> Sent:\tWednesday, February 11, 1998 3:43 PM\n> To:\tMichael Meskes\n> Cc:\tPostgreSQL Hacker\n> Subject:\tRe: [HACKERS] info files\n> \n> On Wed, 11 Feb 1998, Michael Meskes wrote:\n> \n> > There is an info file describing ecpg. But it seems we do not have\n> any other\n> > info file. Shall I install it regardless?\n> \n> \tI don't see why not...have you submitted docs to Thomas yet?\n> \n", "msg_date": "Wed, 11 Feb 1998 15:39:44 +0100", "msg_from": "\"Meskes, Michael\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] info files" }, { "msg_contents": "On Wed, 11 Feb 1998, Meskes, Michael wrote:\n\n> No, I've just added a small man page. Where do I have to submit what in\n> terms of doc? How is documentation handled anyway?\n\n\[email protected] is where you shoudl submit docs to of a\nnon-man nature...Thomas is right now converting everything over to sgml(?)\nso that we have access to various formats at hand...\n\n\n > \n> Michael\n> \n> --\n> Dr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\n> [email protected] \| Europark A2, Adenauerstr. 20\n> [email protected] \| 52146 Wuerselen\n> Go SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\n> Use Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n> \n> > -----Original Message-----\n> > From:\tThe Hermit Hacker [SMTP:[email protected]]\n> > Sent:\tWednesday, February 11, 1998 3:43 PM\n> > To:\tMichael Meskes\n> > Cc:\tPostgreSQL Hacker\n> > Subject:\tRe: [HACKERS] info files\n> > \n> > On Wed, 11 Feb 1998, Michael Meskes wrote:\n> > \n> > > There is an info file describing ecpg. But it seems we do not have\n> > any other\n> > > info file. Shall I install it regardless?\n> > \n> > \tI don't see why not...have you submitted docs to Thomas yet?\n> > \n> \n\n", "msg_date": "Wed, 11 Feb 1998 09:47:41 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] info files" }, { "msg_contents": "> > No, I've just added a small man page. Where do I have to submit what in\n> > terms of doc? How is documentation handled anyway?\n>\n> [email protected] is where you shoudl submit docs to of a\n> non-man nature...Thomas is right now converting everything over to sgml(?)\n> so that we have access to various formats at hand...\n\nI'm converting the existing hardcopy, plus integrating some of the text help\nfiles (e.g. SPI and triggers), plus adding some new information (e.g. data\ntypes), plus updating some of the existing information.\n\nIntegrating the man page information will have to wait until after v6.3 for a\ncouple of reasons:\n\n- the current man page source is the only mechanism for people doing updates\nfor v6.3 to put new information, and there are likely to be changes and\nadditions at the last minute. Once v6.3 is released, everyone will have\ndirect access to the sgml source since it will be in the distribution and in\nthe cvs tree.\n\n- we should have a mechanism to take the sgml source and regenerate the man\npages, so we are working from only one source. Brandon is looking into this.\n\nI can integrate your \"info\" if it is readable as a text file. I'm going to\nfreeze (or at least chill) the source content around Feb 15 to allow time to\ngenerate good hardcopy and to work out conversion and installation issues in\ntime for the v6.3 release.\n\n - Tom\n\n", "msg_date": "Wed, 11 Feb 1998 15:38:21 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] info files" } ]
[ { "msg_contents": "Oracle doesn't count as they don't have integers etc. They only use a\nselfdesigned number datatype that is able to handle that kind of\nnumbers, too.\n\nMichael\n\n--\nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n\n> -----Original Message-----\n> From:\tThe Hermit Hacker [SMTP:[email protected]]\n> Sent:\tWednesday, February 11, 1998 4:14 PM\n> To:\tThomas G. Lockhart\n> Cc:\[email protected]; Postgres Hackers List\n> Subject:\tRe: [HACKERS] Re: [QUESTIONS] Error on PostgreSQL\n> agregate SUM() function??\n> \n> On Wed, 11 Feb 1998, Thomas G. Lockhart wrote:\n> \n> > > PostgreSQL SUMs population column given -1523690296 (overflow)\n> > > While SOLID and MySQL gives 2771277000.\n> > >\n> > > Who are right PostgreSQL or SOLID and MySQL ?\n> > \n> > Duh.\n> \n> Just to add in here...Oracle:\n> \n> SQL> select sum(population) from nations;\n> \n> SUM(POPULATION)\n> ---------------\n> 2771277000\n> \n> SQL> \n> \n> \n> \n", "msg_date": "Wed, 11 Feb 1998 16:39:07 +0100", "msg_from": "\"Meskes, Michael\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] Re: [QUESTIONS] Error on PostgreSQL agregate SUM()\n\tfunction??" } ]
[ { "msg_contents": "I have updated the TODO on our web page with a new item:\n\n\tHow do I test my changes?\n\nYou will find it on our web site with the other FAQ's.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Wed, 11 Feb 1998 11:03:17 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Updated developers TODO" } ]
[ { "msg_contents": "\nAs I just said on patches, I've just had a go at installing the latest 6.3\nunder Irix5.3. A couple of patches were required which are described on\nthe patches list (to keep Marc happy :-)\n\nThe regression tests were largely OK. There were the usual error message\nand rounding differences for int2, int4, oidint2, oidint4 and geometry.\nI've supplied the Irix results files as a uuencoded gzipped tar file\nin a separate message, so these can be incorporated as expected results\nfor Irix.\n\nI also got PST/PDT swapped for abstime, tinterval and horology even though\nI set TZ as specified in the instructions.\n\n\nfloat8, triggers and select_views gave rather more radically different\nresults.\nI've appended the results I obtained in a uuencoded gzipped tar file.\nAs you'll see they are really rather different and I'm not sure why\nor whether they are important differences.\n\n\nAndrew\n\n\nbegin 644 regress_errors.tgz\nM'XL(`(?3X30``^Q=;7/;-K;NU^A7X,N.[=2VB!<\"9+SMWB1UIMV;-KF.NYT[\nM.[L>1J(<;F12I2@GON,??P$\")$\")I\"6&L9D$9SJU<@B`P'EP#@[>#L,T3=+E\nM^+O/28`XS'7!=P``Y,+\\+R0D_ZO(`8`2YF`7$LCX4X=\"^!UP/VNM%V669`\"\nM\\-U5D&91W)SN[?+J/JISWQ1_&?S),B\\XV25?89W0,>A\"N\\:_\"&%A)7X8T8Y\nM\"U'&\\7<^0UTVZ!O'_W]^/SW[WR?@^=GIT_-3</[TV<M3\\.+EJZ?GWL7YLY?[\nM,PADUS@X&:FDO_SVYO3LG/\\Y?U5->0#^\\?3E[Z=OP/Z><^SL[9:#]Q%RC'?-\nM=83)L4=V?=4QPL2ES/,=R'^%WR-GY]JN%7'4H0@G)#+3Z=G9J[,G`#P+IDK8\nM((H7JPS,DO0JR$\"1=\"?![%#\\48?R>9:C':I_U7^6[_@:/T-;TY?GCX_!WM[\nMX.D;,(NNPT/C1<>/3T:\"=\\O[]@:-CCA]?U1#(_'XUMG,`N03V8'KGLA.6ING\nMIB?6/Q$=;+3O@C3YL#P8U;<S6:6'8';\\&+PX>_6KT6`P`W_\\?'IVRI_Q)O_U\nM1Z\"4;8_+@>?I60[]M9:TM3:)PRT:^X/15I[C5KQM!D>Z;2,-W6@?BA<VO\"][\nMEX9WO+%X%?@Q?_G),]SNRDH68,V^8:7,\"I?23EK1[5\"QW\"7+AJF.8\\CR\nMBVC9W0JA$?M!098K1&.[6A2BL55K3U1?:V]N%X6XN[4QA_6]?^+;&[9;6=0\nM_A\\\\S@>\"_.%'4>R&&`33L`W<#@HWI>C,=5;P]F--.T%+%Z]E%AV_;H2X/1)<\nMO/X&4\"L683!YADTN;%&2S0Q'T/>W-0JE=+^_)^GNMPVV?K^!J]5M(T#5-D\nM-X?!G04[_F[[?%ZO]VMVXI>V\"K<32[<6;I'7V\"WA0ZL]R1[ZK8UO?:N;JN\\\nM+2Z.?X,]=\"P%NOQS%:3AQ0S6\"A9\\>!>F88,;IK-6!52X8QZ%WC'QVUTRZ=A+\nML/\\K?P]G!F^7335J<?MO9;8:3'>$M-8QWG'6L.-4XK:&U^)Q[^B&=YJ)&,#\\\nMI0`F2U?QI`LT1<;!0%.?O!8:SFPJO0=H:NN^/30Y,G\\1W#19Q=,NV!09S7X\nM8K!QFTI_&&PJH]?MN%\"<Y9]IUH1-A]%+%;=6LV9\\=O<-,#RFOD-]1+'ONTK\nM[0,8E\"26_:'C<]_@CA%L/8-P1W;Q#9Z`?7\"2_SX0S\\/BXMYW.0RP(_FY#K\nMF#MX\"=B[VTVH9O\"WFG(;YN56=^')6]GG=K(M1:8-$3Z,;3GF7AF&R,/,0\\Q(\nMWF!;CA&E#B,(8D0XX%6VT.?>9S3%R?LQAUP^\\I;;$MZ!A#B!R\"B,/97GA$\nMV79V_ZM<!/W]]4]BZTW361^<WH.<L?3:'Q1K\"G%;KZM-3L.@F^#:9ZH0&&\nM8L$?S-+D:JT_W[58?70$DNLPY8\\^W/&:?V_SFD7R8?\\`I.%R-<]`M`0)?U,R\nM`VD07X;-Y>^?@/W<^!ULEKWFMPN!`/V^21#O92`+WH=@GER5_U?F\":?_J:_\nM;O&F`,3A99#Q7@OBU=7;,&U^[1/UUC;1<1O=073%G#VO;4OI$OMI=/T$\\/]%\nMTQ\"\\O<EE!?(=UT]7TA8M;5'3%FTL'K7H<+.JULW635U]Z&U.2PVD]O^S-+J\\\nM#-/E9SD!T+[_CYC#]/X_<<3Y#XR1W?^_%U)V:)&019R<_MV'H+%^_!F\"?;%\nM'\\A'KXR`.,FXS9W/#_-G\"&3AQZQDZGWO2BDS6<JL+.4P9\\G,\\C<6_+;L2.9'\nME0$OJY$4T5R6I^%'6=9%!)X4BM57ELF9.0JU/4HS7CHII35E;GB.)E\nMF&:B\\HF2]W4P7X6\\:M`YY\"/_WC9)D4BMDJ15\\55(BDIMDKHBJ;M54BJ2\nMTKT-F:WBZ,]5(;J%!LKLA`KKS9XBK1:8O`LG[R\\DRB+E1?@Q6F:C1V]#[G6%\nM1:62%P64Y&OP&7TB#\\'83!Y)\\:IT:/P8SA9\\>>+-)F$TU4:CA[)HA=IQ.W#\nMS07/!/;W\\LZSQYN3=P/Q(Z]L\\0,6/]!F:YMJC'JK,FBJ,2YJO\"RKC/!6-40[\nM\"15]@E21%NNGR%5V(U',Q2183H)I6%9[&LY#45FSVHN=^H(H)KJ,9:T1KY)Z\nM14TU#T>/I,2+5E5[C09CH^W;P9WD'NP_.$D:VWBELB`FB9R[=LKWF%TG+^\nMQ8\\FVZ(,X)IU.P1PJ^3;!T\"M%5R9;H.@;M5<F6^#D&#85Y+SLSDY3RE7D^>\nM@&RUX,-8&L[X)\">>\\/QQ$A_ESZ+X$@C11D77:WKUIM3:Q+`IM+9F;<ILJ[)+\nMD6V5NI08K)?89Q.8&FT.`6EY,=KIS4J%UN5DN4S0#E4R=FL=)I^`-@!03Q5\nM[M(/`HZ3T6^OSG]Y?EI69--/0%0UF5_>2#FHK)-0:G3T_I>5S$#&ZV9EI7?\nMMNYDK>ZDG[IWP)5RE`M4(9-5GX99F75V:&N-N,S]&K#W+6&N0,\"I6.3X%J3\nMX.\"P^NGLU6MUL#@OYJ3FP:SU`:J?(4Q7\"\\C`_L=\\;M`X8.:I+N30N.&\\Q/+Q\nM%F/_;!7'-Q?JE0>C6MLGGY8VFQT84%3RSZ4]X!GIR]>G9T>B%K,P^MP+J;%\nM3X`S=J1@EZKL<#I><HE.#)'?X-)LHJS_<=JE2LO^V24,T=Z&0AN;+W6F7(\nM>$\\8H^8J[U@<%<7AWHIS17&DM^(,[MK3@LBJ.]%8=$<:RWXJ`HSNNM.$<4\nMY_=5'/!%<;!%5W8LS\\O+@[V5E^L%[$TQ0X8L#?-`+EFP-Y4`^2J`7O3#9#K\nM!NQ-.4\"N'+`W[0\"Y=L`6]=C=5L/-`T_3-%G4#F7EV-4^[@6S+$Q'C_(_]S/J\nMX3N&D<OSD_/P,80PL:X8U^L5#H\"AV3CL:FID\"A+.Z8=M2^F@%MI`QZ]B]\nM:PH4ZH+'7D=S75.@T!<T]ON3H5`88;'[U&,`ZWV;TU6@PJ/C?:O<$B1A6/\nM6^W>.HX85A@OL#<ABG&%CL5V3%\\%<EUQ>8']@<)UA?\"1I>/07%,@UQ7,\"^P/\nM%XKB!?8GPSEX-)6X)VCBYHG?A1S1+P^U/\"B=YD6=+7IW),G'3V@!JO.G+';\nMU4EKL.O4'U/6GU$2EIUZ8X_T:]LIUWJGJSO98-XI5WS(.KI$#1:><MW'I#^$\nMA(VG0OV=_K1+3!THMP\"4]FKH(>5&0!P.Z]'60\\KM@._U5V;N&CEC!/OKG_E4\nMPO7'\"/>(D5`CUQLC0GNU^]!E8T2[3@/J33]TZ1@QUIMNYM;?=<?(;W-$/FD`\nM$$5O#`#&LEOMS&(9_KD2:Y<@RZ:KJZN;\"\\[@%>-/K\\(X$RN1XBA%!AQP%<7Y\nMV`&<^I6[+,O\"90;V1X\\6:30)+Z+IHWQ_O_@WSUQE)/':OV>SG`&FX2P0Y[?\\\nMG$:-JX\"JRJU[9+)66\\R5&%@OZC<(2BKU5(#CF`4S%LW4+>N0K#$BYZLPH&\nMO5;,TKJY:A^\"/A_U5,3+<G1(4`[),>'@/_GZ+2JHSZ6/52F/AD5^-^6P-\\6\nM+;HM93K:/.3[_29KI,ZJP?B:?$+.N'ZB+J&2I3%;]032I<IBI^83/5QF%H\nM<V=%\"40MUZMV<F5$0Q7)'8V]0W\"\"2,WY\\+I&5H3\"6\\.E(E\\RW-;OH[5V:55\nM#<HW;(H:R^V_#>#Q4(&_`])U`>'BH;N#&&7ZS4O=36+,)>9^O4L%T_Q:T/4\nMM$W4:X#(])L'Y97`N(POU%BQ+^SSWR;)?'45_ZT4RFC?Z6#6W).[(#45O9_\nM4)EO`=H[M.A;UP^WHWZT\"[&FY\\/M>GYC7XZYK(_@IORA/FY2NHDR]\\$3<).L\nMU+V&R3MQS4`7%L33,??YM+(JG%':K4,S8H??+HV=VG3??>I>^PMC;WX\"&[1\nMIUK\\FEC$6L&N/)]1`OMCE578SL\\@PSND4[_59,S/3BI/ZF9_)_9>QY=.ZOZ'\nM[,T7UU'XH?\\[(WW/XC#,%V/_TF0X]K['_=!U7MHGC4,DO#D.M_'%R%H)5N\nMLW?A(LC>M:>R-`BZG12`E@-),]7?++0T2#(&_Z>32;A<@K,IM[4-W>\"?^T<0\nMP6,?>?@0LV//)P>'599_\\_[[)B6&NGV5?!^+MQ%,'IZN0K`BS0P'-NGM_7\nM)2Z1=4D.(T(&LC`'V[/8#H-,9.-I&@9-\"9]':8DL$S=C/0DD);!`5[`]FD/N\nM4,@LP@])K)B[I9D#0E?!G%8ZBQAN8(2IT35)VZ.?:117009\";SJ/X,N.S\nM_#KZ=76V/,<))'U\"F0YBTEDB45V$+0ULF=),\"VML>\\RZ4&5R')688<MLH.@\nM\"K)I<I-PL[L$KY-EE/&_ZXE+9)ETA'UF()N#[3BN1780M(GLFW\"2-\"4ND'4D\nMLHS2`EG4<Y9R/K&PR`#V>?!/[email protected]=\\UZ/.3DR#K8+<=9),=9'T.+[\"#(\nM1#9)TZAQVO-'</GLZZC?&,]GW4=5WI0GD5V$%1!]H.X.-]`VH-\"?#\"5OK&/\nM]1H48SF+>7;6,PPRD/TIC9:39#ZO3VCZQC[QUP'Z_FL1UPHM=;WK1?U\\&0@\nM>QHG&]ZP)M,:,\\K4K?TH#A+NLMVF!T(&<B^\")4;.S6)WQZ'>IQUO=SE47E\nM=-;W:`XL=BVR`R$#V5^[email protected]$Q?(8B3'6<_7.P)$NLL.U2PHI[C0+DL]\nM`'5`EGM0<CO`@0K8W?RY3Q(>\\O,D_,@3P_&KB<YVO%RY0HTUWF+?K^T-;)G\nMP=6BG,^Z\"$IHRQT!CS!?+DOY!HMNL.0:I&/D<]VB`MMG]0-6:9,+RR5SV/\nM0!LP,D@E@7MWJD3LKY#U+JQ=J$<K+11VUX'48PDDQ&#JY34+FE\\5NJ&+9>\nM,3H>EBRBSLMXGC'88G=][X\"S'\"=+6-UTL>$KGM@&#-:IH0&>[-([='9WJ)H\nM`]GQ+T>T#M\\<V?W\"ZU6+Q)[66>@K]:0'!_??#$L;5$6V%E,C<8$LSKTE:&B<\nM](,A,5ARZ0(B4R\\1D?AS9<2&$GKRD(W#'-_4V)Q'/8,%I7(Z%&(C)6)J'\"I\nM,3YP19:)7>@87.@1:0@(90:;NW#2=T#:%\"#J(=%2@^&NCT)J:N<;DSVY7>(3\nM74/^SF(U]MZZ_6[(%M88RAT!9F`+(:(YWHQP>1ALR:1ZH(4H;R9C6AA\\');;\nM]D9YOF2YAERUI;`V]T[JBSN@QH<N;B(#01SE3&FLXXK=PT<NVOP^6EK9\"NS\nM'D\\Z,5BK7<GRD<'UR\"-?[MK>W^>)[$FR&+=-W5\"5CI+!)NCC2<1,HZ68V_M\nM%)R/)([V?.-]4$=D\"]2,D^282-1<@Z5V]2R.#T!59+TM?6,JG26':1BIW.QQ\nMF)YX4K642(U4:L79\\#==ZE7\\3RGRAC;V=,GT6[(_G-#^B4>:F7?F./P685D\nM&:GD0C+4!II(`\\W=+6.ZE#M>KC79GT0&LG]/WL7+AM/&U9/D/I^WREF5E\"8\nM<SQF+WT,A$QD5W&T\"-.&A&\\RXU2JA]1=/'W>V)/0VC%U&0@VW$'S=$:\"$\nM>-(@^U!O[C#75>,O<NPTYJ')1#:Z#M,K$:\"OCLS3,HSQ42.KSX'Y<L9KCW<\nM-A#JABQ##J-=L<K?2;.EIZO8W?1'YX,9'^-ELNHT8-Z-K^>&JN+<@U_8Y77\nM7`OVI;VV6OPP9\"#[6W`=_?I2@]X;GA0W`/VUWQCC]EH!L,B`]G7T=4\\:KRM\nM5;D_2R%=V\\WRBC4(QV[$#(,,9,^293@-Y@V1@TR=]9$\\!6.>22UF01;6@9\"!\nM[)O@,HJ#AH/D56OL.7CM9C1GT<)`6W\"'0\":R\"W`61/.43UR;$A?(^MBEZERB\nMHX]!>#XK#LQ8<!^<.B'KN_+4B&SG(6L!S4D,I'-1!C7GS_<`(\\T)&Y%ULWW\nM;#C3UP?,?;4JY1NW]E1`\"R.-.M\\(;;?HD4QD5W$R!V?1]++.B3I/Y\\9>CURG\nM(YY>DW>C+:N\\4#(0/8\\6\"Z#_P3IW3'<^.254G78S'\"BF<.IMJUB0&0B:P(\nM_;S<)IH!0YZRL?K^K%Q8].P!BJ&0B:Q870PF=<HUN)4(`?GOK%Q8%1M_5AU\nM'0H9R/XCF,_#2935WWPW=_$\\J]&^U\"?763%%-?.9X=!!K)_!.D5>+-(H_BR\nM!MOC@#&TL,U_\"?IWMHUJ,&0B6PXG[QK])ZJ.^^^<H3U.,OD\\24\"L05W$&0B\nM^R[Y@D?9YO`U1X4H6IG!R];\\<PAOG]#`OM$,A`%BZ;(N`JIR#<EU'N<)(\nMAR5G+G9E^&H_OQ5I\\7U0JL1=G(9-^^YKD;Y\\E@,+&2L\"5<#BWQ;/@9\")[++M\nMHQY:9SFR1IL<5U9<,2_D;W6,1PRD'V6BH7%11+%6<WRHMYY1^)&JKSJ4=R\nM%\"P97`;9I8J!D(GLLO\"=!:%\\[J]G@JR'E8'%\")K\"?O]?B>-<C#(#.B9CB/\nM&C\\04;7&%,H0BR6NZOB,U=CAD(GL/$C#JR2N]Y!?STUK3%\"QF'A89=ES4$,A\nM$]GT9ID%\\RBN/0E5N=>#7^Z!L49ZGJ`XR)KCX=`9A3<<))DC:=2[,>Z/J5\nM$#^\"1=='7E@<J+&K4@]!9A1<,9UM7'_$Q?((N2HR\"\\.9=HB(QFTS[=QR0=`\nM!K(_A\\M%F$9!_6J%WA'(5RIR!PIK%TJ=@3%FN.JR++(J^S#40HN0U)A=<P]\nMAI\"TSSXV6%@&%^PZP_H,00=R>9_=90@INZ/0&,S'\\K[?E\"?CF10!H_3]Z@Y\nM1X6=TB]6MWNACNS$'#_O9M\"X8N8PAF4]7&IP93A7J+]JPEFR.#.5#\"EE!!UD\nM#E;+<LA@L0T64BS=\\$(:7GDPCHOY$#?8,G+Z%`'>'JJ\"ARRDI0EG0<P_^\nM92-]'7T3D;YLG(JOASK$J<@7)M3DU5BK<)0GK$=:UU?'X+2_S&=+LH<S@R6[\nM.#4R%L=4/<W\"%N3L(29:YI,NNF>2[2?+A,2HE>UY6L]R`R6/`,\"L<&2H<F^\nMCE#+W9#UY`B%H;&Z\"-6Z,358.<?30#!I=1'Q#);D:`:WU+F$$29&88P;VAP+\nMA%RC7S#JRI,[W%U'9FHU\\\"/H&HO;3(ZBR/D&ELJV1E;'@QQY26,V-%Z)K_@\nM8SWAP5!'9(G266WU+<%,-0:XJ@/EU98[D9&V2>PX>7X\\C`.LN&/X'Z0U9Z\nM_25`.4L:1(\\8^$N+ZC$+XV>F;LAKZ7B&C$5H5:C:+A9%K+[IV[(%I,7J,=9\nMBF&+(VRAO7?JAJR$#&O]]!QEGNL\\8POK0U!?R%ZB6PY#;+0/@`9R/YW-)\\&\nM#=&@UL\\N4F6--9!^/O9:$`=#E7A0\\VD:-MUW?VZ>EBEFIXY4_&:P,6W2%0\nM!=DTN`J:(FKJ&&Y([-H^(IZIBJBN16?QOL6EG@&3S:&V]=)K))DH:-WQ8V\nM==9QY9=9771MX?)!T4&LOQ7VA\"D`JS=G\\7JU)/>@O20A_.M$HOK0,B,SA<L\nMPP2\\#J:U_O'K]Q]NC)CDZC/\"QI>3L(LLL@,B$]DPC3Y$\\?OZ^'R560]QX;HU\nM)J38(O/%SJC%]Z')C+O8&N:K,LXBHKXTJG<$U%==S,UXA+`]EOI@5$$V7BZ2\nMICNTE7L]/JM^1MBN5`R/A$UXRP`OP9I[>V>BLZR_/2:<0Q\"?7G6`CL<ZA!W\nM44!9F%X]SBJ6;X.0#(2J<1?CK'&E8FT-\"JL/\"5.?&<M0KCK921W[':6'IDH,\nMMV#R/DR#F]J$56NL#K@89\\?DYWJ0M<A#(1O#[6LE&^GK:Z5I]E%J8Q>!U,\nMHEDTT>[4F\\5[@\"5UAC+16+'..JM6.)`N`5W\"&0C?7VM9$8-N@[C50@@)EE#\nM^#2-U9?/#3N04//57=GK5,\\#I!ECAW(EO<Q=)+B9#ZZNRB=8Z'06;4H-:$\nM9FP9R(B-E^N5'!6KL:$V$7B89\");!!/TF36L'!<6:EPF0SU)>[#EMA/2;V\nMJ,PPJ!NR5\"JHBPQK+$^2N_KV!Z0R;`6Q>ST/02:R43IICKQ8L<947O0F^H0;\nM9TDMAM8U'@9U1%9N[KB.KY&E2D&M;SP,JD343(+I%E\\6SC=CI3'6=WA\\&2F\"\nMV%6HH9\"-J/FUDAF=KS5A==8C;W_HT!#E/`C;+8&!D(EL<!6FV^W/JE.IKK$&\nMA66D(^+9=>-AD(GLNV!QU1#!;<T:$Z2F.$;X%N4;VPW:@5`5V73>>$F@$M\\8\nMR;!F_]_>U37'<1S)>SW^BGTS=7=R3'].]UW<@T73IFWYE&Z</AQ!(R(#2YW\nM$/MA&@K^^)M%U6QG#V=@>,\\$&JO%TD=@\\4.=U=75V5B;(MVFEB5A(594#T\nMC<\\5R&RW6F[;_8?I!U--A?ZE9JU`FZP_V`W$B5]T4!]XSL?S&+CH-1H-5:!\nM](#D0%L1+GZ7('U7<^F,]9HS^;L]H/L7$:HA$A]C$PUB2?17:>Y5X\\9%99\nMYE'.LV5`U.;/%9G:_+MV_4.[>==.:56\\_-O0/WN(A$E)WJ445+2U!%!%(6-V\nMLUE.LGI`ML_&H3,Z)8[YLD?F:RDXP4?@LE22T`%-16/0/2-Z\"JO@%J+V%\nM81Y+(Z1P'B'XJH:X&GZ.Y:P#_`;6+P=+S8IBN^C@LRBVBRY]61;VC\";52\\?`\nM!N?@<<$:\\3AR._`DLC&G,>MHSM90N^A(,MP#%XXD-;U+<9:MZ2GX04OW1B[`\nM$/4-.9\\N]FUD5P9G0VUAF$(YN'5R](HY`Q5:E1\\7\"EBZ8K;PGEB25;%)%[O_\nM(C0$1G.CQO\\NB3';NKT)0RK8]5P942T`X>(@<,`Z^<&:OC5+_^6!9^HP\nMZ.7=]7[=FUE4Y^OGW6=5J94/P^DU#?FA!'EBXI7_UC]MG,;LY(M;CU]<2XOQ\nM](LKQ'YAG.8C8'A132M>92@MY56$H4!#5-E+%O7^LI6\\0-2V8=?0.AA6O\nMMK7!7\\4)+^<-'KDXK'/)VN4P&LA:I%^PHTZ_4-?!T-,F@M:V=K&.G/_&#Z$2\nMDJ=U5A?O#_'^6&0)!3U:(@M.;D(ZJE4@6P=>SI(23!YRI%TZI?XRT45O$G\nMPAWP,;PX@S_\\%\\=)SJ:#<6288NJ+?N\"!1C2-(3=/S24#%B4KVL>5`X\":U^S\nMA9@ZU%:E4;(:2V<3Y?F[P(!BHZ@8\\#G+5FX0,T>VBDI/]2L(26X?W%;3[W4N\nM4M]29;Q+@;CE[=2&$Y3))G^#!R!1(/A!BUP&[email protected]!0[1!G;H%B(_`N4)\\\nM!,X5J$G>++>SR<6\\PLT/%I+]VJ.!W3[@I1R#DM[WQP8P^[I9O%DU-U,R7XN<\nM6<5./,ZFG:B2[JRR@,QV%_OMG#C?J^G\\N/DKXN<79EMPR@<G7S;MVM)J7Y\nM%J.JU$@AFC->[IW#W(B049L[M=-ZM)GNVS;&\\'/0+]4(B<J?%!EN3'1\\;L\nM_H=V,[<<Y]4RBG47T^4)IRI$ZZL4(+,7+U;[6;^>C%G+)_2DJ-D/R<U[40!F\nM_]CLEMVZ64T_F,]9RAJER%A3XL:(64Q0+7Y;K/;?NCVTYI!N58JE3R`L^QQ\nMZ$G4&_P<(&KSYPI1FS]7(+/MY<>NF],CQ^Z/H?P(+S2&^F-)A8\"\\1$X5R\"S\nM[V]6R_7<J2?I&Q]N7EF!!,3Y%)=42(53(;@WL]D^RZ414(041,N@,(CWQ[GB\nM)&:5&2I<L.+8>#<N+S94\\U+M%Y)B?&#XC1FH3\\K]?48'?B^1Y;D`I#[]>S:\nMQ:N/-XNH9QY./0)410B%:JS89Y)BGUD>6!2O7$_0J7D\"AJ\"-/T@%,IHX9\\\nMUG'(W[Y+&A8$1=T:&LKL-`W(B[7X!YB=ZOXPJ5)&L1<X)AR!6?E3/SC$^^-<\nM@<PN5ZOEAW8FO8BJ056D]+^#FOO(PHN2)\"X$&;/;BVZ]74Z+5>2=T<.<14$H\nM-]S/IB$6-Y`2FD<`,NJ6=./Q5AMWK$F.2A7<[ALY'!3!C)FZ_LRZZCCPF:=\nMT73W+MML(0!F@[N#V)S9P(U7/IUG%JVM<3!A0\"9C?=G5E,1JDNK<:BX:4`F\nM;1D`9F.X-[..-<DA7'+4!GTH5!5F2\\!IS'JZ>;<&S'E8B\\](%P&T!=OM;I#\nMGN\\%]@AP^L^C?B99^#BYZRD$G[YI-^_;57MS8/:OS::Y[!9_7$Y(DV=:J89R\nMO1'V=A:0I22P$RNUG^UUGVGKROA[N\"X]P/VLH91_D[KT0(+/B/WM.F&!6\nM_&?/`L#LK7O:Q^;FCH<''P%;46=T.&8JM(FLNR?GV3`S.XOWB_7[ZZ:28\\(\nMK%W4FBU)CXDK0<!+\\E!%8,6?'%.R,`LR^:U?+';K.>[GK'V%BSUFR55+@T\nM^9)P7$IR)C]<+V_5\\^[JEFKT(-KMMI'#RDI>A^-GAIEGX$9,QNMK.&AR/O\nM#\\MIB>.9A]0LH\\1/Q0\"9;=>[S5QC=)8WUH:(A.RB-D-0)1.T#\"\"S?51\\O9R;\nMM=D^J\\:ZJ/U(/<140FP10&;%\\?\"<@,S>W_%0D<.\"3Y7DFHO+72KUUA4I^#DO\nM9#\\\",F8WZW9UKWTV4G8QZG3JX:K4F(Q>=\"0-OU#+UOL(0&8W[78W&QSC+9[V\nM)HXR%=JS40HD'#W):\\8D>:P=^Z+44I7\\I9$QVWU<O&C6-]V<B5IR`S>CXVQE\nMJ'Y&JJ\"`3[[Y?B^_GB>6V'=3:)&9-$E);L8AD`9L5_]JR0,=MNENMW,P^.\nMSK,4054V'6C)HRQ('50A0&:7_6%VKBDZCXVYFQ5LZ_HA/4QC8;8$(+/=>K8\nMC_/F/,<!N;A`O('2EC5L9`&;%6?BL@,SN+U;+^ZW&?#\\'^L;:L[VAW+P7\nM`F#V-UVWNUK.):&R:AEV/$R\\LKEAD#BX&&3,[M>[9J8Q.H^-#1NZ)DM2K2-O\nMO9)+`/`[&\\WRY]^6MTLWK3-Q$Z;5[CQC4#JZ^E//=3]$20+508R9KN_SFGS\nMC70J/47`WJE]9LO70#$_@`$%FIUY(.N2_GCRE-9J\\317P[`[MVLUG.FFSE\nM.A76C')0HJ/0%G(F-U>MYME,[W19EJIBFH70<&+VD$.I>7\":QD`9N]V`[]]\nM^,AL1?7&&J38R#X-.GVJH\"J>VK6W,,QC8$$0Q[(TU9!SAI]+%O''(<MW@O!9\nM],Y%!Y]%><_HP(&>KQP-^.>R2RB8.$82@(TX0DG5I[`NG<:LH]6XSA1(J/L#\nMN'\"DEN==JEVU;\",./VCI_L^A9SR)V3N?BB-MK\"QI4-E06QBFM2(91BM'KY@#\nMNU?'H7R%GT<:O\">V,#B&V@FRT;(,#1E<4]#8'%OZ!V8MKA/EK/T$+QRYFB>\nMG$9N_Q8&GJ'ZW[\\GC7<2LUI%SANG0\\_XOW45AUHI#6<CGE)I>NK#X9BWZ.CA\nM2:]&36%:U^/IWA^_%L-93N&^P1M^D%#+T4%W2K&6?()ZU<'#9]'>J_5&:1;\nM[LULYC\\;C;;<06L<+%:&%M'04V1AM:7,T0M4K@/AE.9'1A,_+&2<8C1`]=\nM#=51PNC#XD1F:<:&B+J+7$$CL[(,G,@L)1\\\"A)9T/1LPD.3X5IA^%)S$+)]P\nM`AY3C=,V8ESA'#[\\#B1V?\"9FD%%AY<8<<@/TA43AU)A^\\LB9S;<[]13&3Z6\nMIC-C'R/?SE\"O(@P%&DIQEK%N:(ZW%3RI0DU'SCIZ'55BL^KM<%?Q>5TSAN\\\nM]N>=WO7'X+1+Z.`#>]+'I.Q:]0<H@MQ)E;PM(N4_'8J^Q##DG5/:?4YC5FO\nMQLX?E:_&75S]\\=&/LD:5HY-X3-F`?HBF,434CF<V2)X[XRD5$J+%6MBJIC<J\nM0+\"-H9@T\\'+TJ\\-H+1MZ=<&,&VTE!0(Y[%]G,3LT<`%<HD\\07'OY6JI$)+;\nM`<1,(UD%/=/0WFXBYK!T-TIH(TI0[,:?IX=H@4NA-!^Q(:W#AM!@=L$\nM_0_5\\\"44O:J01>IC0LI<U?V_0MHS4CX.\\ZJ&TV]10:=X1>N2?YB+SM.8K4G?\nM6('M1DUMEBI=U?5#FH906XB&P#3$#]X_2CGX>_F:SL=68.=]8H&$YN>OPL,\nM$I$]S$</D<GM0IV?5Y4JK1<J'X%(16SRNFT-BCG(JDB5?U+G5-+.J,VNR6\nM$F5H!DDO9328::0%3F,6D>)->(A>A%.W@).8'91DZO2'T[KBMS1E\\%2DK3?U\nM8FHR=?%I?=:TNZ^AB$R4/0.AG@7QZM^2AG\"+JZ5X8T=?I#3V5\"Q1=,)FP85\nM[]O0D\\2^)4^XD.`T9FM^#=-;K5D^-0)#GC7\\/#1QT<]Y>(@XBVDF].LGO?;`\nMK.?;7PL?Q=J/!KX#J7I&2`1[_J;)?Z:/DG@73Y]%.PG<\\VAG_&AST99RWL'!\nM$&\\;-GT'0\\>&@,GHVC[\"_GUO9K-,!;.!H1'M<!&K+%8]41?^R>-TYCUBEU\nM4SSKC;HC$!9J'QPG,:OZPR-M:\"GN4UQ4`9EDVF8/MZQ\"[,/C-&;-L9@-8CLS\nMNB0X.%[)<&CX?_%;#3`;.WB$+\\8.)K6'-_#T92]PS'DYWH9!4,<U8\"8V/$4\nM'<9/B:KRYSB)V:G(T5F9GT4!F/U]<_%^7L0MTR3OP_S;??:XIP92HA4:CE`\nM9KN;B_E\\HQ9SAKYU!E=U;3L/N&C_9D!F/U#V_]SKI`\\[\\5SG$(#B2`W!%42\nM\"9<!8/9U\\VX_WQJ=,5M%3E6DG!TY\"PNKQ2!C]GU[M=]LI]7Y7BPWB=E@:\\U3\nMM`UK^\"YCB96QHLCT^-BQ.SVJMM,[[7Y:ARX6B(MQH-GA(10A0\"97:XOYTVV\nMLO[9,\"B0!)BO5W])#HC?@Y`9KN+_<Q2O!CUXOF9RA<5?(-BJ@9%`)@]MMF\nM,]?POABMQO50;YQ4@Z@51T3)BT'.[ZG]O5^=]4??OYP$/WZ?3H$92JX-1>#\nMI#Q?/\\2:Y'+J0,9L_L?VLW<<IRYE\"K%FVIJ;ANB8E&;+P3([,6+U7XF,AXQ\nM:[G^)SD>]D/2\\UX4D-GEZHX<5:H&;GH56)5($N8IR5.5L&,F:WVUF#B$S-\nMHHJONOA[(ZS%57<B4+03`[.`>_;OUNI^ZKYK-#]UH!B<--ZXN2V78VE=<\nMN=\\'S$'R3X\\/9';=+N=<!/)]MJI]3?>Q7MO4PU_5EETB9$5^?`\"S;YK-^\\WR\nM\\MUT#)6YIUDJ[`4]&4U7?\\([email protected]>3D3'&=>EI\\+B526^Z7$\\;`09,QN\nMVV[QIKF<//F\\>?_Q)D50)'?K34P1%/6L:(F@\"@$RVS:;^7TV5]0,/$%3=M%0\nMLZQ7LL>6`6#V;;-<'0S!IQ_,(BA-H5(%+8K:V$CX#A$C7=B$?$HR)A=;Z^[\nMN>QBML]&SO^G\"]HJ.JFI`K(;'MY>3_O#Q6'.JB478PU-R[+:EP&<F8_=MV<\nMW#QF%P=1/&P2'7P%Y!:O$\"\"SW<6\\]4>^&IO/NE(K0]<_<HM7\"H#9[YJ+3?.A\nM74_[;&6Q<<T]SU\"56E/OAQ,GID0,7MYN6H7WVRZ;D)M/L]!#?LLB(\\.[1^0\nMEG)4>2&W!(^!C-GUXE?KRTW;;\"<>S%;CHUA\\ZHV$C^5A9S97;,X5$--;;?Y\nMC<!M.QUT?@P!E=!:#)#9JV:WVU_,.&SE3DSA,\\=#'A+'PU0,=MMVM5R_??O\nM>K0G11^X$1AD0^0^H!0@L^]O9GD=.434AG4JDO83NV[)54\\IR)G]V,Z6DF>Q\nML2(I>`\\B:&+^41B0V>O%W9<\"$$$-Z?^D+<-#4?;90G`:LW$0/$W,DG&&@S9X\nMSSI8SCE).3X\\3F+VH!H3QVJI1X$0E`Y4[\"X0K1=SB(<%,KMK=NWBU<>;1;]I\nM3C_,S)J)6]!J#\"R2U'RV.F'W%B++[+5B\\,AHAY&0CW6W1M,D9(>33]$'P4_\nM5W]V%J.49U+%UY$!`)^3=;1M?@49<_L$[Y9GF9VFEC02F6!(%\";K7L^2,5-\nM!9N2CB'+5?5/5:R#ZP_5%\\?AP!JW&H;HCPL:M_6@<0L_2#F28.!7T$/F\"9/R\nM3\\$DL]J$Z8>?#_OLT47L>/'.^M,5:KB-!9TF[`<&;ZX007A558;U4D/_\\L#3\nMEK,A(\"7,(U]]]<!_N.)Q;V8SI2\\=1G.1?85#$KSMA_R@4O+SGCR/A(S9?3/;\nMBI??XCD5,CTH[:IZ%\"QK2U+/(A'U2`!FO]]?MZO)N]D#OF^3`DGE(I]RO((B\nM-U=SM:J+AX!(^'Q4`+-_/B2@7G?=]>2#62^>8T5\\4/KRU)XG2:A2,&VW:P7\nM;YJ+Y8_+BW2X_>YZ3ZU;J5J&=/`K,.<8I/&]9\"7`#[W%YTZ^V,5D5>NS@P\nMFQ1(N'81[8$C'X.BM.8]`I#9>^>@0!@UA5!\\_)`I6PB`615W5_,/9C?OW.H!\nM,M6:/;8.QU>AM@0`L^XN8O-33SSV\\04'7?Z2,5,(0!F@U[/+<.+D0()>TVY\nM\"(900UE%D)M\"0!F?W5Y1R%Y/F?9[-Z@FPW-8RM-`H7@T^OFQ^:FW>W:9R^:\nMU?+';K.>+%W,F:T,W^&EY9A5J1<IA1DS%YOEA<]M9,/9A&4JGF\"IOLR18J:\nM1M1'\"@$R>]5\\^'=3)61+T:UBWP_:](UJ^+&2^,EA\"H#]V;V=;-N@5E%S,9Q\nMIL+4LL^6@1&SU[-27]D^J\\8NPOT(=UG34`A0&:[;C4K&C3R$:!+N\\I!I$+\nMHR2`@3([;=[F;5^?\"N1WL3AQF:'$+-N#2KW]B#7Y-7(U1RZGW2P.8_?7R\nM0S>?J\\ACXT%M'F-C/@C)I\"T#P.S+?B7^V.QFY%+SV)@\"J!I,7<ADRTK\\5`J0\nMV?U%O\\_./9CWXM6L;YRH]71)(#X\"I0\"8?=4VF]VL!LGH/'M+H[%J'$$9(^?9\nM,H#,=JO5<N8T.W9U(3=H`[=X)&]LE:2@\"@$P^[NO%V[>I?3VX6,$Q97D:3'6\nMX__6%;<#1=3EZTV>E1YH77_-G#.7IXTJM1>XG6I(X!7C+:4/A=Z121&RI5\nMKU5NS:62D!28D4;9RO2H?-6P^=1!TMU!EMSNP=Q&;^L]%SO&3@@C92WABZ\nM,1P'6I]>`2,F+VGLW`]1%\")QMOY)`R6`V1V?;EL9JV8,K^>B@)AO)^M2&33\nMCGWE`%@5EQSPH9L^)E>49`9L7+\\IP`S'[;K\"^;5;^=3B[$3!TRL'SK&6\nM'BL5;H4`F'VSZ2YV8XN>([+56-.<M:\"HJ?F((WT]A0\"8_>ZJW2POYZ+C;,[J\nMFC).X`2NJ>#8IOYFS6KSII;`ZA&`S-Y;S:`_XMPRFTZSE2&[GH/.06J.9D<U\nMI7&HHDR24MY^IF@N1Z9_'N[-;):IT)2NP[R<-D/>^#A\"-34FPI)-U_-&='&_\nM/\":9U7;FX72+=TN:ME!`0<(2&A1::#/6/JW/WM_.5UWC$/T<A-F>$Y707.(=\nM74$8^(6.:W:,\"Y#2])8]H4RML;[#<A@`OYDD!@THSWAR7\\67T2L.>#+$K\nM0/[:5YRG2>1\"LQ^W^PW^]F[GKRFPO'-3OH+,U_)U&;+P/`[#_0L45)XB0T\nM_UG'0C,<#4O/#\\.@-D8[MUER?7&&FL7>8&6.+@0`+//==\"+3?=Q^]6S9__S\nMOR_?_N4_%]^]?/WRQ?>+?UO\\YNV?OETLV[\\M=\\_^]4]O?_WR[>;ORS4?RST\nM?SU;-Q_:3[NK]KK977VZ?>+K'O_^->'VG\\^>5W=^\\Z[:3]<+_[\\ZN7;EXO#\nMYRW^>_&+[56SZ=:_H%_0?]'F7?MIU5TTNV6W_M2LU_MFM6U6S]O&7YE^I>O\nMG]&'?%IH]^EYOSOVV^#BT^%_6_4O9_7LN3I\\K?Y;_8M`(!`(!`!0\"`0\"`0\"\n5@4`@$`@$`H%`<,;X/\\O$GO,`@`(`\n`\nend\n\n----------------------------------------------------------------------------\nDr. Andrew C.R. Martin University College London\nEMAIL: (Work) [email protected] (Home) [email protected]\nURL: http://www.biochem.ucl.ac.uk/~martin\nTel: (Work) +44(0)171 419 3890 (Home) +44(0)1372 275775\n", "msg_date": "Wed, 11 Feb 1998 16:52:30 GMT", "msg_from": "Andrew Martin <[email protected]>", "msg_from_op": true, "msg_subject": "6.3 Snapshot 11Feb98 Regression tests under Irix" } ]
[ { "msg_contents": "\nAs promised here are expected regression outputs from Irix5 for\nint2, int4, oidint2, oidint4 and geometry:\n\nbegin 644 irix.tgz\nM'XL(`';3X30``^T]:W/;1I+YNOH52%)7$A.`GO?#]FKC)+K=5/FBG.]NU0J\nME5)LVF:5(BHDG=A7JOOMU_,`,#-XDH9LWX6S:Q$<=O?T]&L:/0-DN5Z^OO?1\nMW;:,(<EY]E&6981C^XD9LY^^H2P33\"+,1,$?D6$HX\\R?L=\\V?9JL[U<9]E'\nMOUZNM\\OK;KA?-K^^\"W;>=5L:_2^OMV2^>K6]HS$P0L+KNT7_1MF\\UC\\#6\\!8\nM2OQ1ANZ(GZC]R?7_[_\\\\N_CA?O;5Q=FC)V?9DT=?/C[+OOGV\"?GYR9>/3Y[C\nMS)C&[,&1!_OFV^_/+IX8@/,0:I;]QZ/'_SS[/CLY1L>[0&-\"V4X(Q<X8E,VY\nM03B[N#B_N)]E-R]^OMRNEO>SQ7J]6L/\\LD\\,R\"?WLZ>7U\\?;[.9RO5EDGT#/\nM#F,0>1N\\]@=!;P(6L]4UHO+9\\OK%]DG#A)F=+'8O+K:9MO5KNZ7+]8C!_L\nM<O/L^8#4#$@J-=M7CO+]V>.SKYYDQ\\?9H^^SY\\O?%WDUV/RS!T>FY]8$@.?X\nMJ(#V>6';D>FZ=:M\">6V4[J^+X-KT%T[>1Z=\\&R]^F,S.VKG8?5JG6?+^6?9\nMOUZ<_UO%3+;,_O,?9Q=G\\`L8_,/3#(SX_GUC^,`DH/0P.9XQ-B5C[%TQMKI>\nM#/+UUU!>@'#K./K<,7,+6CS!9HQ)AF#[#+']8S4LW7`6@!#(-A!M(,U@1/\nM@+L-S28=^N5Z,2S:AY'Z+$YH5I%=16X965;-$IV4)7;W+(U0T&F+;83Z<>JI\nMM#.579RVV,7;#CM`:?M-A'+/Q!_)?U`^-.9PVF[.4S.S4`@.K',_$MV3+QD\nM9B8LX2CRM2G(#]T;GX;DD([-9MD$G^TB\\4MVR;X=W?[+-:`.;WURG'T;IN\nM/UZ'ZU/`F/?<@#V_;MUFA`D5KEZW1=!C!6JP\"AN9)8.&<H`NJ?#1DZG;3[-\nM\":4S:DXI:YF3X+Q<JOVD\"M>UXZP^WUE)(V=%DEG!ATAF!1\\DFE712#K<!]]_\nM5A/K:I]9V0\\]P:RNW\"H(IE2P+YWJ&\".CGO'?#B?42EU]W2F55(ZH^:4LN:<\nMLN:D\"J>W'6=U[ZX<@T3F<`RGE56U%U^5EA01>-9N:Z]9W7'P6^J6;WO8LG_\nMPU;6_]C[J_\\Q)'!0_V.N_B<.];]WT3KJ?RRH_['.A7;J?[7`FV\"`1<[H11[\nMX$Q>`VP9@V`FF:)@QKM-9T\\\\5^/S;7Q1T+?QM<&6H>^N-L@:M<',+3[AJI.N\nM.?&D[E%AXNZHX@Z:GG[M2;HV+5\\R$:7#T?,XRW8WK&XV,\\V^S#8;MR.MW`]\nM=>FQ=XAI2H]MLF\\K/0:23P6?BCD6ZFYEIUYVV)VSTRQ\\M#'44:(,C;-IG<WH\nMT+3/A-==BU7]O++WR.L(-;<5-=NU7\"DYT/&D%M=6Z+P+5D9IL/XV:7`5'^A\nM^B+M36QH'571=\\?F0/`<62[M47/(R8[5TWY6.JNG3<E%8HMJJ+57^F]3U5%9\nMX][;<^:O7S<6XXC9,))$?(>KM:VP:DP:J[8MM-8_U\"IPV%7%-?UAPL+KGV3^\nMG27:H?GWRZ!/#H[/IBRJ:]641WW-FC(I>K[X'JZ0N]!-MWEXA&RZ15-MV2\nM#L%$4X_E4D0R\"V8>3+PAE?$W8J.+SG]FH736K._6B]JBK6T\"2]+A1;8:3%HD\nMDV$DJ618$=KB1<&/4U6^_TRR<CGU&=R-A\\[<MSCKG2+%<1G]#9+]^\"D\"YP4\nM,(\\J0?-IUE$M(]J)\"/[LAC)U\"%>-.\"K16&<R6!E[F%J9G%.47T#K+/\"([?J5\nM80O^.NI;(#]$G8VASO:E3JT*#.WHIW4_HKZ<#=O=8ZX//YY=4&2-D/KYWG\nM60\\U]K$S@L7U]LW/3D'!EZ/4$VPCG5K[^&-K@IOEZP='\\,?;E.@\"-[G!\\#_#\nMW^(:^%I<.XXLTI#`,&HCNQCJ\"<6>U:ACW2,9YO?UMN38V'T^?QJ=;E5,X.^\nM6+YXN7UP9#^\"T547F=M[(8F]``)(B,2PR)XWQL6AS9IL_M_J^6SNWP$H'__\nM+Z,X//_/S?E_PI@\\[/^]B]:V_W?^S=?A(P#>.MIWIV+89&OOGE:2=FQK]2'R\nMO1`!YU[?\\P&]K';N`Y@W:N1!W?@+.3(K;?.87=ETVR[P;\\1&W;PKVW/SG3W\nM9-7!P-7.'1B-:TFF[/+4RCJKSSX6FFQ!@8IH\\Q]17#E/=G_+8ZN2I+@@&/\nMV<I7!5=^=Z<VMW)_)^`]9CU@M;<BZ7;<AD<WFV[A\\';;;2?9C1;6B!W\"$?S^\nM]6WY[=?U6'Y]Q7<$PS&_KM`;,YQRG+<,AEQ.;P),\\SC:2)4L_\\QS@9CC??O\nM?8STA]-N?XA8B<9N.,/7%^??^:4D'.K!(7?\\\\%J0_]W9$;\"A_(\\@&>1_U.9_\nMG!SROW?1NO,_%N=_':?`8MB=\\[]VQ!'Y7POBN/QOUU-H`U@^_U,:N0L)5^.3\nMP0AMI\\RPA:%=)_!N,D/V?R`S9.\\U,VP9_8/.#-OX_:`SPS:&/[3,L(7']Y<9\nMMC%S1YDA.V2&[[79_._%8O7K8KM^<T<)8'_^QR'O(\\G[7X@4Z)#_O8O6'?6?\nM+JZWB[5-&:#/?3/>;>/$E^?_Y1S71GKW8QSG7>`^P3G@^U7GA.0D^#+G.9WY\nM+Q0N1ZY&)U]\\D7U`3&V6K_/L9@49<H>HOOKFXJO'9XXQ`\"[Y:@WA)RA',QM3\nM/9\\D^D:[?T,H)PC-ZF_(LBX&6>\\6YX?+N%U`NQG_[OSQ#W\\__]9P;OK\\<!/\nM+?1IYH^[5K/I;NEIYZ[F>9_3J.5IAY\\C&88;<:3V!L.J',P8EE[:-_@>LK+\nMS<O5>OG?J^OMY=6)1<BS8ZNIX_OWK<G.HJPBG&^H57==8/-MMXL3/:W(F%H\nM:GYL=C-.5K\\O8`%X&DJS\\%RV9R'XHIN!%(+<8Q5:\".>/MA!5+=M'$W)T-/5\nMJ^LML&1G?S7?V#_9I[;#_+XT#K-9/-TN5]<5RX^_/_N[Y?@J#]FO;RF?O%RL\nM%R#>['J5K6X6:[B_7&?'GQY#-%]GVS<WBTUV?+59O#C.+J^?9<=V5L='?_EA\nM]2K[8WEUE2V66Z\"0O;S\\?0'WQ7`#;7`@T0>O[U:K-]DKS;FAOKR.EN\\OKE:\nM/EUNX49UL\\V/_@+TWY142O1GB^?+:X.^J)EQ!'P%Y/R[LXM'3\\XO.FXOENOM\nMFTA$5CB??FI%961XM=HL-MMA\\3A2@>9N-WVAZ-83'HQ7L?[[H]>8V%:>XJ@C\nM?';[HXWA.:R/;/93@],JP/=Q6L(Z)YZ:J$Q-:>5$PT1+0<?Q2G/X09R-BVG\nM(%(\\2J:[<!IJWUSG)R(7;40KP!%$0^U/1S30_F1$0^WW$C7=HSD-M#\\9IZ'V\nM)^,T\\GT$7!NZ(.@BMJM;WS>.:.3[DQ$-?7\\JHI'O3\\9IZ/M3$8U\\?RJBH?8+\nM_QRQ3=HALK@O!4.SGP!PCA!1#!$JN-)<29)C/J>\"<R40$2P6:OV>XD6>JZU\nMQ%PS0K@BR-,40!1AAI@DLU;M]Q,EAJC2\"A,B),?:$>58,6(I(C.6K7?2Y3/\nMD39E#$4)P4P+1Y,R9B60A,^:]=^/Z<,.&6:,8(4XX(RSRFGDA.%!1:S5NWW\nM$L5&4TAK<T08Q.!(FA>C@#@ITM4-<</W83VQ\"PK(.5E2;OV/XTPJ\\OW)B(:^\nM/Q71R/<GXS3T_:F(1K[_MD3!8#KOJRXWKG3PR^IU6?.`R_:B@?EAFC;V;GQD\nMV_&,<&4(,HHY^\"\\>=IAPF<#J-$S\\_>D$L\\E1D(JC)7@3.:2Q!V&G-`)4\"%4\nM`N7)@=D!?4HUXIH2R71=#CN$J`\"Q3U4^LH#3>?!6B:+&Y-%20^IBC5RC@^\nM(=+$'4!.0YPE\"KC`2F'>([C'C[+JDJ01$4>!Y_M[(C>AY2@S6L`32B>&1.\nMJ[M:XB_N<;/R#G>[OKS>7%U&-[B^J)C]DM[`561N`[348`?NM.HC\\KX0F3=R\nMO0C$5\"1S7[[L`+$5S+F1O;F\"?[,.M17Y5JI%,HR4X7I-A!IR10ZXB8!J9BQ\nMEXZ;5BHRXB8&P9\"IYG5T;P%!N?4LDO-P2C$5\\(I<.E[,I6CRXJB8OYURD68U\nM$)9M3`D(,KTNX&$N>1-$#D7.54`8\"_X;JH5)?IC&B9WY6K27/2)\"^T@0`W\nMQI6:$A`01J&<8.RU-I6B5BKVLU4N&+(R8S#5&M0\"0G-LK1?,%W=1@?&Q,QAW\nM35)>BKVLZ)R0OKWA5.?+^[(YT=Y_2B_'^7YHWQ_E/<W]>G5V1.-L&>J-Q[A\nMFBG<&9%2KTQ\"4S3%3.2&Y<(2(\"A`Q#8>Q\\E8^;(>[%QGC:>/\"7[T4FI@#45\nMAB+.?;C]PIK^8_NQ`V+V\"VT!$@;(N:&Y)$V>8DKUESBZY-BLY,9[NADW.;^=\nM'>Z1DXG9F-O![\"5KDY.G9#^ZY:0@QO@%I>B$+`(%-(M4&<:5M3\"E$OD5;\nM:\"PIV<^`TJYQXK,J3JQ7VYV#1(F3-=NH,)$$\"E?.:2L&?2A@!0'UDIZ0401%\nM\";BB799`C+,[(X<KRCO`/(T!:LCG,JP_Q):I\"&E$V6092IC8<M0TIP\"KLCU\nMN!AD&Y#3VN\"AYJH[V8+,%^[\\,<`867F)=Z+F;I%CQ4P0QQCF;RK5N\"2,6\nM/Z;FOL51F7G=0TSK$0A,$&1B8D1>\\$AR,9@VHF).<%:&!6MQ74.-&^[]1:?<\nM6(==)F#\"@9%>U7LP8YV\\QWHK:J)!;51@\\4'EWJY!I=X+/+&8#XO3=\"/7G!;#\nM9?#I\"SS[M>;6TD0MC/9U)P8?QK<[1CFN\"1L[E_M3:PC3\\UUFOOL2YA8YR?^\nMIL)]HWQ/XB%AS^7$',Y4J5+0\\P:K/_O0!@[M9EX`J/8JH$:W%T:0Y@(0\\O>\nMF#D]JK>QC0;'BCG9#J.D>`82\\&Y%(QJ,(IXH@IRI@F@C!3Q)!C1@L):RF`\nMIGG#@^!24N2,\"P)$Q3#'X&\\_\"GA`$`D_,-:V+DA23B65&DJJ5:$ST)1&,I$\nM8DZ0TEA;<\"X(E5IR1HE2S$D)4RU1`+^<::1U0P2B%&,N:G`\"$[P;)#CE)%\\\nM+&`8X`7O],@E#\"5+N^\"\"-8Q\"@QI3H2@P_MFL8Q-F$\"(:VSV&Z2@%.?&-A@1\nMB'.B.9>=$6$J.F33%+,T,0PRS8'@ZR])844$;W@WABBB76BE/F!\"F0!'2\nM,D(,89B^,%L(?H24<).5D1RG8(FYD<[,;Y\\6$:.,+CS%1CPG/D#,JZW92$\nM+=6Q!C5`^(0TB^M/LA>+WU;9D_/L>/7\\^?&#UC3\"OMK!'S?<E+7W:_<USUQ.\nM<7.Y?>F/$SUZ\\@]7B;=XMQ[PUDT,A8M\\D/EB_+59]W'#!XDP\"N)<GR<S:\nMN'929[E-&,TF]E[TQL.-G8?=7G'E+FKNM7_J@#M)X()RE1H^YVJT9%6<GFGT\nMYUM;M-AX,5&D,.#'N33)F7?L\\J0K3!G[HI`1=?\"#A6K(POZ`+1%<_CQP3K8T\nM7>S?\\.9+A64\":T\">+3?;GR^?/:NRX=``:NTX1(FZVK-5ZSVVG\"G+A.==NHR\nM@?O1S\\O!V4_F\"B0&O[;AL?1V&'?4/'XT]U5V:Y#0G/3--X9KCMMOPY&Z/P-U\nM$U-':VC[UU=7@]H&F'1PR\\\"PHJLY,V?8<'>73KB:;@#2>J<8'<I?U2N+(7\nM#I0Z2&740(/L_GABS@$8((:,=?W4!#E)0+[\\H;N+L[%:A]X0D+8^[TX\\WJ\nMZDUXD?U/=5?[U)PX7EYO:LW6P<N\"=U?,@O.0]675;BOM8X'3CM&#R:59((S\nM#2TQ\\;8\\9=FH7+G=]3:4YZTHU7F$T2CU:8/QHU1G\"4:CU\"<%1J.$$FLL%AV,\nM!1(;BQ)(;\"1*+&QHP02&XD22FPD2BBQQBHZ;&--E$&)C40))39VE$!B(U%\"\nMB8U$B266I!?#-C86)9+8)188N-&B20V\"B666\"?#EL8::0V\\?F+F#7D7KQ\nM[,Y\"-9#>/5;O$ZSWB=;[A.M]XO4^`7N?B+U/R-XG9N\\3M/>)VON$[7WB]CZ!\nM>Y_(O4_HWB=V[Q.\\]XG>^X3O?>+W/@%\\GPB^3PC?)X8/!G$?OGO+/!\"<7P0[\nM1\\T\"057N<9!9T)\"07?!P`'3[L)!#=!10'``N+N0X[)'@6%4D9V<D9(Z29:\nM62=I3K^]=59/VEM8(<G]'C/Q?U''?;HM16$K(EJ)B]93;H3:J\\Y/MP3'_,)\nMZC@AW?)OR\\W[GJM[M'+IR;M<VG/#$R7N#O$6LLCLWMV1W72$9-S[G1ZF9Q\nM_;,II=7SS6[LI,R<9JV>YF50[\\HN-X9,((2:BKX;FL<L+Z!8FQW[7@^ME=\nM;U,._UE]M8C=I2J^@JF8Y[!;YIK>AI]=\"PXMG;Q,?W#VW'9GP\"'R@_J#6\nMG&N%I\"`84PZ?.9[S^K\\/R9@[,35GNFH8BSS\"4A8(\\\"A&#&OX16JD%@5VBUM(\nM$$NA(F2S9>S6KZI+\"\"3S<$0IM0TB!9TC09\"B&FO)%>9FB`29<N;.<59#4J29\nMY:(F)A[=J-&D_\"_F3V3R37%A,,,!)6$VDG8,F]$$\\02XQ.I4F&:+?(\\DAR'\nMSMFLE+_6[C2:XG.!\"&>(2@6R5#DG\\UI>S/.EYK4X$&>Y4C66P&Z\"%18W&\\J0\nM?QKZ/.Q7FD:8`&=KU[+N4D3')\"JX8B@ECR\"B%SUBVU/:(:H$K%E#G+I:H!\nM@;PM/'-4(RJ\"0?:56@:$V:``A@FS$.#>:&M4)F>A^(@$;8$?=%$C!K^[PZ5\nM512AIUQ%2OF[VCK8%4($XFPTAB@<A[:OWEF)K%_Q8G,98!%D8S%#VQPDAO>\nM:4\"+:\"8B/%\"3@>%!%Z,`PT-+I?9`G7&XV@$92#]&5)HP<T`!3#,JI:&?1@<\nMJ\\#!>%S8G1T4$\">4,>4.6$<F+\\$O(V3C(B5(\"A.`A\\U&I@8T?&>@Y<_R<LQ\nMXOA#6^(/2N(/SEE;`KQ,,2BW\"?&<?!AS>!#T^!#&\\&'!;8/6M%4)7@V[H1L\nM\"@VQRC0QAW4C#NA`L\"$A79'S@L4Q1R(>C$V`>9TD0;=R)$\"!`R2&<_'7Y\nM0@V.M62<0WH@D'&AG@D20,3]A!MGO\"\"!($(!L@4`#$@8[;]8RH<[email protected]\nM1TD`8:%%&BO#YH9:Z[I;V>`0H4IE'BMC(8SUS)`:Q!^`\"::,['F;B#(#^6-M\nM91-B&K^/4`G#B?Q`Y1`E0VEA,(+FD=DJI<;@/('L(?B1O`/NVP04:(1^T/9\nM@^BQU2P.,2'V1X@^]..P#V)_SG0C]!/SO#!X%TP/AJ?@@6#_BP+I!-\\L`'D1\nM#%K&?QP0-0L`1'#1B/^!\"JP&\"C>)$)43H2+<,OH'8C09@@G^NC7X&_F/>!=6\nM>0NAS-WAX2[B3Y:S-S+DX+%79DY!!!V\"(]&7_(:PTF1W80>S[M67V-:PE)B,\nM.Q10N!G#4$!\\HT)D\"PY,U\\-'PJF4E&HT>7(4--EMHPUPR>>H:9P\"H=HIKU\nMMCN/#%#!J\\U9W^!9:,VYT'TI8L`S1\">I(FP)8JW6OR3[\"Y^9)B\"\\Z,%J!@E#\nM5V87/I$-.H_P$!>=25N`!RN=H/%#VB:$BIZ$+&17FX@>8D/T-H<W>W.MQ))9\nMFR6W9%$TM6+6L..#\"DU8!0;L#^GW)[]X(;UHL1Z<=MLZ\\PF?,+<V%3\\7+IJ\nMK)QAUA)`P@U1C&F=LS,C\"2#!E.,'W9$]0-V5;82`R\"8J00\\1K;-U>43X/#PG\nM7$;/S)=.VYXD!)#&:2-,[[.]\"4``+V7R)'[IN!UK>\\\"UF2^)<\".W';5J/UVN\nMGUZ9<FV>F20D%:7.]?^_946KG=QL\\)]Z(3+T6DM\\(=^I\\KWEGVFPED\"^KYJ\nMI\\GTEW!N)\\E!^CY_3-#3ZY^F?R%K.<^NG8YD@N%$^]]`8-@QVQ)POXL9>)H]\nM^T_Q:?FCC2.=/]I7III;C]/82LV/=@\\\"G28,#;S$V[Q#LF6RO:_U_&M&_7L]\nMVX4PT-XBTRGU.?A:3U@7(%4P[_R'D$OA?Z5Y#+\\1U`@?G(Q19#-GQ4\\'W\\-9\nMGP)YBLO3'E8TY:%,N[.Q-&]\\=&^\\M$^56=A9>3SS\\OKIHB4+!JB\\XUV535G\nM[K]Z>G[Q]=E%]N4/%=V\\9L>>\"'5O87SX,#IC$BOSUD':R[O8K-62+'OD,G\nMX=[M0__V61,@3Z,M8UBHI(`U$6Z,3%U`)EC&.:@S\\7!SVJC,N`NL;6:!4S&6\nM':D%[%+TL=AC06)I)0\"UE!FZ@6\\'ZO:UN9SH:A)4`A1'),4R[@NKSBL-NK%\nMG\"%8@QGA\"%8$JOJPZNWP?JQ$&M58B33Z)5]AJ3E6\"'(CR+08+,^\\#ZOF,-/&\nM/2']XQ)2:O\\?U>WBL))AS\"!&O9+OF%>%8\\5<(@A-0I\"<^^@X-$J1]Y@UB\nM&&Z>>Z418)$YH9@S4UN21/9JN?84(N;@1(Q24EJ0SCL6JL#!(\"\"0DFB)Y\"\nMWD/ZI%%C43PW^H6<A,/]\"D_G5;[Y#/3$`LL8P[I-J*@_$KUH=52Q[ZYN9Q\nM.&$\".%&B#ZOV2A`=K)F0KRO\"F6]'-:'9T\"_<P&Y$(0:\"!Q2]XY5610D4W,8\nMA1@M(R)23XFQBT34!B'R&(@BQ/RO!<ROM^,_VA'=JA'=JA'=JA'=JA'=JA\nM'=JA'=JA'=JA'=JA'=JA'=JA'=JA'=JA'=JA'=JA'=JAC6G_\"W\"0EE0`R```\n`\nend\n\n----------------------------------------------------------------------------\nDr. Andrew C.R. Martin University College London\nEMAIL: (Work) [email protected] (Home) [email protected]\nURL: http://www.biochem.ucl.ac.uk/~martin\nTel: (Work) +44(0)171 419 3890 (Home) +44(0)1372 275775\n", "msg_date": "Wed, 11 Feb 1998 16:53:50 GMT", "msg_from": "Andrew Martin <[email protected]>", "msg_from_op": true, "msg_subject": "Expected regression output from Irix5" }, { "msg_contents": "On Wed, 11 Feb 1998, Andrew Martin wrote:\n\n> \n> As promised here are expected regression outputs from Irix5 for\n> int2, int4, oidint2, oidint4 and geometry:\n\n\tWhat does 'uname -s' show for an Irix5 system?\n\n\n", "msg_date": "Wed, 11 Feb 1998 13:07:33 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Expected regression output from Irix5" } ]
[ { "msg_contents": "I posted this to questions a few days back but got no response. Can anyone\nhere help me with this? I know it should be straightforward, and that I'm\nmissing something obvious, but precisely what I'm missing is unknown to\nme.\n\nThanks!\n\n- t.\n\n-----------------------------------------------------------------------\n Tripp Lilley, Perspex Imageworks, Inc. ([email protected])\n-----------------------------------------------------------------------\n \"Give me a fast computer, for I intend to go in harm's way\"\n - updating John Paul Jones\n\n---------- Forwarded message ----------\nDate: Mon, 9 Feb 1998 12:42:48 -0500 (EST)\nFrom: Tripp Lilley <[email protected]>\nTo: [email protected]\nSubject: compiling extension functions?\n\nI'm trying to compile and use a 'C' extension function for PG 6.2.1 under\nRed Hat Linux 4.2 (stock distribution). I am executing the following\ncommands to compile my shared library and install my function, but I get\n\"Unable to resolve symbol\" when I try to execute the function.\n\n> gcc -fpic -shared salmon.c -o salmon.so -L/usr/lib -lpq \n-I/usr/include/postgres\n> touch salmon.sql\n> psql -d nim_0_01 -f salmon.sql\ndrop function salmon (int4);\nDROP\n\ncreate function\n salmon (int4)\n returns\n char16\n as\n '/home/tlilley/projects/perspex/scratch/nim/salmon.so'\n language\n 'c';\nCREATE\nEOF\n\n\n> psql -d nim_0_01\n\nnim_0_01=> select salmon( 42 );\nWARN:Load of file /home/tlilley/projects/perspex/scratch/nim/salmon.so\nfailed: Unable to resolve symbol\nnim_0_01=> \n\n\nAnd here's the source for my function:\n\n#include <postgres.h>\n#include <utils/palloc.h>\n\n\nchar16 * salmon (int4 value)\n{\n char16 * buffer = (char16 ) palloc( sizeof( char16 ) );\n\n\n memset( (void ) buffer, 0, sizeof( char16 ) );\n (void) strncpy( buffer, \"hello\", 16 );\n return( buffer );\n}\n\n\nAny help (especially 'working' compile command-lines) would be great!\n\nThanks...\n\n- t.\n\n-----------------------------------------------------------------------\n Tripp Lilley, Perspex Imageworks, Inc. ([email protected])\n-----------------------------------------------------------------------\n \"Give me a fast computer, for I intend to go in harm's way\"\n - updating John Paul Jones\n\n\n", "msg_date": "Wed, 11 Feb 1998 14:41:22 -0500 (EST)", "msg_from": "Tripp Lilley <[email protected]>", "msg_from_op": true, "msg_subject": "compiling extension functions? (fwd)" }, { "msg_contents": "> I'm trying to compile and use a 'C' extension function for PG 6.2.1 under\n> Red Hat Linux 4.2 (stock distribution). I am executing the following\n> commands to compile my shared library and install my function, but I get\n> \"Unable to resolve symbol\" when I try to execute the function.\n>\n> > gcc -fpic -shared salmon.c -o salmon.so -L/usr/lib -lpq\n> -I/usr/include/postgres\n> > touch salmon.sql\n> > psql -d nim_0_01 -f salmon.sql\n> drop function salmon (int4);\n> DROP\n>\n> create function\n> salmon (int4)\n> returns\n> char16\n> as\n> '/home/tlilley/projects/perspex/scratch/nim/salmon.so'\n> language\n> 'c';\n> CREATE\n> EOF\n>\n> > psql -d nim_0_01\n>\n> nim_0_01=> select salmon( 42 );\n> WARN:Load of file /home/tlilley/projects/perspex/scratch/nim/salmon.so\n> failed: Unable to resolve symbol\n> nim_0_01=>\n>\n> And here's the source for my function:\n>\n> #include <postgres.h>\n> #include <utils/palloc.h>\n>\n> char16 * salmon (int4 value)\n> {\n> char16 * buffer = (char16 ) palloc( sizeof( char16 ) );\n>\n> memset( (void ) buffer, 0, sizeof( char16 ) );\n> (void) strncpy( buffer, \"hello\", 16 );\n> return( buffer );\n> }\n>\n> Any help (especially 'working' compile command-lines) would be great!\n\nLook in contrib/int8/ for code and Makefile which were developed on RH Linux\n4.2. ld may not be finding the libraries it needs. You may need to set\nLD_LIBRARY (or something like that, don't remember the name exactly) and/or\nmodify /etc/ld.so.conf then run ldconfig to get the load library database\npointing at the Postgres libraries.\n\nGood luck.\n\n - Tom\n\n", "msg_date": "Thu, 12 Feb 1998 02:31:28 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] compiling extension functions? (fwd)" } ]
[ { "msg_contents": "Bruce Momjian wrote:\n> > \n> > I ran my performance tests some more times and it seems the numbers are not\n> > really comparable. When I run PostgreSQL without -F I get a sync after every\n> > insert. With -F I get no sync at all as all inserts fit well into the\n> > buffer. However, Oracle in comparison does sync. Simply hearing the disk\n> > access it seems as if they sync every two or three seconds.\n> > \n> > Does anyone know a way to really check both DBMSs?\n> \n> Many dbms's do buffered logging, that is they sync after the buffer gets\n> full or after a minute or so. We have the logic to add buffered logging\n> to PostgreSQL and will be doing it later. Right now, we only have\n> non-buffered logging, and no logging.\n\nWill there be (or is there) a provision for using raw disk devices,\nrather than going through the UNIX filesystem? This might be able to\nreduce some of the overhead. Also, if a transaction log were appended\nserially to a seperate raw disk, there would be verry little seeking\nneeded and this (the log) could easily be fsync'd for each transaction\nwhile the data itself is buffered.\n\nOf course not everyone has two raw disks to devote to a database (or\neven one raw disk), but for those who do, this could provide speed and\nreliability.\n\nOcie Mitchell\n\n", "msg_date": "Wed, 11 Feb 1998 12:00:00 -0800 (PST)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Problem with the numbers I reported yesterday (fwd)" } ]
[ { "msg_contents": "> Just to add in here...Oracle:\n> \n> SQL> select sum(population) from nations;\n> \n> SUM(POPULATION)\n> ---------------\n> 2771277000\n> \n> SQL> \n\nHere is Sybase. Note, I declared the population field as int.\n\n1> SELECT * FROM t1\n2> go\n name population \n -------------------- ----------- \n RUSSIA 281170000 \n INDIA 766140000 \n CHINA 1072220000 \n JAPAN 129947000 \n CANADA 25610000 \n U.S.A. 242080000 \n MEXICO 81160000 \n BRAZIL 141450000 \n ARGENTINA 31500000 \n \n(9 rows affected)\n1> select sum(population) from t1\n2> go\nArithmetic overflow occurred.\n\nBut with population defined as numeric(20,0)\n\n1> select * from t2\n2> go\n name population \n -------------------- ----------------------- \n JAPAN 129947000 \n CANADA 25610000 \n U.S.A. 242080000 \n MEXICO 81160000 \n BRAZIL 141450000 \n ARGENTINA 31500000 \n INDIA 766140000 \n CHINA 1072220000 \n RUSSIA 281170000 \n \n(9 rows affected)\n1> select sum(population) from t2\n2> go\n \n ----------------------------------------- \n 2771277000 \n\n\nI sort of like this behavior. This way, a valid sum over fields of\ntype X can always be stored in a field of type X.\n\nOcie Mitchell\n\n\n", "msg_date": "Wed, 11 Feb 1998 12:01:50 -0800 (PST)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "sum(population) under Sybase" }, { "msg_contents": "> Here is Sybase. Note, I declared the population field as int.\n> 1> select sum(population) from t1\n> 2> go\n> Arithmetic overflow occurred.\n>\n> But with population defined as numeric(20,0)\n> 1> select * from t2\n> 2> go\n> name population\n> -------------------- -----------------------\n> JAPAN 129947000\n> CANADA 25610000\n> U.S.A. 242080000\n> MEXICO 81160000\n> BRAZIL 141450000\n> ARGENTINA 31500000\n> INDIA 766140000\n> CHINA 1072220000\n> RUSSIA 281170000\n>\n> (9 rows affected)\n> 1> select sum(population) from t2\n> 2> go\n>\n> -----------------------------------------\n> 2771277000\n>\n> I sort of like this behavior. This way, a valid sum over fields of\n> type X can always be stored in a field of type X.\n\nHow is that? numeric(20,0) guarantees that at least 20 digits can be\nstored. However, the SQL92 standard allows but does not require that\nmore than 20 digits are handled. So the standard does not preclude\noverflow problems. In fact, many implementations will allocate a fixed\namount of storage for the numeric field, so would have trouble coping\nwith summation overflows.\n\n - Tom\n\n", "msg_date": "Thu, 12 Feb 1998 02:44:55 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] sum(population) under Sybase" } ]
[ { "msg_contents": "To: [email protected] \n\n Need a re-launch? Launchmaster will re-launch \n your web site to 50 search engines and indexes for $95. \n Satisfaction guaranteed! \n\nMillions of people use search engines to find web sites every day. But if\n your site isn't listed, no one will find it.\n\nLaunchmaster will re-launch your site by submitting it to the top 50 search \nengines for $95.\n\nHere's how to do it:\n\n1. Complete the form below and return it to us.\n\n2. We'll post your site to 50 search engines within two business days, and \nwe'll send you a complete report when we're finished.\n\n3. Pay nothing now. We'll send you an invoice after we've completed the\nposting. Your satisfaction is guaranteed!\n\nThese are permanent listings. The $95 is a one-time fee.\n\nWHICH SEARCH ENGINES?\n\nHere's the list:\n\nExcite, Alta Vista, Infoseek, Galaxy, HotBot, Lycos, Magellan, Open\nText Web Index, Web Crawler, BizWeb, New Riders WWW Yellow Pages, \nLinkMonster, True North, Northern Light, YelloWWWeb, The Weekly Bookmark,\nScrub The Web, Seven Wonders, Sserv, Starting Point, Web 100, Web Walker,\nNet-Announce, New Page List, One World Plaza, PageHost A-Z, PeekABoo, Project\nCool, Where2Go, World Wide Business Yellow Pages, Wow! Web Wonders!, WWW \nWorm, JumpCity, The Galactic Galaxy, TurnPike, Unlock:The Information \nExchange, Your WebScout, Manufacturers Information Network, Net Happenings,\nNet Mall, Web World Internet Directory, InfoSpace, Jayde Online Directory,\nBC Internet, BizCardz Business Directory, WebVenture, Hotlist, What's New,\nWhatUSeek, JumpLink, Linkcentre Directory.\n\nORDER FORM\n\nHit the REPLY button on your e-mail program and fill out the following information.\n (This information will be posted to the search engines/indexes):\n\n\nContact name: \nCompany Name:\nAddress:\nCity: State/Prov: Zip/Postal Code: \nTelephone: \nFax: \nEmail address: \n\nContact e-mail address (in case we have questions about this order): \n\nURL: http://\nSite Title: \nDescription (250 characters): \n\nKey words (250 characters, in descending order of importance):\n\nIf billing a different address, please complete the following:\n\nAddressee: \nCompany Name:\nAddress:\nCity: State/Prov: Zip/Postal Code: \nTelephone: \nFax: \nEmail address: \n\nTERMS\n\nTerms are net 15 days from date of invoice.\n\n______________________________________________________________________\nLaunchmaster, Inc.\n12 Godfrey Place\nWilton CT 06897\nPhone: (203) 834-5823 \nFax: (203) 834-1897\nE-mail: [email protected]\n", "msg_date": "Wed, 11 Feb 1998 19:22:33 -0500 (EST)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Your Web Site Re-Launch" } ]
