threads
listlengths 1
2.99k
|
|---|
[
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nI've gotten 6.3 initdb to run to a successful completion on my\nAlpha running OSF/1 V3.2c. Forget the change that I sent in\nearlier. While I still think that there's something funky with\nthat code, it doesn not need to be modifed. Actually, the\nmodifications are miniscule. The only files that need to be\nchanged are backend/main/main.c and template/alpha.\n\nThe real trick is to add -Dalpha to the CFLAGS setting. The\nchanges to main.c are only to add some extra includes to support\nsome code that's suddenly being used.\n\nThe #define ASSEMBLER is to prevent most of the code of\nsys/proc.h from being included, as it ends up conflicting with\nsome of the postgresql definitions. This may or may not work on\nother versions of Digital Unix.\n\nAs far as I'm concerned, this is a hack fix. There's still some\nunderlying 32/64 bit assumtions that this is masking. Perhaps\nI'll make that my pet project.\n\nHere are the diffs for the two files that I modified:\n\n*** backend/main/main.c\tMon Mar 16 15:53:26 1998\n- --- backend/main/main.c.orig\tMon Mar 16 16:05:07 1998\n***************\n*** 15,28 ****\n #include <string.h>\n #include <unistd.h>\n\n- - #ifdef alpha\n- - #include <sys/sysinfo.h>\n- - #include <machine/hal_sysinfo.h>\n- - #define ASSEMBLER\n- - #include <sys/proc.h>\n- - #undef ASSEMBLER\n- - #endif\n- -\n #include \"postgres.h\"\n #ifdef USE_LOCALE\n #include <locale.h>\n- --- 15,20 ----\n\n*** template/alpha\tMon Mar 16 16:06:08 1998\n- --- template/alpha.orig\tMon Mar 16 16:11:25 1998\n***************\n*** 5,11 ****\n # This is defined here because a bunch of clients include tmp/c.h,\n # which is where the work is done on HP-UX. It only affects the\n # backend on Ultrix and OSF/1.\n! CFLAGS:-DNOFIXADE -Dalpha\n SHARED_LIB:\n ALL:\n SRCH_INC:\n- --- 5,11 ----\n # This is defined here because a bunch of clients include tmp/c.h,\n # which is where the work is done on HP-UX. It only affects the\n # backend on Ultrix and OSF/1.\n! CFLAGS:-DNOFIXADE\n SHARED_LIB:\n ALL:\n SRCH_INC:\n\n\n\n- --\nDwayne Bailey + WHAT is your name? Sir Galahad\nMIKA Systems, Bingham Farms, MI + WHAT is your quest? I Seek the Holy Grail\[email protected] + What is your favorite color?\nhttp://www.mika.com/~dwayne + Blue ... no, Yelloooooooooooooooooow\n finger [email protected] for PGP Public Key\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQB1AwUBNQ2YLaA2uleK7maRAQG50gMAne7myS15kxEjkC95WexnZKxBobKGFG8L\nNRNv0u7JeNSuDTHR5xf4UDSiacGLXlDvMwhUk83W+GnUdwACsQuX1ASfVfc2mCAP\nIN6HiMK+DQuzpYfrf4gT3sdymQGyPl00\n=F/Mt\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "Mon, 16 Mar 1998 16:22:46 -0500 (EST)",
"msg_from": "Dwayne Bailey <[email protected]>",
"msg_from_op": true,
"msg_subject": "Alpha initdb fixed!"
},
{
"msg_contents": "On Mon, 16 Mar 1998, Dwayne Bailey wrote:\n\n>I've gotten 6.3 initdb to run to a successful completion on my\n>Alpha running OSF/1 V3.2c. Forget the change that I sent in\n>earlier. While I still think that there's something funky with\n>that code, it doesn not need to be modifed. Actually, the\n>modifications are miniscule. The only files that need to be\n>changed are backend/main/main.c and template/alpha.\n>\n>The real trick is to add -Dalpha to the CFLAGS setting. The\n>changes to main.c are only to add some extra includes to support\n>some code that's suddenly being used.\n>\n>The #define ASSEMBLER is to prevent most of the code of\n>sys/proc.h from being included, as it ends up conflicting with\n>some of the postgresql definitions. This may or may not work on\n>other versions of Digital Unix.\n\nI'll try it immediately, but I have a suggestion. On my DU 3.2c system, cc\ndefines automatically the symbols \"__osf__\" and \"__alpha\", and gcc defines\n\"__osf__\", \"__alpha\" and \"__alpha__\". I think it would be easier to change\nevery \"#ifdef alpha\" to \"#ifdef __alpha\", and stop worrying about it in\nthe Makefiles.\n\nCan any of the linux-alpha folks try out which symbols does the compiler\ndefine? And someone who has DU 4.0x installed?\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": "Tue, 17 Mar 1998 09:42:07 +0100 (MET)",
"msg_from": "\"Pedro J. Lobo\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Alpha initdb fixed!"
},
{
"msg_contents": "On Tue, 17 Mar 1998, Pedro J. Lobo wrote:\n\n>On Mon, 16 Mar 1998, Dwayne Bailey wrote:\n>\n>>I've gotten 6.3 initdb to run to a successful completion on my\n>>Alpha running OSF/1 V3.2c. Forget the change that I sent in\n>>earlier. While I still think that there's something funky with\n>>that code, it doesn not need to be modifed. Actually, the\n>>modifications are miniscule. The only files that need to be\n>>changed are backend/main/main.c and template/alpha.\n>>\n>>The real trick is to add -Dalpha to the CFLAGS setting. The\n>>changes to main.c are only to add some extra includes to support\n>>some code that's suddenly being used.\n>>\n>>The #define ASSEMBLER is to prevent most of the code of\n>>sys/proc.h from being included, as it ends up conflicting with\n>>some of the postgresql definitions. This may or may not work on\n>>other versions of Digital Unix.\n>\n>I'll try it immediately, but I have a suggestion. On my DU 3.2c system, cc\n>defines automatically the symbols \"__osf__\" and \"__alpha\", and gcc defines\n>\"__osf__\", \"__alpha\" and \"__alpha__\". I think it would be easier to change\n>every \"#ifdef alpha\" to \"#ifdef __alpha\", and stop worrying about it in\n>the Makefiles.\n\nI've just tried it, and it works partially. The initdb works fine, so I've\ntried to run the regression tests. Here is the output:\n\n==============================================================\nboolean .. ok\nchar .. ok\nchar2 .. ok\nchar4 .. ok\nchar8 .. ok\nchar16 .. ok\nvarchar .. ok\ntext .. ok\nstrings .. ok\nint2 .. failed\nint4 .. failed\noid .. ok\noidint2 .. failed\noidint4 .. failed\noidname .. failed\n[...]\n==============================================================\n\nAll tests after oid fail, because the postmaster dies with this message:\n\n========================\n[...]\nERROR: pg_atoi: error reading \"123456\": Result too large\nERROR: pg_atoi: error in \"asdfasd\": can't parse \"asdfasd\"\nsemget: No space left on device\nThis type of error is usually caused by improper\nshared memory or System V IPC semaphore configuration.\nSee the FAQ for more detailed information\nFATAL 1: AttachSLockMemory: could not attach segment\n=========================\n\nRunning the regression test after starting the postmaster with \"-d 2\"\ngives:\n\n========================\n[...]\n/usr/local/pgsql.beta/bin/postmaster child[0]:\nexecv(/usr/local/pgsql.beta/bin/postgres, -p, -d2, -P4, -F, -e, -B, 256, -v 65536, regression, )\n/usr/local/pgsql.beta/bin/postmaster: BackendStartup: pid 6011 user pgbeta\ndb regression socket 4\nFindBackend: found \"/usr/local/pgsql.beta/bin/postgres\" using argv[0]\nbinding ShmemCreate(key=0, size=2414376)\nsemget: No space left on device\nThis type of error is usually caused by improper\nshared memory or System V IPC semaphore configuration.\nSee the FAQ for more detailed information\n ---debug info---\n Quiet = f\n Noversion = f\n timings = f\n dates = European\n bufsize = 256\n sortmem = 512\n query echo = f\n DatabaseName = [regression]\n ----------------\n\n InitPostgres()..\n/usr/local/pgsql.beta/bin/postmaster: reaping dead processes...\n/usr/local/pgsql.beta/bin/postmaster: CleanupProc: pid 6011 exited with \nstatus 768\n/usr/local/pgsql.beta/bin/postmaster: CleanupProc: reinitializing shared\nmemory and semaphores\nFATAL 1: AttachSLockMemory: could not attach segment\n===========================\n\nI am using these options (they worked fine with 6.2.1 and 6.2.1p6):\n\npostmaster -d 2 -o '-F -e' -B 256 -D/usr/local/pgsql.beta/data\n\nAlso, after the postmaster dies I have to manually remove (using ipcrm) 15\nsemaphores and one shared memory area. Since there is one more semaphore\nowned by root, there are 16 semaphores allocated when the postmaster dies.\nI have looked at my system configuration, and that's the system limit. I\ncan raise it to, say, 32, but the 6.2.1 system worked fine with my current\nconfiguration. I suspect that the postmaster is allocating semaphores and\nnever releasing them.\n\nAny hints?\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": "Tue, 17 Mar 1998 11:40:01 +0100 (MET)",
"msg_from": "\"Pedro J. Lobo\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Alpha initdb fixed!"
},
{
"msg_contents": "On Tue, 17 Mar 1998, Pedro J. Lobo wrote:\n\n>On Tue, 17 Mar 1998, Pedro J. Lobo wrote:\n>\n>I've just tried it, and it works partially. The initdb works fine, so I've\n>tried to run the regression tests. Here is the output:\n>\n>==============================================================\n>boolean .. ok\n>char .. ok\n>char2 .. ok\n>char4 .. ok\n>char8 .. ok\n>char16 .. ok\n>varchar .. ok\n>text .. ok\n>strings .. ok\n>int2 .. failed\n>int4 .. failed\n>oid .. ok\n>oidint2 .. failed\n>oidint4 .. failed\n>oidname .. failed\n>[...]\n>==============================================================\n\nI've done more tests. The problem is that if you start the postmaster\nwithout the '-p' option and without assigning a value to the PGPORT\nenvironment variable, then all the ipc stuff is messed up. No shared\nmemory regions are created, and the semaphores are created but never\nfreed. When a port number is specified, the sempahores (and the shared\nmemory regions) have a 'key' value that contains the port number. Without\nport number, there is no shared memory and the sempahores have 0 as the\nkey value.\n\nI don't know if this behaviour is due to the use of a non-standard port\n(5440), but since it's been specified in configure (--with-pgport=5440) it\nshould work. shouldn't it?\n\nThese are the regression tests when a port number is specified (note that \nyou *must* assign a value to PGPORT before running the tests):\n\n\n===============================================================\nboolean .. ok\nchar .. ok\nchar2 .. ok\nchar4 .. ok\nchar8 .. ok\nchar16 .. ok\nvarchar .. ok\ntext .. ok\nstrings .. ok\nint2 .. failed\nint4 .. failed\noid .. ok\noidint2 .. failed\noidint4 .. failed\noidname .. ok\nfloat4 .. ok\nfloat8 .. failed\nnumerology .. ok\npoint .. ok\nlseg .. ok\nbox .. ok\npath .. ok\npolygon .. ok\ncircle .. ok\ngeometry .. failed\ntimespan .. ok\ndatetime .. failed\nreltime .. ok\nabstime .. failed\ntinterval .. failed\nhorology .. failed\ncomments .. ok\ncreate_function_1 .. ok\ncreate_type .. ok\ncreate_table .. ok\ncreate_function_2 .. ok\nconstraints .. ok\ntriggers .. ok\ncopy .. ok\ncreate_misc .. ok\ncreate_aggregate .. ok\ncreate_operator .. ok\ncreate_view .. ok\ncreate_index .. ok\nsanity_check .. ok\nerrors .. ok\nselect .. ok\nselect_into .. ok\nselect_distinct .. ok\nselect_distinct_on .. ok\nsubselect .. ok\naggregates .. ok\ntransactions .. ok\nrandom .. failed\nportals .. ok\nmisc .. ok\narrays .. ok\nbtree_index .. ok\nhash_index .. ok\nselect_views .. ok\nalter_table .. ok\nportals_p2 .. ok\n==========================================\n\nSome of them fail (most notably int2, int4 and float8), but anyway it's\nbetter than before :-)\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": "Tue, 17 Mar 1998 13:06:59 +0100 (MET)",
"msg_from": "\"Pedro J. Lobo\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nOn Tue, 17 Mar 1998, Pedro J. Lobo wrote:\n\n> \n> I've done more tests. The problem is that if you start the postmaster\n> without the '-p' option and without assigning a value to the PGPORT\n> environment variable, then all the ipc stuff is messed up. No shared\n> memory regions are created, and the semaphores are created but never\n> freed. When a port number is specified, the sempahores (and the shared\n> memory regions) have a 'key' value that contains the port number. Without\n> port number, there is no shared memory and the sempahores have 0 as the\n> key value.\n> \n> I don't know if this behaviour is due to the use of a non-standard port\n> (5440), but since it's been specified in configure (--with-pgport=5440) it\n> should work. shouldn't it?\n\nI got the same results that you did. I was planning on\ninvestigating this morning, but it looks like you beat me to it.\nI ALSO built 6.3 with a non-standard port, so that I could keep\nmy current database live while I work on this.\n\nI'll try your suggestion, but I'll also try rebuilding using the\nstandard port, to see if it makes any difference.\n\nRe: your suggestion to use __alpha and not worry about the\nmakefile, I'm a little uncomfortable with that. DEC's cc will\nactually output different symbols, depending on the use of the\n- -std flag. I'd rather have something that we have explicit\ncontrol over, rather than relying on the compiler like this. I'm\nnot violently opposed to useing __alpha or anything, it's just a\npreference against it.\n\n- --\nDwayne Bailey + WHAT is your name? Sir Galahad\nMIKA Systems, Bingham Farms, MI + WHAT is your quest? I Seek the Holy Grail\[email protected] + What is your favorite color?\nhttp://www.mika.com/~dwayne + Blue ... no, Yelloooooooooooooooooow\n finger [email protected] for PGP Public Key\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQB1AwUBNQ5svqA2uleK7maRAQHJ1gL/ULW54HyDSjLZv++z2j1taxfdchgpPAL1\n9WDrJAdPHmEjm1iAZfQT6gqIpwZ70fp2VpRneqZZyoUw1ZCHE3ufcDHz29t43Rbb\nQJL6lDl99J0R3ZH6rA8JHhd6Mn0uV9YM\n=eZ4b\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "Tue, 17 Mar 1998 07:29:42 -0500 (EST)",
"msg_from": "Dwayne Bailey <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "On Tue, 17 Mar 1998, Pedro J. Lobo wrote:\n\n> I don't know if this behaviour is due to the use of a non-standard port\n> (5440), but since it's been specified in configure (--with-pgport=5440) it\n> should work. shouldn't it?\n\n\tYes, and there was a fix submitted and applied for this...its miss\ndefined in configure...\n\n\n",
"msg_date": "Tue, 17 Mar 1998 08:17:58 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "> I've done more tests. The problem is that if you start the postmaster\n> without the '-p' option and without assigning a value to the PGPORT\n> environment variable, then all the ipc stuff is messed up. No shared\n> memory regions are created, and the semaphores are created but never\n> freed. When a port number is specified, the sempahores (and the shared\n> memory regions) have a 'key' value that contains the port number. Without\n> port number, there is no shared memory and the sempahores have 0 as the\n> key value.\n> \n> I don't know if this behaviour is due to the use of a non-standard port\n> (5440), but since it's been specified in configure (--with-pgport=5440) it\n> should work. shouldn't it?\n> \n> These are the regression tests when a port number is specified (note that \n> you *must* assign a value to PGPORT before running the tests):\n\nLet's get a patch for this alpha fix. Not sure about the pgport problem.\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 09:52:24 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nOn Tue, 17 Mar 1998, Bruce Momjian wrote:\n> \n> Let's get a patch for this alpha fix. Not sure about the pgport problem.\n> \n\nI included a diff in my original report. I can resend it to the\npatches list, if required. However, I would prefer to hear that\nsomebody tested it on DU 4.0. Thus far, AFAIK, only 3.2 has been\ntested.\n\nI'm confident that the patched template/alpha file will be fine,\nbut the corresponding changes to backend/main/main.c leave me\nless comfortable. There's a #define ASSEMBLER there to prevent\nthe loading of wholesale portions of sys/proc.h. I'd like to\nknow if that works as expected on other versions of DU.\n\nThe pgport problem has been identified as a problem with\nconfigure, which had been previously reported. (A report that I\nmust have missed.)\n\n- --\nDwayne Bailey + WHAT is your name? Sir Galahad\nMIKA Systems, Bingham Farms, MI + WHAT is your quest? I Seek the Holy Grail\[email protected] + What is your favorite color?\nhttp://www.mika.com/~dwayne + Blue ... no, Yelloooooooooooooooooow\n finger [email protected] for PGP Public Key\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQB1AwUBNQ6hI6A2uleK7maRAQEZcQMAgBQGn9smBHdf1aIGGz5a22qVSSOE4wBe\nlpvCCvWzc0X09Qa1I2xdr4+Tln5gp1iWUQfi/0jaADuI/RgzRDABTcTjBt2vXY8S\n7z/GKfxsXWie54LyrviDAxqfAGlpI16z\n=rCSI\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "Tue, 17 Mar 1998 11:13:13 -0500 (EST)",
"msg_from": "Dwayne Bailey <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "On Tue, 17 Mar 1998, Bruce Momjian wrote:\n\n> > I've done more tests. The problem is that if you start the postmaster\n> > without the '-p' option and without assigning a value to the PGPORT\n> > environment variable, then all the ipc stuff is messed up. No shared\n> > memory regions are created, and the semaphores are created but never\n> > freed. When a port number is specified, the sempahores (and the shared\n> > memory regions) have a 'key' value that contains the port number. Without\n> > port number, there is no shared memory and the sempahores have 0 as the\n> > key value.\n> > \n> > I don't know if this behaviour is due to the use of a non-standard port\n> > (5440), but since it's been specified in configure (--with-pgport=5440) it\n> > should work. shouldn't it?\n> > \n> > These are the regression tests when a port number is specified (note that \n> > you *must* assign a value to PGPORT before running the tests):\n> \n> Let's get a patch for this alpha fix. Not sure about the pgport problem.\n\n\tThe pgport problem, I *think*, is the one that was configure\nrelated, where the port is set wrong by default.\n\n\tTry this:\n\nIndex: pgsql/src/configure\n===================================================================\nRCS file: /usr/local/cvsroot/pgsql/src/configure,v\nretrieving revision 1.132\nretrieving revision 1.134\ndiff -r1.132 -r1.134\n811c811\n< #define DEF_PGPORT \"${DEF_PGPORT}\"\n---\n> #define DEF_PGPORT \"${withval}\"\n\n\n",
"msg_date": "Tue, 17 Mar 1998 11:33:28 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "On Tue, 17 Mar 1998, Pedro J. Lobo wrote:\n\n> I'll try it immediately, but I have a suggestion. On my DU 3.2c system, cc\n> defines automatically the symbols \"__osf__\" and \"__alpha\", and gcc defines\n> \"__osf__\", \"__alpha\" and \"__alpha__\". I think it would be easier to change\n> every \"#ifdef alpha\" to \"#ifdef __alpha\", and stop worrying about it in\n> the Makefiles.\n> \n> Can any of the linux-alpha folks try out which symbols does the compiler\n> define? And someone who has DU 4.0x installed?\n\n\tLinux/Alpha provides the following useful/relavent symbols:\nlinux\n__alpha\n__alpha__\n__linux\n__linux__\n\n\tI had gone through the pgsql 6.2.1 source trying to fix/replace\nall instances of 'linuxalpha' and such used as defines with '(defined\n__alpha__) && (defined __linux__)'. But I hit a few snags in testing (i.e.\nlack of time), and by the time I got things about sorted out, 6.3 came out\nand changed so much I need to go through again and do it all anew. The\nbaisc problem it looks like you hit as well, is that non-standard define\nnames were used, and then never included in the platform specific defines.\nThis was the reason Linux/Alpha couldn't even get initdb to run (probably\nsame for you). Of course, the regression tests are still not perfect, and\nthere is a good deal of cleanup on the Linux/Alpha end of things as well.\nIt will be a while, but things are moving.\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",
"msg_date": "Tue, 17 Mar 1998 20:02:36 -0600 (CST)",
"msg_from": "Ryan Kirkpatrick <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Alpha initdb fixed!"
},
{
"msg_contents": "> These are the regression tests when a port number is specified (note that\n> you *must* assign a value to PGPORT before running the tests):\n> \n> ===============================================================\n> boolean .. ok\n> char .. ok\n> char2 .. ok\n> char4 .. ok\n> char8 .. ok\n> char16 .. ok\n> varchar .. ok\n> text .. ok\n> strings .. ok\n> int2 .. failed\n> int4 .. failed\n> oid .. ok\n> oidint2 .. failed\n> oidint4 .. failed\n> oidname .. ok\n> float4 .. ok\n> float8 .. failed\n> numerology .. ok\n> point .. ok\n> lseg .. ok\n> box .. ok\n> path .. ok\n> polygon .. ok\n> circle .. ok\n> geometry .. failed\n> timespan .. ok\n> datetime .. failed\n> reltime .. ok\n> abstime .. failed\n> tinterval .. failed\n> horology .. failed\n> comments .. ok\n> create_function_1 .. ok\n> create_type .. ok\n> create_table .. ok\n> create_function_2 .. ok\n> constraints .. ok\n> triggers .. ok\n> copy .. ok\n> create_misc .. ok\n> create_aggregate .. ok\n> create_operator .. ok\n> create_view .. ok\n> create_index .. ok\n> sanity_check .. ok\n> errors .. ok\n> select .. ok\n> select_into .. ok\n> select_distinct .. ok\n> select_distinct_on .. ok\n> subselect .. ok\n> aggregates .. ok\n> transactions .. ok\n> random .. failed\n> portals .. ok\n> misc .. ok\n> arrays .. ok\n> btree_index .. ok\n> hash_index .. ok\n> select_views .. ok\n> alter_table .. ok\n> portals_p2 .. ok\n> ==========================================\n> \n> Some of them fail (most notably int2, int4 and float8), but anyway it's\n> better than before :-)\n\nOooh. I think you might have a running system now! Those int2, int4,\nfloat8 \"failures\" are probably just error message differences and are\nexpected. The date and time stuff may or may not be a problem, and the\ngeometry stuff is probably OK (rounding trouble in the math libraries).\n\nMake sure your date/time stuff looks OK, at least for simple tests; it\nmay be, for example, that your timezone database is just different for\ndates before 1960...\n\n - Tom\n",
"msg_date": "Wed, 18 Mar 1998 06:38:22 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "On Tue, 17 Mar 1998, Dwayne Bailey wrote:\n\n>Re: your suggestion to use __alpha and not worry about the\n>makefile, I'm a little uncomfortable with that. DEC's cc will\n>actually output different symbols, depending on the use of the\n>- -std flag. I'd rather have something that we have explicit\n>control over, rather than relying on the compiler like this. I'm\n>not violently opposed to useing __alpha or anything, it's just a\n>preference against it.\n\nHere's an extract from the DEC's cc man page:\n\n The following table shows which macros are defined for each of the -std\n flags.\n\n -----------------------------------------------\n Macro std0 std std1\n (default)\n -----------------------------------------------\n LANGUAGE_C yes no no\n __LANGUAGE_C__ yes yes yes\n unix yes no no\n __unix__ yes yes yes\n __osf__ yes yes yes\n __alpha yes yes yes\n SYSTYPE_BSD yes no no\n _SYSTYPE_BSD yes yes yes\n LANGUAGE_ASSEMBLY yes yes yes\n __LANGUAGE_ASSEMBLY__ yes yes yes\n -----------------------------------------------\n\nAs you can see, __alpha and __osf__ are always defined. However, I\nunderstand your point. If we define 'alpha' in the template file, we are\nprotected from mind-changing vendors that define __alpha in DU 3.2 and\n__alpha__ in DU 4.0 and alpha__ in DU 5.0 (just an example). From this\npoint of view, the current approach is better. And, it's always easier\n(and safer) to leave things untouched.\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": "Wed, 18 Mar 1998 12:00:47 +0100 (MET)",
"msg_from": "\"Pedro J. Lobo\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "On Wed, 18 Mar 1998, Pedro J. Lobo wrote:\n\n>On Tue, 17 Mar 1998, Dwayne Bailey wrote:\n>\n>>Re: your suggestion to use __alpha and not worry about the\n>>makefile, I'm a little uncomfortable with that. DEC's cc will\n>>actually output different symbols, depending on the use of the\n>>- -std flag. I'd rather have something that we have explicit\n>>control over, rather than relying on the compiler like this. I'm\n>>not violently opposed to useing __alpha or anything, it's just a\n>>preference against it.\n>\n\n[stuff deleted...]\n\n>As you can see, __alpha and __osf__ are always defined. However, I\n>understand your point. If we define 'alpha' in the template file, we are\n>protected from mind-changing vendors that define __alpha in DU 3.2 and\n>__alpha__ in DU 4.0 and alpha__ in DU 5.0 (just an example). From this\n>point of view, the current approach is better. And, it's always easier\n>(and safer) to leave things untouched.\n\nJust a thought: I think we should make a distinction between architecture\n(i.e., define 'alpha') and OS (i.e., define 'osf' or something like that),\nnow that linux runs also on alpha (and NT, if someone ever makes a port).\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": "Wed, 18 Mar 1998 12:08:12 +0100 (MET)",
"msg_from": "\"Pedro J. Lobo\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "On Wed, 18 Mar 1998, Thomas G. Lockhart wrote:\n\n>> hash_index .. ok\n>> select_views .. ok\n>> alter_table .. ok\n>> portals_p2 .. ok\n>> ==========================================\n>> \n>> Some of them fail (most notably int2, int4 and float8), but anyway it's\n>> better than before :-)\n>\n>Oooh. I think you might have a running system now! Those int2, int4,\n\nYes, it seems so.\n\n>float8 \"failures\" are probably just error message differences and are\n>expected.\n\nYes. For int2: Expected:\n! ERROR: pg_atoi: error reading \"100000\": Math result not representable\n\nGot:\n! ERROR: pg_atoi: error reading \"100000\": Result too large\n\nFor int4: Expected:\n! ERROR: pg_atoi: error reading \"1000000000000\": Math result not\nrepresentable\n\nGot:\n! ERROR: pg_atoi: error reading \"1000000000000\": Result too large\n\nThe same goes for oidint2 and oidint4.\n\nFor float8: Expected:\n! ERROR: Bad float8 input format -- overflow\n\nGot:\n! ERROR: floating point exception! The last floating point operation\neither exceeded legal ranges or was a divide by zero\n\nThis one was harmless, but there is another one: Expected:\n QUERY: SELECT '' AS bad, : (f.f1) from FLOAT8_TBL f;\n! bad| ?column?\n! ---+--------------------\n! | 1\n! |7.39912306090513e-16\n! | 0\n! | 0\n! | 1\n! (5 rows)\n!\n\nGot:\n QUERY: SELECT '' AS bad, : (f.f1) from FLOAT8_TBL f;\n! ERROR: exp() result is out of range\n\nCan someone comment on this?\n\n>The date and time stuff may or may not be a problem, and the\n>geometry stuff is probably OK (rounding trouble in the math libraries).\n\nYou are right on the geometry stuff. I am not sure about the date stuff.\nSome are differences of one second between the expected and the actual\nresults, some others are dates that appear displaced by 19 years (for\nexample, expecter year 1997 becomes 2016, expected 1957 becomes 1976...).\nThe diff output is very long on this.\n\n>Make sure your date/time stuff looks OK, at least for simple tests; it\n>may be, for example, that your timezone database is just different for\n>dates before 1960...\n\nThe date/time stuff has never worked completely right. And, if the problem\nlies in postgres, that's ok. Sooner or later it will be fixed. But if, as\nit seems, the problem lies in the timezone databases, we might be in big\ntrouble. Perhaps we could make a test, so we can say for sure \"your\ntimezone database is incorrect, go and ask your verdor for a patch\".\n\nAlso, the test fails form the random stuff:\n*** expected/random.out ma 29 abr 07:23:40 1997\n--- results/random.out ma 17 mar 03:51:57 1998\n***************\n*** 7,18 ****\n QUERY: SELECT count(*) FROM onek where oidrand(onek.oid, 10);\n count\n -----\n! 92\n (1 row)\n\n QUERY: SELECT count(*) FROM onek where oidrand(onek.oid, 10);\n count\n -----\n! 98\n (1 row)\n\n--- 7,18 ----\n QUERY: SELECT count(*) FROM onek where oidrand(onek.oid, 10);\n count\n -----\n! 95\n (1 row)\n\n QUERY: SELECT count(*) FROM onek where oidrand(onek.oid, 10);\n count\n -----\n! 88\n (1 row)\n\n\n----------------------\n\n\nYes, the results are different, but... aren't they random? O:-)\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": "Wed, 18 Mar 1998 12:29:24 +0100 (MET)",
"msg_from": "\"Pedro J. Lobo\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "> This one was harmless, but there is another one: Expected:\n> QUERY: SELECT '' AS bad, : (f.f1) from FLOAT8_TBL f;\n> ! bad| ?column?\n> ! ---+--------------------\n> ! | 1\n> ! |7.39912306090513e-16\n> ! | 0\n> ! | 0\n> ! | 1\n> ! (5 rows)\n> !\n> \n> Got:\n> QUERY: SELECT '' AS bad, : (f.f1) from FLOAT8_TBL f;\n> ! ERROR: exp() result is out of range\n> \n> Can someone comment on this?\n\nI think you are getting a better result than the regression test machine\ngets. That's good.\n\n> Some are differences of one second between the expected and the actual\n> results, some others are dates that appear displaced by 19 years (for\n> example, expecter year 1997 becomes 2016, expected 1957 becomes \n> 1976...). The diff output is very long on this.\n> The date/time stuff has never worked completely right. And, if the \n> problem lies in postgres, that's ok. Sooner or later it will be fixed.\n> But if, as it seems, the problem lies in the timezone databases, we \n> might be in big trouble. Perhaps we could make a test, so we can say \n> for sure \"your timezone database is incorrect, go and ask your verdor \n> for a patch\".\n\nNo, you still have date/time trouble, and it looks as though the\ntimezone stuff is not being set properly. By definition, it is a problem\nwith your machine, since the code works on several other platforms, and\nno, it isn't likely to get fixed eventually unless you pursue it, since\nit does work on the ~20 other OS/processor combinations listed as\nsupported platforms.\n\nOK, what I meant by \"timezone database\" trouble would have been sort of\nobvious in that only dates from times before computers existed would\nhave shown problems, and then usually 1 hour differences due to daylight\nsavings time settings. That is not what you are seeing.\n\nThe 19 year differences usually seem to come from mis-handling the\nHAVE_INT_TIMEZONE compile-time option. How is yours set? Try changing it\nin config.h and see if it helps.\n\n> Yes, the results are different, but... aren't they random? O:-)\n\nRight. OK for random to be different.\n\n - Tom\n",
"msg_date": "Wed, 18 Mar 1998 15:11:52 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "On Wed, 18 Mar 1998, Thomas G. Lockhart wrote:\n\n>> Got:\n>> QUERY: SELECT '' AS bad, : (f.f1) from FLOAT8_TBL f;\n>> ! ERROR: exp() result is out of range\n>> \n>> Can someone comment on this?\n>\n>I think you are getting a better result than the regression test machine\n>gets. That's good.\n\nOk.\n\n>> Some are differences of one second between the expected and the actual\n>> results, some others are dates that appear displaced by 19 years (for\n>> example, expecter year 1997 becomes 2016, expected 1957 becomes \n>> 1976...). The diff output is very long on this.\n>> The date/time stuff has never worked completely right. And, if the \n>> problem lies in postgres, that's ok. Sooner or later it will be fixed.\n>> But if, as it seems, the problem lies in the timezone databases, we \n>> might be in big trouble. Perhaps we could make a test, so we can say \n>> for sure \"your timezone database is incorrect, go and ask your verdor \n>> for a patch\".\n>\n>No, you still have date/time trouble, and it looks as though the\n>timezone stuff is not being set properly. By definition, it is a problem\n>with your machine, since the code works on several other platforms, and\n>no, it isn't likely to get fixed eventually unless you pursue it, since\n>it does work on the ~20 other OS/processor combinations listed as\n>supported platforms.\n\nYou have misinterpreted me. What I mean is that if the problem lies in\npostgres, we can hunt it and fix it, but if the problem lies in the\ntimezone libraries then it is out of our hands. Of course, the problem\nisn't going to vanish into nothingness by itself (although it would be\nvery nice, wouldn't it? :-)\n\n>OK, what I meant by \"timezone database\" trouble would have been sort of\n>obvious in that only dates from times before computers existed would\n>have shown problems, and then usually 1 hour differences due to daylight\n>savings time settings. That is not what you are seeing.\n>\n>The 19 year differences usually seem to come from mis-handling the\n>HAVE_INT_TIMEZONE compile-time option. How is yours set? Try changing it\n>in config.h and see if it helps.\n\nI am going to be offline for 4 days, until next Monday. I will dig into\nthat problem then.\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": "Wed, 18 Mar 1998 17:33:16 +0100 (MET)",
"msg_from": "\"Pedro J. Lobo\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Unix Domain Sockets error (was Re: [HACKERS] Alpha initdb fixed!)"
},
{
"msg_contents": "Thomas G. Lockhart wrote:\n\n> The 19 year differences usually seem to come from mis-handling the\n> HAVE_INT_TIMEZONE compile-time option. How is yours set? Try changing it\n> in config.h and see if it helps.\n>\n\nCouldn't this be tested for, just like there is a \"flex test\" which finds\nout if flex is ok or not?\nCan the configure script find out and add HAVE_INT_TIMEZONE if appropriate?\n\n/* m */\n\n\n",
"msg_date": "Wed, 18 Mar 1998 23:29:58 +0100",
"msg_from": "Mattias Kregert <[email protected]>",
"msg_from_op": false,
"msg_subject": "Timezone problems / HAVE_INT_TIMEZINE"
},
{
"msg_contents": "> Couldn't this be tested for, just like there is a \"flex test\" which \n> finds out if flex is ok or not? Can the configure script find out and \n> add HAVE_INT_TIMEZONE if appropriate?\n\nUh, it does a test already by trying to compile a program referencing a\nglobal integer variable called \"timezone\". Somehow a few systems will\ncompile that but don't really have a useful integer timezone\n(RH5.0/glibc2.0 is one of those).\n\nI'm wondering if we could change the sense of the test, to try instead\nto test for the presence of a timezone field in the tm structure? That\nmight fix the glibc2.0 port (assuming it still has problems at v2.0.7;\nhaven't tested recently) but I don't know which other ports might break.\n\nCan we experiment with this Marc?? Post-megapatch of course :)\n\n - Tom\n",
"msg_date": "Thu, 19 Mar 1998 03:10:20 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Timezone problems / HAVE_INT_TIMEZINE"
},
{
"msg_contents": "On Thu, 19 Mar 1998, Thomas G. Lockhart wrote:\n\n> > Couldn't this be tested for, just like there is a \"flex test\" which \n> > finds out if flex is ok or not? Can the configure script find out and \n> > add HAVE_INT_TIMEZONE if appropriate?\n> \n> Uh, it does a test already by trying to compile a program referencing a\n> global integer variable called \"timezone\". Somehow a few systems will\n> compile that but don't really have a useful integer timezone\n> (RH5.0/glibc2.0 is one of those).\n> \n> I'm wondering if we could change the sense of the test, to try instead\n> to test for the presence of a timezone field in the tm structure? That\n> might fix the glibc2.0 port (assuming it still has problems at v2.0.7;\n> haven't tested recently) but I don't know which other ports might break.\n> \n> Can we experiment with this Marc?? Post-megapatch of course :)\n\n\tSounds reasonable to me...so you want the test changed to:\n\n===========================================================================\n#include <stdio.h>\n#include <time.h>\n\nmain() { struct tm *tmstruct; printf(\"%s\\n\", tmstruct->timezone); }\n===========================================================================\n\n\tAnd, if the compile fails...how is HAVE_INT_TIMEZONE set? to\nFALSE?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 19 Mar 1998 00:24:37 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Timezone problems / HAVE_INT_TIMEZINE"
},
{
"msg_contents": "> Sounds reasonable to me...so you want the test changed to:\n========================================================================\n> #include <stdio.h>\n> #include <time.h>\n> \n> main() { struct tm *tmstruct; printf(\"%s\\n\", tmstruct->timezone); }\n> ========================================================================\n\nThe structure member looks like tm->tm_gmtoff (an integer). There would\nneed to be other calls to set it up, unless something like\n\n main() {struct tm tmstruct, *tm = &tmstruct; tm->tm_gmtoff = 0; }\n\nwould be acceptable.\n\n> And, if the compile fails...how is HAVE_INT_TIMEZONE set? to\n> FALSE?\n\nActually, if the test fails, then we need to #undef HAVE_INT_TIMEZONE,\nalthough if it would be easier to set it to FALSE then I can pretty\neasily fix up the sources to use that.\n\n - Tom\n",
"msg_date": "Thu, 19 Mar 1998 05:32:51 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: Timezone problems / HAVE_INT_TIMEZINE"
},
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nThomas G. Lockhart wrote:\n\n> The 19 year differences usually seem to come from mis-handling the\n> HAVE_INT_TIMEZONE compile-time option. How is yours set? Try changing it\n> in config.h and see if it helps.\n>\n\nAs far as I've been able to determine, the correct setting for\nHAVE_INT_TIMEZONE (1) is being used in the Alpha port. It does\nin fact define 'long timezone' (not 'int timezone') as being\navailable, as part of the tzset() man page. I have to admit that\nI'm not familiar with the way that this is supposed to work, so\nthis may seem kind of dumb, but I did some experimenting on the\nvalue of 'timezone' and 'tzname', since the contents of those\nvariable weren't documented anywhere that I could find in DEC's\nman pages. I of course now know that tzname[0] is the base\ntimezone name, tzname[1] is the dst name, and timezone is the\nnumber of seconds offset from GMT.\n\nHowever, what I also discovered in that these values are not set\nuntil after the tzset() routine is called. Is that normal\nbehavior? Doing a grep for tzset in the PG sources revealed\nthat it's only called for a few SQL commands. Is it called\nanywhere as part of startup processing, and I'm just missing it?\nOr is the DEC implementation the only one that requires an\nexplicit tzset() call before the use of these variables?\n\n- --\nDwayne Bailey + WHAT is your name? Sir Galahad\nMIKA Systems, Bingham Farms, MI + WHAT is your quest? I Seek the Holy Grail\[email protected] + What is your favorite color?\nhttp://www.mika.com/~dwayne + Blue ... no, Yelloooooooooooooooooow\n finger [email protected] for PGP Public Key\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQB1AwUBNREaJqA2uleK7maRAQGvdwL9F5t3M1dK8Qf9MVWGa3CfKguegHyG/f9+\n1Oe3OETtA5gI0GLUJkxgpVBQFMzT6kczju1AR6l7JcM2N+wXMk1lE5ULrLH96axd\nT8sLQwkdjTWhNsnBBFulyocyoLPF7TzK\n=SbKH\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "Thu, 19 Mar 1998 08:14:04 -0500 (EST)",
"msg_from": "Dwayne Bailey <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Timezone problems / HAVE_INT_TIMEZINE"
},
{
"msg_contents": "> However, what I also discovered in that these values are not set\n> until after the tzset() routine is called. Is that normal\n> behavior? Doing a grep for tzset in the PG sources revealed\n> that it's only called for a few SQL commands. Is it called\n> anywhere as part of startup processing, and I'm just missing it?\n> Or is the DEC implementation the only one that requires an\n> explicit tzset() call before the use of these variables?\n\nAFAIK tzset() is called automagically by all time-related libc routines \nwhen they detect it is not set yet (at least I think with Linux it is \ndone this way. It's been a long time since I looked at that).\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",
"msg_date": "Thu, 19 Mar 1998 16:10:09 +0100 (MET)",
"msg_from": "Maarten Boekhold <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Timezone problems / HAVE_INT_TIMEZINE"
},
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nOn Thu, 19 Mar 1998, Maarten Boekhold wrote:\n\n> AFAIK tzset() is called automagically by all time-related libc routines \n> when they detect it is not set yet (at least I think with Linux it is \n> done this way. It's been a long time since I looked at that).\n\nThat would explain it then. I was just accessing the variables\ndirectly, without any intervening calls.\n\nIt's a moot point, anyway. I put explicit calls in to the\nstartup, and it made no difference in the result. It's likely to\nbe a 32/64 bit issue somewhere that I haven't located yet. It\nreally shouldn't be that hard to track down. Since the output is\ndifferent from the input by a consistance amount (19 years +- a\nfew days) it can only be in one of 4 places, AFAIK: parsing\ninput, storing value, retrieving value, or generating output. My\nbet is on the retrieve phase, but we'll see.\n\n- --\nDwayne Bailey + WHAT is your name? Sir Galahad\nMIKA Systems, Bingham Farms, MI + WHAT is your quest? I Seek the Holy Grail\[email protected] + What is your favorite color?\nhttp://www.mika.com/~dwayne + Blue ... no, Yelloooooooooooooooooow\n finger [email protected] for PGP Public Key\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQB1AwUBNRE5eqA2uleK7maRAQGqPQMAgajIzCAK8cBRmqCHw83mVyI8i5YI7yo4\nj0jhJXG3vEauLST0B+6ompKw0+KQvRoOfgFWOoyqelZ08zo6qCBrJJmuAbGSM1/b\nEbBtsORCpSymqaeDIIPHoPdaq+jG9c8e\n=BiGQ\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "Thu, 19 Mar 1998 07:27:39 -0800 (PST)",
"msg_from": "Dwayne Bailey <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Timezone problems / HAVE_INT_TIMEZINE"
},
{
"msg_contents": "> It's a moot point, anyway. I put explicit calls in to the\n> startup, and it made no difference in the result. It's likely to\n> be a 32/64 bit issue somewhere that I haven't located yet. It\n> really shouldn't be that hard to track down. Since the output is\n> different from the input by a consistance amount (19 years +- a\n> few days) it can only be in one of 4 places, AFAIK: parsing\n> input, storing value, retrieving value, or generating output. My\n> bet is on the retrieve phase, but we'll see.\n\nDidn't this stuff work for v6.2.1, even on Alpha? afaik nothing around\nthis adt code changed recently...\n\n - Tom\n\nI moved to another job recently so left my dozen Alphas and don't have\naccess to man pages on them :( Have you tried compiling with\nHAVE_INT_TIMEZONE disabled?\n",
"msg_date": "Thu, 19 Mar 1998 15:42:43 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Timezone problems / HAVE_INT_TIMEZINE"
},
{
"msg_contents": "Applied.\n\n> \n> -----BEGIN PGP SIGNED MESSAGE-----\n> \n> I've gotten 6.3 initdb to run to a successful completion on my\n> Alpha running OSF/1 V3.2c. Forget the change that I sent in\n> earlier. While I still think that there's something funky with\n> that code, it doesn not need to be modifed. Actually, the\n> modifications are miniscule. The only files that need to be\n> changed are backend/main/main.c and template/alpha.\n> \n> The real trick is to add -Dalpha to the CFLAGS setting. The\n> changes to main.c are only to add some extra includes to support\n> some code that's suddenly being used.\n> \n> The #define ASSEMBLER is to prevent most of the code of\n> sys/proc.h from being included, as it ends up conflicting with\n> some of the postgresql definitions. This may or may not work on\n> other versions of Digital Unix.\n> \n> As far as I'm concerned, this is a hack fix. There's still some\n> underlying 32/64 bit assumtions that this is masking. Perhaps\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 22:53:33 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Alpha initdb fixed!"