[ { "msg_contents": "Hello,\n\nI have made patches that allow PostgreSQL to handle multi-byte\ncharacters. The patches makes most commands including \"regexp\" and\n\"like\" 8 bit, multi-byte aware. You have a choice of one of followings\nfor coding schema at the compile time:\n\n\t1. single byte 7 bit ASCII/8 bit characters(mostly used in Europ)\n\t2. EUC (extend Unix Code) for Japanese, Chinese and Korean\n\t3. UNICODE (UTF-8)\n\t4. Mule internal code\n\nI only modified include/regexp, backend/regex and\nbackend/utils/adt/like.c. So the patches would not affect other parts\nof PostgreSQL, I believe.\nNote that I changed regexp coming with PostgreSQL rather than\nborrowing GNU's regex to avoid license issues.\n\nShould I post the patches for 6.3b now or wait after v6.3?\n--\nTatsuo Ishii\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 11:10:58 +0900", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "MB patch" }, { "msg_contents": "On Thu, 12 Feb 1998 [email protected] wrote:\n\n> Hello,\n> \n> I have made patches that allow PostgreSQL to handle multi-byte\n> characters. The patches makes most commands including \"regexp\" and\n> \"like\" 8 bit, multi-byte aware. You have a choice of one of followings\n> for coding schema at the compile time:\n> \n> \t1. single byte 7 bit ASCII/8 bit characters(mostly used in Europ)\n> \t2. EUC (extend Unix Code) for Japanese, Chinese and Korean\n> \t3. UNICODE (UTF-8)\n> \t4. Mule internal code\n> \n> I only modified include/regexp, backend/regex and\n> backend/utils/adt/like.c. So the patches would not affect other parts\n> of PostgreSQL, I believe.\n> Note that I changed regexp coming with PostgreSQL rather than\n> borrowing GNU's regex to avoid license issues.\n> \n> Should I post the patches for 6.3b now or wait after v6.3?\n\n\tPlease hold these until post v6.3 ... only two weeks left for\nrelease, and with the currently reproducable bug that has been found,\nthrowing in new code isn't a good idea :(\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 13 Feb 1998 00:45:12 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] MB patch" }, { "msg_contents": "On Fri, 13 Feb 1998, The Hermit Hacker wrote:\n\n> On Thu, 12 Feb 1998 [email protected] wrote:\n> \n> > \t1. single byte 7 bit ASCII/8 bit characters(mostly used in Europ)\n> > \t2. EUC (extend Unix Code) for Japanese, Chinese and Korean\n> > \t3. UNICODE (UTF-8)\n> > \t4. Mule internal code\n> > \n> > I only modified include/regexp, backend/regex and\n> > backend/utils/adt/like.c. So the patches would not affect other parts\n> > of PostgreSQL, I believe.\n> > Note that I changed regexp coming with PostgreSQL rather than\n> > borrowing GNU's regex to avoid license issues.\n> > \n> > Should I post the patches for 6.3b now or wait after v6.3?\n> \n> \tPlease hold these until post v6.3 ... only two weeks left for\n> release, and with the currently reproducable bug that has been found,\n> throwing in new code isn't a good idea :(\n\nAnother reason is that the UTF code may be handy for the JDBC driver -\noccasionally I get questions about how Unicode is handled. Currently I'm\nsaying \"It isn't yet\".\n\nHowever, this is something I wont be able to get round to until 6.3 is\nout.\n\n-- \nPeter T Mount [email protected] or [email protected]\nMain Homepage: http://www.demon.co.uk/finder\nWork Homepage: http://www.maidstone.gov.uk Work EMail: [email protected]\n\n", "msg_date": "Fri, 13 Feb 1998 19:54:00 +0000 (GMT)", "msg_from": "Peter T Mount <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] MB patch" } ]
[ { "msg_contents": "\nRemember that heap_modifytuple error that I was reporting for\ni386_solaris? Well, I hit it with my server in Toronto today trying to\nget a v6.3 server up for Julie to test ODBC against...which blew my mind,\nsince I hadn't experienced any problems here at home.\n\nAfter investigating further, it appears the problem was a versioning\none...when I was running 'initdb', it was taking 'postgres' out of my\npath, which is different (v6.2.1 vs v6.3) then the one that I wanted.\n\nOnce I set the path correctly, I got a clean initdb from i386_solaris as\nwell...will try regression tests tomorrow.\n\nWith that in mind, would it make more sense to have a --pgbase vs --pglib\noption to initdb? So that we'd use:\n\n\t$pgbase/bin/postgres with $pgbase/lib/..?\n\n\t--pgdata is still seperate, for obvious reasons. The bin/lib\ndirectories should be central...the data directory can be pretty much\ncreated anywhere.\n\n\tComments?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 12 Feb 1998 00:38:39 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "i386_solaris port...problem found..." } ]
[ { "msg_contents": "Hello,\n\nconfigure on PPC/Linux fails because of an incorrect template\nfile(configure tries to envoke gcc with -m486 option!). Following\npatches should fix the problem.\n\nP.S.\tI have tested on the 2/11 snapshot.\n--\nTatsuo Ishii\[email protected]\n--------------------------------------------------------------\ndiff -c -N template.orig/.similar template/.similar\n* template.orig/.similar\tTue Feb 3 17:00:32 1998\n--- template/.similar\tThu Feb 12 15:14:13 1998\n***********\n* 17,22 **\n powerpc-ibm-aix3.2.5=aix_325\n powerpc-ibm-aix4.1.4.0=aix_41\n powerpc-ibm-aix4.1.5.0=aix_41\n! powerpc-unknown-linux-gnu=linux-elf\n sparc-sun-solaris=sparc_solaris-gcc\n sparc-unknown-linux-gnu=linux-elf-sparc\n--- 17,22 ----\n powerpc-ibm-aix3.2.5=aix_325\n powerpc-ibm-aix4.1.4.0=aix_41\n powerpc-ibm-aix4.1.5.0=aix_41\n! powerpc-unknown-linux-gnu=linux-elf-ppc\n sparc-sun-solaris=sparc_solaris-gcc\n sparc-unknown-linux-gnu=linux-elf-sparc\ndiff -c -N template.orig/linux-elf-ppc template/linux-elf-ppc\n* template.orig/linux-elf-ppc\tThu Jan 1 09:00:00 1970\n--- template/linux-elf-ppc\tThu Feb 12 15:14:41 1998\n*************\n* 0 ****\n--- 1,10 ----\n+ AROPT:crs\n+ CFLAGS:-O2\n+ SHARED_LIB:-fpic\n+ ALL:\n+ SRCH_INC:/usr/include/ncurses /usr/include/readline\n+ SRCH_LIB:\n+ USE_LOCALE:no\n+ DLSUFFIX:.so\n+ YFLAGS:-d\n+ YACC:bison -y\n", "msg_date": "Thu, 12 Feb 1998 16:36:39 +0900", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "configure fails on PPC/Linux" } ]
[ { "msg_contents": "Bruce Momjian writes:\n> Strange. No changes since before the weekend.\n\nI'm not sure when I tried the cursor access for the last time. I definitely\ntried it on Sunday, but I believe I never tried it after the last major\nchange that caused me to do a initdb.\n\nMichael\n\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Thu, 12 Feb 1998 09:05:12 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] cursor access doesn't work anmore" } ]
[ { "msg_contents": "IMO this is a rather serious bug:\n\n{\nexec sql begin declare section;\n\tlong i;\nexec sql end declare section;\n\t...\n\t{\n\t\texec sql begin declare section;\n \t\tshort i;\n\t\texec sql end declare section;\n\t\t...\n\t}\n\t...\n\tsome DB access into i;\n\nThe code produced by ecpg now lists i with size 2 bytes, i.e. it still\nthinks i is a short.\n\nI believe I have found a way to work around this and disregard variable\ndefinitions as soon as they are no longer valid.\n\nTo my big surprise though I noticed that ORACLE's proc suffers from the same\nbug. I hope it's okay with you that I don't care about compatibility and fix\nthis. :-)\n\nMichael\n\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Thu, 12 Feb 1998 09:15:58 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "preprocessor bug?" }, { "msg_contents": "> To my big surprise though I noticed that ORACLE's proc suffers from the same\n> bug. I hope it's okay with you that I don't care about compatibility and fix\n> this. :-)\n\nFix it, please.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 09:26:41 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] preprocessor bug?" } ]
[ { "msg_contents": "subscribe\n\n", "msg_date": "Thu, 12 Feb 1998 16:30:56 +0800", "msg_from": "\"Bu Dongbo\" <[email protected]>", "msg_from_op": true, "msg_subject": "subscribe" } ]
[ { "msg_contents": "Hi,\n\nAccoding to the HISTORY file it is now possible to set the max-tuple\nsize for a database. (I'm using yesterdays cvs snapshot on Linux).\n\nCould anyone help me find out how to configure this value?\n\nThanks Maurice.\n\n", "msg_date": "Thu, 12 Feb 1998 10:50:29 +0100", "msg_from": "\"Maurice Gittens\" <[email protected]>", "msg_from_op": true, "msg_subject": "Installation data blocksize and max tuple size configuration" }, { "msg_contents": "> \n> Hi,\n> \n> Accoding to the HISTORY file it is now possible to set the max-tuple\n> size for a database. (I'm using yesterdays cvs snapshot on Linux).\n> \n> Could anyone help me find out how to configure this value?\n> \n\nActually, Darren King is still working on it.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 09:27:10 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Installation data blocksize and max tuple size\n\tconfiguration" } ]
[ { "msg_contents": "> On Wed, 11 Feb 1998, Andrew Martin wrote:\n> \n> > \n> > As promised here are expected regression outputs from Irix5 for\n> > int2, int4, oidint2, oidint4 and geometry:\n> \n> \tWhat does 'uname -s' show for an Irix5 system?\n> \n> \nIrix5.x or 6.x gives: IRIX\nIrix6.x (on an Origin2000) gives: IRIX64\n(i.e. a 64bit machine)\n\nAndrew\n\n----------------------------------------------------------------------------\nDr. Andrew C.R. Martin University College London\nEMAIL: (Work) [email protected] (Home) [email protected]\nURL: http://www.biochem.ucl.ac.uk/~martin\nTel: (Work) +44(0)171 419 3890 (Home) +44(0)1372 275775\n", "msg_date": "Thu, 12 Feb 1998 12:35:13 GMT", "msg_from": "Andrew Martin <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Expected regression output from Irix5" } ]
[ { "msg_contents": "> Hi,\n> \n> Accoding to the HISTORY file it is now possible to set the max-tuple\n> size for a database. (I'm using yesterdays cvs snapshot on Linux).\n> \n> Could anyone help me find out how to configure this value?\n> \n> Thanks Maurice.\n\nI didn't get this done in time to make the 6.3 beta.\n\ndarrenk\n", "msg_date": "Thu, 12 Feb 1998 08:10:14 -0500", "msg_from": "[email protected] (Darren King)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Installation data blocksize and max tuple size\n\tconfiguration" } ]
[ { "msg_contents": "Fixed with my lates upload, abour half an hour ago.\n\nMichael\n\n--\nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n\n> -----Original Message-----\n> From:\tBruce Momjian [SMTP:[email protected]]\n> Sent:\tThursday, February 12, 1998 3:27 PM\n> To:\[email protected]\n> Cc:\[email protected]\n> Subject:\tRe: [HACKERS] preprocessor bug?\n> \n> > To my big surprise though I noticed that ORACLE's proc suffers from\n> the same\n> > bug. I hope it's okay with you that I don't care about compatibility\n> and fix\n> > this. :-)\n> \n> Fix it, please.\n> \n> -- \n> Bruce Momjian\n> [email protected]\n", "msg_date": "Thu, 12 Feb 1998 15:34:17 +0100", "msg_from": "\"Meskes, Michael\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] preprocessor bug?" } ]
[ { "msg_contents": "\nunsubscribe\n", "msg_date": "Thu, 12 Feb 1998 12:25:07 -0500", "msg_from": "\"Nguyen, Thuan\" <[email protected]>", "msg_from_op": true, "msg_subject": "None" } ]
[ { "msg_contents": "[similar report submitted previously, but this is more complete]\n\nThere is something that looks like shared memory corruption going on,\nwhich I first noticed by accident the other day, in the 1998-02-09\nsnapshot. It's still there today, with the 1998-02-12 one, and looks\nlike the following on my Sun SS2 under NetBSD/sparc 1.3 (I've created\na simple test case here, for easy testing elsewhere):\n\nFirst, I run initdb, start a postmaster, create a user 'tih', stop the\npostmaster, restart the postmaster with '-d', thus:\n\n barsoom:postgres> postmaster -i -d\n FindBackend: searching PATH ...\n FindBackend: found \"/usr/local/pgsql/bin/postgres\" using PATH\n\nNext, I create a database 'words', thus:\n\n barsoom:tih> createdb words\n barsoom:tih>\n\nThe postmaster says:\n\n postmaster: BackendStartup: pid 6542 user tih db template1 socket 5\n postmaster: reaping dead processes...\n postmaster: CleanupProc: pid 6542 exited with status 0\n\nI fire up psql, thus:\n\n barsoom:tih> psql words\n words=>\n\nThe postmaster goes:\n\n postmaster: BackendStartup: pid 6549 user tih db words socket 5\n\nIn psql, I then do the following:\n\n words=> create table dictionary (entry char(64));\n CREATE\n words=> create unique index dict_by_entry on dictionary (entry);\n CREATE\n words=> copy dictionary from '/usr/share/dict/words';\n\nThe postmaster generates no output at this, and the copy starts as it\nshould. There is much disk activity. Next, while this is running,in\nanother terminal window, as the same user 'tih', I do:\n\n barsoom:tih> createdb \n Connection to database 'template1' failed.\n PQexec() -- There is no connection to the backend.\n createdb: database creation failed on tih.\n barsoom:tih>\n\nWhen this happens, the postmaster generates the following output:\n\n postmaster: BackendStartup: pid 6560 user tih db template1 socket 5\n ERROR: cannot write block 171 of dict_by_entry [words] blind\n postmaster: reaping dead processes...\n postmaster: CleanupProc: pid 6560 exited with status 0\n\nLooking at processes running on the system at this time, I see:\n\n 6549 p6 R+ 2:01.88 /usr/local/pgsql/bin/postgres -p -Q -P5 -v 65536 words\n\nThis is the backend doing the copy. It is spinning furiously, eating\nCPU like there was no tomorrow -- but there is no more disk activity.\nThe terminal window where I initiated the copy operation looks as\nthough it were proceeding normally. So now I attempt to perform the\ndatabase creation again, thus (in the second terminal):\n\n barsoom:tih> createdb\n\nNothing happens -- it just hangs there. The postmaster says:\n\n postmaster: BackendStartup: pid 6595 user tih db template1 socket 5\n\nLooking with ps again, I can see that this backend is now also running\nwild, sharing the CPU half and half with the one with PID 6549...\n\nNote that I'm trying to create a different database when it breaks;\nthe only possible interaction is through the shared memory that I\nunderstand is maintained by the postmaster on behalf of the backends.\nAs for seeing this on other platforms, I certainly hope it's\nrepeatable elsewhere, but it's not unreasonable to assume that it\ncould cause different symptoms on other platforms, including quiet\ndata corruption...\n\nThe whole thing is completely repeatable here -- any ideas can be\nverified quickly and easily -- and with enthusiasm. :-)\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n\n", "msg_date": "Thu, 12 Feb 1998 19:27:53 +0100 (CET)", "msg_from": "Tom I Helbekkmo <[email protected]>", "msg_from_op": true, "msg_subject": "Shared memory corruption?" }, { "msg_contents": "I saw this here too. I ran the regression tests, and while doing it,\ntried to create a database. No idea on a cause.\n\n> \n> [similar report submitted previously, but this is more complete]\n> \n> There is something that looks like shared memory corruption going on,\n> which I first noticed by accident the other day, in the 1998-02-09\n> snapshot. It's still there today, with the 1998-02-12 one, and looks\n> like the following on my Sun SS2 under NetBSD/sparc 1.3 (I've created\n> a simple test case here, for easy testing elsewhere):\n> \n> First, I run initdb, start a postmaster, create a user 'tih', stop the\n> postmaster, restart the postmaster with '-d', thus:\n> \n> barsoom:postgres> postmaster -i -d\n> FindBackend: searching PATH ...\n> FindBackend: found \"/usr/local/pgsql/bin/postgres\" using PATH\n> \n> Next, I create a database 'words', thus:\n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 14:39:26 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Shared memory corruption?" }, { "msg_contents": "Vadim, I may need your help on this one. I can reproduce it by runinng\nthe regression test, and doing a shell 'while' loop that continuously\ncreates databases:\n\n\twhile :\n\tdo\n\t\tsh -c 'createdb $$'\n\tdone\n\nI get the errors too. I have no idea on a cause. I would hope it is\nnot the new deadlock code, or locking fixes I did. I think the message\ncomes from smgrblindwrt. Is it possible our new speedups are causing\nit?\n\n\n\n> \n> [similar report submitted previously, but this is more complete]\n> \n> There is something that looks like shared memory corruption going on,\n> which I first noticed by accident the other day, in the 1998-02-09\n> snapshot. It's still there today, with the 1998-02-12 one, and looks\n> like the following on my Sun SS2 under NetBSD/sparc 1.3 (I've created\n> a simple test case here, for easy testing elsewhere):\n> \n> First, I run initdb, start a postmaster, create a user 'tih', stop the\n> postmaster, restart the postmaster with '-d', thus:\n> \n> barsoom:postgres> postmaster -i -d\n> FindBackend: searching PATH ...\n> FindBackend: found \"/usr/local/pgsql/bin/postgres\" using PATH\n> \n> Next, I create a database 'words', thus:\n> \n> barsoom:tih> createdb words\n> barsoom:tih>\n> \n> The postmaster says:\n> \n> postmaster: BackendStartup: pid 6542 user tih db template1 socket 5\n> postmaster: reaping dead processes...\n> postmaster: CleanupProc: pid 6542 exited with status 0\n> \n> I fire up psql, thus:\n> \n> barsoom:tih> psql words\n> words=>\n> \n> The postmaster goes:\n> \n> postmaster: BackendStartup: pid 6549 user tih db words socket 5\n> \n> In psql, I then do the following:\n> \n> words=> create table dictionary (entry char(64));\n> CREATE\n> words=> create unique index dict_by_entry on dictionary (entry);\n> CREATE\n> words=> copy dictionary from '/usr/share/dict/words';\n> \n> The postmaster generates no output at this, and the copy starts as it\n> should. There is much disk activity. Next, while this is running,in\n> another terminal window, as the same user 'tih', I do:\n> \n> barsoom:tih> createdb \n> Connection to database 'template1' failed.\n> PQexec() -- There is no connection to the backend.\n> createdb: database creation failed on tih.\n> barsoom:tih>\n> \n> When this happens, the postmaster generates the following output:\n> \n> postmaster: BackendStartup: pid 6560 user tih db template1 socket 5\n> ERROR: cannot write block 171 of dict_by_entry [words] blind\n> postmaster: reaping dead processes...\n> postmaster: CleanupProc: pid 6560 exited with status 0\n> \n> Looking at processes running on the system at this time, I see:\n> \n> 6549 p6 R+ 2:01.88 /usr/local/pgsql/bin/postgres -p -Q -P5 -v 65536 words\n> \n> This is the backend doing the copy. It is spinning furiously, eating\n> CPU like there was no tomorrow -- but there is no more disk activity.\n> The terminal window where I initiated the copy operation looks as\n> though it were proceeding normally. So now I attempt to perform the\n> database creation again, thus (in the second terminal):\n> \n> barsoom:tih> createdb\n> \n> Nothing happens -- it just hangs there. The postmaster says:\n> \n> postmaster: BackendStartup: pid 6595 user tih db template1 socket 5\n> \n> Looking with ps again, I can see that this backend is now also running\n> wild, sharing the CPU half and half with the one with PID 6549...\n> \n> Note that I'm trying to create a different database when it breaks;\n> the only possible interaction is through the shared memory that I\n> understand is maintained by the postmaster on behalf of the backends.\n> As for seeing this on other platforms, I certainly hope it's\n> repeatable elsewhere, but it's not unreasonable to assume that it\n> could cause different symptoms on other platforms, including quiet\n> data corruption...\n> \n> The whole thing is completely repeatable here -- any ideas can be\n> verified quickly and easily -- and with enthusiasm. :-)\n> \n> -tih\n> -- \n> Popularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n> \n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 15:09:24 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Shared memory corruption?" }, { "msg_contents": "> \n> Bruce Momjian wrote:\n> > \n> > Vadim, I may need your help on this one. I can reproduce it by runinng\n> > the regression test, and doing a shell 'while' loop that continuously\n> > creates databases:\n> > \n> > while :\n> > do\n> > sh -c 'createdb $$'\n> > done\n> > \n> > I get the errors too. I have no idea on a cause. I would hope it is\n> > not the new deadlock code, or locking fixes I did. I think the message\n> > comes from smgrblindwrt. Is it possible our new speedups are causing\n> > it?\n> \n> I'll try to deal with this in the next week.\n> I'm going to update CVS with subselect support right now\n> and I'll try to fix bugs after this.\n\nGreat. Thanks.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 21:29:37 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Shared memory corruption?" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> Vadim, I may need your help on this one. I can reproduce it by runinng\n> the regression test, and doing a shell 'while' loop that continuously\n> creates databases:\n> \n> while :\n> do\n> sh -c 'createdb $$'\n> done\n> \n> I get the errors too. I have no idea on a cause. I would hope it is\n> not the new deadlock code, or locking fixes I did. I think the message\n> comes from smgrblindwrt. Is it possible our new speedups are causing\n> it?\n\nI'll try to deal with this in the next week.\nI'm going to update CVS with subselect support right now\nand I'll try to fix bugs after this.\n\nVadim\n", "msg_date": "Fri, 13 Feb 1998 09:30:18 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Shared memory corruption?" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> Vadim, I may need your help on this one. I can reproduce it by runinng\n> the regression test, and doing a shell 'while' loop that continuously\n> creates databases:\n> \n> while :\n> do\n> sh -c 'createdb $$'\n> done\n> \n> I get the errors too. I have no idea on a cause. I would hope it is\n> not the new deadlock code, or locking fixes I did. I think the message\n> comes from smgrblindwrt. Is it possible our new speedups are causing\n> it?\n\nI can reproduce it. Keep looking...\nBTW, did you compile without --enable-cassert ?\n(Should be ON by default in beta-s...)\nI got some interest assertion from BufferAlloc, without CASSERT you should get\ndead spinlock from there.\n\nVadim\n", "msg_date": "Thu, 19 Feb 1998 15:02:37 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Shared memory corruption?" }, { "msg_contents": "> \n> Bruce Momjian wrote:\n> > \n> > Vadim, I may need your help on this one. I can reproduce it by runinng\n> > the regression test, and doing a shell 'while' loop that continuously\n> > creates databases:\n> > \n> > while :\n> > do\n> > sh -c 'createdb $$'\n> > done\n> > \n> > I get the errors too. I have no idea on a cause. I would hope it is\n> > not the new deadlock code, or locking fixes I did. I think the message\n> > comes from smgrblindwrt. Is it possible our new speedups are causing\n> > it?\n> \n> I can reproduce it. Keep looking...\n> BTW, did you compile without --enable-cassert ?\n> (Should be ON by default in beta-s...)\n> I got some interest assertion from BufferAlloc, without CASSERT you should get\n> dead spinlock from there.\n\nI always have asserts on.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 19 Feb 1998 09:24:36 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Shared memory corruption?" } ]
[ { "msg_contents": "\nA while back we talked about exec/shmem/mmap.\nI am thinking about messing it up and make some test\naround this issues when I get some spare time.\n\nI think I figured out that most memory,\nat least the buffers is allocated in shmem.\n\nWhat are the primary reason for this?\nSpeed? Footprint? Synchonization/Locking? Other?\n\n regards,\n-- \n---------------------------------------------\nG�ran Thyni, sysadm, JMS Bildbasen, Kiruna\n\n", "msg_date": "12 Feb 1998 19:44:05 -0000", "msg_from": "Goran Thyni <[email protected]>", "msg_from_op": true, "msg_subject": "shmem/mmap Q" }, { "msg_contents": "> \n> \n> A while back we talked about exec/shmem/mmap.\n> I am thinking about messing it up and make some test\n> around this issues when I get some spare time.\n> \n> I think I figured out that most memory,\n> at least the buffers is allocated in shmem.\n> \n> What are the primary reason for this?\n> Speed? Footprint? Synchonization/Locking? Other?\n\nThe exec() adds 0.01 seconds on a start connection/do select/disconnect\ntest that takes 0.09 seconds. That's a lot.\n\nSeems like the exec() is not necessary because it is the same binary as\nthe parent, and we can skip loading all those shared memory segments\ninto the address space. I also think we may be able to do some\ninitialization in the postmaster that can be given to each backend.\n\nOnce you get the shared memory stuff figured out, the only other issue\nis initialized variables or static local varaibles that are changed by\nthe postmaster. This should be minimal because the postmaster doesn't\ndo much.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 12 Feb 1998 16:14:19 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] shmem/mmap Q" } ]
[ { "msg_contents": "\nunsubscribe\n", "msg_date": "Thu, 12 Feb 1998 17:18:58 -0500", "msg_from": "\"Nguyen, Thuan\" <[email protected]>", "msg_from_op": true, "msg_subject": "None" } ]
[ { "msg_contents": "unscribe \n\n\n\n\n\n\n\nunscribe", "msg_date": "Fri, 13 Feb 1998 10:05:44 +0800", "msg_from": "\"Dr. Huaimin Wang\" <[email protected]>", "msg_from_op": true, "msg_subject": "None" } ]
[ { "msg_contents": "On Wed, 11 Feb 1998, Kenji T. Hollis wrote:\n\n>> Brian:\n>> \n>> > > I wrote a message about 4 months ago asking about an Alpha Linux version\n>> > > of PostGreSQL. Last I heard, they were working on it and it would be\n>> > > released in early January with a 64-bit clean version of the server.\n>> > > \n>> > > This is my third request.\n>> > \n>> > So, fix it your own self and stop whining. Can't fix it? Too bad, I\n>> > guess you'll just have to wait until they get around to it.\n>> \n>> Is this how you guys normally handle inquiries? Idiot replies like this?\n>> Oh by the way, I tried fixing it. Next question?\n>\n>\tNo, this isn't...generally we tend to try and point you in the\n>right direction towards determining the problem...\n>\n>\t...first question is, how many ppl out there are running under a\n>Linux/Alpha platform, or at least are trying? I don't have one myself, so\n>can only be of limited help...\n>\n\nI have been trying to build Postgres 6.3 on an Alpha, running Digital\nStandard Unix, and have been finding a similar problem to Ken Hollis,\nonce I succeeded in getting the thing to compile cleanly\n[...]\n>\n>ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n>ERROR: BuildFuncTupleDesc: function mkoidname(opaque, opaque) does not exist\n>\n\nGet a similar problem to the above\n\nThis problem has also occurred\n 1. using native cc instead of gcc\n 2. encouraging native cc to restrict image to 32 bits (-taso flag) at\n ld time.\n\nForcing pointers to 32 bits with native cc by use of the -xtaso_short\nCC and LD flags dies with errors related to varargs datatypes, so I\nhaven't managed to investigate this further.\n\nI originally also encountered these errors with postgres 6.2.1, but\ngiven the iminent release of 6.3 and significant other distractions, I\ndecided to wait for this before revisiting the issue.\n\nI have already posted this problem to the pgsql-ports mailing list,\nbut it is less that totally evident from our remote location what is\nthe best strategy, so I await developments here on 64 bit issues with\ninterest.\n\nregards\n\nPeter A. Stockwell\n\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 17:03:13 +1300 (NZDT)", "msg_from": "Peter Stockwell <[email protected]>", "msg_from_op": true, "msg_subject": "postgres initdb on ALPHA/Digital Unix" }, { "msg_contents": "On Fri, 13 Feb 1998, Peter Stockwell wrote:\n\n> On Wed, 11 Feb 1998, Kenji T. Hollis wrote:\n> \n> I have already posted this problem to the pgsql-ports mailing list,\n> but it is less that totally evident from our remote location what is\n> the best strategy, so I await developments here on 64 bit issues with\n> interest.\n\n\tall I can say to both you and Kenji...please stay vocal in here.\nAs Bruce has already shown, about the best we can offer is to try to help\npoint you in the right directions towards fixing this, but the fix has to\ncome from you guys. Use this list (or, preferrably, pgsql-ports) to\ndiscuss this issues amongst yourself...we'll throw in ideas as we can\nthink of them, but you are the ones that have to act on the ideas...\n\n\tPeter, how much experience do you have with gdb? I don't know if\nthis would help any (Bruce, comments?), but if you are reasonably familiar\nwith gdb, make a debug (-g) postmaster/postgres binary, and when it core\ndumps (I believe Kenji said that it did core dump), do an analysis of the\ncore using gdb. If nothing more then doing a 'where' inside of gdb to\nshow us the sequence of functions leading to the dump, and then hopefully\none of us will be able to suggest where to go next, as well as what\nvariables should be printed out at those points to give an idea of what\nmay be wrong...\n\n\tHell, if it helps any, I'll do the manual aspect of this if\nsomeone wants to give me access to their machine, and try to work with\nBruce on this shrug\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 13 Feb 1998 00:35:39 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] postgres initdb on ALPHA/Digital Unix" }, { "msg_contents": "> > I have already posted this problem to the pgsql-ports mailing list,\n> > but it is less that totally evident from our remote location what is\n> > the best strategy, so I await developments here on 64 bit issues with\n> > interest.\n\n> \tHell, if it helps any, I'll do the manual aspect of this if\n> someone wants to give me access to their machine, and try to work with\n> Bruce on this shrug\n\nDoes this problem only appear on ALPHA machines or also on other 64 bit \narchitectures? I might be able to find a 64 bit machine on my university \non which I can do a compile/initdb/gdb run if that helps....\n\nsparc stations 20 are 32 bits aren't they? Then that machine does not \nqualify. We are getting an ultra-sparc soon though, and there might be \nsome alpha's around.\n\nMaarten\n\n_____________________________________________________________________________\n\| TU Delft, The Netherlands, Faculty of Information Technology and Systems \|\n\| Department of Electrical Engineering \|\n\| Computer Architecture and Digital Technique section \|\n\| [email protected] \|\n-----------------------------------------------------------------------------\n\n", "msg_date": "Fri, 13 Feb 1998 11:34:16 +0100 (MET)", "msg_from": "Maarten Boekhold <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] postgres initdb on ALPHA/Digital Unix" }, { "msg_contents": "On Fri, 13 Feb 1998, Maarten Boekhold wrote:\n\n> > > I have already posted this problem to the pgsql-ports mailing list,\n> > > but it is less that totally evident from our remote location what is\n> > > the best strategy, so I await developments here on 64 bit issues with\n> > > interest.\n> \n> > \tHell, if it helps any, I'll do the manual aspect of this if\n> > someone wants to give me access to their machine, and try to work with\n> > Bruce on this shrug\n> \n> Does this problem only appear on ALPHA machines or also on other 64 bit \n> architectures? I might be able to find a 64 bit machine on my university \n> on which I can do a compile/initdb/gdb run if that helps....\n\n\tWe just also had a report from one Dec ALPHA user that he's been\nusing PostgreSQL since the v6.0 days with no problems...most curious...\n\n> sparc stations 20 are 32 bits aren't they? Then that machine does not \n> qualify. We are getting an ultra-sparc soon though, and there might be \n> some alpha's around.\n\n\tI hate to sound stupid, but an Ultra1...that's a 64bit machine?\nIf so, then I too have been running it on a 64bit machine, with absolutely\nno problems, for almost 6 months now...\n\n\n", "msg_date": "Fri, 13 Feb 1998 08:18:33 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] postgres initdb on ALPHA/Digital Unix" }, { "msg_contents": "> > sparc stations 20 are 32 bits aren't they? Then that machine does not \n> > qualify. We are getting an ultra-sparc soon though, and there might be \n> > some alpha's around.\n> \n> \tI hate to sound stupid, but an Ultra1...that's a 64bit machine?\n> If so, then I too have been running it on a 64bit machine, with absolutely\n> no problems, for almost 6 months now...\n\nYup, see http://www.sun.com/sparc/Net.Engine/ for more info.....\n\nMaarten \n\n_____________________________________________________________________________\n\| TU Delft, The Netherlands, Faculty of Information Technology and Systems \|\n\| Department of Electrical Engineering \|\n\| Computer Architecture and Digital Technique section \|\n\| [email protected] \|\n-----------------------------------------------------------------------------\n\n", "msg_date": "Fri, 13 Feb 1998 15:02:47 +0100 (MET)", "msg_from": "Maarten Boekhold <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] postgres initdb on ALPHA/Digital Unix" }, { "msg_contents": "> > I hate to sound stupid, but an Ultra1...that's a 64bit machine?\n> > If so, then I too have been running it on a 64bit machine, with absolutely\n> > no problems, for almost 6 months now...\n>\n> Yup, see http://www.sun.com/sparc/Net.Engine/ for more info.....\n\nI'm not familiar with the 64-bit Sparc machines, but is it possible that the OS\nand compilers default to a 32-bit mode? I believe that they did at one time...\n\n - Tom\n\n", "msg_date": "Fri, 13 Feb 1998 15:33:36 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] postgres initdb on ALPHA/Digital Unix" }, { "msg_contents": "On Fri, 13 Feb 1998, Thomas G. Lockhart wrote:\n\n> > > I hate to sound stupid, but an Ultra1...that's a 64bit machine?\n> > > If so, then I too have been running it on a 64bit machine, with absolutely\n> > > no problems, for almost 6 months now...\n> >\n> > Yup, see http://www.sun.com/sparc/Net.Engine/ for more info.....\n> \n> I'm not familiar with the 64-bit Sparc machines, but is it possible that\n> the OS and compilers default to a 32-bit mode? I believe that they did\n> at one time... \n\n\tI got a private email stating that it isn't until v2.6 that the\nOS actually starts to really use the 64bit stuff (I'm summarizing\nhere)...I'm doing my testing on both x86 and sparc v2.6, and so far, no\nnoticeable problems...\n\n\n", "msg_date": "Fri, 13 Feb 1998 12:17:07 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] postgres initdb on ALPHA/Digital Unix" }, { "msg_contents": ">>>>> \"scrappy\" == The Hermit Hacker <[email protected]> writes:\n\n > On Fri, 13 Feb 1998, Maarten Boekhold wrote:\n >> > > I have already posted this problem to the pgsql-ports\n >> mailing list, > > but it is less that totally evident from our\n >> remote location what is > > the best strategy, so I await\n >> developments here on 64 bit issues with > > interest.\n >> \n >> > Hell, if it helps any, I'll do the manual aspect of this if >\n >> someone wants to give me access to their machine, and try to\n >> work with > Bruce on this shrug\n >> \n >> Does this problem only appear on ALPHA machines or also on\n >> other 64 bit architectures? I might be able to find a 64 bit\n >> machine on my university on which I can do a compile/initdb/gdb\n >> run if that helps....\n\n > \tWe just also had a report from one Dec ALPHA user that he's\n > been using PostgreSQL since the v6.0 days with no\n > problems...most curious...\n\n >> sparc stations 20 are 32 bits aren't they? Then that machine\n >> does not qualify. We are getting an ultra-sparc soon though,\n >> and there might be some alpha's around.\n\n > \tI hate to sound stupid, but an Ultra1...that's a 64bit\n > machine? If so, then I too have been running it on a 64bit\n > machine, with absolutely no problems, for almost 6 months now...\n\nSparc Ultra are 64 bits, but Solaris is only a 32 bit OS. The only 64\nbit machines (with 64 bit OS's) would be some SGI's and Alphas.\n\n\nKent S. Gordon\nArchitect\niNetSpace Co.\nvoice: (972)851-3494 fax:(972)702-0384 e-mail:[email protected]\n\n\n\n", "msg_date": "Fri, 13 Feb 1998 14:55:17 -0600 (CST)", "msg_from": "\"Kent S. Gordon\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] postgres initdb on ALPHA/Digital Unix" } ]
[ { "msg_contents": "Just tried 'em, and like 'em :) Great work Vadim!\n\n - Tom\n\n", "msg_date": "Fri, 13 Feb 1998 06:57:55 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Subselects" }, { "msg_contents": "Thomas G. Lockhart wrote:\n> \n> Just tried 'em, and like 'em :) Great work Vadim!\n\nThanks. I'm currently writing message about implementation and opened\nissues, will post it in the next 2-3 hrs...\n\nVadim\n", "msg_date": "Fri, 13 Feb 1998 14:19:19 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Subselects" } ]
[ { "msg_contents": "Hi,\n\nI know this is a bit off topic, but it's concerning something\nthat I believe most ppl find a bit annoying.\n Like most other mailinglists the postgresql ones suffer from\noccational \"unsubscribe\" msges in the list itself. Wouldn't it\nbe possible to write some filter that stop these mail and return\nunsubscription instructions to the sender?\n\nRegards,\nPatrik Kudo\n\n", "msg_date": "Fri, 13 Feb 1998 08:19:24 +0100 (CET)", "msg_from": "Patrik Kudo <[email protected]>", "msg_from_op": true, "msg_subject": "Mailfilter" } ]
[ { "msg_contents": "ALL_SUBLINK type was returned for x Op (subquery).\nJust fixed, not in today snapshot...\nWork arround: (x) Op (subquery)...\n ^ ^\nVadim\n", "msg_date": "Fri, 13 Feb 1998 15:32:05 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": true, "msg_subject": "little bug in gram.y" } ]
[ { "msg_contents": "It seems it didn't make it into the source tree so far. In fact I'm not even\nsure I really send it out. Since I'd like to do some more work now I can\neasily include yesterday's changes again. But if the patch somehow makes it\ninto the archive we will get 'already applied' errors. \n\nSo please do not apply it anymore in case it is still in the queue. I will\nupload a new version later on.\n\nMichael\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Fri, 13 Feb 1998 09:39:45 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "My second patch from yesterday" } ]
[ { "msg_contents": "I removed my old 6.3 DB and called initdb. After starting postmaster I tried\nto create a user to no avail:\n\ngauss:~> createuser meskes\nPQexec() -- Request was sent to backend, but backend closed the channel\nbefore responding.\n This probably means the backend terminated abnormally before or\nwhile processing the request.\ncreateuser: database access failed.\n\nMichael\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Fri, 13 Feb 1998 10:01:25 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "backend terminates" } ]
[ { "msg_contents": "I'm reposting this, since the first reply didn't make it to the list.\n\nOn Thu, 12 Feb 1998, Bruce Momjian wrote:\n\n>> I created a small program to return the size of values. They are:\n>> \n>> Size of short: 2\n>> Size of char: 1\n>> Size of int: 4\n>> Size of long: 8\n>> \n>> Does this help?\n>\n>Yep. Does anyone else run OS's with the size of long greater than int? \n>Does Dec Unix Alpha? Are people running that?\n\nYes, I'm running PostgreSQL since the 6.0 days, now 6.2.1p6, on a DEC\nAlpha with Digital Unix 3.2c, and it has always worked like a charm. I use\nthe DEC C compiler, if it makes any difference. The type sizes are exactly\nthose above.\n\n\tPedro.\n\n-------------------------------------------------------------------\nPedro José Lobo Perea Tel: +34 1 336 78 19\nCentro de Cálculo Fax: +34 1 331 92 29\nEUIT Telecomunicación - UPM e-mail: [email protected]\n\n\n", "msg_date": "Fri, 13 Feb 1998 10:43:14 +0100 (MET)", "msg_from": "\"Pedro J. Lobo\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha (repost)" }, { "msg_contents": "> >Yep. Does anyone else run OS's with the size of long greater than int?\n> >Does Dec Unix Alpha? Are people running that?\n>\n> Yes, I'm running PostgreSQL since the 6.0 days, now 6.2.1p6, on a DEC\n> Alpha with Digital Unix 3.2c, and it has always worked like a charm. I use\n> the DEC C compiler, if it makes any difference. The type sizes are exactly\n> those above.\n\nThere is a possibility that the v6.3beta has trouble with your type of\nmachine. Since you have had success in the past, and (presumeably) would like\nto use v6.3, would you have time to try testing the new package to see if it\ninstalls for your machine?\n\nWe have enough reports of possible trouble (and a new port support scheme)\nthat each platform needs to be tested before v6.3 is released.\n\n - Tom\n\n", "msg_date": "Fri, 13 Feb 1998 13:20:03 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha (repost)" }, { "msg_contents": "On Fri, 13 Feb 1998, Thomas G. Lockhart wrote:\n\n>> Yes, I'm running PostgreSQL since the 6.0 days, now 6.2.1p6, on a DEC\n>> Alpha with Digital Unix 3.2c, and it has always worked like a charm. I use\n>> the DEC C compiler, if it makes any difference. The type sizes are exactly\n>> those above.\n>\n>There is a possibility that the v6.3beta has trouble with your type of\n>machine. Since you have had success in the past, and (presumeably) would like\n>to use v6.3, would you have time to try testing the new package to see if it\n>installs for your machine?\n\nOf course I will... when I can find the time :-( In fact, I want to make\nsome tests with hash indices, to see if they work properly or not. I hope\nto have the time to grab the last snapshot, build it and run the\nregression tests on Monday. I'll let you know as soon I have any result\n(positive or not).\n\n>We have enough reports of possible trouble (and a new port support scheme)\n>that each platform needs to be tested before v6.3 is released.\n\nSure. BTW, should we move this discussion to pgsql-ports?\n\n\tPedro.\n\n-------------------------------------------------------------------\nPedro José Lobo Perea Tel: +34 1 336 78 19\nCentro de Cálculo Fax: +34 1 331 92 29\nEUIT Telecomunicación - UPM e-mail: [email protected]\n\n", "msg_date": "Fri, 13 Feb 1998 14:30:41 +0100 (MET)", "msg_from": "\"Pedro J. Lobo\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] PostGreSQL v6.2.1 for Linux Alpha (repost)" } ]
[ { "msg_contents": "\nI looked in gram.y and scan.l but couldn't figure out how to unreserve\n\"user\" so I can use it as the name of a table. Is there a way to do\nthis?\n", "msg_date": "Fri, 13 Feb 1998 02:38:23 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "reserved words.." }, { "msg_contents": "> I looked in gram.y and scan.l but couldn't figure out how to unreserve\n> \"user\" so I can use it as the name of a table. Is there a way to do\n> this?\n\nYes, and since I'm working there I'll make the change.\n\nfyi, look for the line starting with \"ColId:\" in gram.y, and add \"USER\"\nas one of the clauses just like the other keywords are already.\n\n - Tom\n\n", "msg_date": "Fri, 13 Feb 1998 14:02:31 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] reserved words.." }, { "msg_contents": "\nand somehow I missed that...? doh!\n\nOn Fri, 13 February 1998, at 14:02:31, Thomas G. Lockhart wrote:\n\n> > I looked in gram.y and scan.l but couldn't figure out how to unreserve\n> > \"user\" so I can use it as the name of a table. Is there a way to do\n> > this?\n> \n> Yes, and since I'm working there I'll make the change.\n> \n> fyi, look for the line starting with \"ColId:\" in gram.y, and add \"USER\"\n> as one of the clauses just like the other keywords are already.\n> \n> - Tom\n> \n", "msg_date": "Fri, 13 Feb 1998 07:05:26 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] reserved words.." } ]
[ { "msg_contents": "\nthis is the latest snapshot under linux 2.0.33\n\nwhen I create a view with an aggregate such as count (with a group by\nclause) I get a core dump. The first three lines of the gdb backtrace\nfollows:\n\n#0 0x80a9982 in ExecEvalAggreg (agg=0x820fa20, econtext=0x8213a48, \n isNull=0xbfffd6cf \"\") at execQual.c:201\n#1 0x80ab271 in ExecEvalExpr (expression=0x820fa20, econtext=0x8213a48, \n isNull=0xbfffd6cf \"\", isDone=0xbfffd7c3 \"\\001\\fZ\\031\\b\\030K!\\b\")\n at execQual.c:1199\n#2 0x80ab69b in ExecTargetList (targetlist=0x8213498, nodomains=2, \n targettype=0x8214990, values=0x8214b48, econtext=0x8213a48, \n isDone=0xbfffd7c3 \"\\001\\fZ\\031\\b\\030K!\\b\") at execQual.c:1484\n\ncode at point:\n\n201 isNull = econtext->ecxt_nulls[agg->aggno];\n\necontext->ecxt_nulls, is, well, a NULL pointer and hard to index :)\ni'll take a look but I don't think I'll find it -- i notice that 6.2.1\npretends to make a view with aggregates but then it comes up empty\nwhen selected.\n\n--brett\nlet me know if you need more info\n", "msg_date": "Fri, 13 Feb 1998 03:19:44 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "v6.3 snapshot core dump" }, { "msg_contents": "Brett McCormick wrote:\n> \n> this is the latest snapshot under linux 2.0.33\n> \n> when I create a view with an aggregate such as count (with a group by\n> clause) I get a core dump. The first three lines of the gdb backtrace\n> follows:\n...\n> \n> 201 isNull = econtext->ecxt_nulls[agg->aggno];\n> \n> econtext->ecxt_nulls, is, well, a NULL pointer and hard to index :)\n> i'll take a look but I don't think I'll find it -- i notice that 6.2.1\n> pretends to make a view with aggregates but then it comes up empty\n> when selected.\n> \n> --brett\n> let me know if you need more info\n\nDid this work in previous 6.3 snapshots ?\n\nVadim\n", "msg_date": "Fri, 13 Feb 1998 19:11:26 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] v6.3 snapshot core dump" }, { "msg_contents": "\nOn Fri, 13 February 1998, at 19:11:26, Vadim B. Mikheev wrote:\n\n> > econtext->ecxt_nulls, is, well, a NULL pointer and hard to index :)\n> > i'll take a look but I don't think I'll find it -- i notice that 6.2.1\n> > pretends to make a view with aggregates but then it comes up empty\n> > when selected.\n> > \n> > --brett\n> > let me know if you need more info\n> \n> Did this work in previous 6.3 snapshots ?\n> \n> Vadim\n\nI never attempted it before.\n\n", "msg_date": "Fri, 13 Feb 1998 04:40:24 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] v6.3 snapshot core dump" }, { "msg_contents": "I will look into this. I made many fixes in 6.3 for aggregates and\nviews, but I did not test views of aggregates. I will do that soon.\n\nGood test case.\n\n> \n> \n> this is the latest snapshot under linux 2.0.33\n> \n> when I create a view with an aggregate such as count (with a group by\n> clause) I get a core dump. The first three lines of the gdb backtrace\n> follows:\n> \n> #0 0x80a9982 in ExecEvalAggreg (agg=0x820fa20, econtext=0x8213a48, \n> isNull=0xbfffd6cf \"\") at execQual.c:201\n> #1 0x80ab271 in ExecEvalExpr (expression=0x820fa20, econtext=0x8213a48, \n> isNull=0xbfffd6cf \"\", isDone=0xbfffd7c3 \"\\001\\fZ\\031\\b\\030K!\\b\")\n> at execQual.c:1199\n> #2 0x80ab69b in ExecTargetList (targetlist=0x8213498, nodomains=2, \n> targettype=0x8214990, values=0x8214b48, econtext=0x8213a48, \n> isDone=0xbfffd7c3 \"\\001\\fZ\\031\\b\\030K!\\b\") at execQual.c:1484\n> \n> code at point:\n> \n> 201 *isNull = econtext->ecxt_nulls[agg->aggno];\n> \n> econtext->ecxt_nulls, is, well, a NULL pointer and hard to index :)\n> i'll take a look but I don't think I'll find it -- i notice that 6.2.1\n> pretends to make a view with aggregates but then it comes up empty\n> when selected.\n> \n> --brett\n> let me know if you need more info\n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 13:25:15 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] v6.3 snapshot core dump" } ]