}
] |
[
{
"msg_contents": "Redirected to Questions list. (It is cool to say that.)\n\n> -----Original Message-----\n> hello all,\n> \n> yep, i'm new to postgres so please bare with me. if this isn't the\n> correct place to ask these questions, please point me in the right\n> direction.\nI know how you feel.\n\n> first and foremost,\n> \n> i need to have a field (say pin) that is the primary key of a given\n> relation. each time a record gets added to this relation, the pin\n> increases by 1. thus making an auto increasing primary key (i.e. a\n> unique identifier for each record that isn't explicitly increased by\n> 1,\n> the dbms does it for me.)\nWhat you are looking for is a sequence. Try\nman create_sequence (or '\\h create sequence' in psql).\nex. \n create sequence myseq start 0 minvalue 0 cycle;\n create table mytbl (\n mykey int4 default nextval('myseq') primary key, \n mydata1 varchar(20), \n ...);\nwhen you do an insert you'll have to explicitly skip mykey.\nex. \n insert into mytbl(mydata1, ...) \n values ('Hello, world');\n\n> second,\n> \n> i have a platform related installation question, which list should be\n> used.\nI'll let someone else answer this on from fear of being wrong.\n\n> thankx,\n> \n> -brian ([email protected])\n> \n\t-DEJ\n",
"msg_date": "Mon, 16 Mar 1998 15:51:46 -0600",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] AUTOINDEXING AND HOROLOGY REGRESSION ERROR ON PGSQL\n\t6.3, LINUX-ELF i686"
}
] |
[
{
"msg_contents": "Is this possible? Or is it already being done?\n\n<a lot snipped>\n> > > How do you assign function to your objects?\n<snip>\n> I don't know if tying functions to classes has been added\n> to the TO-DO list. You may want to post to the hacker list and\n> suggest\n> it.\n> \nAccess to the function would then be gained by calling them through the\nclass. (i.e. select EMP.totalPeople();, or update CORPEMP set\nCORPEMP.totals = DIVEMP.totalPeople();)\nJust an OO thought. (*ahhhh* It's that concurrency/data-hiding/coupling\nthing again.)\nThen one could disallow select/update on a table but all execute on\nspecific functions. I do realize that this can be accomplished through\ntriggers, but this has other potentials...\n",
"msg_date": "Mon, 16 Mar 1998 19:41:38 -0600",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [QUESTIONS] The Complete and Ccomprehensive list of Newbie qu\n\testion"
},
{
"msg_contents": "Jackson, DeJuan wrote:\n> \n> Is this possible? Or is it already being done?\n> \n> <a lot snipped>\n> > > > How do you assign function to your objects?\n> <snip>\n> > I don't know if tying functions to classes has been added\n> > to the TO-DO list. You may want to post to the hacker list and\n> > suggest\n> > it.\n> > \n> Access to the function would then be gained by calling them through the\n> class. (i.e. select EMP.totalPeople();, or update CORPEMP set\n> CORPEMP.totals = DIVEMP.totalPeople();)\n> Just an OO thought. (*ahhhh* It's that concurrency/data-hiding/coupling\n> thing again.)\n> Then one could disallow select/update on a table but all execute on\n> specific functions. I do realize that this can be accomplished through\n> triggers, but this has other potentials...\n\nLooks sort of like our procedural language, except that instead of\nhaving a \"pool\" of functions, functions can be \"member functions\" of a\nclass. In order to do this \"all the way\", it seems like we will need\nall instances to appear in their parent classes -- Select\npeople.average_age() should hit all instances of people, even those\nthat have been subclassed into teachers, police, lawyers, etc.\n\nOcie Mitchell\n\n",
"msg_date": "Mon, 16 Mar 1998 17:58:56 -0800 (PST)",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [QUESTIONS] The Complete and Ccomprehensive list of\n\tNewbie qu"
},
{
"msg_contents": ">\n> Is this possible? Or is it already being done?\n>\n> <a lot snipped>\n> > > > How do you assign function to your objects?\n> <snip>\n> > I don't know if tying functions to classes has been added\n> > to the TO-DO list. You may want to post to the hacker list and\n> > suggest\n> > it.\n> >\n> Access to the function would then be gained by calling them through the\n> class. (i.e. select EMP.totalPeople();, or update CORPEMP set\n> CORPEMP.totals = DIVEMP.totalPeople();)\n> Just an OO thought. (*ahhhh* It's that concurrency/data-hiding/coupling\n> thing again.)\n> Then one could disallow select/update on a table but all execute on\n> specific functions. I do realize that this can be accomplished through\n> triggers, but this has other potentials...\n\n Currently, we don't have ACL's on functions. But at least for\n the ones called by ExecMakeFunctionResult(), this really\n makes sense as soon as we implement a uid/euid model for\n function/view execution.\n\n The uid/euid for functions, triggers and views is already on\n my personal TODO and I hope to make it for 6.4, so we can\n offer view and function/trigger creation (restricted to\n trusted languages) to ordinary users.\n\n When adding ACL's to functions, I think the ACL's at all\n should be moved into a separate catalog, containing the Oid,\n owner and the ACL.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Tue, 17 Mar 1998 08:19:24 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] RE: [QUESTIONS] The Complete and Ccomprehensive list of\n\tNewbie qu"
}
] |
[
{
"msg_contents": "I was playing around with my latest compile and tried to make a unique\nindex on two columns, using a hash method. Both of these (more than\none column and unique) are currently not allowed for hash indexes.\n\nI thought about this for a bit and realized that making a NESTED hash\nindex (index on a and b also serves as an index on a) would be a\ntrick, but allowing the unique clause should not be a problem.\n\nTherefore, I would like to try implementing unique constraints on hash\nindexes. Has this come up before? Are there any reasons not to\nsupport this? As far as I understand, specifying an index method is\nnon standard (above and beyond standard) to begin with.\n\nOcie Mitchell\n\n",
"msg_date": "Mon, 16 Mar 1998 17:45:15 -0800 (PST)",
"msg_from": "[email protected]",
"msg_from_op": true,
"msg_subject": "Unique index using hash?"
},
{
"msg_contents": "> \n> I was playing around with my latest compile and tried to make a unique\n> index on two columns, using a hash method. Both of these (more than\n> one column and unique) are currently not allowed for hash indexes.\n> \n> I thought about this for a bit and realized that making a NESTED hash\n> index (index on a and b also serves as an index on a) would be a\n> trick, but allowing the unique clause should not be a problem.\n> \n> Therefore, I would like to try implementing unique constraints on hash\n> indexes. Has this come up before? Are there any reasons not to\n> support this? As far as I understand, specifying an index method is\n> non standard (above and beyond standard) to begin with.\n> \n\nIt is on the TODO list:\n\n\t* add UNIQUE capability to non-btree indexes\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 22:02:49 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Unique index using hash?"
}
] |
[
{
"msg_contents": "> Looks sort of like our procedural language, except that instead of\n> having a \"pool\" of functions, functions can be \"member functions\" of a\n> class. In order to do this \"all the way\", it seems like we will need\n> all instances to appear in their parent classes -- Select\n> people.average_age() should hit all instances of people, even those\n> that have been subclassed into teachers, police, lawyers, etc.\n> \n> Ocie Mitchell\nAgreed. I meant member functions. But I wouldn't take away the global\npool of functions either. This satisfies my since of OOP.\n",
"msg_date": "Mon, 16 Mar 1998 20:14:09 -0600",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] RE: [QUESTIONS] The Complete and Ccomprehensive lis\n\tt of Newbie qu"
}
] |
[
{
"msg_contents": "> > I am not absolutly happy with this because it looks like the \n> > statement is not working correctly for 100%. It comes back with an \n> > strange error message if the '%' sign is at the end of '40'!!..\n> >\n> > select * from my_table where cast(my_int as text) like '40%';\n> > ERROR: transformExpr: does not know how to transform node 103\n> >\n> > Any ideas!..\n> Thomas, I believe this is the problem with my processing a node twice.\n> Do you have a fix for that, or should I make one?\n\ntgl=> select * from t where cast(i as text) like '40%';\n i\n----\n4030\n(1 row)\n\nSo, the patch for \"function() BETWEEN\" helps, but I'm still not sure it\nis safe, since I don't know how these nodes might be handled deeper in\nthe backend.\n\nAlso, one of my recent testing patches slightly breaks the type\nconversion stuff so the numerology regression test fails on one of the\ninserts and I haven't figured out which piece caused that. Nothing\nshould be affected in the CVS tree though.\n\n - Tom\n",
"msg_date": "Tue, 17 Mar 1998 02:52:04 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [QUESTIONS] UPDATE statement ORACLE 6 compatible"
},
{
"msg_contents": "> \n> tgl=> select * from t where cast(i as text) like '40%';\n> i\n> ----\n> 4030\n> (1 row)\n> \n> So, the patch for \"function() BETWEEN\" helps, but I'm still not sure it\n> is safe, since I don't know how these nodes might be handled deeper in\n> the backend.\n> \n> Also, one of my recent testing patches slightly breaks the type\n> conversion stuff so the numerology regression test fails on one of the\n> inserts and I haven't figured out which piece caused that. Nothing\n> should be affected in the CVS tree though.\n\nShip the patch over here and I will have a look at 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": "Mon, 16 Mar 1998 22:21:53 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: [QUESTIONS] UPDATE statement ORACLE 6 compatible"
},
{
"msg_contents": "> > So, the patch for \"function() BETWEEN\" helps, but I'm still not sure \n> > it is safe, since I don't know how these nodes might be handled \n> > deeper in the backend.\n> \n> Ship the patch over here and I will have a look at it.\n\nOK, here it is...\n\n - Tom",
"msg_date": "Tue, 17 Mar 1998 04:08:21 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: [QUESTIONS] UPDATE statement ORACLE 6 compatible"
},
{
"msg_contents": "> > Thomas, I believe this is the problem with my processing a node \n> > twice. Do you have a fix for that, or should I make one?\n> So, the patch for \"function() BETWEEN\" helps, but I'm still not sure\n> it is safe, since I don't know how these nodes might be handled deeper \n> in the backend.\n\nBruce, did I remember to actually enclose the patch for you? Looks to me\nlike I didn't, so here it is again...\n\n - Tom",
"msg_date": "Fri, 20 Mar 1998 03:27:47 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: [QUESTIONS] UPDATE statement ORACLE 6 compatible"
},
{
"msg_contents": "> \n> This is a multi-part message in MIME format.\n> --------------5B2D8E5180AAC1149F41A1FA\n> Content-Type: text/plain; charset=us-ascii\n> Content-Transfer-Encoding: 7bit\n> \n> > > So, the patch for \"function() BETWEEN\" helps, but I'm still not sure \n> > > it is safe, since I don't know how these nodes might be handled \n> > > deeper in the backend.\n> > \n> > Ship the patch over here and I will have a look at it.\n> \n> OK, here it is...\n\nApplied. There is a nice comment, and it just prevents an elog(). I\nthink your logic is correct. There are several cases, I think, where\nstrings are used multiple places. Also, you are just preventing\nreprocessing of the args. We can perhaps improve this later if a\nproblem comes up.\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 22:59:46 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: [QUESTIONS] UPDATE statement ORACLE 6 compatible"
}
] |
[
{
"msg_contents": "\n[ I'm forwarding this to the hackers list, as this looks like a problem in\nthe backend ]\n\nOn Mon, 16 Mar 1998, Uri Blumenthal wrote:\n\n> > I've tried to reproduce this, but it's not failing.\n> \n> Lucky you are...\n> \n> > This sounds like the backend dying while in the middle of the fastpath\n> > call. If you run the backend with debugging enabled, does anything appear\n> > when this fault occurs?\n> \n> You bet. Here's the output of \"-d 3 -o /tmp/pgsql.log\":\n> \n> NOTICE: DateStyle is Postgres with US (NonEuropean) conventions\n> NOTICE: buffer leak [77] detected in BufferPoolCheckLeak()\n> NOTICE: buffer leak [77] detected in BufferPoolCheckLeak()\n> NOTICE: buffer leak [77] detected in BufferPoolCheckLeak()\n> NOTICE: buffer leak [77] detected in BufferPoolCheckLeak()\n> Failed Assertion(\"!(BufferIsValid(buffer)):\", File: \"bufmgr.c\", Line: 1294)\n> !(BufferIsValid(buffer)) (0) [No such file or directory]\n\nUri, what platform are you running the backend? This looks like something\nserious is going wrong here...\n\n> And here's the Java dump:\n\n[ dump snipped - it's just confirming the above ]\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 Mar 1998 07:10:04 +0000 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Problem with JDBC interface in Postgresql-6.3"
}
] |
[
{
"msg_contents": "\nOn Mon, 16 Mar 1998, Uri Blumenthal wrote:\n\n> >>>>> \"Peter\" == Peter T Mount <[email protected]> writes:\n> \n> Peter> I've tried to reproduce this, but it's not failing.\n> \n> Darn... I recompiled and re-installed PostgreSQL-6.3, and\n> now I can't reproduce this yesterday's error either. (:-(\n> \n> [but see below.]\n> \n> Opening large object 19201\n> Opening test destination object\n> Copying large object to file\n> object size=6839\n> copied 512 bytes, 6327 left\n> copied 1024 bytes, 5815 left\n> copied 1536 bytes, 5303 left\n> copied 2048 bytes, 4791 left\n> copied 2560 bytes, 4279 left\n> copied 3072 bytes, 3767 left\n> copied 3584 bytes, 3255 left\n> copied 4096 bytes, 2743 left\n> copied 4608 bytes, 2231 left\n> copied 5120 bytes, 1719 left\n> copied 5632 bytes, 1207 left\n> copied 6144 bytes, 695 left\n> copied 6656 bytes, 183 left\n> copied 6839 bytes, 0 left\n> The whole object copied successfully (6839 bytes).\n> Closing object\n\nweird... at least that problems now gone ;-)\n\n> But now I'm trying \"example.datestyle\" and it fails me:\n> \n> PostgreSQL datestyle test v6.3 rev 1\n> \n> Connecting to Database URL = jdbc:postgresql:foo\n> Connected...Now creating a statement\n> \n> Running tests:\n> Test 0 - postgres,european failed, returned 08-01-1998\n> Test 1 - postgres,us failed, returned 01-08-1998\n> Test 2 - iso failed, returned 1998-01-08\n> Test 3 - sql,european failed, returned 08/01/1998\n> Test 4 - sql,us failed, returned 01/08/1998\n> Test 5 - german failed, returned 08.01.1998\n> Now closing the connection\n\nJust ran here:\n\nRunning tests:\nTest 0 - postgres,european passed, returned 08-01-1998\nTest 1 - postgres,us passed, returned 01-08-1998\nTest 2 - iso passed, returned 1998-01-08\nTest 3 - sql,european passed, returned 08/01/1998\nTest 4 - sql,us passed, returned 01/08/1998\nTest 5 - german passed, returned 08.01.1998\nNow closing the connection\n\n\n> Surprising thing is, that the date appears to be the same,\n> so it's beyond me why \n> \tstandard.equals(rs.getDate(1)) \n> \n> returns \"False\".\n> \n> Any idea?\n\nAgain, are your client & backend on different machines/platforms, and what\nplatform(s) are you using?\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 Mar 1998 07:14:18 +0000 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Problem with JDBC interface in Postgresql-6.3"
}
] |
[
{
"msg_contents": "Ocie Mitchell wrote:\n\n> I was playing around with my latest compile and tried to make a unique\n\n> index on two columns, using a hash method. Both of these (more than\n> one column and unique) are currently not allowed for hash indexes.\n>\n> I thought about this for a bit and realized that making a NESTED hash\n> index (index on a and b also serves as an index on a) would be a\n> trick, but allowing the unique clause should not be a problem.\n\nIt can be complicated (especially for extensible hashing) but the\ntheory\nfor this seems to be on in the database handbooks under the name of\n'segmented hash' or some like fancy name.\n\nThe trick is to hash each field separately and then have a concatenation\n\nof the hash values.\n\nso assuming that for fields (a,b,c) values (120, 'friday', 3.1415927)\nhash\nto 'aa', 'bb', 'cc' the hash value of the whole tuple will be 'aabbcc'\n\nthen the index can be used not only when selecting = a,b,c but also\nwhen\nselecting on _any_ of the fields in this index. For example when\nselecting\nfor b='friday' one would examine only buckets where the middle is 'bb'\n\n> Therefore, I would like to try implementing unique constraints on hash\n\n> indexes. Has this come up before? Are there any reasons not to\n> support this? As far as I understand, specifying an index method is\n> non standard (above and beyond standard) to begin with.\n\nWhile you are at it could you please comment if the GIST indexes are\nused or\nat least easily usable?\n\nI have asked this question once in each two months for about a year now,\nbut\nit seems that noone is in the position to answer it.\n\nCheers,\n\nHannu Krosing\n\n",
"msg_date": "Tue, 17 Mar 1998 10:16:59 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Multi field hash indexes"
},
{
"msg_contents": "Hannu Krosing writes:\n> It can be complicated (especially for extensible hashing) but the\n> theory\n> for this seems to be on in the database handbooks under the name of\n> 'segmented hash' or some like fancy name.\n> \n> The trick is to hash each field separately and then have a concatenation\n> \n> of the hash values.\n\nMore or less.\n\n> so assuming that for fields (a,b,c) values (120, 'friday', 3.1415927)\n> hash\n> to 'aa', 'bb', 'cc' the hash value of the whole tuple will be 'aabbcc'\n\nI had a research project on this several years ago and we found that all\nthese multidimensional hash trees have one or the other problem. But one\nindex type named Balanced Multidimensional Hash Tree (from the top of my\nhead) performed excellently by combining hash and tree structures. \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 Mar 1998 14:00:01 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Multi field hash indexes"
},
{
"msg_contents": "Hannu Krosing wrote:\n> \n> Ocie Mitchell wrote:\n> \n> > I was playing around with my latest compile and tried to make a unique\n> \n> > index on two columns, using a hash method. Both of these (more than\n> > one column and unique) are currently not allowed for hash indexes.\n> >\n> > I thought about this for a bit and realized that making a NESTED hash\n> > index (index on a and b also serves as an index on a) would be a\n> > trick, but allowing the unique clause should not be a problem.\n> \n> It can be complicated (especially for extensible hashing) but the\n> theory\n> for this seems to be on in the database handbooks under the name of\n> 'segmented hash' or some like fancy name.\n> \n> The trick is to hash each field separately and then have a concatenation\n> \n> of the hash values.\n> \n> so assuming that for fields (a,b,c) values (120, 'friday', 3.1415927)\n> hash\n> to 'aa', 'bb', 'cc' the hash value of the whole tuple will be 'aabbcc'\n> \n> then the index can be used not only when selecting = a,b,c but also\n> when\n> selecting on _any_ of the fields in this index. For example when\n> selecting\n> for b='friday' one would examine only buckets where the middle is 'bb'\n> \n\nHMM, this doesn't feel right. If I have an index on four int4s\na,b,c,d and I only know d, then it seems like searching for these in\nthe hash table could be as much work, or more work than a table scan.\nNow if a hash on two columns implicitly make a seperate hash on each,\nand then took their intersection, that might be fast. \n\nThe kind of thing I am thinking that these two-or-more-column hash\nindexes would be use for supporting intersection tests, where the\norder of the items doesn't matter, but it is good to be able to come\nup with a quick answer to the questions:\n\nselect x from t where y=5;\nselect y from t where x in (1, 4, 6, 7);\nselect count(*) from t where x=1 and y=10;\n\n\nI was originally thinking that this would be supported like the btree\nindexes are now -- an index on (a,b,c,d) serves as in index on a,\n(a,b), (a,b,c) and (a,b,c,d), but it doesn't serve as an index on b,\nor (b,c), etc. My original idea was that the first item in the index\nwould define a hash table whose entries were hash tables on the second\nitem, etc. I now think that this would waste quite a bit of space,\nand would have the same restriction as btrees, which is unnatural.\n\n\n> > Therefore, I would like to try implementing unique constraints on hash\n> \n> > indexes. Has this come up before? Are there any reasons not to\n> > support this? As far as I understand, specifying an index method is\n> > non standard (above and beyond standard) to begin with.\n> \n> While you are at it could you please comment if the GIST indexes are\n> used or\n> at least easily usable?\n\nWhat are GIST indexes?\n\n\nOcie Mitchell\n",
"msg_date": "Tue, 17 Mar 1998 10:40:05 -0800 (PST)",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Multi field hash indexes"
},
{
"msg_contents": "> I was originally thinking that this would be supported like the btree\n> indexes are now -- an index on (a,b,c,d) serves as in index on a,\n> (a,b), (a,b,c) and (a,b,c,d), but it doesn't serve as an index on b,\n> or (b,c), etc. My original idea was that the first item in the index\n> would define a hash table whose entries were hash tables on the second\n> item, etc. I now think that this would waste quite a bit of space,\n> and would have the same restriction as btrees, which is unnatural.\n\nThis is a standard restriction. If you need an index on a lower-level\nfield, create one. I don't think you are going to be able to improve on\n(a,b), (a,b,c). If you allowed (b,c) that is another index.\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": "Tue, 17 Mar 1998 14:05:07 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Multi field hash indexes"
},
{
"msg_contents": "Bruce Momjian wrote:\n> \n> > I was originally thinking that this would be supported like the btree\n> > indexes are now -- an index on (a,b,c,d) serves as in index on a,\n> > (a,b), (a,b,c) and (a,b,c,d), but it doesn't serve as an index on b,\n> > or (b,c), etc. My original idea was that the first item in the index\n> > would define a hash table whose entries were hash tables on the second\n> > item, etc. I now think that this would waste quite a bit of space,\n> > and would have the same restriction as btrees, which is unnatural.\n> \n> This is a standard restriction. If you need an index on a lower-level\n> field, create one. I don't think you are going to be able to improve on\n> (a,b), (a,b,c). If you allowed (b,c) that is another index.\n\nWhat I meant is that this is a perfectly reasonable restriction for\nbtree indexes. If the index is built on (a,b,c,d), then finding the\nnodes where a=5 gives you a few branches of the tree, and if you\nfurther restrict this to b=10, then you need only look at these\nbranches. The same could be said for ranges of a values.\n\nHash tables are different in that they do not have this tree structure\n(unless we impose this structure on them. The problem I see with\nbuilding this tree structure as I described earlier is that it has the\npossibility for a lot of wasted space (even more than btrees). Also,\nif we can provide the added functionality that a hash index on (a,b)\nis also an index on b, with equal or lesser memory usage than the\nhash-tree, wouldn't this be a \"good thing\".\n\nAnyway, I think this needs some thinking. I will look at the\nalternatives and post something a bit later.\n\nOcie\n",
"msg_date": "Tue, 17 Mar 1998 11:41:46 -0800 (PST)",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Multi field hash indexes"
},
{
"msg_contents": "[email protected] wrote:\n\n> Hannu Krosing wrote:\n> >\n> > The trick is to hash each field separately and then have a concatenation\n> >\n> > of the hash values.\n> >\n> > so assuming that for fields (a,b,c) values (120, 'friday', 3.1415927)\n> > hash\n> > to 'aa', 'bb', 'cc' the hash value of the whole tuple will be 'aabbcc'\n> >\n> > then the index can be used not only when selecting = a,b,c but also\n> > when\n> > selecting on _any_ of the fields in this index. For example when\n> > selecting\n> > for b='friday' one would examine only buckets where the middle is 'bb'\n> >\n>\n> HMM, this doesn't feel right. If I have an index on four int4s\n> a,b,c,d and I only know d, then it seems like searching for these in\n> the hash table could be as much work, or more work than a table scan.\n\nI was assuming that the hash table scan would be cheaper than the table scan\n\n> > > Therefore, I would like to try implementing unique constraints on hash\n> >\n> > > indexes. Has this come up before? Are there any reasons not to\n> > > support this? As far as I understand, specifying an index method is\n> > > non standard (above and beyond standard) to begin with.\n> >\n> > While you are at it could you please comment if the GIST indexes are\n> > used or\n> > at least easily usable?\n>\n> What are GIST indexes?\n\nSome kind of generalised binary tree indexes that should make it easy to\ndefine additional indexing strategies.\n\nThere is a directory access/gist in src/backend, and they are briefly\nmentioned in the PostgreSQL programmers guide.\n\nThey seem to be offspring of some independent (of postgres) Berkeley project.\n\nHannu\n\n",
"msg_date": "Tue, 17 Mar 1998 21:55:42 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: Multi field hash indexes"
},
{
"msg_contents": "Hannu Krosing wrote:\n> \n> [email protected] wrote:\n> \n> > Hannu Krosing wrote:\n> > >\n> > > The trick is to hash each field separately and then have a concatenation\n> > >\n> > > of the hash values.\n> > >\n> > > so assuming that for fields (a,b,c) values (120, 'friday', 3.1415927)\n> > > hash\n> > > to 'aa', 'bb', 'cc' the hash value of the whole tuple will be 'aabbcc'\n> > >\n> > > then the index can be used not only when selecting = a,b,c but also\n> > > when\n> > > selecting on _any_ of the fields in this index. For example when\n> > > selecting\n> > > for b='friday' one would examine only buckets where the middle is 'bb'\n> > >\n> >\n> > HMM, this doesn't feel right. If I have an index on four int4s\n> > a,b,c,d and I only know d, then it seems like searching for these in\n> > the hash table could be as much work, or more work than a table scan.\n> \n> I was assuming that the hash table scan would be cheaper than the table scan\n\nI think I see what you're getting at. The different parts are\nseparately hashed and then concatenated. I was thinking the other way\nconcatenate and THEN hash, which would spread the values all over the\ntable. \n\nOne problem with this (don't know how severe) is that a hash index on\na can answer 'select blah from t where a=10' by computing one hash\nvalue and scaning the list at that point, whereas with an index on\n(a,b), several hash values must be computed (one for each possible\nresulting hash value of b).\n\n> \n> > > > Therefore, I would like to try implementing unique constraints on hash\n> > >\n> > > > indexes. Has this come up before? Are there any reasons not to\n> > > > support this? As far as I understand, specifying an index method is\n> > > > non standard (above and beyond standard) to begin with.\n> > >\n> > > While you are at it could you please comment if the GIST indexes are\n> > > used or\n> > > at least easily usable?\n> >\n> > What are GIST indexes?\n> \n> Some kind of generalised binary tree indexes that should make it easy to\n> define additional indexing strategies.\n> \n> There is a directory access/gist in src/backend, and they are briefly\n> mentioned in the PostgreSQL programmers guide.\n> \n> They seem to be offspring of some independent (of postgres) Berkeley project.\n\nInteresting. I saw something in Dr Dobbs Journal (the last bastion of\nprogramming magazines) about trinary trees. The idea is that there is\na less than branch for items less than the given node and one for\nthose greater and another for those equal. This can be used to index\nitems by using the less than/ greater than links to find the first\nletter, then taking the 'equal' link and looking for the second\nletter, etc. The problem is that these are not self balancing (like\nred black trees or btrees), but maybe there is a way to make them so.\n\nOcie\n",
"msg_date": "Tue, 17 Mar 1998 13:07:11 -0800 (PST)",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Multi field hash indexes"
},
{
"msg_contents": "[email protected] wrote:\n\n> Hannu Krosing wrote:\n> >\n> > [email protected] wrote:\n> >\n> > > Hannu Krosing wrote:\n> > > >\n> > > > The trick is to hash each field separately and then have a concatenation\n> > > >\n> > > > of the hash values.\n> > > >\n> > > > so assuming that for fields (a,b,c) values (120, 'friday', 3.1415927)\n> > > > hash\n> > > > to 'aa', 'bb', 'cc' the hash value of the whole tuple will be 'aabbcc'\n> > > >\n> > > > then the index can be used not only when selecting = a,b,c but also\n> > > > when\n> > > > selecting on _any_ of the fields in this index. For example when\n> > > > selecting\n> > > > for b='friday' one would examine only buckets where the middle is 'bb'\n> > > >\n> > >\n> > > HMM, this doesn't feel right. If I have an index on four int4s\n> > > a,b,c,d and I only know d, then it seems like searching for these in\n> > > the hash table could be as much work, or more work than a table scan.\n> >\n> > I was assuming that the hash table scan would be cheaper than the table scan\n>\n> I think I see what you're getting at. The different parts are\n> separately hashed and then concatenated. I was thinking the other way\n> concatenate and THEN hash, which would spread the values all over the\n> table.\n>\n> One problem with this (don't know how severe) is that a hash index on\n> a can answer 'select blah from t where a=10' by computing one hash\n> value and scaning the list at that point, whereas with an index on\n> (a,b), several hash values must be computed (one for each possible\n> resulting hash value of b).\n\nThe idea is that after computing the hash value you still have to find that value\nand\nthat values are arranged in some way that makes finding consecutive values cheap\n(like ordered or tree or matrix (using sparse files)), at least for values with the\nsame\nstart or possibly for all other kinds of same parts.\n\n> > > > > indexes. Has this come up before? Are there any reasons not to\n> > > > > support this? As far as I understand, specifying an index method is\n> > > > > non standard (above and beyond standard) to begin with.\n> > > >\n> > > > While you are at it could you please comment if the GIST indexes are\n> > > > used or\n> > > > at least easily usable?\n> > >\n> > > What are GIST indexes?\n> >\n> > Some kind of generalised binary tree indexes that should make it easy to\n> > define additional indexing strategies.\n> >\n> > There is a directory access/gist in src/backend, and they are briefly\n> > mentioned in the PostgreSQL programmers guide.\n> >\n> > They seem to be offspring of some independent (of postgres) Berkeley project.\n>\n> Interesting. I saw something in Dr Dobbs Journal (the last bastion of\n> programming magazines) about trinary trees. The idea is that there is\n> a less than branch for items less than the given node and one for\n> those greater and another for those equal. This can be used to index\n> items by using the less than/ greater than links to find the first\n> letter, then taking the 'equal' link and looking for the second\n> letter, etc. The problem is that these are not self balancing (like\n> red black trees or btrees), but maybe there is a way to make them so.\n\nThe information about GIST is at:\n\nhttp://GiST.CS.Berkeley.EDU:8000/gist/\n\nwith more on different indexing and sorting schemes at:\n\nhttp://s2k-ftp.CS.Berkeley.EDU:8000/personal/jmh/\n\nAnd there is more interesting reading at the Berkely databese site at :\n\nhttp://epoch.cs.berkeley.edu:8000/\n\n\nHannu\n\n",
"msg_date": "Wed, 18 Mar 1998 08:58:21 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: Multi field hash indexes"
}
] |
[
{
"msg_contents": "> The file can be found in\n> \n> \tftp.postgresql.org:/pub/patches/.test/v6.3p1.tar.gz\n> \n> and is hidden from public view... \n> \n> *NOTE: This is not to be discussed in any other mailing list, period, or\n> you will lose your secret decoder ring, and we'll have to cut off\n> your hands so that you can't do the secret handshake anymore*\n> \n> \n> \nGeesh! Why the secrecy? Surely the more people who test it the better. Look\nat the number of bugs that have turned up in the last 2 weeks since 6.3\nwas released - none of them *terribly* serious, but a fair number none\nthe less.\n\nHas Marc suddenly taken some sort of commercial attitude???? (Give the\ncustomer a product - who cares if it hasn't been tested by as many\npeople as possible - we can charge for an upgrade :-)\n\n\nYours puzzled!\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": "Tue, 17 Mar 1998 10:50:32 GMT",
"msg_from": "Andrew Martin <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] First mega-patch..."
},
{
"msg_contents": "On Tue, 17 Mar 1998, Andrew Martin wrote:\n\n> > The file can be found in\n> > \n> > \tftp.postgresql.org:/pub/patches/.test/v6.3p1.tar.gz\n> > \n> > and is hidden from public view... \n> > \n> > *NOTE: This is not to be discussed in any other mailing list, period, or\n> > you will lose your secret decoder ring, and we'll have to cut off\n> > your hands so that you can't do the secret handshake anymore*\n> > \n> > \n> > \n> Geesh! Why the secrecy? Surely the more people who test it the better. Look\n> at the number of bugs that have turned up in the last 2 weeks since 6.3\n> was released - none of them *terribly* serious, but a fair number none\n> the less.\n> \n> Has Marc suddenly taken some sort of commercial attitude???? (Give the\n> customer a product - who cares if it hasn't been tested by as many\n> people as possible - we can charge for an upgrade :-)\n\n\tNo, Marc's attitude hasn't changed at all...pgsql-hackers is the\nplace where *development* issues take place, and it is the *only* list\nwhere discussing *any* pre-release software. If someone on\npgsql-questions (a *GENERIC*, how to *USE* PostgreSQL list) is interested,\nthey are more then welcome to join pgsql-hackers...\n\n\n",
"msg_date": "Tue, 17 Mar 1998 08:16:14 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] First mega-patch..."