[ { "msg_contents": "Is the ix86/linux \"confirmed\" port using libc 5 or glibc 2.x (aka \nlibc6)? As most major distributions are going to release all new \nversions with glibc I think it's prudent to test on both \"platforms\" \nindependantly.\n\n-Fred\n\n\n> Bruce, are you planning on keeping a ToDo list for the 6.3 release\n> as you have in (at least) the past few releases? That will make sure\n> that we don't leave something out. As a start:\n> \n> Commit sgml document sources(Thomas)\n> Commit html documents(Thomas)\n> Commit postscript documents(Thomas)\n> Fix aggregates in views reported by Brett\n> Include subselect regression tests(Thomas, unless someone else wants\n> it :) Fix crashes from backend collisions reported by Tom H. Fix\n> isinf.c irix port problem reported by Andrew(Marc) Check triggers\n> regression test(Vadim) Check select_views regression test Test ecpg\n> new struct features(Michael) Test big-endian/little-endian\n> operation(Tatsuo?) I don' know if we need this...\n> \n> Confirm ports for v6.3 (need this soon for docs):\n> \n> _ aix4.2 - already fixed? what are actual versions??\n> _ aix3.5\n> _ bsdi\n> _ freebsd,etc (I don't know the different bsd variants)\n> _ 68k/bsdXXX (Macintosh port) never successful? Needed 68k locking\n> code?? _ dgux _ hpux9.0.x _ hpux10.20 _ alpha/unix - currently\n> broken? _ alpha/linux - currently broken? _ ppc/mklinux(Tatsuo?)\n> * ix86/linux(Thomas) - confirmed working\n> _ ix86/solaris\n> _ irix5\n> _ irix6\n> _ nextstep - not tested? not working??\n> _ sco\n> _ sparc/solaris\n> _ sparc/sunos\n> _ svr4(which platforms?)\n> _ ultrix4\n> _ univel\n> \n> If a platform does not get on this list, and have a confirmed\n> compile/regression test, then it will go on the unsupported list for\n> v6.3 :( The regression tests do not have to pass completely (since\n> there are a couple of remaining beta problems) but the port should\n> compile, run, and produce reasonable results for most tests.\n> \n> Can people commit to checking on each platform? (I know some have\n> been done already...)\n> \n> \n", "msg_date": "Fri, 13 Feb 1998 11:24:12 +0000", "msg_from": "\"Frederick W. Reimer\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "Bruce, are you planning on keeping a ToDo list for the 6.3 release as\nyou have in (at least) the past few releases? That will make sure that\nwe don't leave something out. As a start:\n\nCommit sgml document sources(Thomas)\nCommit html documents(Thomas)\nCommit postscript documents(Thomas)\nFix aggregates in views reported by Brett\nInclude subselect regression tests(Thomas, unless someone else wants it\n:)\nFix crashes from backend collisions reported by Tom H.\nFix isinf.c irix port problem reported by Andrew(Marc)\nCheck triggers regression test(Vadim)\nCheck select_views regression test\nTest ecpg new struct features(Michael)\nTest big-endian/little-endian operation(Tatsuo?) I don' know if we need\nthis...\n\nConfirm ports for v6.3 (need this soon for docs):\n\n_ aix4.2 - already fixed? what are actual versions??\n_ aix3.5\n_ bsdi\n_ freebsd,etc (I don't know the different bsd variants)\n_ 68k/bsdXXX (Macintosh port) never successful? Needed 68k locking\ncode??\n_ dgux\n_ hpux9.0.x\n_ hpux10.20\n_ alpha/unix - currently broken?\n_ alpha/linux - currently broken?\n_ ppc/mklinux(Tatsuo?)\n* ix86/linux(Thomas) - confirmed working\n_ ix86/solaris\n_ irix5\n_ irix6\n_ nextstep - not tested? not working??\n_ sco\n_ sparc/solaris\n_ sparc/sunos\n_ svr4(which platforms?)\n_ ultrix4\n_ univel\n\nIf a platform does not get on this list, and have a confirmed\ncompile/regression test, then it will go on the unsupported list for\nv6.3 :( The regression tests do not have to pass completely (since\nthere are a couple of remaining beta problems) but the port should\ncompile, run, and produce reasonable results for most tests.\n\nCan people commit to checking on each platform? (I know some have been\ndone already...)\n\n", "msg_date": "Fri, 13 Feb 1998 16:08:45 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "v6.3 release ToDo list and supported ports" }, { "msg_contents": "> Is the ix86/linux \"confirmed\" port using libc 5 or glibc 2.x (aka\n> libc6)? As most major distributions are going to release all new\n> versions with glibc I think it's prudent to test on both \"platforms\"\n> independently.\n\nYes, that is a good point. There is a known bug in the glibc2 math\nlibrary which breaks the date/time routines:\n\nselect '4 hours'::timespan;\n---------\n@ 4 hours\n(1 row)\n\ncomes out instead as\ntgl=> select '4 hours'::timespan;\n----------------------------\n@ 3 hours 59 mins 60.00 secs\n(1 row)\n\nOliver was working on patches. Lost his e-mail message; is Oliver still\nhere? I've got a RH5.0 linux system at work now, and the shipped Postgres\ninstallation has this problem. I'd like to see it fixed...\n\nBruce, can you add this to the v6.3 ToDo (assuming you agree to take it\non :)\n\n - Tom\n\n\n", "msg_date": "Fri, 13 Feb 1998 16:57:18 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "On Fri, 13 Feb 1998, Thomas G. Lockhart wrote:\n\n> _ freebsd,etc (marc) (I don't know the different bsd variants)\n\n\tFreeBSD is confirmed recently, but will be doing a new build this\nweekend now that all the subselect code is in place...I also have access\nto a Net/BSD machine that I'll be connecting to over the weekend and doing\na build/test on there also...\n\n> _ ix86/solaris(marc) - confirmed working\n\n> _ sparc/solaris(marc) - confirmed working\n\n> _ sparc/sunos\n\n\t- I have an old SunOS machine here at work that I can most likely\nrun it on, if nobody else wishes to volunteer...I don't know if it even\nhas a compiler on it, its only used as a console server in our machine\nroom :)\n\n\n", "msg_date": "Fri, 13 Feb 1998 12:23:18 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "> \n> Bruce, are you planning on keeping a ToDo list for the 6.3 release as\n> you have in (at least) the past few releases? That will make sure that\n> we don't leave something out. As a start:\n\nYea, I have stuff in my mailbox. Now that subselects are done, I can\nstart putting together a list. This helps me get started.\n\n> \n> Commit sgml document sources(Thomas)\n> Commit html documents(Thomas)\n> Commit postscript documents(Thomas)\n> Fix aggregates in views reported by Brett\n> Include subselect regression tests(Thomas, unless someone else wants it\n> :)\n> Fix crashes from backend collisions reported by Tom H.\n> Fix isinf.c irix port problem reported by Andrew(Marc)\n> Check triggers regression test(Vadim)\n> Check select_views regression test\n> Test ecpg new struct features(Michael)\n> Test big-endian/little-endian operation(Tatsuo?) I don' know if we need\n> this...\n> \n> Confirm ports for v6.3 (need this soon for docs):\n> \n> _ aix4.2 - already fixed? what are actual versions??\n> _ aix3.5\n> _ bsdi\n> _ freebsd,etc (I don't know the different bsd variants)\n> _ 68k/bsdXXX (Macintosh port) never successful? Needed 68k locking\n> code??\n> _ dgux\n> _ hpux9.0.x\n> _ hpux10.20\n> _ alpha/unix - currently broken?\n> _ alpha/linux - currently broken?\n> _ ppc/mklinux(Tatsuo?)\n> * ix86/linux(Thomas) - confirmed working\n> _ ix86/solaris\n> _ irix5\n> _ irix6\n> _ nextstep - not tested? not working??\n> _ sco\n> _ sparc/solaris\n> _ sparc/sunos\n> _ svr4(which platforms?)\n> _ ultrix4\n> _ univel\n> \n> If a platform does not get on this list, and have a confirmed\n> compile/regression test, then it will go on the unsupported list for\n> v6.3 :( The regression tests do not have to pass completely (since\n> there are a couple of remaining beta problems) but the port should\n> compile, run, and produce reasonable results for most tests.\n> \n> Can people commit to checking on each platform? (I know some have been\n> done already...)\n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 14:27:37 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: v6.3 release ToDo list and supported ports" }, { "msg_contents": "> Oliver was working on patches. Lost his e-mail message; is Oliver still\n> here? I've got a RH5.0 linux system at work now, and the shipped Postgres\n> installation has this problem. I'd like to see it fixed...\n> \n> Bruce, can you add this to the v6.3 ToDo (assuming you agree to take it\n> on :)\n\nTo be honest, I there has been so much going on, I am not sure what to\ndocument, especially with the ports. I don't think I ever did a good\njob of tracking which ports did and did not work. I usually\nconcentrated on show-stopper bugs like crashes and broken features.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 14:30:14 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "> Oliver was working on patches. Lost his e-mail message; is Oliver still\n> here? I've got a RH5.0 linux system at work now, and the shipped Postgres\n> installation has this problem. I'd like to see it fixed...\n> \n> Bruce, can you add this to the v6.3 ToDo (assuming you agree to take it\n> on :)\n\nMarc and Thomas, there are so many patches going around, I can't even\nkeep track that they are all getting applied. Thank goodness Marc is on\ntop of it.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 14:31:41 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "On Fri, 13 Feb 1998, Bruce Momjian wrote:\n\n> > Oliver was working on patches. Lost his e-mail message; is Oliver still\n> > here? I've got a RH5.0 linux system at work now, and the shipped Postgres\n> > installation has this problem. I'd like to see it fixed...\n> > \n> > Bruce, can you add this to the v6.3 ToDo (assuming you agree to take it\n> > on :)\n> \n> Marc and Thomas, there are so many patches going around, I can't even\n> keep track that they are all getting applied. Thank goodness Marc is on\n> top of it.\n\n\tI'm trying...there are some that I let \"slip\" I'm uncertain of\nthem. If anyone has submitted a patch recently that hasn't been\napplied, it may have slipped through the cracks inadvertantly...please\nrepost it with a subject that includes \"REPOST\" in it...\n\n\tBruce/Thomas/Vadim...if you see one of those go through, please\ntake a scan as it, cause there is a good chance that I didn't apply it due\nto uncertainty...\n\n\n", "msg_date": "Fri, 13 Feb 1998 14:41:40 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "> \n> On Fri, 13 Feb 1998, Bruce Momjian wrote:\n> \n> > > Oliver was working on patches. Lost his e-mail message; is Oliver still\n> > > here? I've got a RH5.0 linux system at work now, and the shipped Postgres\n> > > installation has this problem. I'd like to see it fixed...\n> > > \n> > > Bruce, can you add this to the v6.3 ToDo (assuming you agree to take it\n> > > on :)\n> > \n> > Marc and Thomas, there are so many patches going around, I can't even\n> > keep track that they are all getting applied. Thank goodness Marc is on\n> > top of it.\n> \n> \tI'm trying...there are some that I let \"slip\" I'm uncertain of\n> them. If anyone has submitted a patch recently that hasn't been\n> applied, it may have slipped through the cracks inadvertantly...please\n> repost it with a subject that includes \"REPOST\" in it...\n> \n> \tBruce/Thomas/Vadim...if you see one of those go through, please\n> take a scan as it, cause there is a good chance that I didn't apply it due\n> to uncertainty...\n\nProblem is, I have never been able to determine what patches you are\nuncertain about.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 14:45:09 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "On Fri, 13 Feb 1998, Bruce Momjian wrote:\n\n> Problem is, I have never been able to determine what patches you are\n> uncertain about.\n\n\tOkay, I'll start replying to the list about patches that I'm not\npersonally comfortable with applying...most look okay, and I'll apply\nthem, but some of htem...\n\n\n", "msg_date": "Fri, 13 Feb 1998 14:52:58 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "> \n> On Fri, 13 Feb 1998, Bruce Momjian wrote:\n> \n> > Problem is, I have never been able to determine what patches you are\n> > uncertain about.\n> \n> \tOkay, I'll start replying to the list about patches that I'm not\n> personally comfortable with applying...most look okay, and I'll apply\n> them, but some of htem...\n\nUsually we pipe up when we see a bad patch, but not always.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 15:09:42 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "At 8:08 AM -0800 2/13/98, Thomas G. Lockhart wrote:\n>Bruce, are you planning on keeping a ToDo list for the 6.3 release as\n\n>_ 68k/bsdXXX (Macintosh port) never successful? Needed 68k locking\n>code??\n\nThat's NetBSD/m68k properly speaking since the difference between an Amiga,\nan HP 300 and a Macintosh is only visible inside the kernel. Userland\nbinaries are identical for all 68000 NetBSD ports. Most likely my problems\nwould also exist for OpenBSD/m68k.\n\nI wrote the 68k locking code and fixed the include file so it would\nrecognize that the code only existed for certain specific CPU's, not for\nall NetBSD ports. This was not strictly necessary to get it to work; just\nwhat I could do.\n\nAs of 6.2.x Postgres does not work with NetBSD/m68k. I have not had time\nto test with 6.3beta. It dies during initdb, just like the linux/alpha\nversion.\n\nTom Helbekkmo reports that Postgres works just fine on NetBSD/sparc (though\nyou should check he's tried a recient version there), and I feel fairly\nconfident of NetBSD/i386 in view of the FreeBSD contengent.\n\nSignature failed Preliminary Design Review.\nFeasibility of a new signature is currently being evaluated.\[email protected], or [email protected]\n\n\n", "msg_date": "Fri, 13 Feb 1998 12:31:32 -0800", "msg_from": "\"Henry B. Hotz\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "Here are the items.\n\n---------------------------------------------------------------------------\n\nORDER BY NULLs problem?\npg_user not world-readable, no initdb password option(Todd Brandys)\nmake psql \\d optionally user pg_user(Bruce)\noptimizer memory exhaustion with many OR's\nmax tuple size patch(Darren)\nIs GROUP BY duplicates fixed?\nDo we have a self-join optimizer performance problem?\nGet interfaces fixed for new protocol\nProfiling improvements?\nDo we have a problem with GROUP BY without ORDER BY?\nProblem with tables >2Gb\nDo we want to add timestamps to elog messages?\nScanKeyData missing initializations\nCan we improve vacuum locking issues?\nAlpha/64-bit issues, mkoidname() problem\n'Can not write block blind' error on createdb and regression collision\nViews on aggregates fail\nlarge objects memory exhaustion\nsubselect issues, NOT IN (empty query)\nCommit sgml document sources(Thomas)\nCommit html documents(Thomas)\nCommit postscript documents(Thomas)\nInclude subselect regression tests\nFix isinf.c irix port problem reported by Andrew(Marc)\nCheck triggers regression test(Vadim)\nCheck select_views regression test\nTest ecpg new struct features(Michael)\nTest big-endian/little-endian operation(Tatsuo?)?\nLinux library issues?\n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 16:29:27 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Open 6.3 issues" }, { "msg_contents": "On Fri, 13 Feb 1998, Bruce Momjian wrote:\n\n> > \n> > On Fri, 13 Feb 1998, Bruce Momjian wrote:\n> > \n> > > Problem is, I have never been able to determine what patches you are\n> > > uncertain about.\n> > \n> > \tOkay, I'll start replying to the list about patches that I'm not\n> > personally comfortable with applying...most look okay, and I'll apply\n> > them, but some of htem...\n> \n> Usually we pipe up when we see a bad patch, but not always.\n\n\tHrmmm, so reasonably safe to just apply a patch and expect that if\nits a problem, someone will yell at me about it? :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sat, 14 Feb 1998 02:17:00 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "On Fri, Feb 13, 1998 at 12:31:32PM -0800, Henry B. Hotz wrote:\n\n> Tom Helbekkmo reports that Postgres works just fine on NetBSD/sparc\n> (though you should check he's tried a recient version there), and I\n> feel fairly confident of NetBSD/i386 in view of the FreeBSD contengent.\n\nI rebuild everything under NetBSD/sparc and NetBSD/vax at least twice\nper week, and will try to increase the rebuild rate as the release\ndraws near. If I can just get around to replacing the power supply in\nmy old i386 box, I can do the same for NetBSD/i386.\n\nHowever: reading regression test output is a royal pain. We're\ntalking almost 100 kilobytes of 'diff' output per platform, most of\nwhich is last digit differences in floating point numbers and slightly\ndifferent wording of (expected) error messages. I cannot guarantee\nthat I don't miss actual problems in there that mean something\nrelevant. So far things look pretty good, though.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "Sat, 14 Feb 1998 13:34:15 +0100", "msg_from": "Tom I Helbekkmo <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "> > Tom Helbekkmo reports that Postgres works just fine on NetBSD/sparc\n> > (though you should check he's tried a recient version there), and I\n> > feel fairly confident of NetBSD/i386 in view of the FreeBSD contengent.\n>\n> I rebuild everything under NetBSD/sparc and NetBSD/vax at least twice\n> per week, and will try to increase the rebuild rate as the release\n> draws near. If I can just get around to replacing the power supply in\n> my old i386 box, I can do the same for NetBSD/i386.\n>\n> However: reading regression test output is a royal pain. We're\n> talking almost 100 kilobytes of 'diff' output per platform, most of\n> which is last digit differences in floating point numbers and slightly\n> different wording of (expected) error messages. I cannot guarantee\n> that I don't miss actual problems in there that mean something\n> relevant. So far things look pretty good, though.\n\nThat counts as a working port for v6.3, since there are not likely to be\nchanges that break individual ports from here on. If there is a change which\nmight, then we will post a \"call for port testing\" on hackers. No need to do\nexplicit port testing on a daily/weekly basis once a port has been shown to\nwork in the last few weeks, although it would be a good idea to do it once\nmore perhaps a few days before the release.\n\nI'll take on the list of tested ports if someone has not already, since I'll\nwant to fold it into the docs anyway. Will repost my current list soon.\n\n - Tom\n\n", "msg_date": "Sat, 14 Feb 1998 14:20:43 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "> \n> On Fri, 13 Feb 1998, Bruce Momjian wrote:\n> \n> > > \n> > > On Fri, 13 Feb 1998, Bruce Momjian wrote:\n> > > \n> > > > Problem is, I have never been able to determine what patches you are\n> > > > uncertain about.\n> > > \n> > > \tOkay, I'll start replying to the list about patches that I'm not\n> > > personally comfortable with applying...most look okay, and I'll apply\n> > > them, but some of htem...\n> > \n> > Usually we pipe up when we see a bad patch, but not always.\n> \n> \tHrmmm, so reasonably safe to just apply a patch and expect that if\n> its a problem, someone will yell at me about it? :)\n\nYep.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sat, 14 Feb 1998 12:42:43 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "Thomas G. Lockhart wrote:\n> \n> Include subselect regression tests(Thomas, unless someone else wants it :)\n\nOh, thanks!\n\nVadim\n", "msg_date": "Sun, 15 Feb 1998 16:35:19 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "> \n> Thomas G. Lockhart wrote:\n> > \n> > Include subselect regression tests(Thomas, unless someone else wants it :)\n> \n> Oh, thanks!\n\nI think someone suggested using the MySQL crashme test for subselects. \nMySQL doesn't have subselects.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sun, 15 Feb 1998 12:55:47 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> >\n> > Thomas G. Lockhart wrote:\n> > >\n> > > Include subselect regression tests(Thomas, unless someone else wants it :)\n> >\n> > Oh, thanks!\n> \n> I think someone suggested using the MySQL crashme test for subselects.\n> MySQL doesn't have subselects.\n ^^^^^^^^^^^^^\nMe. I know, but I thought that crashme tests more features than\nMySQL has itself.\n\nVadim\n", "msg_date": "Mon, 16 Feb 1998 08:51:17 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" }, { "msg_contents": "> \n> Bruce Momjian wrote:\n> > \n> > >\n> > > Thomas G. Lockhart wrote:\n> > > >\n> > > > Include subselect regression tests(Thomas, unless someone else wants it :)\n> > >\n> > > Oh, thanks!\n> > \n> > I think someone suggested using the MySQL crashme test for subselects.\n> > MySQL doesn't have subselects.\n> ^^^^^^^^^^^^^\n> Me. I know, but I thought that crashme tests more features than\n> MySQL has itself.\n\nGood question. I don't know the answer myself.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Mon, 16 Feb 1998 15:50:37 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported ports" } ]
[ { "msg_contents": "Hi,\n\n Someone changed the parser to build a TypeName node on CREATE\n FUNCTION in any case. As a side effect, ALL! functions\n created got the proretset attribute to true. Thus for a\n SELECT the parser wrapped an Iter node around the Expr and\n since singleton functions set isDone the Iter returns no\n tuple up.\n\n\nUntil later, 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* define.c.orig\tFri Feb 13 12:14:17 1998\n--- define.c\tFri Feb 13 12:14:38 1998\n***********\n* 94,100 ***\n \t\tTypeName setType = (TypeName ) returnType;\n \n \t\tprorettype_p = setType->name;\n! \t\treturnsSet_p = true;\n \t}\n \telse\n \t{\n--- 94,100 ----\n \t\tTypeName setType = (TypeName ) returnType;\n \n \t\tprorettype_p = setType->name;\n! \t\treturnsSet_p = setType->setof;\n \t}\n \telse\n \t{\n", "msg_date": "Fri, 13 Feb 1998 12:41:14 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "CREATE FUNCTION broken" }, { "msg_contents": "> Someone changed the parser to build a TypeName node on CREATE\n> FUNCTION in any case. As a side effect, ALL! functions\n> created got the proretset attribute to true. Thus for a\n> SELECT the parser wrapped an Iter node around the Expr and\n> since singleton functions set isDone the Iter returns no\n> tuple up.\n\nAh. I broke it (though the regression tests did not find the problem). What\nI changed was the code in gram.y, which used to just create a string node\nfor the return type clause _unless_ the return type was a \"SETOF type\". In\nthat case a Typename node was created, and the setof attribute was\nexplicitly set.\n\nWhat you found is that farther along, the setof attribute was forced to be\ntrue if _any_ Typename node is present.\n\nIt looks like your patch will completely fix things, and is better than my\nreverting the gram.y code. Can you suggest a small test case to include in\nthe regression suite?\n\nUnless there are objections from others (with a preference for reverting\nthe gram.y code) I'll go ahead and apply Jan's patch.\n\n - Tom\n\n", "msg_date": "Fri, 13 Feb 1998 13:50:04 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CREATE FUNCTION broken" }, { "msg_contents": "\nTom wrote:\n>\n> > Someone changed the parser to build a TypeName node on CREATE\n> > FUNCTION in any case. As a side effect, ALL! functions\n> > created got the proretset attribute to true. Thus for a\n> > SELECT the parser wrapped an Iter node around the Expr and\n> > since singleton functions set isDone the Iter returns no\n> > tuple up.\n>\n> Ah. I broke it (though the regression tests did not find the problem). What\n> I changed was the code in gram.y, which used to just create a string node\n> for the return type clause _unless_ the return type was a \"SETOF type\". In\n> that case a Typename node was created, and the setof attribute was\n> explicitly set.\n>\n> What you found is that farther along, the setof attribute was forced to be\n> true if _any_ Typename node is present.\n\n Haven't spend time to analyze if other places might have been\n affected by that - just thought we should trust the parser\n about the SETOF flag in TypeName node instead of knowing it\n better deep down in the utility commands.\n\n ;-)\n\n>\n> It looks like your patch will completely fix things, and is better than my\n> reverting the gram.y code. Can you suggest a small test case to include in\n> the regression suite?\n\n Small test case - hmmm.\n\n The regression tests found it - but you wouldn't expect it\n there. It's in the trigger test, where at some places SELECT\n set_ttdummy(0) returns 0 columns instead of one.\n\n Anyway - add a little function in regress.c returning a\n basetype value. Then add tests that use it in SELECT queries.\n\n int32\n allways_one()\n {\n return 1;\n }\n\n\n SELECT allways_one() AS one;\n\n SELECT a, allways_one() AS one FROM t;\n\n>\n> Unless there are objections from others (with a preference for reverting\n> the gram.y code) I'll go ahead and apply Jan's patch.\n\n Even if reverting the gram.y code - my patch could only make\n things better.\n\n\nUntil later, 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", "msg_date": "Fri, 13 Feb 1998 16:34:51 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] CREATE FUNCTION broken" }, { "msg_contents": "> > > Someone changed the parser to build a TypeName node on CREATE\n> > > FUNCTION in any case. As a side effect, ALL! functions\n> > > created got the proretset attribute to true. Thus for a\n> > > SELECT the parser wrapped an Iter node around the Expr and\n> > > since singleton functions set isDone the Iter returns no\n> > > tuple up.\n> > It looks like your patch will completely fix things, and is better than my\n> > reverting the gram.y code. Can you suggest a small test case to include in\n> > the regression suite?\n> The regression tests found it - but you wouldn't expect it\n> there. It's in the trigger test, where at some places SELECT\n> set_ttdummy(0) returns 0 columns instead of one.\n\nAh! This might be all of the problem with the trigger regression test then? I\nhad wanted Vadim to look at it because I wasn't sure what the behavior should\nbe. Does this test look good to you now?\n\n> Even if reverting the gram.y code - my patch could only make\n> things better.\n\nYes, and scrappy already applied it :)\n\n - Tom\n\n", "msg_date": "Fri, 13 Feb 1998 16:16:52 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CREATE FUNCTION broken" }, { "msg_contents": "\nTom wrote:\n>\n> > > > Someone changed the parser to build a TypeName node on CREATE\n> > > > FUNCTION in any case. As a side effect, ALL! functions\n> > > > created got the proretset attribute to true. Thus for a\n> > > > SELECT the parser wrapped an Iter node around the Expr and\n> > > > since singleton functions set isDone the Iter returns no\n> > > > tuple up.\n> > > It looks like your patch will completely fix things, and is better than my\n> > > reverting the gram.y code. Can you suggest a small test case to include in\n> > > the regression suite?\n> > The regression tests found it - but you wouldn't expect it\n> > there. It's in the trigger test, where at some places SELECT\n> > set_ttdummy(0) returns 0 columns instead of one.\n>\n> Ah! This might be all of the problem with the trigger regression test then? I\n> had wanted Vadim to look at it because I wasn't sure what the behavior should\n> be. Does this test look good to you now?\n\n Looks better now. In some places the triggers.sql has\n comments that an error is expected. And these errors now\n happen :-)\n\n But for the different NOTICE messages I get I'm not sure too.\n Who's the one who created the trigger test's (Vadim)? Could\n that guy please have a look at the results now and update the\n expected/triggers.out to what's really expected?\n\n In addition to the triggers I get a failed on the float8\n tests too. The test in question is\n\n SELECT '' AS bad, f.f1 ^ '1e200' from FLOAT8_TBL f;\n\n ====== float8 ======\n 200c200,208\n < ERROR: pow() result is out of range\n ---\n > bad\|?column?\n > ---+--------\n > \|0\n > \|NaN\n > \|NaN\n > \|NaN\n > \|NaN\n > (5 rows)\n >\n\n Content of table is\n\n QUERY: SELECT '' AS five, FLOAT8_TBL.;\n five\|f1\n ----+--------------------\n \|0\n \|1004.3\n \|-34.84\n \|1.2345678901234e+200\n \|1.2345678901234e-200\n (5 rows)\n\n What's correct on the overflow - NaN or ERROR?\n\n>\n> > Even if reverting the gram.y code - my patch could only make\n> > things better.\n>\n> Yes, and scrappy already applied it :)\n>\n> - Tom\n>\n>\n\n\nUntil later, 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", "msg_date": "Fri, 13 Feb 1998 17:53:44 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] CREATE FUNCTION broken" }, { "msg_contents": "I added typmod to the TypeName structure, but am not aware of adding any\nTypeName structure instance to anything.\n\nNice you have a patch for us.\n\n> \n> Hi,\n> \n> Someone changed the parser to build a TypeName node on CREATE\n> FUNCTION in any case. As a side effect, ALL! functions\n> created got the proretset attribute to true. Thus for a\n> SELECT the parser wrapped an Iter node around the Expr and\n> since singleton functions set isDone the Iter returns no\n> tuple up.\n> \n> \n> Until later, 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> * define.c.orig\tFri Feb 13 12:14:17 1998\n> --- define.c\tFri Feb 13 12:14:38 1998\n> ***********\n> * 94,100 ***\n> \t\tTypeName setType = (TypeName ) returnType;\n> \n> \t\tprorettype_p = setType->name;\n> ! \t\treturnsSet_p = true;\n> \t}\n> \telse\n> \t{\n> --- 94,100 ----\n> \t\tTypeName setType = (TypeName ) returnType;\n> \n> \t\tprorettype_p = setType->name;\n> ! \t\t*returnsSet_p = setType->setof;\n> \t}\n> \telse\n> \t{\n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 13:28:07 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CREATE FUNCTION broken" }, { "msg_contents": "Jan Wieck wrote:\n> \n> Tom wrote:\n> >\n> > Ah! This might be all of the problem with the trigger regression test then? I\n> > had wanted Vadim to look at it because I wasn't sure what the behavior should\n> > be. Does this test look good to you now?\n> \n> Looks better now. In some places the triggers.sql has\n> comments that an error is expected. And these errors now\n> happen :-)\n> \n> But for the different NOTICE messages I get I'm not sure too.\n> Who's the one who created the trigger test's (Vadim)? Could\n> that guy please have a look at the results now and update the\n> expected/triggers.out to what's really expected?\n\nI'll take a look...\n\nVadim\n", "msg_date": "Sun, 15 Feb 1998 16:26:40 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CREATE FUNCTION broken" } ]
[ { "msg_contents": "\n-----Original Message-----\nFrom: Vadim B. Mikheev <[email protected]>\nTo: Maurice Gittens <[email protected]>\nCc: [email protected] <[email protected]>\nDate: vrijdag 13 februari 1998 18:36\nSubject: Re: [HACKERS] Memory leaks?\n\n\n>Maurice Gittens wrote:\n>> \n>> Ok, I have an application which uses posgresql.\n>> \n>> While testing the system (inserting many records into\n>> a database (with large objects) the system ran out of memory\n>> (the system has 64M RAM + 128M SWAP). Sorting the processes\n>> according to memory usage showed me a postgres process with 160M of\n>> memory (+-80%) used.\n>\n\n>Version ?\n>OS ?\n>INSERTs are inside BEGIN/END ? Was -DTBL_FREE_CMD_MEMORY used if YES and\n>version is < 6.3 ?\n>\n>Vadim\n\nSorry, should have been more complete.\n\n- Postgresql Version: 6.3 (recent (this week) snapshop)\n- OS: Linux 2.0.33\n- INSERT's are within BEGIN/END brackets\n- TBL_FREE_CMD_MEMORY is defined in config.h (I guess it's the default).\n\nHope this helps.\n\n\n", "msg_date": "Fri, 13 Feb 1998 12:51:55 +0100", "msg_from": "\"Maurice Gittens\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Memory leaks?" }, { "msg_contents": "Maurice Gittens wrote:\n> \n> - Postgresql Version: 6.3 (recent (this week) snapshop)\n> - OS: Linux 2.0.33\n> - INSERT's are within BEGIN/END brackets\n> - TBL_FREE_CMD_MEMORY is defined in config.h (I guess it's the default).\n ^^^^^^^^^^^^^^^^ - in 6.3\nWell, this means that problem lies in large objects...\nI never used them... I assume that new relation and index\nis created for each LO. If this is right then we have memory leak\nin relation cache (heap_close doesn't flush relation from cache).\nAlso, as I see now, LO interface uses its own memory context:\nTBL_FREE_CMD_MEMORY can't help here -> all allocation must be\nfreed \"manually\".\n\nUnfortunately, I haven't time...\n\nVadim\n", "msg_date": "Fri, 13 Feb 1998 19:31:50 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Memory leaks?" } ]
[ { "msg_contents": "Hi,\n\nI just posted another set of patches. Please test this version a little bit\nbefore we release. I will be on the road for most of next week, but should\nbe back early enough to fix some bugs before the end of the month.\n\necpg should now be able to understand struct definitions as well as simple\ntypes. It even works with nested structs, but you have to put the struct\ndefinition into the sql declare section so ecpg can parse it. Do not use\ntypedefs, they won't work. Neither will #define's on datatypes.\n\nI checked the output manually and it looks good. gcc agrees that my test\nprograms have a correct form. But I cannot test whether they run correctly\nas I cannot even create a user right now.\n\nMichael\n\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Fri, 13 Feb 1998 14:30:05 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "actual ecpg version" } ]
[ { "msg_contents": "> > Modified Files:\n> > linux-elf\n> > Log Message:\n> >\n> > Remove -m486 option, since it is platform specific...this should technically\n> > be removed everywhere, but so far linux-elf is the only platform that is\n> > being reported a problem with...\n>\n> Hmm. I don't think there was a problem reported, only a complaint that the\n> -m486 was suboptimal for 586 and 686 machines. The proposed solution was not in\n> keeping with the Postgres philosophy of not requiring root privilege for an\n> install, since it involved modifying or adding /usr/lib compiler settings\n> files. With the flag removed, standard compiler settings will generate 386\n> code. Yuck...\n\nAndrew, could you put a new entry into the Linux FAQ suggesting that\n\n CFLAGS+= -m486\n\nor (for the new compilers that most people are not yet running)\n\n CFLAGS+= -mpentium\nor\n CFLAGS+= -mpentiumpro\n\nbe added to Makefile.custom? Otherwise people are going to get 386 code generation\nfor most installations...\n\n - Tom\n\n", "msg_date": "Fri, 13 Feb 1998 13:36:23 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [COMMITTERS] 'pgsql/src/template linux-elf'" } ]
[ { "msg_contents": "This is some implementation notes and opened issues...\n\nFirst, implementation uses new type of parameters - PARAM_EXEC - to deal\nwith correlation Vars. When query_planner() is called, it first tries to\nreplace all upper queries Var referenced in current query with Param of \nthis type. Some global variables are used to keep mapping of Vars to \nParams and Params to Vars. \n\nAfter this, all current query' SubLinks are processed: for each SubLink \nfound in query' qual union_planner() (old planner() function) will be \ncalled to plan corresponding subselect (union_planner() calls \nquery_planner() for \"simple\" query and supports UNIONs). After subselect \nare planned, optimizer knows about is this correlated, un-correlated or \n_undirect_ correlated (references some grand-parent Vars but no parent \nones: uncorrelated from the parent' point of view) query. \n\nFor uncorrelated and undirect correlated subqueries of EXPRession or \nEXISTS type SubLinks will be replaced with \"normal\" clauses from\nSubLink->Oper list (I changed this list to be list of EXPR nodes,\nnot just Oper ones). Right sides of these nodes are replaced with\nPARAM_EXEC parameters. This is second use of new parameter type.\nAt run-time these parameters get value from result of subquery\nevaluation (i.e. - from target list of subquery). Execution plan of\nsubquery itself becomes init plan of parent query. InitPlan knows\nwhat parameters are to get values from subquery' results and will be\nexecuted \"on-demand\" (for query select * from table where x > 0 and\ny > (select max(a) from table_a) subquery will not be executed at all\nif there are no tuples with x > 0 _and_ y is not used in index scan).\n\nSubLinks for subqueries of all other types are transformed into\nnew type of Expr node - SUBPLAN_EXPR. Expr->args are just correlation\nvariables from _parent_ query. Expr->oper is new SubPlan node.\n\nThis node is used for InitPlan too. It keeps subquery range table,\nindices of Params which are to get value from _parent_ query Vars\n(i.e. - from Expr->args), indices of Params into which subquery'\nresults are to be substituted (this is for InitPlans), SubLink\nand subquery' execution plan.\n\nPlan node was changed to know about dependencies on Params from\nparent queries and InitPlans, to keep list of changed Params\n(from the above) and so be re-scanned if this list is not NULL.\nAlso, added list of InitPlans (actually, all of them for current\nquery are in topmost plan node now) and other SubPlans (from\nplan->qual) - to initialize them and let them know about changed\nParams (from the list of their \"interests\").\n\nAfter all SubLinks are processed, query_planner() calls qual'\ncanonificator and does \"normal\" work. By using Params optimizer \nis mostly unchanged.\n\nWell, Executor. To get subplans re-evaluated without ExecutorStart()\nand ExecutorEnd() (without opening and closing relations and indices\nand without many palloc() and pfree() - this is what SQL-funcs does\non each call) ExecReScan() now supports most of Plan types...\n\nExplanation of EXPLAIN.\n\nvac=> explain select * from tmp where x >= (select max(x2) from test2 \nwhere y2 = y and exists (select * from tempx where tx = x));\nNOTICE: QUERY PLAN:\n\nSeq Scan on tmp (cost=40.03 size=101 width=8)\n SubPlan\n ^^^^^^^ subquery is in Seq Scan' qual, its plan is below\n -> Aggregate (cost=2.05 size=0 width=0)\n InitPlan\n ^^^^^^^^ EXISTS subsubquery is InitPlan of subquery\n -> Seq Scan on tempx (cost=4.33 size=1 width=4)\n -> Result (cost=2.05 size=0 width=0)\n ^^^^^^ EXISTS subsubquery was transformed into Param\n and so we have Result node here\n -> Index Scan on test2 (cost=2.05 size=1 width=4)\n\n\nOpened issues.\n\n1. No read permissions checking (easy, just not done yet).\n2. readfuncs.c can't read subplan-s (easy, not critical, because of\n we currently nowhere use ascii representation of execution plans).\n3. ExecReScan() doesn't support all plan types. At least support for\n MergeJoin has to be implemented.\n4. Memory leaks in ExecReScan().\n5. I need in advice: if subquery introduced with NOT IN doesn't return\n any tuples then qualification is failed, yes ?\n6. Regression tests !!!!!!!!!!!!!!!!!!!!\n (Could we use data/queries from MySQL' crash.me ?\n Copyright-ed ? Could they give us rights ?)\n7. Performance.\n - Should be good when subquery is transformed into InitPlan.\n - Something should be done for uncorrelated subqueries introduced\n with ANY/ALL - keep thinking. Currently, subplan will be re-scanned\n for each parent tuple - very slow...\n\nResults of some test. TMP is table with x,y (int4-s), x in 0-9,\ny = 100 - x, 1000 tuples (10 duplicates of each tuple). TEST2 is table \nwith x2, y2 (int4-s), x2 in 1-99, y2 = 100 -x2, 10000 tuples (100 dups).\n\n Trying \n\nselect * from tmp where x >= (select max(x2) from test2 where y2 = y);\n \n and\n\nbegin;\nselect y as ty, max(x2) as mx into table tsub from test2, tmp \nwhere y2 = y group by ty;\nvacuum tsub;\nselect x, y from tmp, tsub where x >= mx and y = ty;\ndrop table tsub;\nend;\n\n Without index on test2(y2):\n\nSubSelect -> 320 sec\nUsing temp table -> 32 sec\n\n Having index\n\nSubSelect -> 17 sec (2M of memory)\nUsing temp table -> 32 sec (12M of memory: -S 8192)\n\nVadim\n", "msg_date": "Fri, 13 Feb 1998 20:50:50 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": true, "msg_subject": "Subselects are in CVS..." }, { "msg_contents": "Vadim B. Mikheev writes:\n> 5. I need in advice: if subquery introduced with NOT IN doesn't return\n> any tuples then qualification is failed, yes ?\n\nDo you mean something like this:\n\nselect * from table1 where x not in (select x from table2)\n\ntable1.x: a,b\n\ntable2.x is empty\n\nThe correct answer IMO is 'a,b' in this case.\n\nMichael\n\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Fri, 13 Feb 1998 14:53:29 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Subselects are in CVS..." }, { "msg_contents": "Impressive, just one comment:\n\n> 5. I need in advice: if subquery introduced with NOT IN doesn't return\n> any tuples then qualification is failed, yes ?\n\nI think this is backwards. Consider the following statements:\n\n\"1 not in (1,2)\" FALSE\n\"1 not in (2)\" TRUE\n\"1 not in ()\" ?\n\nI would tend to think that the ? should be TRUE. I.E. every value\nqualifies as not in an empty list of tuples. The second item was\ntrue, so why should removing the 2 from the list make the statement\nfalse?\n\nOcie Mitchell\n\n", "msg_date": "Fri, 13 Feb 1998 10:14:34 -0800 (PST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Subselects are in CVS..." }, { "msg_contents": "Want to add this as a README in the optimizer/planner directory?\n\n> \n> This is some implementation notes and opened issues...\n> \n> First, implementation uses new type of parameters - PARAM_EXEC - to deal\n> with correlation Vars. When query_planner() is called, it first tries to\n> replace all upper queries Var referenced in current query with Param of \n> this type. Some global variables are used to keep mapping of Vars to \n> Params and Params to Vars. \n> \n> After this, all current query' SubLinks are processed: for each SubLink \n> found in query' qual union_planner() (old planner() function) will be \n> called to plan corresponding subselect (union_planner() calls \n> query_planner() for \"simple\" query and supports UNIONs). After subselect \n> are planned, optimizer knows about is this correlated, un-correlated or \n> _undirect_ correlated (references some grand-parent Vars but no parent \n> ones: uncorrelated from the parent' point of view) query. \n> \n> For uncorrelated and undirect correlated subqueries of EXPRession or \n> EXISTS type SubLinks will be replaced with \"normal\" clauses from\n> SubLink->Oper list (I changed this list to be list of EXPR nodes,\n> not just Oper ones). Right sides of these nodes are replaced with\n> PARAM_EXEC parameters. This is second use of new parameter type.\n> At run-time these parameters get value from result of subquery\n> evaluation (i.e. - from target list of subquery). Execution plan of\n> subquery itself becomes init plan of parent query. InitPlan knows\n> what parameters are to get values from subquery' results and will be\n> executed \"on-demand\" (for query select * from table where x > 0 and\n> y > (select max(a) from table_a) subquery will not be executed at all\n> if there are no tuples with x > 0 _and_ y is not used in index scan).\n> \n> SubLinks for subqueries of all other types are transformed into\n> new type of Expr node - SUBPLAN_EXPR. Expr->args are just correlation\n> variables from _parent_ query. Expr->oper is new SubPlan node.\n> \n> This node is used for InitPlan too. It keeps subquery range table,\n> indices of Params which are to get value from _parent_ query Vars\n> (i.e. - from Expr->args), indices of Params into which subquery'\n> results are to be substituted (this is for InitPlans), SubLink\n> and subquery' execution plan.\n> \n> Plan node was changed to know about dependencies on Params from\n> parent queries and InitPlans, to keep list of changed Params\n> (from the above) and so be re-scanned if this list is not NULL.\n> Also, added list of InitPlans (actually, all of them for current\n> query are in topmost plan node now) and other SubPlans (from\n> plan->qual) - to initialize them and let them know about changed\n> Params (from the list of their \"interests\").\n> \n> After all SubLinks are processed, query_planner() calls qual'\n> canonificator and does \"normal\" work. By using Params optimizer \n> is mostly unchanged.\n> \n> Well, Executor. To get subplans re-evaluated without ExecutorStart()\n> and ExecutorEnd() (without opening and closing relations and indices\n> and without many palloc() and pfree() - this is what SQL-funcs does\n> on each call) ExecReScan() now supports most of Plan types...\n> \n> Explanation of EXPLAIN.\n> \n> vac=> explain select * from tmp where x >= (select max(x2) from test2 \n> where y2 = y and exists (select * from tempx where tx = x));\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on tmp (cost=40.03 size=101 width=8)\n> SubPlan\n> ^^^^^^^ subquery is in Seq Scan' qual, its plan is below\n> -> Aggregate (cost=2.05 size=0 width=0)\n> InitPlan\n> ^^^^^^^^ EXISTS subsubquery is InitPlan of subquery\n> -> Seq Scan on tempx (cost=4.33 size=1 width=4)\n> -> Result (cost=2.05 size=0 width=0)\n> ^^^^^^ EXISTS subsubquery was transformed into Param\n> and so we have Result node here\n> -> Index Scan on test2 (cost=2.05 size=1 width=4)\n> \n> \n> Opened issues.\n> \n> 1. No read permissions checking (easy, just not done yet).\n> 2. readfuncs.c can't read subplan-s (easy, not critical, because of\n> we currently nowhere use ascii representation of execution plans).\n> 3. ExecReScan() doesn't support all plan types. At least support for\n> MergeJoin has to be implemented.\n> 4. Memory leaks in ExecReScan().\n> 5. I need in advice: if subquery introduced with NOT IN doesn't return\n> any tuples then qualification is failed, yes ?\n> 6. Regression tests !!!!!!!!!!!!!!!!!!!!\n> (Could we use data/queries from MySQL' crash.me ?\n> Copyright-ed ? Could they give us rights ?)\n> 7. Performance.\n> - Should be good when subquery is transformed into InitPlan.\n> - Something should be done for uncorrelated subqueries introduced\n> with ANY/ALL - keep thinking. Currently, subplan will be re-scanned\n> for each parent tuple - very slow...\n> \n> Results of some test. TMP is table with x,y (int4-s), x in 0-9,\n> y = 100 - x, 1000 tuples (10 duplicates of each tuple). TEST2 is table \n> with x2, y2 (int4-s), x2 in 1-99, y2 = 100 -x2, 10000 tuples (100 dups).\n> \n> Trying \n> \n> select * from tmp where x >= (select max(x2) from test2 where y2 = y);\n> \n> and\n> \n> begin;\n> select y as ty, max(x2) as mx into table tsub from test2, tmp \n> where y2 = y group by ty;\n> vacuum tsub;\n> select x, y from tmp, tsub where x >= mx and y = ty;\n> drop table tsub;\n> end;\n> \n> Without index on test2(y2):\n> \n> SubSelect -> 320 sec\n> Using temp table -> 32 sec\n> \n> Having index\n> \n> SubSelect -> 17 sec (2M of memory)\n> Using temp table -> 32 sec (12M of memory: -S 8192)\n> \n> Vadim\n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 15:57:34 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Subselects are in CVS..." }, { "msg_contents": "Bruce Momjian wrote:\n> \n> Want to add this as a README in the optimizer/planner directory?\n> \n> >\n> > This is some implementation notes and opened issues...\n\nI'll do.\n\nVadim\n", "msg_date": "Sun, 15 Feb 1998 18:41:29 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Subselects are in CVS..." }, { "msg_contents": "Added to backend/optimizer/plan/README.\n\n> \n> This is some implementation notes and opened issues...\n> \n> First, implementation uses new type of parameters - PARAM_EXEC - to deal\n> with correlation Vars. When query_planner() is called, it first tries to\n> replace all upper queries Var referenced in current query with Param of \n> this type. Some global variables are used to keep mapping of Vars to \n> Params and Params to Vars. \n> \n> After this, all current query' SubLinks are processed: for each SubLink \n> found in query' qual union_planner() (old planner() function) will be \n> called to plan corresponding subselect (union_planner() calls \n> query_planner() for \"simple\" query and supports UNIONs). After subselect \n> are planned, optimizer knows about is this correlated, un-correlated or \n> _undirect_ correlated (references some grand-parent Vars but no parent \n> ones: uncorrelated from the parent' point of view) query. \n> \n> For uncorrelated and undirect correlated subqueries of EXPRession or \n\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Mon, 16 Mar 1998 00:37:00 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Subselects are in CVS..." } ]