}
] |
[
{
"msg_contents": "Can anyone explain the following:\n\nhannu=> select 1 + 2 + 3 + 4;\n?column?\n--------\n 10\n(1 row)\n\nworks fine, but:\n\nhannu=> select '1' || '2' || '3';\nERROR: parser: parse error at or near \"||\"\nhannu=> select ('1' || '2') || '3';\nNOTICE: there is more than one operator || for types\nNOTICE: unknown and unknown. You will have to retype this query\nERROR: using an explicit cast\nhannu=> select ('1'::text || '2') || '3';\n?column?\n--------\n 123\n(1 row)\n\nI understand the 'there is more than one operator' errors, but\nwhat is so different between operators + and || ?\n\nHannu\n\n",
"msg_date": "Tue, 17 Mar 1998 13:28:53 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": true,
"msg_subject": "Possible bug in parsing"
},
{
"msg_contents": "> I understand the 'there is more than one operator' errors, but\n> what is so different between operators + and || ?\n\n?? They are different operators, that's why. The \"+\" is addition, and\nthe \"||\" is defined in SQL92 to be string concatenation.\n\n - Tom\n",
"msg_date": "Wed, 18 Mar 1998 06:54:31 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Possible bug in parsing"
},
{
"msg_contents": "\n\nThomas G. Lockhart wrote:\n\n> > I understand the 'there is more than one operator' errors, but\n> > what is so different between operators + and || ?\n>\n> ?? They are different operators, that's why. The \"+\" is addition, and\n> the \"||\" is defined in SQL92 to be string concatenation.\n>\n>\n\nYes I know that, but why are they so different that I can have\n\n ARG1 op ARG2 op ARG3\n\nwith + but not with ||\n\nHannu\n\n\n\n",
"msg_date": "Wed, 18 Mar 1998 09:31:40 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Possible bug in parsing"
},
{
"msg_contents": "Thomas G. Lockhart wrote:\n\n> > I understand the 'there is more than one operator' errors, but\n> > what is so different between operators + and || ?\n>\n> ?? They are different operators, that's why. The \"+\" is addition, and\n> the \"||\" is defined in SQL92 to be string concatenation.\n\nI actually found this while trying to find out how characters in strings\nare escaped\nand if \\0 is supported in them.\n\nIs there some docs on escaping characters, and if not then where do you\nthink it should go in docs ?\n\nHannu\n\n\n\n",
"msg_date": "Wed, 18 Mar 1998 09:42:41 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Possible bug in parsing"
},
{
"msg_contents": "> > > I understand the 'there is more than one operator' errors, but\n> > > what is so different between operators + and || ?\n> \n> Yes I know that, but why are they so different that I can have\n> ARG1 op ARG2 op ARG3\n> with + but not with ||\n\nAh! I *thought* that was a silly question, and as usual that's because I\ndidn't understand the question :)\n\nThey behave differently because the \"+\" operator is allowed to be\n\"left-associative\", since its behavior is well-defined mathematically,\nwhile the \"||\" operator falls into a broad class of multi-character\noperators in the parser which are \"non-associative\". So, the parser does\nnot know whether to evaluate left-to-right or right-to-left and throws\nan error instead.\n\n> I actually found this while trying to find out how characters in\n> strings are escaped and if \\0 is supported in them.\n> \n> Is there some docs on escaping characters, and if not then where do \n> you think it should go in docs ?\n\nI believe that there are no docs on escaping characters and I would\nsuggest that docs for it should go into the user's guide in the section\non character data types (since it applies to all of them). That would be\nin doc/src/sgml/datatype.sgml, probably in a \"Sect2\" at the end of the\n\"Sect1\" on \"Character Types\" and just before the \"Sect1\" on \"Date/Time\nTypes\".\n\nIf you find it difficult to work with that source document, you can just\ntype the information and I can help to integrate it into the doc.\n\nbtw, I think that the escape conventions for strings need some work; the\nfact that the characters are re-escaped when output makes it easy to do\na dump/reload but sort of defeats the usefulness of escaping anything in\nthe first place.\n\n - Tom\n",
"msg_date": "Wed, 18 Mar 1998 14:44:43 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Possible bug in parsing"
}
] |
[
{
"msg_contents": "Is there a standard that defines how to return information from an embedded\nSQL statement? ORACLE uses the sqlca. In particular it uses sqlca.sqlerrd[2]\nfor affected tuples. Is this correct with the standards? How do other\nsystems do that?\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, 17 Mar 1998 16:53:24 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "standards question"
},
{
"msg_contents": "> \n> Is there a standard that defines how to return information from an embedded\n> SQL statement? ORACLE uses the sqlca. In particular it uses sqlca.sqlerrd[2]\n> for affected tuples. Is this correct with the standards? How do other\n> systems do that?\n\nI believe this is standard for embedded SQL.\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 13:53:32 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] standards question"
},
{
"msg_contents": "Michael Meskes wrote:\n> \n> Is there a standard that defines how to return information from an \n> embedded SQL statement? ORACLE uses the sqlca. In particular it uses \n> sqlca.sqlerrd[2] for affected tuples. Is this correct with the \n> standards? How do other systems do that?\n\nThere is a part of the SQL92 standard which deals with this. My Date &\nDarwen book has a chapter on it.\n\n - Tom\n",
"msg_date": "Wed, 18 Mar 1998 06:56:13 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] standards question"
},
{
"msg_contents": "Thomas G. Lockhart writes:\n> There is a part of the SQL92 standard which deals with this. My Date &\n> Darwen book has a chapter on it.\n\nCould you please enlighten me Tom? :-)\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 Mar 1998 09:23:33 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] standards question"
},
{
"msg_contents": "> Could you please enlighten me Tom? :-)\n\n*grin* As you could see I was trying to get out of doing work...\n\nOK, here is a short series of extracts from the book:\n\nSummary list of SQL statements having to do with SQL descriptor areas:\n ALLOCATE DESCRIPTOR, DEALLOCATE DESCRIPTOR\n GET DESCRIPTOR, SET DESCRIPTOR\n DESCRIBE\n\n\"SQLDA\" is a useful abbreviation (for the SQL descriptor area) but it is\nnot an official standard term. Note clearly, moreover, that several SQL\nproducts do already support some kind of SQLDA, and even refer to it by\nthat name, but those SQLDAs are typically not the same as the \"SQLDA\"\ndefined in the standard. In particular, the SQLDA in the standard is\n*encapsulated* (meaning that its internal structure is not specified in\nthe standard and is thus *a fortiori* hidden from the user), whereas the\nsame is typically not true of the SQLDAs in current products.\n\nWe list below some of the more important of these components (the\nretrievable and settable components of the SQL92 SQLDA):\n\nNAME\tcharacter string\nUNAMED\texact numeric with scale 0 (tgl- I'll call this int)\nTYPE\tint\nLENGTH\tint\nRETURNED_LENGTH int\nRETURNED_OCTET_LENGTH int\nPRECISION int\nSCALE\tint\nDATA\t(depends on type, length, etc)\nINDICATOR int\n\nIf the INDICATOR value is negative, the DATA value is undefined\n(\"null\"). The UNAMED component is set to 1 if the NAME value is\nimplementation-dependent (loosely, if the corresponding column is\nunnamed).\n\nSyntax:\n ALLOCATE DESCRIPTOR [LOCAL|GLOBAL] descriptor [WITH MAX occurences]\n\n(descriptor is specified inside single quotes). The GLOBAL is default if\nnot specified.\n\nThere are several ways to populate a descriptor:\n\n1) Specify\n INTO SQL DESCRIPTOR descriptor\nin an EXECUTE statement (in which case the prepared statement must be a\nsingle-row select).\n\n2) Specify the same in a dynamic FETCH statement.\n\n3) Specify\n USING SQL DESCRIPTOR descriptor\nin a DESCRIBE INPUT or DESCRIBE OUTPUT statement. This is the\nrecommended method for the most highly generalized type of application.\n\n4) Execute\n SET DESCRIPTOR descriptor ...\n\nEach such statement will assign values to components within a *single\nitem descriptor* within the SQLDA.\n\n...\n\nGET DESCRIPTOR descriptor target = COUNT\n\nwhere target is a parameter or host variable of type exact numeric with\nscale 0 (int??).\n\n\nMy book has quite a bit more information on this, but much more will be\ntoo much to type :(\n\n - Tom\n",
"msg_date": "Wed, 18 Mar 1998 16:07:05 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] standards question"
},
{
"msg_contents": ">>>>> \"meskes\" == Michael Meskes <[email protected]> writes:\n\n > Is there a standard that defines how to return information from\n > an embedded SQL statement? ORACLE uses the sqlca. In particular\n > it uses sqlca.sqlerrd[2] for affected tuples. Is this correct\n > with the standards? How do other systems do that? \n\n sqlca is fairly standard in most databases. It was the recommended method of\ndoing exception handling in the past. The SQL2 (SQL/92) standard went\nto using SQLSTATE variable and GET DIAGNOSISTICS statement.\n\n > Michael -- Dr. Michael Meskes, Project-Manager | topsystem\n > Systemhaus GmbH [email protected] | Europark A2,\n > Adenauerstr. 20 [email protected] | 52146 Wuerselen Go SF49ers!\n > Go Rhein Fire! | Tel: (+49) 2405/4670-44 Use Debian GNU/Linux!\n > | Fax: (+49) 2405/4670-10\n\n-- \nKent S. Gordon\nArchitect\niNetSpace Co.\nvoice: (972)851-3494 fax:(972)702-0384 e-mail:[email protected]\n",
"msg_date": "Wed, 18 Mar 1998 13:28:44 -0600 (CST)",
"msg_from": "\"Kent S. Gordon\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] standards question"
}
] |
[
{
"msg_contents": "> On Tue, 17 Mar 1998, Andrew Martin wrote:\n> > Geesh! Why the secrecy? Surely the more people who test it the\n> better. Look\n> > at the number of bugs that have turned up in the last 2 weeks since\n> 6.3\n> > was released - none of them *terribly* serious, but a fair number\n> none\n> > the less.\n> > \n> > Has Marc suddenly taken some sort of commercial attitude???? (Give\n> the\n> > customer a product - who cares if it hasn't been tested by as many\n> > people as possible - we can charge for an upgrade :-)\n> \n> \tNo, Marc's attitude hasn't changed at all...pgsql-hackers is the\n> place where *development* issues take place, and it is the *only* list\n> where discussing *any* pre-release software. If someone on\n> pgsql-questions (a *GENERIC*, how to *USE* PostgreSQL list) is\n> interested,\n> they are more then welcome to join pgsql-hackers...\n> \nHey, I resemble that remark. But seriously I don't understand the need,\nin limiting the discussion of new features or patches to just this list.\nThere are HACKERS-type things that pass through the QUESTIONS list all\nthe time (usually posted to both). Of course I'll admit that I've never\ntried to manage anything with this scope or deal with the standard\nPostgres users (I know we can be annoying), so I do see the potential\nfor the abuse. (Start accent here --) But here in Texas we don't shoot\nthe horse until he's broken his leg (-- end accent).\n",
"msg_date": "Tue, 17 Mar 1998 11:11:57 -0600",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] First mega-patch..."
},
{
"msg_contents": "On Tue, 17 Mar 1998, Jackson, DeJuan wrote:\n\n> Hey, I resemble that remark. But seriously I don't understand the need,\n> in limiting the discussion of new features or patches to just this list.\n> There are HACKERS-type things that pass through the QUESTIONS list all\n\n\tRight, and they should be redirected as appropriate, *to* the\nappropriate list...\n\n> the time (usually posted to both). Of course I'll admit that I've never\n> tried to manage anything with this scope or deal with the standard\n> Postgres users (I know we can be annoying), so I do see the potential\n> for the abuse. (Start accent here --) But here in Texas we don't shoot\n> the horse until he's broken his leg (-- end accent).\n\n\tI'd prefer to never have to shoot the horse myself...waste of a\nfine animal :)\n\n\tSeriously, at least in my case, I run procmail to filter my\nmail...I tend to read everything that goes to hackers, while skim through\nthe other lists as appropriate...I'm subscribed to *all* the lists, which\nmakes for several hundred messages per day...I like to automate the\nsorting as much as possible...\n\n\n",
"msg_date": "Tue, 17 Mar 1998 12:37:20 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] First mega-patch..."
},
{
"msg_contents": "> > \tNo, Marc's attitude hasn't changed at all...pgsql-hackers is the\n> > place where *development* issues take place, and it is the *only* list\n> > where discussing *any* pre-release software. If someone on\n> > pgsql-questions (a *GENERIC*, how to *USE* PostgreSQL list) is\n> > interested,\n> > they are more then welcome to join pgsql-hackers...\n> > \n> Hey, I resemble that remark. But seriously I don't understand the need,\n> in limiting the discussion of new features or patches to just this list.\n> There are HACKERS-type things that pass through the QUESTIONS list all\n> the time (usually posted to both). Of course I'll admit that I've never\n> tried to manage anything with this scope or deal with the standard\n> Postgres users (I know we can be annoying), so I do see the potential\n> for the abuse. (Start accent here --) But here in Texas we don't shoot\n> the horse until he's broken his leg (-- end accent).\n\nI think it is just an attempt to limit the number of postings to the\nquestions group, so it does not get to busy and people drop off.\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 13:59:37 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] First mega-patch...h"
}
] |
[
{
"msg_contents": "> \tRight, and they should be redirected as appropriate, *to* the\n> appropriate list...\n> \nThat systems seems to be working well. But, the fact that we don't yell\nat someone for posting a question to the hackers list or a bug/fix to\nthe questions list, but gently redirect the discussion from that list to\nthe appropriate one, while letting the original list listen in for\npossible new information, promotes participation. \n\n(Start favorite inspirational music here --) Maybe I think of the future\ntoo much, but I think that it's good for other's to not feel excluded.\nThere are walls and there are boundaries. (-- Musical climax --)\nBoundaries guide, while walls separate. (-- music fades)\n\nWell, like Tom needs people to help with the docs, which I plan on doing\nif 'I ever get the time (read as \"If there is an act of God to change\ndays to 29 hours instead of 24\")'. Otherwise, the only people you'd\never get on the hackers list are crazy coders with little or no social\nlife, like me, who like clear watch cases so they can see all the moving\nparts.\n\n> \tI'd prefer to never have to shoot the horse myself...waste of a\n> fine animal :)\n> \nWell now they have all these techniques so that the leg can be set and\nthe horse can live a fairly normal life afterwards, but that's a whole\nother topic.\n\n> \tSeriously, at least in my case, I run procmail to filter my\n> mail...I tend to read everything that goes to hackers, while skim\n> through\n> the other lists as appropriate...I'm subscribed to *all* the lists,\n> which\n> makes for several hundred messages per day...I like to automate the\n> sorting as much as possible...\n> \nI understand, the desire to sort your mail as well as the need to\nredirect the questions. But I don't see the harm in letting a standard\nuser go googlie-eyed over the technical prowess and dedication of the\nPostgres Developers.\n\nJust my $0.02,\n\t\t-DEJ (who now has $0.00)\n",
"msg_date": "Tue, 17 Mar 1998 12:25:23 -0600",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] First mega-patch..."
},
{
"msg_contents": "On Tue, 17 Mar 1998, Jackson, DeJuan wrote:\n\n> > \tRight, and they should be redirected as appropriate, *to* the\n> > appropriate list...\n> > \n> That systems seems to be working well. But, the fact that we don't yell\n> at someone for posting a question to the hackers list or a bug/fix to\n> the questions list, but gently redirect the discussion from that list to\n> the appropriate one, while letting the original list listen in for\n> possible new information, promotes participation. \n\n\tI don't yell at anyone...except those on pgsql-hackers :) \n\n> > \tI'd prefer to never have to shoot the horse myself...waste of a\n> > fine animal :)\n> > \n> Well now they have all these techniques so that the leg can be set and\n> the horse can live a fairly normal life afterwards, but that's a whole\n> other topic.\n\n\tExactly...but think how you feel laid up with a cast, unable to\nromp in the field *grin*\n\n> I understand, the desire to sort your mail as well as the need to\n> redirect the questions. But I don't see the harm in letting a standard\n> user go googlie-eyed over the technical prowess and dedication of the\n> Postgres Developers.\n\n\tI don't either...they are most welcome to join us over here...\n\n",
"msg_date": "Tue, 17 Mar 1998 13:41:04 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] First mega-patch..."
}
] |
[
{
"msg_contents": "Hi,\n\nUsing the ElectringFence debugging library I found what seems to be a\nbuffer overrun when creating databases. I don't think I understand it all\nwell enough\n(yet?) to fix it however.\n\nDoing a \"create database test;\" causes the buffer overrun to appear.\nWithout the debugging library I notice no malfunction.\n\nIn the function createdb a call to the function pg_exec_query which\ninserts the info about the new database into pg_database eventually\ndies in the function DataFill called from heap_formtuple.\n\nI've noticed that the function pg_exec_query appears twice on the\ncallstack so I guess this function is reentrant right?\n\nI hope this is useful to somebody.\n\nWith regards from Maurice.\n\n\n",
"msg_date": "Tue, 17 Mar 1998 19:41:58 +0100",
"msg_from": "\"Maurice Gittens\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Buffer overrun"
}
] |
[
{
"msg_contents": "> \tI don't either...they are most welcome to join us over here...\nOk I give. Uncle. Uncle! UNCLE!!!\n\n",
"msg_date": "Tue, 17 Mar 1998 12:45:34 -0600",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] First mega-patch..."
}
] |
[
{
"msg_contents": "\nThings have been nice and quiet, so here is a patch that, unless anything\nelse goes in in between now and Sunday, will be the official post-release\npatch...if something is missing from it, you only hvae yourself to blame\nfor not testing it :)\n\nftp.postgresql.org/pub/patches/.test/v6.3p1beta2.gz\n\nHidden like the last one :) Should apply easier, as its a true patch this\ntime\n\n\n",
"msg_date": "Tue, 17 Mar 1998 15:35:41 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "New patch to try..."
},
{
"msg_contents": "> \n> \n> Things have been nice and quiet, so here is a patch that, unless anything\n> else goes in in between now and Sunday, will be the official post-release\n> patch...if something is missing from it, you only hvae yourself to blame\n> for not testing it :)\n> \n> ftp.postgresql.org/pub/patches/.test/v6.3p1beta2.gz\n> \n> Hidden like the last one :) Should apply easier, as its a true patch this\n> time\n> \n\nBoth David and I have patches ready for tonight. David's is for\nspinlock looping, and mine will test UNIONs to make sure each query has\nthe same number of target fields and the same types:\n\ntest=> select usename from pg_user union select usename, usename from pg_user;\nERROR: Each UNION query must have the same number of columns.\ntest=> select usename from pg_user union select usesysid from pg_user;\nERROR: Each UNION query must have identical target types.\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": "Tue, 17 Mar 1998 17:06:22 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] New patch to try..."
}
] |
[
{
"msg_contents": "> \n> This message is in MIME format. The first part should be readable text,\n> while the remaining parts are likely unreadable without MIME-aware tools.\n> Send mail to [email protected] for more info.\n> \n> --0-295419167-890168702=:27180\n> Content-Type: TEXT/PLAIN; charset=US-ASCII\n> \n> -----BEGIN PGP SIGNED MESSAGE-----\n> \n> The attached two patches should be all that's required to get 6.3\n> initdb running properly on a DEC Alpha. These patches have been\n> tested against versions 3.2 and 4.0 of Digital Unix.\n> \n> The two patched files are template/alpha and backend/main/main.c\n> \n> (This is my first patch submission, so be gently with me <G>)\n\nCongratulations. Certainly has to be put in the coming mega-patch. \nStrange that the fix is so small, just some config lines, and a #define\nASSEMBLER around sys/proc.h. Amazing.\n\nGlad the cause wasn't some of my code additions.\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 17:09:22 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [PATCHES] Fixes for initdb on DEC Alpha"
},
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nOn Tue, 17 Mar 1998, Bruce Momjian wrote:\n\n> \n> Congratulations. Certainly has to be put in the coming mega-patch. \n> Strange that the fix is so small, just some config lines, and a #define\n> ASSEMBLER around sys/proc.h. Amazing.\n\nActually, all that my \"patch\" does it to enable a pre-existing\npatch. With -Dalpha, the macro LONGALIGN() is defined to aling,\nnot on long boundaries as you would expect, but on int\nboundaries. Once I found that, it was clear how to fix the\nproblem. The comments in the header file were quite clear that\nthis was to prevent exactly the sorts of problems that we were\nseeing.\n\n> \n> Glad the cause wasn't some of my code additions.\n> \n\nNo additions at all. Just a removal of a #define\n\n- --\nDwayne Bailey + WHAT is your name? Sir Galahad\nMIKA Systems, Bingham Farms, MI + WHAT is your quest? I Seek the Holy Grail\[email protected] + What is your favorite color?\nhttp://www.mika.com/~dwayne + Blue ... no, Yelloooooooooooooooooow\n finger [email protected] for PGP Public Key\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQB1AwUBNQ71/aA2uleK7maRAQGnWwL/dJegGLZZW47G0MyBUiF8kMU1ujX+odRk\nrvSiJTjuNpqKb8Kxty8srMyzOagQfggAndDLCvT2Pt4+oZcMdguj6Kx33Lv+24Xa\n6mA0Mm+Y9M0lcuLNCmvf63earPTwy5I1\n=ryQD\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "Tue, 17 Mar 1998 17:15:16 -0500 (EST)",
"msg_from": "Dwayne Bailey <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [PATCHES] Fixes for initdb on DEC Alpha"
}
] |
[
{
"msg_contents": "hello,\n\nthis has been mentioned recently, although i didn't find a solution.\nthe date/time 'now' default still shows the table creation time/date in\n6.3.\n\nany thoughts of why?\n\n-brian rickabaugh\n\n\n\n",
"msg_date": "Tue, 17 Mar 1998 23:09:07 -0500",
"msg_from": "\"PostgreSQL On Peanuts.roanoke.edu\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "default time/data 'now' broken in 6.3? linux-elf 2.0.33, i686 "
}
] |
[
{
"msg_contents": "Hi everybody,\n\nThe electricfence library is a tool which helps us to detect\nbuffer overruns in C programs.\n\nLinking Postgresql with this library gives at least 3 errors I have found\nuntil now.\n\nThe errors manifest themselves in sql statements like:\n1. create table test(f1 int);\nThis one allways fails.\n2. select * from pg_shadow\nOnly some tables give errors so the problem seems to be related\nto the type/alignment of fields in the tables.\n\nThe errors are all found in the DataFill function called from\nheap_from_tuple.\n\nI'm using a snapshot version of postgresql downloaded around\nthe release date 6.3.\n\nThis all is on Linux (Redhat 4.0) with Electric Fence version 2.0.5.\n\nDo you guys feel comfortable with a database which has reproducible\nbuffer overuns?\n\nI think it would be great is somebody who has more experience with\npostgresql than I were to look at this.\n\nI am now going to do some work which I actually get paid to do!\n\nWith regards from Maurice.\n\n\n",
"msg_date": "Wed, 18 Mar 1998 12:33:49 +0100",
"msg_from": "\"Maurice Gittens\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Buffer overuns with the Electric fence debugging library"
},
{
"msg_contents": "> \n> Hi everybody,\n> \n> The electricfence library is a tool which helps us to detect\n> buffer overruns in C programs.\n> \n> Linking Postgresql with this library gives at least 3 errors I have found\n> until now.\n> \n> The errors manifest themselves in sql statements like:\n> 1. create table test(f1 int);\n> This one allways fails.\n> 2. select * from pg_shadow\n> Only some tables give errors so the problem seems to be related\n> to the type/alignment of fields in the tables.\n> \n> The errors are all found in the DataFill function called from\n> heap_from_tuple.\n> \n> I'm using a snapshot version of postgresql downloaded around\n> the release date 6.3.\n> \n> This all is on Linux (Redhat 4.0) with Electric Fence version 2.0.5.\n> \n> Do you guys feel comfortable with a database which has reproducible\n> buffer overuns?\n> \n> I think it would be great is somebody who has more experience with\n> postgresql than I were to look at this.\n> \n> I am now going to do some work which I actually get paid to do!\n> \n> With regards from Maurice.\n> \n> \n> \n> \n\nCan you send us the lines that cause the problem? I will look at them. \nI have electric fence here on BSDI, but it really takes some time to\nrun.\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 10:46:48 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Buffer overuns with the Electric fence debugging\n library"
},
{
"msg_contents": "On Wed, 18 Mar 1998, Maurice Gittens wrote:\n\n> Do you guys feel comfortable with a database which has reproducible\n> buffer overuns?\n\n\tNope, but only a very few ppl appear to have the software to\ndetect them, and with the size of the code base for this, we sort of rely\non those with those tools to help out...\n\n\tEfforts you can provide in this area would be much appreciated,\neven if only insofar as providing points to sections of code that should\nbe checked...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Thu, 19 Mar 1998 00:12:08 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Buffer overuns with the Electric fence debugging\n library"
},
{
"msg_contents": "Marc G. Fournier writes:\n> On Wed, 18 Mar 1998, Maurice Gittens wrote:\n> \n> > Do you guys feel comfortable with a database which has reproducible\n> > buffer overuns?\n> \n> \tNope, but only a very few ppl appear to have the software to\n> detect them, and with the size of the code base for this, we sort of rely\n> on those with those tools to help out...\n> \n> \tEfforts you can provide in this area would be much appreciated,\n> even if only insofar as providing points to sections of code that should\n> be checked...\n\nI agree with Marcs comments here, but would like to add that ElectricFence is\nfreely available at least on Linux systems. It is a long way from Purify,\nbut it beats sitting around scratching your head.\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": "Wed, 18 Mar 1998 20:29:01 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Buffer overuns with the Electric fence debugging\n library"
},
{
"msg_contents": "On Wed, 18 Mar 1998, David Gould wrote:\n\n> I agree with Marcs comments here, but would like to add that ElectricFence is\n> freely available at least on Linux systems. It is a long way from Purify,\n> but it beats sitting around scratching your head.\n\n\tURL for the software? I'd like to see if they at least have a\nFreeBSD port that I can use :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Fri, 20 Mar 1998 02:18:43 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Buffer overuns with the Electric fence debugging\n library"
},
{
"msg_contents": "> \n> On Wed, 18 Mar 1998, David Gould wrote:\n> \n> > I agree with Marcs comments here, but would like to add that ElectricFence is\n> > freely available at least on Linux systems. It is a long way from Purify,\n> > but it beats sitting around scratching your head.\n> \n> \tURL for the software? I'd like to see if they at least have a\n> FreeBSD port that I can use :)\n\nI just got it off the CD myself, but I ftp.redhat.com has it in the\nredhat 5.0 distribution area. Look in the SRPMS for:\n\nElectricFence-2.0.5-5.src.rpm \n\nOr, if you don't have rpm, AltaVista turned up some bsd entries too.\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 01:20:02 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Buffer overuns with the Electric fence debugging\n library"
}
] |
[
{
"msg_contents": "I just compiled and ran my first ecpg program with indicator variables. It\nneeds lots of work before I can submit a patch, and I will have to increase\nthe major number of libecpg. Anyway, I used indicator variables with\nthe following syntax:\n\nselect ... into :variable:indicator\n\nNow I checked Oracle and it accepts\n\nselect ... into :variable indicator :indicator \n\nas well. I like this because I think this it is better readable. However,\nI'm not sure which version the standard supports. Currently ecpg eaccepts\nboth.\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, 18 Mar 1998 13:56:58 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "another standards question"
},
{
"msg_contents": "> ... I used indicator variables with\n> the following syntax:\n> \n> select ... into :variable:indicator\n> \n> Now I checked Oracle and it accepts\n> \n> select ... into :variable indicator :indicator\n> \n> as well. I like this because I think this it is better readable.\n> I'm not sure which version the standard supports. Currently ecpg \n> accepts both.\n\nGood. It looks like the standard allows both, including spaces between\n\":variable\" and \":indicator\" even when \"INDICATOR\" is omitted :(\n",
"msg_date": "Wed, 18 Mar 1998 16:23:26 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] another standards question"
},
{
"msg_contents": "Thomas G. Lockhart writes:\n> Good. It looks like the standard allows both, including spaces between\n> \":variable\" and \":indicator\" even when \"INDICATOR\" is omitted :(\n\nNo problem. That's the way it's implemented right now. I just have to do\nsome parser rewriting to get rid of a shift/reduce conflict when allowinf\nindicators in inserts etc.\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, 19 Mar 1998 11:41:06 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] another standards question"
}
] |
[
{
"msg_contents": "> \n> The errors are all found in the DataFill function called from\n> heap_from_tuple.\n> \n\nWhere specifically in DataFill?\n\n> \n> Do you guys feel comfortable with a database which has reproducible\n> buffer overuns?\n> \n> I think it would be great is somebody who has more experience with\n> postgresql than I were to look at this.\n> \n\nThis might be by design. Check out include/access/htup.h and look at\nhow the tuple is defined. Basically, there is a struct with the \nnecessary header fields that is used to cast an area of memory that\nhas been [p]alloc'd. See heap_addheader as an example...\n\nNote that this technique has been declared as \"Bad\" by the C standards\ncommittee, but it works and there are other things more broken.\n\ndarrenk\n",
"msg_date": "Wed, 18 Mar 1998 10:39:51 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Buffer overuns with the Electric fence debugging\n library"
}
] |
[
{
"msg_contents": "Hi,\n\nAs you requested I've included more information about the buffer overruns\ndetected by Electric Fence.\n\nFirst (in postgres interactive mode) I enter the statement: \"create database ipay;\" Here stack dump of the situation just before the crash.\nThe memmove call gives buffer overrun. \n\n177 data_length = VARSIZE(DatumGetPointer(value[i]));\n178 memmove(data, DatumGetPointer(value[i]), data_length);\n179 data += data_length;\n180 break;\n181 case sizeof(char):\n\nHere is a stacktrace.\n\n#0 DataFill (data=0x408c8fd4 \"\", tupleDesc=0x4088aff4, value=0x40896ff4,\n nulls=0xbfffd5c0 \" @\", infomask=0x408c8fcc, bit=0x0) at heaptuple.c:178\n#1 0x80600d0 in heap_formtuple (tupleDescriptor=0x4088aff4, value=0x40896ff4,\n nulls=0xbfffd5c0 \" @\") at heaptuple.c:860\n#2 0x8098940 in ExecTargetList (targetlist=0x4085aff4, nodomains=3,\n targettype=0x4088aff4, values=0x40896ff4, econtext=0x40888fd4,\n isDone=0xbfffd653 \"\\001h���fo\\t\\b,o\\207@�\\217\\207@\\003\") at execQual.c:1576\n#3 0x80989aa in ExecProject (projInfo=0x40894fe8,\n isDone=0xbfffd653 \"\\001h���fo\\t\\b,o\\207@�\\217\\207@\\003\") at execQual.c:1643\n#4 0x809e260 in ExecResult (node=0x40876fb8) at nodeResult.c:163\n#5 0x8096f66 in ExecProcNode (node=0x40876fb8, parent=0x40876fb8)\n at execProcnode.c:249\n#6 0x80965b4 in ExecutePlan (estate=0x4087afcc, plan=0x40876fb8,\n parseTree=0x404fcfbc, operation=CMD_INSERT, numberTuples=0,\n direction=ForwardScanDirection, printfunc=0x80615d0 <debugtup>)\n at execMain.c:727\n#7 0x80960f1 in ExecutorRun (queryDesc=0x40878ff0, estate=0x4087afcc,\n feature=3, count=0) at execMain.c:236\n#8 0x80e882b in ProcessQueryDesc (queryDesc=0x40878ff0) at pquery.c:332\n#9 0x80e8896 in ProcessQuery (parsetree=0x404fcfbc, plan=0x40876fb8,\n argv=0x0, typev=0x0, nargs=0, dest=Debug) at pquery.c:378\n#10 0x80e755d in pg_exec_query_dest (\n query_string=0xbfffd9d0 \"insert into pg_database (datname, datdba, datpath)\nvalues ('ipay', '500', 'ipay');\", argv=0x0, typev=0x0, nargs=0, dest=Debug)\n\nThe query \"select * from pg_shadow\" crashes in the same function. However\nthis time the offending line is the one with the folling assigment.\n\n195 *(int32 *) data = (att[i]->attbyval ?\n(gdb)\n\nProgram received signal SIGSEGV, Segmentation fault.\n0x805f524 in DataFill (data=0x40acafd4 \"\", tupleDesc=0x40ab2ff4,\n value=0x40ac8fe0, nulls=0xbfffda78 \" n \\001\\003\\021\\b\",\n infomask=0x40acafcc, bit=0x40acafcf \"�\") at heaptuple.c:195\n195 *(int32 *) data = (att[i]->attbyval ?\n(gdb)\n\nThe stack trace for this one follows:\n\n#0 DataFill (data=0x40acafd4 \"\", tupleDesc=0x40ab2ff4, value=0x40ac8fe0,\n nulls=0xbfffda78 \" n \\001\\003\\021\\b\", infomask=0x40acafcc,\n bit=0x40acafcf \"�\") at heaptuple.c:195\n#1 0x80600d0 in heap_formtuple (tupleDescriptor=0x40ab2ff4, value=0x40ac8fe0,\n nulls=0xbfffda78 \" n \\001\\003\\021\\b\") at heaptuple.c:860\n#2 0x8098940 in ExecTargetList (targetlist=0x40a84ff4, nodomains=8,\n targettype=0x40ab2ff4, values=0x40ac8fe0, econtext=0x40aaefd4,\n isDone=0xbfffdb0b \"\\001\\034U��+�\\t\\b,\\017\"@�a\\t\\b0U��\\206o\\t\\b,\\017\"@�/�@\\001\") at execQual.c:1576\n#3 0x80989aa in ExecProject (projInfo=0x40ac6fe8,\n isDone=0xbfffdb0b \"\\001\\034U��+�\\t\\b,\\017\"@�a\\t\\b0U��\\206o\\t\\b,\\017\"@�/�@\\001\") at execQual.c:1643\n#4 0x8098a8f in ExecScan (node=0x40a80fb8, accessMtd=0x809e3b0 <SeqNext>)\n at execScan.c:153\n#5 0x809e42b in ExecSeqScan (node=0x40a80fb8) at nodeSeqscan.c:138\n#6 0x8096f86 in ExecProcNode (node=0x40a80fb8, parent=0x40a80fb8)\n at execProcnode.c:261\n#7 0x80965b4 in ExecutePlan (estate=0x40aa4fcc, plan=0x40a80fb8,\n parseTree=0x404a1fbc, operation=CMD_SELECT, numberTuples=0,\n direction=ForwardScanDirection, printfunc=0x80615d0 <debugtup>)\n at execMain.c:727\n#8 0x80960f1 in ExecutorRun (queryDesc=0x40aa2ff0, estate=0x40aa4fcc,\n feature=3, count=0) at execMain.c:236\n#9 0x80e882b in ProcessQueryDesc (queryDesc=0x40aa2ff0) at pquery.c:332\n\nThe statement \"create sequence test;\" also gives a buffer overrun in the \nfunction the call to memmove in DataFill.\nSo it might be that this happens often.\n\nI hope this information is useful.\n\nThanks with regards from Maurice.\n",
"msg_date": "Wed, 18 Mar 1998 16:50:32 +0100",
"msg_from": "Maurice Gittens <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Buffer overruns with the Electric Fence debugging library"
}
] |
[
{
"msg_contents": "On Wed, 18 Mar 1998, David Harvey-George wrote:\n\n> I found that triggers don't get inherited... or rather the triggers \n> for the superclass don't get checked when inserting or updating rows \n> when enforcing referential intregrity. This is a bit of a pain as it \n> means I have to duplicate a lot of triggers. Any idea if this can be \n> fixed? Presumably this would mean 'fixing' to the TRIGGER handling \n> code to look up the inheritance stack calling the appropriate \n> triggers?\n> \n> Maybe like inheriting PRIMARY KEY attributes this is more of a \n> feature?\n\nI did ask about this last weekend (while looking at some features for\nJDBC). This looks like a similar problem.\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": "Wed, 18 Mar 1998 19:06:15 +0000 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [QUESTIONS] Inheriting Triggers"
}
] |
[
{
"msg_contents": "hello,\n\nwhile running regression tests on the above system, the horology test\nfailed. does anyone have an idea why?\n\nthankx,\n\n-brian rickabaugh\n\n([email protected])\n\n\n\n",
"msg_date": "Wed, 18 Mar 1998 20:04:49 -0500",
"msg_from": "\"PostgreSQL On Peanuts.roanoke.edu\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "regression testing on linux-elf 2.0.33, postgresql 6.3, i686 "
},
{
"msg_contents": "> while running regression tests on the above system, the horology test\n> failed. does anyone have an idea why?\n\nWhat is the symptom? Did you read the last item in the FAQ_Linux file?\nDoes it help, or are your problems different than described there??\n\n - Tom\n",
"msg_date": "Thu, 19 Mar 1998 03:14:41 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] regression testing on linux-elf 2.0.33, postgresql 6.3,\n\ti686"
}
] |
[
{
"msg_contents": "> > 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.\n\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\nI was afraid that would be the answer. Basically, this never worked. The\nproblem is that in an indeterminate number of places the code manipulates\nmultiple distinct but related structures in a sequence. To leave the server\nin a consistant state all these updates need to be done in the right order\nso that if the sequence in interrupted by a signal the partially updated\nstate is consistant. Unhappily, the original coders did not always have this\nin mind. An example:\n\n cleaning up after a scan\n 1 - release buffer pointed by scan descriptor\n 2 - release scan descriptor\n\nIf a signal is taken just after one, the abort code will see the scan\ndescriptor and call the cleanup for it resulting in:\n\ncleaning 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\nThese sequences either _all_ have to identified and fixed, or made atomic\nsomehow, which is a biggish job.\n\nOr the system has to acknowledge signals at only clearly defined points.\n\nMy preference is to have signal handlers only set a global flag to indicate\nthat a signal was seen. Then one just sprinkles check_for_interrupts() calls\nin all the likely places: step to next node, fetch next page, call function,\netc.\n\nThe way this shows up in real life is strange unreproduceable errors on busy\nsystems, especially when backends are killed or transactions are forced to\nabort.\n\nFixing this is a bit of a project, but the result is that a lot of mystery\nbugs go away.\n\n> I considered the possibility that your select() could return EINTR, so I\n> was thinking of suggesting something like\n> \n> \tdo { } while (lock-no-set && (select(),true) )\n> \n> or something like that so the return value of select is always true. I\n> also recommend making a S_LOCK macro, and inside the while loop, call\n> the OS-specific lock stuff, so you don't have to code the select() for\n> each platform. So you have two macros, the S_LOCK macro with the while,\n> and inside the while, you call S_LOCK_SPIN() which is defined for each\n> platform. More centralized and less error-prone. \n\nYou may like what I have done then. S_LOCK() becomes a function:\n\nS_LOCK(lock)\n{\n do {\n while (!S_LOCK_FREE(lock)) /* non interlocked test to avoid */\n { /* hammering the bus and caches */\n select( a_semi_random_time_delay_with_backoff );\n }\n } while (TAS(lock));\t\t/* TAS: test and set\n}\n\nS_LOCK_FREE(), and TAS() are platform specific macros that invoke the\nplatform specific implementation. There are default implementations for\nall the platform interface macros, so most of the time all that is needed\nis to create a tas() function in asm and the rest falls into place.\n\nNow I have to go home and test it.\n\nBtw, feel free to forward this to the list if you feel it is of interest. I\ndidn't because I didn't want to quote you from private mail.\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": "Wed, 18 Mar 1998 19:36:05 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: [PATCHES] patches for 6.2.1p6"
}
] |
[
{
"msg_contents": "I'm planning to modify some string functions so that they would be\naware of multi-byte strings if compiled with the multi-byte\ncapability. Followings are files I'm going to modify. I would like to\nhear your opinions if you have any.\n\no character_length()\n\nIt seems that the function is implemented as textlen() in\nutils/adt/varlena.c or as varcharlen() in varchar.c. Current\nimplementaion returns an octet length rather than a char length. So I\nwill change them. However, there might be necessity for getting an\noctet length in some applications. Maybe this is a good chance to add\nSQL92's octet_length().\n\no lower()/upper()\n\nImplemented in oracle_compat.c. One thing I have noticed is that it\nuses toupper()/tolower(). For ASCII, they are fine. But on some\nplatforms (I guess SysV) they might have some problems:\n\n\tchar c;\t/* c is an 8-bit letter and this platform uses char as\n\t\t signed char */\n\ttoupper(c);\t/* may cause segfault or any other bad thing */\n\nSo I will change like:\n\n\ttoupper((unsigned char)c);\n\no position()\n\nImplemented as textpos() in varlena.c.\n\no substring()\n\nImplemented as text_substr() in varlena.c.\n\n--\nTatsuo Ishii\[email protected]\n",
"msg_date": "Thu, 19 Mar 1998 13:09:58 +0900",
"msg_from": "[email protected]",
"msg_from_op": true,
"msg_subject": "multi-byte aware char_length() etc."