[ { "msg_contents": "> From: Gerhard Reithofer <[email protected]>\n> ...\n> Subject: [HACKERS] Unsubscribe\n> \n> unsubscribe [email protected]\n> end\n\nAnother aix user leaving the list...\n\ndarrenk\n", "msg_date": "Fri, 13 Feb 1998 10:00:49 -0500", "msg_from": "[email protected] (Darren King)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Unsubscribe" }, { "msg_contents": "> Another aix user leaving the list...\n\nWhew. I saw the subject and the sender and thought \"Oh no! Darren's\nleaving. There goes our AIX support...\".\n\nAnyway, can you confirm or reconfirm or coordinate confirming which AIX\nsystems are supported for v6.3?\n\n", "msg_date": "Fri, 13 Feb 1998 16:22:32 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Unsubscribe" }, { "msg_contents": "On Fri, 13 Feb 1998, Darren King wrote:\n\n> > From: Gerhard Reithofer <[email protected]>\n> > ...\n> > Subject: [HACKERS] Unsubscribe\n> > \n> > unsubscribe [email protected]\n> > end\n> \n> Another aix user leaving the list...\n\n\tWhy?\n\n\n", "msg_date": "Fri, 13 Feb 1998 12:10:20 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Unsubscribe" } ]
[ { "msg_contents": ">Well, this means that problem lies in large objects...\n>I never used them... I assume that new relation and index\n>is created for each LO. If this is right then we have memory leak\n>in relation cache (heap_close doesn't flush relation from cache).\n>Also, as I see now, LO interface uses its own memory context:\n>TBL_FREE_CMD_MEMORY can't help here -> all allocation must be\n>freed \"manually\".\n>\n>Unfortunately, I haven't time...\n>\n\n\nOK, part of the problem might be the lo_create function which\ncreates a large object using the functions inv_create and\ninv_close. \n\ninv_create uses TupleDescInitEntry to create tuple descriptors (I think)\nbut it never seems to free these tuple descriptors.\n\nThe function FreeTupleDesc seems to have been created for this purpose.\nAs I'm new to postgresql this I'm wondering if I'm on the right track.\n\nAnyway I'll give it a try and see if the problem goes a way.\n\nRegards,\n Maurice\n\n", "msg_date": "Fri, 13 Feb 1998 16:28:39 +0100", "msg_from": "\"Maurice Gittens\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Memory leaks?" } ]
[ { "msg_contents": "I'm a Linux developer that are making a huge server, and i'm trying out\ndatabases.\nAre the postgres librarys not thread-safe?\nIf not, when will it be?\n\nSincerly Magnus Nï¿½slund\n\n\n", "msg_date": "Fri, 13 Feb 1998 17:16:31 +0100", "msg_from": "\"Magnus N��slund\" <[email protected]>", "msg_from_op": true, "msg_subject": "pgsql and threads don't match" }, { "msg_contents": "Magnus Näslund wrote:\n\n> I'm a Linux developer that are making a huge server, and i'm trying out\n> databases.\n> Are the postgres librarys not thread-safe?\n> If not, when will it be?\n\nThe Postgres backend is single-threaded software. The client stuff is\nprobably thread-safe, and both frontend and backend work with thread-safe\nlibraries. Clients can open multiple connections to backends, so I would\nthink that you could do so from multiple threads. You might need to take a\nmutex on the connection-opening step; not sure about that. Also, there is\na separate process per backend connection, so multiple simultaneous\nconnections are allowed.\n\nGood luck.\n\n - Tom\n\n", "msg_date": "Fri, 13 Feb 1998 17:02:14 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] pgsql and threads don't match" }, { "msg_contents": "On Fri, 13 Feb 1998, Magnus Nslund wrote:\n\n> I'm a Linux developer that are making a huge server, and i'm trying out\n> databases.\n> Are the postgres librarys not thread-safe?\n> If not, when will it be?\n\nI'm assuming you mean the C libraries?\n\nI can't say about them, but the Java/JDBC is _not_ thread safe - yet. It's\none of the 4 main things to work on once 6.3 is released.\n\n-- \nPeter T Mount [email protected] or [email protected]\nMain Homepage: http://www.demon.co.uk/finder\nWork Homepage: http://www.maidstone.gov.uk Work EMail: [email protected]\n\n", "msg_date": "Sat, 14 Feb 1998 10:54:29 +0000 (GMT)", "msg_from": "Peter T Mount <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] pgsql and threads don't match" } ]
[ { "msg_contents": "> \tSomeone want to send me a patch to fix this? or a list of files\n> that need to be changed?\n\nOnly occurrs in\nsrc/include/storage/s_lock.h:#if defined(__AIX)\nsrc/include/utils/dt.h:#if defined(__AIX)\nsrc/include/utils/nabstime.h:#if defined(__AIX)\n\nSimply delete one underscore, only occurs once per file, so no patch.\n\nThanks, I don't have CVS on AIX here \nAndreas\n", "msg_date": "Fri, 13 Feb 1998 17:23:38 +0100", "msg_from": "Zeugswetter Andreas SARZ <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] AIX port s_lock.h __AIX --> _AIX" }, { "msg_contents": "On Fri, 13 Feb 1998, Zeugswetter Andreas SARZ wrote:\n\n> > \tSomeone want to send me a patch to fix this? or a list of files\n> > that need to be changed?\n> \n> Only occurrs in\n> src/include/storage/s_lock.h:#if defined(__AIX)\n> src/include/utils/dt.h:#if defined(__AIX)\n> src/include/utils/nabstime.h:#if defined(__AIX)\n> \n> Simply delete one underscore, only occurs once per file, so no patch.\n\n\tPerfect, change is done...\n\n", "msg_date": "Fri, 13 Feb 1998 12:12:09 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] AIX port s_lock.h __AIX --> _AIX" } ]
[ { "msg_contents": "> 5. I need in advice: if subquery introduced with NOT IN doesn't return\n> any tuples then qualification is failed, yes ?\n\nInformix treats the subselect as NULL if no rows are returned.\nTherefore all parent rows that are not null are returned.\n\nselect * from taba where a not in (<a select returning no row>); --\nis same as\nselect * from taba where a is not null;\n\nAndreas\n", "msg_date": "Fri, 13 Feb 1998 18:00:25 +0100", "msg_from": "Zeugswetter Andreas SARZ <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Subselects open issue Nr. 5" }, { "msg_contents": "Michael Meskes wrote:\n> \n> > 5. I need in advice: if subquery introduced with NOT IN doesn't return\n> > any tuples then qualification is failed, yes ?\n> \n> Do you mean something like this:\n> \n> select * from table1 where x not in (select x from table2)\n> \n> table1.x: a,b\n> \n> table2.x is empty\n> \n> The correct answer IMO is 'a,b' in this case.\n\nOk. I'll fix this. As I see, this is exactly what Oracle 6 does, but\n\nZeugswetter Andreas SARZ wrote:\n> \n> Informix treats the subselect as NULL if no rows are returned.\n> Therefore all parent rows that are not null are returned.\n> \n> select * from taba where a not in (<a select returning no row>); --\n> is same as\n> select * from taba where a is not null;\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\nOracle returns tuples with A being NULL!!! and more of that (table B is empty):\n\nSQL> select count() from a where x > ALL (select from b);\n\n COUNT()\n----------\n 2\n\nand result is the same for all OP-s with ALL modifier... And\n\nSQL> select count() from a where x in (select * from b);\n\n COUNT(*)\n----------\n 0\nhaving tuple with NULL in X...\n\nWho's right ?\nWhat standard says ?\n\nVadim\n", "msg_date": "Sun, 15 Feb 1998 18:33:09 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Subselects open issue Nr. 5" } ]
[ { "msg_contents": "> On Fri, 13 Feb 1998, Darren King wrote:\n> \n> > > From: Gerhard Reithofer <[email protected]>\n> > > ...\n> > > Subject: [HACKERS] Unsubscribe\n> > > \n> > > unsubscribe [email protected]\n> > > end\n> > \n> > Another aix user leaving the list...\n> \n> \tWhy?\n> \n\nSeems the account has disappeared. The cc of the above message\nto his address bounced back.\n\nToo bad...he seemed to be the only one looking after the use\nof gcc for the aix ports.\n\ndarrenk\n", "msg_date": "Fri, 13 Feb 1998 12:18:20 -0500", "msg_from": "[email protected] (Darren King)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Unsubscribe" }, { "msg_contents": "On Fri, 13 Feb 1998, Darren King wrote:\n\n> > On Fri, 13 Feb 1998, Darren King wrote:\n> > \n> > > > From: Gerhard Reithofer <[email protected]>\n> > > > ...\n> > > > Subject: [HACKERS] Unsubscribe\n> > > > \n> > > > unsubscribe [email protected]\n> > > > end\n> > > \n> > > Another aix user leaving the list...\n> > \n> > \tWhy?\n> > \n> \n> Seems the account has disappeared. The cc of the above message\n> to his address bounced back.\n> \n> Too bad...he seemed to be the only one looking after the use\n> of gcc for the aix ports.\n\n\tLet's hope that he re-subscribes using a different account...:(\n\n\n", "msg_date": "Fri, 13 Feb 1998 12:43:31 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Unsubscribe" } ]
[ { "msg_contents": "> \n> > Another aix user leaving the list...\n> \n> Whew. I saw the subject and the sender and thought \"Oh no! Darren's\n> leaving. There goes our AIX support...\".\n\nThere are still a few active aix people on the lists. One's that pop\nto mind right now...Andreas...is Frank Dana still around? Who else?\n\nI remember seeing aix in a few bug & install reports the last year or two.\n\n> Anyway, can you confirm or reconfirm or coordinate confirming which AIX\n> systems are supported for v6.3?\n\nI'm using 4.1.4.0 successfully. Frank was keeping 3.2.5 alive/kicking.\nOne of the install/bug reports during the last year was for 4.2 I think.\nI've run the 4.1.4.0 binary on a 4.2 system, but have yet to compile a\n4.2 binary.\n\nOne of these days I might even get the chance to compile and run postgres\non a 4.3 system, which is IBM's 64-bit release.\n\nSo...\n\nAIX 3.2.5 - Frank?\n 4.1.4.0 - Working ok.\n 4.2 + - Anyone?\n \ndarrenk\n", "msg_date": "Fri, 13 Feb 1998 12:18:37 -0500", "msg_from": "[email protected] (Darren King)", "msg_from_op": true, "msg_subject": "aix systems for 6.3" } ]
[ { "msg_contents": "Are you working on an initdb option for passwords, so we don't have\npg_user world-unreadable by default?\n\n> \n> > What, pg_user is not readable by world anymore? This could be a problem.\n> \n> It has to be this way, otherwise it would be possible for user to see other\n> users' passwords in pg_user. I spoke to you all about this when I first started.\n> I was going to make a separate relation (pg_password), but I was convinced not\n> to since there is a one to one correlation between users and passwords. At this\n> point I sent email to the effect that pg_user could no longer be readable by\n> the group 'public'. If it was readable by public, then the passwords would have\n> to be encrypted in pg_user. If this is the case, then the frontends will have\n> to pass an unencrypted password over the network. Again this degrades the\n> security of PostgreSQL.\n> \n> The real solution to this problem would be to create a pg_privileges relation,\n> overhauling the privileges system entirely. Then we could just restrict access\n> to the password column of pg_user. However, I would suggest that the entire\n> pg_privileges table be cached in shared memory to speed things up. I am unsure\n> if the catalog table are cached in shared memory or not (They really should be,\n> but then this would probably require some logging to files in case of system\n> crash).\n> \n> In the meantime, there should really be nothing that the average user will need\n> from pg_user. The '\\d' is the only problem I have encountered thus far, and I\n> hope to solve that problem soon. Therefore, if you really, really need something\n> from pg_user, then you need to have select privileges given to you explicitly,\n> or you could explicitly give them to public. This would, however, give public\n> the ability to see user passwords (If you are using, HBA only, then just give\n> public the select over pg_user).\n> \n> Todd A. Brandys\n> [email protected]\n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Fri, 13 Feb 1998 15:11:01 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: New pg_pwd patch and stuff" } ]
[ { "msg_contents": "\nOn Thu, 12 Feb 1998, The Hermit Hacker wrote:\n\n> On Thu, 12 Feb 1998, Peter T Mount wrote:\n> \n> > Getting most of the rest of DatabaseMetaData is my main priority after 6.3\n> > is released. However, there are some methods that cannot yet be\n> > implemented because postgresql itself isn't yet up to it.\n> \n> \tLike? Maybe you should coordinate with Bruce, as far as the TODO\n> list is concerned, and have a JDBC section, that has notes attached to\n> those that require something from the PostgreSQL section? Or annotate te\n> PostgreSQL section pieces with \"needed for JDBC driver\"?\n\nThere's quite a few.\n\nBelow is a list of what I think we still have to work on.\n\nMost of the text here are notes from the source. Most of these were put in\nby either myself, or Adrian. \n\nDatabaseMetaData:\n\n\tWe have some methods that should return the available functions. \n\tThey are:\n\n\t\tgetNumericFunctions, getStringFunctions,\n\t\tgetSystemFunctions, getTimeDateFunctions\n\n\tOk, our functions are in pg_proc, but how to differenciate them?\n\n\tsupportsTableCorrelationNames()\n\n\tAre table correlation names supported? A JDBC-Compliant driver\n\talways returns true. We return false here. Do we support this?\n\n\tsupportsCoreSQLGrammar()\n\n\tDoes this driver support the Core ODBC SQL grammar. We need SQL-92\n\tconformance for this. We return false here.\n\n\tsupportsExtendedSQLGrammar()\n\n\tDoes this driver support the Extended (Level 2) ODBC SQL grammar. \n\tWe don't conform to the Core (Level 1), so we can't conform to the\n\tExtended SQL Grammar.\n\n\tsupportsANSI92EntryLevelSQL()\n\n\tDoes this driver support the ANSI-92 entry level SQL grammar? \n\tAll JDBC Compliant drivers must return true. I think we have to\n\tsupport outer joins for this to be true. \n\nPreparedStatement:\n\n\tThis e A SQL statement is pre-compiled and stored in a\n\tPreparedStatement object. This object can then be used to\n\tefficiently execute this statement multiple times. \n\n\tWhat we actually do in the driver, is store the statement, then on\n\teach execute, builds a query, then sends it to the backend.\n\n\tI think some one recently (last week) said they were doing\n\tsomething with the parser to store pre-parsed statements?\n\nCallableStatement: This we don't implement at all! As this uses stored\nprocedures, do we support them, and if so, do we have some examples?\n\nThis is from the JDBC API docs: \n\nCallableStatement is used to execute SQL stored procedures. \n \n JDBC provides a stored procedure SQL escape that allows stored\n procedures to be called in a standard way for all RDBMS's. This escape\n syntax has one form that includes a result parameter and one that does\n not. If used, the result parameter must be registered as an OUT \n parameter. The other parameters may be used for input, output or both.\n Parameters are refered to sequentially, by number. The first parameter\n is 1. \n \n {?= call <procedure-name>[<arg1>,<arg2>, ...]} \n {call <procedure-name>[<arg1>,<arg2>, ...]}\n \n \n IN parameter values are set using the set methods inherited from\n PreparedStatement. The type of all OUT parameters must be registered\n prior to executing the stored procedure; their values are retrieved\n after execution via the get methods provided here. \n \n A Callable statement may return a ResultSet or multiple ResultSets.\n Multiple ResultSets are handled using operations inherited from \n Statement. \n \n For maximum portability, a call's ResultSets and update counts should\n be processed prior to getting the values of output parameters.\n\n-- \nPeter T Mount [email protected] or [email protected]\nMain Homepage: http://www.demon.co.uk/finder\nWork Homepage: http://www.maidstone.gov.uk Work EMail: [email protected]\n\n\n", "msg_date": "Fri, 13 Feb 1998 22:08:37 +0000 (GMT)", "msg_from": "Peter T Mount <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [QUESTIONS] full JDBC driver ?" }, { "msg_contents": "Peter T Mount wrote:\n> \n> On Thu, 12 Feb 1998, The Hermit Hacker wrote:\n> \n> > On Thu, 12 Feb 1998, Peter T Mount wrote:\n> >\n> > > Getting most of the rest of DatabaseMetaData is my main priority after 6.3\n> > > is released. However, there are some methods that cannot yet be\n> > > implemented because postgresql itself isn't yet up to it.\n> >\n> > Like?\n\nIsn't the \"right\" way to do this to implement the standard\nINFORMATION_SCHEMA, then all frontends have a decent model to work to?\n\nPhil\n", "msg_date": "Sat, 14 Feb 1998 00:05:24 +0000", "msg_from": "Phil Thompson <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [QUESTIONS] full JDBC driver ?" }, { "msg_contents": "On Sat, 14 Feb 1998, Phil Thompson wrote:\n\n> Peter T Mount wrote:\n> > \n> > On Thu, 12 Feb 1998, The Hermit Hacker wrote:\n> > \n> > > On Thu, 12 Feb 1998, Peter T Mount wrote:\n> > >\n> > > > Getting most of the rest of DatabaseMetaData is my main priority after 6.3\n> > > > is released. However, there are some methods that cannot yet be\n> > > > implemented because postgresql itself isn't yet up to it.\n> > >\n> > > Like?\n> \n> Isn't the \"right\" way to do this to implement the standard\n> INFORMATION_SCHEMA, then all frontends have a decent model to work to?\n> \n> Phil\n\nPossibly. Where possible, we have either default or assumed values where\nthere is no (visible) way of getting these values.\n\nI did miss one bit out of the last post: There is a problem with\ngetColumnPrivileges(), in that it will never work while pg_user is not\nreadable by public.\n\nI know that someone (vadim?) was trying to work out where to store the\npasswords (other than in pg_user).\n\nAnyhow, now I've finished the courses my employer keeps sending me on, I\ncan now spend more time on this ;-)\n\n-- \nPeter T Mount [email protected] or [email protected]\nMain Homepage: http://www.demon.co.uk/finder\nWork Homepage: http://www.maidstone.gov.uk Work EMail: [email protected]\n\n", "msg_date": "Sat, 14 Feb 1998 11:03:54 +0000 (GMT)", "msg_from": "Peter T Mount <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [QUESTIONS] full JDBC driver ?" } ]
[ { "msg_contents": "Hi All,\n\nI know I'm not capable of locating or fixing the problem with\nthis so it's a it of a cheek asking but...\n\nIs it time to increase the priority of the TODO item:-\n\n* Allow indexes to be used with OR clauses\n\nThis would seem particularly important with the new subselect\nfunctionality. (Great work vadim)\n\nMany subselect-using queries use the \"IN\" keyword which we\ntranslate to a string of \"OR\"s so could potentially benefit\nfrom index useage.\n\nKeith.\n\n", "msg_date": "Fri, 13 Feb 1998 23:45:04 +0000 (GMT)", "msg_from": "Keith Parks <[email protected]>", "msg_from_op": true, "msg_subject": "Indexes and OR clauses." } ]
[ { "msg_contents": "tip1: I find many string functions calls like strcpy, strncpy\n, strcat etc.. in the code.\n\nIs not faster if you use memcpy, memset, sizeof() functions calls? \nEspecially watch those areas inside the for loop and while loop.\n\ntip 2: Inside for loops, where functions are called\nit will be faster to declare and assign in one step\nlike\n int ii=22;\ninstead of \n int ii;\n ii = 22;\n\n\n\n\n\n_________________________________________________________\nDO YOU YAHOO!?\nGet your free @yahoo.com address at http://mail.yahoo.com\n\n", "msg_date": "Fri, 13 Feb 1998 16:13:10 -0800 (PST)", "msg_from": "al dev <[email protected]>", "msg_from_op": true, "msg_subject": "performace speed up suggestions." } ]
[ { "msg_contents": "Could somebody apply this little patch for me? (don't have access to CVS)\nThanks!\n\nkeep up the great work!\nCheers,\n-Pailing\n\n--- contrib/spi/refint.c.old\tSat Feb 14 00:04:15 1998\n+++ contrib/spi/refint.c\tSat Feb 14 00:04:42 1998\n@@ -423,7 +423,7 @@\n \t\t\t\tsprintf(sql, \"update %s set \", relname);\n \t\t\t\tfor (i = 1; i <= nkeys; i++)\n \t\t\t\t{\n-\t\t\t\t\tsprintf(sql + strlen(sql), \"%s null%s\",\n+\t\t\t\t\tsprintf(sql + strlen(sql), \"%s = null%s\",\n \t\t\t\t\t\t\targs2[i], (i < nkeys) ? \", \" : \"\");\n \t\t\t\t}\n \t\t\t\tstrcat(sql, \" where \");\n", "msg_date": "14 Feb 1998 00:26:56 -0500", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "a little patch to contrib/spi/refint.c" } ]
[ { "msg_contents": "At 4:08 PM 98.2.13 +0000, Thomas G. Lockhart wrote:\n>Bruce, are you planning on keeping a ToDo list for the 6.3 release as\n>you have in (at least) the past few releases? That will make sure that\n>we don't leave something out. As a start:\n\n>Test big-endian/little-endian operation(Tatsuo?) I don' know if we need\n>this...\n\nOk. I can do it. By the way, These are platforms I have access to:\nFreeBSD 2.2.1-RELEASE\nFreeBSD 2.2.5-RELEASE\nSparc/Solaris 2.6\nSparc/SunOS 4.1.4\nPPC/MkLinux\n\nI will run the regression tests on all of them.\n\n>_ ppc/mklinux(Tatsuo?)\n\nI have already tested the 2/11 snapshot. I found a small problem regarding\ntemplate (already reported to hackers list). Another more serious problem\nwas the trigger regression test. Simply the backend died. I will test more\nrecent snapshot.\n--\nTatsuo Ishii\[email protected]\n---\nTatsuo Ishii\[email protected]\n\n\n", "msg_date": "Sat, 14 Feb 1998 23:12:14 +0900", "msg_from": "[email protected] (Tatsuo Ishii)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] v6.3 release ToDo list and supported ports" } ]
[ { "msg_contents": "The 'configure' script currently says\n\n\tchecking setting HBA... enabled\n\nwhich is hardly needed any more. I just grepped through the whole\ntree, and there is no occurrence of \"NOHBA\" outside of configure -- as\nthere shouldn't be now that the pg_hba.conf file is used to select the\nauthentication scheme(s) to be used. (Which, by the way, was a Smart\nMove, in my opinion. Things were too confusing as they were.)\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "Sat, 14 Feb 1998 15:19:55 +0100", "msg_from": "Tom I Helbekkmo <[email protected]>", "msg_from_op": true, "msg_subject": "Time to drop --{en\|dis}able-hba from configure" }, { "msg_contents": "\nGood point...done\n\nOn Sat, 14 Feb 1998, Tom I Helbekkmo wrote:\n\n> The 'configure' script currently says\n> \n> \tchecking setting HBA... enabled\n> \n> which is hardly needed any more. I just grepped through the whole\n> tree, and there is no occurrence of \"NOHBA\" outside of configure -- as\n> there shouldn't be now that the pg_hba.conf file is used to select the\n> authentication scheme(s) to be used. (Which, by the way, was a Smart\n> Move, in my opinion. Things were too confusing as they were.)\n> \n> -tih\n> -- \n> Popularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sat, 14 Feb 1998 14:02:13 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Time to drop --{en\|dis}able-hba from configure" } ]
[ { "msg_contents": "Hope you don't mind my marking something that's for the contrib\ndirectory with REPOST like this, but Marc did ask us to, and I\ndid post this back on Februaray 1, so...\n\nI'm including an update to my user defined IP and MAC address type\nimplementation that's in contrib/ip_and_mac/. This one works right\nwith 6.3, avoids the problems I ran into earlier with LIKE, and\nincludes a bit of extra functionality.\n\nHere's an idea for the TODO list, by the way: make it doable without\nresorting to black magic to make a user defined type indexable! :-)\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"", "msg_date": "Sat, 14 Feb 1998 15:31:47 +0100", "msg_from": "Tom I Helbekkmo <[email protected]>", "msg_from_op": true, "msg_subject": "REPOST update of contributed IP and MAC types" }, { "msg_contents": "On Sat, 14 Feb 1998, Tom I Helbekkmo wrote:\n\n> Hope you don't mind my marking something that's for the contrib\n> directory with REPOST like this, but Marc did ask us to, and I\n> did post this back on Februaray 1, so...\n\n\tWell, a repost is generally if I forgot to add it :)\n\n> I'm including an update to my user defined IP and MAC address type\n> implementation that's in contrib/ip_and_mac/. This one works right\n> with 6.3, avoids the problems I ran into earlier with LIKE, and\n> includes a bit of extra functionality.\n\n\tupdated...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sat, 14 Feb 1998 13:59:28 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] REPOST update of contributed IP and MAC types" } ]
[ { "msg_contents": "Probably fixed already, since it's uncompilable, but anyway:\n\nThe file 'backend/utils/adt/arrayfuncs.c' won't compile with the\nFebruary 14th snapshot, because of an inconsistency between the\ndeclaration and implementation of ReadArrayStr(). As far as I can\ntell, the predeclaration is wrong. I assume this is what was meant:\n\n* backend/utils/adt/arrayfuncs.c.ORIG\tSat Feb 14 09:00:31 1998\n--- backend/utils/adt/arrayfuncs.c\tSat Feb 14 15:52:29 1998\n***********\n* 44,52 ***\n /-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-/\n static int\t_ArrayCount(char str, int dim[], int typdelim);\n static char _ReadArrayStr(char arrayStr, int nitems, int ndim, int dim[],\n! \t\t\t FmgrInfo inputproc, Oid typelem, char typdelim,\n! \t\t\t int typlen, bool typbyval, char typalign,\n! \t\t\t int nbytes, int16 typmod);\n \n #ifdef LOARRAY\n static char \n--- 44,52 ----\n /-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-/\n static int\t_ArrayCount(char str, int dim[], int typdelim);\n static char _ReadArrayStr(char arrayStr, int nitems, int ndim, int dim[],\n! \t\t\t FmgrInfo inputproc, Oid typelem, int16 typmod,\n! \t\t\t char typdelim, int typlen, bool typbyval,\n! \t\t\t char typalign, int nbytes);\n \n #ifdef LOARRAY\n static char \n\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n\n", "msg_date": "Sat, 14 Feb 1998 16:30:02 +0100 (CET)", "msg_from": "Tom I Helbekkmo <[email protected]>", "msg_from_op": true, "msg_subject": "Declaration glitch in arrayfuncs.c" }, { "msg_contents": "Gee, I noticed the compile was broken, but did not realize is was me. \nThanks.\n\n\n> \n> Probably fixed already, since it's uncompilable, but anyway:\n> \n> The file 'backend/utils/adt/arrayfuncs.c' won't compile with the\n> February 14th snapshot, because of an inconsistency between the\n> declaration and implementation of ReadArrayStr(). As far as I can\n> tell, the predeclaration is wrong. I assume this is what was meant:\n> \n> backend/utils/adt/arrayfuncs.c.ORIG\tSat Feb 14 09:00:31 1998\n> --- backend/utils/adt/arrayfuncs.c\tSat Feb 14 15:52:29 1998\n> ***********\n> * 44,52 ***\n> /-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-/\n> static int\t_ArrayCount(char str, int dim[], int typdelim);\n> static char _ReadArrayStr(char arrayStr, int nitems, int ndim, int dim[],\n> ! \t\t\t FmgrInfo inputproc, Oid typelem, char typdelim,\n> ! \t\t\t int typlen, bool typbyval, char typalign,\n> ! \t\t\t int nbytes, int16 typmod);\n> \n> #ifdef LOARRAY\n> static char \n> --- 44,52 ----\n> /-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-/\n> static int\t_ArrayCount(char str, int dim[], int typdelim);\n> static char _ReadArrayStr(char arrayStr, int nitems, int ndim, int dim[],\n> ! \t\t\t FmgrInfo inputproc, Oid typelem, int16 typmod,\n> ! \t\t\t char typdelim, int typlen, bool typbyval,\n> ! \t\t\t char typalign, int nbytes);\n> \n> #ifdef LOARRAY\n> static char *\n> \n> \n> -tih\n> -- \n> Popularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n> \n> \n> \n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sat, 14 Feb 1998 20:12:22 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Declaration glitch in arrayfuncs.c" } ]
[ { "msg_contents": "Here is my current list for porting status for the v6.3 release. I may\nhave missed at least a few reports, e.g hpux, irix??\n\nSince the porting support has changed for v6.3, if a system is not\ntested it should be assumed to be broken. Any regression test done since\nFeb 1 will count as \"confirmed working\", as long as the test ran to\ncompletion and for the most part behaved properly.\n\nAny machine which does not get an installation and a regression test for\nv6.3beta will move to the unsupported list. Also, let us know if you\nhave an interest in a port even though you cannot actually do the work\nto confirm it; that may encourage someone else to volunteer.\n\nMarc/Bruce, can you help me clarify the bsdi/freebsd/netbsd/bsdxxx\nentries? I'm not sure which are unique and what the names should be...\n\n - Tom\n\n* aix/4.1.4.0-4.2 - confirmed working when built on 4.1.4.0 (Darren\nKing)\n_ aix/3.5 - not yet tested? close enough to 4.1 to count?? (Frank\nDana?)\n_ bsdi\n_ FreeBSD/2.2.1,2.2.5 - in progress (Tatsuo)\n? NetBSD/i386 version? - not yet tested but should work?\nx NetBSD/m68k Amiga, HP300, Mac - not yet working... (Henry Hotz)\n* NetBSD/sparc version? confirmed working (Tom Helbekkmo)\n* NetBSD/vax version? confirmed working (Tom Helbekkmo)\n* dgux/5.4R4.11 - patches submitted (Brian Gallew)\n_ hpux/9.0.x\n_ hpux/10.20\n_ irix5\n_ irix6/MIPS\n_ dec/alpha - currently broken? confirmed working on v6.2.1 (Pedro)\n_ linux/alpha - currently broken?\n* linux/i386 - confirmed working (Thomas)\n? linux/i386/glibc2 - minor library breakage; in progress (Oliver)\n_ mklinux/ppc - in progress (Tatsuo)\n_ nextstep - worked with patches on v1.0.9; not working now?\n_ sco/i386\n_ solaris/i386 - confirmed working (Marc)\n* solaris/sparc/2.5.1 - confirmed working (Marc)\n_ solaris/sparc/2.6 - in progress (Tatsuo)\n_ sunos/sparc/4.1.4 - in progress (Tatsuo)\n_ svr4/MIPS - dcosx and sinix/seimens-nixdorf worked on v6.1 (Frank\nRidderbusch?)\n_ ultrix4 - no recent reports? obsolete port??\nx univel - not working now; in progress? (Billy G. Allie)\n\n", "msg_date": "Sat, 14 Feb 1998 17:03:32 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Valid ports for v6.3" }, { "msg_contents": "On Sat, 14 Feb 1998, Thomas G. Lockhart wrote:\n\n> Here is my current list for porting status for the v6.3 release. I may\n> have missed at least a few reports, e.g hpux, irix??\n\n\tNeil, can you organize with Thomas about having a sub page from\nthe main WWW page that lists these? Something visible? Including the\ndisclaimer below about testing?\n\n> Since the porting support has changed for v6.3, if a system is not\n> tested it should be assumed to be broken. Any regression test done since\n> Feb 1 will count as \"confirmed working\", as long as the test ran to\n> completion and for the most part behaved properly.\n> \n> Any machine which does not get an installation and a regression test for\n> v6.3beta will move to the unsupported list. Also, let us know if you\n> have an interest in a port even though you cannot actually do the work\n> to confirm it; that may encourage someone else to volunteer.\n> \n> _ FreeBSD/2.2.1,2.2.5 - in progress (Tatsuo)\n\n\tFreeBSD 2.x and 3.x are the current working models, with 2.x being\nour \"stable\" distribution, while 3.x is our development one. From what\nI'm seeing, 3.x looks like it might be moving towards a release, as we're\nstarting to generate snapshot's of that also, but I'm not sure.\n\n\tFrom my experience so far, 2.x and 3.x operate the same way as far\nas PostgreSQL is concerned, as I'm running development under 3.x, but my\nproduction platform is under 2.x...\n\n> ? NetBSD/i386 version? - not yet tested but should work?\n\n\tIf I can remember my password on the machine in question, I'll get\nonline and do a compile/regression test on the i386 model this\nafternoon...\n\n> _ solaris/i386 - confirmed working (Marc)\n> * solaris/sparc/2.5.1 - confirmed working (Marc)\n> _ solaris/sparc/2.6 - in progress (Tatsuo)\n\n\tconfirmed working...its my model machine at the office, and what\nthe regression.SunOS file under src/test/regress is based off of...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sat, 14 Feb 1998 13:45:34 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Valid ports for v6.3" }, { "msg_contents": "In message <[email protected]>, \"Thomas G. Lockhart\" writes:\n> Here is my current list for porting status for the v6.3 release. I may\n> have missed at least a few reports, e.g hpux, irix??\n> \n> Since the porting support has changed for v6.3, if a system is not\n> tested it should be assumed to be broken. Any regression test done since\n> Feb 1 will count as \"confirmed working\", as long as the test ran to\n> completion and for the most part behaved properly.\n> \n> Any machine which does not get an installation and a regression test for\n> v6.3beta will move to the unsupported list. Also, let us know if you\n> have an interest in a port even though you cannot actually do the work\n> to confirm it; that may encourage someone else to volunteer.\n> \n> Marc/Bruce, can you help me clarify the bsdi/freebsd/netbsd/bsdxxx\n> entries? I'm not sure which are unique and what the names should be...\n> \n> - Tom\n> \n> * aix/4.1.4.0-4.2 - confirmed working when built on 4.1.4.0 (Darren\n> King)\n> _ aix/3.5 - not yet tested? close enough to 4.1 to count?? (Frank\n> Dana?)\n> _ bsdi\n> _ FreeBSD/2.2.1,2.2.5 - in progress (Tatsuo)\n> ? NetBSD/i386 version? - not yet tested but should work?\n> x NetBSD/m68k Amiga, HP300, Mac - not yet working... (Henry Hotz)\n> * NetBSD/sparc version? confirmed working (Tom Helbekkmo)\n> * NetBSD/vax version? confirmed working (Tom Helbekkmo)\n> * dgux/5.4R4.11 - patches submitted (Brian Gallew)\n> _ hpux/9.0.x\n> _ hpux/10.20\n> _ irix5\n> _ irix6/MIPS\n> _ dec/alpha - currently broken? confirmed working on v6.2.1 (Pedro)\n> _ linux/alpha - currently broken?\n> * linux/i386 - confirmed working (Thomas)\n> ? linux/i386/glibc2 - minor library breakage; in progress (Oliver)\n> _ mklinux/ppc - in progress (Tatsuo)\n> _ nextstep - worked with patches on v1.0.9; not working now?\n> _ sco/i386\n> _ solaris/i386 - confirmed working (Marc)\n> * solaris/sparc/2.5.1 - confirmed working (Marc)\n> _ solaris/sparc/2.6 - in progress (Tatsuo)\n> _ sunos/sparc/4.1.4 - in progress (Tatsuo)\n> _ svr4/MIPS - dcosx and sinix/seimens-nixdorf worked on v6.1 (Frank\n> Ridderbusch?)\n> _ ultrix4 - no recent reports? obsolete port??\n> x univel - not working now; in progress? (Billy G. Allie)\n> \n> \n\nI don't see linux/sparc here. The last snapshot I tried was from 2/6.\nThat ran well. I just grapped today's (2/14) snapshot and will try that,\nand let you know. I'm using 2.0.29 kernel. \n\nOne issue I've heard of is that _SC_OPEN_MAX is used in \nbackend/storage/file/fd.c. I later kernels, /usr/include/asm/unistd.h\nwas changed. _SC_OPEN_MAX has #ifdef __KERNEL__ around it. 2.0.29 \ndoesn't have the #ifdef, so I don't have this issue.\n\nI'm not what the correct fix should be.\n\nTom Szybist\[email protected]\n", "msg_date": "Sat, 14 Feb 1998 15:49:14 -0500", "msg_from": "\"Thomas A. Szybist\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Valid ports for v6.3 " }, { "msg_contents": "> Here is my current list for porting status for the v6.3 release. I may\n> have missed at least a few reports, e.g hpux, irix??\n> \n> x univel - not working now; in progress? (Billy G. Allie)\n> \nYes, it's currently being worked on.\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": "Sat, 14 Feb 1998 16:55:40 -0500", "msg_from": "\"Billy G. Allie\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Valid ports for v6.3 " }, { "msg_contents": " ? NetBSD/i386 version? - not yet tested but should work?\n\nI'm trying to compile 6.3 on NetBSD/i386 v1.3. A couple of problems\ncrop up with the compile. I get the following warnings/errors:\n\n ... (lots of stuff deleted from compilation output) ...\n\n gmake[2]: Entering directory `/usr/local/NetBSD/pkgsrc/databases/postgresql/work/pgsql/src/backend/bootstrap'\n /usr/bin/yacc -d bootparse.y\n grep -v \"^#\" boot.sed > sedfile\n sed -f sedfile < y.tab.c > bootparse.c\n mv y.tab.h bootstrap_tokens.h\n rm -f y.tab.c sedfile\n gcc -I../../include -I../../backend -I/usr/local/include -O2 -pipe -Wall -Wmissing-prototypes -I.. -Wno-error -c bootparse.c -o bootparse.o\n y.tab.c: In function `Int_yyparse':\n y.tab.c:378: warning: implicit declaration of function `Int_yylex'\n y.tab.c:417: warning: implicit declaration of function `Int_yyerror'\n flex bootscanner.l\n grep -v \"^#\" boot.sed > sedfile\n sed -f sedfile < lex.yy.c > bootscanner.c\n rm -f lex.yy.c sedfile\n gcc -I../../include -I../../backend -I/usr/local/include -O2 -pipe -Wall -Wmissing-prototypes -I.. -Wno-error -c bootscanner.c -o bootscanner.o\n lex.Int_yy.c:683: warning: no previous prototype for `Int_yylex'\n bootscanner.l:137: warning: no previous prototype for `Int_yyerror'\n\n ... (lots of stuff deleted) ...\n\n gmake[2]: Entering directory `/usr/local/NetBSD/pkgsrc/databases/postgresql/work/pgsql/src/backend/parser'\n /usr/bin/yacc -d gram.y\n /usr/bin/yacc: f - maximum table size exceeded\n gmake[2]: * [parse.h] Error 2\n /usr/bin/yacc -d gram.y\n /usr/bin/yacc: f - maximum table size exceeded\n gmake[2]: * [gram.c] Error 2\n\nBoth sets of problems seem to relate to processing parsers with yacc.\nDo I need bison instead? If so, perhaps this should be listed as a\nrequirement in the INSTALL docs.\n\nThanks for your help. I'll verify this port as soon as I resolve the\nparser problems.\n\nCheers,\nBrook\n\n\n\n\n", "msg_date": "Sat, 14 Feb 1998 16:10:12 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Valid ports for v6.3 -- NetBSD/i386 compile errors" }, { "msg_contents": "On Sat, 14 Feb 1998, Brook Milligan wrote:\n\n> ? NetBSD/i386 version? - not yet tested but should work?\n> \n> I'm trying to compile 6.3 on NetBSD/i386 v1.3. A couple of problems\n> crop up with the compile. I get the following warnings/errors:\n> \n> ... (lots of stuff deleted from compilation output) ...\n> \n> gmake[2]: Entering directory `/usr/local/NetBSD/pkgsrc/databases/postgresql/work/pgsql/src/backend/bootstrap'\n> /usr/bin/yacc -d bootparse.y\n> grep -v \"^#\" boot.sed > sedfile\n> sed -f sedfile < y.tab.c > bootparse.c\n> mv y.tab.h bootstrap_tokens.h\n> rm -f y.tab.c sedfile\n> gcc -I../../include -I../../backend -I/usr/local/include -O2 -pipe -Wall -Wmissing-prototypes -I.. -Wno-error -c bootparse.c -o bootparse.o\n> y.tab.c: In function `Int_yyparse':\n> y.tab.c:378: warning: implicit declaration of function `Int_yylex'\n> y.tab.c:417: warning: implicit declaration of function `Int_yyerror'\n> flex bootscanner.l\n> grep -v \"^#\" boot.sed > sedfile\n> sed -f sedfile < lex.yy.c > bootscanner.c\n> rm -f lex.yy.c sedfile\n> gcc -I../../include -I../../backend -I/usr/local/include -O2 -pipe -Wall -Wmissing-prototypes -I.. -Wno-error -c bootscanner.c -o bootscanner.o\n> lex.Int_yy.c:683: warning: no previous prototype for `Int_yylex'\n> bootscanner.l:137: warning: no previous prototype for `Int_yyerror'\n\n\tThese \"prototype\" errors are normal and can safely be ignored...\n\n> ... (lots of stuff deleted) ...\n> \n> gmake[2]: Entering directory `/usr/local/NetBSD/pkgsrc/databases/postgresql/work/pgsql/src/backend/parser'\n> /usr/bin/yacc -d gram.y\n> /usr/bin/yacc: f - maximum table size exceeded\n> gmake[2]: * [parse.h] Error 2\n> /usr/bin/yacc -d gram.y\n> /usr/bin/yacc: f - maximum table size exceeded\n> gmake[2]: * [gram.c] Error 2\n\n\tRequires bison to be installed instead of yacc...\n\n> Both sets of problems seem to relate to processing parsers with yacc.\n> Do I need bison instead? If so, perhaps this should be listed as a\n> requirement in the INSTALL docs.\n\n\tIt doesn't appear to be a seperate requirement on all systems...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sat, 14 Feb 1998 20:48:34 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PORTS] Re: [HACKERS] Valid ports for v6.3 -- NetBSD/i386 compile\n\terrors" }, { "msg_contents": "> \tRequires bison to be installed instead of yacc...\n> \n> > Both sets of problems seem to relate to processing parsers with yacc.\n> > Do I need bison instead? If so, perhaps this should be listed as a\n> > requirement in the INSTALL docs.\n> \n> \tIt doesn't appear to be a seperate requirement on all systems...\n\nNeeded on bsdi. I recommend we make it come with the distribution like\nwe do with scan.c.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sat, 14 Feb 1998 20:16:25 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PORTS] Re: [HACKERS] Valid ports for v6.3 -- NetBSD/i386 compile\n\terrors" }, { "msg_contents": "> > gmake[2]: Entering directory `/usr/local/NetBSD/pkgsrc/databases/postgresql/work/pgsql/src/backend/parser'\n> > /usr/bin/yacc -d gram.y\n> > /usr/bin/yacc: f - maximum table size exceeded\n> > gmake[2]: * [parse.h] Error 2\n> > /usr/bin/yacc -d gram.y\n> > /usr/bin/yacc: f - maximum table size exceeded\n> > gmake[2]: * [gram.c] Error 2\n>\n> Requires bison to be installed instead of yacc...\n>\n> > Both sets of problems seem to relate to processing parsers with yacc.\n> > Do I need bison instead? If so, perhaps this should be listed as a\n> > requirement in the INSTALL docs.\n>\n> It doesn't appear to be a seperate requirement on all systems...\n\nStan Brown suggested trying a -N switch (not sure which system he got this from):\n\n -N<secondary><n>\n Allow the sizes of certain internal yacc tables to\n be reset. secondary is one of the letters from\n the set {B a m s p n e c l w} and specifies the\n table; n is the new size. Tables that can be\n reset by using secondary letters are as follows:\n a a-array size; default is 12000.\n m mem array size; default is 12000\n s number of states; default is 1000.\n p number of productions; default is\n 800\n n number of non-terminals; default is\n 600.\n e temp-space size; default is 1250.\n c name-space size; default is 5000.\n l look-ahead set table size; default i\n 650.\n w working set table size; default is\n 650.\n\nIt would be great if someone would show exactly what is needed for the BSD yacc systems to avoid a requirement for bison. The problem is that in the last\ncouple of weeks the parser finally grew to exceed some internal limit in BSD yacc. Any solution might be applicable to other yacc'ers...\n\nAlso, if this fails we can try packaging \"gram.c\" with the distribution; I think that bison is similar to flex in generating library-independent C code. Would\nlike to resolve this in the next few days...\n\n - Tom\n\n", "msg_date": "Sun, 15 Feb 1998 01:19:49 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [PORTS] Re: [HACKERS] Valid ports for v6.3 -- NetBSD/i386 compile\n\terrors" }, { "msg_contents": " > > gmake[2]: Entering directory `/usr/local/NetBSD/pkgsrc/databases/postgresql/work/pgsql/src/backend/parser'\n > > /usr/bin/yacc -d gram.y\n > > /usr/bin/yacc: f - maximum table size exceeded\n > > gmake[2]: * [parse.h] Error 2\n > > /usr/bin/yacc -d gram.y\n > > /usr/bin/yacc: f - maximum table size exceeded\n > > gmake[2]: * [gram.c] Error 2\n >\n > Requires bison to be installed instead of yacc...\n >\n > > Both sets of problems seem to relate to processing parsers with yacc.\n > > Do I need bison instead? If so, perhaps this should be listed as a\n > > requirement in the INSTALL docs.\n >\n > It doesn't appear to be a seperate requirement on all systems...\n\n Stan Brown suggested trying a -N switch (not sure which system he got this from):\n\nAt least on NetBSD/i386 v1.3 yacc does not have a -N switch.\n\n Also, if this fails we can try packaging \"gram.c\" with the\n distribution; I think that bison is similar to flex in generating\n library-independent C code. \n\nThis seems to be the best solution. The situation with flex and yacc\nis much the same as with the configuration system. These programs\ntake specs and generate compilable files that should be system\nindependent (am I wrong about that last point?). Just as we don't\nrequire people to have the autoconf tool to generate the configure\nscript, perhaps we shouldn't require flex or bison either. Of course,\nthe original lexer and parser files should be shipped as well for\ncompleteness, just as configure.in is.\n\nJust my thoughts. I'll get bison in the meantime so I can do the\ntesting.\n\nCheers,\nBrook\n\n", "msg_date": "Sat, 14 Feb 1998 22:17:22 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PORTS] Re: [HACKERS] Valid ports for v6.3 -- NetBSD/i386 compile\n\terrors" }, { "msg_contents": "On Sat, 14 Feb 1998, Thomas G. Lockhart wrote:\n\n> Here is my current list for porting status for the v6.3 release. I may\n> have missed at least a few reports, e.g hpux, irix??\n...\n> _ linux/alpha - currently broken?\n\n\tLinux/Alpha is indeed currently broken (i.e. I would not use it\nyet for production work), but getting better every day! :)\n\n----------------------------------------------------------------------------\n\| \"For to me to live is Christ, and to die is gain.\" \|\n\| --- Philippians 1:21 (KJV) \|\n----------------------------------------------------------------------------\n\| Ryan Kirkpatrick \| Boulder, Colorado \| [email protected] \|\n----------------------------------------------------------------------------\n\| http://www-ugrad.cs.colorado.edu/~rkirkpat/ \|\n----------------------------------------------------------------------------\n\n\n", "msg_date": "Sun, 15 Feb 1998 12:11:29 -0600 (CST)", "msg_from": "Ryan Kirkpatrick <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Valid ports for v6.3" } ]