},
{
"msg_contents": "> I'm planning to modify some string functions so that they would be\n> aware of multi-byte strings if compiled with the multi-byte\n> capability. Followings are files I'm going to modify. I would like to\n> hear your opinions if you have any.\n> \n> o character_length()\n> \n> It seems that the function is implemented as textlen() in\n> utils/adt/varlena.c or as varcharlen() in varchar.c. Current\n> implementaion returns an octet length rather than a char length. So I\n> will change them. However, there might be necessity for getting an\n> octet length in some applications. Maybe this is a good chance to add\n> SQL92's octet_length().\n\nYes.\n\n> o lower()/upper()\n> \n> Implemented in oracle_compat.c. One thing I have noticed is that it\n> uses toupper()/tolower(). For ASCII, they are fine. But on some\n> platforms (I guess SysV) they might have some problems:\n> \n> char c; /* c is an 8-bit letter and this platform uses char as\n> signed char */\n> toupper(c); /* may cause segfault or any other bad thing */\n> \n> So I will change like:\n> \n> toupper((unsigned char)c);\n\nI would like to move these routines, as you clean them up, to varlena.c\nor whatever Postgres-specific source file is appropriate. Let's leave\noracle_compat.c for non-standard, Oracle-specific functions. Perhaps\neventually we can move any of those which remain to the contrib\ndirectory, assuming that there are good equivalent functions available\nin SQL92. \n\nSort of annoying having oracle_compat when Oracle doesn't return the\nfavor by having a \"postgres_compat\". Well, maybe DataBlades are the same\nthing?? :)\n\n> o position()\n> \n> Implemented as textpos() in varlena.c.\n> \n> o substring()\n> \n> Implemented as text_substr() in varlena.c.\n\nThese two are OK. I'm not yet clear on where in the parser these varlena\nfunctions are matched up with both text and varchar() types. We may need\nto do something different as we keep working on getting the\ntext/varchar/char behavior improved.\n",
"msg_date": "Thu, 19 Mar 1998 05:50:40 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] multi-byte aware char_length() etc."
},
{
"msg_contents": "Hi, here are patches I promised (against 6.3.2):\n\n* character_length(), position(), substring() are now aware of \n\t multi-byte characters\n* add octet_length()\n* add --with-mb option to configure\n* new regression tests for EUC_KR\n (contributed by \"Soonmyung. Hong\" <[email protected]>)\n* add some test cases to the EUC_JP regression test\n* fix problem in regress/regress.sh in case of System V\n* fix toupper(), tolower() to handle 8bit chars\n\nnote that:\n\no patches for both configure.in and configure are\nincluded. maybe the one for configure is not necessary.\n\no pg_proc.h was modified to add octet_length(). I used OIDs\n(1374-1379) for that. Please let me know if these numbers are not\nappropriate.\n\n>> I'm planning to modify some string functions so that they would be\n>> aware of multi-byte strings if compiled with the multi-byte\n>> capability. Followings are files I'm going to modify. I would like to\n>> hear your opinions if you have any.\n>> \n>> o character_length()\n>> \n>> It seems that the function is implemented as textlen() in\n>> utils/adt/varlena.c or as varcharlen() in varchar.c. Current\n>> implementaion returns an octet length rather than a char length. So I\n>> will change them. However, there might be necessity for getting an\n>> octet length in some applications. Maybe this is a good chance to add\n>> SQL92's octet_length().\n>\n>Yes.\n>\n>> o lower()/upper()\n>> \n>> Implemented in oracle_compat.c. One thing I have noticed is that it\n>> uses toupper()/tolower(). For ASCII, they are fine. But on some\n>> platforms (I guess SysV) they might have some problems:\n>> \n>> char c; /* c is an 8-bit letter and this platform uses char as\n>> signed char */\n>> toupper(c); /* may cause segfault or any other bad thing */\n>> \n>> So I will change like:\n>> \n>> toupper((unsigned char)c);\n>\n>I would like to move these routines, as you clean them up, to varlena.c\n>or whatever Postgres-specific source file is appropriate. Let's leave\n>oracle_compat.c for non-standard, Oracle-specific functions. Perhaps\n>eventually we can move any of those which remain to the contrib\n>directory, assuming that there are good equivalent functions available\n>in SQL92. \n>\n>Sort of annoying having oracle_compat when Oracle doesn't return the\n>favor by having a \"postgres_compat\". Well, maybe DataBlades are the same\n>thing?? :)\n>\n>> o position()\n>> \n>> Implemented as textpos() in varlena.c.\n>> \n>> o substring()\n>> \n>> Implemented as text_substr() in varlena.c.\n>\n>These two are OK. I'm not yet clear on where in the parser these varlena\n>functions are matched up with both text and varchar() types. We may need\n>to do something different as we keep working on getting the\n>text/varchar/char behavior improved.\n>\n\n----------------------------- cut here --------------------------\nbegin 644 cl.patch.gz\nM'XL(\"(>Z/C4``V-L+G!A=&@`[%QI6]O(LO[,_(7SI>W(89,7R8LDP`E@S+X-\nMD$G.\\\\P=CRS+6\"!+'DEFR3*__51UMQ8;V9@,2<@YERS8K>[JJK?>KJ[NEK3G\nM=,R[%3)P_>#2,_V_['RM4\"[(Q8YK%,^:&UM'S4*__4O'ZG9)WGBDVHI4H']F\nMK%:H%:1?EI:69JP^=^0Z9&/@$;E$I-)*N;Q2D8BD:>HO^7S^$1ES%[TA:ULF\nMDKQ2T5;*-=9V:?2'ZB.)%4*_9!)B\"8@E_:$=6/GV?6`N'&TNDH$>&#UR>B@3\nMUM,<(4>Z!]HQV83`WSGZ<Z$'_M`E>W[/LJ+\"(&_A]W7?TPN&6[@:4$NP=_B=\nMG]8[H=W[P\\'`]8*X<[#0LHDL?6WW:6#41$EE:%!Y>T[@N9VA$5BN@R49<K3)\nM<;!\\8CF!\"83JD*[K$=VVW5O+N22GS(SS7P])X)*>[G1L,V$,R#!ZNJ<;O<#T\nMB&\\&/E@&\\G2?--\\U%IIW5\"9YYUAWI.%VS$41/QOPB8`H<C2T401V[3FZ3?!\"\nM@;RW@AX)>F:LWKT[)(;ND*&?[)QW'?9,!1'/O#3O!E3ZX=Y!LT`N0)#I@&0T\nMQ[_W`[,/#5W?=-#JC@G-^Y8#=NL!=HH6N?V!!68&5M\\L,*!0\"%-%MWV7=*T[\nM$RQU^U#JN6W;[/O0RC'`\"NSE%@U0\\VTK(#Y\\!U$4*S*F,*AKV$.JU][YB:I6\nMM0)9V\".W[M#N$,>%QOH]>H*XW:@;$-+3;T\"@\"?JC&IT\"V2-70S]`!;J6UP=+\nM@AZS!<$`\"OK@;Q*84,73L9&#JB!`.NF:MV3;`WAZL6YH\"FK0-A'O#K,F\"`$H\nMD%/;U,$/MAE0SE.24=)/(-D%\\V3(^\"=1#9I/]'<:TSJ3J(9D![)-HYKIZ`!Q\nM9P+9(J8GO!>2342F440I);I#AYKOIY*/\"II(OS3R@6[_S[H$Z])\"G5P2*]5$\nMK-MU;Y%%((IAN-%%^_7!P+Y'RY/1122&9^I`+]\\SBD?ZM=D%_`L&&.?VF1HZ\nML5']3`S=2NB9>NC=%O<N[?VV9WJQX_DE]+?KF(`K;PV\\;>V?0C3?UP>Z8X)U\nM4,++&\\=0WNA98\\4'9U!\\X(*^3J+TXCV=)ZS;J/3=\\5[C9*L)Q7P,++R[V,ZK\nMBZCSN\\-F:^_XHGEVO'$X-S<R*ICVS3N]/[!-KF4.!D-`QG!AMC,#6+6]+@(*\nM)B*%.B;ZNR/BEQ[B#>7@:><2'&O>&>8@`#\\'E)$L*F`=\"`8@Z@&!Q^A;8#J>\nMF5T3&63Z-`B!^ZMR(@J-NW\\V#S_P[UR*@S-X`>*6-W08^2^'X.N(KOTV<0<X\nM^E=8S;E<HE8^C_7R_?8_H<W<)-[,32#.7#ISYM*I,S>%.W-/)<_<)/;,Q?1A\nM,%'G3\\8K63>B6L?R:<C^/CQ+\"SM55:R56=BA<V%55,LA#><H;\"S*T^%`T+U<\nM[JX%0'CWORP3^!OF?B)-_EB$-QVPPS#[IA/X11KYH>+<4JQERS:=RZ\"W`/,<\nMY)H64@X_^\\.V'WA@UL(BT0%\"!T:\"?HN?()ZC\")(ZHW+I>J=#7/@>1-+C\\L@9\nMG.`XP\")7L6H.!/:QZ._329XQ$\"H15&`!FH&.[6$`H+3O2?;<=9W^_=\"Y+,#(\nM=2ZS9*T'O];MH:-[EE\\`)/JF[W8#S'6OO3<)I2A6=)8Q8'KP42<<A8PNX[JP\nM9@!F.+'Q.1SK%/GO@M_#4I2&B)VS,?A;W#1P@4JFAU@'+J0N^#&1&ZM(*X35\nM9[[%185<Y9X%VN**@Z9`ANL!(W6$`$:0FTQ_<!4D,9ZP-0EOCFT]DTZ$'208\nM@V!\\N@WQWC\\5^53\"<NWDT/UE;Y:U(ZXK9EL^0LT95Y\")FC,N(A,M_LDZ$L3,\nMO0>P:7.9-J^!A&E+2>GQM229M)BD/RB[6\"[\"HO)?PN:*LU[XU\\)F<FTW!\\7K\nMIZ5NX]7>>W9ITO(.KT`2%@2#E6+Q]O:V$%TJ#DP7`FZ1MRG&/\"JRKJ`/X410\nM!5TXR,QC+QG\"\"LO\"<>[77\".W(9QDLKEWN:O,FK`/%<C9UN;1^<*9:>LXSB%9\nMWM(#O8TCXDAW]$L:E/C(6$1!QZ'4!(<7$KK&Z!5<[[+(VJSLO%5V=[`[1!2+\nM=K8_'_RNVEC$IB*FYKZP!^K]D;O.;>1.<T)N=U]=_2AX)45X#2;L\"IZ\\M5,7\nM#@1-N!;>@I7'0AO^[0A:YE5H8L)JL-83LL(V5,;&!T)/N,ID2_OR&C3\"$I&+\nMVQ>NA#]02.0SZM$\"RAA79E=0A.U,ENF?R3:ZVE^JC8H+Q^5M[2\\*$-H$;3-9\nM^,PGUO`KETTC\\]@2A2(@YLHY'^P0<M>K?PL'&TN;U.;,J]Q.9BWW-O=O^%\\#\nME4]0%MN'D.3TC8B\"_/A61$S=2E&67AAUQWF&50Z$)4$2KI-NR55S_P>H[`HG\nMW+4]8,O^=F-K7O!6OP@-((#P4[HW-5)I41;\".JI0,VAKH<X'Q('0?F\"8N+I[\nM]\"OV%6M9347@>+U:%??J:EW-K-RF-:NE-T-@Z.(8KJ8UI=Z$A`!A>\"NL\"PH%\nM3!;+8H5[=@O\\M\"V<\\$#%ANZ^4(3_VW2@[G`?+C)NC,X!5,;Q]M]K5_M?#C)Q\nM]!N;*0+=*UQ^9&Q#C1-#A@D`QT-'Q^K=6BG7REF`:Q5M%':2X30K7&/C+N4Y\nM_%^XTB')*^@&9?JP7>RT^WZ2X[35LF\"A?B`^'_.1BY3?\"A94.#[Z<E2&\"F7X\nMDZR^+PAO;.T2HW?8*.X\\,<B@:Z/?*896%<%DUG?[Z.[HU]2>0?82E3H.16XI\nM=Y53<_NYT%8:4QD:)Z/FCT;]2/).W!7\\'9V''KCJ:4;1:9A][K=A-BZPO0-P\nM;=2[1]�Q!H2H(HYJ$6E\"2+0MO1Z9+WO`MJP33908'K.`UU?(RE/\\Q8A:L\nM8BZ'SD=K0![H$EJ5QM1<G0[UR3*!J>2N^S&=P.%X0E$X&>Z.>3#J9Q?&U.O<\nM3NY5SLX5(59>\"2)EP37.FYDLA(RR<!\":;G1HFQ1MX,JH\\-\"T&$JO>0-0/FC)\nM\\J7\\0\")K8P\"1A`P6=ZDSMF'$KS.'K65@,D##TA;SL761?U8<G.#Y%(7_QC8N\nM6%<T4M&Y3KL0O*T]N9&PF*\\F*'\\J?PAZF\"_DRN\"N/8AI'[;&!C\";*,.%.K)G\nM)+4A\\5I]9Y/)C:(\\4*S1W5=9MSB%H$`Z5\\\"RE,A1E&(3O2)62N&\"\\R>*_,NT\nMJ5)XRFJ6!8_C9%>Q*,+C$)WTH[D%RP286[7<+I1?CXQGX/^[W\"+(@C#.0M=8\nM5G&2[`D<%&84\"L\\=5@5_)&:FIA_>45[]R))3C&N\"M]X^7$\\.)L[![T1FG*1`\nM(ILQXGT6J%DM;I<P^TBD2P>4U^,#GY#'=VBH&L@#/D(8:\"><_D\\<2QF>(*CU\nME99PO+:Y,42'9;()^2>C&932`QE5`5*U-4``2T(Y-)\"7FDO\"P?ZOFD:C).;\\\nM)[Q'<\"E-);['\\$U+Y#19U)3$=I(F:B.CFTT2.MAZ*ASO?\\%L()?-[5'.P;2Q\nMM4*M`!7HT@>\">*6V]FZ,;2RS^'OKZNCCL<=&!-VH&$OYJYF=.C(G\\Q:DE^K\"\nM>I75?.(6%-/XP9\"=HUILGDI_1U)'MYWP,G:/D,/'[!)^L!PKZ+293*$C7&V]\nM$90]\\>A]K%F2T319>9]3<I<YGS(`MU\\248*+9]HLF'>%*;3F.TYL&RL>8^-;\nM+['27.I3=[:HZ*K@;_FU#&5BU\"T?,Y.Z908^XAFF5,(]XF346<^)3;&H=[;P\nM_RVY\\MKUU59%6\\^/T\"1UHRQ2E>Z3\\2@AB,T3S%>.'&&#CT@QR1,:M8X%+Q8?\nM\\[5<E*L/]LZHFZ]P0R.GT5#'-][H:FL-5OS:WS`F;)JPAXN_>*\\FK.F%(GB\\\nM71K;5A-']]2F^\\<+<9VT\\S8R`US:;ENW\"Y8S>?\\MO?[D7;CI]2?OQ4UOE[(C\nM5YNR(Y<N;&Q?K@)R*E/VY2J**%7E.$RR`B4,E!]:AWN;YW6RSCZL8Y$D\\<*\\\nM#9_Q4O/#Q=E&5`'X]`K_\\4/@Y!X)WQZA5XV>Z^*^,#^.P2).B>8N900+$\"([\nM61'Y28HXNO6ZC$=GZT>;ZXQYKY[U!P5.I5B\\8S^565&UZ81Z4&TZCQY4_PKZ\nM1#)26%.=S!I9$<L)SN#72L@8W6CU3'M0SPK\\$PT(^3PC0]XS<?Z&4,3)8=Q[\nMEFU;!F$7H@VT;*JH92HJG$[6PM.]-Y.I!H*6)^I$!0TNL5Y]#?]WAOTW_-`+\nM#YIT$$=\\6!`&PP&9HE9\"6F#8N$.']SO`1ZB?>M*E2&)54<)=IJ[%V&O\"D\"`H\nMU\"%96'L;UWC\"U@._A;?:!/<#,YLM%`JTEI$ETIO7-<C%:,-L3(BJ4ETADR60\nM-Z^KK$>\\U-)M2_?K`G[&&QQP:SHKQ%>RD\"I0-U>5,FBMA8[^!EHKSZQU&O:U\nMJB;6:LG;&R;9`#DF/>,$#08VGG@'[BQFU&KEA!EI0K@EKT@#:^%)==`SO8A!\nM84TW+'*'05QZJ_ODTKHQ'3S\\M+IL4LP*G[!B*ZRU#-U^R9(Z=K^*[email protected]\nMN=%M8.Y([2Q+D&L2`*,D#OZ?'9CJ3P=,&H,48)`2,VCJ.`BM?'?>;!V>-#8.\nMF[,@I:12*\"%C(E`\\RMJNH=L<*'ZP'Y9-@HFU;+%:$V!B=1A0(_49@Q1@D!(S\nMZ)L`DT:A%PU,*H.4BJBHI2?$H,:_SUIG34R`9H))45-@2LAX#\"8^&8_`Q,L>\nM@8G5FATF5I_S1ZD!+/(3(M`385%+/QLLJ>Q155'1'L2?Q[#::FZW3G=.3\\XN\nM9L)*2PM!\"1G38S5+JD8C-2^;&J=9G=FB-*O+J:.51%4NC8<>R/VF8G*T^1\"+\nMY10LTJ(.M*48+$_\"`&_N2=H/WQ.V+S^PO=].L7OYH=W]=I86XY*>YSV\\!DUZ\nMEL/-ML]L'?69K:,^LW749[Z.^CRRCEJ,=G[9#RRJ8J&K;.^`TXMFO4$VFPUO\nM\\:5WZT65(Q\"3/ZM4Q-)B*.C3`XB)Z7FNMX*<)O]VAY#(8_8WQ%M,B>X0W;L<\nMTALE^\"U)#VZ?8JO(IR\\@J;KR*C'OK(!(J^0+4Y$I;/JZ@;]'T.\"F(4ULWZ3?\nM1G$);Z2+H8\"_0$CZ^WD'J2JEQ?D7.4C3`IE:4D15*C]A&L29_J)Q.!LVTH14\nM`04\\DFG\"RFTTR82\"Z?FE8<^86AHVBUAJ61;5LOJ$R>XIQI=K+];X5\"94)%&M\nM1`MB5/1#7>#ZS@C-:?-L-FPJDW)(*N&1(6-Z8^C0DNG#!6K,.%B@)F='K2:J\nM2I0*?5M`E+2TZ`4!DLH86&NHM8@QYAT-6Z'2,R\"T<7[>/+L@C=UFXV#O>&<F\nMH%)7'^.\"'DLB8<KVS3#LAEED6/A(&LFKS9Y'\\@:<5\"J02HM(]7TPT]+(]7-@\nMELH[%<*V&CT)./>*-.\\\"/\">B>4G7\\J#O6]?K8#:2O32,K$A\\E^!MS)#&M$VB\nMX^W2EY[>)X[>YX\\Z0'9#;Y''0\\W.L-^_)]!P%?<945)=D\"?Y!>].%GBUF9RA\nM)J>%D=;<`P`?@D&R%,3?LW]2$?/\"_/RG>11]TT+U6XW&/,5S_L_?.:39)*8)\nM71<,'3KL+(YH1ECNA)S4(`?0I/@^_Y\\*3VTDQ_CA>*;Q%50$?)7XAL:I`'\\M\nMOM\\*WNI+@I?2M5P3M4H40G\\N.\"NEEP1G*ENKFJC-N+4:3A*(>R-\\U-,C\"T*C\nM083&]N'&SCD1#K?HAT7TR;4_$THC>Z]?TTE\\4`$=F7=!W6!3&ZO-G^JS'+QJ\nM#`;`\"D/'XZ+\\I4CR)R(Q`Z/`GS8*\"'C#ZA\"LQA:Z0!1*0[4B:MJ#[8[OBI.J\nMOFR<4OF%=P$I6L@OPW-]O\\6T!2OJ0H+D!GS$RTQ#?/#N#;VG!8=&06`:D[6U\nMYLDV@/.*/O&H*>4$0GA4^(H]!&FRXH[9]0L]+._KEK.P^,DS@Z'G+)065[]`\nM(1,%8_`36:##D$*>0!S/QWZG^PS0W?7OV47T3765UP[+%XE,2[^0UZ]9+I3W\nM(\\T3XW345NJ2^KWI\\ZBVUR6W)D:M^8`^G:F3P+-N+.B'!S`1KR/\\^/@7Y$WW\nM9(@/`$)%\"EO$@0(5\"%8M%(JQ%KC?@HH6.^9-T1G:-E>,DENKB%*I%)TH/[>7\nM-/5;>@DTE__KW90ZMK0R<UN4:DS>YK,</]!M]D@4)O5A#5;`'RA<2482,!''\nM.7_VV+PSC6&`V;I?>+B%QP/B,T1#\"$(<J'Q8=:;8J\"E?'1O3NN21<F2?,6T4\nM1!/L[$-F*DB<.HPQ.\\?O2&,6Z]D(F$78T](.2.*?*XV32I(*9)6U.)'[GR4K\nMP)#FKO\\=MLII@_6EL#4MTL(`4X\"]4N(Q-9@15G$F,YV.U1V=I=`!WGU]7OC4\nM:.0-XPO)-^,ITIA?G3R1X>J6360@(&T>@^)H&L.9[#/!FS4'J`V)YPRL(:7/\nM:``2G\\RHL2E7'9</V+(,)L?[Y=_(9-PW?P$FIWM=1@AD-?8Z-/3U&[/%AF4]\nMR\\=GEE[DA0_8]H#J$4JZ@6]Y\\\"'#GG'@U%(&3IJTIXT<&/*7SQ?H\\7[64K4<\nM\\^8'@U:57C1HZ<RK(?-JZ@PG=3WVLAC,1=D,,H#1X;F&Z?ONC--1K39R&^`T\nM>=&^\\8E#SH>.+])W6.#KGGR8K$Y/Z2X+3E$=RS,-?$E(<H\\X#XICK6R<@G=X\nM20(Y^%KG7%*12UIYAC.[9X!!DUX>#.GL4!`654ON,9XP/8_-#Q<B,0P,P*\"[\nMS]_*!4NMH.>YP\\M>\\C5=WKQ/!KKGFY[(I6`215^,%=:&!3Y;J,[email protected]!F7\nM>E'Y&MO\\+_3>0.GYO1/H=Z2)>2\"?84@XN;\"49C!XJ,3(-!!==8=!%C2;./.H\nMRA-F'KYGXGGU/^GLD[\\A\\W^0I>5YDNSOSZ23/[)<#-HDO;K\"F`VQ7)0DJ9+<\nMUWP9+H1P\\+.X4)+D'^/\"]%&)NRB(7C0JO3[)>]U(_%(TK+/)/\"D?>'K'8N\\J\nMR0,XV:>,MYFW5GZ\\LTJ5'SG>)`V<4Y;B\\?;MG2/)L^Y._GCGR-H+&DE2&4=2\nM64L\\)\"55T7^UZF@FP.[OVMWXK=GX\\*$^CQ]:C8$]]/%?O:M#?C7/:NX=GU]L\nM'!Z>;ESLUK/%H>\\5AT9[1<#OZ<?.$)LQ0E>B%&SJN1/:AEGAUYT^A:V_Q1D4\nMF/`,1]!ZT&MQ\")]MH2#55`!8B:;`GQ9@17YQ`*<R6I9JHB3+\\DR,[MKFW=>!\nMC2V_\"=\"R_$S'J8?-#\\_&8AG#A%Q19V+Q2P3U.<+#,X*:SMP*1'^Y6HU3JZ5%\nM0H]UVG4;[QVF/>#-PQ.W$=\"R^_M;3Q_@\"6?>9J=\"[=DPJI;',$J3Q.%BWUHW\nM.LR.D290,M^:YZT^$WQ%\\/Q]KE!<SN=:K4$K-__G3$\"C8\"'NX!_AS=BK5`%8\nM-9$6?5]@%>V_#]AT!E=AR2XGWW,6[HK1)[NS`O[\"5)!]C8PG)+HR6](IUV9=\nM@1>7R,D-)&06?<<XO65+CM_O!F,Z</$Q4-P%T6]<JX./!=`#F@*PI,@$O*?O\nMNZ9O;HAN%P`1I&]=]N!_^DXD]E)M/-&ECY5B(-1I9YR\":ADIJ#W<+WQN9%3E\nMIT$FG4.8,,E*=-[+6+^`+WQ9#>646'K^Z)FYK#SKF7GJ8HJ-N=0!AAORV=%Y\nM%%?/Y5)TA\\T_,DY37XQQZ9[\\#WE?WM\\VCJ3]M_,I..[TQHYEAP?`(YGTMN-X\nMNKV3:VUG.CN[LQI9HFV-)5&O2,5QI_/=WRJ`ETC0!BE*I++I7R>V1(\"HZT$5\nM4\"@X6./2D3N/'BU9CR8X/NET(>B?\"A:^T[V4G.#?=<]J<YL,#?#0T(U2\"55E\nM.6#HHJ7_YCD@U`E#`T?22*X\\D>+(XZ]O#_]ZO#_NW;C?^/T(_\\,_^B;'($V\\\nMH538:<BO5&25>I;YI_A3.#]O/]:WEYB3&<NQN^[CKV''W[KPYN6]'L,`(#4(\nM*:5\\R[*:B/;\\OV=6B[6<:,AZ2RH&G?4FZ/]5\"IAXVY6$3`:I*4OZ]/`=N##U\nM8:J)F&H94J%H6WEKU93@7Q]OQ7ILJ<AK4VXM93@95`S[H>6*^$SJ6;3\"$=:G\nMP8[9T8BJRRVFM(^K1*UAA:[email protected]`JA\\%!;U:-R=!P13RN`7V1Q\\&2\nM<7Y:<8D.T$L,.>AM'TN-&D\"W1I:*M1;ST8@AMQ_C3T?#H!J36=,5L;FFG1@V\nMQ/ITEX(W3$RY?9AV,M:L:0>F-L:*-=@TD-&.E`:[0>_*K\\9HUG1%C+;J830;\nM8GT:;#O`6(=*:7`[&>MD=QB:9JQ8@W&1CJJJE`9_00968S1KNB)&V_4PF@VQ\nM-@VFNHI584TI#6XE8ZE>4T!1&V.%&DQU\"HPV-\"D-'D[[%?F,+5?$YNR.644V\nMXPCK4U\\*$05-BAAO'%=I33%%75P5ZRZ%Z)B:<OD;,)+9N#*;9^/5\\-FL*3YF\nM0ZQ/?6T(+:@ME\\/13L;646>D5L:*-1@=-9I44MS4E\"]:E[MV?EJ;$IL:S&RF\nMKFUZ.IVIU33#U<-;H1Z;!-PUD\\BM_M[UJE8EP9:KX3&IR5G#$=:GP2;,;Z8E\nMM_K;1JY:-<UN=7%5K+N6AER66_V]&/K>I!J;6=,5\\;DF;XT-L3[U=:R.9JER\nM:\\\"M9*REUK0.7!MCA1ILX4QGZ9I4I@12@65%>$JA?SF4JIP/[\\C.0IE.BI(2\nM\\<LGW2?LZ7;D(_*;]G#?W2)RI>*K\\BRW+[ZQ/!/KG6X`#PU3.H<3Z2J=IFCI\nMLFF*R!7,*&1%<Y2OCZ+Z.8\\JI,Q9!FE-RERHL!`J6*8NG19:C=E4]HA6K<S&\nM,+@ES!9K.FX)62:I@+#]^<QWY=+SK-RV3:Z;(L3@7[<+,[C:V@0X)UDHKCKG\nM[.PJ]D9S3JR#&`U8EB.-MIRR\\A!@J4U`0+VEY6K`6T?O:+9*I?&V*KL=V>/E\nM=;+;5HW6L%NH[3;FY-BJ7$;XHL$'[FS<[TVE@,/.)<_D^RE\"CO#[=D$'TUT;\nMLV]L0RZ9?`GFY=)D-IQY!9H(CI>=G!)]$`A\"TDHC`5Y$T0`2Z.WQO;CR&A#-\nMVL26!M[*_#:<)OA-S-;P6ZSO!#PVFZH5D/=ZZ&-I'SGP(,4N6]1/$7B$W[<+\nM/+CR8AJT;9D5D+<4\\W+9RAO./+$F4G3!3\"J-O\"%IY9&`-N*#F>WQP4+EA2C7\nM=C1IY*W,;UMO@M^VTQI^B_4=J[O9CE$!>6=N;X\"\\E4,/I]AOBSLJ@H_H@7;A\nM!]-?1S,Z\\)?<_<!+\\,_1LEF?&\\\\_L3XZX(DYJJ#<9P$>1+25!P2G\"5?,4=OC\nMBG$%UB'H=0PB#<\"5&>[HM`F&&^TY9\"_4>,=040\"F-`+?SH:!VXWFP0I08F0=\nMN?MZE,\"4Q>;M`!>NVYBLYR3[%ZMG;2ZW[CMD;8$6@Q_G$+)ZW,:+B#:EA`C7\nM0A.\"6\\=2UX\"PIFPERN99(]8BW'!TDAI3\"^I?$?]IRR(>!\\L].4FYISIHM%H>\nM9<0TAW)-[BO$\"WO?';X]WE;^]!*Z'W[95O9[^<]'T^L>OB(<XY:$>WWARR7<\nM.+GZ5XM]%*$T?-<N?QJU2U=5K0-_Q3DF+>(T#*MXXV.C.\"W6<:RAY3CRF2E`\nM5GF$EZZ?5:L/[;0K,04TR0$MU^434RKP6E>;V!L!JMJS-R+2<V`+XWT\\1S^Z\nM[RZ?12,?R\\&$GDWL3_=0!!+C=D$$5U-\"@54TGNKK9Q4IGKLVAE5B+=,M8!U1\nMI=%T7-Z^#=DJC;7:M]&>NGA<2:D!G#9-:2RMP&G:P*P%-+5GUA+KN*DCY^6J\nMS2T:]V`DAP]F]DC:0A=%`#$8M0LAN)[:-G#+D2L85XU;3G$NY`9Q2ZQKE@K<\nML^5KWPY&Y<W<(DV8N:VUQLRYHCIF1]=4^6*Z%5BMJ0TD3,);VY,P*=1R306'\nM2],JG4KQ^C=@:#(X`;(M/F/!NRD\\9<&^;A=>,*75#`TX9U0ZFU*\"<\\8]IU,V\nMCW-B'=3`G]+T$B=4&&7E(4!KPJG2]/8X5:'B0CBJT1)G5\"JRFS2Q&J#1EJ\\&\nM:,1&]E<YI3+QY=PR$&XA:&`?18@!W[4++KB^6@089E<YG\"+-,*LXS7&C&\";6\nM.(KNE2E_)@7(*F_M9B/^E=D>_XHKJPT1JN;('TBIQ&N[@4Q(H*H]^X)B/7?0\nM(Y.\\GV#1R(?3OAQ0.,7N&/91!!3P7;N`@BDKX%Y'UR6O,ZC&,%TOSF+<*(85\nM:!RX4N#A22,KD%7:VG6U\"3\\*8K;66#M75ATB5=V0/W%2B==Z`RF.0%5[4AR%\nM>JX;X(+II,IIDY,/1W)`812[8-A'$5#`=^T\"\"JZLI@H,,ZN<,)%FF%E\\NF2C\nM&\";6.`)^U,+=>0]8.Y!5WMI)$WZ4WK+\\*AUOZ-(7KM-;!:^M!DZ3Z'K+\\[Q@\[email protected]\"V55.DXPD/3\"[V`,;%3M@HY;Y7TQ3#=7HP%]5SHY(<LM0BT^-;!\"WQ+IF\nMHP?ER)\\4&55PH.Q&'\"BG/0X45U0-0E%#ES\\C4H'5AM;`Z1`@JMVG0V\"`*K)>\nM_G1(>LM:+BW3R%4-7^RC>-^[96F97%>)\"0RC\\F<^*C\",9D]Y;\"C#\"C0._\">#\nM:\"4R!<KG5AI&$_Z38;3'?^+*2B$2-4RK1*I`!5Y3V5,AM?+:I*WAM5C/3?2^\nMDGV$,M[71`XG<C>U+'11Z'U-VH427%-M![B5[`2L@%NY>MX;R2VQKEGH0-GR\nMI^-&D_)F;C7B0-GM<:\"XHCH0@A)-_K1=!583M8G,8:*V)W-8J.5$-9'U>A5$\nME?.\\B%;L>=WC>+7-[V**BG?IZ239`5@!MW)WW6TDM\\2ZIH'G1'1+'E'+.TY$\nM:\\)Q(GI['\">NJ`1\"4$(->42MP&K21.HPH>U)'19K.7&0];0\"HO:]\\;0GEY5)\nM:+'G%7936,F6?=TNO.!*:U'@7+(+L\"+.W7-V>`,Y)]9!B@Z5)7\\2BU-6'@+,\nM1IPJLSU.%5=<&\\)2XL@?QZK*[B9.$@-A[4D?%FN[@UZ94^5,UJNSUW*@X12[\nM9=A'$6+`=^V\"\"Z:O5+,[.M6K',N291C5BQ.`-XIA0HVC*KA75),_F05DE;9V\nMJC;A7U&M/?X55U8=(E1JR!_-JL1KHXG<86JT)W=8K.<&>&245#F;U9_=3>4<\nM,FK<XY\"Q7@K],?RV77#!5=;4@&UFE8-9)=AF%B<\";QS;Q-I'P*VB5/Y4%B.L\nMO.V3)KPJ2MOC585*\"W$KM>4/957DMM5$+C&UVY-++-9U\"UTRN\\J9K\"M7;K.&\nMVL4N&?91!!?P7;O`@JFKJ9*.;FI5SF3),LQ4BQ.\"-XIA8HVST:]RY,]D`5GE\nMK=UIQ*]RVN-7<675($(U=?DS655X;6I-Y!*;>GMRB85Z;NH:\\K[*F:P/O\\GA\nMA%[LB4$713#QX;=VH0375&(!MVB5`UFRW*+%V<`;Q*T\"70,?RB3RI[$^_%;>\nMS(TF'\"B3M,>!XHI*(30U3?G#6%5839M()3;-]J02B[7<1-9;9JJ:X'T(<?CN\nM[$0Y4J[=WL\"=`:\",Y\"YKA-=DUP5$/4E?L,Q;=?U@4,L=X?BLI!9)EUB*/_\\S\nMC!+$<7#]4_;3WNP*/^4F@*>-+55/E2I<B1QR)XXW4PZ6=/TE>3F([0-C.3.)\nMY9*&ER.O%_#><&@,LX+9W4MF[/WI-$_#3\\\\&[N=GD_EH!.8>?^O-@VT@K!!!\nMD@0OZ%T$(/!QC!]\\&.X,9MJKF3M5]C\\K3_Y7>;KW1$F_+YQA.=#\\KK#.H<WV\nM?4C#M-3\"0,U*`K6U<\\-*<K`:Y8985QR*-IRKLOTX9RQ@(H#,J1?\\H)S-)^_/\nM%'+P1?&#&1CGP;4R\\%Q?F7B!,G#[H][,5<;N^&D'1Q',>K,[+`Z/UGL@/P=*\nM7[F2M@\\^FDBR7`_0#\\=+NC-5KEM#J25]UTDQI4(96UA=RM*<2C(^.3M2]`/U\nM0%<B+,I3?CESW:4HERY+)4;&1,98V,DBM)*,UT*I=$6H8DK%,L;SZ5:Z]NWS\nM\"C.4(9N\\EHRNCW<TA!,4?(_]G)R]>?_;\\>E.?U?9>=)[HOSYI8(_@YN(_\\!O\nM3WY_LAL*#$^)6^E\"LE6&;99W<\"H/6\\Q]+-ILT1A%W2_#0-E1=SF>RSC=%EG^\nMF`D;ZL&SQ`?'86\"+>-9ZI\"0*'S.:\"P*SS2TK1L<*%%C+YR!6I$`L$]Q/L&SY\nMB@W^G?_LMC<$CP!>`8HW])4/[\\]./AUH8>+5\\&(D=:4/L\"*;'\"?7=VD?]L[O\nM8J_=ZS5[LE9YG$`.H,WY.6<V9DT85ECHDMBJ?.6(&@7G9$.^[TMPMG1*1TG!\nMB2W007!/;K?P@<3?3OYR_.GD_/BULN/O*O^N_(:_G9T?GG\\\\8Y\\\\5[07518M\nMK\"1)#.4\"(:$(@,*O8@QZ:)FB0&BX1+&`GK9F@LHF=TNLDM+41=CKI50H8QMS\nMZFTC1MF!]RA<\";MTPY53-M3KP6R[<.6TT+*CEC*V:VO9W;+XO26M\\W$X_C7;\nMIE[>-0PI7'4$:^OMB&\"YK>'1`KP).;2U!C6.9`_!;YC&D?*A];HT+LE.:=^:\nMB8TGL&VK#9A'-QSSS/9BGMDFS+,1\\YPV8)Z]X9@G79UG_1KGM!GS',`\\1VL#\nMYCF;C7F.=#'3M6N@TY*=\"H9Y#E8QPCNN&\\<\\)U?L:,,T3KJDZ_HUSF@QYCFX\nMIN[0%F\">DRL-NV$:6&';8UT:2-J$>;@=XEAMP+Q<;=T-TSBSO9AGM1GS\\*H1\nMQVD#YN5N)-DP#;3;BWEV>S#/4%6U8ZA:\\Y@'(]EHS(/QMQ7S0+[MQ3P8G`X:\nMJ!N2^:2WWHS]#+WZ4D<E4;LSBK78A[1ZX;X.-%BS6NE5BVQP^S9LX\"YQ)+-$\nM*W\"79,^5;11W*Y_^%.NR:0*W+3VERV5W&*'YZG880QYG=U`-U4$;=.R4EI0?\nM=U(#>QWC%O/?MC@=(?\\?3%Z8#P?=`$^?I/;9Y=3>R9YC$78EK?W8H,NZ6+,!\nM.++'Q@IS$J+\\/P/OT32T$H5%*G-?T[)U6C>3^YHN6U3Y7NX+;4'3-9!&B>H-\nMPPD.2H[_N3-O8>,2:,];U,7OQ5Y?3M\"-Q7'!YS>W\"CMAQEC>#9_H=N,?7W\"G\nM5UIF2Q3\"_L:&#P.Z]+SPH](XJR4EL5>`LR'_^#!?*!<SMW>S,%5H!\"!6*U'=\nMHHQ:Y0X'_A]2JR5J?M>B5DGU[S6KE1B]3$2OQ)-R_5[_`44;>P.W*^>T:OGJ\nM*;QQN1F#-UKSE&%5.,>432*\\5,[.7Q]U?ST^?'U\\>A8:-OI.6N(#ULSQG+^T\nM.1Q?WD7*<5RL\\P[I&+HJ7SO4N[R49G]V.8^W+<=]UF:]S)>_?+*DNNNZ\"LPV\nMY,N-RC-;S^U8;@JS*VQ55M)T7;>0^?)QFC_\\719I0*;9+'+>N!S[>:,U\\]^H\nM'VFXLE.(S'13/C(KPV^:C<4VA]_FTN&7I+Z;\"#:67\"QV>^T\"S3,E&([=@VNE\nM-QGPH);_.N[=*1=><*U<N$HX**G:\\L;B?9T57E)V&1[[6K<\\:\\\"OA2\\8.WX*\nMC<@V08B.7.2S*B$ZZG<O1*?\"05P9(8HMTT*?*SFGY@>S>3]0@K'R-)A6.2$\"\nM/=JK/B&\"%.76<`W<R3.2@UO+4V*H*S_5DZ5$+\".'<LK*H&=\"_M\"/UP6Y87B1\nMQ4B:G\",P.:GNI8V-]]8-QNLU-:/.`VD+6&G@XB->&5<&*VL3F;&P//E]B4RZ\nM5'$)D0FMSE!M$*%&Q,BH!-/]GX)QUW?[U;!%TU>&+;%H7G(\"%V'2`,`WB\"J&\nMR:7),E8'_D5DB:5GX&Q`J.2N.BXH8J^_>Y)+M8:1C7$7N@C-2%:GRX<]D>KF\nMUD+QUYF+1T\"CP2C/%!/D5TF<9/7E#1F/.),XH<\">_^'9(%%MA%\\/_W;</7EW\nMWCT_>7O\\]_?OCA4M5&8*+JEA:9*[^^6E;&9O[UI\"RA6\"K35)V5S]):,5I2RV\nM;8H0EMR#B2]<?-]]2G#E!L@S[W(`\\4%O=B55!@S>F5U`S/=34ATJ3F1%*A$!\nM.7R/4F-HGOJ,Z<C3X'?X-#WRG6#:@0]WJVK.ZDO-RFK.+\\?GJ#CO__+Z\\+^Z\nM>O?P])>S&\"9LG`R2JT#7HC\"YHVK+*HQ=<7&L50KCM&=\"$2N,&'$L]`5MLQKB\nMS\"=##]QM=SR7JKQM&+E2Y>D>2BI-S24WAM-^'.00%7`8;\\^L9%9EN4)R]<@K\nM<P4O6UT%5\\2Z8SL<?$+=28V:_UW-T6Z1(7U\\=_+^7??L^.W'=Q'B$HV\":NBQ\nM8U8/T41KCZ.R2+10\\D0UT#X6:V\\+C.1/<2+QE\\<\\U)G.O*ON51_KQ'U9+!3W\nMP!+\"XZ]'1_O]_C=EXKH#7]D/9KW!,`#>]T:21B9:[+FO5^D%@XBF;JIU76L'\nM+$2<]H+`G4U>;A_.`P]I.GCZY,N3[1*+\"Z\"UY7'A*@CN.\"9$[U7.3]X?P>SR\nM(;0%'0\"`D,6BXO>J05X+UJL$AFCYZ/^,$A#9A'-))1!C`U,*([T^D2L8*6(6\nML&CB+:B\")%72==-3T:<[&P^]+%E(U&&HV111WDR'WZLE0KHBN1P18LE0#8BB\nM\\C<FV/L7PT#IC]S>!*NP]L=3.2NCV1PL04?2AG4YG_2[O%&7]5\"'57$A6Q;P\nMPY:_$Z$J/^QLWD+K^%&@+VC))DUJGQ:,@&=/EESO)M(U:A[%5?>YV&RT34=;\nMN\"&@OF%)%S))AB7FGHG<LV+N2=;[A\";+^[_EZGT6L2^[#TEPMXZJ,>.E*7+6\nM78-5@B*QS#!3,G7C;!KN<[TQK,=QO3EY]?[5?X1W4^!/H4T?>-OW`$H8(6#D\nMI7B7BC^\\POGCNC<9C-R9W$H,64B]?*##LDD[5S4Z,)*F5T<Z9OP%(R\".Z\"F>\nM8$E=C=MRR=*%8S`;+UFZ_&D8D62%-DQ5P%VJQ;B+:X/#2D$YZ,S*MOQ2HGCY\nMV1L.%H`6KV<U\\'K6J-QZ=1)6>*2DB`2Q5'2[@Z.1W+7\\/)T!T9=RMF)D%Q6C\nMUN5\\JK#5FBW#D#VF^NRI<G;G!^XXNI\\E\\)1PV:;;]8/Y!7@C_9GGLV2Q:V_J\nM7L*D>*=<NK<*1\"J!QVRI$T8EM]?#_C4,<<(&.!KV`^5V&,`GU[V8?3N[+Y0+\nM=^3='BC*TV>A9N(&%36IY+9D*3'F]J$V28S2SFS38A1;)^XV4W.IT\\W47.]-\nM8FEAYZX1,ZBM`T7V4N>>J;W>\"[L**1++#+?^J*.E9!8M4&P_SG>XK?QI<;&M\nMT&B[`V\\ZF\\BEB-/<7F#4NIS1AJW6;+32H=[ZC#9DA`A[316,U-2LE$JO7^\"F\nMELT;V2\"!F]*KX$T+7&CQ)N;MF4G>7A5,,Y/LO/5A6DAC'J5-3-DSDY2]2A0E\nMB7G-4B26&;KR)DDRG/D18^P$,U6#WT'IE<_PZ0V&EF/OLXOA'7SXK_%4N1F.\nM1IC,RE9V+Z?]4<_']0MV^KC0FA^'W<N9LY&]F\"-N7LZ>HV9K-F@BFP2P/H..\nM.\"&$<#S[;%I)\\G3+M\"%W>'JSM$'Z`'7CVB#&\"M,`[3\"=I?!]#7EK.32,B!0`\nM/*:(F`Y=\"N#7D!`B1Y)8:@[XX:F;'C,V?8EKBF\"I\\0_=@;(*TW5$-?8VQW2E\nM;Z-LW'297EL:^.*I2R^;$;JE9YWQC1*Z_,6<C0M=:/F69J,2+.6/6WH#_G@Q\nM7EL$''*++N606Z0!AUP>K_%.22\"QR\",'?PM4IC_N^3<*WM4U'0[0$GSXY[+_\nM.:C9@,E&N]\\6W2SW&R_C-\"R[R/U>K^BMC?:U\\0+0#1&]&`4L\\+4M:RE?V[):\nMY6M;#OC:MKJ4KVVKK?:U+<=&$F.I#;R)^^B!DW'^<\"*W`66KV7U$WK:<.;(V\nMZ[5%6VW?'B)C@PB#;1T\\#-N(M;1>$>K9/<1-$:%T$FNS(A1:I:V;*-*E=@]M\nMHX'=0T9A'DAMJ@,]=*F]0YLVL'>8IT<L+\\P#MA=V>^\\SORLWF,W]WI7<P6([\nME_R;M\"]GAG&[-9NB]/'5]9EBS`HAHF(:L[VPU5VC.'.YRYLF3NG#I<V+4VRM\nM%KIU-ET*7>T&W+J8RCS\".BK$X(ZF+8.PCKKZ(W.2-`GEYN`94D>3K=/AST\"M\nMO+&443JY`Z-1ZW(F&;9:KT$Z6OM<G9`1(G1UL-R*0V0+<9028Z[2RB:)T6C?\nMBKU8C&+K-,#'<Y(ZI940R&C`QPMI%&`JA=C822J!5J+(;&`U6D216&;40@IM\nM>;_UVO.#26\\LY^HXIL#5B7LH[>Q$+==LEF8KW9V(&4*$M0T0JT/D_==R8K6S\nM56PW4:R.[`&*-HA5;+UX;,]9R%<MCT]V,SY?1&<.=8FJ:1WXRUH\"=:$Y;0]5\nM(MG!\"%6@4M<DD5?>!R+L`CM1XW*&V8`'!$-OGP=4Z,<2U3!!A,221-DR(B39\nM+(3-$2%I7P:\"M`]+\\`9`(&&9_`/\"+OE;-_84N+!$-0D09\"V3?0!]-)!](.G!\nMPN`,))!(U^YT@VXO\\*3*2$'G64<G:5]R`R1JMV9KM-KGY,2L$&*J[72(IJK2\nM13I+B1/3'C=:G([LO5?-BU-LK0[Z=HZU%+XZ#?AV,95YB-4T\"U16-Y:!6$W7\nMVD*34&Z:9B*-ENS*:S!S9S-/ZBYKZ#;K[,3-2R[:A<W6:Y-:\"U,N(TZ($%8C\nM.HB2&K*KKZ5$2;*W1V^6**EL-:_&12FV4D)0M.HRZ*HUD6H:$2D`5PO\"9,TR\nMEP)7B[2$)+'4&(FV[$H`=#R82U7)XH2+&I<V1FBT9E.TVK<2P/D@Q%0'ID==\nME5T)*\"%\"715,CILA0KR<<2-$*+9*QT:1+K42H*M-[/TP$O-(JNLP.>C&4BL!\nMNM[,Y)`E2\"@Q73>00-F5`.@V\\*1JG$+/N5O[>./21@B-UFR$TA=PK]4(@0\\B\nM'-4)1/^Z*1O]EQ$AS8;^FR-\"Z<NN&Q:AV\"HI1/PZ72KBUVD#$3\\G48\"C%H3[\nMNKU4N*_;#83[`H+$$K/0F;%+Q/J!-Y>T0EO@S;#6%<QPOFX[M%OISR`C1%AJ\nMJ!#G&UJ).%]:C(8J\"/,W18R&ULH@7R!&H74:*D&Q+A7C&VHS;AS2F`=4O'V-\nMX+5F2P`JNV&M#12)9::#6X/7Z$DC:K_GNY*UL(%V@6L3=5#:(,.&:[9)HY4N\nM3L@+(;I2\"B(U37ET+252FCW$O'DB-=MWD+E0I&*KI1:*V%@*:<UF'+V03`'8\nMVA`D&[:S%-C:#1S2+2!*+#D;71_'D,3;_L5,JE(?])OU>UC3<@:)3=9LBD[[\nM/![D@@A7B6IW\"-$<25R5%AW1LB>1-T)T>\"'4!HA.:(4$,QR)9BZ#GT1KP*]#\nM`O/(23#;CQ!]&>0DI($CN3ERQ-(R-\"#/B*/^6%&Y*>'0P-(F7J+)]]DDUO[8\nM'\\DE.R[>()7M(;10QIF+[N?>[\"4WSO&3[A/VW!]@A0/ER=V/!\\_V]G_L=J?=\nM'Y_\\4\\JFL</'2<=+6397$?08\\!*64$76R,/<08L-Y&&!7CIH=HLWYOB]SV[W\nMS<FK\\!J),U1M_BN0JRCQ9Y(@2TH<GWO_V9W-AO!5;W*G7/7[.EXEX,[P]H@8\nM1Q%[>UB_'QY2V\"XPATG6P6^N,O==#J47;K^'O[#+5H=7UWCE:@#(\"SS+WE#1\nM8R\\+]<R$X(-8B]?UU,P3L\\3^:\\,\\$>L-Q=DGJ4G%9ZU*H$W7.P=E3$HP$^$\"\nM+4EJ+BU!FKW>^>@^TL12Y*3*WZ^&=;KE$#.W%,V:EDRP'ZZYWC8,NWV+T,@%\nMD2]/<06::O*7P4D+C^86H#=\">+2%2\\\\\"X0DMD>*Z,UUNW9DVL>Z,!.8QE.*B\nM,UUNT9DVL>B<(T<L+9TB>=KRWOR,W90D[XE2/;MTF>[A/D^4/=<.3Y2K\".Y<\nM4VHM[\\V7YF&NX,\\&\\E\"LEX:!>NFLT)LO=?U3&[QY2B'\"25T`M0J>E+E+J8W>\nM/,7,9IID-O-9JQ)HKSF_^4%OGF+Z+TTRG)<@;<UYSJ6]>8J;#WA%4\"C%'Y2/\nMONNCNBO0'G[P83SCJ<^T:P3^\"+Q\\/$87AZDQ?.I%NLD0AWD_#\"?]`];=_CXZ\nM-MX\\V/\\2/^K#F-W977\"-6(ICZR@7\\T`9>/#-Q`N@UWFHCZS/@_N`_),<?.?*\nM(WR*0)N-\\@B_BJ]*YV/&`<\\6QG\\+W+@:?G8G!X_2MQ)]Q2>Z7Q@RA[`<RHQI\nMDZD\"F.\"E/*$V?:<L-K7LGO=:6\"S4:I.`KV72]#HW#N$OPYD?=)1@=@=HB*,,\nMK@&K+X<CEWO?$T_QIVY_\"/8T&,[<?N#!D^R3RZ$[.)#&=I.4O^?\\\\9<N?V<7\nM>^Z&G_\\4@@R;@&#<+_D=6--I?@S)'<$`)O&WP-5M&%@1/IG)0@KT+H(G^#A&\nM)SX,4+\"7_[R:N5-E_[/RY'^5IWM/E/3[_IF2W/[OW\"N#-MOWX1@W%`N\\.]-.\nM+^>O56J6;!S?N-22-:)&I2:T/4L#N+.2.JF%OE/B/66X\")/7K`<R+.E167J)\nMJ]:9\\\\/N-5>^PJ^/ORX.`>.Z8.A-OE6;]JTUU'D-N9EF+J,9?3U<6,`7*)_0\nM/[R:]<9^SE8B+D][P?4!-T\"\\-(!824G8]8E.&C!7+[HUE+2M5W1\"*[0-O4-L\nM8B=6B+(`KW_[GSMS+(<#XYTMW\"S_SVW@;/CD]ME\\\\OY,H=M/=\\-/\"MR#>#H_\nM5<9S8,*%\"S,_^C`0[5[<0?S@3WM]J0I6,.+T`J)DQZ&#P?]`_U]RVOI\"2?T&\nMW8'\"<O8-N6L<L49&36WI:H!Y-:VBBS99O2Y\"'Z?=B<?8B>'\"0B!D4QNTR\"()\nM(+1>BTR]]5IDROH9-6F1M?HC;T5:),8FKE5:&IOBM;RM!!P3,:``LRS?DN6V\nM[&FKFKAMKKYX&^-V@<7:!'CK6&F+71UO;=D<X9IXNX8R(V+>BO78L3K$48VT\nM'O^@Q(Q4;MV9\"U'W<!1N[KP^/IJX@>+/IU-O%APHRB$$\\C>]V>CGOG\\P'U\\<\nMN(.YXO?N_+@OC,\\/1]/KGC)QW8&O#+`\\\"-MKVL$?DW#>_3+T@]V#1_>`+$;D\nMK/TD\\'I\\X1I_E4)51\\WFLHAZ*EH%QR^?=)\\D3=JQ&L[LQ<$3JH[NI.WE.Y5A\nM[L3L=R!#H5UBW76\"==<?BD\"C((;)H6S,XJB;DX[#51W3\"1W#DM[4J,86HT0V\nM1!OW-1RLJ^3H\\?I%K/+5XDE';T\\^#U<#+#;DD#C(7Y:^-<0H2^UP.%@_TEFH\nM'XE_HK6NQTFWC!,)PKT$5$Z-K\"PH,XR70V:2380O[.X^>&9/M!\"CN=9AT4MG\nMH>AEVZ1@9J^;^*ZD(+8-@EA''UZK34\\)W'DI/2_0$A<.M6*ZM!`GK8<70VO@\nMC56BG'TKYTS,LG>2<WG+SBEK/J+W\\)R)N^1.<D1O6?K6?%JO_)QI(RXD-4%Y\nM)',.2C'Q4\\O/,_>S.PG\\9*'Z<N:-%6_J3O@VG:M\\4@9#?SKJW<6=>!/E9#;\\\nMHM`#O0,6U?=F`_8PH.>P?^/>_=P?]68W!\\#@!R.AL+C[Q9WL+0E`4C;#>+&/\nM2O<D\\+;K35MUG/;E'\"^P0Y!\\3%5-[U`UJ<KZG6@5T)5-?=Y,K:)X]\\$F:94(\nMNT`:!+5LF:QHZ*.!K.@%2G-S$%6)\"G219=*CH7D#Z='%=(GE9U\"D4[L_7LOW\nM7\"I26&S)W'O`(#EK-[+9U`6=%44)\\!W$!XN-VA$C<#TS->\"_:=T?J37(?U-P\nM]>KWP7^Q/1`#[4$^/QLGTY(1\",4K4S8E`@FUU`&N6/(9VI6X8FUVCC95*<Z%\nM9CP79B;32E,(;5>R-E5MG!?M>%ZL@T:[W5G;.$\"@V:&+\\=G^R/?`V0WB7`>6\nM2`L_`$BZ'!21Z@L/?*L>?(+[43Q!DK<'!>SY_AS@D&58#MF0)RZX8WSGBO70\nM47R/[UDM,/I!MSKL2@[B[>Q27-2Z9'$0WFK-3K33OIM?0T:(@C)-`_/1=&TQ\nM*/MN5$G+W;RV0:JDM?#N-;$J\"5%*PWA?TY:Y?1::-W#[;$AC?J[1#!LH(LO<\nM/@O-&ZA`+J)(+#,#O$TM.0UXC_<?6709OS]JPYQT#C%R9FQDDQ=$/15Y_/QK\nMK*P2-FJ'N\\]5BH(CJR5'#=O\"<#-[FN<[8+A8XXF&-FU)QU?AS/CX4_?XT_GI\nM(4^B*QE9:&6N9&Q#O*69%G#),J3CK5JX9)58CVQC_*51G`%I/`/&DV>EB6,-\nM)Q-*15Z:A;.A'<^&RU%GKW=:+!US:9:)U%J9F?$'Y<H;N[__/!X>S\"?#?7<V\nMZDVNW,D!*\"*F`0)63CU_^$49^N#Z]EW?QST.;Z(</OOXZ4%/=^:.O<]RFQ!:\nMOG@.;URR`@MKM&8WMXT%=!@?1`&3KH)-ZPM%W-ND!WJ^#L^FZ('>QEH\\(CT0\nMXH.N$M2+I?:=]$:J\\3`2\\_\"N&RH0M%0]'FC>1#V>/$%BB>D4\"939:>)=EO*\\\nMPR;,76:F+V>^^5H]N7Z*O&[V+5:<X2W:X7-S5:+@8^M49D]I?9S.5_39;$Z+\nM=1SC>;U$=1\\^2Y7UU/4-J_`#TH>P6R]1X:<B7S:\\R@_5\"<YJJ2H_X7Q8:3)H\nM69T?JELXPZ7J_\"Q#7,LK_8`JVDAL=F7O!^75V>L3_.O9^S-%/]#B%?KAM,_+\nMN+R?NI/7/.'J04_5OP8UD\\-?*YOUSMN6O$L#VZS93;7:5P\">L4$8K3A@OX::\nM75YL3NI.=FMO0Z1NJ.W;UA-)76S[CHE:H\"\\3FQAJ`X7R&85YX#9TB,$-W5XF\nM-#'T!G:6\\O0(Y65H%M)G2$0FK,=2[C)OP;Q<,'8IJS7T[/G?3\"=%CC)\\!VXR\nM?[@=7C+7'\\RI,H@C$8^LA[\\D=YOMIO)7K,\\8:1N&_(E<G(7*^MJ&OED'<JF!\nM4;%!Y0_D5N(*W9R=)K'N&#AWD7CN\"B>]2I\"_YKM>'@Q`#!/G,3.>QY:AS6S/\nM[I!8CIBQ:5BIG+7%.@L<_QCX,7>&90>A_O24P?#RTIVYH$?PIF'O8A3F\"/V@\nM`..8ALYZDY_GH_EL?G`6]\";0Q^#@^/7'CG)SD:K+<)!#L`7(/>F[1WP':^A-\nMWD$;<-T8`)\\<'<NAN)7=,[^WRR),A^\\`TT5-VX'P7',=0'2BIM+&OCMI.MF5\nMV.]2FF);M=#_M`WI^1JH+#\\SV9NSN\\]TGJ@&ZKPC/5]7X0I1-V<E5:P[M@VZ\nMX\\3U%@2J7W&&<]J3(<#U`4^:$3VN:5`;I63-Q\\]*S^5$UY#R.!=B\"___I&FA\nM(;`28H7(G%Y9XO@)+:4@F>C9+1MQ7T58#-\\!%B^T:0<(<W4B`+J$QHD%ZV(J\nMS<;VWPE3\"S07O%!BR-]=!^0IC[]^ZGXX/6:/?\"L-Z,;F;`%Q/:3HVIGR5]DM\nMSR*Z.9&[6*OP;\"XA<;WYM#54G`+6?%+WX<G.A#\"<6'%Q]AI(M-H3C8NE:D-T\nM3IPX>W/K_.A-]^3=T>N3TY<3#S]@AG'IADBX'?1'H*:%\\%<(WKR=%%CGKKL.\nMWRF[?<*W)!@?<.3KW4,ATK=>)Q7V&7TK+J@/`S-;45\"?61I5P8&F>IQ6VP*U\nMH[FKNC=)[:CTC=UK5CNJZZU0.R'ZX7UU-'5?W58H=L1/]`H?AS_[+Y0!T\\JM\nMA=>BTH(7D+QW:ZO8>\\6ZP-S3C'J54\\M<LIR@HR*_-7H`G%?6JAU.*T<`W!A(\nM77/71M;GLN<VGO5B*\\#\\.9IX=@\\ZPQ%MI1=^Y\"_)>Y0KILU_K.*$T;;YF7CU\nM'JB^+AU]5&>X=&Y>K0QON]=+L;:`J:J)U_O7^[V/FVK.QXVL[Y$K!L#?N!FN\nMAW0Q@)3K<;,&S\\.Q6^%Y,),WL02!J9N)P]NLQIFYF@$;I'&F=,V`]6J<J;?C\nMICDAYIF8#V0F-9:XPW63^%LW][E;?ZWF;=U(.UMF+CTHWT^QKW73/E>+FSTF\nM!)A)\":>V,3VWF[_A3!=K/H8:IDE*^+<WE;PM4SH/J4YORS1;MF=HXLZZZ:@E\nMW-N*_);>7:^5WW;+=RYM7*FW2;C_PT1BX\\JVG40<_H\\_OV8!],\\_/N8__'C%\nM/_YX=MR%X!H^#W]*?_'A^#3Z!G^$K_;PJ[?8S=NXBU\\/_W9\\].D3?!;^%'UQ\nM\\N[L_/`-]A#^E/[BS?L/YV?A5^QG^/)D,G\"_/,>C:`%,D?[_&^V;!\\:!_LR?\nM]9_%$CD83AYAOHVRWW_XR>?:`?M/_LD#\\T!C?)5NL?46SSQ/9XJN*IKZW#\">\nM$TO1',=FLGBPFZW?`$A9<QV;$^B!\\.8B:6NVV=$<+9$U?J\"G$SB4PZ/NV[-?\nMNJ?'9Q_?G.\\,AC[F(PU0]_%_$&'XP-&OQT=_/7GWRPZ@;H`3PMM7N_S;P]-?\nMNK^=G/^Z,[[HP\"?XY[^5Z'[C\\<7+/[N3OH?5B']2W`GVKHSGHV\"X?W$7N-$]\nM0LH_.DK4%O[=XG?L/<8N0-G9E7KPZ?''H^Y_?/CC^->C[M&[/_\"WOYZR?\\Y_\nM^^/CNY.C]Z^/_WC[\\<TQ.,_GQZ?O#M_LAGU&?]Z^>IET^@*[S+-@FP]RP(\\9\nMQ$]GNWK!FC_=S71R?'KZ_G3GOY'Y_^7->?4PI'%TASMQO=G5?(R)7X''=MAB\nM)BG>%',F%&_\"MMHXH1U.:(<3VN&$=D)\".PN$_B,<!Q^5Z_?Z^.\\\"N>'X_Q')\nM*$-U)'E&*/[_\"*N\\CZ+-0APM(&8/)*=`;W,V3$H,G8<'\\.WKX[]T/_SRX?WI\[email protected],NSCUF77<N^S7<MA>(`CI'UZA?)_R90`80\\?OKB;]GP??AWZRG3H\nM]M']+U+#Y+6[0A/0':VC.V9B`H;J=(SD,G+H]^SCJ[/S'8YTN^F/0I#+?8;P\nM%BH___`M:Q=2^W3)/V$WOWJC`3#Y$FLD('^/]@!.@SE\\,)\\$0\\94/!7#D4&!\nM8$\"YO;Y#[F)!N<F30+D7(\"]Z_1MW,G@V<Z_<+\\_FT*%_T+\\?*85-[H?,>YO<\nMCYWW-LV\"*'FNV0^`J+\"_K7-09-:/IFCF<\\UY3JUB-#5TTC'2-\\H^!3-^\":X!\nM_/*-RQ_WRL%,0%\"S>3\\`7^)/\\!CNPV_M/!U?Z+>XY[Z[L_M\"]'F77?$^<B?L\nM`>A2F5YU^3?!Q>@%?X$?]()A?^&;U\"]HP?_]#Q@3?_%7^,:=]_\\UY2_H*(N_\nMQR_\\UDD_WK]>?#SZO>#QF]GBX]'O1;U/,KU/BA^?!Y>IA^/?1(_\"E.*FGDU^\nM31[^%K*0I47@,R'?=V!^!92>3_SAU01PG^5%/,6:^YV8M2AJ-/&O3.X[3S,L\nM?_OJ'P>)>%E3>#ZE%]EWQJ.2?7F'N:BH''*C2\"E3.)P._L*'Q&`0E)FD+[;-\nM*O->I&OX8BXK&'[AD'T$Q*]L;@E'BE,1__T2QNHK+U\\J9V?Z;OB0@H_`\"_47\nM[-=OS&5=>-3(/FJ('OTW1?UBJ_?WNOB=%G['_N;N_0[GS1[[4$3YOZ:2M(?]\nM+?#+W[VW[YO9ZOKN7Z]\"9D)!M$YF_<DJ:\"?MIQV`LASET4MBZM,#!=IRP]5R\nME$4MW+#%E_X#-`H;N;E&1EGB$?F7D/O)6??-D08#VWUX^.S9#Z=_$SZ>%SY[\nM7\"_Q+'0M?'Q1!;]N/7L:U1H^/#LZ.<$\\OU+Z(^^V;.&[@*W@S`=*;R&.\"V;A\nM?3,])9RTGCWHY)3H#?K:@C_0(0L>>LIH.!X&(%7HZ@H^\"-\\&(F4OBUZ`;^`D\nM^\\R)#I]FV8RI]]UZLP%/CUR?T]7)3\"TEG;!.!N%+.F6=S.13TDGK9#\"VA-/6\nM68\"H4BY<9]'`0X]N3RFEE$TX@&6&6*C?;?(B8_#-V!>CZ0$CVXNANABEQQ?`\nMF)SG4S!6-C[>XIZAA:.\"]T'\\#\"_J^1'+=W-##:6R.%CX4':\\R;RJ<O\"]O1Z.\nM7&4G?\"[&<_:KLO<RX09_X$4\"X'M[#\\^`RU/*W[>#MW!'Q1F'HSM%>??QS1LE\nM<&?CX:07X'Z*B\"4/J\"%[+-0\\5.I[V,0^$'`,5\\E96^6G!6=$*>8<>WK_9=1=\nMBM<C.>[BAS(K*&PIX5EO$#SS9KT^8%/?&T][@>QJ2F%SN965!YO+K;(\\V$U^\nMQ45_:-GZP;XSJR^Z^EQ3BU=?*.V8Z4J%6Z&&C/?W$<VVOC(7X>DTF'5!CGM[\nMH!N!-_)NW=D.?KBWQV2Z%0;D6^&^#OSS(LR[ZB0G/LIUO[.@\\+L/O4U$G$,[\nMFEJ:NOET*DE=V'])\\GC_-9\"G&?#^Y);=+>@\"7O\"WP]/7A^>'.QR#>(_A`%+?\nMPF_L*QS@@_3O[X]?%!$8;J'O/-V)7K*O:\"SP>`+__?&'(OABZPEK'++GX?<S\nMCFQMA2DK1>WR6LG;,7:R;9J07U%B0*W\\ND>>3;#OGN&4X>:#W3#FE@/TS[T9\nM=E8>RN.&94$\\UQ#@6R\\#W[D.:@'NN-?,]J/A/->=8LC6M8Z>VGG!7],5JWZ`\nMEPPON7^<[,WS!?K(T<,=>A:B/F-+[D^5HU\\/3W=V6>DPT/CP-U8QC-W#,^V!\nMAPU.#OI!A^_.3I2S_WR#\\=MDT)L-#L+6O*>ASZLSC'J3F_UI;S``TD)WZ/;:\nM\\V,_\"I[SIVY_>#F$!X83Y>CT^/#\\6#D_?/7F^*\"`;M+1-2=-N=G1DYLKP<TQ\nM=/CW8HHD,L>)>4J]V15WS)$?F*\"@O'T%/V]EES#8)E_H/'6444>Y93Y-R$_\\\nM#NST3Q\\\\=B[PQ/];;S0<[&#G;)<0[,F[VF$;E1UE^U5O`'[??#3:9;T#,]TO\nMX6G\"F3L%U7`G&/UZD^T.\\P:9A[4P.O3\"+OKGL[F+E.!KV/A2CMV6G\\*G^($0\nM84:)3P?A_>V\"3^?OAINT(W3E.)7PFX]N7/Q;[,9MH1>WE7&1?V#H$7^N[\"P.\nM=7>!<:$_S>6S%\\K'ZP=ND!'2'HN>[N&S4@.?P3\"$VU\"4=`Q33^UHVEJ':\"2K\nM7:'1UJU>E<D.QR-)>5;)'M*F$?_P[.3OQ^Q#F(3@UU]?GY[]O4EM*QC2O6H7\nM,JI([RK;=RD!2%`0+FZ4GE.!IEZE.94WK#\"G+C:L,*<N=E#7G,I[+3>G:G8G\nM*HO($WHZ>IS+D$RB_#73WFCD]3/)SN%W%_/A*!A.,`6.K9+>7KLP?6+%/Y:(\nM,G#QPG\"F%;YRY?'U)[EYNLX_;*;>4CZ>'9\\J)\\_>*Z?O/YZ?O#L^VUKXH_#G\nMZOPCKB.J.<!NS4S%:9%;$H!)(4;MXW(FC'AA0:;7#^9@;.EU,7P^HDY1=GB5\nM4_`T1BY+?.GQG(_0]B*'AC7BWLNS%-\"'[][!?Y6G0;2V)T2)0`32[S\"3?3B9\nMSGEB$N]N>S<*!;)($.1Q`#4H&L9NJ\"R+S,&DS?V(X*!W`_Y:<,M?]I1Y<Q'#\nM0MZ$]R9R!O:]\";2/8`K8$7>$7+GU#KB7I:I@#'K*PY22#G!BV%^W>);T\\*(_\nMM8E9<KH-\\I-MT.A4*Z&>:<?X7EW=BU0EFGE!7_;R^C*]OO,+%&:ODL+L*?&R\nM*IO_TT-(M.:>J3\\03?QY:4=];N^&6YX2K$/O)&)8/*9P&9BM/*_-PD5XK%N`\nMQU&E+>8*:YBBE227A\\-[YP7A1:P,CL,L3%_QIB[,<2ZFV#'B0I%V^,M]3\\$%\nM\\[email protected]+H[PT,(_@$3(=>?0Q9'YC-<^3/`WO->X,\\]Y<2_'@[9\nM-+]/]G4M1@TN[_\"'KC_'5>M0\"7APVN%ZHHRC'^+-G\"U\\#-7O*5N(4QB\"P/08\nM[G$O&#ABRY#9%D><Z8*''\\%*N-2$:SGA&':YAXCK/3MCY<^8&\\\"1)E0EWB\":\nM.3B,1-H5KEMEX$*`.\\*]A<SJ5R?!@=2@05/'RNT0-'_BW6)Q&V^,%1=_=V?>\nM_D7/9Q%^;\\821_$^\"Z9KC.V,W#$@%ML<$^F:H1H=0[53ND8<B,.2TSM;6SCV\nM'88;RCBU0)JA<2I:RMOCIW]VAB_5%\\,_CU\\,]_9BZ)PN[!!-=Q-('\".S.#_O\nMZVXBV=TD[\"[+:V6/T9,.`?F:8ZP4W-/<B>0*#?@6RE8D>UR5Y'3'#Q0'N)K5\nM,?1T@&N!5=O)B6C$A9/Q=.1R*T:;/OO/-XZN?'A_=G)^\\OX=(%3DR1Y$#4[=\nMR^?*H?++').0SSWEG+=2SL+%H8[R&J'@W^\"?V2W.NF\";%H>/?7B%-^[Y_[^\\\nM8VV-X[I^5G[%]>*BV=4HGGMG]F$9%>S8E'X(;=/F0V.$6%GC6+6EG>PCM8W(\nM?Q'U!!H\"M1M4/V.GL8O24@5,6P2%-/WBTE`G).1#Z;><<Q\\S=V;O/+5N8[H@\nM[>SLG.<]Y]QS[FOYK46GN\\BZ,_1X/4\\`HU01G]HR%HQ9AD&RCNVZT1&5F`@'\nM2:^GMKP/UPS7G')GYQ$B4-_Q#^S$\"SFY!.7N/G6;)3W?<IHGY'`$\\'6\\H_E%\nM,5^JC##SEG8@]7`K&(&6X#\\S1J_#\"&)4N`MVV=9VSGB0<.)1:MJT\"DJ'CA'U\nMJ'0J6>92)YYATQE+0MX1H@PXWG@V/W(ZB]-<(+39&FU<]0?GHX?$Z-:<A>L&\nMFL9U`U8J\\6Q&B1YMVM`.B#H:SJ,B-`L4<C8\\D%]#&^\"W+(='-G,>&?#(D$<6\nM\\<A,/+)DPL@UJ0DJ)E*8?I.EPAWXI5H<.Q=<EF&>0JM)XG+SIA7PIR\"(0L\\8\nM7(8+B+MB^B35IJ)[;2'59;3X)DYQ6SKG6^<V`U\"O#4JU!1FQ%*^IBV)&4PBM\nMQ)+3.E*\\`!M(S)C,K0W]_D4Y>R+>`HK)./9ITP9Z?NC[%IBIM`7Y,:U$Y5]`\nM+5DU8M13F7C>:)(<>S@&F6(?4ET<?`B&.,\"1/Y:4\"98_DE0(EC^U7@@^/8K4\nM=@I&D3)QIJ?2Z9+#\\N9ENC;3.UAQ(QI!/GWFQR^]8C5$,BP*1.X6&ULC7,#T\nM@^^?QF6%K-TC%E&I/X_6/_S>J_`=I03J&_AC^-<F#?AC[08(Z1\"0D#C@*P.R\nM!K[<%+@%.5$\"B.+$0,[M=A4Y58B8:.([<Y&F1I`0G>*\"HJCG_%EBDIZ'9-<&\nM@TM;_ER6E+1#&O!'.XJHP]^)45`L*OPW)KBI+#,'HNWCS(9_6KX)GSR;=AQ/\nMZV^,S4(YPVIN*4])U/%8&8[3BLIHH%Y,N$P3%5\"OU$P@MHO4XTF/?,)N&<)Q\nME5<H^G&->\"1[;0XR%9_9Z!U7.<?J>&/3'P7]+6(F3VFOH[N'@;S1(+N4VK3+\nMHF-QC&IP.!?(P+B_&:\"[2`X\\G0/<C]<`/KP4#XV1?\\D_-R;KD(@CDE6.Q3I*\nMFPWH:M/!`A=39IJ#V^5>(#27P0?!IB`&72@^U\"B1QL\"12D;I=EEI+@S>H/B(\nMYXD/P8E;A9.T<2I.M#E%4ZN4-5>WQPT%VWF5S]C'INII495T'<I[CS*V@GB`\nM*9L<91%C/&QVH7.C73<Z5OK_W'!-0>U_9[P5N'F&!EP0\\;.\"/L\\,N,VO'KZ]\nMHM%5$U\\E.^!V188*FVT63'6J,U70>B:NOK4QJ$Q!DYY++5?/I*'*E3-94.6J\nMF2SH=#'37G+*%C-IE(9:)N>(\"_>X[>D3`/`Q.C94+'-0R]63E;58JAX/5=AD\nM^@X.5:R*U91)5!Q3?VO=B+,*/G%G:L0B#<J=*\\E\"SN8)X]-YNQ=R``K6]O,Y\nM\"\"Z0/OHFK\\M8?GHQ0CG+3T.5L_PLJ'*6GP4];?FTK.6G41HLOYUM^?@S!5['\nMT88C^0V:LGZL7&4T?]V/UQ)!D(JN6<K$728!N%%7@DBN5TK:5?208663AFXB\nM'KW0'UT0CUMR6Z)<,4-:Y*)_145V\"20PR?YA0V$>B6TGZY/-S2O\\$FKT_G@\\\nMOA)L#M8S\"AVOT[7QF/]8K9V>OE->5ZND5T&O$J**8HWK^*8TF[UHS*Q:M1@L\nM3[?'6B1:I93UT[I>S[&]GCXHWF-PHV=2%R]/7Q_+&0XA>72=D%Q[W\"_QN!`K\nMN=#\"J\"7#U'I*1?%S\\5Q,'FDY1U-^LO9$T2@G'O2%G2+<'ZGW%T<%D3$#*#\\P\nM%@#EQ\\4\"X.FPJ%:U98;%#(P)3!W$Y.5%1=>.C[$49QHN)U]D.-G:PAE@20+G\nM@-^8^/B[?7@JHGHMIU^-\"&&#?%<#?A$;9O0\"WS]^EO2/OGR*'%DF?;)R@A_C\nM=H3,;:ZA:\"-U-N++I[;'0W)R\\3727[S*3VR5H_GJ.?Y31^<WA#MQ>90S?8OE\nM:9P%B58:\\`Y2K33RY\"IM_?[E`/)K?_T8'GCQLP\"//*W@!@;H\"OZ0`UW!,7*P\nM3'N(YU;Q$`/J1`+1!3]9\\CHY.^KP9'?M'$#7[D6I<_A>^'7XY_#)SKWP07@]\nM_\"S\\2_B+\\&'XJ_#V]A_>/46=A_\\`D]J.C`M[\"XN1X>#G(WELW(]>/?/*3Y>(\nMK)!:=E19RR+0VG]\\ZVF3X+9B<O/@[MW=';R!`1O?B?$%I#_^>_@>\\/3[[8_^\nM?7N?]J]]<.V?VP(A0\"YFOA;R+@$R(>X>B'H[/`B?AM>`YN[.28?>?Y(4EU]2\nMEH8LH2@!Z:8A#\\+/.?3?=N[!,P_V7W/HK^]ET;1<J>D%DZ;U4GN&6M;1UM?U\nMX30N7LRKJW<)WZFK?8U^?ANHQ476?+@WCYN&9M<.D%@`]O^^M9-N;:T[M?7M\nM%&I:)%W0_7%3%QJ&`@QGX[T9*9Q3X$&MMM'74;KZNH[:U==U%*^^3JF^7J=]\nM<9CHM..#TVOAX5TGDYUOA&ON)Q<FHM>#_I,MM9TE5G0:8!&A=,[I++5S]F1H\nM-3C^;@V1+2Y-=7TX\"`@_KX+LO_/I+Q^__>!?][]`2^0K=9=P9=HEL>HU_IJ<\nM'O@C7+9*SES>&(V/Q-C.#7U<II;&1RS^'VL/FWSRX<']1W_=?:**0IO<^>KZ\nMY_3D;[Z^<4OL/J,=,9*;1+J!3:PAY9\\I27(V&6&NN38>^KX@6A81,R+\"6I18\nMBF,=EQRNA;IIH(/QPT5'UORC1WMW?[>_^^7[=_8>WKS]P1_G;3+_^&UPI$_O\nMSU=&<^/@YL[>1^_?022??`@>]=5GU9'L_?;FQXCHUI\\0S=/'X%S_N9M`\\V;_\nMW&2RF3*#5$03,2OY\"\">F1RJEKVW>L'UL6$-X6IBZX':9UMRV5%LBGU./*<UP\nMHEPQQL=BV;>EX*G'<J-X6F:Y&RFRX^58G<]*'<4\"4!2@'O]OM5\"`J\\^O`)<V\nM+OID?A70KC[?`GSGN6*?7[Y%9'PYN_OEC8.59R;`K`)\"5`=6%K3U'$M:.CV\"\nMKU3%#I<5AC.2@!5&,LR`^;M@2R`PC%]42KN26!.[8\"D>Q:R2.%/&15V;:D,7\nM^#$:)$N:6IS\\2U.+D_Q$4\"B$XY?@BWH^?182^>LKY:%;&>`+=<9)*A;\\-6K3\nMND5695^`I%OWA2I50A+%LRD0DC32M0$8:\\ZAX8G:@$:U0691,-O\\_O\")_4PR\nK^L.F\\H?+X0^;O!NS]KQTO69:6S>;K)O$U<N=ZJ8L-3.`;P`X2;Q*,+4\"`$PR\n`\nend\n",
"msg_date": "Thu, 23 Apr 1998 13:05:41 +0900",
"msg_from": "[email protected]",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] multi-byte aware char_length() etc. "
},
{
"msg_contents": "> \n> Hi, here are patches I promised (against 6.3.2):\n> \n> * character_length(), position(), substring() are now aware of \n> \t multi-byte characters\n> * add octet_length()\n> * add --with-mb option to configure\n> * new regression tests for EUC_KR\n> (contributed by \"Soonmyung. Hong\" <[email protected]>)\n> * add some test cases to the EUC_JP regression test\n> * fix problem in regress/regress.sh in case of System V\n> * fix toupper(), tolower() to handle 8bit chars\n> \n> note that:\n> \n> o patches for both configure.in and configure are\n> included. maybe the one for configure is not necessary.\n> \n> o pg_proc.h was modified to add octet_length(). I used OIDs\n> (1374-1379) for that. Please let me know if these numbers are not\n> appropriate.\n\nYou obviously used /include/catalog/unused_oids for that. Perfect.\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, 23 Apr 1998 09:52:06 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] multi-byte aware char_length() etc."