[ { "msg_contents": "I have asked Massimo if he is more able to get involved in development.\nHis reply is attached. He has submitted a major patch that we need to\ndiscuss as a group.\n\n---------------------------------------------------------------------------\n\n\n> > > > I am submitting this patch for people to review.\n> > > > \n> > > > It fixes several problems in the lock manager. Let me show you how to\n> > > > reproduce them. Try this before applying the patch and you will see the\n> > > > old bugs.\n> > > > \n> > > \n> > > I'm sorry for the delay but I have been very busy in the last weeks.\n> > > I have an old patch which does something similar. I have submitted it\n> > > and many others to the patches mailing list. Plase have a look at it.\n> > \n> > I will certainly check the archives. I remember a lock debug patch we\n> > applied several months ago, but I don't rememeber anything else. I will\n> > find it.\n> > \n> > Massimo, you certainly know the code. It would be nice if you had more\n> > time to get involved in the development effort. I know you send in\n> > patches from time to time, but it would be nice if you could be more\n> > involved in discussions and fielding bug/feature requests.\n> > \n> > -- \n> > Bruce Momjian\n> > [email protected]\n> > \n> \n> Hi Bruce,\n> \n> I'm sorry for the delay in answering your message but I have been forced\n> off line for three weeks (forced at home with all the university modems\n> unusable). But also under normal conditions I can usually read my e-mail\n> once for week and in I'm in general too busy to follow all the discussions\n> in the mailing lists. For these reasons I can't partecipate actively to\n> the postgres developement. The only thing I can reasonably do is find some\n> bugs and submit patches when I need them fixed urgently.\n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sat, 14 Feb 1998 21:27:45 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Massimo assistance" }, { "msg_contents": "As people start understanding certain areas of the code, would you\nplease send in a README file for that directory, so future developers\ncan be aided by your knowledge of how things work?\n\nI have done README files for the parser and optimizer, but many more\nneed to be done or added to.\n\nVadim is adding one to the optimizer/planner about subselects, and\nperhaps Massimo can add to the lock README file. A good README would\nhave helped me a lot in adding the deadlock detection code. I had to\nfigure it out myself instead.\n\nThese README's can be added up to the last day of release because they\ndon't affect the code.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sun, 15 Feb 1998 13:05:25 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Adding README files to source" } ]
[ { "msg_contents": "Here is a description of the major patch from Massimo. The irony of\nthis is that the mail message is dated January 27th, when application of\nthe patch would have been easier because we were not in beta testing. I\nhave a copy of the patch here on my machine.\n\nWhat do people want to do with this? I have reviewed the patch, and it\nlooks good, but it may take work to merge in because it is against\n6.2.1p7, not 6.3 beta, and it does introduce quite a bit of new code.\n\n---------------------------------------------------------------------------\n\nForwarded message:\n> From [email protected] Sat Feb 14 09:31:23 1998\n> From: Massimo Dal Zotto <[email protected]>\n> Message-Id: <[email protected]>\n> Subject: [PATCHES] patches for 6.2.1p6\n> To: [email protected] (Pgsql Patches)\n> Date: Tue, 27 Jan 1998 22:46:38 +0100 (MET)\n> X-Mailer: ELM [version 2.4 PL24 ME4]\n> MIME-Version: 1.0\n> Content-Type: multipart/mixed; boundary=%#%record%#%\n> Sender: [email protected]\n> Precedence: bulk\n> \n> --%#%record%#%\n> Content-Type: text/plain; charset=iso-8859-1\n> Content-Transfer-Encoding: 8bit\n> Content-Length: 6638 \n> \n> Hi hackers,\n> \n> I have old patches for version 6.2.1p6 which fix some problems and add\n> new features. Here is a short description of each patch file:\n> \n> \n> assert.patch\n> \n> \tadds a switch to turn on/off the assert checking if enabled at compile\n> \ttime. You can now compile postgres with assert checking and disable it\n> \tat runtime in a production environment.\n> \n> async-unlisten.patch\n> \n> \tdeclares Async_Unlisten() external so that it can be called by user\n> \tmodules.\n> \n> exec-limit.patch\n> \n> \tremoves the #ifdef NOT_USED around ExecutorLimit(). It is used.\n> \n> exitpg.patch\n> \n> \tlimits recursive calls to exitpg() preventing an infinite loop\n> \tif an error is found inside exitpg.\n> \n> libpgtcl-listen.patch\n> \n> \tJust a change from upper to lowercase of an sql command in libpgtcl,\n> \ttotally harmless.\n> \n> new-locks.patch\n> \n> \tAfter long studying and many debugging sessions I have finally\n> \tunderstood how the low level locks work.\n> \tI have completely rewritten lock.c cleaning up the code and adding\n> \tbetter assert checking. I have also added some fields to the lock\n> \tand xid tags for better support of user locks. This patch includes\n> \talso a patch submitted by Bruce Momjian which changes the handling\n> \tof lock priorities. It can however be disabled if an option is set\n> \tin pg_options, see tprintf.patch (Bruce patch works by building\n> \tthe queue in reverse priority order, my old patch kept the queue in\n> \tdecreasing order and traversed it from the other side).\n> \n> pg-flush.patch\n> \n> \tremoves an unnecessary flush in libpq reducing network traffic and\n> \tincreasing performance.\n> \n> relname.patch\n> \n> \tan utility which returns the relname corresponding to a given oid.\n> \tUseful for debug messages (see vacum.patch).\n> \n> sequence.patch\n> \n> \tadded a setval() function which enables othe owner of a sequence\n> \tto set its value without need to delete and recreate it.\n> \n> sinval.patch\n> \n> \tfixes a problem in SI cache which causes table overflow if some\n> \tbackend is idle for a long time while other backends keep adding\n> \tentries.\n> \tIt uses the new signal handling implemented in tprintf.patch.\n> \tI have also increacasesed the max number of backends from 32 to 64 and\n> \tthe table size from 1000 to 5000.\n> \n> spin-lock.patch\n> \n> \tI'm not sure if this is really useful, but it seems stupid to have\n> \ta backend wasting cpu cycles in a busy loop while the process which\n> \tshould release the lock is waiting for the cpu. So I added a call\n> \tto process_yield() if the spin lock can't obtained.\n> \tThis has been implemented and tested only on Linux. I don't know if\n> \tother OS have process_yield(). If someone can check please do it.\n> \n> tprintf.patch\n> \n> \tadds functions and macros which implement a conditional trace package\n> \twith the ability to change flags and numeric options of running\n> \tbackends at runtime.\n> \tOptions/flags can be specified in the command line and/or read from\n> \tthe file pg_options in the data directory.\n> \tRunning backends can be forced to update their options from this file\n> \tby sending them a SIGHUP signal (this is the convention used by most\n> \tunix daemons so I changed the meaning of SIGHUP).\n> \tOptions can be debugging flags used by the trace package or any other\n> \tnumeric\tvalue used by the backend, for example the deadlock_timeout.\n> \tHaving flags and options specified at runtime and changed while the\n> \tbackends are running can greatly simplify the debugging and tuning\n> \tof the database. New options can be defined in utils/misc/trace.c and\n> \tinclude/utils/trace.h. As an example of the usage of this package\n> \tsee lock.c and proc.c which make use of new runtime options.\n> \n> \tOld files using int flags or variables can be easily changed to\n> \tuse the new package by substituting the old variable with a #define\n> \tlike in the following example:\n> \t\n> \t /* int my_flag = 0; /\n> \t #include \"trace.h\"\n> \t #define my_flag pg_options[OPT_MYFLAG]\n> \n> \tI have done it in postgres.c and some other files and now I can turn\n> \ton/off any single debug flag on the fly with a simple shell script.\n> \tI have removed the IpcConfigTip() from ipc.c, it should better be\n> \tdescribed in the postgres manual instead of being printed on stderr.\n> \n> \tThis patch provides also a new format of debugging messages which\n> \tare always in a single line with a timestamp and the backend pid:\n> \n> \t #timestamp #pid #message\n> \t 980127.17:52:14.173 [29271] StartTransactionCommand\n> \t 980127.17:52:14.174 [29271] ProcessUtility: drop table t;\n> \t 980127.17:52:14.186 [29271] SIIncNumEntries: table is 70% full\n> \t 980127.17:52:14.186 [29286] Async_NotifyHandler\n> \t 980127.17:52:14.186 [29286] Waking up sleeping backend process\n> \t 980127.19:52:14.292 [29286] Async_NotifyFrontEnd\n> \t 980127.19:52:14.413 [29286] Async_NotifyFrontEnd done\n> \t 980127.19:52:14.466 [29286] Async_NotifyHandler done\n> \n> \tThis improves the readability of the log and allows one to understand\n> \texactly which backend is doing what and at which time. It also makes\n> \teasier to write simple awk or perl scripts which monitor the log to\n> \tdetect database errors or problem, or to compute transaction times.\n> \n> \tThe patch changes also the meaning of signals used by postgres, as\n> \tdescribed by the following table:\n> \n> \t\t postmaster\t\t\t\tbackend\n> \n> \tSIGHUP kill(,sighup)\t\t\tread_pg_options\n> \tSIGINT\t kill(,sigint), die\t\t\tdie\n> \tSIGCHLD\t reaper\t\t\t\t-\n> \tSIGTTIN\t ignored\t\t\t\t-\n> \tSIGTTOU\t ignored\t\t\t\t-\n> \tSIGQUIT\t die\t\t\t\t\thandle_warn\n> \tSIGTERM\t kill(,sigterm), kill(,9), die\tdie\n> \tSIGCONT\t dumpstatus\t\t\t\t-\n> \tSIGPIPE\t ignored\t\t\t\tdie\n> \tSIGFPE\t -\t\t\t\t\tFloatExceptionHandler\n> \tSIGTSTP\t -\t\t\t\t\tignored (alive test)\n> \tSIGUSR1\t kill(,sigusr1), die\t\tquickdie\n> \tSIGUSR2\t kill(,sigusr2)\t\t\tAsync_NotifyHandler\n> \t\t\t\t\t\t\t(also SI buffer flush)\n> \n> \tThe main changes to the old implementation are SIGQUIT instead of\n> \tSIGHUP to handle warns, SIGHUP to reread pg_options and redirection\n> \tto all backends of SIGHUP, SIGINT, SIGTERM, SIGUSR1 and SIGUSR2.\n> \tIn this way some of the signals sent to the postmaster can be sent\n> \tautomatically to all the backends. To shut down postgres one needs\n> \tonly to send a SIGTERM to postmaster and it will stop automatically\n> \tall the backends. This new signal handling mechanism is also used\n> \tto prevent SI cache table overflows: when a backend detects the SI\n> \ttable full at 70% it simply sends a signal to the postmaster which\n> \twill wake up all idle backends and make them flush the cache.\n> \n> vacuum.patch\n> \n> \tadds a debug message to vacuum that prints the name of a table or\n> \tindex before* vacuuming it, if the verbose keyword is set.\n> \tThis is useful to know which table is causing troubles if a\n> \tvacuum all crashes. Currently table information is printed only\n> \tat the end of each vacuum operation and is never printed if the\n> \tvacuum crashes.\n> \n> -- \n> Massimo Dal Zotto\n\n\n-- \nBruce Momjian\[email protected]\n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sat, 14 Feb 1998 21:31:46 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Massimo patch" }, { "msg_contents": "On Sat, 14 Feb 1998, Bruce Momjian wrote:\n\n> Here is a description of the major patch from Massimo. The irony of\n> this is that the mail message is dated January 27th, when application of\n> the patch would have been easier because we were not in beta testing. I\n> have a copy of the patch here on my machine.\n> \n> What do people want to do with this? I have reviewed the patch, and it\n> looks good, but it may take work to merge in because it is against\n> 6.2.1p7, not 6.3 beta, and it does introduce quite a bit of new code.\n\n\tI sent Massimo and email (CC'd to the list) explaining the fact\nthat we are in Beta mode, and that some of these patches are currently\nquestionable for v6.3 (the lock patch being one)...with another 2 weeks of\nbeta testing before release, my opinion is that these involve changes that\nwill not have sufficient time to test prior to release, especially with at\nleast one major bug still existing...\n\n\tThe other question that I do have is how appropriate some of these\npatches are to v6.3...2 weeks, in my opinion, isn't a suitable amount of\ntime, in which to accurately test these, and should wait until after the\nrelease...\n\n\tUnless anyone can really argue this, those that add features\n(assert.patch) should be omitted...those that fix a bug are appropriate.\nNot that I haven't looked at the patch itself, only at the descriptions\npresented...\n\n\n\n\n> > assert.patch\n> > \n> > \tadds a switch to turn on/off the assert checking if enabled at compile\n> > \ttime. You can now compile postgres with assert checking and disable it\n> > \tat runtime in a production environment.\n\n\tNew feature...not a bug fix...\n\n> > async-unlisten.patch\n> > \n> > \tdeclares Async_Unlisten() external so that it can be called by user\n> > \tmodules.\n\n\tNew feature...not a bug fix...\n\n> > exec-limit.patch\n> > \n> > \tremoves the #ifdef NOT_USED around ExecutorLimit(). It is used.\n\n\tif it has a \"#ifdef NOT_USED\" around it, then how is it used? \n\n> > exitpg.patch\n> > \n> > \tlimits recursive calls to exitpg() preventing an infinite loop\n> > \tif an error is found inside exitpg.\n\n\tPotential bug...but how is it triggered?\n\n> > libpgtcl-listen.patch\n> > \n> > \tJust a change from upper to lowercase of an sql command in libpgtcl,\n> > \ttotally harmless.\n\t\n\t??\n\n> > new-locks.patch\n> > \n> > \tAfter long studying and many debugging sessions I have finally\n> > \tunderstood how the low level locks work.\n> > \tI have completely rewritten lock.c cleaning up the code and adding\n> > \tbetter assert checking. I have also added some fields to the lock\n> > \tand xid tags for better support of user locks. This patch includes\n> > \talso a patch submitted by Bruce Momjian which changes the handling\n> > \tof lock priorities. It can however be disabled if an option is set\n> > \tin pg_options, see tprintf.patch (Bruce patch works by building\n> > \tthe queue in reverse priority order, my old patch kept the queue in\n> > \tdecreasing order and traversed it from the other side).\n\n\tI don't understand this one, but it sounds like its negates the\nwork you just did? Again, sounds like a feature change, not a bug fix,\nsince you have recently re-written this...\n\n> > pg-flush.patch\n> > \n> > \tremoves an unnecessary flush in libpq reducing network traffic and\n> > \tincreasing performance.\n\n\tDidn't we just do a protocol rewrite? \n\n> > relname.patch\n> > \n> > \tan utility which returns the relname corresponding to a given oid.\n> > \tUseful for debug messages (see vacum.patch).\n\n\tIs this a new program? src/bin/relname?\n\n> > sequence.patch\n> > \n> > \tadded a setval() function which enables othe owner of a sequence\n> > \tto set its value without need to delete and recreate it.\n\n\tFeature change, not a bug...\n\n> > sinval.patch\n> > \n> > \tfixes a problem in SI cache which causes table overflow if some\n> > \tbackend is idle for a long time while other backends keep adding\n> > \tentries.\n> > \tIt uses the new signal handling implemented in tprintf.patch.\n> > \tI have also increacasesed the max number of backends from 32 to 64 and\n> > \tthe table size from 1000 to 5000.\n\n\tBug fix...\n\n> > spin-lock.patch\n> > \n> > \tI'm not sure if this is really useful, but it seems stupid to have\n> > \ta backend wasting cpu cycles in a busy loop while the process which\n> > \tshould release the lock is waiting for the cpu. So I added a call\n> > \tto process_yield() if the spin lock can't obtained.\n> > \tThis has been implemented and tested only on Linux. I don't know if\n> > \tother OS have process_yield(). If someone can check please do it.\n\n\tSounds like a bug fix to me...\n\n> > tprintf.patch\n> > \n> > \tadds functions and macros which implement a conditional trace package\n> > \twith the ability to change flags and numeric options of running\n> > \tbackends at runtime.\n> > \tOptions/flags can be specified in the command line and/or read from\n> > \tthe file pg_options in the data directory.\n> > \tRunning backends can be forced to update their options from this file\n> > \tby sending them a SIGHUP signal (this is the convention used by most\n> > \tunix daemons so I changed the meaning of SIGHUP).\n> > \tOptions can be debugging flags used by the trace package or any other\n> > \tnumeric\tvalue used by the backend, for example the deadlock_timeout.\n> > \tHaving flags and options specified at runtime and changed while the\n> > \tbackends are running can greatly simplify the debugging and tuning\n> > \tof the database. New options can be defined in utils/misc/trace.c and\n> > \tinclude/utils/trace.h. As an example of the usage of this package\n> > \tsee lock.c and proc.c which make use of new runtime options.\n\n\tDefinitely have to say new feature...\n\n> > vacuum.patch\n> > \n> > \tadds a debug message to vacuum that prints the name of a table or\n> > \tindex before vacuuming it, if the verbose keyword is set.\n> > \tThis is useful to know which table is causing troubles if a\n> > \tvacuum all crashes. Currently table information is printed only\n> > \tat the end of each vacuum operation and is never printed if the\n> > \tvacuum crashes.\n\n\tAgain, a feature more then a bug fix...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sun, 15 Feb 1998 00:29:06 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Massimo patch" }, { "msg_contents": "> Here is a description of the major patch from Massimo. The irony of\n> this is that the mail message is dated January 27th, when application of\n> the patch would have been easier because we were not in beta testing. I\n> have a copy of the patch here on my machine.\n>\n> What do people want to do with this? I have reviewed the patch, and it\n> looks good, but it may take work to merge in because it is against\n> 6.2.1p7, not 6.3 beta, and it does introduce quite a bit of new code.\n\nJeesh. Sure looks nice. Without knowing how fundamental the code changes are, it\nis hard to say for sure, but I would think we would want to put this in, unless\nit leads to major breakage. Someone would need to take the patches, merge them\nall in as a group, and validate the regression tests to see where we are on them.\nI'd even think it would be worth doing if it delays release a couple of weeks,\nwhich it may not.\n\nMore comments??\n\n - Tom\n\n", "msg_date": "Sun, 15 Feb 1998 04:37:03 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Massimo patch" }, { "msg_contents": "> spin-lock.patch\n> \n> \tI'm not sure if this is really useful, but it seems stupid to have\n> \ta backend wasting cpu cycles in a busy loop while the process which\n> \tshould release the lock is waiting for the cpu. So I added a call\n> \tto process_yield() if the spin lock can't obtained.\n> \tThis has been implemented and tested only on Linux. I don't know if\n> \tother OS have process_yield(). If someone can check please do it.\n\nMassimo brings up a good point. Most of our s_lock.h locking does asm\nmutex loops looking for a lock. Unless we are using a multi-cpu\nmachine, there is no way this is going to change while we are spinning.\n\nLinux has process_yield(), but most OS's don't. Is there a\nplatform-independent way to relinquish the cpu if the first attempt at\nthe spinlock fails? Would a select() of 1 microsecond work?\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sun, 15 Feb 1998 00:27:17 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "spin locks" }, { "msg_contents": "On Sun, 15 Feb 1998, Bruce Momjian wrote:\n\n> > spin-lock.patch\n> > \n> > \tI'm not sure if this is really useful, but it seems stupid to have\n> > \ta backend wasting cpu cycles in a busy loop while the process which\n> > \tshould release the lock is waiting for the cpu. So I added a call\n> > \tto process_yield() if the spin lock can't obtained.\n> > \tThis has been implemented and tested only on Linux. I don't know if\n> > \tother OS have process_yield(). If someone can check please do it.\n> \n> Massimo brings up a good point. Most of our s_lock.h locking does asm\n> mutex loops looking for a lock. Unless we are using a multi-cpu\n> machine, there is no way this is going to change while we are spinning.\n\n\tI'm not quite sure I follow this...in a multi-cpu environment,\nwould process_yield() introduce a problem? raised eyebrow\n\n> Linux has process_yield(), but most OS's don't. Is there a\n> platform-independent way to relinquish the cpu if the first attempt at\n> the spinlock fails? Would a select() of 1 microsecond work?\n\n\tThere is nothing wrong with introducing an OS specific\noptimization to the code...we can add a HAVE_PROCESS_YIELD to config.h and\nif a system has it, use it...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sun, 15 Feb 1998 02:11:04 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] spin locks" }, { "msg_contents": "> \n> On Sat, 14 Feb 1998, Bruce Momjian wrote:\n> \n> > Here is a description of the major patch from Massimo. The irony of\n> > this is that the mail message is dated January 27th, when application of\n> > the patch would have been easier because we were not in beta testing. I\n> > have a copy of the patch here on my machine.\n> > \n> > What do people want to do with this? I have reviewed the patch, and it\n> > looks good, but it may take work to merge in because it is against\n> > 6.2.1p7, not 6.3 beta, and it does introduce quite a bit of new code.\n> \n> \tI sent Massimo and email (CC'd to the list) explaining the fact\n> that we are in Beta mode, and that some of these patches are currently\n> questionable for v6.3 (the lock patch being one)...with another 2 weeks of\n> beta testing before release, my opinion is that these involve changes that\n> will not have sufficient time to test prior to release, especially with at\n> least one major bug still existing...\n> \n> \tThe other question that I do have is how appropriate some of these\n> patches are to v6.3...2 weeks, in my opinion, isn't a suitable amount of\n> time, in which to accurately test these, and should wait until after the\n> release...\n> \n> \tUnless anyone can really argue this, those that add features\n> (assert.patch) should be omitted...those that fix a bug are appropriate.\n> Not that I haven't looked at the patch itself, only at the descriptions\n> presented...\n> \n> \n> \n> \n> > > assert.patch\n> > > \n> > > \tadds a switch to turn on/off the assert checking if enabled at compile\n> > > \ttime. You can now compile postgres with assert checking and disable it\n> > > \tat runtime in a production environment.\n> \n> \tNew feature...not a bug fix...\n\nOld feature, I submitted this patch more than six months ago, I'm using it\nfrom that date without problems. It simply adds a global variable to a macro.\n\n> > > async-unlisten.patch\n> > > \n> > > \tdeclares Async_Unlisten() external so that it can be called by user\n> > > \tmodules.\n> \n> \tNew feature...not a bug fix...\n\nJust declaring external an existing function shouldn't break anything.\n\n> \n> > > exec-limit.patch\n> > > \n> > > \tremoves the #ifdef NOT_USED around ExecutorLimit(). It is used.\n> \n> \tif it has a \"#ifdef NOT_USED\" around it, then how is it used? \n\nIt is used by a loadable module in contrib.\n\n> \n> > > exitpg.patch\n> > > \n> > > \tlimits recursive calls to exitpg() preventing an infinite loop\n> > > \tif an error is found inside exitpg.\n> \n> \tPotential bug...but how is it triggered?\n\nI don't know exactly how it is triggered but I have seen it happen. This\npatch tries to limit damages.\n\n> \n> > > libpgtcl-listen.patch\n> > > \n> > > \tJust a change from upper to lowercase of an sql command in libpgtcl,\n> > > \ttotally harmless.\n> \t\n> \t??\n\nNot really important, it just changes 'LISTEN' to 'listen' so that when I\nlook at the postgres log the query is shown in lowercase like the others.\nTotally harmless.\n\n> \n> > > new-locks.patch\n> > > \n> > > \tAfter long studying and many debugging sessions I have finally\n> > > \tunderstood how the low level locks work.\n> > > \tI have completely rewritten lock.c cleaning up the code and adding\n> > > \tbetter assert checking. I have also added some fields to the lock\n> > > \tand xid tags for better support of user locks. This patch includes\n> > > \talso a patch submitted by Bruce Momjian which changes the handling\n> > > \tof lock priorities. It can however be disabled if an option is set\n> > > \tin pg_options, see tprintf.patch (Bruce patch works by building\n> > > \tthe queue in reverse priority order, my old patch kept the queue in\n> > > \tdecreasing order and traversed it from the other side).\n> \n> \tI don't understand this one, but it sounds like its negates the\n> work you just did? Again, sounds like a feature change, not a bug fix,\n> since you have recently re-written this...\n\nI rewrote part of the work I did one year ago, cleaned up the original lock\ncode and did some small changes. I admit that this should be handled with\ncare, but, again, I'm using it from months and it seems to work well.\n\n> \n> > > pg-flush.patch\n> > > \n> > > \tremoves an unnecessary flush in libpq reducing network traffic and\n> > > \tincreasing performance.\n> \n> \tDidn't we just do a protocol rewrite? \n\nI don't know about it, but Bruce approved the patch some months ago, so it\nhas probably been included in the new protocol.\n\n> \n> > > relname.patch\n> > > \n> > > \tan utility which returns the relname corresponding to a given oid.\n> > > \tUseful for debug messages (see vacum.patch).\n> \n> \tIs this a new program? src/bin/relname?\n\nIt is just a new utility function which returns the name of a relation given\nits oid. It may be used to show relname information in user messages and\ntraces.\n\n> \n> > > sequence.patch\n> > > \n> > > \tadded a setval() function which enables othe owner of a sequence\n> > > \tto set its value without need to delete and recreate it.\n> \n> \tFeature change, not a bug...\n> \n> > > sinval.patch\n> > > \n> > > \tfixes a problem in SI cache which causes table overflow if some\n> > > \tbackend is idle for a long time while other backends keep adding\n> > > \tentries.\n> > > \tIt uses the new signal handling implemented in tprintf.patch.\n> > > \tI have also increacasesed the max number of backends from 32 to 64 and\n> > > \tthe table size from 1000 to 5000.\n> \n> \tBug fix...\n\nYes, and very nasty.\n\n> \n> > > spin-lock.patch\n> > > \n> > > \tI'm not sure if this is really useful, but it seems stupid to have\n> > > \ta backend wasting cpu cycles in a busy loop while the process which\n> > > \tshould release the lock is waiting for the cpu. So I added a call\n> > > \tto process_yield() if the spin lock can't obtained.\n> > > \tThis has been implemented and tested only on Linux. I don't know if\n> > > \tother OS have process_yield(). If someone can check please do it.\n> \n> \tSounds like a bug fix to me...\n\nNo, it is just a performance optimization. And I'm not sure how much we could\ngain from this patch, so I ask a discussion about it.\n\n> \n> > > tprintf.patch\n> > > \n> > > \tadds functions and macros which implement a conditional trace package\n> > > \twith the ability to change flags and numeric options of running\n> > > \tbackends at runtime.\n> > > \tOptions/flags can be specified in the command line and/or read from\n> > > \tthe file pg_options in the data directory.\n> > > \tRunning backends can be forced to update their options from this file\n> > > \tby sending them a SIGHUP signal (this is the convention used by most\n> > > \tunix daemons so I changed the meaning of SIGHUP).\n> > > \tOptions can be debugging flags used by the trace package or any other\n> > > \tnumeric\tvalue used by the backend, for example the deadlock_timeout.\n> > > \tHaving flags and options specified at runtime and changed while the\n> > > \tbackends are running can greatly simplify the debugging and tuning\n> > > \tof the database. New options can be defined in utils/misc/trace.c and\n> > > \tinclude/utils/trace.h. As an example of the usage of this package\n> > > \tsee lock.c and proc.c which make use of new runtime options.\n> \n> \tDefinitely have to say new feature...\n\nYes, but very handy. I highly recommend it, maybe after 6.3.\n\n> \n> > > vacuum.patch\n> > > \n> > > \tadds a debug message to vacuum that prints the name of a table or\n> > > \tindex before vacuuming it, if the verbose keyword is set.\n> > > \tThis is useful to know which table is causing troubles if a\n> > > \tvacuum all crashes. Currently table information is printed only\n> > > \tat the end of each vacuum operation and is never printed if the\n> > > \tvacuum crashes.\n> \n> \tAgain, a feature more then a bug fix...\n\nJust a debug message, should not break anything.\n\n\nI would like also to add a new patch for the oracle_compat patch added in\nversion 6.2.1p7, which is definitely broken. Try this and you will see:\n\ntest=> create table t (x text);\nCREATE\ntest=> insert into t values('1');\nINSERT 17422 1\ntest=> insert into t values('12');\nINSERT 17423 1\ntest=> insert into t values('123');\nINSERT 17424 1\ntest=> insert into t values('1234');\nINSERT 17425 1\ntest=> insert into t values('12345');\nINSERT 17426 1\ntest=> insert into t values('123456');\nINSERT 17427 1\ntest=> insert into t values('1234567');\nINSERT 17428 1\ntest=> insert into t values('12345678');\nINSERT 17429 1\ntest=> insert into t values('123456789');\nINSERT 17430 1\ntest=> insert into t values('1234567890');\nINSERT 17431 1\ntest=> insert into t values('12345678901');\nINSERT 17432 1\ntest=> insert into t values('123456789012');\nINSERT 17433 1\ntest=> insert into t values('1234567890123');\nINSERT 17434 1\ntest=> insert into t values('12345678901234');\nINSERT 17435 1\ntest=> insert into t values('123456789012345');\nINSERT 17436 1\ntest=> insert into t values('1234567890123456');\nINSERT 17437 1\ntest=> select x from t;\n x\n----------------\n 1\n 12\n 123\n 1234\n 12345\n 123456\n 1234567\n 12345678\n 123456789\n 1234567890\n 12345678901\n 123456789012\n 1234567890123\n 12345678901234\n 123456789012345\n1234567890123456\n(16 rows)\n\ntest=> select substr(x,1) from t;\nsubstr \n----------------\n1 \n12 \n123 \n12347 \n12345 \n123456 \n1234567 \n12345678 \n123456789 \n1234567890 \n12345678901 \n123456789012? \n1234567890123 \n12345678901234 \n123456789012345 \n1234567890123456\n(16 rows)\n\nNote the extra character after 1234 and 123456789012. The bug is random and\nis present also in the latest version sent me by Bruce. The following patch\ncorrects the problem. BTW, why are the results aligned differently by psql?\n\n* src/backend/utils/adt/oracle_compat-621p7.c\tWed Jan 28 00:11:16 1998\n--- src/backend/utils/adt/oracle_compat.c\tTue Feb 3 14:48:42 1998\n***********\n* 510,526 ***\n \t\t\t ptr_ret;\n \tint\t\t\tlen;\n \n \tif ((string == (text ) NULL) \|\|\n! \t\t(m <= 0) \|\| (n <= 0) \|\|\n! \t\t((len = VARSIZE(string) - VARHDRSZ - m + 1) <= 0))\n \t\treturn string;\n \n! \tlen = len + 1 < n ? len + 1 : n;\n \n \tret = (text ) palloc(VARHDRSZ + len);\n \tVARSIZE(ret) = VARHDRSZ + len;\n \n! \tptr = VARDATA(string) + m - 1;\n \tptr_ret = VARDATA(ret);\n \n \twhile (len--)\n--- 510,529 ----\n \t\t\t ptr_ret;\n \tint\t\t\tlen;\n \n+ \t/ Make offset 0-based /\n+ \tm--;\n+ \n \tif ((string == (text ) NULL) \|\|\n! \t\t(m < 0) \|\| (n <= 0) \|\|\n! \t\t((len = VARSIZE(string) - VARHDRSZ - m) <= 0))\n \t\treturn string;\n \n! \tlen = (len < n) ? len : n;\n \n \tret = (text ) palloc(VARHDRSZ + len);\n \tVARSIZE(ret) = VARHDRSZ + len;\n \n! \tptr = VARDATA(string) + m;\n \tptr_ret = VARDATA(ret);\n \n \twhile (len--)\n\n\n-- \nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n\| Massimo Dal Zotto e-mail: [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, 16 Feb 1998 10:43:46 +0100 (MET)", "msg_from": "Massimo Dal Zotto <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Massimo patch" }, { "msg_contents": "Peanut Gallery Vote on tprintf patch:\n\nI would vote to put this into 6.3.\n\nRATIONALE: It seems to offer database _users_ the potential of\nproviding more useful information about (potential) bugs to\nthe development team. Seems like it could help speed up\nthe development process. Getting it into the code base\nsooner rather than later seems desireable.\n\nMarc Zuckman\[email protected]\n\n", "msg_date": "Mon, 16 Feb 1998 11:57:33 -0500 (EST)", "msg_from": "Marc Howard Zuckman <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Massimo patch" }, { "msg_contents": "On Mon, 16 Feb 1998, Marc Howard Zuckman wrote:\n\n> Peanut Gallery Vote on tprintf patch:\n> \n> I would vote to put this into 6.3.\n> \n> RATIONALE: It seems to offer database _users_ the potential of\n> providing more useful information about (potential) bugs to\n> the development team. Seems like it could help speed up\n> the development process. Getting it into the code base\n> sooner rather than later seems desireable.\n\n\tExcept, it is a new feature, not a bug fix...except for a very\nselect few features that overlapped the beta freeze (subselects being the\nonly one that I can thnk of), nothing new is going into this release...\n\n\tTo add to it, the patches are against v6.2.1, which would mean\nhaving to spend time away from current bug fixes to safely be able to\nmerge them in.\n\n\tAs I said before, all of the patches look good and I want to get\nthem in, but not until after v6.3 is released.\n\n\tMassimo, it would really help things if you could submit a new\npatch, after v6.3 is released, against v6.3...I've kept the original\npatch, and will manually apply it if I have to, but...:)\n\n\n\n", "msg_date": "Mon, 16 Feb 1998 12:10:09 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Massimo patch" }, { "msg_contents": "> \n> Hi hackers,\n> \n> I have old patches for version 6.2.1p6 which fix some problems and add\n> new features. Here is a short description of each patch file:\n> \n> \n> assert.patch\n> \n> \tadds a switch to turn on/off the assert checking if enabled at compile\n> \ttime. You can now compile postgres with assert checking and disable it\n> \tat runtime in a production environment.\n> \n> async-unlisten.patch\n> \n> \tdeclares Async_Unlisten() external so that it can be called by user\n> \tmodules.\n> \n> exec-limit.patch\n> \n> \tremoves the #ifdef NOT_USED around ExecutorLimit(). It is used.\n> \n> exitpg.patch\n> \n> \tlimits recursive calls to exitpg() preventing an infinite loop\n> \tif an error is found inside exitpg.\n> \n> libpgtcl-listen.patch\n> \n> \tJust a change from upper to lowercase of an sql command in libpgtcl,\n> \ttotally harmless.\n> \n> new-locks.patch\n> \n> \tAfter long studying and many debugging sessions I have finally\n> \tunderstood how the low level locks work.\n> \tI have completely rewritten lock.c cleaning up the code and adding\n> \tbetter assert checking. I have also added some fields to the lock\n> \tand xid tags for better support of user locks. This patch includes\n> \talso a patch submitted by Bruce Momjian which changes the handling\n> \tof lock priorities. It can however be disabled if an option is set\n> \tin pg_options, see tprintf.patch (Bruce patch works by building\n> \tthe queue in reverse priority order, my old patch kept the queue in\n> \tdecreasing order and traversed it from the other side).\n> \n> pg-flush.patch\n> \n> \tremoves an unnecessary flush in libpq reducing network traffic and\n> \tincreasing performance.\n> \n> relname.patch\n> \n> \tan utility which returns the relname corresponding to a given oid.\n> \tUseful for debug messages (see vacum.patch).\n> \n> sequence.patch\n> \n> \tadded a setval() function which enables othe owner of a sequence\n> \tto set its value without need to delete and recreate it.\n> \n> sinval.patch\n> \n> \tfixes a problem in SI cache which causes table overflow if some\n> \tbackend is idle for a long time while other backends keep adding\n> \tentries.\n> \tIt uses the new signal handling implemented in tprintf.patch.\n> \tI have also increacasesed the max number of backends from 32 to 64 and\n> \tthe table size from 1000 to 5000.\n> \n> spin-lock.patch\n> \n> \tI'm not sure if this is really useful, but it seems stupid to have\n> \ta backend wasting cpu cycles in a busy loop while the process which\n> \tshould release the lock is waiting for the cpu. So I added a call\n> \tto process_yield() if the spin lock can't obtained.\n> \tThis has been implemented and tested only on Linux. I don't know if\n> \tother OS have process_yield(). If someone can check please do it.\n> \n> tprintf.patch\n> \n> \tadds functions and macros which implement a conditional trace package\n> \twith the ability to change flags and numeric options of running\n> \tbackends at runtime.\n> \tOptions/flags can be specified in the command line and/or read from\n> \tthe file pg_options in the data directory.\n> \tRunning backends can be forced to update their options from this file\n> \tby sending them a SIGHUP signal (this is the convention used by most\n> \tunix daemons so I changed the meaning of SIGHUP).\n> \tOptions can be debugging flags used by the trace package or any other\n> \tnumeric\tvalue used by the backend, for example the deadlock_timeout.\n> \tHaving flags and options specified at runtime and changed while the\n> \tbackends are running can greatly simplify the debugging and tuning\n> \tof the database. New options can be defined in utils/misc/trace.c and\n> \tinclude/utils/trace.h. As an example of the usage of this package\n> \tsee lock.c and proc.c which make use of new runtime options.\n> \n> \tOld files using int flags or variables can be easily changed to\n> \tuse the new package by substituting the old variable with a #define\n> \tlike in the following example:\n> \t\n> \t / int my_flag = 0; /\n> \t #include \"trace.h\"\n> \t #define my_flag pg_options[OPT_MYFLAG]\n> \n> \tI have done it in postgres.c and some other files and now I can turn\n> \ton/off any single debug flag on the fly with a simple shell script.\n> \tI have removed the IpcConfigTip() from ipc.c, it should better be\n> \tdescribed in the postgres manual instead of being printed on stderr.\n> \n> \tThis patch provides also a new format of debugging messages which\n> \tare always in a single line with a timestamp and the backend pid:\n> \n> \t #timestamp #pid #message\n> \t 980127.17:52:14.173 [29271] StartTransactionCommand\n> \t 980127.17:52:14.174 [29271] ProcessUtility: drop table t;\n> \t 980127.17:52:14.186 [29271] SIIncNumEntries: table is 70% full\n> \t 980127.17:52:14.186 [29286] Async_NotifyHandler\n> \t 980127.17:52:14.186 [29286] Waking up sleeping backend process\n> \t 980127.19:52:14.292 [29286] Async_NotifyFrontEnd\n> \t 980127.19:52:14.413 [29286] Async_NotifyFrontEnd done\n> \t 980127.19:52:14.466 [29286] Async_NotifyHandler done\n> \n> \tThis improves the readability of the log and allows one to understand\n> \texactly which backend is doing what and at which time. It also makes\n> \teasier to write simple awk or perl scripts which monitor the log to\n> \tdetect database errors or problem, or to compute transaction times.\n> \n> \tThe patch changes also the meaning of signals used by postgres, as\n> \tdescribed by the following table:\n> \n> \t\t postmaster\t\t\t\tbackend\n> \n> \tSIGHUP kill(,sighup)\t\t\tread_pg_options\n> \tSIGINT\t kill(,sigint), die\t\t\tdie\n> \tSIGCHLD\t reaper\t\t\t\t-\n> \tSIGTTIN\t ignored\t\t\t\t-\n> \tSIGTTOU\t ignored\t\t\t\t-\n> \tSIGQUIT\t die\t\t\t\t\thandle_warn\n> \tSIGTERM\t kill(,sigterm), kill(,9), die\tdie\n> \tSIGCONT\t dumpstatus\t\t\t\t-\n> \tSIGPIPE\t ignored\t\t\t\tdie\n> \tSIGFPE\t -\t\t\t\t\tFloatExceptionHandler\n> \tSIGTSTP\t -\t\t\t\t\tignored (alive test)\n> \tSIGUSR1\t kill(,sigusr1), die\t\tquickdie\n> \tSIGUSR2\t kill(,sigusr2)\t\t\tAsync_NotifyHandler\n> \t\t\t\t\t\t\t(also SI buffer flush)\n> \n> \tThe main changes to the old implementation are SIGQUIT instead of\n> \tSIGHUP to handle warns, SIGHUP to reread pg_options and redirection\n> \tto all backends of SIGHUP, SIGINT, SIGTERM, SIGUSR1 and SIGUSR2.\n> \tIn this way some of the signals sent to the postmaster can be sent\n> \tautomatically to all the backends. To shut down postgres one needs\n> \tonly to send a SIGTERM to postmaster and it will stop automatically\n> \tall the backends. This new signal handling mechanism is also used\n> \tto prevent SI cache table overflows: when a backend detects the SI\n> \ttable full at 70% it simply sends a signal to the postmaster which\n> \twill wake up all idle backends and make them flush the cache.\n> \n> vacuum.patch\n> \n> \tadds a debug message to vacuum that prints the name of a table or\n> \tindex before* vacuuming it, if the verbose keyword is set.\n> \tThis is useful to know which table is causing troubles if a\n> \tvacuum all crashes. Currently table information is printed only\n> \tat the end of each vacuum operation and is never printed if the\n> \tvacuum crashes.\n> \n> -- \n> Massimo Dal Zotto\n\nMassimo, now that 6.3 is released, any chance of getting these patches\nagainst the 6.3 source code?\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Sun, 15 Mar 1998 22:07:20 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> > Hi hackers,\n> > \n> > I have old patches for version 6.2.1p6 which fix some problems and add\n> > new features. Here is a short description of each patch file:\n\n> > spin-lock.patch\n> > \n> > \tI'm not sure if this is really useful, but it seems stupid to have\n> > \ta backend wasting cpu cycles in a busy loop while the process which\n> > \tshould release the lock is waiting for the cpu. So I added a call\n> > \tto process_yield() if the spin lock can't obtained.\n> > \tThis has been implemented and tested only on Linux. I don't know if\n> > \tother OS have process_yield(). If someone can check please do it.