},
{
"msg_contents": "\nApplied...can you check that I haven't missed anything?\n\nOn Thu, 23 Apr 1998 [email protected] wrote:\n\n> Hi, here are patches I promised (against 6.3.2):\n> \n> * character_length(), position(), substring() are now aware of \n> \t multi-byte characters\n> * add octet_length()\n> * add --with-mb option to configure\n> * new regression tests for EUC_KR\n> (contributed by \"Soonmyung. Hong\" <[email protected]>)\n> * add some test cases to the EUC_JP regression test\n> * fix problem in regress/regress.sh in case of System V\n> * fix toupper(), tolower() to handle 8bit chars\n> \n> note that:\n> \n> o patches for both configure.in and configure are\n> included. maybe the one for configure is not necessary.\n> \n> o pg_proc.h was modified to add octet_length(). I used OIDs\n> (1374-1379) for that. Please let me know if these numbers are not\n> appropriate.\n> \n\n",
"msg_date": "Mon, 27 Apr 1998 13:07:17 -0400 (EDT)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] multi-byte aware char_length() etc. "
},
{
"msg_contents": ">Applied...can you check that I haven't missed anything?\n>\n>On Thu, 23 Apr 1998 [email protected] wrote:\n>\n>> Hi, here are patches I promised (against 6.3.2):\n\nThanks. I will check after Apr 27 version of\npostgresql.snapshot.tar.gz gets uploaded.\n(I don't have CVSup settings yet)\n--\nTatsuo Ishii\[email protected]\n",
"msg_date": "Tue, 28 Apr 1998 09:52:19 +0900",
"msg_from": "[email protected]",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] multi-byte aware char_length() etc. "
}
] |
[
{
"msg_contents": "Hi,\n\nI want to embedded Tix with Postgres. But I don't know how to do it ?\nWould anyone give me some *hints*?\n\nThanks in advance,\nDoug.\n\n",
"msg_date": "Thu, 19 Mar 1998 14:15:39 +0800",
"msg_from": "Doug Lo <[email protected]>",
"msg_from_op": true,
"msg_subject": "Tix + Postgres."
}
] |
[
{
"msg_contents": "Oracle, like DB/2 can' t set it to on. The autocommit ***on*** in ORA ODBC is a trick.\nIt simply silently adds commit once in a while (after every statement).\nBut please to make the vokabulary easy don't refer to begin, but to 'begin work'\nin this context. 'begin' will mean begin a statement block, not a transaction block.\n\nAndreas\n\n> How do you set autocommit off in ORACLE embedded SQL? I only know this\n> from ODBC.\n\n\n",
"msg_date": "Thu, 19 Mar 1998 16:42:15 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] Begin statement again"
},
{
"msg_contents": "Zeugswetter Andreas writes:\n> Oracle, like DB/2 can' t set it to on. The autocommit ***on*** in ORA ODBC is a trick.\n> It simply silently adds commit once in a while (after every statement).\n\nI know. The problem we were talking about was the automatic beginning of a\ntransaction. No need to say \"begin work\" it just starts.\n\n> But please to make the vokabulary easy don't refer to begin, but to 'begin work'\n> in this context. 'begin' will mean begin a statement block, not a transaction block.\n\nI know that, too. But currently in postgresql begin means the same as begin\nwork. Correct me if I'm wrong.\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, 20 Mar 1998 13:57:02 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Begin statement again"
}
] |
[
{
"msg_contents": ">From my experience, it *** is *** impossible. I would be glad if somebody\ntold me different, but to my understanding the doc is wrong here.\n\nAndreas\n\n>> Currently only 'null' and 'single value'. The executor\n>> doesn't accept anything else for non-sql language functions.\n>> PL functions are treated by the executor like 'C' functions.\n>\n>Actually what I understood from the docs was thatit is 'terribly complicated' and 'beyond the\n>scope of this tutorial', but not impossible ;)\n\n\n\n\n",
"msg_date": "Thu, 19 Mar 1998 16:57:19 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] Re: PL/PgSQL discussion"
},
{
"msg_contents": "\nAndreas Zeugswetter wrote:\n>\n> >From my experience, it *** is *** impossible. I would be glad if somebody\n> told me different, but to my understanding the doc is wrong here.\n>\n> Andreas\n>\n> >> Currently only 'null' and 'single value'. The executor\n> >> doesn't accept anything else for non-sql language functions.\n> >> PL functions are treated by the executor like 'C' functions.\n> >\n> >Actually what I understood from the docs was thatit is 'terribly complicated' and 'beyond the\n> >scope of this tutorial', but not impossible ;)\n\n I have been thinking about that quite a time and when David\n Gould said 'looks like PL functions want to _be_ SQL\n functions', my first action was scratching my head. Then I\n thought it would blow up the parser and executor with things\n they wheren't designed for and dropped these thoughts.\n\n Anyway - thanks David - it was a push into the right\n direction. Not that I think PL functions should become real\n SQL functions processed by the backends main parser and\n executor. But they might become something between 'C' and\n 'SQL'.\n\n The executor could check in ExecMakeFunctionResult() that the\n function is a PL function (simply by looking at\n fcache->func.fn_plhandler). So it is possible that the\n Executor treats PL functions somewhat like\n postquel_function().\n\n My PL/pgSQL is now at the state, that I can write functions.\n I'll leave triggers for later and take a look if that all is\n possible. Would be really nice to enable tuples and sets of\n tuples as PL return types and finally beeing able to do a\n\n DECLARE ret EMP%ROWTYPE;\n BEGIN\n ...\n FOR ret IN SELECT ... LOOP\n RETURN ret AND RESUME;\n ENDLOOP;\n ...\n END;\n\n from PL/pgSQL. Well, returning and resume later might cause\n some trouble with SPI, but let's see.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Thu, 19 Mar 1998 18:10:46 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PL/PgSQL discussion"
},
{
"msg_contents": "\nI wrote:\n>\n>\n> Andreas Zeugswetter wrote:\n> >\n> > >From my experience, it *** is *** impossible. I would be glad if somebody\n> > told me different, but to my understanding the doc is wrong here.\n> >\n> > Andreas\n> >\n> > >> Currently only 'null' and 'single value'. The executor\n> > >> doesn't accept anything else for non-sql language functions.\n> > >> PL functions are treated by the executor like 'C' functions.\n> > >\n> > >Actually what I understood from the docs was thatit is 'terribly complicated' and 'beyond the\n> > >scope of this tutorial', but not impossible ;)\n>\n> The executor could check in ExecMakeFunctionResult() that the\n> function is a PL function (simply by looking at\n> fcache->func.fn_plhandler). So it is possible that the\n> Executor treats PL functions somewhat like\n> postquel_function().\n\n That way it might be possible. The call interface to the PL\n handler must be extended and the PL handler cannot use SPI\n any more. The memory management of SPI makes it impossible\n to return from the function and resume later. Leaving the\n function while connected to SPI would possibly corrupt the\n SPI stack.\n\n And the handling of functions returning tuples is really a\n mess. The executor calls postquel_function() first to get a\n tuple table slot, and then again for each attribute used via\n nested dot expression. It's up to the function itself then to\n create the correct projection from the functions targetlist.\n\n And there must be something in the parser/optimizer too. For\n an 'sql' function, the fcache has a prepared tuple table slot\n on the actual call. But for 'C' or 'PL' functions, it is NULL\n on the call to ExecMakeFunctionResult().\n\n For now, I'll let the PL interface as is and try to finish\n PL/pgSQL in the trigger area. Someday I'll get back to here\n and give it a real try to enable tuples and sets as return\n from functions. That all is far too much work for me right\n now cause the next 4 weeks I have to do some homework (bought\n a new house and that's built now).\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n",
"msg_date": "Fri, 20 Mar 1998 14:50:03 +0100 (MET)",
"msg_from": "[email protected] (Jan Wieck)",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PL/PgSQL discussion"
}
] |
[
{
"msg_contents": "\nI had thought that char2-16 add _no_ functionality over the char() and\nvarchar() types; Tatsuo points out at least one capability which they\nhave. Are there any others?\n\nThey give and take a char * pointer to a C function like\ncreate function upper(char16)\nreturning char16 as '/u/my/upper.so' language 'sql';\nwhereas char() gives a varlena pointer.\n\nAndreas\n\n\n\n\n",
"msg_date": "Thu, 19 Mar 1998 17:55:32 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] varchar() vs char16 performance"
},
{
"msg_contents": "> I had thought that char2-16 add _no_ functionality over the char() and\n> varchar() types; Tatsuo points out at least one capability which they\n> have. Are there any others?\n> \n> They give and take a char * pointer to a C function like\n> create function upper(char16)\n> returning char16 as '/u/my/upper.so' language 'sql';\n> whereas char() gives a varlena pointer.\n\nI don't really see this as a big deal since, for example, only 16 bytes\nare allocated for a char16, so it is not guaranteed to be zero delimited\nand you have to make a working copy to use libc functions anyway. Also,\nthat is really an implementation detail or annoyance rather than a\nuser-visible feature.\n\nWith the macros that are provided for the varlena structure\nmanipulations, things are pretty convenient. Are there more macros which\ncould be helpful here??\n\n - Tom\n",
"msg_date": "Thu, 19 Mar 1998 17:27:38 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] varchar() vs char16 performance"
},
{
"msg_contents": "> I don't really see this as a big deal since, for example, only 16 bytes\n> are allocated for a char16, so it is not guaranteed to be zero delimited\n> and you have to make a working copy to use libc functions anyway. Also,\n> that is really an implementation detail or annoyance rather than a\n> user-visible feature.\n\nAgreed.\n\n> With the macros that are provided for the varlena structure\n> manipulations, things are pretty convenient. Are there more macros which\n> could be helpful here??\n\n>From postgres.h:\n\t\n\t#define VARSIZE(PTR) (((struct varlena *)(PTR))->vl_len)\n\t#define VARDATA(PTR) (((struct varlena *)(PTR))->vl_dat)\n\t#define VARHDRSZ sizeof(int32)\n\nI have gone throught the code before 6.3, and changed many hard-coded\n4's to use these macros.\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 14:50:11 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] varchar() vs char16 performance"
}
] |
[
{
"msg_contents": "> I had thought that char2-16 add _no_ functionality over the char() and\n> varchar() types; Tatsuo points out at least one capability which they\n> have. Are there any others?\n> \n> They give and take a char * pointer to a C function like\n> create function upper(char16)\n> returning char16 as '/u/my/upper.so' language 'sql';\n> whereas char() gives a varlena pointer.\n> \n\nThe char[248] types rip out just fine.\n\nBut that char16 is a whole new beast. It's tentacles are everywhere...\n\n>From comments in various files, the char16 was the original name type\nand was then replaced with 'name'. But there are still a few places\nof inconsistency in the code, namely (*bad pun*) in the cache code.\n\nThere is this eqproc array in catcache.c that is a hack that has to\nmatch the oids of the types from oid 16 to 30, except that F_CHAR16EQ\nis still where F_NAMEEQ should be. Tried renaming it last night, but\ninitdb would blowup the first insert, so there is some other effect in\nthe caching code.\n\nStill other incomplete conversions. In pg_proc.h the arg types for\nchar16eq, lt, le, gt, ge & ne are names (oid 19) when they should be\nchar16 (oid 20)! But char16eq is correctly defined to take char16\nin pg_operator.h.\n\nI'll work on this some more tonite.\n\ndarrenk\n",
"msg_date": "Thu, 19 Mar 1998 12:29:18 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: AW: [HACKERS] varchar() vs char16 performance"
},
{
"msg_contents": "> \n> > I had thought that char2-16 add _no_ functionality over the char() and\n> > varchar() types; Tatsuo points out at least one capability which they\n> > have. Are there any others?\n> > \n> > They give and take a char * pointer to a C function like\n> > create function upper(char16)\n> > returning char16 as '/u/my/upper.so' language 'sql';\n> > whereas char() gives a varlena pointer.\n> > \n> \n> The char[248] types rip out just fine.\n> \n> But that char16 is a whole new beast. It's tentacles are everywhere...\n> \n> >From comments in various files, the char16 was the original name type\n> and was then replaced with 'name'. But there are still a few places\n> of inconsistency in the code, namely (*bad pun*) in the cache code.\n\nYes, you are correct. That was the original name length.\n\nI thought I fixed all the cache name16 references before 6.3. That is\nwhy we can now have index names over 16 characters. Can you confirm\nthis is still a problem in 6.3.\n\n> \n> There is this eqproc array in catcache.c that is a hack that has to\n> match the oids of the types from oid 16 to 30, except that F_CHAR16EQ\n> is still where F_NAMEEQ should be. Tried renaming it last night, but\n> initdb would blowup the first insert, so there is some other effect in\n> the caching code.\n> \n> Still other incomplete conversions. In pg_proc.h the arg types for\n> char16eq, lt, le, gt, ge & ne are names (oid 19) when they should be\n> char16 (oid 20)! But char16eq is correctly defined to take char16\n> in pg_operator.h.\n> \n> I'll work on this some more tonite.\n\nGreat.\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 14:57:13 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] varchar() vs char16 performance"
}
] |
[
{
"msg_contents": "\n>> I had thought that char2-16 add _no_ functionality over the char() and\n>> varchar() types; Tatsuo points out at least one capability which they\n>> have. Are there any others?\n>> \n>> They give and take a char * pointer to a C function like\n>> create function upper(char16)\n>> returning char16 as '/u/my/upper.so' language 'sql';\n>> whereas char() gives a varlena pointer.\n\n>I don't really see this as a big deal since, for example, only 16 bytes\n>are allocated for a char16, so it is not guaranteed to be zero delimited\n>and you have to make a working copy to use libc functions anyway. Also,\n>that is really an implementation detail or annoyance rather than a\n>user-visible feature.\n\nI thought almost all postgresql users write their C extensions ;-) \nit is ***the*** feature of postgresql. I am not saying that this is a real problem.\nIt is just one thing that will be different, but ok for me :-)\n\n>With the macros that are provided for the varlena structure\n>manipulations, things are pretty convenient. Are there more macros which\n>could be helpful here??\n\n\n\n\n",
"msg_date": "Thu, 19 Mar 1998 19:04:18 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] varchar() vs char16 performance"
}
] |
[
{
"msg_contents": "\n\n>On Wed, 18 Mar 1998, David Harvey-George wrote:\n>\n>> I found that triggers don't get inherited... or rather the triggers\n>> for the superclass don't get checked when inserting or updating rows\n>> when enforcing referential intregrity. This is a bit of a pain as it\n>> means I have to duplicate a lot of triggers. Any idea if this can be\n>> fixed? Presumably this would mean 'fixing' to the TRIGGER handling\n>> code to look up the inheritance stack calling the appropriate\n>> triggers?\n>>\n>> Maybe like inheriting PRIMARY KEY attributes this is more of a\n>> feature?\n>\nI'm glad that I'm not the only one who is of the opinion that inheritance is\nneeded for triggers and the like.\n\nRegards,\nMaurice\n\n",
"msg_date": "Thu, 19 Mar 1998 19:50:24 +0100",
"msg_from": "\"Maurice Gittens\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: [QUESTIONS] Inheriting Triggers"
}
] |
[
{
"msg_contents": "Hi.\n\nIm running Postgresql v6.2, perl5.004_04, the latest DBI, and the latest\nDBD::Pg (0.69). I have a script that just executes a query, and tries to\nget the results, and it seems that this particular query doesnt work. Here\nis the code that does not work with DBD::Pg.\n\nuse strict;\nuse DBI;\n\nmy $dbh = DBI->connect(\"dbi:Pg:dbname=gkgis\");\n\nmy $query;\n$query = \"SELECT project, user, created, \";\n$query .= \" 'now'::datetime-datetime(created) AS otime \";\n$query .= \"FROM project_todo \";\n$query .= \"WHERE openincident='TRUE' \";\n$query .= \"ORDER BY created\";\n\nprint \"QUERY : $query\";\nmy $sth = $dbh->prepare($query);\nmy $rc = $sth->execute();\n\n\nThis produces no accessible results! I know the query is okay because I\ncan cut and paste teh query (that the script prints) into a psql session\nand it dumps out the results.\n\nWhat makes me think its a bug in DBD is that if I change the above script\nto have:\n\nmy $query = \"SELECT * FROM project_todo\";\n...\n\ninstead of the long query above, it produces results. So the connection\npart is okay (ie access is granted etc). If I use Pg instead of DBI the\nlong query works, but I would prefer to use DBI. Does anyone have any\nideas what would cause this? Its a quite strange problem, adn I dont know\nhow to debug DBD::Pg :)\n\nMike\n",
"msg_date": "Thu, 19 Mar 1998 14:50:08 -0600 (CST)",
"msg_from": "Michael J Schout <[email protected]>",
"msg_from_op": true,
"msg_subject": "Bug in DBD::Pg v0.69?"
},
{
"msg_contents": "Michael J Schout wrote:\n> \n> Hi.\n> \n> Im running Postgresql v6.2, perl5.004_04, the latest DBI, and the latest\n> DBD::Pg (0.69). I have a script that just executes a query, and tries to\n> get the results, and it seems that this particular query doesnt work. Here\n> is the code that does not work with DBD::Pg.\n> \n> use strict;\n> use DBI;\n> \n> my $dbh = DBI->connect(\"dbi:Pg:dbname=gkgis\");\n> \n> my $query;\n> $query = \"SELECT project, user, created, \";\n> $query .= \" 'now'::datetime-datetime(created) AS otime \";\n> $query .= \"FROM project_todo \";\n> $query .= \"WHERE openincident='TRUE' \";\n> $query .= \"ORDER BY created\";\n> \n> print \"QUERY : $query\";\n> my $sth = $dbh->prepare($query);\n> my $rc = $sth->execute();\n> \n> This produces no accessible results! I know the query is okay because I\n> can cut and paste teh query (that the script prints) into a psql session\n> and it dumps out the results.\n> \n> What makes me think its a bug in DBD is that if I change the above script\n> to have:\n> \n> my $query = \"SELECT * FROM project_todo\";\n> ...\n> \n> instead of the long query above, it produces results. So the connection\n> part is okay (ie access is granted etc). If I use Pg instead of DBI the\n> long query works, but I would prefer to use DBI. Does anyone have any\n> ideas what would cause this? Its a quite strange problem, adn I dont know\n> how to debug DBD::Pg :)\n> \n> Mike\n\nthis is a bug in the scanner, which parses the statement for\nplaceholders.\nIt recognizes the following styles as placeholders: '?', ':1', ':foo',\nbut it does not distiniguish between ':' and '::'.\nI'll fix that in the next release.\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": "Fri, 20 Mar 1998 07:23:23 +0100",
"msg_from": "Edmund Mergl <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Bug in DBD::Pg v0.69?"
}
] |
[
{
"msg_contents": "> > So, how about we fix the storage manager instead?\n> \n> I don't recall, what exactly breaks when going over 2 gig? I don't have the\n> disk space available, otherwise I'd debug this. I can still try if I knew\n> what the problem was...this code isn't all that complex.\n\nI agree.\n\n> OK...here is a patch that will cause the magnetic disk storage manager to\n> not try to split files in 2 gig chunks. It will just try to get another\n> block.\n> \n> If applied, everything is just as before. But if LET_OS_MANAGE_FILESIZE\n> is defined, the chaining disappears and the file just keeps on going,\n> and going, and going, til the OS barfs.\n> \n> Are there #defines in the system includes that could be used to determine\n> a max file size? If so, then I'd think that this would be something\n> to add to configure. If files over 2 gig are not allowed, then the old\n> code would compile.\n> \n> Anyway, if the patch looks ok to the powers-that-be or if there is some\n> thing else to be changed, let me know and I'll resubmit it to PATCHES.\n> \n> Compiled and regressed ok with and without LET_OS_MANAGE_FILESIZE, but\n> then again there aren't any regression tables over 2 gig. :)\n> \n\nWell, BSD/OS goes over 2gig, but the postgreSQL code uses lseek, which\nreturns long, so even though I can handle larger files, the lseek()\ncan't because long is 32-bits. Looks like only Alpha can handle those\nfiles based on our current code.\n\nThanks to our success, this looks like something we will have to deal\nwith.\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 21:47:48 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] tables >2GB"
},
{
"msg_contents": "Applied.\n\n> I don't recall, what exactly breaks when going over 2 gig? I don't have the\n> disk space available, otherwise I'd debug this. I can still try if I knew\n> what the problem was...this code isn't all that complex.\n> \n> OK...here is a patch that will cause the magnetic disk storage manager to\n> not try to split files in 2 gig chunks. It will just try to get another\n> block.\n> \n> If applied, everything is just as before. But if LET_OS_MANAGE_FILESIZE\n> is defined, the chaining disappears and the file just keeps on going,\n> and going, and going, til the OS barfs.\n> \n> Are there #defines in the system includes that could be used to determine\n> a max file size? If so, then I'd think that this would be something\n> to add to configure. If files over 2 gig are not allowed, then the old\n> code would compile.\n> \n> Anyway, if the patch looks ok to the powers-that-be or if there is some\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 23:20:53 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] tables >2GB"
},
{
"msg_contents": "* Bruce Momjian\n|\n| Well, BSD/OS goes over 2gig, but the postgreSQL code uses lseek, which\n| returns long, so even though I can handle larger files, the lseek()\n| can't because long is 32-bits.\n\nAre you sure? In NetBSD, lseek() is declared to return an off_t,\nwhich again is defined to be a 64bit quantity. I would assume that\nBSD/OS did it the same way -- in fact, I'd be surprised if not.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n",
"msg_date": "20 Mar 1998 17:40:01 +0100",
"msg_from": "Tom Ivar Helbekkmo <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] tables >2GB"
},
{
"msg_contents": "> \n> * Bruce Momjian\n> |\n> | Well, BSD/OS goes over 2gig, but the postgreSQL code uses lseek, which\n> | returns long, so even though I can handle larger files, the lseek()\n> | can't because long is 32-bits.\n> \n> Are you sure? In NetBSD, lseek() is declared to return an off_t,\n> which again is defined to be a 64bit quantity. I would assume that\n> BSD/OS did it the same way -- in fact, I'd be surprised if not.\n\nOops, you are right:\n\n\ttypedef quad_t off_t;\n\nI thought they added fgetpos() only for 64-bit quantities, and did not\nchange the return value of lseek. However:\n\n\tsys/types.h:76: typedef quad_t off_t; /* file offset*/\n\nso you are right, but our code:\n\n\tfd.c:110: long seekPos;\n\tfd.c:263: fileP->seekPos = (long) lseek(fileP->fd, 0L, SEEK_CUR);\n\nso it still will not work because the code is not defining seekPos as\noff_t. We need to get this code cleaned up/fixed.\n\nHow could they make such a mistake and assume it is a long, unless this\nthing gets passed around in the backend, and they don't want to\nreference off_t all over the place? That code needs cleanup.\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 12:29:18 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] tables >2GB"
},
{
"msg_contents": "* Bruce Momjian\n|\n| How could they make such a mistake and assume it is a long, [...]\n\n\"All the world's a VAX!\" Getting rid of (and tidying up after)\nexplicit \"long lseek();\" declarations is a major part of porting\nold software to new UNIXes.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n",
"msg_date": "21 Mar 1998 13:52:29 +0100",
"msg_from": "Tom Ivar Helbekkmo <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] tables >2GB"
},
{
"msg_contents": "> \n> * Bruce Momjian\n> |\n> | How could they make such a mistake and assume it is a long, [...]\n> \n> \"All the world's a VAX!\" Getting rid of (and tidying up after)\n> explicit \"long lseek();\" declarations is a major part of porting\n> old software to new UNIXes.\n\nIt's on our TODO list now.\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 09:56:58 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] tables >2GB"
}
] |
[
{
"msg_contents": "OK, tonight's mega-patch should be pretty close to the final patch. We\nhad to have the Alpha fix, and I put in the other fixes that looked easy\nto understand. Most were port-specific cleanup people had posted.\n\nIf it breaks, let Marc (Oh, no, I mean the hackers list) know.\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 23:27:55 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "tonight's mega-patch"
},
{
"msg_contents": "Bruce Momjian writes:\n> \n> OK, tonight's mega-patch should be pretty close to the final patch. We\n> had to have the Alpha fix, and I put in the other fixes that looked easy\n> to understand. Most were port-specific cleanup people had posted.\n> \n> If it breaks, let Marc (Oh, no, I mean the hackers list) know.\n\nI have decided based on some thought and some discussion with Bruce to hold\noff trying to get the S_LOCK backoff change into the mega-patch. While it\nis a nice idea, and I even have it all coded, I would like to do some extra\ntesting and paw around some other code a bit as there is a possibility of it\nsurfacing some problems with respect to the way signals are handled.\n\nIt could I suppose be put out for people to experiment with...\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 00:43:43 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] tonight's mega-patch"
},
{
"msg_contents": "On Thu, 19 Mar 1998, Bruce Momjian wrote:\n\n> OK, tonight's mega-patch should be pretty close to the final patch. We\n> had to have the Alpha fix, and I put in the other fixes that looked easy\n> to understand. Most were port-specific cleanup people had posted.\n\n\tBuilding a new patch right now...its not automatic...meant to do\none this afternoon, but had a server at work causing me headaches :(\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Fri, 20 Mar 1998 19:25:09 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] tonight's mega-patch"
}
] |
[
{
"msg_contents": "Marc, does the mega-patch get regenerated automatically every night? If\nnot, people will not be able to see tne new patch items until you\nregenerate one.\n\nThanks.\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 23:47:33 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Mega-patch tonight"
}
] |
[
{
"msg_contents": "\nexpr ? trueval : falseval\n\nif expr evaluated is true, the entire expression is trueval evaluated,\notherwise it is falseval evaluated.\n\nIs it possible to have this sort of grammar logic in the parser?\n\n--brett\n",
"msg_date": "Fri, 20 Mar 1998 00:07:03 -0800 (PST)",
"msg_from": "Brett McCormick <[email protected]>",
"msg_from_op": true,
"msg_subject": "expr ? trueval : falseval"
},
{
"msg_contents": "Brett McCormick wrote:\n> \n> expr ? trueval : falseval\n> \n> if expr evaluated is true, the entire expression is trueval evaluated,\n> otherwise it is falseval evaluated.\n> \n> Is it possible to have this sort of grammar logic in the parser?\n\nI suspect that this is covered in the SQL92 CASE expression, which we\ndon't yet support.\n\n - Tom\n",
"msg_date": "Fri, 20 Mar 1998 17:02:55 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] expr ? trueval : falseval"
},
{
"msg_contents": "> \n> \n> expr ? trueval : falseval\n> \n> if expr evaluated is true, the entire expression is trueval evaluated,\n> otherwise it is falseval evaluated.\n> \n> Is it possible to have this sort of grammar logic in the parser?\n\nGood question. Answer is no, I think, unless you can do some fancy\nthings with functions.\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 12:03:31 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] expr ? trueval : falseval"
}
] |
[
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nI've finally figured out why the Alpha fails when\nHAVE_INT_TIMEZONE is defined. As far as I'm concerned, somebody\nat DEC should be shot. The problem has nothing to do with\npostgresql code. The symbol 'timezone' is defined in two\ndifferent libraries, as two very different things! libc.a\ndefines it as a global variable of type long, as documented in\nthe timezone(3) manpage. However, libbsd.a defines a function\nnamed timezone. Since we link explicitly to -lbsd, guess which\ndefinition gets linked into the program? We're using the address\nof that function as our timezone offset. Well, the low four\nbytes of the address, anyway.\n\nAs far as I can tell, we don't require any of the routines used\nin libbsd.a. I'm about to do some more testing to confirm this,\nso hopefully a patch will be on its way soon. Up to this point,\nall I've recompiled under this new model is postgres, which links\nfine, and runs without error. The only failure on the date tests\nthat I noticed seemed to be using PST when it should have been\nPDT. I suspect that's an artifact of where I am (Michigan)\nwhich, according to the /etc/zoneinfo/localtime source, didn't\nobserve DST between 1968 and 1973.\n\n- --\nDwayne Bailey + WHAT is your name? Sir Galahad\nMIKA Systems, Bingham Farms, MI + WHAT is your quest? I Seek the Holy Grail\[email protected] + What is your favorite color?\nhttp://www.mika.com/~dwayne + Blue ... no, Yelloooooooooooooooooow\n finger [email protected] for PGP Public Key\n\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQB1AwUBNRKDVKA2uleK7maRAQE3AQMAmJ41hAwHc/H1m+8XxoxCxwvxcS2S3Eki\nfDQI7Bhd2QAPM8O8ufUaCejmSEbc2cBrJrYETukXzaQ9WdqcD1wZydLQv9h6i/4i\n+g8lq48LxkBYT5QDlW3glHL7uNV+X6YS\n=XQms\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "Fri, 20 Mar 1998 09:55:08 -0500 (EST)",
"msg_from": "Dwayne Bailey <[email protected]>",
"msg_from_op": true,
"msg_subject": "DEC Alpha and HAVE_INT_TIMEZONE"
},
{
"msg_contents": "> \n> -----BEGIN PGP SIGNED MESSAGE-----\n> \n> I've finally figured out why the Alpha fails when\n> HAVE_INT_TIMEZONE is defined. As far as I'm concerned, somebody\n> at DEC should be shot. The problem has nothing to do with\n> postgresql code. The symbol 'timezone' is defined in two\n> different libraries, as two very different things! libc.a\n> defines it as a global variable of type long, as documented in\n> the timezone(3) manpage. However, libbsd.a defines a function\n> named timezone. Since we link explicitly to -lbsd, guess which\n> definition gets linked into the program? We're using the address\n> of that function as our timezone offset. Well, the low four\n> bytes of the address, anyway.\n\nI assume they are supporting both uses for the symbol. I recommend\nhard-coding the stuff into the port, and hopefully it will work. Please\nsend it patches soon. If it is alpha-specific, we can fold it into the\nmega-patch, because alpha was broken in 6.3 anyway.\n\n> \n> As far as I can tell, we don't require any of the routines used\n> in libbsd.a. I'm about to do some more testing to confirm this,\n> so hopefully a patch will be on its way soon. Up to this point,\n> all I've recompiled under this new model is postgres, which links\n> fine, and runs without error. The only failure on the date tests\n> that I noticed seemed to be using PST when it should have been\n> PDT. I suspect that's an artifact of where I am (Michigan)\n> which, according to the /etc/zoneinfo/localtime source, didn't\n> observe DST between 1968 and 1973.\n\nYea, I see this sometimes too on BSDI.\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 12:19:42 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] DEC Alpha and HAVE_INT_TIMEZONE"
},
{
"msg_contents": "> > I've finally figured out why the Alpha fails when\n> > HAVE_INT_TIMEZONE is defined. As far as I'm concerned, somebody\n> > at DEC should be shot. The problem has nothing to do with\n> > postgresql code. The symbol 'timezone' is defined in two\n> > different libraries, as two very different things! libc.a\n> > defines it as a global variable of type long, as documented in\n> > the timezone(3) manpage. However, libbsd.a defines a function\n> > named timezone. Since we link explicitly to -lbsd, guess which\n> > definition gets linked into the program? We're using the address\n> > of that function as our timezone offset. Well, the low four\n> > bytes of the address, anyway.\n> \n> I assume they are supporting both uses for the symbol. I recommend\n> hard-coding the stuff into the port, and hopefully it will work. \n> > As far as I can tell, we don't require any of the routines used\n> > in libbsd.a. I'm about to do some more testing to confirm this,\n> > so hopefully a patch will be on its way soon. Up to this point,\n> > all I've recompiled under this new model is postgres, which links\n> > fine, and runs without error. The only failure on the date tests\n> > that I noticed seemed to be using PST when it should have been\n> > PDT. I suspect that's an artifact of where I am (Michigan)\n> > which, according to the /etc/zoneinfo/localtime source, didn't\n> > observe DST between 1968 and 1973.\n> \n> Yea, I see this sometimes too on BSDI.\n\nOne of the other ports (AIX?) had this problem and they could just\nremove the libbsd from the link arguments.\n\n - Tom\n",
"msg_date": "Sat, 21 Mar 1998 04:42:24 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] DEC Alpha and HAVE_INT_TIMEZONE"
}
] |
[
{
"msg_contents": "Hi all\n\nA very simple question I hope !\n\nI installed postgreSQL 6.3 and on starting it \"postmaster -i\", I get:\n\ngenx:/home/postgres:postgres> FATAL: StreamServerPort: bind() failed:\nerrno=98\n Is another postmaster already running on that port?\n If not, wait a few seconds and retry.\npostmaster: cannot create INET stream port\n\nI've checked that there isn't another copy of postmaster running. I also\nchecked /tmp/ for any postgres files but none there. Does anyone know\nwhat may be the problem here.\n\nI'm running Linux RH4.2 with 2.0.30 kernel.\n\nThanks\nNishi\n\n",
"msg_date": "Fri, 20 Mar 1998 09:06:09 -0600",
"msg_from": "\"Kapoor, Nishikant X\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "newbie question"
}
] |
[
{
"msg_contents": "I have noted with 2000+ records in a table (using a btree unique index on\nthe primary key, btw) performance of a query takes 15+ minutes to respond.\n\nThe query is in the structure of this\n\nselect foo from bar where foo in (select foo from bar where bazzfoo like\n'Word%');\n\nJust reporting a possible bug..\n\nWith smaller amounts of records, responce time is within a few seconds.\n\nOther information you might want:\n\npostgres version: 6.3\nos: BSDI 3.1\n\nThanks\[email protected]\n\n",
"msg_date": "Fri, 20 Mar 1998 11:31:17 -0500",
"msg_from": "Coronach <[email protected]>",
"msg_from_op": true,
"msg_subject": "psql nested queries with 2000+ records"
},
{
"msg_contents": "> \n> I have noted with 2000+ records in a table (using a btree unique index on\n> the primary key, btw) performance of a query takes 15+ minutes to respond.\n> \n> The query is in the structure of this\n> \n> select foo from bar where foo in (select foo from bar where bazzfoo like\n> 'Word%');\n\nVery strange. 15+ minutes? Wow, that is terrible, even longer than a\nsequential scan of the table. Try EXPLAIN and tell us what it says in\nthe two cases.\n\n> \n> Just reporting a possible bug..\n> \n> With smaller amounts of records, responce time is within a few seconds.\n> \n> Other information you might want:\n> \n> postgres version: 6.3\n> os: BSDI 3.1\n\nSame OS here.\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 12:24:05 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "At 12:24 PM 3/20/98 -0500, Bruce Momjian wrote:\n\n>Very strange. 15+ minutes? Wow, that is terrible, even longer than a\n>sequential scan of the table. Try EXPLAIN and tell us what it says in\n>the two cases.\n\nThis is within the 2717 record database.\n\nexplain select name from games where name in (select name from games where\nname like 'A%');\n\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=207.95 size=446 width=12)\n SubPlan\n -> Seq Scan on games (cost=207.95 size=1 width=12)\n\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=207.95 size=446 width=12)\n SubPlan\n -> Seq Scan on games (cost=207.95 size=1 width=12)\n\nEXPLAIN\n\nThis is within the 24 record database of the same type of data\n\nexplain select name from games where name in (select name from game\ns where name like 'A%');\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=207.95 size=446 width=12)\n SubPlan\n -> Seq Scan on games (cost=207.95 size=1 width=12)\n\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=207.95 size=446 width=12)\n SubPlan\n -> Seq Scan on games (cost=207.95 size=1 width=12)\n\nEXPLAIN\n\nObviously, I get the same information, any suggestions?\n\[email protected]\n\n",
"msg_date": "Fri, 20 Mar 1998 17:24:50 -0500",
"msg_from": "Coronach <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "OK, I think I see the problem. Indexes are not being used in this case.\nBasically for every row in the outer query, you are doing a sequential\nscan in the inner table.\n\nCan you try this:\n\nselect * into games2 from games;\nexplain select name from games where name in (select name from games2 where\nname like 'A%');\n\n\nI would like to know if the use of the same table in the subquery is\ncausing the problem. I assume you have run vacuum and vacuum analyze.\n\n\n> \n> At 12:24 PM 3/20/98 -0500, Bruce Momjian wrote:\n> \n> >Very strange. 15+ minutes? Wow, that is terrible, even longer than a\n> >sequential scan of the table. Try EXPLAIN and tell us what it says in\n> >the two cases.\n> \n> This is within the 2717 record database.\n> \n> explain select name from games where name in (select name from games where\n> name like 'A%');\n> \n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=207.95 size=446 width=12)\n> SubPlan\n> -> Seq Scan on games (cost=207.95 size=1 width=12)\n> \n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=207.95 size=446 width=12)\n> SubPlan\n> -> Seq Scan on games (cost=207.95 size=1 width=12)\n> \n> EXPLAIN\n> \n> This is within the 24 record database of the same type of data\n> \n> explain select name from games where name in (select name from game\n> s where name like 'A%');\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=207.95 size=446 width=12)\n> SubPlan\n> -> Seq Scan on games (cost=207.95 size=1 width=12)\n> \n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=207.95 size=446 width=12)\n> SubPlan\n> -> Seq Scan on games (cost=207.95 size=1 width=12)\n> \n> EXPLAIN\n> \n> Obviously, I get the same information, any suggestions?\n> \n> [email protected]\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": "Fri, 20 Mar 1998 22:38:10 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": ">select * into games2 from games;\n\namusements=> select * into games2 from games;\nERROR: parser: parse error at or near \"games2\"\n\nBut, I just dumped it and changed a few things to fit the scenario...\n\n>explain select name from games where name in (select name from games2 where\n>name like 'A%');\n\namusements=> explain select name from games where name in (select name from\ngame\ns2 where name like 'A%');\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=207.95 size=446 width=12)\n SubPlan\n -> Seq Scan on games2 (cost=10.23 size=1 width=12)\n\nEXPLAIN\n\nUpon execution of the query, it still took quite a long time. \n\n>causing the problem. I assume you have run vacuum and vacuum analyze.\n\nYep, ran vacuum on it and also had it work with newly generated databases too.\n\n\nAny other suggestions would be welcome, thanks\n\[email protected]\n\n",
"msg_date": "Sat, 21 Mar 1998 01:12:07 -0500",
"msg_from": "Coronach <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "Coronach wrote:\n> \n> I have noted with 2000+ records in a table (using a btree unique index on\n> the primary key, btw) performance of a query takes 15+ minutes to respond.\n> \n> The query is in the structure of this\n> \n> select foo from bar where foo in (select foo from bar where bazzfoo like\n> 'Word%');\n\nEXPLAIN select foo from bar where bazzfoo like 'Word%';\n\nIs index used ?\n\nVadim\n",
"msg_date": "Sat, 21 Mar 1998 21:16:07 +0700",
"msg_from": "\"Vadim B. Mikheev\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "> \n> >select * into games2 from games;\n> \n> amusements=> select * into games2 from games;\n> ERROR: parser: parse error at or near \"games2\"\n\nOops, we haven't released 6.3p1 yet, so 'select * into table games2 from\ngames;'\n\n> \n> But, I just dumped it and changed a few things to fit the scenario...\n> \n> >explain select name from games where name in (select name from games2 where\n> >name like 'A%');\n> \n> amusements=> explain select name from games where name in (select name from\n> game\n> s2 where name like 'A%');\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=207.95 size=446 width=12)\n> SubPlan\n> -> Seq Scan on games2 (cost=10.23 size=1 width=12)\n> \n> EXPLAIN\n> \n> Upon execution of the query, it still took quite a long time. \n> \n> >causing the problem. I assume you have run vacuum and vacuum analyze.\n> \n> Yep, ran vacuum on it and also had it work with newly generated databases too.\n\nTry it without the LIKE, with just an equals.\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 09:54:54 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "At 09:54 AM 3/21/98 -0500, Bruce Momjian wrote:\n\n>Try it without the LIKE, with just an equals.\n\namusements=> explain select name from games where name in (select name from\ngame\ns where mfr = '');\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=207.95 size=446 width=12)\n SubPlan\n -> Seq Scan on games (cost=207.95 size=1 width=12)\n\nEXPLAIN\n\n--\n\namusements=> explain select name from games where name in (select name from\ngame\ns2 where mfr = '');\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=207.95 size=446 width=12)\n SubPlan\n -> Seq Scan on games2 (cost=10.23 size=1 width=12)\n\nEXPLAIN\n\nThe first query approved to be lengthy, but the last query was executed and\nreturned rows in just a few.\n\nI then did an explain on a query that refered to the games table within the\nsub query.\n\namusements=> explain select name from games where name in (select name from\ngame\ns where mfr = '');\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=207.95 size=446 width=12)\n SubPlan\n -> Seq Scan on games (cost=207.95 size=1 width=12)\n\nEXPLAIN\n\n\nFor the nature of the searches that the front end send to postgres (the\ndatabase is used with a web interface), it is nec. to use the like\nexpression. Where should I go from here?\n\nThanks once again,\n\[email protected]\n\n",
"msg_date": "Sat, 21 Mar 1998 12:31:08 -0500",
"msg_from": "Coronach <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "What does this show in explain? Does it use an index?\n\n\tselect name from games where mfr = ''\n\n> \n> At 09:54 AM 3/21/98 -0500, Bruce Momjian wrote:\n> \n> >Try it without the LIKE, with just an equals.\n> \n> amusements=> explain select name from games where name in (select name from\n> game\n> s where mfr = '');\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=207.95 size=446 width=12)\n> SubPlan\n> -> Seq Scan on games (cost=207.95 size=1 width=12)\n> \n> EXPLAIN\n\nI ran this test:\n\n\ttest=> explain select * from pg_proc where oid in (select oid from\n\tpg_proc where oid =3);\n\tNOTICE: QUERY PLAN:\n\t\n\tSeq Scan on pg_proc (cost=48.90 size=91 width=122)\n\t SubPlan\n\t -> Index Scan on pg_proc (cost=2.05 size=1 width=4)\n\nand it used the index in the subquery. Please let us about the above\ntest.\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 22:17:25 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "Aparently, the postgresql server was moved from the BSDI box to a linux\n2.0.30 box, so the numbers that where given where not from the bsdi box. So\nnow I am doing the queries personally. I'm sorry for the run around.\n\nThis is the responce from the explain on the BSDI 3.1 box.\n\namusements=> explain select name from games where name in (select name from\ngame\ns where name like 'A%');\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=0.00 size=0 width=12)\n SubPlan\n -> Seq Scan on games (cost=0.00 size=0 width=12)\n\nEXPLAIN\n\namusements=> explain select name from games where name in (select name from\ngame\ns2 where mfr = '');\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=0.00 size=0 width=12)\n SubPlan\n -> Seq Scan on games2 (cost=0.00 size=0 width=12)\n\nEXPLAIN\n\namusements=> explain select * from games where mfr = '';\nNOTICE: QUERY PLAN:\n\nSeq Scan on games (cost=0.00 size=0 width=160)\n\nEXPLAIN\n\nOnce again, I appologize for this misinformation and hope this sheds some\nlight to the original problem. I'm making a note to change permissions on\nwho can do what from now on.\n\nOn a side note, the query was done on the 2.0.30 box in 5 minutes compared\nto the BSDI box at 15+.\n\[email protected]\n\n",
"msg_date": "Sat, 21 Mar 1998 23:21:39 -0500",
"msg_from": "Coronach <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "> amusements=> explain select * from games where mfr = '';\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=0.00 size=0 width=160)\n> \n> EXPLAIN\n\nOK. This is the base of the problem. You are vacuum. How varied is the\ndata in the mfr field. Is the usually just a few unique values. If so,\nthe index really does you no good, and perhaps that is why it is not\nbeing used.\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, 22 Mar 1998 00:30:01 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
},
{
"msg_contents": "[email protected] blushes and says:\n> Aparently, the postgresql server was moved from the BSDI box to a linux\n> 2.0.30 box, so the numbers that where given where not from the bsdi box. So\n> now I am doing the queries personally. I'm sorry for the run around.\n> \n> This is the responce from the explain on the BSDI 3.1 box.\n> \n> amusements=> explain select name from games where name in (select name from\n> game\n> s where name like 'A%');\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=0.00 size=0 width=12)\n> SubPlan\n> -> Seq Scan on games (cost=0.00 size=0 width=12)\n> \n> EXPLAIN\n> \n> amusements=> explain select name from games where name in (select name from\n> game\n> s2 where mfr = '');\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=0.00 size=0 width=12)\n> SubPlan\n> -> Seq Scan on games2 (cost=0.00 size=0 width=12)\n> \n> EXPLAIN\n> \n> amusements=> explain select * from games where mfr = '';\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on games (cost=0.00 size=0 width=160)\n> \n> EXPLAIN\n> \n> Once again, I appologize for this misinformation and hope this sheds some\n> light to the original problem. I'm making a note to change permissions on\n> who can do what from now on.\n> \n> On a side note, the query was done on the 2.0.30 box in 5 minutes compared\n> to the BSDI box at 15+.\n\nWell of course... ;-)\n\nI am sure this has been mentioned, but could you post the exact schema for\nthese tables and their indexes? It might even be worth dropping and\nrecreating the indexes just to be _sure_.\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": "Sun, 22 Mar 1998 01:00:56 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] psql nested queries with 2000+ records"
}
] |
[
{
"msg_contents": "Ok,\n\nI think I've found the problem.\nThe problem as I've identified it, seems to be a result of the fact that\npointers returned by palloc are not alligned in accordance with\nthe DOUBLEALIGN macro.\n\nThe function ComputeDataSize returns a tuple size\ncomputed using an initial data size of 0 bytes. (which is\ndouble aligned according to the DOUBLEALIGN macro).\nThis needless to say, makes sense.\n\nHowever the function DataFill is called with an expression similar to:\n\nDataFill((char *)tuple + tuple->t_hoff, ...) ;\n\nSo it happens that the address for tuple data passed to the function\nDataFill,\nis not double aligned in the general case.\nSo every now and then (when the tuple data pointer is not double aligned)\nthe DataFill function actually uses memory more than it should,\nbecause the alignment macros bump the data pointer while no corresponding\nspace has been reserved by the function ComputeDataSize.\n\nI think the source of the problem lies in the fact that palloc does not\nreturn memory addresses which are double aligned.\n\nI don't know how to solve the problem of making palloc return\n MAXALIGNED addresses in a _portable_ (compiler, os, cpu ) way.\n\nCan any one confirm this analyses and maybe give suggestions for a\nportable solution?\n\nWith regards from Maurice.\n\n\n",
"msg_date": "Fri, 20 Mar 1998 19:30:55 +0100",
"msg_from": "\"Maurice Gittens\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: Buffer overruns with the Electric Fence debugging\n\tlibrary (Solved?)"
}
] |
[
{
"msg_contents": "We are preparing a 6.3 patch. Is this patch ready to be used by the\nuser community, or are you still working on it?\n\n> \n> Hi!\n> \n> I started adding the Having Clause and it works quite fine for\n> sequential scans! (I think it will also work with hash, index, etc but\n> I did not check it out! I made some High level changes which should\n> work for all access methods, but maybe I'm wrong. Please let me know.)\n> \n> Now it is possible to make queries like:\n> \n> select s.sname, max(p.pid), min(p.pid)\n> from part p, supplier s\n> where s.sid=p.sid\n> group by s.sname\n> having max(pid)=6 and min(pid)=1 or avg(pid)=4;\n> \n> Having does not work yet for queries that contain a subselect statement \n> in the Having clause, I'll try to fix this in the next days.\n> \n> If there are some bugs, please let me know, I'll start to read the\n> mailinglists now!\n> \n> Now here is the patch against the original 6.3 version (no snapshot!!):\n> \n> It is encoded using uuencode and called: \n> postgresql-6.3.tar.gz_to_having.diff.gz.uu\n> \n> If your original 6.3 source tree is in:\n> tmp/postgresql-6.3/ \n> perform the following steps to apply the patch:\n> \n> $ cd tmp/\n> $ uudecode postgresql-6.3.tar.gz_to_having.diff.gz.uu\n> $ gunzip postgresql-6.3.tar.gz_to_having.diff.gz\n> $ patch < postgresql-6.3.tar.gz_to_having.diff\n> \n> Hopefully you will see only \"Hunk succeeded...\" messages :-)\n> \n> Regards\n> \n> Stefan\n> \n> \n> -----------------------Cut here----------------------------------------\n> begin 644 postgresql-6.3.tar.gz_to_having.diff.gz\n> M'XL(\"&QL\"34``W!O<W1G<F5S<6PM-BXS+G1A<BYG>E]T;U]H879I;F<N9&EF\n> M9@#-6VUSVD@2_DQJ?\\0XF\\H!%B`A\\5Y.%>>0\"W>.G;79[*7VMCA9#*\"+D!2-\n> MY(3-^K]?3\\^,)%X-6:>RE`/2:*:G>[K[F>[6)`Q8/(LH^^A5FE6SQB*G=FL[\n> M'Z@_J='/U$GB(,*+GQ+;JSJD6JV%QXV8N-,IC;K$F=L1:>HM4R.>ZU-2UYM/\n> MRN4R.8Y<831/R\"MZ2^I-HEM=T^@:)C$ZG?:32J5R/'>%5Y%+W@!CG(K5M9I=\n> MHRW(E5<_R&I=-[6ZH1-L((3%=NPZY*4=)PNX'0#5P9WM]6<P_ZPH?DC9GLTT\n> M>!9&YX$?T\\\\Q*5-'7&GD-@@\\4G;99>)Y)2#QY<D)@;^\";\")G1/6MO*#.YWCL\n> M0RO[%4A67L\"7'_S6@U&%B,9)Y*_W!582NM'Y'O[A6J$P>H?`=>6;\"7.,)(6O\n> M$F6;GII66VLV3*$G+FO3:D&#I60MX\">=1,N:GMO1;(@<YQJQ[67@TU+OR2GA\n> M?_R1.R7%DV(Y>UC\"YB]B7)\\Q&L5%EYR=$1W&;6/3,!M-S3\";.4:AJ<6;6BFK\n> MI(R7^8]HKHGE8HD7P^I.;8_1GN1M&D34=N9%Q[,31C7R$6P=N?L\"'0H$OX00\n> MZ?B!=(ES'%(L7@832LHEXDW=B,624DE+UZS42Q=!T3@C<910L0J%6^#@0^_)\n> M(=[HPU1@7X>#2S9@!5O:[::$%K-^,+*DM+8`2^MH8,FHK>)*H]VUC#VXTM%,\n> M*S,\"?MM0!O\"CZSM>`LIXRF`&>T9KM\\ET,8NJ\\Z<K3Y/8]5@MM#TO<+8_8TOF\n> M@%50_O0T]S0(8W?A_DZCFM`R$\\/AK\\91#@P01'J5^,[0GP:DLMV4]8YF`'1F\n> MAFSHFE%OY#U.(($!QO:J?W$ST%::ZZJ9VY7PLD(-B-V47\\-WC=].`O*%9`Y8\n> MVW2,G'<49C1&3*/$Y6Q/HV!!N'IV^-2*8XD)`&@$@3,<B,B##6CZW/)5`SX:\n> M3S@(2*]\"$+O\\^>)BN^?70<7U%96;3:UNU7/+=8\\_^'V?.NP.N:7(!/R><,<G\n> M<1)Z5,@<SRG^\"Y*81B3T;%\\C=AAZ2P*F@D]LQ'80@VWEU6JU-*O=RGBUV@VM\n> MH9LYA\"HP+XB)[4^(E-R-JP]@%ZI1?=84C?)^FKL>X)#\"I>+4_3P.0G?\"BD(9\n> M7)(J`IN68M+)&2(0>?Y<8AS_K/<_.>-Z*2DL+^\"B\\FENP+_HB*]<$32)%\\I&\n> M@4RF:X?A115^QS>.[6//&UB!7.^A#UN6._E[PO$IUXXP+;!S;8L#W_2%VLX.\n> MF4M!O0!?WB)!_&T4_(\\Z<3&$7^ZP&GGNJLU+&!\"W7\"5@:1]\"9\\#`%P^_?!H]\n> MB-,[AZV@M='2+0G7S0?A>A?)==\"N=QO&H:\"]D^8&=)OF;NAN=+1FSH_AUFBE\n> MKD$$?C[FIY;\"8\\YIX)KT_:GMSU+7DN'<A<L@1H-[P'WGPWANW[G^;,S=8`Q@\n> M,>:&)K=Z%2N($2RYY0LR@NB&QKRI!#0P>N`WW)++:(+9+4=`CX\\](Y?#\"^5<\n> MW-:&K)]&(N_LJ,1)R4B$B`D&?APMTTFA3T\\B'G[7RM*9R^2U?4+ZG`@B\\HEJ\n> MKXF+;#PPL;!C9PZ!8^1.BD7>!,&,XF)3NET3GL]A406``BD_(,_YKQW'<#EU\n> MJ3=A))BJQ\\A458U=8VZ-B5+E14KH+,=YY058ZR18X*\\?2+[4AB(6%@%R!W92\n> MP!?\"5]UEXRGLV6(V%;^1/_X`)N/Q9BL?6R!@$IN/`-?76E,-XBBQ7!SL1W.7\n> M$?BS\"9\\Z=@._2JXA-(J8>T=AQW'$E@,](MB.N.OS'<N-F2(`\"DD6U(]9M5J5\n> M;2A5&M%\"FL'WA'!U)6$80YZ$614*.6/T`6:=HFK0Q'/$XCTN(:/>N+3-5H2Q\n> MB%U*;?6*?JJ>3!%Y\\Q=I5,X#U'@/N&2JFY!\\/X?%-\"7+K4.,3.[A^0'NHLA>\n> M7M-IGC_5ABX>P>\\9*:9MZ=R]/;8`B\\/[P]@CC8&\"FBEC8$486L`3\".4F]'/N\n> MP1XSX<QR)YHF84@C'!A&QQK)HYC(-J:\\X-/W9>K!>?9-DXK!V=\\SU>-,8C,V\n> M\\_?.=*`G`B0&X9%(YI,`S,>&T/`0Z[T-XGG>A#TZE7&Q.YO'@I0_82LVR[=?\n> MW\"BP,[B72KU!Q#%O\"\\(-R.NMCA34UX9BXXZQ.)PO\"<YYDAN)L?&CF\")0WF,:\n> M6#A`MK_)[$CZS]I+'AO?PF:]X-MAOO$<0#O.HR7$89\"EP@X[!Y9.E(:W[-XX\n> MR0[;NPJ\"D)V03Q2,S.?$\"`0A$X]'(79\\LF([U`MFQ<'U]=6U1I[N7HXN.5^G\n> M!(:;ZWGR'_^I5,P69N\\WH\\V!SQ562[/QM[#*&.W*,+KX4T)Y2!?:$:.B&+@M\n> M>C;:AF:HW)((\\7DP)4M*(A^2`9;4):O*CEC_*N`^R$U'#!GS+B61JI,S7H`L\n> M%!@X`]=RS-4\\YJ@54=^A;.O0GDBX5:HMZS&Z9AI&EIQ_?RZ)+!2OU$A.T*^*\n> MN.:5%T*]/V49KRC(`F\"D,;O:O\"4MLJYEX(E\\#+F!</$$00DC1,['`]NMLJ7Y\n> M]IGC3]WI,L6A#?8T3-IS7*R7+Q^:H23'D51\"S/OWK;G$E#U]-*&6O3'8CB*I\n> M&\"G9V,L]!\\DXHE2%;IZ$*\"7%/5[=K]GD-D^RZAW-,HU<)MHPM4:CG1HM5V?B\n> M@$&QM'3,'5W4T4]57>WP?!EL`XR4'5L-R(:M5`/JAF&T93F@;1U;#TB);JD'\n> M'/QV:\"]-K`?4>3V@;G7YQ:YZ0-OL:&TK5Q$0#5DA]%Z]N#C-5U@GU-OJ_?@4\n> M3.&:+H([D8=R:^*))LO5]]!X\"!^[3X/H>5'-]FUO^3M]4',;W5<UIILM76JL\n> MU;8>TM@ZL75-F5VX.%!3&[1^H1/4$+&(T>@:[:ZYYR5>\"_:<5KN9:0@;.GJJ\n> MH8_14OGCA0A`8HC:&+C](@LDBB$6YB#$B!=QOGM:9D,RB)TY`K_,0;'R'<LJ\n> MA4_9$Q6C[$CQD7`&GX>1%_WS]$E*B=TDMQ>N_X'C8B@+C^$XU[XB$@NB6-#)\n> MSWR3MJJ)#[#$&<14AYJA[+MB@Y9AJ%<^=<,\\T`0%(:SRH?VUB%[O6GK7/-;^\n> M)*$\\/(#QM;K&OC<]AJ[5C:8*>$CNBPQ?#BY'PU?#\\_YH>'4)R(^>_^PUM2=<\n> MXEK\"HIH70.)1<^Y8%`1Q+9P!>[M9T^Y(O2KPKZ;7:WE)F1/98;CDKW+)LQP/\n> MKX<WHZOK]PIV^C^/7E]=PU[VLC\\:P,^;_C^OKLG@'3`J(B,ISFJ]\\]N(LT1Q\n> M&IDX<&EV]597;S^&.-NT9;9@$X6O3JHO^%0,;>/':-4YK/<0T4FE7#F'?X1<\n> M!C%$A,$\"(OE@$;H>C1@8;O`!HD6?-V&IBU__]T=NQ']#4V95(H*K'\\6[3/(4\n> MEXJ!P=/:PF6P.BX+_\"IS%Q!WIF_JR,T'ZM$8B$$^&B8Q$6LGTE(^0!/\\GP?A\n> M4B1>Q?,2+&;;TOAWAW\\#\"+Z\":(3<!-/X$TQ'7@6)/[%Y(4\\C0]^IHM*S55%J\n> M_RZK\\AU78:NM=*RF9G8:=64K/X(!N]/\\Q0\\I^T:G>:A:1[\"_)[email protected])0/\n> M?__^;?_Z9C\"&[_X;8C,,`GQ[@5D(%MQ$&97$`;FEL`:,T8D0W/6A;;G$98$E\n> MY=33WFP>)-Z$Q_D04BQ#2NX\"=T+*4NNI?.K%]$/R_77EV::_9J/3TII-W53Z\n> M<VS8X:Q&L\\M%%'N,U0&[$V#T5!UPX1],PR^O1L/S`>3AK_OOAI?_P'1[R5]\"\n> M)V$(FR.=](@[\\R&UD1G44WF$`C[+Y9WM57E0!ULJW+#P5_TWO%==>M+-Q+&*\n> M>W&H)>58:N0(CM<2O:\\50*:W7R7\"-B5T+-W0X*N5!UPU0,3-&`^)]];XAH%,\n> M`4%XP0M`)`A=B`1=G_]!$._(K-QE9!)X'I@(\\)]'$:O3?M@-?\\!)4212X;)Y\n> MU._)-K;2QI6229`#QV\\LP==S?&\"0LSPB(%JN'?@S6LWTQ)_5.31/$*0PL,<@\n> M+;?7?T60MMP,TG1]3Y!F=0P-OE(KA#]9`!\"6WR5$^H@]YH5I:>&/A@?/GH$3\n> M/:NK>W$4Y`^PFL&;MZ/WY5I!?KX0[(GG3(@`A8QW:7_'\\O[HR/\"5PCRL6OSA\n> M6?.A*41^P.K),3,[.F8>:J(Y:EN2V6/M=)U:9JQZLVM9NXW5-`W--+,2JCC-\n> MHS)26%#U3@CO2^G9)5`3'L$4)Y;4*-5*Q&$^+,K!\\Y6LL\"\\J:.(Y(5DM29XD\n> M$D\"H^%)9]:/S5?@*QO;H(:*?(C>FZO<U%LH?/NRR<]B*@34,2X&@(?2XS\\!V\n> MT5PW,P\"R@W/6O32Q=H(06X?T=0\\R&IV&5M>-U-BXPHI1/*;\\`,<8LCFL(434\n> MPVBY\\B*:C*/$HZS$3X>*ESOY\\W'\\'0/!JJ\\:BJ<V+OA-\\?S-R_'-X&)P/M+(\n> MZA1XCR]/-1'?\\Q)J6G<!T(*@<H+'VV\"BR*5W_`P@BR'QE*5/56!V02><8$D<\n> M693\"*8O]2PE7^#/2;<6-5D,SVWI6!A,-F?CI6X(P\"B:)$X\\_)A2F8^J`#U\\A\n> M/P;@5OS\"`UV=WLRQ$<5<)/O6H_((,E^D:WZL9G3KR9,_:@DR6(CB'8>F34C5\n> MS'8GQW<;^.[DCDXJ?DY/>^I>K*]2'<PRIW8X#D+J1ZF&^;MFCR<=I>P(ZI/T\n> MC/,ZC4S]_`5FIGUY_%N=B=XJ@F58FF7D1+\",EF;5S141UJQN]_Q\"2!3(\\0)&\n> M-WDMY;2R*HVTCQSC<L*M?#?;IM9LI\\40>;:,$'S7=RV`9>7%'YHNIAT5G#S?\n> M4=86BVE'42PMY?]#0[[_E4^+J\\XNLZ`V)$%*_=NX;K4Z6JMMY>N]=6CH;!IZ\n> M>AA>V3?_'PVP8L*[>BM]\\4V0](S4)93-*/0X/?6315Z(H7\\7?(\"8]@6Y'OQR\n> :/1P-QL/+=U?_&HS?]/\\MO>/_/<H'1Y`S```X\n> `\n> end\n> \n> -----------------------Cut here----------------------------------------\n> \n> \n> -- \n> +------------------------------------------------------------------------+\n> + Simkovics Stefan +\n> + Student an der TU Wien (Informatik) +\n> + Tel.: 02239/3367 +\n> + email: [email protected] | [email protected] +\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": "Fri, 20 Mar 1998 17:14:29 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Added Having Clause"
},
{
"msg_contents": "Applied. \n\nThis adds HAVING to the system, and is our first contribution from\nStefan, who is doing his thesis on adding ANSI features to PostgreSQL. \n(Stefan, correct me if I am wrong on this.) He plans to finish by June.\n\n\n> \n> Hi!\n> \n> I started adding the Having Clause and it works quite fine for\n> sequential scans! (I think it will also work with hash, index, etc but\n> I did not check it out! I made some High level changes which should\n> work for all access methods, but maybe I'm wrong. Please let me know.)\n> \n> Now it is possible to make queries like:\n> \n> select s.sname, max(p.pid), min(p.pid)\n> from part p, supplier s\n> where s.sid=p.sid\n> group by s.sname\n> having max(pid)=6 and min(pid)=1 or avg(pid)=4;\n> \n> Having does not work yet for queries that contain a subselect statement \n> in the Having clause, I'll try to fix this in the next days.\n> \n> If there are some bugs, please let me know, I'll start to read the\n> mailinglists now!\n> \n> Now here is the patch against the original 6.3 version (no snapshot!!):\n> \n> It is encoded using uuencode and called: \n> postgresql-6.3.tar.gz_to_having.diff.gz.uu\n> \n> If your original 6.3 source tree is in:\n> tmp/postgresql-6.3/ \n> perform the following steps to apply the patch:\n> \n> $ cd tmp/\n> $ uudecode postgresql-6.3.tar.gz_to_having.diff.gz.uu\n> $ gunzip postgresql-6.3.tar.gz_to_having.diff.gz\n> $ patch < postgresql-6.3.tar.gz_to_having.diff\n> \n> Hopefully you will see only \"Hunk succeeded...\" messages :-)\n> \n> Regards\n> \n> Stefan\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": "Mon, 30 Mar 1998 11:36:02 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: Added Having Clause"
}
] |
[
{
"msg_contents": "\nftp.postgresql.org/pub/patches/.test/v6.3p1beta3.gz\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Fri, 20 Mar 1998 19:57:34 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "Newest Patch...try this one..."
},
{
"msg_contents": "At 7:57 PM 98.3.20 -0400, The Hermit Hacker wrote:\n>ftp.postgresql.org/pub/patches/.test/v6.3p1beta3.gz\n\nI have tested the patch on my FreeBSD 2.2.5 box. \nHere are quick reports.\n\no the patch command needs -p1 and -l option. It would be nice\n if this is mentioned in README or whatever. \no an unnecceary patch (I guess) for \n backend/optimizer/geqo/geqo_paths.c included. This cause some\n compile errors.\nTatsuo Ishii\[email protected]\n\n",
"msg_date": "Sat, 21 Mar 1998 23:10:20 +0900",
"msg_from": "[email protected] (Tatsuo Ishii)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Newest Patch...try this one..."
},
{
"msg_contents": "On Sat, 21 Mar 1998, Tatsuo Ishii wrote:\n\n> o an unnecceary patch (I guess) for \n> backend/optimizer/geqo/geqo_paths.c included. This cause some\n> compile errors.\n\n\tThis patch represents what has changed between the tar file that\nis v6.3 and the source tree that is current...if there is an unnecessary\npatch, then it should most likely be removed from current tree altogether,\nno?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Sat, 21 Mar 1998 14:42:02 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Newest Patch...try this one..."
},
{
"msg_contents": "At 2:42 PM 98.3.21 -0400, The Hermit Hacker wrote:\n>On Sat, 21 Mar 1998, Tatsuo Ishii wrote:\n>\n>> o an unnecceary patch (I guess) for \n>> backend/optimizer/geqo/geqo_paths.c included. This cause some\n>> compile errors.\n>\n> This patch represents what has changed between the tar file that\n>is v6.3 and the source tree that is current...if there is an unnecessary\n>patch, then it should most likely be removed from current tree altogether,\n>no?\n\nI don't know why this happens, but...\n\nIt seems that the patch is trying to create a new file of geqo_path.c\nwhich is identical to existing geqo_path.c in the 6.3 tar file.\nAs a result, dupliation of it is performed by the patch.\nTo confirm above, take a look at the RCS id of geqo_path.c\nin the tar file and the patch.\nYou will find they are same (1.8 created on 1998/2/26).\n--\nTatsuo Ishii\[email protected]\n\n",
"msg_date": "Sun, 22 Mar 1998 11:39:36 +0900",
"msg_from": "[email protected] (Tatsuo Ishii)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Newest Patch...try this one..."
},
{
"msg_contents": "> \n> At 7:57 PM 98.3.20 -0400, The Hermit Hacker wrote:\n> >ftp.postgresql.org/pub/patches/.test/v6.3p1beta3.gz\n> \n> I have tested the patch on my FreeBSD 2.2.5 box. \n> Here are quick reports.\n> \n> o the patch command needs -p1 and -l option. It would be nice\n> if this is mentioned in README or whatever. \n> o an unnecceary patch (I guess) for \n> backend/optimizer/geqo/geqo_paths.c included. This cause some\n> compile errors.\n> Tatsuo Ishii\n> [email protected]\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": "Sat, 21 Mar 1998 22:32:13 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Newest Patch...try this one..."
},
{
"msg_contents": "> o an unnecceary patch (I guess) for \n> backend/optimizer/geqo/geqo_paths.c included. This cause some\n> compile errors.\n> Tatsuo Ishii\n> [email protected]\n\nI am confused about this item. cvs shows no changes in this file since\nthe 6.3 release.\n\n\n---------------------------------------------------------------------------\n\nRCS file: /usr/local/cvsroot/pgsql/src/backend/optimizer/geqo/geqo_paths.c,v\nWorking file: geqo_paths.c\nhead: 1.8\nbranch:\nlocks: strict\naccess list:\nsymbolic names:\n release-6-3: 1.8\nkeyword substitution: kv\ntotal revisions: 8; selected revisions: 8\ndescription:\n----------------------------\nrevision 1.8\ndate: 1998/02/26 04:32:23; author: momjian; state: Exp; lines: +2 -2\npgindent run before 6.3 release, with Thomas' requested changes.\n----------------------------\nrevision 1.7\ndate: 1997/09/08 21:44:32; author: momjian; state: Exp; lines: +7 -7\nUsed modified version of indent that understands over 100 typedefs.\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 22:32:49 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Newest Patch...try this one..."
},
{
"msg_contents": ">I just put the final patch in place (unannounced) for anyone to try out...\n>\n>ftp://ftp.postgresql.org/pub/postgresql-6.3-6.3.1.gz\n>\n>Unless there are any outstandign problems with either that patch or the\n>full distribution, I will put out an announcement of those two Monday\n>morning...\n\nIt seems still the problem below exists in postgresql-6.3-6.3.1.gz.\n(I haven't checked postgresql-6.3.1.tar.gz yet)\n\n>>> o an unnecceary patch (I guess) for \n>>> backend/optimizer/geqo/geqo_paths.c included. This cause some\n>>> compile errors.\n>>\n>> This patch represents what has changed between the tar file that\n>>is v6.3 and the source tree that is current...if there is an unnecessary\n>>patch, then it should most likely be removed from current tree altogether,\n>>no?\n>\n>I don't know why this happens, but...\n>\n>It seems that the patch is trying to create a new file of geqo_path.c\n>which is identical to existing geqo_path.c in the 6.3 tar file.\n>As a result, dupliation of it is performed by the patch.\n>To confirm above, take a look at the RCS id of geqo_path.c\n>in the tar file and the patch.\n>You will find they are same (1.8 created on 1998/2/26).\n",
"msg_date": "Mon, 23 Mar 1998 10:01:37 +0900",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Newest Patch...try this one... "
},
{
"msg_contents": ">>I just put the final patch in place (unannounced) for anyone to try out...\n>>\n>>ftp://ftp.postgresql.org/pub/postgresql-6.3-6.3.1.gz\n>>\n>>Unless there are any outstandign problems with either that patch or the\n>>full distribution, I will put out an announcement of those two Monday\n>>morning...\n>\n>It seems still the problem below exists in postgresql-6.3-6.3.1.gz.\n>(I haven't checked postgresql-6.3.1.tar.gz yet)\n\npostgresql-6.3.1.tar.gz has passed the regression test on following\nplatforms:\n\nSparc/Solaris2.6\nFreeBSD 2.2.1R\nPPC/MkLinux\n--\nTatsuo Ishii\[email protected]\n",
"msg_date": "Mon, 23 Mar 1998 13:01:25 +0900",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: Final Patch?"
},
{
"msg_contents": "On Mon, 23 Mar 1998 [email protected] wrote:\n\n> >I just put the final patch in place (unannounced) for anyone to try out...\n> >\n> >ftp://ftp.postgresql.org/pub/postgresql-6.3-6.3.1.gz\n> >\n> >Unless there are any outstandign problems with either that patch or the\n> >full distribution, I will put out an announcement of those two Monday\n> >morning...\n> \n> It seems still the problem below exists in postgresql-6.3-6.3.1.gz.\n> (I haven't checked postgresql-6.3.1.tar.gz yet)\n\n\tI don't know how, but for some reason, the geqo_paths.c that was\nuntar'd from the 6.3 tar file came out zero length...rebuilding the patch\nnow...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Mon, 23 Mar 1998 00:59:55 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Newest Patch...try this one... "
},
{
"msg_contents": "\n\n> It seems that the patch is trying to create a new file of geqo_path.c\n> which is identical to existing geqo_path.c in the 6.3 tar file.\n> As a result, dupliation of it is performed by the patch.\n> To confirm above, take a look at the RCS id of geqo_path.c\n> in the tar file and the patch.\n> You will find they are same (1.8 created on 1998/2/26).\n\n\tFixed...I will be regenerating the patch as well as the tar file\ntomorrow morning, prior to announcing...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Mon, 23 Mar 1998 01:13:02 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Newest Patch...try this one..."
},
{
"msg_contents": "On Mon, 23 Mar 1998 [email protected] wrote:\n\n> Sparc/Solaris2.6\n> FreeBSD 2.2.1R\n> PPC/MkLinux\n\nadd RH/Linux 5.0 Intel\n\nCristian\n--\n----------------------------------------------------------------------\nCristian Gafton -- [email protected] -- Red Hat Software, Inc.\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n UNIX is user friendly. It's just selective about who its friends are.\n\n",
"msg_date": "Wed, 1 Apr 1998 21:11:20 -0500 (EST)",
"msg_from": "Cristian Gafton <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: Final Patch?"
}
] |
[
{
"msg_contents": "\nHave managed to delete char2,4,8 and 16. The char16 was tricky\nsince it was still one of the builtins. Removed fine everywhere\nbut the bootstrap code. Had to leave a dummy entry in the Procid\narray so all of the types after char16 would be indexed correctly.\nCould remove this dummy by moving types 21-30 down one, but that\nwould entail renumbering the catalog. With that in mind, I can\nlive with a filler entry in an array. Heaven knows there are bigger\nand badder hacks elsewhere in the code dying to be discovered. :)\n\nThree points now...regression tests, docs and interfaces.\n\n1. Regression tests make very extensive use of char16. Seems to\nme that these should be changed to a name. Same for tutorial and\nany other example references in the code. Seem ok?\n\n2. I'll take my best shot at the docs.\n\n3. Also take a look/see at the interfaces that reference the char16.\nPerhaps would be better for me to contact the maintainers of those\ninterfaces as they would know what to do better than I.\n\nOnce I do one and two, I'll submit the patch.\n\ndarrenk\n",
"msg_date": "Fri, 20 Mar 1998 19:45:46 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "char types gone."