\n\nThe generic way to do this is\n\n select( NULL_FDSET, NULL_FDSET, NULL_FDSET, &delaytime, NULL);\n\nDelay time may be 0, but a random value between 0 and say 30 msec seems\nto be optimal. Hard busy wait spinlocks cause huge performance problems with\nheavily loaded systems and lots of postgres backends. Basically one backend\nends up with the lock and gets scheduled out holding it, every else queues\nup busywaiting behind this one. But the backend holding the lock cannot\nrelease it until all the other backeds waiting on the lock exhaust a full\ntimeslice busywaiting. Get 20 of these guys going (like on a busy website) and\nthe system pretty much stops doing any work at all.\n\nI say we should get this in as soon as we can.\n\n> > -- \n> > Massimo Dal Zotto\n> \n> Massimo, now that 6.3 is released, any chance of getting these patches\n> against the 6.3 source code?\n> \n> -- \n> Bruce Momjian \| 830 Blythe Avenue\n> [email protected] \| Drexel Hill, Pennsylvania 19026\n\n\n-dg\n\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - I realize now that irony has no place in business communications.\n\n", "msg_date": "Mon, 16 Mar 1998 15:27:02 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "On Mon, 16 Mar 1998, David Gould wrote:\n\n> The generic way to do this is\n> \n> select( NULL_FDSET, NULL_FDSET, NULL_FDSET, &delaytime, NULL);\n> \n> Delay time may be 0, but a random value between 0 and say 30 msec seems\n> to be optimal. Hard busy wait spinlocks cause huge performance problems with\n> heavily loaded systems and lots of postgres backends. Basically one backend\n> ends up with the lock and gets scheduled out holding it, every else queues\n> up busywaiting behind this one. But the backend holding the lock cannot\n> release it until all the other backeds waiting on the lock exhaust a full\n> timeslice busywaiting. Get 20 of these guys going (like on a busy website) and\n> the system pretty much stops doing any work at all.\n> \n> I say we should get this in as soon as we can.\n\nCan you submit an appropriate patch that can be included in the mega-patch\nto be created on Sunday?\n\nThanks...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 16 Mar 1998 22:47:57 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \n> On Mon, 16 Mar 1998, David Gould wrote:\n> \n> > The generic way to do this is\n> > \n> > select( NULL_FDSET, NULL_FDSET, NULL_FDSET, &delaytime, NULL);\n> > \n> > Delay time may be 0, but a random value between 0 and say 30 msec seems\n> > to be optimal. Hard busy wait spinlocks cause huge performance problems with\n> > heavily loaded systems and lots of postgres backends. Basically one backend\n> > ends up with the lock and gets scheduled out holding it, every else queues\n> > up busywaiting behind this one. But the backend holding the lock cannot\n> > release it until all the other backeds waiting on the lock exhaust a full\n> > timeslice busywaiting. Get 20 of these guys going (like on a busy website) and\n> > the system pretty much stops doing any work at all.\n> > \n> > I say we should get this in as soon as we can.\n> \n> Can you submit an appropriate patch that can be included in the mega-patch\n> to be created on Sunday?\n\nJust a warning that this is not going to be easy. We have OS-specific\ncode for spinlocks in include/storage/s_lock.h and\nbackend/storage/buffer/s_lock.c. So each S_LOCK macro call has to have\nits test-and-set logic de-coupled with its while-lock-fail-try-again\nlogic. Most of them are easy, but some like VAX:\n\n#define S_LOCK(addr) __asm__(\"1: bbssi $0,(%0),1b\": :\"r\"(addr))\n\nare hard to de-couple. Now, I did not know we supported NetBSD on VAX. \nDoes it work, anyone? Can I remove it?\n\nThis is going to be pretty tough to test on every platform we support,\nso if it is done now, it will have to be done carefully.\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Mon, 16 Mar 1998 22:20:48 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> > Can you submit an appropriate patch that can be included in the \n> > mega-patch to be created on Sunday?\n> \n> Just a warning that this is not going to be easy. We have OS-specific\n> code for spinlocks in include/storage/s_lock.h and\n> backend/storage/buffer/s_lock.c. So each S_LOCK macro call has to\n> have its test-and-set logic de-coupled with its \n> while-lock-fail-try-again logic. \n> Most of them are easy, but some like VAX:\n> \n> #define S_LOCK(addr) __asm__(\"1: bbssi $0,(%0),1b\": :\"r\"(addr))\n> \n> are hard to de-couple. Now, I did not know we supported NetBSD on \n> VAX. Does it work, anyone? Can I remove it?\n\nNetBSD on VAX in on our supported list, and was verified for v6.3 by Tom\nHelbekkmo.\n\n> This is going to be pretty tough to test on every platform we support,\n> so if it is done now, it will have to be done carefully.\n\nIs this behavior in v6.2.x? In any case, if it is anything but minimally\ntrivial, it should be given a test on every supported platform, since it\nhits the heart of the platform-specific code, doesn't it? Seems like it\nshould be put into the CVS tree and shaken out until the next release...\n\n - Tom\n", "msg_date": "Tue, 17 Mar 1998 03:46:30 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "On Tue, 17 Mar 1998, Thomas G. Lockhart wrote:\n\n> > This is going to be pretty tough to test on every platform we support,\n> > so if it is done now, it will have to be done carefully.\n> \n> Is this behavior in v6.2.x? In any case, if it is anything but minimally\n> trivial, it should be given a test on every supported platform, since it\n> hits the heart of the platform-specific code, doesn't it? Seems like it\n> should be put into the CVS tree and shaken out until the next release...\n\n\tNot realizing what was involved, I have to agree here...after I\nget a post-release patch out on Sunday? :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 16 Mar 1998 23:49:04 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \n> > > Can you submit an appropriate patch that can be included in the \n> > > mega-patch to be created on Sunday?\n\n> > are hard to de-couple. Now, I did not know we supported NetBSD on \n> > VAX. Does it work, anyone? Can I remove it?\n> \n> NetBSD on VAX in on our supported list, and was verified for v6.3 by Tom\n> Helbekkmo.\n> \n> > This is going to be pretty tough to test on every platform we support,\n> > so if it is done now, it will have to be done carefully.\n> \n> Is this behavior in v6.2.x? In any case, if it is anything but minimally\n> trivial, it should be given a test on every supported platform, since it\n> hits the heart of the platform-specific code, doesn't it? Seems like it\n> should be put into the CVS tree and shaken out until the next release...\n\nYea, that is what I was hinting at.\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Mon, 16 Mar 1998 22:54:53 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> > > > Can you submit an appropriate patch that can be included in the \n> > > > mega-patch to be created on Sunday?\n> \n> > > are hard to de-couple. Now, I did not know we supported NetBSD on \n> > > VAX. Does it work, anyone? Can I remove it?\n> > \n> > NetBSD on VAX in on our supported list, and was verified for v6.3 by Tom\n> > Helbekkmo.\n> > \n> > > This is going to be pretty tough to test on every platform we support,\n> > > so if it is done now, it will have to be done carefully.\n> > \n> > Is this behavior in v6.2.x? In any case, if it is anything but minimally\n> > trivial, it should be given a test on every supported platform, since it\n> > hits the heart of the platform-specific code, doesn't it? Seems like it\n> > should be put into the CVS tree and shaken out until the next release...\n> \n> Yea, that is what I was hinting at.\n> \n> -- \n> Bruce Momjian \| 830 Blythe Avenue\n> [email protected] \| Drexel Hill, Pennsylvania 19026\n\n\nI tend to agree but am willing to compromise.\n\nCan we do only the easy platforms at this time and then fix the others later?\n\nSince S_LOCK is a macro, it could be\n\n#define S_LOCK s_lock_with_backoff\n\non the easy platforms and\n\n#define S_LOCK original_definition\n\non the tricky or hard to test platforms\n\nIf this will work, I am willing to hack this together tomorrow.\nWhat is the time frame for accepting a patch like this?\n\n-dg\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468\nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - I realize now that irony has no place in business communications.\n\n\n", "msg_date": "Mon, 16 Mar 1998 20:04:23 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "David Gould wrote:\n> \n> > > > > Can you submit an appropriate patch that can be included in the\n> > > > > mega-patch to be created on Sunday?\n> >\n> > > > are hard to de-couple. Now, I did not know we supported NetBSD on\n> > > > VAX. Does it work, anyone? Can I remove it?\n> > >\n> > > NetBSD on VAX in on our supported list, and was verified for v6.3 by Tom\n> > > Helbekkmo.\n> > >\n> > > > This is going to be pretty tough to test on every platform we support,\n> > > > so if it is done now, it will have to be done carefully.\n> > >\n> > > Is this behavior in v6.2.x? In any case, if it is anything but minimally\n> > > trivial, it should be given a test on every supported platform, since it\n> > > hits the heart of the platform-specific code, doesn't it? Seems like it\n> > > should be put into the CVS tree and shaken out until the next release...\n> >\n> > Yea, that is what I was hinting at.\n> >\n> > --\n> > Bruce Momjian \| 830 Blythe Avenue\n> > [email protected] \| Drexel Hill, Pennsylvania 19026\n> \n> I tend to agree but am willing to compromise.\n> \n> Can we do only the easy platforms at this time\n> If this will work, I am willing to hack this together tomorrow.\n> What is the time frame for accepting a patch like this?\n\nWhat do you think it would take to exercise the patch heavily enough on\nany of the affected platforms to ensure that it behaves no worse than\nthe current code? Is a test case easy to set up and run? If so, you can\nprobably get ~5 platforms tested in the next few days, and reduce the\nrisk of including this in the mega-patch.\n\nAlternatively, we could do this the day _after_ the mega-patch, so that\nthe two are decoupled, and have a nice \"slock patch\" a few days later.\n\nAt the moment, there aren't any large backend patches waiting to go (I\nthink Vadim is still sleeping to recover from v6.3 :)\n\n - Tom\n", "msg_date": "Tue, 17 Mar 1998 04:16:10 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "On Mon, 16 Mar 1998, David Gould wrote:\n\n> If this will work, I am willing to hack this together tomorrow.\n> What is the time frame for accepting a patch like this?\n\n\tAssuming that its clean (clean meaning that Bruce fully approves\nof it, as this is his area of the code...well, one of them\ngrin)...tomorrow would be great :) If Bruce has any doubts though, it\ndoesn't go in until after I do the patch I want to do...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Tue, 17 Mar 1998 00:51:23 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> David Gould wrote:\n> > \n> > > > > > Can you submit an appropriate patch that can be included in the\n> > > > > > mega-patch to be created on Sunday?\n> > >\n> > > > > are hard to de-couple. Now, I did not know we supported NetBSD on\n> > > > > VAX. Does it work, anyone? Can I remove it?\n> > > >\n> > > > NetBSD on VAX in on our supported list, and was verified for v6.3 by Tom\n> > > > Helbekkmo.\n> > > >\n> > > > > This is going to be pretty tough to test on every platform we support,\n> > > > > so if it is done now, it will have to be done carefully.\n> > > >\n> > > > Is this behavior in v6.2.x? In any case, if it is anything but minimally\n> > > > trivial, it should be given a test on every supported platform, since it\n> > > > hits the heart of the platform-specific code, doesn't it? Seems like it\n> > > > should be put into the CVS tree and shaken out until the next release...\n> > >\n> > > Yea, that is what I was hinting at.\n> > >\n> > > --\n> > > Bruce Momjian \| 830 Blythe Avenue\n> > > [email protected] \| Drexel Hill, Pennsylvania 19026\n> > \n> > I tend to agree but am willing to compromise.\n> > \n> > Can we do only the easy platforms at this time\n> > If this will work, I am willing to hack this together tomorrow.\n> > What is the time frame for accepting a patch like this?\n> \n> What do you think it would take to exercise the patch heavily enough on\n> any of the affected platforms to ensure that it behaves no worse than\n> the current code? Is a test case easy to set up and run? If so, you can\n> probably get ~5 platforms tested in the next few days, and reduce the\n> risk of including this in the mega-patch.\n\nSimple but clumsy test:\n\nMultiple backends all doing single row inserts into private tables. Run it\nfor a fixed time period and add up the total number of rows. Do this for both\na small (eg 2) and large (eg 20 or 40) number of backends.\n\npsuedo perl /sh / sql follows\n# driver <number of users> <delay time> <run time>\nsynchtime = timenow\nstarttime = synctime + delaytime;\nendtime = starttime + runtime\nfor user 1 to number of users\n singleuser synchtime delay_time run_time\n\n# singleuser <userno> <starttime> <endtime>\nexec sql 'create table user$user (i int);'\n\nsleep until start time\nn = 1;\nwhile timenow < endtime\n exec sql \"insert into user$user values($n);\"\n\nprint \"user $user inserted $n\"\n\nThe figure of merit is the total number of inserts. Oh and that the system\ndoes not fall over.\n\nBut, it may be best to leave this until after the mega patch. I am not sure\nI want to share the blame ;-).\n\nJust to fill me in, where does the mega patch fall in with the next release\nsnapshot? That is, if this misses the mega patch is it waiting until 6.4?\n\n-dg\n\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - I realize now that irony has no place in business communications.\n\n", "msg_date": "Mon, 16 Mar 1998 22:38:14 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "On Mon, 16 Mar 1998, David Gould wrote:\n\n> Just to fill me in, where does the mega patch fall in with the next release\n> snapshot? That is, if this misses the mega patch is it waiting until 6.4?\n\n\tThat is pretty much up to you, actually. There is nothing wrong\nwith a clean patch for this being placed in the patches directory, if it\ncan be done shortly after the post-release patch. Basically, starting\nApril 1st (or so), development basically starts up again, so backtracking\na patch from v6.4-alpha to v6.3 release can prove difficult :(\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Tue, 17 Mar 1998 03:40:44 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \n> On Mon, 16 Mar 1998, David Gould wrote:\n> \n> > If this will work, I am willing to hack this together tomorrow.\n> > What is the time frame for accepting a patch like this?\n> \n> \tAssuming that its clean (clean meaning that Bruce fully approves\n> of it, as this is his area of the code...well, one of them\n> grin)...tomorrow would be great :) If Bruce has any doubts though, it\n> doesn't go in until after I do the patch I want to do...\n\nDavid, go for it. The code is all local in two files, and I think you\ncan basically change all the do{test-and-set} while(lock-is-false)\nloops to:\n\n\tdo{test-and-set} while(lock-is-false && select ())\n\nPretty easy. No need to test multiple platforms. The ones where the\nloop is integrated into the asm(), leave them for later.\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Tue, 17 Mar 1998 09:20:00 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> But, it may be best to leave this until after the mega patch. I am not \n> sure I want to share the blame ;-).\n> \n> Just to fill me in, where does the mega patch fall in with the next \n> release snapshot? That is, if this misses the mega patch is it waiting \n> until 6.4?\n\nI would guess that we could post a separate patch any time soon,\nespecially since the changes are apparently isolated to only a few\nplaces in the code. In the last release, we posted ~7 patches, each\nindependent of the others, and generated on our local source trees. Each\nof the patches was, however, fairly simple, quite often only one or a\nfew lines of change, and were intended as bug fixes. Also, they were\neasily tested. In any case, at a minimum the regression test should be\nrun (and passed!).\n\n - Tom\n", "msg_date": "Tue, 17 Mar 1998 16:45:10 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "* Bruce Momjian\n\|\n\| Just a warning that this is not going to be easy. We have OS-specific\n\| code for spinlocks in include/storage/s_lock.h and\n\| backend/storage/buffer/s_lock.c. So each S_LOCK macro call has to have\n\| its test-and-set logic de-coupled with its while-lock-fail-try-again\n\| logic. Most of them are easy, but some like VAX:\n\| \n\| #define S_LOCK(addr) __asm__(\"1: bbssi $0,(%0),1b\": :\"r\"(addr))\n\| \n\| are hard to de-couple. Now, I did not know we supported NetBSD on VAX. \n\| Does it work, anyone? Can I remove it?\n\nYes, it works. No, please don't break it. Heck, I only just got it\nin in time for 6.3! :-) The not-so-busy-waiting-spinlock stuff can be\nput in on a platform at a time -- I'll expand the VAX version to do\nthe right thing once someone has done another platform, so I can see\nwhat's the preferred way of doing it.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "18 Mar 1998 22:50:35 +0100", "msg_from": "Tom Ivar Helbekkmo <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> * Bruce Momjian\n> \| Just a warning that this is not going to be easy. We have OS-specific\n> \| code for spinlocks in include/storage/s_lock.h and\n> \| backend/storage/buffer/s_lock.c. So each S_LOCK macro call has to have\n> \| its test-and-set logic de-coupled with its while-lock-fail-try-again\n> \| logic. Most of them are easy, but some like VAX:\n> \| \n> \| #define S_LOCK(addr) __asm__(\"1: bbssi $0,(%0),1b\": :\"r\"(addr))\n> \| \n> \| are hard to de-couple. Now, I did not know we supported NetBSD on VAX. \n> \| Does it work, anyone? Can I remove it?\n> \n> Yes, it works. No, please don't break it. Heck, I only just got it\n> in in time for 6.3! :-) The not-so-busy-waiting-spinlock stuff can be\n> put in on a platform at a time -- I'll expand the VAX version to do\n> the right thing once someone has done another platform, so I can see\n> what's the preferred way of doing it.\n>\n> -tih\n\nI won't. I hope.\n\nSeriously, if you want to, please create a function to emulate the following:\n\n/\n tas(lock)\n \n Access to platform specific test_and_set functionality. Given pointer to\n * lock attempts to acquire the lock atomically.\n \n Returns 0 for success, nonzero for failure.\n /\ntypedef slock_t unsigned char;\t\t/ or whatever works on the platform /\n\nint tas(slock_t lock)\n{\n slock_t\ttmp;\n\n /* atomic, interlocked /\n tmp = lock;\n lock = -1; \t\t\t/ any nonzero will do here /\n\n return (tmp != 0);\n}\n\nGiven this, I can fold the VAX right into the grand scheme, just like a\nnormal computer (;-)).\n\n-dg\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - I realize now that irony has no place in business communications.\n\n\n\n", "msg_date": "Wed, 18 Mar 1998 15:51:51 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \n> Bruce Momjian\n> \|\n> \| Just a warning that this is not going to be easy. We have OS-specific\n> \| code for spinlocks in include/storage/s_lock.h and\n> \| backend/storage/buffer/s_lock.c. So each S_LOCK macro call has to have\n> \| its test-and-set logic de-coupled with its while-lock-fail-try-again\n> \| logic. Most of them are easy, but some like VAX:\n> \| \n> \| #define S_LOCK(addr) __asm__(\"1: bbssi $0,(%0),1b\": :\"r\"(addr))\n> \| \n> \| are hard to de-couple. Now, I did not know we supported NetBSD on VAX. \n> \| Does it work, anyone? Can I remove it?\n> \n> Yes, it works. No, please don't break it. Heck, I only just got it\n> in in time for 6.3! :-) The not-so-busy-waiting-spinlock stuff can be\n> put in on a platform at a time -- I'll expand the VAX version to do\n> the right thing once someone has done another platform, so I can see\n> what's the preferred way of doing it.\n\nOK, now I know that the VAX stuff is still used and supported. Good. \nWe don't have good platform-specific information on NetBSD and Linux\nports.\n\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Wed, 18 Mar 1998 21:35:56 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> OK, now I know that the VAX stuff is still used and supported. Good.\n> We don't have good platform-specific information on NetBSD and Linux\n> ports.\n\n?? \n\nHave you checked the hardcopy or html versions of the \"Supported\nPlatforms\" section in the Administrator's Guide? _Every_ entry in this\nwas updated and refreshed at the end of February.\n\nWhat other info should we be collecting for this?\n\n - Tom\n", "msg_date": "Thu, 19 Mar 1998 03:24:25 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \n> > OK, now I know that the VAX stuff is still used and supported. Good.\n> > We don't have good platform-specific information on NetBSD and Linux\n> > ports.\n> \n> ?? \n> \n> Have you checked the hardcopy or html versions of the \"Supported\n> Platforms\" section in the Administrator's Guide? _Every_ entry in this\n> was updated and refreshed at the end of February.\n> \n> What other info should we be collecting for this?\n> \n> - Tom\n> \n\nSorry, haven't looked there yet.\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Wed, 18 Mar 1998 23:33:49 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \n> > \n> > Hi hackers,\n> > \n> > I have old patches for version 6.2.1p6 which fix some problems and add\n> > new features. Here is a short description of each patch file:\n> > \n> > ...\n> > \n> > -- \n> > Massimo Dal Zotto\n> \n> Massimo, now that 6.3 is released, any chance of getting these patches\n> against the 6.3 source code?\n\nI have already applied all my old patches against 6.3 except the lock\npatch. The lock code has changed and I have to verify if my patches are\nstill compatible. It will take some time and I haven't very much.\nI found also an interesting bug in the notify code. I will post the new\npatches when I find some spare time to verify them.\nCould you please send me some documentation on the new lock and deadlock\ncode in the meantime ?\n\nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n\| Massimo Dal Zotto e-mail: [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": "Thu, 19 Mar 1998 18:25:04 +0100 (MET)", "msg_from": "Massimo Dal Zotto <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> I have already applied all my old patches against 6.3 except the lock\n> patch. The lock code has changed and I have to verify if my patches are\n\nGreat.\n\n> still compatible. It will take some time and I haven't very much.\n> I found also an interesting bug in the notify code. I will post the new\n> patches when I find some spare time to verify them.\n> Could you please send me some documentation on the new lock and deadlock\n> code in the meantime ?\n\nHere is some comments from storage/lmgr/lock.c:DeadLockCheck(). This\nwas a real mind-bender for me. It finds holders of the lock it has been\nwaiting on, and checks to see if any of those holders is waiting on the\nlock I own. If not, it then checks all the holders of locks these new\nprocesses are waiting on, and checks to see if they are waiting on my\nlock, and it continues until it has traced all backend process id's\nrelated to my lock. I have a static checked_procs[] array that keeps\ntrack of what I have checked so I don't loop. The code is recursive.\n\nThe new wait queue handling is documented in storage/lmgr/lock.c:ProcSleep().\n\n---------------------------------------------------------------------------\n\n * This code takes a list of locks a process holds, and the lock that\n * the process is sleeping on, and tries to find if any of the processes\n * waiting on its locks hold the lock it is waiting for. If no deadlock\n * is found, it goes on to look at all the processes waiting on their locks.\n \n We have already locked the master lock before being called.\n /\nbool\nDeadLockCheck(SHM_QUEUE lockQueue, LOCK findlock, bool skip_check)\n\n\n---------------------------------------------------------------------------\n\n /\n * This is our only check to see if we found the lock we want.\n \n The lock we are waiting for is already in MyProc->lockQueue so we\n * need to skip it here. We are trying to find it in someone\n * else's lockQueue.\n /\n if (lock == findlock && !skip_check)\n\n \n---------------------------------------------------------------------------\n\n /\n * For findlock's wait queue, we are interested in\n * procs who are blocked waiting for a write-lock on\n * the table we are waiting on, and already hold a\n * lock on it. We first check to see if there is an\n * escalation deadlock, where we hold a readlock and\n * want a writelock, and someone else holds readlock\n * on the same table, and wants a writelock.\n \n Basically, the test is, \"Do we both hold some lock on\n * findlock, and we are both waiting in the lock\n * queue?\"\n /\n\n---------------------------------------------------------------------------\n\n For non-MyProc entries, we are looking only\n * waiters, not necessarily people who already\n * hold locks and are waiting. Now we check for\n * cases where we have two or more tables in a\n * deadlock. We do this by continuing to search\n * for someone holding a lock\n /\n if (DeadLockCheck(&(proc->lockQueue), findlock, false))\n\n---------------------------------------------------------------------------\n\n /\n * If the first entries in the waitQueue have a greater priority than\n * we have, we must be a reader, and they must be a writers, and we\n * must be here because the current holder is a writer or a reader but\n * we don't share shared locks if a writer is waiting. We put\n * ourselves after the writers. This way, we have a FIFO, but keep\n * the readers together to give them decent priority, and no one\n * starves. Because we group all readers together, a non-empty queue\n * only has a few possible configurations:\n \n [readers] [writers] [readers][writers] [writers][readers]\n * [writers][readers][writers]\n \n In a full queue, we would have a reader holding a lock, then a writer\n * gets the lock, then a bunch of readers, made up of readers who\n * could not share the first readlock because a writer was waiting,\n * and new readers arriving while the writer had the lock.\n \n /\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Thu, 19 Mar 1998 14:47:10 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \n> > \n> > On Mon, 16 Mar 1998, David Gould wrote:\n> > \n> > > If this will work, I am willing to hack this together tomorrow.\n> > > What is the time frame for accepting a patch like this?\n> > \n> > \tAssuming that its clean (clean meaning that Bruce fully approves\n> > of it, as this is his area of the code...well, one of them\n> > grin)...tomorrow would be great :) If Bruce has any doubts though, it\n> > doesn't go in until after I do the patch I want to do...\n> \n> David, go for it. The code is all local in two files, and I think you\n> can basically change all the do{test-and-set} while(lock-is-false)\n> loops to:\n> \n> \tdo{test-and-set} while(lock-is-false && select ())\n> \n> Pretty easy. No need to test multiple platforms. The ones where the\n> loop is integrated into the asm(), leave them for later.\n> \n> -- \n> Bruce Momjian \| 830 Blythe Avenue\n> [email protected] \| Drexel Hill, Pennsylvania 19026\n> + If your life is a hard drive, \| (610) 353-9879(w)\n> + Christ can be your backup. \| (610) 853-3000(h)\n> \n> \n> \n\nI'am against a generic patch using select(). If we have sched_yield() on an\narchitecture I don't see why dont't use it. Here is the patch for Linux.\nIt has been tested for two months by 100 users without any problem.\nThe only thing I would add is a more general configuration test in configure\nto include the proper include files.\n\n* src/include/storage/s_lock.h.orig\tSat Oct 18 22:39:21 1997\n--- src/include/storage/s_lock.h\tWed Nov 19 23:11:14 1997\n***********\n* 294,300 ***\n--- 294,314 ----\n /\n \n #if defined(NEED_I386_TAS_ASM)\n+ #include <unistd.h>\n+ #include <sched.h>\n \n+ #ifdef _POSIX_PRIORITY_SCHEDULING\n+ #define\tS_LOCK(lock)\tdo \\\n+ \t\t\t\t\t\t{ \\\n+ \t\t\t\t\t\t\tslock_t\t\t_res; \\\n+ \t\t\t\t\t\t\tdo \\\n+ \t\t\t\t\t\t\t{ \\\n+ \t\t\t\t\t\t\t\t__asm__(\"xchgb %0,%1\": \"=q\"(_res), \\\n+ \t\t\t\t\t\t\t\t\t\t\"=m\"(lock):\"0\"(0x1)); \\\n+ \t\t\t\t\t\t\t\tif (_res) sched_yield(); \\\n+ \t\t\t\t\t\t\t} while (_res != 0); \\\n+ \t\t\t\t\t\t} while (0)\n+ #else\n #define\tS_LOCK(lock)\tdo \\\n \t\t\t\t\t\t{ \\\n \t\t\t\t\t\t\tslock_t\t\t_res; \\\n************\n* 303,308 ***\n--- 317,323 ----\n \t\t\t\t__asm__(\"xchgb %0,%1\": \"=q\"(_res), \"=m\"(lock):\"0\"(0x1)); \\\n \t\t\t\t\t\t\t} while (_res != 0); \\\n \t\t\t\t\t\t} while (0)\n+ #endif\n \n #define\tS_UNLOCK(lock)\t((lock) = 0)\n \n\nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n\| Massimo Dal Zotto e-mail: [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": "Fri, 20 Mar 1998 11:59:01 +0100 (MET)", "msg_from": "Massimo Dal Zotto <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> I'am against a generic patch using select(). If we have sched_yield() on an\n> architecture I don't see why dont't use it. Here is the patch for Linux.\n> It has been tested for two months by 100 users without any problem.\n> The only thing I would add is a more general configuration test in configure\n> to include the proper include files.\n\nI understand your issue. Unfortunately, only Linux has sched_yield(),\nas far as I know. Perhaps we can implement sched_yield/select based on\nthe platform. However, if someone is holding a spinlock, does\nsched_yield() give the other process enough time to finish with the\nspinlock before we start checking it again. Seems select() allows us to\ncontrol the time we wait before checking again.\n\nAlso, it looks like the s_lock.h file is going to change pretty\nradically from David's change, so when he is done, we can put some\nOS-specific stuff if you wish.\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Fri, 20 Mar 1998 12:06:55 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "Massimo Dal Zotto writes:\n> > David, go for it. The code is all local in two files, and I think you\n> > can basically change all the do{test-and-set} while(lock-is-false)\n> > loops to:\n> > \n> > \tdo{test-and-set} while(lock-is-false && select ())\n> > \n> > Pretty easy. No need to test multiple platforms. The ones where the\n> > loop is integrated into the asm(), leave them for later.\n> > \n> > -- \n> > Bruce Momjian \| 830 Blythe Avenue\n> \n> I'am against a generic patch using select(). If we have sched_yield() on an\n> architecture I don't see why dont't use it. Here is the patch for Linux.\n> It has been tested for two months by 100 users without any problem.\n> The only thing I would add is a more general configuration test in configure\n> to include the proper include files.\n> \n> src/include/storage/s_lock.h.orig\tSat Oct 18 22:39:21 1997\n> --- src/include/storage/s_lock.h\tWed Nov 19 23:11:14 1997\n> ***********\n> * 294,300 ***\n> --- 294,314 ----\n> /\n> \n> #if defined(NEED_I386_TAS_ASM)\n> + #include <unistd.h>\n> + #include <sched.h>\n> \n> + #ifdef _POSIX_PRIORITY_SCHEDULING\n> + #define\tS_LOCK(lock)\tdo \\\n> + \t\t\t\t\t{ \\\n> + \t\t\t\t\t\tslock_t\t\t_res; \\\n> + \t\t\t\t\t\tdo \\\n> + \t\t\t\t\t\t{ \\\n> + \t\t\t\t\t\t\t__asm__(\"xchgb %0,%1\": \"=q\"(_res), \\\n> + \t\t\t\t\t\t\t\t\t\"=m\"(lock):\"0\"(0x1)); \\\n> + \t\t\t\t\t\t\tif (_res) sched_yield(); \\\n> + \t\t\t\t\t\t} while (_res != 0); \\\n> + \t\t\t\t\t} while (0)\n> + #else\n> #define\tS_LOCK(lock)\tdo \\\n> \t\t\t\t\t{ \\\n> \t\t\t\t\t\tslock_t\t\t_res; \\\n> ************\n> * 303,308 ***\n> --- 317,323 ----\n> \t\t\t__asm__(\"xchgb %0,%1\": \"=q\"(_res), \"=m\"(lock):\"0\"(0x1)); \\\n> \t\t\t\t\t\t} while (_res != 0); \\\n> \t\t\t\t\t} while (0)\n> + #endif\n> \n> #define\tS_UNLOCK(lock)\t((lock) = 0)\n> \n> +----------------------------------------------------------------------+\n> \| Massimo Dal Zotto e-mail: [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\n\n\nI am perfectly happy to use sched_yield() on Linux. My goal however is to\nmake the concept work on all platforms.\n\nThere was a recent thread on comp.os.linux.system on context switch times\nof Linux vs NTthat revealed that sched_yield() is fairly costly as it causes\nthe scheduler to do a full scan of the process table and recalculate the\nthe priorities of all processes. Probably not a problem, but it should\nbe it should probably be benchmarked both ways.\n\nFinally, even though this appears to work, there is a possible stability\nproblem with both approaches. Here is some of the discussion I had about\nthat with Bruce:\n\n-----------------------------------------------------------------------\nDavid:\n> Right now, there is not much chance of catching a signal while waiting for\n> a spinlock. This is good, cause the code that waits for spinlocks tends to\n> be doing funny things with buffers and locks and shared stuff like that.\n> We don't catch signals because we don't make syscalls. But, once this goes in,\n> we will be calling select() a _lot_ and so it kinda becomes the likely place\n> for a signal to get delivered. Without thinking about it and looking at the\n> code a bit longer, I am not sure it is prudent to rush this in. I still want\n> it in as soon a possible, but possible includes free from harmful side effects.\nBruce:\n> Well, signals are handled in the backend by tcop/postgres.c. In\n> most/all? cases, a signal causes a longjump() out of the code and\n> releases locks and aborts the transaction.\n\nDavid:\n> I was afraid that would be the answer. Basically, this never worked. The\n> problem is that in an indeterminate number of places the code manipulates\n> multiple distinct but related structures in a sequence. To leave the server\n> in a consistant state all these updates need to be done in the right order\n> so that if the sequence in interrupted by a signal the partially updated\n> state is consistant. Unhappily, the original coders did not always have this\n> in mind. An example:\n> \n> cleaning up after a scan\n> 1 - release buffer pointed by scan descriptor\n> 2 - release scan descriptor\n> \n> If a signal is taken just after one, the abort code will see the scan\n> descriptor and call the cleanup for it resulting in:\n> \n> cleaning up after a scan (take 2)\n> 1 - release buffer pointed by scan descriptor\n> - Whoopsie, buffer already released!\n> 2 - release scan descriptor\n> \n> These sequences either _all_ have to identified and fixed, or made atomic\n> somehow, which is a biggish job.\n> \n> Or the system has to acknowledge signals at only clearly defined points.\n> My preference is to have signal handlers only set a global flag to indicate\n> that a signal was seen. Then one just sprinkles tests of the flag in\n> all the likely places: step to next node, fetch next page, call function, etc.\n> \n> The way this shows up in real life is strange unreproduceable errors on busy\n> systems, especially when backends are killed or transactions are forced to\n> abort.\n> \n> Fixing this is a bit of a project, but the result is that a lot of mystery\n> bugs go away.\n\n--------------------------------------------------------------------------\n\n-dg\n\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - I realize now that irony has no place in business communications.\n\n", "msg_date": "Fri, 20 Mar 1998 11:15:47 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "On Fri, 20 Mar 1998, Bruce Momjian wrote:\n\n> > I'am against a generic patch using select(). If we have sched_yield() on an\n> > architecture I don't see why dont't use it. Here is the patch for Linux.\n> > It has been tested for two months by 100 users without any problem.\n> > The only thing I would add is a more general configuration test in configure\n> > to include the proper include files.\n> \n> I understand your issue. Unfortunately, only Linux has sched_yield(),\n> as far as I know. Perhaps we can implement sched_yield/select based on\n> the platform. \n\n\tWhat's the possibility of doing this similar to how we do some of\nthe other functions (dl_open comes immediately to mind)...make a\npg_sched_yield function and use that, which is built based on the various\nplatforms?\n\n\tRight now, I don't believe we have anything* in place, so have\npg_sched_yield() return 0 (or an equivalent) for every platform except for\nLinux...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 20 Mar 1998 19:27:52 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \n> On Fri, 20 Mar 1998, Bruce Momjian wrote:\n> \n> > > I'am against a generic patch using select(). If we have sched_yield() on an\n> > > architecture I don't see why dont't use it. Here is the patch for Linux.\n> > > It has been tested for two months by 100 users without any problem.\n> > > The only thing I would add is a more general configuration test in configure\n> > > to include the proper include files.\n> > \n> > I understand your issue. Unfortunately, only Linux has sched_yield(),\n> > as far as I know. Perhaps we can implement sched_yield/select based on\n> > the platform. \n> \n> \tWhat's the possibility of doing this similar to how we do some of\n> the other functions (dl_open comes immediately to mind)...make a\n> pg_sched_yield function and use that, which is built based on the various\n> platforms?\n> \n> \tRight now, I don't believe we have anything in place, so have\n> pg_sched_yield() return 0 (or an equivalent) for every platform except for\n> Linux...\n> \n> Marc G. Fournier \n> Systems Administrator @ hub.org \n> primary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \nI appreciate all the help, but I think I have a solution for this. Details\nnext week...\n\n-dg\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - I realize now that irony has no place in business communications.\n\n", "msg_date": "Fri, 20 Mar 1998 17:05:32 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> \tWhat's the possibility of doing this similar to how we do some of\n> the other functions (dl_open comes immediately to mind)...make a\n> pg_sched_yield function and use that, which is built based on the various\n> platforms?\n> \n> \tRight now, I don't believe we have anything in place, so have\n> pg_sched_yield() return 0 (or an equivalent) for every platform except for\n> Linux...\n> \n\nProbably even easier. Just use #ifdef linux around the select or\nsched_yield.\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Fri, 20 Mar 1998 22:40:54 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "On Sun, 22 Mar 1998, Mattias Kregert wrote:\n\n> The Hermit Hacker wrote:\n> >\n> > What's the possibility of doing this similar to how we do some of\n> > the other functions (dl_open comes immediately to mind)...make a\n> > pg_sched_yield function and use that, which is built based on the various\n> > platforms?\n> > \n> > Right now, I don't believe we have anything in place, so have\n> > pg_sched_yield() return 0 (or an equivalent) for every platform except for\n> > Linux...\n> \n> But sched_yield() is not Linux-specific:\n> -- The sched_yield() function relinquishes the processor for the\n> -- running process.\n> -- IEEE Std 1003.1b-1993, ��13.3.5. (POSIX real-time standard 1003.lb)\n> \n> Except from Linux, I can find references to sched_yield() in LynxOS,\n> DECthreads thread library, AIX 4.1 and up (libc), Solaris (thread.h\n> (c)1994 Sun\n> Microsystems), Unix98, GNU, C EXECUTIVE(r) and PSX(tm) real time kernels\n> ...\n> This is just a quick search.\n> \n> Perhaps we should enable sched_yield() for every OS except for... well,\n> what's the\n> name of that OS which does not have sched_yield()... FreeBSD ;)\n> \n> After all, sched_yield() is five years old. Any reasonable OS should\n> have it.\n\n\tYou missed my point...so far as I've heard, there are two ways of\ndoing what is being proposed...either using sched_yield() on those systems\nthat support it, or select() on those that don't. If you are going to\nbuild a patch for this, it should look something like:\n\n#ifdef HAVE_SCHED_YIELD\n\t<insert sched_yield() code here>\n#else\n\t<insert select() code here>\n#endif\n\n\tTotally 'configure' configurable, and non-system dependent :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sat, 21 Mar 1998 21:33:16 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: sched_yield()" }, { "msg_contents": "The Hermit Hacker wrote:\n>\n> What's the possibility of doing this similar to how we do some of\n> the other functions (dl_open comes immediately to mind)...make a\n> pg_sched_yield function and use that, which is built based on the various\n> platforms?\n> \n> Right now, I don't believe we have anything in place, so have\n> pg_sched_yield() return 0 (or an equivalent) for every platform except for\n> Linux...\n\nBut sched_yield() is not Linux-specific:\n-- The sched_yield() function relinquishes the processor for the\n-- running process.\n-- IEEE Std 1003.1b-1993, ï¿½13.3.5. (POSIX real-time standard 1003.lb)\n\nExcept from Linux, I can find references to sched_yield() in LynxOS,\nDECthreads thread library, AIX 4.1 and up (libc), Solaris (thread.h\n(c)1994 Sun\nMicrosystems), Unix98, GNU, C EXECUTIVE(r) and PSX(tm) real time kernels\n...\nThis is just a quick search.\n\nPerhaps we should enable sched_yield() for every OS except for... well,\nwhat's the\nname of that OS which does not have sched_yield()... FreeBSD ;)\n\nAfter all, sched_yield() is five years old. Any reasonable OS should\nhave it.\n\n/* m /\n", "msg_date": "Sun, 22 Mar 1998 02:50:12 +0100", "msg_from": "Mattias Kregert <[email protected]>", "msg_from_op": false, "msg_subject": "Re: sched_yield()" }, { "msg_contents": "> On Sun, 22 Mar 1998, Mattias Kregert wrote:\n> \n> > The Hermit Hacker wrote:\n> > >\n> > > What's the possibility of doing this similar to how we do some of\n> > > the other functions (dl_open comes immediately to mind)...make a\n> > > pg_sched_yield function and use that, which is built based on the various\n> > > platforms?\n> > > \n> > > Right now, I don't believe we have anything* in place, so have\n> > > pg_sched_yield() return 0 (or an equivalent) for every platform except for\n> > > Linux...\n> > \n> > But sched_yield() is not Linux-specific:\n> > -- The sched_yield() function relinquishes the processor for the\n> > -- running process.\n> > -- IEEE Std 1003.1b-1993, ��13.3.5. (POSIX real-time standard 1003.lb)\n> > \n> > Except from Linux, I can find references to sched_yield() in LynxOS,\n> > DECthreads thread library, AIX 4.1 and up (libc), Solaris (thread.h\n> > (c)1994 Sun\n> > Microsystems), Unix98, GNU, C EXECUTIVE(r) and PSX(tm) real time kernels\n> > ...\n> > This is just a quick search.\n> > \n> > Perhaps we should enable sched_yield() for every OS except for... well,\n> > what's the\n> > name of that OS which does not have sched_yield()... FreeBSD ;)\n> > \n> > After all, sched_yield() is five years old. Any reasonable OS should\n> > have it.\n> \n> \tYou missed my point...so far as I've heard, there are two ways of\n> doing what is being proposed...either using sched_yield() on those systems\n> that support it, or select() on those that don't. If you are going to\n> build a patch for this, it should look something like:\n> \n> #ifdef HAVE_SCHED_YIELD\n> \t<insert sched_yield() code here>\n> #else\n> \t<insert select() code here>\n> #endif\n> \n> \tTotally 'configure' configurable, and non-system dependent :)\n> \n> Marc G. Fournier \n> Systems Administrator @ hub.org \n> primary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \nOk, I will add a configuration option\n\n USE_SCHED_YIELD\n\nto the select patch I am working on. This can be enabled by configure.\nAssuming someone can find the header files needed on all the platforms.\n\nHowever, we should not assume that sched_yield() even where available is\nthe automatic \"right thing\". It might be, but...\n\nThe situation that either the select() or the sched_yield() style of\nspinlock back off is meant to help is when there are a number of processes\nbusywaiting on the same spinlock.\n\nOn Linux, sched_yield() triggers the scheduler to to a full priority re-calc\nfor all processses. This is slightly expensive especially with a long run\nqueue. Having a bunch of processes pound on sched_yield() might be actually\nworse than to use select(). At the very least it needs testing.\n\nSecondly, the select() backoff patch I am working on starts out with a zero\ntimeout and backs off incrementally by increasing the timeout value on\nsubsequent iterations. The idea is to break up convoys and avoid big piles of\nprocesses pounding on a spinlock. This cannot be done with sched_yield().\n\nWhich is better? Well, golly gosh, I have no idea. I know that the select()\nflavor effectively solves the problem caused by hard loop busy waiting.\nWithout some testing it is kinda hard to say more than that. I will try to\nfit in some testing, but if someone has a favorite many process workload\nand would like to try comparing both flavors it would be useful.\n\n-dg\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - I realize now that irony has no place in business communications.\n\n", "msg_date": "Sat, 21 Mar 1998 19:02:56 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: sched_yield()" }, { "msg_contents": "> \n> \tYou missed my point...so far as I've heard, there are two ways of\n> doing what is being proposed...either using sched_yield() on those systems\n> that support it, or select() on those that don't. If you are going to\n> build a patch for this, it should look something like:\n> \n> #ifdef HAVE_SCHED_YIELD\n> \t<insert sched_yield() code here>\n> #else\n> \t<insert select() code here>\n> #endif\n> \n> \tTotally 'configure' configurable, and non-system dependent :)\n\nYep, I like it.\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Sat, 21 Mar 1998 22:59:03 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: sched_yield()" }, { "msg_contents": "On Sat, 21 Mar 1998, David Gould wrote:\n\n> Ok, I will add a configuration option\n> \n> USE_SCHED_YIELD\n> \n> to the select patch I am working on. This can be enabled by configure.\n> Assuming someone can find the header files needed on all the platforms.\n> \n> However, we should not assume that sched_yield() even where available is\n> the automatic \"right thing\". It might be, but...\n> \n> The situation that either the select() or the sched_yield() style of\n> spinlock back off is meant to help is when there are a number of processes\n> busywaiting on the same spinlock.\n> \n> On Linux, sched_yield() triggers the scheduler to to a full priority re-calc\n> for all processses. This is slightly expensive especially with a long run\n> queue. Having a bunch of processes pound on sched_yield() might be actually\n> worse than to use select(). At the very least it needs testing.\n> \n> Secondly, the select() backoff patch I am working on starts out with a zero\n> timeout and backs off incrementally by increasing the timeout value on\n> subsequent iterations. The idea is to break up convoys and avoid big piles of\n> processes pounding on a spinlock. This cannot be done with sched_yield().\n> \n> Which is better? Well, golly gosh, I have no idea. I know that the select()\n> flavor effectively solves the problem caused by hard loop busy waiting.\n> Without some testing it is kinda hard to say more than that. I will try to\n> fit in some testing, but if someone has a favorite many process workload\n> and would like to try comparing both flavors it would be useful.\n\n\tOkay, we have two differing viewpoints here...from what I've been\nable to read, the select() solution will work on all platforms, while\nthe sched_yield() will work on some systems, but not all.\n\n\tI personally like the \"work on all platform\" solution, but that's\njust me :)\n\n\tI may have missed it, but I'm curious as to under what\ncircumstance sched_yield() is better then the select() solution? The\n\"con\", as I see it, is that sched_yield() doesn't work everywhere...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sun, 22 Mar 1998 00:00:01 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: sched_yield()" }, { "msg_contents": "> Perhaps we should enable sched_yield() for every OS except for... well,\n> what's the\n> name of that OS which does not have sched_yield()... FreeBSD ;)\n> \n> After all, sched_yield() is five years old. Any reasonable OS should\n> have it.\n\nGee, I just checked and BSDI has it. However, it appears to work only\non threaded applications:\n\n #include <pthread.h>\n...\n If other threads are ready to run, the sched_yield() function forces the\n current thread to suspend itself temporarily and let them execute.\n\n\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Sat, 21 Mar 1998 23:01:12 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: sched_yield()" }, { "msg_contents": "> Secondly, the select() backoff patch I am working on starts out with a zero\n> timeout and backs off incrementally by increasing the timeout value on\n> subsequent iterations. The idea is to break up convoys and avoid big piles of\n> processes pounding on a spinlock. This cannot be done with sched_yield().\n\nHard to beat the backoff argument. I vote we only use select().\n\n-- \nBruce Momjian \| 830 Blythe Avenue\[email protected] \| Drexel Hill, Pennsylvania 19026\n + If your life is a hard drive, \| (610) 353-9879(w)\n + Christ can be your backup. \| (610) 853-3000(h)\n", "msg_date": "Sat, 21 Mar 1998 23:03:41 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: sched_yield()" }, { "msg_contents": "On Sat, 21 Mar 1998, Bruce Momjian wrote:\n\n> > Secondly, the select() backoff patch I am working on starts out with a zero\n> > timeout and backs off incrementally by increasing the timeout value on\n> > subsequent iterations. The idea is to break up convoys and avoid big piles of\n> > processes pounding on a spinlock. This cannot be done with sched_yield().\n> \n> Hard to beat the backoff argument. I vote we only use select().\n\n\tI haven't heard any compelling arguments so far as to why\nsched_yield() is better then select(), so I tend to vote the same way...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sun, 22 Mar 1998 01:44:50 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: sched_yield()" }, { "msg_contents": "\nOn Sun, 22 Mar 1998, Mattias Kregert wrote:\n\n> Except from Linux, I can find references to sched_yield() in LynxOS,\n> DECthreads thread library, AIX 4.1 and up (libc), Solaris (thread.h\n> (c)1994 Sun\n> Microsystems), Unix98, GNU, C EXECUTIVE(r) and PSX(tm) real time kernels\n> ...\n> This is just a quick search.\n\n This seems to be part of Posix threads, which means that sched_yield\nshould only be callable by a thread.... I'm surprised that normal\nprocesses can call sched_yield. In fact, in most of the environments\nyou've mentioned, it certainly can't be used outside a thread.\n\nTom\n\n", "msg_date": "Sat, 21 Mar 1998 22:32:15 -0800 (PST)", "msg_from": "Tom <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: sched_yield()" }, { "msg_contents": "David Gould wrote:\n\n> Seriously, if you want to, please create a function to emulate the following:\n> \n> /\n> tas(lock)\n> \n> Access to platform specific test_and_set functionality. Given pointer to\n> * lock attempts to acquire the lock atomically.\n> \n> Returns 0 for success, nonzero for failure.\n> /\n> typedef slock_t unsigned char;\t\t/ or whatever works on the platform /\n> \n> int tas(slock_t lock)\n> {\n> slock_t\ttmp;\n> \n> /* atomic, interlocked /\n> tmp = lock;\n> lock = -1; \t\t\t/ any nonzero will do here /\n> \n> return (tmp != 0);\n> }\n> \n> Given this, I can fold the VAX right into the grand scheme, just like a\n> normal computer (;-)).\n\nHmpf! The true worth of a computer is a function of its weight! :-)\n\nSorry this took a while, but anyway, this should do it for the VAX (in\nfact, it's more or less the version of the code that I figured I'd use\nuntil Bruce asked me to bum it down maximally for performance, only\nnow with the return values from tas() swapped). I include the macros\nthat would fit the current (6.3) locking scheme:\n\ntypedef unsigned char slock_t;\n\nint tas(slock_t lock) {\n\tregister ret;\n\nasm(\"\tmovl $1, r0\n\tbbssi $0,(%1),1f\n\tclrl r0\n1:\tmovl r0,%0\"\n\t: \"=r\"(ret)\t/* return value, in register /\n\t: \"r\"(lock)\t/ argument, 'lock pointer', in register /\n\t: \"r0\");\t/ inline code uses this register /\n\n\treturn ret;\n}\n\n#define\tS_LOCK(addr)\t\tdo { while (tas(addr)) ; } while (0)\n#define\tS_UNLOCK(addr)\t\t((addr) = 0)\n#define\tS_INIT_LOCK(addr)\t((addr) = 0)\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "31 Mar 1998 23:20:40 +0200", "msg_from": "Tom Ivar Helbekkmo <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "> David Gould wrote:\n> \n> >Seriously, if you want to, please create a function to emulate the following:\n> > \n> > /\n> > * tas(lock)\n> > \n> > Access to platform specific test_and_set functionality. Given pointer to\n> > * lock attempts to acquire the lock atomically.\n> > \n> > Returns 0 for success, nonzero for failure.\n> > /\n> > typedef slock_t unsigned char;\t\t/ or whatever works on the platform /\n> > \n> > int tas(slock_t lock)\n> > {\n> > slock_t\ttmp;\n> > \n> > /* atomic, interlocked /\n> > tmp = lock;\n> > lock = -1; \t\t\t/ any nonzero will do here /\n> > \n> > return (tmp != 0);\n> > }\n> > \n> > Given this, I can fold the VAX right into the grand scheme, just like a\n> > normal computer (;-)).\n> \n> Hmpf! The true worth of a computer is a function of its weight! :-)\n> \n> Sorry this took a while, but anyway, this should do it for the VAX (in\n> fact, it's more or less the version of the code that I figured I'd use\n> until Bruce asked me to bum it down maximally for performance, only\n> now with the return values from tas() swapped). I include the macros\n\nWhat do you mean \"now with the return values from tas() swapped\"? I think\nyour code looks ok, but just want to be sure we are following the same\ngrand plan...\n\n> that would fit the current (6.3) locking scheme:\n> \n> typedef unsigned char slock_t;\n> \n> int tas(slock_t lock) {\n> \tregister ret;\n> \n> asm(\"\tmovl $1, r0\n> \tbbssi $0,(%1),1f\n> \tclrl r0\n> 1:\tmovl r0,%0\"\n> \t: \"=r\"(ret)\t/* return value, in register /\n> \t: \"r\"(lock)\t/ argument, 'lock pointer', in register /\n> \t: \"r0\");\t/ inline code uses this register /\n> \n> \treturn ret;\n> }\n> \n> #define\tS_LOCK(addr)\t\tdo { while (tas(addr)) ; } while (0)\n> #define\tS_UNLOCK(addr)\t\t((addr) = 0)\n> #define\tS_INIT_LOCK(addr)\t((addr) = 0)\n> \n> -tih\n> -- \n\nThanks, this is just what I was looking for. I will fold it in to my changes.\nI have gotten a little snowed under with other tasks, but I expect to finalize\nmy patch next week and will post it.\n\n-dg\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - Linux. Not because it is free. Because it is better.\n\n", "msg_date": "Wed, 1 Apr 1998 11:56:56 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": " David Gould\n\|\n\| What do you mean \"now with the return values from tas() swapped\"? I think\n\| your code looks ok, but just want to be sure we are following the same\n\| grand plan...\n\nI just meant that my original code (which has been posted before) had\nthe tas() function implemented so that it returned 0 on failure, not\non success, as you asked for. Thus, I had to swap the sense of the\nreturn value. In practice, I changed\n\n\tclrl r0\t\t\t; clear register r0\n\tbbssi $0,(%1),1f\t; branch on bit set else set\n\tincl r0\t\t\t; increment register r0\n1:\tmovl r0,%0\t\t; return register r0\n\n\t[...]\n\n#define\tS_LOCK(addr)\t\tdo { while (!tas(addr)) ; } while (0)\n\n...into...\n\n\tmovl $1, r0\t\t; set register r0 to 1\n\tbbssi $0,(%1),1f\t; branch on bit set else set\n\tclrl r0\t\t\t; clear register r0\n1:\tmovl r0,%0\t\t; return register r0\n\n\t[...]\n\n#define\tS_LOCK(addr)\t\tdo { while (tas(addr)) ; } while (0)\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "02 Apr 1998 06:52:33 +0200", "msg_from": "Tom Ivar Helbekkmo <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" }, { "msg_contents": "Tom Ivar Helbekkmo:\n> * David Gould\n> \|\n> \| What do you mean \"now with the return values from tas() swapped\"? I think\n> \| your code looks ok, but just want to be sure we are following the same\n> \| grand plan...\n> \n> I just meant that my original code (which has been posted before) had\n> the tas() function implemented so that it returned 0 on failure, not\n> on success, as you asked for. Thus, I had to swap the sense of the\n> return value. In practice, I changed\n> \n> \tclrl r0\t\t\t; clear register r0\n> \tbbssi $0,(%1),1f\t; branch on bit set else set\n> \tincl r0\t\t\t; increment register r0\n> 1:\tmovl r0,%0\t\t; return register r0\n> \n> \t[...]\n> \n> #define\tS_LOCK(addr)\t\tdo { while (!tas(addr)) ; } while (0)\n> \n> ...into...\n> \n> \tmovl $1, r0\t\t; set register r0 to 1\n> \tbbssi $0,(%1),1f\t; branch on bit set else set\n> \tclrl r0\t\t\t; clear register r0\n> 1:\tmovl r0,%0\t\t; return register r0\n> \n> \t[...]\n> \n> #define\tS_LOCK(addr)\t\tdo { while (tas(addr)) ; } while (0)\n> \n\nThats what I thought, but it has been a few years (not saying how many ;-) )\nsince I wrote any Macro-32 so I figured I should check.\n\nThe tas() definition is not to return success or failure so much as to\nreturn the _previous_state_ of the lock. So if you test_and_set the lock\nthe test part returns true if it was previously locked and false if it was\nunlocked. In either case, it is locked after the tas() (the set part). \nOnly, if it was previously unlocked, someone else owns the lock so we\nhave to wait for them to unlock it.\n\n-dg\n\nDavid Gould [email protected] 510.628.3783 or 510.305.9468 \nInformix Software (No, really) 300 Lakeside Drive Oakland, CA 94612\n - Linux. Not because it is free. Because it is better.\n\n\n", "msg_date": "Wed, 1 Apr 1998 23:06:36 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6" } ]
[ { "msg_contents": "Bruce Momjian wrote:\n> \n> > > We have a new atttypmod field that holds the length for the column as\n> > > defined at table-creation time for char() and varchar(). Is that the\n> > > number you want?\n> > >\n> > > --\n> > > Bruce Momjian\n> > > [email protected]\n> >\n> >\n> > for varchar() and also for the datatype text it would be\n> > nice to know the actual length. If this is not possible,\n> > then the length at creation time is still better than -1.\n> \n> If you use binary cursors, you can use VARSIZE and VARDATA to see the\n> size and data for a string type.\n> \n> --\n> Bruce Momjian\n> [email protected]\n\nFor the perl interface (DBD-Pg) I can not use cursors at all,\nbecause they can not be nested.\n\nEdmund\n-- \nEdmund Mergl mailto:[email protected]\nIm Haldenhau 9 http://www.bawue.de/~mergl\n70565 Stuttgart fon: +49 711 747503\nGermany gsm: +49 171 2645325\n", "msg_date": "Sun, 15 Feb 1998 07:46:35 +0100", "msg_from": "Edmund Mergl <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] wishlist: PQfsize" }, { "msg_contents": "Hi all,\n\nI would like to add the following item to the \nwishlist:\n\ncurrently the libpq-function PQfsize returns the \nsize in bytes of the field with the given field-\nindex. If the field is a variable-length field, \nPQfsize returns -1.\n\nIn general as well as for some internal functionality\nof the perl interface it would be useful, if PQfsize\nwould return the number of bytes even for variable-\nlength fields.\n\nEdmund\n-- \nEdmund Mergl mailto:[email protected]\nIm Haldenhau 9 http://www.bawue.de/~mergl\n70565 Stuttgart fon: +49 711 747503\nGermany gsm: +49 171 2645325\n", "msg_date": "Sun, 15 Feb 1998 09:34:05 +0100", "msg_from": "Edmund Mergl <[email protected]>", "msg_from_op": true, "msg_subject": "wishlist: PQfsize" }, { "msg_contents": "> \n> Hi all,\n> \n> I would like to add the following item to the \n> wishlist:\n> \n> currently the libpq-function PQfsize returns the \n> size in bytes of the field with the given field-\n> index. If the field is a variable-length field, \n> PQfsize returns -1.\n> \n> In general as well as for some internal functionality\n> of the perl interface it would be useful, if PQfsize\n> would return the number of bytes even for variable-\n> length fields.\n\nWe have a new atttypmod field that holds the length for the column as\ndefined at table-creation time for char() and varchar(). Is that the\nnumber you want?\n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sun, 15 Feb 1998 12:54:48 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] wishlist: PQfsize" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> >\n> > Hi all,\n> >\n> > I would like to add the following item to the\n> > wishlist:\n> >\n> > currently the libpq-function PQfsize returns the\n> > size in bytes of the field with the given field-\n> > index. If the field is a variable-length field,\n> > PQfsize returns -1.\n> >\n> > In general as well as for some internal functionality\n> > of the perl interface it would be useful, if PQfsize\n> > would return the number of bytes even for variable-\n> > length fields.\n> \n> We have a new atttypmod field that holds the length for the column as\n> defined at table-creation time for char() and varchar(). Is that the\n> number you want?\n> \n> --\n> Bruce Momjian\n> [email protected]\n\n\nfor varchar() and also for the datatype text it would be\nnice to know the actual length. If this is not possible,\nthen the length at creation time is still better than -1.\n\nEdmund\n-- \nEdmund Mergl mailto:[email protected]\nIm Haldenhau 9 http://www.bawue.de/~mergl\n70565 Stuttgart fon: +49 711 747503\nGermany gsm: +49 171 2645325\n", "msg_date": "Sun, 15 Feb 1998 20:56:17 +0100", "msg_from": "Edmund Mergl <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] wishlist: PQfsize" }, { "msg_contents": "> > We have a new atttypmod field that holds the length for the column as\n> > defined at table-creation time for char() and varchar(). Is that the\n> > number you want?\n> > \n> > --\n> > Bruce Momjian\n> > [email protected]\n> \n> \n> for varchar() and also for the datatype text it would be\n> nice to know the actual length. If this is not possible,\n> then the length at creation time is still better than -1.\n\nIf you use binary cursors, you can use VARSIZE and VARDATA to see the\nsize and data for a string type.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Sun, 15 Feb 1998 18:26:26 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] wishlist: PQfsize" } ]
[ { "msg_contents": " Do we have a way, at the moment, supported or otherwise, to \"share\"\ntables across multiple databases? For instance, some of the catalogs\nare the same from one database to another, but could I create a new\ntable and share it (and possibly some indexes on it) between more than\none database?\n Thanks in advance...\n\n-Brandon :)\n", "msg_date": "Sun, 15 Feb 1998 04:01:23 -0600 (CST)", "msg_from": "Brandon Ibach <[email protected]>", "msg_from_op": true, "msg_subject": "Shared tables" }, { "msg_contents": "> Do we have a way, at the moment, supported or otherwise, to \"share\"\n> tables across multiple databases? For instance, some of the catalogs\n> are the same from one database to another, but could I create a new\n> table and share it (and possibly some indexes on it) between more than\n> one database?\n\nNo, that is not directly supported by the backend. You might be able to\ndo some trickery using soft links...\n\nAlso, I've been thinking about what it would take to allow the backend to\nconnect to another backend as a client. 'Course, I don't know much about\nit, but it would be neat to be able to do...\n\n - Tom\n\n", "msg_date": "Sun, 15 Feb 1998 15:02:22 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Shared tables" }, { "msg_contents": "Thomas G. Lockhart said:\n> \n> > Do we have a way, at the moment, supported or otherwise, to \"share\"\n> > tables across multiple databases? For instance, some of the catalogs\n> > are the same from one database to another, but could I create a new\n> > table and share it (and possibly some indexes on it) between more than\n> > one database?\n> \n> No, that is not directly supported by the backend. You might be able to\n> do some trickery using soft links...\n> \n> Also, I've been thinking about what it would take to allow the backend to\n> connect to another backend as a client. 'Course, I don't know much about\n> it, but it would be neat to be able to do...\n> \n Sounds interesting... :) As for table sharing, I could do hard or\nsoft links, but the question is, would that create the potential for\nPostgres corrupting things, given that it isn't wise to the situation,\nand could therefore defeat its own locking mechanisms? Of course,\nthis would probably just be a read-only table (just some data I'd like\nto have access to from multiple databases), so I could probably get\naway with just creating the links. Anybody have a reason against\nthis?\n\n-Brandon :)\n", "msg_date": "Tue, 17 Feb 1998 04:02:05 -0600 (CST)", "msg_from": "Brandon Ibach <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Shared tables" } ]
[ { "msg_contents": "I just tried this on Debian with glibc 2.0.6 and it works fine. That is\nmy 6.3 postgresql release outputs 4 hours.\n\nMichael\n\n--\nDr. Michael Meskes, Projekt-Manager \| topystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Use Debian GNU/Linux! \| Tel: (+49) 2405/4670-44\n\n> ----------\n> From: \tThomas G. Lockhart[SMTP:[email protected]]\n> Sent: \tFreitag, 13. Februar 1998 17:57\n> To: \[email protected]; Bruce Momjian\n> Cc: \[email protected]\n> Subject: \t[HACKERS] Re: [PORTS] v6.3 release ToDo list and\n> supported ports\n> \n> > Is the ix86/linux \"confirmed\" port using libc 5 or glibc 2.x (aka\n> > libc6)? As most major distributions are going to release all new\n> > versions with glibc I think it's prudent to test on both \"platforms\"\n> > independently.\n> \n> Yes, that is a good point. There is a known bug in the glibc2 math\n> library which breaks the date/time routines:\n> \n> select '4 hours'::timespan;\n> ---------\n> @ 4 hours\n> (1 row)\n> \n> comes out instead as\n> tgl=> select '4 hours'::timespan;\n> ----------------------------\n> @ 3 hours 59 mins 60.00 secs\n> (1 row)\n> \n> Oliver was working on patches. Lost his e-mail message; is Oliver\n> still\n> here? I've got a RH5.0 linux system at work now, and the shipped\n> Postgres\n> installation has this problem. I'd like to see it fixed...\n> \n> Bruce, can you add this to the v6.3 ToDo (assuming you agree to take\n> it\n> on :)\n> \n> - Tom\n> \n> \n> \n", "msg_date": "Sun, 15 Feb 1998 12:01:00 +0100", "msg_from": "\"Meskes, Michael\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported po rts" }, { "msg_contents": "I have a \"custom\" slackware with at least the following updates:\n\ngcc-2.8.0\nglibc-2.0.6\n\nAs per the glibc upgrade notes out there on the web, I've cleared \n(actually moved) the /usr/include header files before installing \nglibc, and have reinstalled some \"standard\" extra libraries (such as \nncurses) after that.\n\nI get the 3 hours 59 minutes 60.00 seconds.\n\nI also get a bunch of other problems in other tests. Some are just \ndifferences in the error message, but others are actual differences \nin the float/double numbers produced. Others don't show an error and \ngive a number where one \"should\" have been produced. This is all \nvauge now, but I will recompile the latest snapshot and give a \ncomplete update maybe tonight, if not tomorrow.\n\nCould some of the other problems be due to gcc-2.8.0, which I'm not \nsure is being used by anyone else for postgress (yet)? Should we \ninclude the compiler used in the porting updates (I think so). For \ninstance, I like to use gcc on the Sparc/Solaris2.5.1 and HP-UX/10.20 \nboxes I have because it makes porting things easier between those \nboxes and my Linux systems. Isn't it possible that postgres compiles \nfine using the vendor-supplied compilers, but would fail on gcc, or \nvisa versa?\n\n\n- Fred\n\n\n> I just tried this on Debian with glibc 2.0.6 and it works fine. That\n> is my 6.3 postgresql release outputs 4 hours.\n> \n> Michael\n> \n> --\n> Dr. Michael Meskes, Projekt-Manager \| topystem Systemhaus GmbH\n> [email protected] \| Europark A2, Adenauerstr.\n> 20 [email protected] \| 52146 Wuerselen Go\n> SF49ers! Use Debian GNU/Linux! \| Tel: (+49) 2405/4670-44\n> \n> > ----------\n> > From: \tThomas G. Lockhart[SMTP:[email protected]]\n> > Sent: \tFreitag, 13. Februar 1998 17:57\n> > To: \[email protected]; Bruce Momjian\n> > Cc: \[email protected]\n> > Subject: \t[HACKERS] Re: [PORTS] v6.3 release ToDo list and\n> > supported ports\n> > \n> > > Is the ix86/linux \"confirmed\" port using libc 5 or glibc 2.x (aka\n> > > libc6)? As most major distributions are going to release all new\n> > > versions with glibc I think it's prudent to test on both \"platforms\"\n> > > independently.\n> > \n> > Yes, that is a good point. There is a known bug in the glibc2 math\n> > library which breaks the date/time routines:\n> > \n> > select '4 hours'::timespan;\n> > ---------\n> > @ 4 hours\n> > (1 row)\n> > \n> > comes out instead as\n> > tgl=> select '4 hours'::timespan;\n> > ----------------------------\n> > @ 3 hours 59 mins 60.00 secs\n> > (1 row)\n> > \n> > Oliver was working on patches. Lost his e-mail message; is Oliver\n> > still\n> > here? I've got a RH5.0 linux system at work now, and the shipped\n> > Postgres\n> > installation has this problem. I'd like to see it fixed...\n> > \n> > Bruce, can you add this to the v6.3 ToDo (assuming you agree to take\n> > it\n> > on :)\n> > \n> > - Tom\n> > \n> > \n> > \n> \n", "msg_date": "Mon, 16 Feb 1998 11:00:24 +0000", "msg_from": "\"Frederick W. Reimer\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported p" }, { "msg_contents": "On Mon, 16 Feb 1998, Frederick W. Reimer wrote:\n\n> Could some of the other problems be due to gcc-2.8.0, which I'm not \n\n\tI'm using v2.8.0 on the Solaris Sparc tests I do here...but\nnothing else yet...\n\n\n", "msg_date": "Mon, 16 Feb 1998 11:47:51 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported p" } ]
[ { "msg_contents": "Eeek,\n\nI'd almost forgotten about this suggestion.\n\nI'll start looking through my mailbox and see if there are any\ninteresting snippets of SQL that might form a basis for this.\n\nKeith.\n\nThe Hermit Hacker <[email protected]>\n> Keith Parks <[email protected]>\n> \n> On Tue, 13 Jan 1998, Keith Parks wrote:\n> \n> > Hi All,\n> > \n> > I wonder if it might be useful to maintain a \"bugs\" regression test, where\n> > we could include various tests that people have supplied to highlight bugs\n> > in the system and that have subsequently been fixed?\n> > \n> > I'm thinking along the lines of the current varchar() bug that can be\n> > easily demonstrated in a couple of lines of sql.\n> > \n> > Ideally the varchar() bug test could be included in the varchar tests\n> > but it may be easier to collect such misc things in a seperate file\n> > that could be regularly updated as we get examples.\n> > \n> > The people testing and fixing the problems could supply the additions\n> > to the bugs.sql and bugs.out files as we find/fix them?\n> \n> \tSounds cool to me...someone want to submit a \"start\" to this?\n> \n> Marc G. Fournier \n> Systems Administrator @ hub.org \n> primary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n> \n> \n\n", "msg_date": "Sun, 15 Feb 1998 21:50:06 +0000 (GMT)", "msg_from": "Keith Parks <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] how about a \"bugs\" regression test?" } ]
[ { "msg_contents": "Folks,\n\nOn the spin lock and multiple CPU's, this should not be\na compile time issue, but a runtime issue. What do you \nthink?\n\nJordan Henderson\n\n", "msg_date": "Sun, 15 Feb 1998 21:05:09 -0500", "msg_from": "Jordan Henderson <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] spin locks" }, { "msg_contents": "On Sun, 15 Feb 1998, Jordan Henderson wrote:\n\n> Folks,\n> \n> On the spin lock and multiple CPU's, this should not be\n> a compile time issue, but a runtime issue. What do you \n> think?\n\n\tAnd you are proposing...?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sun, 15 Feb 1998 22:18:59 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] spin locks" }, { "msg_contents": "Folks,\n\nOn the thread in regards to spinlocks and multiple CPU's.\nThe line of thought appeared to be select a compile\ntime option to determine behavior, and whether to yield or not.\n\nI am thinking that if it comes to having alternate code, the\nsystem should be able to make the determination at runtime,\nnot compile time. I don't know if all of the platforms supported\nhave a version of the sysinfo utility, but here is how, at runtime\nit gets the number of CPUS available:\n\n** EXCERPTED FROM SYSINFO SOURCE 3.3.1 *\n/\n * Use sysconf() to find number of CPU's.\n /\nextern char GetNumCpuSysconf()\n{\n int\t\t\t\tNum = -1;\n static char\t\t NumStr = NULL;\n\n if (NumStr)\n\treturn(NumStr);\n\n#if\tdefined(_SC_NPROCESSORS_CONF)\n Num = (int) sysconf(_SC_NPROCESSORS_CONF);\n if (Num >= 0) {\n\tNumStr = itoa(Num);\n\tif (NumStr)\n\t NumStr = strdup(NumStr);\n }\n#endif\t/ _SC_NPROCESSORS_CONF */\n\n return(NumStr);\n}\n\nWhat I would propose, if the decision is made to yield,\nthat at initialization time, the number of CPU's available\nare determined, and a flag set, or, an indirect jump\nchanged. This would allow the software to have both\npersonalities, depending on which system it found it\nself running on.\n\nThoughts?\nJordan Henderson\n\n", "msg_date": "Mon, 16 Feb 1998 09:21:34 -0500", "msg_from": "Jordan Henderson <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] spin locks" }, { "msg_contents": "> \n> Folks,\n> \n> On the spin lock and multiple CPU's, this should not be\n> a compile time issue, but a runtime issue. What do you \n> think?\n> \n> Jordan Henderson\n> \n> \n> \n\nYes.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Mon, 16 Feb 1998 15:50:52 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] spin locks" }, { "msg_contents": "> \n> > \n> > Folks,\n> > \n> > On the spin lock and multiple CPU's, this should not be\n> > a compile time issue, but a runtime issue. What do you \n> > think?\n> > \n> > Jordan Henderson\n> > \n> > \n> > \n> \n> Yes.\n\nYes, if we have the tprintf patch we can also add an option flag and change\nit un the fly while the backends are running.\n\n-- \nMassimo Dal Zotto\n\n+----------------------------------------------------------------------+\n\| Massimo Dal Zotto e-mail: [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\n", "msg_date": "Mon, 16 Feb 1998 23:33:00 +0100 (MET)", "msg_from": "Massimo Dal Zotto <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] spin locks" } ]
[ { "msg_contents": "Hi,\n\nWhy does makefiles/Makefile.linux have this CFLAGS:\n\nCFLAGS = -I$(SRCDIR)/include -I/usr/include/ncurses -\nI/usr/include/readline -Dlinuxalpha -Wall -Wstrict-prototypes -O2 -\nfomit-frame-pointer\n\nnow every linux box seems to be Alpla?\n--\nTatsuo Ishii\[email protected]\n", "msg_date": "Mon, 16 Feb 1998 16:37:04 +0900", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "makefiles/Makefile.linux of 2/15 snapshop" }, { "msg_contents": "> Why does makefiles/Makefile.linux have this CFLAGS:\n>\n> CFLAGS = -I$(SRCDIR)/include -I/usr/include/ncurses -\n> I/usr/include/readline -Dlinuxalpha -Wall -Wstrict-prototypes -O2 -\n> fomit-frame-pointer\n>\n> now every linux box seems to be Alpla?\n\nThat is a known \"feature\". Marc, are you going to back out the last\nchanges or should I put in a new version from a few days ago?\n\n - Tom\n\n", "msg_date": "Tue, 17 Feb 1998 15:31:48 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] makefiles/Makefile.linux of 2/15 snapshop" }, { "msg_contents": "On Tue, 17 Feb 1998, Thomas G. Lockhart wrote:\n\n> > Why does makefiles/Makefile.linux have this CFLAGS:\n> >\n> > CFLAGS = -I$(SRCDIR)/include -I/usr/include/ncurses -\n> > I/usr/include/readline -Dlinuxalpha -Wall -Wstrict-prototypes -O2 -\n> > fomit-frame-pointer\n> >\n> > now every linux box seems to be Alpla?\n> \n> That is a known \"feature\". Marc, are you going to back out the last\n> changes or should I put in a new version from a few days ago?\n\n\tPut in a new version...:)\n\n\n", "msg_date": "Tue, 17 Feb 1998 10:37:31 -0500 (EST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] makefiles/Makefile.linux of 2/15 snapshop" } ]
[ { "msg_contents": "It appears they didn't make it in. The struct features are not in the cvs\nsource so far. Did they get lost? I still have a copy.\n\nMichael\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Mon, 16 Feb 1998 08:58:56 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "My patches from Friday" } ]
[ { "msg_contents": "2/15 snapshot did not pass a regression test.\n\nregression=> CREATE TABLE DEFAULT_TBL (i int DEFAULT 100,x text DEFAULT 'vadim', f float8 DEFAULT 123.456);\nCREATE\nregression=> INSERT INTO DEFAULT_TBL (i, f) VALUES (2, 987.654);\nPQexec() -- Request was sent to backend, but backend closed the channel before responding.\n This probably means the backend terminated abnormally before or while processing the request.\nProgram received signal SIGABRT, Aborted.\n0x2ad246f8 in kill ()\n\nhere is the back trace info.\n\n(gdb) where\n#0 0x2ad246f8 in kill ()\n#1 0x2ad245b8 in raise ()\n#2 0x2ad251cc in abort ()\n#3 0x2de584 in form ()\n#4 0x2de498 in ExcUnCaught ()\n#5 0x2de52c in ExcRaise ()\n#6 0x2ddc0c in ExceptionalCondition ()\n#7 0x2e1e90 in MemoryContextAlloc ()\n#8 0x2e21f4 in palloc ()\n#9 0x269274 in _copyConst ()\n#10 0x26aa80 in copyObject ()\n#11 0x26a2b0 in _copyTargetEntry ()\n#12 0x26aba0 in copyObject ()\n#13 0x26ac1c in copyObject ()\n#14 0x27f480 in preprocess_targetlist ()\n#15 0x27c20c in union_planner ()\n#16 0x27c134 in planner ()\n#17 0x2b2144 in pg_parse_and_plan ()\n#18 0x2b22c0 in pg_exec_query_dest ()\n#19 0x2b224c in pg_exec_query ()\n#20 0x2b3284 in PostgresMain ()\n#21 0x26764c in main ()\n#22 0x2195c0 in _start ()\n--\nTatsuo Ishii\[email protected]\n", "msg_date": "Mon, 16 Feb 1998 17:49:28 +0900", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "MkLinux port broken (2/15 snapshot)" } ]
[ { "msg_contents": "I still cannot connect to the actual snapshot database. initdb works fine,\nbut createuser causes an error:\n\nPQexec() -- Request was sent to backend, but backend closed the channel\nbefore responding.\n This probably means the backend terminated abnormally before or\nwhile processing the request.\ncreateuser: database access failed.\n\nSomehow this hinders my debugging work. :-)\n\nMichael\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Mon, 16 Feb 1998 09:51:32 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "Current snapshot problems" } ]
[ { "msg_contents": "Could anyone please enable ecpg compilation in src/interfaces/Makefile again\nas the CVS source now compiles flawlessly.\n\nAnyway, I currently have a list of 5-6 open bugs in ecpg. So if anyone wants\nto help, step forward. :-)\n\nJoking aside, there is one bug or better missing feature that I need some\ninput on. Does anyone know what the standards say to the prepare command? Is\ndynamic SQL a standard?\n\nMichael\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Mon, 16 Feb 1998 09:53:16 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "ecpg stuff" }, { "msg_contents": "> Joking aside, there is one bug or better missing feature that I need some\n> input on. Does anyone know what the standards say to the prepare command? Is\n> dynamic SQL a standard?\n\nYes it is. I have the Date/Darwen book on the SQL standard, and can look up\nanswers in the book if you like...\n\n - Tom\n\n", "msg_date": "Tue, 17 Feb 1998 15:45:16 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] ecpg stuff" }, { "msg_contents": "Thomas G. Lockhart writes:\n> > Joking aside, there is one bug or better missing feature that I need some\n> > input on. Does anyone know what the standards say to the prepare command? Is\n> > dynamic SQL a standard?\n> \n> Yes it is. I have the Date/Darwen book on the SQL standard, and can look up\n> answers in the book if you like...\n\nYes, I like that. Could you please also lookup:\n\n- the whenever statement\n- and check resp. tell me whether the cursor behaviour is correct. Currently\nthe declare statement is send to the backend via PQexec. The open statement\nis ignored and the fetch is executed as fetch via PQexec. I think the data\nshouldn't be processed before the cursor is opened. But I do not know\nwhat PostgreSQL does with the declare command.\n\nMichael\n\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Wed, 18 Feb 1998 09:12:47 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] ecpg stuff" }, { "msg_contents": "Michael Meskes wrote:\n> \n> - and check resp. tell me whether the cursor behaviour is correct. Currently\n> the declare statement is send to the backend via PQexec. The open statement\n> is ignored and the fetch is executed as fetch via PQexec. I think the data\n> shouldn't be processed before the cursor is opened. But I do not know\n> what PostgreSQL does with the declare command.\n\nDECLARE: parser + optimizer + ExecutorStart (initializes plan nodes:\nchecks permissions, opens tables & indices). \nIs OPEN statement in standard ?\nIf yes then we could call ExecutorStart() for the OPEN someday.\n\nVadim\n", "msg_date": "Wed, 18 Feb 1998 15:55:40 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] ecpg stuff" } ]
[ { "msg_contents": "Yes, and Oracle7 also.\n\nI think with NULL values Andreas is right. The whole statement should be\nNULLed. That to me is the intuitive behaviour.\n\nMichael\n\n--\nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n\n> -----Original Message-----\n> From:\tVadim B. Mikheev [SMTP:[email protected]]\n> Sent:\tSunday, February 15, 1998 12:33 PM\n> To:\tZeugswetter Andreas SARZ\n> Cc:\t'[email protected]'; Michael Meskes; [email protected]\n> Subject:\tRe: [HACKERS] Re: Subselects open issue Nr. 5\n> \n> Ok. I'll fix this. As I see, this is exactly what Oracle 6 does, but\n> \n> Zeugswetter Andreas SARZ wrote:\n> > \n> > Informix treats the subselect as NULL if no rows are returned.\n> > Therefore all parent rows that are not null are returned.\n> > \n> > select * from taba where a not in (<a select returning no row>);\n> --\n> > is same as\n> > select * from taba where a is not null;\n> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n> Oracle returns tuples with A being NULL!!! and more of that (table B\n> is empty):\n> \n> SQL> select count() from a where x > ALL (select from b);\n> \n> COUNT()\n> ----------\n> 2\n> \n> and result is the same for all OP-s with ALL modifier... And\n> \n> SQL> select count() from a where x in (select * from b);\n> \n> COUNT()\n> ----------\n> 0\n> having tuple with NULL in X...\n> \n> Who's right ?\n> What standard says ?\n> \n> Vadim\n", "msg_date": "Mon, 16 Feb 1998 11:22:48 +0100", "msg_from": "\"Meskes, Michael\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] Re: Subselects open issue Nr. 5" }, { "msg_contents": "Meskes, Michael wrote:\n> \n> Yes, and Oracle7 also.\n> \n> I think with NULL values Andreas is right. The whole statement should be\n> NULLed. That to me is the intuitive behaviour.\n\nNot sure.\nIMHO, any element, either with defined value or with undefined value (NULL),\ncan't be contained by empty set.\n\nHm, btw, just curious, what Informix returns for\n\nselect from taba where (a,b) not in (<a select returning no row>);\n\nhaving in taba tuples with (a,b) in\n\n(NULL, a_value)\n(NULL, NULL)\n\n? \nDoes it return all tuples except for (NULL,NULL) ?\n\nVadim\n", "msg_date": "Mon, 16 Feb 1998 18:56:37 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Subselects open issue Nr. 5" } ]
[ { "msg_contents": "On Sun, 15 Feb 1998, Marin D wrote:\n\n> \n> Hi!\n> \n> You can try to typecast to float8\n> \n> I have no idea how MySql is doing SUM() - probably it uses unsigned int\n> for SUM() so this gives u twice bigger upper limit - try it with MySql\n> with sum exceeding 2^32. I suppose the result will be negative ... \n> \n\nWell, I asked this question to MySql mailing list:\n\n> sferac> What's meaning \"inf\" as a result during a SUM() ?\n> sferac> Is it an overflow warning, and if so what's the max range of SUM()\n> sferac> before to have an overflow ?\n>\n> mysql> update t set myfloat=myfloat9.1;\n> sferac> Query OK, 1415 rows affected (4.94 sec)\n>\n> mysql> select sum(myfloat) from t;\n> sferac> 1 row in set (0.04 sec)\n>\n> sferac> +--------------+\n> sferac> \| sum(myfloat) \|\n> sferac> +--------------+\n> sferac> \| Inf \|\n> sferac> +--------------+\n>\n> mysqld does all normal calculation with doubles (bit functions are\n> done with longlong); The range of a double is typical something like:\n>\n> #define DBL_MAX 1.7976931348623157E+308\n> #define DBL_MIN 2.2250738585072014E-308\n>\n> Yours,\n> Monty\n\nIsn't it a good solution to implement PostgreSQL SUM() function ?\nWhat do you think about ?\n\n____________________________Jose' Soares Da Silva_____________________________\n Via Bazzanese, 69 ),,,( www.sferacarta.com\n Casalecchio R.(BO) Italy +\|o o\|+ [email protected]\n Fax. ++39 51 6131537 /. .\\ Tel. ++39 51 591054\n---------------------------------(_---_)--------------------------------------\n\"...when men cease to destroy the animal race the lion\n and the lamb can dwell together...\" (Joseph Smith)\n\n", "msg_date": "Mon, 16 Feb 1998 10:39:24 +0000 (GMT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: [BUGS] agregate function sum error" }, { "msg_contents": "> > You can try to typecast to float8\n> >\n> > I have no idea how MySql is doing SUM() - probably it uses unsigned int\n> > for SUM() so this gives u twice bigger upper limit - try it with MySql\n> > with sum exceeding 2^32. I suppose the result will be negative ...\n> >\n>\n> Well, I asked this question to MySql mailing list:\n>\n> > sferac> What's meaning \"inf\" as a result during a SUM() ?\n> > sferac> Is it an overflow warning, and if so what's the max range of SUM()\n> > sferac> before to have an overflow ?\n> >\n> > mysql> update t set myfloat=myfloat9.1;\n> > sferac> Query OK, 1415 rows affected (4.94 sec)\n> >\n> > mysql> select sum(myfloat) from t;\n> > sferac> 1 row in set (0.04 sec)\n> >\n> > sferac> +--------------+\n> > sferac> \| sum(myfloat) \|\n> > sferac> +--------------+\n> > sferac> \| Inf \|\n> > sferac> +--------------+\n> >\n> > mysqld does all normal calculation with doubles (bit functions are\n> > done with longlong); The range of a double is typical something like:\n> >\n> > #define DBL_MAX 1.7976931348623157E+308\n> > #define DBL_MIN 2.2250738585072014E-308\n\nThe limits are the same as for Postgres. The difference is that Postgres throws\nan error on floating overflows, rather than returning \"infinity\". I can think of\narguments for both behaviors, but one problem is that at least one older platform\n(vax) does not have IEEE arithmetic so can't work with \"infinity\" :(\n\nWe do not currently use \"longlong\" for any integers in the backend, because I\ndidn't know how this is implemented/supported for _every_ platform on our\nsupported list. I had put an \"int8\" package into the contrib directory to test\nthis, and to get feedback for other platforms, but have gotten only one response\nand it was for a platform (alpha) which I already knew would work.\n\nPerhaps I'll make a concerted effort to get this tested for v6.4; scrappy had\nsuggested just putting int8s into the backend and then working out the porting\nissues and I think I'll do that :) Then, we can use int8 variables in other\nplaces where necessary for v6.5...\n\n - Tom\n\n", "msg_date": "Tue, 17 Feb 1998 15:55:34 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [BUGS] agregate function sum error" }, { "msg_contents": "\n\n\n\nOn Tue, 17 Feb 1998, Thomas G. Lockhart wrote:\n\n> > > You can try to typecast to float8\n> > >\n> > > I have no idea how MySql is doing SUM() - probably it uses unsigned int\n> > > for SUM() so this gives u twice bigger upper limit - try it with MySql\n> > > with sum exceeding 2^32. I suppose the result will be negative ...\n> > >\n> >\n> > Well, I asked this question to MySql mailing list:\n> >\n> > > sferac> What's meaning \"inf\" as a result during a SUM() ?\n> > > sferac> Is it an overflow warning, and if so what's the max range of SUM()\n> > > sferac> before to have an overflow ?\n> > >\n> > > mysql> update t set myfloat=myfloat9.1;\n> > > sferac> Query OK, 1415 rows affected (4.94 sec)\n> > >\n> > > mysql> select sum(myfloat) from t;\n> > > sferac> 1 row in set (0.04 sec)\n> > >\n> > > sferac> +--------------+\n> > > sferac> \| sum(myfloat) \|\n> > > sferac> +--------------+\n> > > sferac> \| Inf \|\n> > > sferac> +--------------+\n> > >\n> > > mysqld does all normal calculation with doubles (bit functions are\n> > > done with longlong); The range of a double is typical something like:\n> > >\n> > > #define DBL_MAX 1.7976931348623157E+308\n> > > #define DBL_MIN 2.2250738585072014E-308\n> \n> The limits are the same as for Postgres. The difference is that Postgres throws\n> an error on floating overflows, rather than returning \"infinity\". I can think of\n> arguments for both behaviors, but one problem is that at least one older platform\n> (vax) does not have IEEE arithmetic so can't work with \"infinity\" :(\n> \n> We do not currently use \"longlong\" for any integers in the backend, because I\n> didn't know how this is implemented/supported for _every_ platform on our\n> supported list. I had put an \"int8\" package into the contrib directory to test\n> this, and to get feedback for other platforms, but have gotten only one response\n> and it was for a platform (alpha) which I already knew would work.\n> \n> Perhaps I'll make a concerted effort to get this tested for v6.4; scrappy had\n> suggested just putting int8s into the backend and then working out the porting\n> issues and I think I'll do that :) Then, we can use int8 variables in other\n> places where necessary for v6.5...\n> \n> - Tom\n> \n> \n\n_______________________________________________________________________________\n Aleksandr Belinsky\n Sternberg Astronomical Institute, Moscow University (Russia)\n E-mail: [email protected] http://zeus.sai.msu.su/~aleks/\n _____________________________________________\n\n", "msg_date": "Tue, 17 Feb 1998 21:31:03 +0300 (MSK)", "msg_from": "\"Aleksandr A. Belinsky\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [BUGS] agregate function sum error" }, { "msg_contents": "On Tue, 17 Feb 1998, Thomas G. Lockhart wrote:\n\n> > > You can try to typecast to float8\n> > >\n> > > I have no idea how MySql is doing SUM() - probably it uses unsigned int\n> > > for SUM() so this gives u twice bigger upper limit - try it with MySql\n> > > with sum exceeding 2^32. I suppose the result will be negative ...\n> > >\n> >\n> > Well, I asked this question to MySql mailing list:\n> >\n> > > sferac> What's meaning \"inf\" as a result during a SUM() ?\n> > > sferac> Is it an overflow warning, and if so what's the max range of SUM()\n> > > sferac> before to have an overflow ?\n> > >\n> > > mysql> update t set myfloat=myfloat9.1;\n> > > sferac> Query OK, 1415 rows affected (4.94 sec)\n> > >\n> > > mysql> select sum(myfloat) from t;\n> > > sferac> 1 row in set (0.04 sec)\n> > >\n> > > sferac> +--------------+\n> > > sferac> \| sum(myfloat) \|\n> > > sferac> +--------------+\n> > > sferac> \| Inf \|\n> > > sferac> +--------------+\n> > >\n> > > mysqld does all normal calculation with doubles (bit functions are\n> > > done with longlong); The range of a double is typical something like:\n> > >\n> > > #define DBL_MAX 1.7976931348623157E+308\n> > > #define DBL_MIN 2.2250738585072014E-308\n> \n> The limits are the same as for Postgres. The difference is that Postgres throws\n> an error on floating overflows, rather than returning \"infinity\". \n\nI think overflows on float8 is OK, the problem is about overflows on int4, \nint2 and float4.\nIf PostgreSQL does all calculation using float8 (even on int2, int4 and float4)\nwe have acceptable overflows. Why we don't do this in this way ?\n\nI can think of\n> arguments for both behaviors, but one problem is that at least one older platform\n> (vax) does not have IEEE arithmetic so can't work with \"infinity\" :(\n> \n> We do not currently use \"longlong\" for any integers in the backend, because I\n> didn't know how this is implemented/supported for _every_ platform on our\n> supported list. I had put an \"int8\" package into the contrib directory to test\n> this, and to get feedback for other platforms, but have gotten only one response\n> and it was for a platform (alpha) which I already knew would work.\n> \n> Perhaps I'll make a concerted effort to get this tested for v6.4; scrappy had\n> suggested just putting int8s into the backend and then working out the porting\n> issues and I think I'll do that :) Then, we can use int8 variables in other\n> places where necessary for v6.5...\n Ciao, Jose'\n\n", "msg_date": "Thu, 19 Feb 1998 12:28:59 +0000 (GMT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [BUGS] agregate function sum error" }, { "msg_contents": "> \n> On Tue, 17 Feb 1998, Thomas G. Lockhart wrote:\n> \n> > > > You can try to typecast to float8\n> > > >\n> > > > I have no idea how MySql is doing SUM() - probably it uses unsigned int\n> > > > for SUM() so this gives u twice bigger upper limit - try it with MySql\n> > > > with sum exceeding 2^32. I suppose the result will be negative ...\n> > > >\n> > >\n> > > Well, I asked this question to MySql mailing list:\n> > >\n> > > > sferac> What's meaning \"inf\" as a result during a SUM() ?\n> > > > sferac> Is it an overflow warning, and if so what's the max range of SUM()\n> > > > sferac> before to have an overflow ?\n> > > >\n> > > > mysql> update t set myfloat=myfloat9.1;\n> > > > sferac> Query OK, 1415 rows affected (4.94 sec)\n> > > >\n> > > > mysql> select sum(myfloat) from t;\n> > > > sferac> 1 row in set (0.04 sec)\n> > > >\n> > > > sferac> +--------------+\n> > > > sferac> \| sum(myfloat) \|\n> > > > sferac> +--------------+\n> > > > sferac> \| Inf \|\n> > > > sferac> +--------------+\n> > > >\n> > > > mysqld does all normal calculation with doubles (bit functions are\n> > > > done with longlong); The range of a double is typical something like:\n> > > >\n> > > > #define DBL_MAX 1.7976931348623157E+308\n> > > > #define DBL_MIN 2.2250738585072014E-308\n> > \n> > The limits are the same as for Postgres. The difference is that Postgres throws\n> > an error on floating overflows, rather than returning \"infinity\". \n> \n> I think overflows on float8 is OK, the problem is about overflows on int4, \n> int2 and float4.\n> If PostgreSQL does all calculation using float8 (even on int2, int4 and float4)\n> we have acceptable overflows. Why we don't do this in this way ?\n\nCan you imagine the problems we would have? Integral types and floats\nare two different things. They behave differently in division,\nrounding, and overflows. For those reasons, and performance, we can't\njust use floats all the time.\n\n\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Thu, 19 Feb 1998 11:35:03 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [BUGS] agregate function sum error" }, { "msg_contents": "On Thu, 19 Feb 1998, Bruce Momjian wrote:\n\n> > \n> > On Tue, 17 Feb 1998, Thomas G. Lockhart wrote:\n> > \n> > > > > You can try to typecast to float8\n> > > > >\n> > > > > I have no idea how MySql is doing SUM() - probably it uses unsigned int\n> > > > > for SUM() so this gives u twice bigger upper limit - try it with MySql\n> > > > > with sum exceeding 2^32. I suppose the result will be negative ...\n> > > > >\n> > > >\n> > > > Well, I asked this question to MySql mailing list:\n> > > >\n> > > > > sferac> What's meaning \"inf\" as a result during a SUM() ?\n> > > > > sferac> Is it an overflow warning, and if so what's the max range of SUM()\n> > > > > sferac> before to have an overflow ?\n> > > > >\n> > > > > mysql> update t set myfloat=myfloat9.1;\n> > > > > sferac> Query OK, 1415 rows affected (4.94 sec)\n> > > > >\n> > > > > mysql> select sum(myfloat) from t;\n> > > > > sferac> 1 row in set (0.04 sec)\n> > > > >\n> > > > > sferac> +--------------+\n> > > > > sferac> \| sum(myfloat) \|\n> > > > > sferac> +--------------+\n> > > > > sferac> \| Inf \|\n> > > > > sferac> +--------------+\n> > > > >\n> > > > > mysqld does all normal calculation with doubles (bit functions are\n> > > > > done with longlong); The range of a double is typical something like:\n> > > > >\n> > > > > #define DBL_MAX 1.7976931348623157E+308\n> > > > > #define DBL_MIN 2.2250738585072014E-308\n> > > \n> > > The limits are the same as for Postgres. The difference is that Postgres throws\n> > > an error on floating overflows, rather than returning \"infinity\". \n> > \n> > I think overflows on float8 is OK, the problem is about overflows on int4, \n> > int2 and float4.\n> > If PostgreSQL does all calculation using float8 (even on int2, int4 and float4)\n> > we have acceptable overflows. Why we don't do this in this way ?\n> \n> Can you imagine the problems we would have? Integral types and floats\n> are two different things. They behave differently in division,\n> rounding, and overflows. For those reasons, and performance, we can't\n> just use floats all the time.\n> \nI see..., then what we need is int8 as Tom said, to avoid to write code as:\n\nselect cast mysmall as int4 * cast mysmall as int4 from mytable;\nselect mysmall::int4 * mysmall::int4 from mytable;\nselect int4(mysmall) * int4(mysmall) from mytable;\nselect sum(int4(mysmall)) from mytable;\n\nthat has nothing to do with SQL standard, and isn't too friendly,\nthose conversions should be done automaticly by PostgreSQL whitout\nuser's help as MySQL, Solid and other databases do.\n-----------\nPS: I installed int8 in my Linux box, but I have an error during insert time:\n\ninsert into qtest values('123','456');\nERROR: fmgr_info: function 159745: cache lookup failed\n\ninsert into qtest values('123','4567890123456789');\nERROR: fmgr_info: function 159745: cache lookup failed\n\ninsert into qtest values('4567890123456789','123');\nERROR: fmgr_info: function 159745: cache lookup failed\n\ninsert into qtest values('4567890123456789','4567890123456789');\nERROR: fmgr_info: function 159745: cache lookup failed\n\ninsert into qtest values('4567890123456789','-4567890123456789');\nERROR: fmgr_info: function 159745: cache lookup failed\n\n Ciao, Jose'\n\n", "msg_date": "Fri, 20 Feb 1998 11:24:18 +0000 (GMT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [BUGS] agregate function sum error" }, { "msg_contents": "> > > > The limits are the same as for Postgres. The difference is that Postgres throws\n> > > > an error on floating overflows, rather than returning \"infinity\".\n> > >\n> > > I think overflows on float8 is OK, the problem is about overflows on int4,\n> > > int2 and float4.\n\nWhat would it take to have backends start signalling ERROR (not core dump :) on integer\noverflows? It would be nice to have the compile-time option; perhaps it is already\nsomewhere in the backend.\n\nThe SQL92 standard requires an abort on floating overflows, though _just_ because the\nstandard says it doesn't mean we have to do it. We should do what makes sense. However,\nonce an IEEE value becomes \"Inf\" further math operations keep it at \"Inf\", which\npropagates the problem farther into your calculation. Probably better to signal the\nerror and let the user fix the overflow problem first. However, if you want to generate\ncompile-time patches...\n\n> > > If PostgreSQL does all calculation using float8 (even on int2, int4 and float4)\n> > > we have acceptable overflows. Why we don't do this in this way ?\n> > Can you imagine the problems we would have? Integral types and floats\n> > are two different things. They behave differently in division,\n> > rounding, and overflows. For those reasons, and performance, we can't\n> > just use floats all the time.\n> >\n> I see..., then what we need is int8 as Tom said, to avoid to write code as:\n>\n> select cast mysmall as int4 * cast mysmall as int4 from mytable;\n> select mysmall::int4 * mysmall::int4 from mytable;\n> select int4(mysmall) * int4(mysmall) from mytable;\n> select sum(int4(mysmall)) from mytable;\n>\n> that has nothing to do with SQL standard, and isn't too friendly,\n> those conversions should be done automaticly by PostgreSQL whitout\n> user's help as MySQL, Solid and other databases do.\n\nWell, not all real databases do this, though I see your point. There is a performance\ntradeoff between always promoting types when doing arithmetic and leaving the type\nconsistant. Also, \"hidden\" promotions may be less intuitive for some users, and lead to\nunexpected behavior elsewhere.\n\n> PS: I installed int8 in my Linux box, but I have an error during insert time:\n> insert into qtest values('123','456');\n> ERROR: fmgr_info: function 159745: cache lookup failed\n\nHmm. Works on my machine (980217 cvs tree):\n\ncreate table qtest(q1 int8, q2 int8);\nCREATE\n...\ninsert into qtest values('123','4567890123456789');\nINSERT 1018571 1\n...\nselect * from qtest;\n q1\| q2\n----------------+-----------------\n 123\| 456\n 123\| 4567890123456789\n4567890123456789\| 123\n4567890123456789\| 4567890123456789\n4567890123456789\|-4567890123456789\n(5 rows)\n\nDid you try from a clean database? How about from a clean install??\n\n - Tom\n\n", "msg_date": "Fri, 20 Feb 1998 13:55:12 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [BUGS] agregate function sum error" }, { "msg_contents": "On Fri, 20 Feb 1998, Thomas G. Lockhart wrote:\n\n> > > > > The limits are the same as for Postgres. The difference is that Postgres throws\n> > > > > an error on floating overflows, rather than returning \"infinity\".\n> > > >\n> > > > I think overflows on float8 is OK, the problem is about overflows on int4,\n> > > > int2 and float4.\n> \n> What would it take to have backends start signalling ERROR (not core dump :) on integer\n> overflows? It would be nice to have the compile-time option; perhaps it is already\n> somewhere in the backend.\n> \n> The SQL92 standard requires an abort on floating overflows, though _just_ because the\n> standard says it doesn't mean we have to do it. We should do what makes sense. However,\n> once an IEEE value becomes \"Inf\" further math operations keep it at \"Inf\", which\n> propagates the problem farther into your calculation. Probably better to signal the\n> error and let the user fix the overflow problem first. However, if you want to generate\n> compile-time patches...\n> \n> > > > If PostgreSQL does all calculation using float8 (even on int2, int4 and float4)\n> > > > we have acceptable overflows. Why we don't do this in this way ?\n> > > Can you imagine the problems we would have? Integral types and floats\n> > > are two different things. They behave differently in division,\n> > > rounding, and overflows. For those reasons, and performance, we can't\n> > > just use floats all the time.\n> > >\n> > I see..., then what we need is int8 as Tom said, to avoid to write code as:\n> >\n> > select cast mysmall as int4 * cast mysmall as int4 from mytable;\n> > select mysmall::int4 * mysmall::int4 from mytable;\n> > select int4(mysmall) * int4(mysmall) from mytable;\n> > select sum(int4(mysmall)) from mytable;\n> >\n> > that has nothing to do with SQL standard, and isn't too friendly,\n> > those conversions should be done automaticly by PostgreSQL whitout\n> > user's help as MySQL, Solid and other databases do.\n> \n> Well, not all real databases do this, though I see your point. There is a performance\n> tradeoff between always promoting types when doing arithmetic and leaving the type\n> consistant. Also, \"hidden\" promotions may be less intuitive for some users, and lead to\n> unexpected behavior elsewhere.\n> \nHi Tom,\nApologies for my insistence and for my bad english, I like to watch things\nfrom point of view of end users. \n(are databases for end users or programmers ??)\nWe programmers, often forget that end users don't want to know what's happening\nbehind their own screen, they don't want write ostrogothics sentences ala C\nto query a database, they had learned SQL whith so much effort and now they\nwant to query database without thought about dificulties, that C find when\nit need convert numbers from one type to another.\nI thing that one of the bigest dificulty of PostgreSQL is how it treats numbers.\nSQL standard says that a number is comparable to, and compatible with, all\nother numbers (that is: all numbers are mutually comparable and mutually\nassignable). I think it means directly, without force it with the CAST or\nfunctions.\nI can't imagine end users using things like this:\n UPDATE my_table SET my_small=int2(my_int4);\n SELECT SUM(float8(my_int) FROM my_table;\n SELECT mysmall::int4 * mysmall::int4 FROM my_table;\n \nEnd users hate uneccessaries or non sense messages like the following\nwhen he/she writes a sentence like this:\n\n UPDATE my_table SET my_int=my_small;\n\nand PostgreSQL show him/her this message ?\n---------------------------------------------------------------------\nWARN: type of my_small does not match target column my_int\n---------------------------------------------------------------------\n- What message is this ? If PostgreSQL understand that I want to do:\n UPDATE my_table SET my_int=my_small::int4;\nwhy it doesn't do that for me?\n\n SELECT my_small * my_float FROM my_table;\n---------------------------------------------------------------------\nNOTICE:there is no operator * for types int2 and float8\nNOTICE:You will either have to retype this query using an\nNOTICE:explicit cast, or you will have to define the operator\nWARN:* for int2 and float8 using CREATE OPERATOR\n---------------------------------------------------------------------\n\n- Well, if PostgreSQL know what I want why it doesn't do it for me ?\nI know that conversion is more expensive than warnning but it's more\nfriendly. What do you think? ;-)\n Regards, Jose'\n\n", "msg_date": "Mon, 23 Feb 1998 13:36:10 +0000 (GMT)", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [BUGS] agregate function sum error" }, { "msg_contents": "> I thing that one of the bigest dificulty of PostgreSQL is how it treats numbers.\n> SQL standard says that a number is comparable to, and compatible with, all\n> other numbers (that is: all numbers are mutually comparable and mutually\n> assignable). I think it means directly, without force it with the CAST or\n> functions.\n\nI agree that the numeric (and other) type coersions need some more work. One of the reasons\nwhy we have been reluctant to embed _too_ many automatic coersions into the backend is that\nit does not extend well into the type-extensible features of Postgres. But I hope to look at\nthis some for v6.4.\n\n> and PostgreSQL show him/her this message ?\n> WARN: type of my_small does not match target column my_int\n\nthis is an example of behavior which I think we _can_ change...\n\n> - Well, if PostgreSQL know what I want why it doesn't do it for me ?\n> I know that conversion is more expensive than warnning but it's more\n> friendly. What do you think? ;-)\n\nI agree up to a point. Once a user has chosen to use an \"optimized data type\" like \"int2\"\nrather than \"integer\" or \"int4\" then they have also chosen to give up some capabilities of\nthe generic integer type such as dynamic range. However, some extra conversion stuff can and\nshould happen, and I have some ideas on how to do this. If any hacker wants to work on it,\nlet's start talking after the upcoming v6.3 release...\n\n - Tom\n\n", "msg_date": "Tue, 24 Feb 1998 13:29:27 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [BUGS] agregate function sum error" } ]
[ { "msg_contents": "Hi, all.\n\nThis one is only to call the developer's attention to a message I've\nposted to psql-ports regarding 6.3 beta problems on DEC alpha /Digital\nUnix.\n\nI know that the \"medicine men\" :-) also read psql-ports, but... just in\ncase.\n\n\tPedro.\n\n-------------------------------------------------------------------\nPedro José Lobo Perea Tel: +34 1 336 78 19\nCentro de Cálculo Fax: +34 1 331 92 29\nEUIT Telecomunicación - UPM e-mail: [email protected]\n\n", "msg_date": "Mon, 16 Feb 1998 12:53:05 +0100 (MET)", "msg_from": "\"Pedro J. Lobo\" <[email protected]>", "msg_from_op": true, "msg_subject": "DEC alpha/6.3beta problems reported on 'ports'" } ]
[ { "msg_contents": "Hi, all.\n\nI have compiled the Feb. 15 snapshot on a DEC alpha and, besides other\nproblems that I've reported in the psql-ports list, there is a bug in\nconfigure that has been there for quite a while.\n\nI use this command to do the 'configure' thing:\n\n./configure --prefix=/usr/local/pgsql.beta --enable-locale\n--with-pgport=5440 --enable-cassert --with-compiler=cc\n\nI use --with-compiler=cc because I have both DEC C ang GNU gcc installed,\nand I've found DEC C to generate better code (faster) than gcc.\n\nThe problem is that when the configure script tries to figure out if we\nare using gcc, to check if it needs the -traditional flag, it dies (the\nscript).\n\nThis is the patch I use as a workaround:\n\n----------------------------------------------------\n* configure.orig do 15 feb 09:00:37 1998\n--- configure lu 16 feb 11:37:21 1998\n***********\n* 3224,3230 ***\n fi\n rm -f conftest\n\n! if test $ac_cv_prog_gcc = yes; then\n echo $ac_n \"checking whether ${CC-cc} needs -traditional\"\"... $ac_c\" 1>&6\n echo \"configure:3230: checking whether ${CC-cc} needs -traditional\" >&5\n if eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\";\nthen\n--- 3224,3230 ----\n fi\n rm -f conftest*\n\n! if test x$ac_cv_prog_gcc = xyes; then\n echo $ac_n \"checking whether ${CC-cc} needs -traditional\"\"... $ac_c\" 1>&6\n echo \"configure:3230: checking whether ${CC-cc} needs -traditional\" >&5\n if eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\";\nthen\n-----------------------------------------------------------\n\nThe problem is that $ac_cv_prog_gcc is empty, instead of 'no' or\nwhatsoever. The patch is not a solution, because configure is generated\nfrom configure.in, and I don't know how to patch it to get a working\n'configure'.\n\nIs there any autoconf guru that has a solution? Marc?\n\n\tPedro.\n\n-------------------------------------------------------------------\nPedro José Lobo Perea Tel: +34 1 336 78 19\nCentro de Cálculo Fax: +34 1 331 92 29\nEUIT Telecomunicación - UPM e-mail: [email protected]\n\n", "msg_date": "Mon, 16 Feb 1998 13:02:06 +0100 (MET)", "msg_from": "\"Pedro J. Lobo\" <[email protected]>", "msg_from_op": true, "msg_subject": "configure doesn't handle --with-compiler correctly" }, { "msg_contents": "On Mon, 16 Feb 1998, Pedro J. Lobo wrote:\n\n> The problem is that $ac_cv_prog_gcc is empty, instead of 'no' or\n> whatsoever. The patch is not a solution, because configure is generated\n> from configure.in, and I don't know how to patch it to get a working\n> 'configure'.\n> \n> Is there any autoconf guru that has a solution? Marc?\n\n\tThe only solution that I have is to re-install autoconf after\nmodifying the appropriate line in acspecific.m4 ... now every time I do an\nautoconf at this end, it will have the 'x's in front of the appropraite\nvariables...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 16 Feb 1998 21:31:19 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] configure doesn't handle --with-compiler correctly" } ]
[ { "msg_contents": "Edmund Mergl <[email protected]> wrote:\n\n> Bruce Momjian wrote:\n> >\n> > >\n> > > Hi all,\n> > >\n> > > I would like to add the following item to the\n> > > wishlist:\n> > >\n> > > currently the libpq-function PQfsize returns the\n> > > size in bytes of the field with the given field-\n> > > index. If the field is a variable-length field,\n> > > PQfsize returns -1.\n> > >\n> > > In general as well as for some internal functionality\n> > > of the perl interface it would be useful, if PQfsize\n> > > would return the number of bytes even for variable-\n> > > length fields.\n> >\n> > We have a new atttypmod field that holds the length for the column\nas\n> > defined at table-creation time for char() and varchar(). Is that\nthe\n> > number you want?\n\nI think this number should be available in protocol as several\nfront-ends want to know it.\n\nIt can of course be read from the column definition tables but that\nwould mean that the front-end\nhas to parse the query as well.\n\nWe would need to supply reasonable defaults for results of expressions,\nmeaning that length of a\nconcatenation of varchar(10) and\nvarchar(20) should be returned as 30 but if it is currently undoable it\ncould be returned as -1\n(variable) for now.\n\n> for varchar() and also for the datatype text it would be\n> nice to know the actual length. If this is not possible,\n> then the length at creation time is still better than -1.\n>\nThe actual length is given in each data tuple and can be different for\neach tuple.\n\nWhat would be interesting is maximum length (for example the PostODBC\ndoes weird things (it\nexecutes the query and measures each string\n:) to determine this, and last time I checked did it in a way that could\nbreak delphi as different\nwhere clause got different lengths)\n\nI suggested a hack to fix it (at the time of v6.1), that would open a\nbinary cursor (as these are\nzero padded to max length) and read\nonly enough records to get a non-null record for each variable field,\nbut I'm not sure it got\nimplemented as I lost interest in ODBC\nsoon after it.\n\nI'm not sure if it should be included in protocol as a protocol change,\nor would some additional\ncommand that gives this info for an\nopen cursor be preferrable.\n\nHannu\n\n\n", "msg_date": "Mon, 16 Feb 1998 14:07:16 +0200", "msg_from": "Hannu Krosing <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Adding field length to protocol." } ]
[ { "msg_contents": "Guess what !\n\nIt (Informix 9.12 and DB/2 4.1) says syntax error (at the first comma).\n(Even looked up the Manuals)\nHaha Hihi Hoho. I guess you beat them here Vadim+Bruce+Tom. * hear the cork\npopping ? \n\nAndreas\n\nPS.: from the logical point of view, I think all rows from x should qualify\nfor a where (a,b) not in (empty set)\nbecause for me NULL is not an empty set, at least it is treated as a value\nin a unique index.\nOn the other hand you could argue: the whole set is NULL so a not in ()\nshould filter where a not null. \nI guess no standard has thought about that so far. (Tom ?)\nSummary: I guess it is for us to decide. So I would do exactly as you said\nand return all except (NULL,NULL)\n\n\tVadim B. Mikheev wrote:\n> Meskes, Michael wrote:\n> > \n> > Yes, and Oracle7 also.\n> > \n> > I think with NULL values Andreas is right. The whole statement should be\n> > NULLed. That to me is the intuitive behaviour.\n> \n> Not sure.\n> IMHO, any element, either with defined value or with undefined value\n> (NULL),\n> can't be contained by empty set.\n> \n> Hm, btw, just curious, what Informix returns for\n> \n> select from taba where (a,b) not in (<a select returning no row>);\n> \n> having in taba tuples with (a,b) in\n> \n> (NULL, a_value)\n> (NULL, NULL)\n> \n> ? \n> Does it return all tuples except for (NULL,NULL) ?\n> \n> Vadim\n> \n> \n", "msg_date": "Mon, 16 Feb 1998 14:19:14 +0100", "msg_from": "Zeugswetter Andreas SARZ <[email protected]>", "msg_from_op": true, "msg_subject": "AW: [HACKERS] Re: Subselects open issue Nr. 5" }, { "msg_contents": "Zeugswetter Andreas SARZ wrote:\n> \n> Guess what !\n> \n> It (Informix 9.12 and DB/2 4.1) says syntax error (at the first comma).\n\n...like SyBase 11... \n\n> (Even looked up the Manuals)\n> Haha Hihi Hoho. I guess you beat them here Vadim+Bruce+Tom. * hear the cork\n> popping ? *\n> \n> Andreas\n> \n> PS.: from the logical point of view, I think all rows from x should qualify\n> for a where (a,b) not in (empty set)\n> because for me NULL is not an empty set, at least it is treated as a value\n> in a unique index.\n> On the other hand you could argue: the whole set is NULL so a not in ()\n> should filter where a not null.\n> I guess no standard has thought about that so far. (Tom ?)\n> Summary: I guess it is for us to decide. So I would do exactly as you said\n> and return all except (NULL,NULL)\n\nI prefer \"logical point of view\" and vote for Oracle-like behaviour.\nBTW, it's easy to implement...\n\nVadim\n", "msg_date": "Mon, 16 Feb 1998 20:53:17 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: [HACKERS] Re: Subselects open issue Nr. 5" }, { "msg_contents": "> I prefer \"logical point of view\" and vote for Oracle-like behaviour.\n> BTW, it's easy to implement...\n\nSorry I haven't had time to look up the standard. However, in the absence of\nthat I vote for the \"logical point of view\" also. This is better called the\n\"intuitive point of view\"?\n\n - Tom\n\n", "msg_date": "Tue, 17 Feb 1998 15:41:50 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: AW: [HACKERS] Re: Subselects open issue Nr. 5" } ]
[ { "msg_contents": "Hm, new one. Oracle 6:\n\nSQL> select * from a where (x,z) in (select y, l from b);\n\n X Z\n---------- ----------\n 1 2\n\nSQL> select * from a where (x,z) = ANY (select y, l from b);\n\n X Z\n---------- ----------\n 1 2\n\nSQL> select * from a where (x,z) >= ANY (select y, l from b);\nselect * from a where (x,z) >= ANY (select y, l from b)\n *\nERROR at line 1:\nORA-00920: invalid relational operator\n\n: only '=' and '<>' are allowed if there are more than one \nexpression on the left side of clause with subselect...\n\nIs this in standard or Oracle' preference ?\n\nVadim\nP.S. Maybe I should ask in DB newsgroups ?..\n", "msg_date": "Mon, 16 Feb 1998 21:09:37 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": true, "msg_subject": "Subselects open issue Nr. NEW" } ]
[ { "msg_contents": "Just compiled and checked the latest snapshot on AIX 4.1.5 using aix_41\ntemplate.\nAll ok ! except:\n\trounding errors\t\n\tResult too large instead of Math result not representable\n\t> ERROR: exp() result is out of range \t\tin float8,\nthis is correct (error on reference platform)\n\t< \|Sat May 10 23:59:12 1947 PST\n\t---\n\t> \|Sun May 11 00:59:12 1947 PDT\t\tin all time related\nstuff PDT instead of PST\n\nso can be stated as supported platform.\n\nMy gcc 2.8 stuff also works on 4.1.5, all except dynamic loading (probably\nonly a problem with my system \nheader files)\n\nAIX 4.2 and up will prbly want a new template since it supports libdl.a. I\nthink this will be an easy port,\nmaybe even generic will work.\nshared libs will need:\n\tld -G -bexpall -bnoentry refint.o -o refint.so -lc -lm\n\nAndreas\n", "msg_date": "Mon, 16 Feb 1998 15:43:26 +0100", "msg_from": "Zeugswetter Andreas SARZ <[email protected]>", "msg_from_op": true, "msg_subject": "AIX 4.1.5 is OK as of snapshot 16.2.98" } ]
[ { "msg_contents": "Ok, \n\nI think large objects are leaking memory because the large object functions\nin the backend use their own GlobalMemoryContext (called Filesystem), which\n(according to a quick grep) is never freed.\n\nSupposing this is true and I ensure that the large object subsystem always\nuses the current memory context for it's memory allocations.\n\nWhat might go wrong? (Or why did the designers decide to use a\nGlobalMemoryContext for large objects?).\n\nI simple don't understand why one would create a special memory context\nfor large objects without some special reason. \nOr should I just try it and see is anything breaks?\n\nThanks for any comments.\nMaurice\n\n\n", "msg_date": "Mon, 16 Feb 1998 16:54:05 +0100", "msg_from": "\"Maurice Gittens\" <[email protected]>", "msg_from_op": true, "msg_subject": "Memory leaks for large objects" }, { "msg_contents": "On Mon, 16 Feb 1998, Maurice Gittens wrote:\n\n> Ok, \n> \n> I think large objects are leaking memory because the large object functions\n> in the backend use their own GlobalMemoryContext (called Filesystem), which\n> (according to a quick grep) is never freed.\n> \n> Supposing this is true and I ensure that the large object subsystem always\n> uses the current memory context for it's memory allocations.\n> \n> What might go wrong? (Or why did the designers decide to use a\n> GlobalMemoryContext for large objects?).\n> \n> I simple don't understand why one would create a special memory context\n> for large objects without some special reason. \n> Or should I just try it and see is anything breaks?\n\nI was wondering the same thing when I was looking at that part of the code\na couple of months back. It would be interesting to see if anything did\nbreak.\n\n-- \nPeter T Mount [email protected] or [email protected]\nMain Homepage: http://www.demon.co.uk/finder\nWork Homepage: http://www.maidstone.gov.uk Work EMail: [email protected]\n\n", "msg_date": "Tue, 17 Feb 1998 21:10:17 +0000 (GMT)", "msg_from": "Peter T Mount <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Memory leaks for large objects" }, { "msg_contents": "Peter T Mount wrote:\n\n> On Mon, 16 Feb 1998, Maurice Gittens wrote:\n>\n> > Ok,\n> >\n> > I think large objects are leaking memory because the large object functions\n> > in the backend use their own GlobalMemoryContext (called Filesystem), which\n> > (according to a quick grep) is never freed.\n> >\n> > Supposing this is true and I ensure that the large object subsystem always\n> > uses the current memory context for it's memory allocations.\n> >\n> > What might go wrong? (Or why did the designers decide to use a\n> > GlobalMemoryContext for large objects?).\n> >\n> > I simple don't understand why one would create a special memory context\n> > for large objects without some special reason.\n> > Or should I just try it and see is anything breaks?\n>\n> I was wondering the same thing when I was looking at that part of the code\n> a couple of months back. It would be interesting to see if anything did\n> break.\n\nDoes the large object I/O persist across transactions? If so, then storage would\nneed to be outside of the usual context, which is reset after every transaction.\nIs there a place where the large object context could be freed, but is not at\nthe moment?\n\n - Tom\n\n", "msg_date": "Wed, 18 Feb 1998 06:54:18 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Memory leaks for large objects" } ]
[ { "msg_contents": "gcc-2.8.0 is know to be buggy. But that shouldn't bring you the glibc\nbug, should it?\n\nMichael\n\n--\nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n\n> -----Original Message-----\n> From:\tFrederick W. Reimer [SMTP:[email protected]]\n> Sent:\tMonday, February 16, 1998 12:00 PM\n> To:\tMeskes, Michael; [email protected];\n> [email protected]\n> Subject:\tRE: [HACKERS] Re: [PORTS] v6.3 release ToDo list and\n> supported p\n> \n> I have a \"custom\" slackware with at least the following updates:\n> \n> gcc-2.8.0\n> glibc-2.0.6\n> \n> As per the glibc upgrade notes out there on the web, I've cleared \n> (actually moved) the /usr/include header files before installing \n> glibc, and have reinstalled some \"standard\" extra libraries (such as \n> ncurses) after that.\n> \n> I get the 3 hours 59 minutes 60.00 seconds.\n> \n> I also get a bunch of other problems in other tests. Some are just \n> differences in the error message, but others are actual differences \n> in the float/double numbers produced. Others don't show an error and \n> give a number where one \"should\" have been produced. This is all \n> vauge now, but I will recompile the latest snapshot and give a \n> complete update maybe tonight, if not tomorrow.\n> \n> Could some of the other problems be due to gcc-2.8.0, which I'm not \n> sure is being used by anyone else for postgress (yet)? Should we \n> include the compiler used in the porting updates (I think so). For \n> instance, I like to use gcc on the Sparc/Solaris2.5.1 and HP-UX/10.20 \n> boxes I have because it makes porting things easier between those \n> boxes and my Linux systems. Isn't it possible that postgres compiles \n> fine using the vendor-supplied compilers, but would fail on gcc, or \n> visa versa?\n> \n> \n> - Fred\n> \n> \n> > I just tried this on Debian with glibc 2.0.6 and it works fine. That\n> > is my 6.3 postgresql release outputs 4 hours.\n> > \n> > Michael\n> > \n> > --\n> > Dr. Michael Meskes, Projekt-Manager \| topystem Systemhaus GmbH\n> > [email protected] \| Europark A2, Adenauerstr.\n> > 20 [email protected] \| 52146 Wuerselen Go\n> > SF49ers! Use Debian GNU/Linux! \| Tel: (+49) 2405/4670-44\n> > \n> > > ----------\n> > > From: \tThomas G. Lockhart[SMTP:[email protected]]\n> > > Sent: \tFreitag, 13. Februar 1998 17:57\n> > > To: \[email protected]; Bruce Momjian\n> > > Cc: \[email protected]\n> > > Subject: \t[HACKERS] Re: [PORTS] v6.3 release ToDo list and\n> > > supported ports\n> > > \n> > > > Is the ix86/linux \"confirmed\" port using libc 5 or glibc 2.x\n> (aka\n> > > > libc6)? As most major distributions are going to release all\n> new\n> > > > versions with glibc I think it's prudent to test on both\n> \"platforms\"\n> > > > independently.\n> > > \n> > > Yes, that is a good point. There is a known bug in the glibc2 math\n> > > library which breaks the date/time routines:\n> > > \n> > > select '4 hours'::timespan;\n> > > ---------\n> > > @ 4 hours\n> > > (1 row)\n> > > \n> > > comes out instead as\n> > > tgl=> select '4 hours'::timespan;\n> > > ----------------------------\n> > > @ 3 hours 59 mins 60.00 secs\n> > > (1 row)\n> > > \n> > > Oliver was working on patches. Lost his e-mail message; is Oliver\n> > > still\n> > > here? I've got a RH5.0 linux system at work now, and the shipped\n> > > Postgres\n> > > installation has this problem. I'd like to see it fixed...\n> > > \n> > > Bruce, can you add this to the v6.3 ToDo (assuming you agree to\n> take\n> > > it\n> > > on :)\n> > > \n> > > - Tom\n> > > \n> > > \n> > > \n> > \n", "msg_date": "Mon, 16 Feb 1998 17:00:13 +0100", "msg_from": "\"Meskes, Michael\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] Re: [PORTS] v6.3 release ToDo list and supported p" } ]
[ { "msg_contents": "> Did we ever go anywhere on this one? I just got into mp3's here\n> at home...something like this might be cool for that, or for large images\n> libraries?\n\nI just didn't see the extraordinary usefulness of it myself. I figured\na named file would reside in a directory as easily as we could move it\nin and out of one big indexed file.\n\nThere's also the 64k size limit per record, which is easily gotten\naround, but again, the original file has no such restrictions.\n\nIf you tell me it'll be useful I'll do it up in a function and send it\nto you. But I don't see the compelling need for it yet.\n\n", "msg_date": "Mon, 16 Feb 1998 10:49:54 -0600", "msg_from": "Alton Moore III <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: development" } ]
[ { "msg_contents": "\nI get the same problem on my alpha running Digital Unix 4.0 with the\nlatest snapshot.. I'll start looking into it as well..\n\n--brett\n\n--\nHi All,\n\nI am looking into the problem with initdb running on a linux alpha. For \nsome reason, when the postgres backend is running in -boot mode, when it \ngets to parsing the last 10 or so directives in lib/local...bki, it bombs \nout, claiming that function mkoidname(opaque, opaque) is not found.\n", "msg_date": "Mon, 16 Feb 1998 15:56:15 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "mkoidname(opaque, opaque)" } ]
[ { "msg_contents": "\nis the snapshot going to be patched to allow 'user' as a valid table\nname?\n", "msg_date": "Mon, 16 Feb 1998 16:11:10 -0800", "msg_from": "Brett McCormick <[email protected]>", "msg_from_op": true, "msg_subject": "'user' as reserved word & snapshot" }, { "msg_contents": "> is the snapshot going to be patched to allow 'user' as a valid table\n> name?\n\nYes, also a few other reserved words. Waiting to see if others come up\nin the next few days...\n\n - Tom\n\n", "msg_date": "Tue, 17 Feb 1998 15:18:36 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] 'user' as reserved word & snapshot" } ]
[ { "msg_contents": "\nThese are some tests that configure can perform, that look like they might\nhelp somewhat...\n\n\n - Macro: AC_C_LONG_DOUBLE\n If the C compiler supports the `long double' type, define\n `HAVE_LONG_DOUBLE'. Some C compilers that do not define\n `__STDC__' do support the `long double' type; some compilers that\n define `__STDC__' do not support `long double'.\n\n - Macro: AC_CHECK_SIZEOF (TYPE [, CROSS-SIZE])\n Define `SIZEOF_UCTYPE' to be the size in bytes of the C (or C++)\n builtin type TYPE, e.g. `int' or `char '. If `type' is unknown\n to the compiler, it gets a size of 0. UCTYPE is TYPE, with\n lowercase converted to uppercase, spaces changed to underscores,\n and asterisks changed to `P'. If cross-compiling, the value\n CROSS-SIZE is used if given, otherwise `configure' exits with an\n error message.\n\n For example, the call\n AC_CHECK_SIZEOF(int )\n\n defines `SIZEOF_INT_P' to be 8 on DEC Alpha AXP systems.\n\n - Macro: AC_INT_16_BITS\n If the C type `int' is 16 bits wide, define `INT_16_BITS'. This\n macro is obsolete; it is more general to use\n `AC_CHECK_SIZEOF(int)' instead.\n\n - Macro: AC_LONG_64_BITS\n If the C type `long int' is 64 bits wide, define `LONG_64_BITS'.\n This macro is obsolete; it is more general to use\n `AC_CHECK_SIZEOF(long)' instead.\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 16 Feb 1998 21:05:54 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Do these help for the Alpha ports...?" } ]
[ { "msg_contents": "Michael Hirohama wrote:\n> \n> [...]\n> >SQL> select * from a where (x,z) >= ANY (select y, l from b);\n> >select * from a where (x,z) >= ANY (select y, l from b)\n> > \n> >ERROR at line 1:\n> >ORA-00920: invalid relational operator\n> >\n> >: only '=' and '<>' are allowed if there are more than one\n> >expression on the left side of clause with subselect...\n> >\n> >Is this in standard or Oracle' preference ?\n> \n> This is a mathematical law.\n> \n> Vectors cannot be strictly ordered. \"Partial orderings\" are possible.\n> \n> Let A be (1, 2)\n> Let B be (4, 7)\n> Let C be (3, 5)\n> Let D be (5, 10)\n> \n> A is smallest; D is largest; how do B and C relate?\n\nI understand this. And this is how it works currently:\n\nselect from tab where (A,B) >= ANY (select X, Y from tab2);\n\nmeans: select tuples where A >= X _and_ B >= Y for some tuple from tab2.\n ^^^^^\n 'AND' is used for all Op-s except for '<>' when 'OR' is used.\n\nQuestion is \"should we drop this feature (?) or leave it as is ?\"\n\nComments ?\n\nVadim\n", "msg_date": "Tue, 17 Feb 1998 10:31:16 +0700", "msg_from": "\"Vadim B. Mikheev\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Subselects open issue Nr. NEW" }, { "msg_contents": "At 10:31 +0700 2/17/98, Vadim B. Mikheev wrote:\n[...]\n>\n>I understand this. And this is how it works currently:\n>\n>select * from tab where (A,B) >= ANY (select X, Y from tab2);\n>\n>means: select tuples where A >= X _and_ B >= Y for some tuple from tab2.\n> ^^^^^\n> 'AND' is used for all Op-s except for '<>' when 'OR' is used.\n>\n>Question is \"should we drop this feature (?) or leave it as is ?\"\n>\n>Comments ?\n>\n>Vadim\n\nI recommend dropping this feature and only supporing =ANY and <>ANY.\nSupporing the relational operators cannot be optimized in the general case.\n\nIf I needed to perform the <=ANY query efficiently, I might use a\nfunctional index on 'A' and 'B' using an R-tree method.\n\nRegards,\n\n--\nMichael Hirohama <[email protected]>\n\n\n", "msg_date": "Mon, 16 Feb 1998 20:14:40 -0800", "msg_from": "Michael Hirohama <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Subselects open issue Nr. NEW" }, { "msg_contents": "Vadim B. Mikheev writes:\n> I understand this. And this is how it works currently:\n> \n> select * from tab where (A,B) >= ANY (select X, Y from tab2);\n> \n> means: select tuples where A >= X _and_ B >= Y for some tuple from tab2.\n> ^^^^^\n> 'AND' is used for all Op-s except for '<>' when 'OR' is used.\n> \n> Question is \"should we drop this feature (?) or leave it as is ?\"\n\nIf we have it, please keep it. Yes, it's not really standard but with some\ndocu that doesn't hurt. It's still better than an error message IMO. Also\nthis does make sense as a shortcut version of asking both to be greater or\nequal. Did you ever try to do something like that in Oracle? I would have\nloved this feature.\n\nMichael\n\n-- \nDr. Michael Meskes, Project-Manager \| topsystem Systemhaus GmbH\[email protected] \| Europark A2, Adenauerstr. 20\[email protected] \| 52146 Wuerselen\nGo SF49ers! Go Rhein Fire! \| Tel: (+49) 2405/4670-44\nUse Debian GNU/Linux! \| Fax: (+49) 2405/4670-10\n", "msg_date": "Tue, 17 Feb 1998 15:50:00 +0100 (CET)", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Subselects open issue Nr. NEW" } ]
[ { "msg_contents": "\nMoved to [email protected]\n\n\nOn Mon, 16 Feb 1998, Bruce Momjian wrote:\n\n> Sure doesn't, but I think it is the same problem with pg_database. The\n> problem is that there is no permission control over tables finer than\n> insert/delete/update permissions. That is, if you say someone can add\n> and delete views by inserting/deleting in pg_rewrite, there is no way to\n> prevent him from deleting other people's views, at least as it is now\n> set up.\n> \n> With pg_database, people used to be able to create databases, but not\n> delete them, that is, we gave them insert permission, but no delete\n> permission. But that meant people could create databases, but not\n> delete them. In 6.3, I enabled normal users with create database\n> permission to delete databases, but now they can delete anyone's\n> database, or at least the record in pg_database if they directly modify\n> pg_database rather than trying the 'drop database' command.\n> \n> No way around that as the current system is configured, and I am not\n> sure how to fix this.\n> \n> With pg_rewrite, the problem is harder because we don't have a 'is able\n> to create views' permission that we can give permission to certain\n> people.\n\n\tEvery database/table/view has an owner associated with them, no? \nCan a check not be put in that checks the owner of the database/table/view\nagainst who is trying to delete it, with pg_superuser having \"override\" \nprivileges? Obviously, the security level of the database itself comes\ninto play, but we have most (if not all?) of those in place and settable\nby the administrator... \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 16 Feb 1998 23:42:37 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [QUESTIONS] Trouble creating view\\" }, { "msg_contents": "> > With pg_rewrite, the problem is harder because we don't have a 'is able\n> > to create views' permission that we can give permission to certain\n> > people.\n> \n> \tEvery database/table/view has an owner associated with them, no? \n> Can a check not be put in that checks the owner of the database/table/view\n> against who is trying to delete it, with pg_superuser having \"override\" \n> privileges? Obviously, the security level of the database itself comes\n> into play, but we have most (if not all?) of those in place and settable\n> by the administrator... \n\nI think the simple answer is no. We don't look at the DATA in the table\nto determine if the person can delete a particular row, and that is what\nwe would require. The way the 'drop database' does it is to check the\nowner of the database, then issue the deletes only if they match or if\nit is the superuser. However, there is not restriction on the user\naccessing pg_database directly. We would need some kind of access\nchecking filter inside the executor that would look at each row about to\nbe modified, and determine if it is legal.\n\n-- \nBruce Momjian\[email protected]\n", "msg_date": "Tue, 17 Feb 1998 00:08:35 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [QUESTIONS] Trouble creating view\\" } ]

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.