},
{
"msg_contents": "> Have managed to delete char2,4,8 and 16.\n> Three points now...regression tests, docs and interfaces.\n> \n> 1. Regression tests make very extensive use of char16. Seems to\n> me that these should be changed to a name. Same for tutorial and\n> any other example references in the code. Seem ok?\n\nAny reason not to use \"text\" for these cases? Or even varchar()? \"name\"\nis probably really only an internal Postgres type. We should have a\nregression test for it if we don't already. Let me know if you want me\nto help...\n\n> 2. I'll take my best shot at the docs.\n\nWhere? The man pages or the sgml sources? I'm shooting for getting the\nsgml sources right, and obsoleting the man page _sources_, regenerating\nthem from the sgml instead.\n\n - Tom\n",
"msg_date": "Mon, 23 Mar 1998 15:57:48 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "Re-applied this patch, now that 6.3.* is closed.\n\n> \n> \n> Have managed to delete char2,4,8 and 16. The char16 was tricky\n> since it was still one of the builtins. Removed fine everywhere\n> but the bootstrap code. Had to leave a dummy entry in the Procid\n> array so all of the types after char16 would be indexed correctly.\n> Could remove this dummy by moving types 21-30 down one, but that\n> would entail renumbering the catalog. With that in mind, I can\n> live with a filler entry in an array. Heaven knows there are bigger\n> and badder hacks elsewhere in the code dying to be discovered. :)\n> \n> Three points now...regression tests, docs and interfaces.\n> \n> 1. Regression tests make very extensive use of char16. Seems to\n> me that these should be changed to a name. Same for tutorial and\n> any other example references in the code. Seem ok?\n> \n> 2. I'll take my best shot at the docs.\n> \n> 3. Also take a look/see at the interfaces that reference the char16.\n> Perhaps would be better for me to contact the maintainers of those\n> interfaces as they would know what to do better than I.\n> \n> Once I do one and two, I'll submit the patch.\n> \n> darrenk\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, 26 Apr 1998 00:03:42 -0400 (EDT)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "Bruce Momjian writes:\n> \n> OK, tonight's mega-patch should be pretty close to the final patch. We\n> had to have the Alpha fix, and I put in the other fixes that looked easy\n> to understand. Most were port-specific cleanup people had posted.\n> \n> If it breaks, let Marc (Oh, no, I mean the hackers list) know.\n\nI have decided based on some thought and some discussion with Bruce to hold\noff trying to get the S_LOCK backoff change into the mega-patch. While it\nis a nice idea, and I even have it all coded, I would like to do some extra\ntesting and paw around some other code a bit as there is a possibility of it\nsurfacing some problems with respect to the way signals are handled.\n\nIt could I suppose be put out for people to experiment with...\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": "Fri, 20 Mar 1998 17:00:14 -0800",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": true,
"msg_subject": "None"
}
] |
[
{
"msg_contents": "> \n> On Wed, 18 Mar 1998, David Gould wrote:\n> \n> > I agree with Marcs comments here, but would like to add that ElectricFence i\ns\n> > freely available at least on Linux systems. It is a long way from Purify,\n> > but it beats sitting around scratching your head.\n> \n> URL for the software? I'd like to see if they at least have a\n> FreeBSD port that I can use :)\n\nI just got it off the CD myself, but I ftp.redhat.com has it in the\nredhat 5.0 distribution area. Look in the SRPMS for:\n\nElectricFence-2.0.5-5.src.rpm \n\nOr, if you don't have rpm, AltaVista turned up some bsd entries too.\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": "Fri, 20 Mar 1998 17:02:05 -0800",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Buffer overuns with the Electric fence debugging\n library"
}
] |
[
{
"msg_contents": "Here are the changes. I have updated the HISTORY and TODO files in the\ndistribution. I assume we don't want to update all the stuff like the\nbug template to reflect 6.3p1, do we?\n\n---------------------------------------------------------------------------\n\nCHANGES IN THE 6.3p1 PATCH \n--------------------------\necpg cleanup/fixes, now version 1.1(Michael Meskes)\npg_user cleanup(Bruce)\nlarge object fix for pg_dump and tclsh([email protected])\nLIKE fix for multiple adjacent underscores\nfix for redefining builtin functions(Thomas)\nultrix4 cleanup\nupgrade to pg_access 0.83\nupdated CLUSTER manual page\nmulti-byte character set support, see doc/README.mb(Tatsuo)\nconfigure --with-pgport fix\npg_ident fix\nbig-endian fix for backend communications(Tatsuo)\nSUBSTR() and substring() fix(Thomas)\nseveral jdbc fixes(Peter)\nlibpgtcl improvements, see libptcl/README(Randy Kunkee)\nFix for \"Datasize = 0\" error(Vadim)\nPrevent \\do from wrapping(Bruce)\nRemove duplicate Russian character set entries\nSunos4 cleanup\nAllow optional TABLE keyword in LOCK and SELECT INTO(Thomas)\nCREATE SEQUENCE options to allow a negative integer(Thomas)\nAdd \"PASSWORD\" as an allowed column identifier(Thomas)\nAdd checks for UNION target fields(Bruce)\nFix Alpha port(Dwayne Bailey)\nFix for text arrays containing quotes(Doug Gibson)\nSolaris compile fix(Albert Chin-A-Young)\nBetter identify tcl and tk libs and includes(Bruce)\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 00:34:59 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "6.3p1 CHANGES"
},
{
"msg_contents": "At 0:34 AM 98.3.21 -0500, Bruce Momjian wrote:\n>Here are the changes. I have updated the HISTORY and TODO files in the\n>distribution. I assume we don't want to update all the stuff like the\n>bug template to reflect 6.3p1, do we?\n>\n>---------------------------------------------------------------------------\n>\n>CHANGES IN THE 6.3p1 PATCH \n>--------------------------\n>ecpg cleanup/fixes, now version 1.1(Michael Meskes)\n>pg_user cleanup(Bruce)\n>large object fix for pg_dump and tclsh([email protected])\n>LIKE fix for multiple adjacent underscores\n>fix for redefining builtin functions(Thomas)\n>ultrix4 cleanup\n>upgrade to pg_access 0.83\n>updated CLUSTER manual page\n>multi-byte character set support, see doc/README.mb(Tatsuo)\n>configure --with-pgport fix\n>pg_ident fix\n>big-endian fix for backend communications(Tatsuo)\n\nThis was done by Kataoka not by me. Could you change above?\n\nTatsuo Ishii\[email protected]\n\n",
"msg_date": "Sat, 21 Mar 1998 23:10:24 +0900",
"msg_from": "[email protected] (Tatsuo Ishii)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] 6.3p1 CHANGES"
},
{
"msg_contents": "The SUBSTRING() fix was due to Jan W.\nThe LIKE fix also fixed having a function call as the target of\nBETWEEN/AND...\n\n - Tom\n---------------------------------------------------------------------------\n> \n> CHANGES IN THE 6.3p1 PATCH\n> --------------------------\n> ecpg cleanup/fixes, now version 1.1(Michael Meskes)\n> pg_user cleanup(Bruce)\n> large object fix for pg_dump and tclsh([email protected])\n> LIKE fix for multiple adjacent underscores\n> fix for redefining builtin functions(Thomas)\n> ultrix4 cleanup\n> upgrade to pg_access 0.83\n> updated CLUSTER manual page\n> multi-byte character set support, see doc/README.mb(Tatsuo)\n> configure --with-pgport fix\n> pg_ident fix\n> big-endian fix for backend communications(Tatsuo)\n> SUBSTR() and substring() fix(Thomas)\n> several jdbc fixes(Peter)\n> libpgtcl improvements, see libptcl/README(Randy Kunkee)\n> Fix for \"Datasize = 0\" error(Vadim)\n> Prevent \\do from wrapping(Bruce)\n> Remove duplicate Russian character set entries\n> Sunos4 cleanup\n> Allow optional TABLE keyword in LOCK and SELECT INTO(Thomas)\n> CREATE SEQUENCE options to allow a negative integer(Thomas)\n> Add \"PASSWORD\" as an allowed column identifier(Thomas)\n> Add checks for UNION target fields(Bruce)\n> Fix Alpha port(Dwayne Bailey)\n> Fix for text arrays containing quotes(Doug Gibson)\n> Solaris compile fix(Albert Chin-A-Young)\n> Better identify tcl and tk libs and includes(Bruce)\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",
"msg_date": "Sat, 21 Mar 1998 14:17:43 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] 6.3p1 CHANGES"
},
{
"msg_contents": "> >pg_ident fix\n> >big-endian fix for backend communications(Tatsuo)\n> \n> This was done by Kataoka not by me. Could you change above?\n\nDone.\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 10:16:06 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] 6.3p1 CHANGES"
},
{
"msg_contents": "> \n> The SUBSTRING() fix was due to Jan W.\n> The LIKE fix also fixed having a function call as the target of\n> BETWEEN/AND...\n> \n\ndone.\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 22:03:14 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] 6.3p1 CHANGES"
}
] |
[
{
"msg_contents": "Hi,\n\nIt turns of that one of the bugs I detected with Electric Fence\nis caused by an assumption that memory return by palloc and malloc \nis aligned on 8 bytes boundaries (double alignment). \nAdjusting this to correspond with the reality of the malloc\nimplementation on my pentium based linux box fixes the problem.\n\nThe following simple patch to include/utils/memutils.h\nwill fix the problem.\n\n85c85\n< #if ! defined(sco)\n---\n> #if ! defined(sco) && !defined(linux)\n\nThere is still another buffer overrun which I will try to find. \nThe remaining one is triggered by creating a sequence.\n\nI hope this patch will be applied if it makes sense to \nyou guys.\n\nWith regards from Maurice.\n\n\n",
"msg_date": "Sat, 21 Mar 1998 19:44:12 +0100",
"msg_from": "Maurice Gittens <[email protected]>",
"msg_from_op": true,
"msg_subject": "patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "\nSome things just require confirmation...can someone else (Thomas?) running\nLinux comment on this before I apply it? I just find it kinda unexpected\nto see something that doesn't work for SCO doesn't work for Linux too :)\n\nOn Sat, 21 Mar 1998, Maurice Gittens wrote:\n\n> Hi,\n> \n> It turns of that one of the bugs I detected with Electric Fence\n> is caused by an assumption that memory return by palloc and malloc \n> is aligned on 8 bytes boundaries (double alignment). \n> Adjusting this to correspond with the reality of the malloc\n> implementation on my pentium based linux box fixes the problem.\n> \n> The following simple patch to include/utils/memutils.h\n> will fix the problem.\n> \n> 85c85\n> < #if ! defined(sco)\n> ---\n> > #if ! defined(sco) && !defined(linux)\n> \n> There is still another buffer overrun which I will try to find. \n> The remaining one is triggered by creating a sequence.\n> \n> I hope this patch will be applied if it makes sense to \n> you guys.\n> \n> With regards from Maurice.\n> \n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Sat, 21 Mar 1998 19:55:39 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "Maurice Gittens wrote:\n> \n> Hi,\n> \n> It turns of that one of the bugs I detected with Electric Fence\n> is caused by an assumption that memory return by palloc and malloc \n> is aligned on 8 bytes boundaries (double alignment). \n> Adjusting this to correspond with the reality of the malloc\n> implementation on my pentium based linux box fixes the problem.\n> \n> The following simple patch to include/utils/memutils.h\n> will fix the problem.\n> \n> 85c85\n> < #if ! defined(sco)\n> ---\n> > #if ! defined(sco) && !defined(linux)\n\nI may be off on this one, but I thought that memory alignment was a\ncpu, and not necessarily an OS issue. I.E. Solaris x86 might show the\n\"misalignment\", while Linux Alpha would not.\n\nOcie Mitchell\n",
"msg_date": "Sat, 21 Mar 1998 16:37:34 -0800 (PST)",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "On Sat, 21 Mar 1998 [email protected] wrote:\n\n> Maurice Gittens wrote:\n> > \n> > Hi,\n> > \n> > It turns of that one of the bugs I detected with Electric Fence\n> > is caused by an assumption that memory return by palloc and malloc \n> > is aligned on 8 bytes boundaries (double alignment). \n> > Adjusting this to correspond with the reality of the malloc\n> > implementation on my pentium based linux box fixes the problem.\n> > \n> > The following simple patch to include/utils/memutils.h\n> > will fix the problem.\n> > \n> > 85c85\n> > < #if ! defined(sco)\n> > ---\n> > > #if ! defined(sco) && !defined(linux)\n> \n> I may be off on this one, but I thought that memory alignment was a\n> cpu, and not necessarily an OS issue. I.E. Solaris x86 might show the\n> \"misalignment\", while Linux Alpha would not.\n\n\tUmmm...you are comparing both two different CPU and two different\nOSs here...more appropriately to your argument would be Linux/x86 vs\nLinux/Alpha, wouldn't it? If it was a CPU issue...?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Sat, 21 Mar 1998 21:26:18 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> \n> \n> Some things just require confirmation...can someone else (Thomas?) running\n> Linux comment on this before I apply it? I just find it kinda unexpected\n> to see something that doesn't work for SCO doesn't work for Linux too :)\n\nOn BSDI, malloc man pages says:\n\n The allocated space is suitably aligned (after possible pointer coercion)\n for storage of any type of object.\n\nI don't believe this is a standard, and I think Linux may not follow it.\nI think I heard somewhere they don't. We certainly need to have this\nchecked before use in a patch.\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 22:58:03 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "On Sat, 21 Mar 1998, Bruce Momjian wrote:\n\n> > \n> > \n> > Some things just require confirmation...can someone else (Thomas?) running\n> > Linux comment on this before I apply it? I just find it kinda unexpected\n> > to see something that doesn't work for SCO doesn't work for Linux too :)\n> \n> On BSDI, malloc man pages says:\n> \n> The allocated space is suitably aligned (after possible pointer coercion)\n> for storage of any type of object.\n> \n> I don't believe this is a standard, and I think Linux may not follow it.\n> I think I heard somewhere they don't. We certainly need to have this\n> checked before use in a patch.\n> \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>From the linux man page:\n For calloc() and malloc(), the value returned is a pointer\n to the allocated memory, which is suitably aligned for any\n kind of variable, or NULL if the request fails.\n\n realloc() returns a pointer to the newly allocated memory,\n which is suitably aligned for any kind of variable and may\n be different from ptr, or NULL if the request fails or if\n size was equal to 0. If realloc() fails the original\n block is left untouched - it is not freed or moved.\n\n\n\n\nMarc Zuckman\[email protected]\n\n_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\n_ Visit The Home and Condo MarketPlace\t\t _\n_ http://www.ClassyAd.com\t\t\t _\n_\t\t\t\t\t\t\t _\n_ FREE basic property listings/advertisements and searches. _\n_\t\t\t\t\t\t\t _\n_ Try our premium, yet inexpensive services for a real\t _\n_ selling or buying edge!\t\t\t\t _\n_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\\_\n\n",
"msg_date": "Sat, 21 Mar 1998 23:57:51 -0500 (EST)",
"msg_from": "Marc Howard Zuckman <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> >From the linux man page:\n> For calloc() and malloc(), the value returned is a pointer\n> to the allocated memory, which is suitably aligned for any\n> kind of variable, or NULL if the request fails.\n\nOK, then why is electric fence telling him to make DOUBLEALIGN match\nINTALIGN? You would think that if this change made sense, we would be\nhaving some Linux problems, but we don't.\n\n---------------------------------------------------------------------------\n\n\n#if ! defined(sco)\n#define DOUBLEALIGN(LEN)\\\n (((long)(LEN) + (sizeof (double) - 1)) & ~(sizeof (double) -1))\n\n#define MAXALIGN(LEN)\\\n (((long)(LEN) + (sizeof (double) - 1)) & ~(sizeof (double) -1))\n#else\n#define DOUBLEALIGN(LEN) INTALIGN(LEN)\n#define MAXALIGN(LEN) INTALIGN(LEN)\n#endif\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, 22 Mar 1998 00:21:22 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> > Some things just require confirmation...can someone else (Thomas?) running\n> > Linux comment on this before I apply it? I just find it kinda unexpected\n> > to see something that doesn't work for SCO doesn't work for Linux too :)\n> \n> On BSDI, malloc man pages says:\n> \n> The allocated space is suitably aligned (after possible pointer coercion)\n> for storage of any type of object.\n> \n> I don't believe this is a standard, and I think Linux may not follow it.\n> I think I heard somewhere they don't. We certainly need to have this\n> checked before use in a patch.\n\nThe alignment of a pointer returned from malloc() is platform specific. The\nBSDI man page above gets the sense of it right, and as far as I am aware\nall mallocs conform to this restriction. In particular x86 Linux malloc\nbehaves exactly that way.\n\nOf course on x86 there is no special requirement any type to be aligned at all.\nDouble do not need to be 8 byte aligned. Ints do not need to be on an even\naddress etc. So malloc can return a pointer to any address it wants on an x86\nand still conform to the restriction.\n\nOn a Sparc or an Alpha, objects must to be aligned to the sizeof the type. \nSo shorts are 2 byte aligned, ints 4 byte, and doubles 8 byte. Since malloc\nhas no idea what you are allocating it has to return an 8 byte aligned pointer.\n\nBut all this is a bit of a red herring. The original bug if I recall was\ncaused by the caller assuming that the return from palloc was 8 byte aligned.\nSince the caller \"knew\" that the pointer was 8 byte aligned it then went on\nto do some \"fancy\" pointer arithmetic to arrive at an offset. This is what\nis broken, not the Linux malloc().\n\nThat said, I have a plan in mind to get a nice performance boost for palloc().\nAs a side effect, I will guarantee 8 byte alignment of pointers returned\nfrom palloc().\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": "Sun, 22 Mar 1998 00:39:42 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> That said, I have a plan in mind to get a nice performance boost for palloc().\n> As a side effect, I will guarantee 8 byte alignment of pointers returned\n> from palloc().\n\nOooh, sounds 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": "Sun, 22 Mar 1998 11:55:58 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> \n> > >From the linux man page:\n> > For calloc() and malloc(), the value returned is a pointer\n> > to the allocated memory, which is suitably aligned for any\n> > kind of variable, or NULL if the request fails.\n> \n> OK, then why is electric fence telling him to make DOUBLEALIGN match\n> INTALIGN? You would think that if this change made sense, we would be\n> having some Linux problems, but we don't.\n> \n> ---------------------------------------------------------------------------\n> \n> \n> #if ! defined(sco)\n> #define DOUBLEALIGN(LEN)\\\n> (((long)(LEN) + (sizeof (double) - 1)) & ~(sizeof (double) -1))\n> \n> #define MAXALIGN(LEN)\\\n> (((long)(LEN) + (sizeof (double) - 1)) & ~(sizeof (double) -1))\n> #else\n> #define DOUBLEALIGN(LEN) INTALIGN(LEN)\n> #define MAXALIGN(LEN) INTALIGN(LEN)\n> #endif\n> \n\nHi Hackers,\n\nit seems that Linux malloc, or better libc-5.4.23 malloc, is doing what is\ndeclared in the man page, i.e. returning \"memory which is suitably aligned\nfor any kind of variable\". I have some malloc traces and all malloc results\nare aligned to double size (8 bytes): \n\nop\tcaller\t\tptr\t\tsize\n--------------------------------------------\nmalloc\t400c32b2\t080a06c8\t25\nmalloc\t400c32b2\t080a06e8\t13\nmalloc\t400c32b2\t080a0700\t26\nmalloc\t400c32b2\t080a0720\t11\nmalloc\t400c32b2\t080a0730\t50\nmalloc\t400c32b2\t080a0768\t14\nmalloc\t400c32b2\t080a0780\t17\nmalloc\t400c32b2\t080a0798\t24\nmalloc\t400c32b2\t080a07b8\t27\nmalloc\t400c32b2\t080a07d8\t817\nmalloc\t400c32b2\t080a0b10\t11\nmalloc\t400c32b2\t080a0b20\t14\nmalloc\t400c32b2\t080a0b38\t164\nmalloc\t400c32b2\t080a0be0\t11\nmalloc\t400c32b2\t080a0bf0\t14\nmalloc\t400c32b2\t080a0c08\t25\nmalloc\t400c32b2\t080a0c28\t16\nmalloc\t400c32b2\t080a0c40\t35\nmalloc\t400c32b2\t080a0c68\t9\nmalloc\t400c32b2\t080a0c78\t7\n\nSo I don't see where is the problem.\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, 26 Mar 1998 16:16:55 +0100 (MET)",
"msg_from": "Massimo Dal Zotto <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
}
] |
[
{
"msg_contents": "Brett McCormick <[email protected]> wrote:\n \n> expr ? trueval : falseval\n> \n> if expr evaluated is true, the entire expression is trueval evaluated,\n> otherwise it is falseval evaluated.\n> \n> Is it possible to have this sort of grammar logic in the parser?\n\nFor achieving this kind of functionality, you can also define a \nfunction like Oracle does. Actually the Oracle way is more general.\n\nThis function (i can't remember its name, but lets assume it is \ncalled CHOICE) behaves as follows:\n\nit has as its first argument a testvalue followed by pairs of arguments, \nmatch and value, and possibly a default value as last argument.\n\nthe function is evaluated so that a value whose match equals \ntestvalue is returned. \n\nso for your case you would call:\n\nCHOICE(expr,'T',trueval,'F',falseval)\n\nbut it is much more general, for example for getting a nice table of \namounts of something bought quarterly, you do the following.\n\nselect\n name as \"Name\",\n sum(choice(quarter(buy_time),'1',amount,0)) as \"Q1\",\n sum(choice(quarter(buy_time),'2',amount,0)) as \"Q2\",\n sum(choice(quarter(buy_time),'3',amount,0)) as \"Q3\",\n sum(choice(quarter(buy_time),'4',amount,0)) as \"Q4\",\n sum(amount) as \"Year total\"\nfrom \n buyings\ngroup by name;\n\nand get the following\n\nName | Q1 | Q2 | Q3 | Q4 | Total\n-------+-----+-----+-----+-----+--------\ncats | 0 | 0 | 3 | 7 | 10\ndogs | 1 | 2 | 3 | 4 | 10\nducks | 1 | 1 | 1 | 1 | 4\n\n\nWhat holds us back from defining a function like this now is \ndifficulty of defining functions that can accept arguments and \ncan return values of an opaque type (or actually type defined \nat invocation time)\n\nOr actually there is (or at least was a little while back) a function \nCOUNT that can take any type of argument and return an integer, so \n_this_ should be possible to define.\n\nBut I have not found any reference how to tell a function that it has \nvariable number of arguments and that it can itself tell what it returns \nbased on what arguments are given. I think that this is quite hard given \nthe current implementation.\n\nHannu\n",
"msg_date": "Sat, 21 Mar 1998 21:58:56 +0200",
"msg_from": "Hannu Krosing <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: expr ? trueval : falseval"
},
{
"msg_contents": "Hannu Krosing wrote:\n> This function (i can't remember its name, but lets assume it is\n> called CHOICE) behaves as follows:\n\nThe function name is DECODE and after using it a few times, I find that\nit is a timesaver.\n\nBTW, is anybody in charge of the oracle-compat portion. If not, I might\nbe interested?\n\nAlso, is anybody addressing the ALTER USER .. IN GROUP ... and CREATE\nUSER .. IN GROUP ... (basically anything pertaining to groups)? I've\nstarted hacking away at it to make it work.\n\nThanks,\n\n-- \nStephane Lajeunesse.\nOracle and Sybase DBA\n",
"msg_date": "Sun, 22 Mar 1998 19:25:11 -0500",
"msg_from": "Stephane Lajeunesse <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: expr ? trueval : falseval"
},
{
"msg_contents": "> \n> Hannu Krosing wrote:\n> > This function (i can't remember its name, but lets assume it is\n> > called CHOICE) behaves as follows:\n> \n> The function name is DECODE and after using it a few times, I find that\n> it is a timesaver.\n> \n> BTW, is anybody in charge of the oracle-compat portion. If not, I might\n> be interested?\n\nEdmund Mergl <[email protected]>. He is still involved, by hasn't made\nany additions to those functions since maybe September.\n\n> \n> Also, is anybody addressing the ALTER USER .. IN GROUP ... and CREATE\n> USER .. IN GROUP ... (basically anything pertaining to groups)? I've\n> started hacking away at it to make it work.\n\nGood. We need help there. See the FAQ on groups, if you haven't\nalready.\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, 22 Mar 1998 21:01:26 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Re: expr ? trueval : falseval"
}
] |
[
{
"msg_contents": "> \n> \tI have configured 6.3 as follows:\n> \n> \t--prefix=/opt/postgresql-6.3 --with-tcl --with-x\n> \n> \tWhen prompted, I supplide the appropriate directories for the tk/tcl\n> \tinclude/library directories. Still when I look in\n> \t~src/interfaces/libpgtcl there is no library creted. When I cd to that\n> \tdirectory and do a gmake, I get a bunch of unsatsfide symbols. Platofmr\n> \tis HP-UX 10.20 PA-RISC.\n> \n> \tCan someone give me a clue here? I wan't to try the new pgaccess on my\n> \t6.3 database.\n\nWait for the 6.3 patch, due in a few days. Configure should then find\nyour libraries, and compile. I fixed this yesterday.\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, 22 Mar 1998 00:16:51 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [QUESTIONS] How do I build libpgtcl?"
}
] |
[
{
"msg_contents": ">\n>Of course on x86 there is no special requirement any type to be aligned at\nall.\n>Double do not need to be 8 byte aligned. Ints do not need to be on an even\n>address etc. So malloc can return a pointer to any address it wants on an\nx86\n>and still conform to the restriction.\nYes and I think it returns 4 byte aligned objects for performance reasons.\n(This is more efficient than storing double/int on 1/2 byte boundaries.)\n\n>\n>On a Sparc or an Alpha, objects must to be aligned to the sizeof the type.\n>So shorts are 2 byte aligned, ints 4 byte, and doubles 8 byte. Since malloc\n>has no idea what you are allocating it has to return an 8 byte aligned\npointer.\n>\n>But all this is a bit of a red herring. The original bug if I recall was\n>caused by the caller assuming that the return from palloc was 8 byte\naligned.\n>Since the caller \"knew\" that the pointer was 8 byte aligned it then went on\n>to do some \"fancy\" pointer arithmetic to arrive at an offset. This is what\n>is broken, not the Linux malloc().\nYes, I think so too.\n>\n>That said, I have a plan in mind to get a nice performance boost for\npalloc().\n>As a side effect, I will guarantee 8 byte alignment of pointers returned\n>from palloc().\n>\nDoes this imply that more memory will be used that strictly necesary under\nlinux?\n\nWith regards from Maurice.\n\n\n\n",
"msg_date": "Sun, 22 Mar 1998 10:42:30 +0100",
"msg_from": "\"Maurice Gittens\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> >That said, I have a plan in mind to get a nice performance boost for\n> palloc().\n> >As a side effect, I will guarantee 8 byte alignment of pointers returned\n> >from palloc().\n> >\n> Does this imply that more memory will be used that strictly necesary under\n> linux?\n> \n> With regards from Maurice.\n\nActually, perhaps less memory, because malloc memory is not aligned on\ndouble boundaries, but this is all still just conjecture.\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, 22 Mar 1998 15:00:56 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> \n> > >That said, I have a plan in mind to get a nice performance boost for\n> > palloc().\n> > >As a side effect, I will guarantee 8 byte alignment of pointers returned\n> > >from palloc().\n> > >\n> > Does this imply that more memory will be used that strictly necesary under\n> > linux?\n> > \n> > With regards from Maurice.\n> \n> Actually, perhaps less memory, because malloc memory is not aligned on\n> double boundaries, but this is all still just conjecture.\n\n\nOk, you've convinced me. Sheesh...\n\nChange the description to \"nice performance boost and use less memory\".\n\n;-)\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": "Sun, 22 Mar 1998 14:28:07 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
}
] |
[
{
"msg_contents": ">\n>OK, then why is electric fence telling him to make DOUBLEALIGN match\n>INTALIGN? You would think that if this change made sense, we would be\n>having some Linux problems, but we don't.\n\nDon't you think it's dangerous to presume that a problem doesn't exist \njust because you've never noticed it?\nUmm, -:)\n\n\nWith regards from Maurice.\n\n\n\n",
"msg_date": "Sun, 22 Mar 1998 10:45:32 +0100",
"msg_from": "\"Maurice Gittens\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> \n> >\n> >OK, then why is electric fence telling him to make DOUBLEALIGN match\n> >INTALIGN? You would think that if this change made sense, we would be\n> >having some Linux problems, but we don't.\n> \n> Don't you think it's dangerous to presume that a problem doesn't exist \n> just because you've never noticed it?\n> Umm, -:)\n\nYes. I am not staying it doesn't exist, but that we need to research it\nfurther before applying a change that will affect so much 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, 22 Mar 1998 15:01:43 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for memory overrun on Linux(i386)"
}
] |
[
{
"msg_contents": "Hi !\n\nPlease help, the FreeBSD port update to version 6.3 drives me to\ndespair ;-)\n\n--with-tcl configure switch doesn't work to build libpgtcl\n\nI only get it to work if I set in the make environment\n\n\tUSE_TCL=true\n\nconfigure --with-tcl doesn't seem to set USE_TCL=true in\nsrc/Makefile.global, so\n\tsrc/interfaces/Makefile,\nwhich includes\n\tsrc/Makefile.global\nfails, since USE_TCL isn't set to true. Workaround was to explicitely\nset USE_TCL=true when starting the make.\n\nAnother problem: on FreeBSD-current (and I think this is true on\nFreeBSD-2.2.x as well)\n\nThe make install fails when running the initdb script.\n\nAdding template1 database to pg_database...\nRunning: postgres -boot -C -F -D/usr/local/pgsql/data -Q template1 < /tmp/create\n.13645\n \nvacuuming template1\ncreating public pg_user view\nloading pg_description\n*** Error code 1\n\nThe Postgresql 6.2 version needed a patch so that this works\n--- bin/initdb/initdb.sh.orig Wed Apr 2 07:09:13 1997\n+++ bin/initdb/initdb.sh Wed Apr 2 07:09:33 1997\n@@ -344,6 +344,5 @@\n \n if [ $debug -eq 0 ]; then\n echo \"vacuuming template1\"\n- echo \"vacuum\" | postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n- grep -v \"^DEBUG:\"\n+ echo \"vacuum\" | postgres -o /dev/null -F -Q -D$PGDATA template1 > /dev/null\n \n fi\n\nI enhanced this patch and now use \n\t\"-o /dev/null\"\nand avoid\n\t\"2>&1\"\n\nseveral lines below where similar commands can be found, but this\ndoesn't solve the problem.\n\n--- bin/initdb/initdb.sh.orig Sun Mar 22 12:47:51 1998\n+++ bin/initdb/initdb.sh Sun Mar 22 13:06:36 1998\n@@ -348,8 +348,7 @@\n \n # If the COPY is first, the VACUUM generates an error, so we vacuum first\n echo \"vacuuming template1\"\n-echo \"vacuum\" | postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n- grep -v \"^DEBUG:\"\n+echo \"vacuum\" | postgres -o /dev/null -F -Q -D$PGDATA template1 > /dev/null\n \n echo \"COPY pg_shadow TO '$PGDATA/pg_pwd' USING DELIMITERS '\\\\t'\" |\\\n postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n@@ -383,6 +382,6 @@\n echo \"loading pg_description\"\n echo \"copy pg_description from '$TEMPLATE_DESCR'\" | postgres -F -Q -D$PGDATA te\nmplate1 > /dev/null\n echo \"copy pg_description from '$GLOBAL_DESCR'\" | postgres -F -Q -D$PGDATA temp\nlate1 > /dev/null\n-echo \"vacuum analyze\" | postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n+echo \"vacuum analyze\" | postgres -o /dev/null -F -Q -D$PGDATA template1 > /dev/\nnull |\\\n grep -v \"^DEBUG:\"\n \nAny help appreciated !\n\nReply-To set, since I haven't subscribed to your list.\n\n-- \nAndreas Klemm http://www.FreeBSD.ORG/~andreas\npowered by ,,symmetric multiprocessor FreeBSD''\n",
"msg_date": "Sun, 22 Mar 1998 13:56:23 +0100",
"msg_from": "Andreas Klemm <[email protected]>",
"msg_from_op": true,
"msg_subject": "urgent: pgsql 6.3 FreeBSD port update trouble,\n --with-tcl doesn't work"
},
{
"msg_contents": "On Sun, 22 Mar 1998, Andreas Klemm wrote:\n\n> Hi !\n> \n> Please help, the FreeBSD port update to version 6.3 drives me to\n> despair ;-)\n> \n> --with-tcl configure switch doesn't work to build libpgtcl\n> \n> I only get it to work if I set in the make environment\n> \n> \tUSE_TCL=true\n\nThis has been fixed in the upcoming patch...I'm going to be bundling up a\npostgresql-v6.3.1.tar.gz and postgresql-v6.3-v6.3.1.gz file this\nafternoon...\n\n> Another problem: on FreeBSD-current (and I think this is true on\n> FreeBSD-2.2.x as well)\n\n\tOdd, since my testing has always been done on FreeBSD-current, and\nI've already done a couple of builds under FreeBSD 2.2-STABLE :(\n\n\tBut, looking at your patches for initdb.sh, I've applied them, as\nI think that it provides a non-shell dependent solution, which I think is\ncleaner...\n\n\tI'm just about to do a build on my system, to make sure\neverythingis okay before I bundle...but the new version should be up (and\nannounced) later this evening...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n\n",
"msg_date": "Sun, 22 Mar 1998 14:26:06 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
},
{
"msg_contents": "\n\nI just updated the ftp site such that there is a postgresql-6.3.1.tar.gz\nfile, which corrects alot of post-release bugs that cropped up, including\nthose you listed below...\n\nLet me know if this help...it hasn't been officially released yet, but,\nunless anyone comes up with an critical problems with it, that tar file\nwill be what is announced tomorrow morning...\n\n\n\nOn Sun, 22 Mar 1998, Andreas Klemm wrote:\n\n> Hi !\n> \n> Please help, the FreeBSD port update to version 6.3 drives me to\n> despair ;-)\n> \n> --with-tcl configure switch doesn't work to build libpgtcl\n> \n> I only get it to work if I set in the make environment\n> \n> \tUSE_TCL=true\n> \n> configure --with-tcl doesn't seem to set USE_TCL=true in\n> src/Makefile.global, so\n> \tsrc/interfaces/Makefile,\n> which includes\n> \tsrc/Makefile.global\n> fails, since USE_TCL isn't set to true. Workaround was to explicitely\n> set USE_TCL=true when starting the make.\n> \n> Another problem: on FreeBSD-current (and I think this is true on\n> FreeBSD-2.2.x as well)\n> \n> The make install fails when running the initdb script.\n> \n> Adding template1 database to pg_database...\n> Running: postgres -boot -C -F -D/usr/local/pgsql/data -Q template1 < /tmp/create\n> .13645\n> \n> vacuuming template1\n> creating public pg_user view\n> loading pg_description\n> *** Error code 1\n> \n> The Postgresql 6.2 version needed a patch so that this works\n> --- bin/initdb/initdb.sh.orig Wed Apr 2 07:09:13 1997\n> +++ bin/initdb/initdb.sh Wed Apr 2 07:09:33 1997\n> @@ -344,6 +344,5 @@\n> \n> if [ $debug -eq 0 ]; then\n> echo \"vacuuming template1\"\n> - echo \"vacuum\" | postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n> - grep -v \"^DEBUG:\"\n> + echo \"vacuum\" | postgres -o /dev/null -F -Q -D$PGDATA template1 > /dev/null\n> \n> fi\n> \n> I enhanced this patch and now use \n> \t\"-o /dev/null\"\n> and avoid\n> \t\"2>&1\"\n> \n> several lines below where similar commands can be found, but this\n> doesn't solve the problem.\n> \n> --- bin/initdb/initdb.sh.orig Sun Mar 22 12:47:51 1998\n> +++ bin/initdb/initdb.sh Sun Mar 22 13:06:36 1998\n> @@ -348,8 +348,7 @@\n> \n> # If the COPY is first, the VACUUM generates an error, so we vacuum first\n> echo \"vacuuming template1\"\n> -echo \"vacuum\" | postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n> - grep -v \"^DEBUG:\"\n> +echo \"vacuum\" | postgres -o /dev/null -F -Q -D$PGDATA template1 > /dev/null\n> \n> echo \"COPY pg_shadow TO '$PGDATA/pg_pwd' USING DELIMITERS '\\\\t'\" |\\\n> postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n> @@ -383,6 +382,6 @@\n> echo \"loading pg_description\"\n> echo \"copy pg_description from '$TEMPLATE_DESCR'\" | postgres -F -Q -D$PGDATA te\n> mplate1 > /dev/null\n> echo \"copy pg_description from '$GLOBAL_DESCR'\" | postgres -F -Q -D$PGDATA temp\n> late1 > /dev/null\n> -echo \"vacuum analyze\" | postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n> +echo \"vacuum analyze\" | postgres -o /dev/null -F -Q -D$PGDATA template1 > /dev/\n> null |\\\n> grep -v \"^DEBUG:\"\n> \n> Any help appreciated !\n> \n> Reply-To set, since I haven't subscribed to your list.\n> \n> -- \n> Andreas Klemm http://www.FreeBSD.ORG/~andreas\n> powered by ,,symmetric multiprocessor FreeBSD''\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Sun, 22 Mar 1998 16:02:52 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
},
{
"msg_contents": "> \n> Hi !\n> \n> Please help, the FreeBSD port update to version 6.3 drives me to\n> despair ;-)\n> \n> --with-tcl configure switch doesn't work to build libpgtcl\n> \n> I only get it to work if I set in the make environment\n> \n> \tUSE_TCL=true\n> \n> configure --with-tcl doesn't seem to set USE_TCL=true in\n> src/Makefile.global, so\n> \tsrc/interfaces/Makefile,\n> which includes\n> \tsrc/Makefile.global\n> fails, since USE_TCL isn't set to true. Workaround was to explicitely\n> set USE_TCL=true when starting the make.\n\nWe are working on a patch for this.\n\n> \n> Another problem: on FreeBSD-current (and I think this is true on\n> FreeBSD-2.2.x as well)\n> \n> The make install fails when running the initdb script.\n> \n> Adding template1 database to pg_database...\n> Running: postgres -boot -C -F -D/usr/local/pgsql/data -Q template1 < /tmp/create\n> .13645\n> \n> vacuuming template1\n> creating public pg_user view\n> loading pg_description\n> *** Error code 1\n> \n> The Postgresql 6.2 version needed a patch so that this works\n> --- bin/initdb/initdb.sh.orig Wed Apr 2 07:09:13 1997\n> +++ bin/initdb/initdb.sh Wed Apr 2 07:09:33 1997\n> @@ -344,6 +344,5 @@\n> \n> if [ $debug -eq 0 ]; then\n> echo \"vacuuming template1\"\n> - echo \"vacuum\" | postgres -F -Q -D$PGDATA template1 2>&1 > /dev/null |\\\n> - grep -v \"^DEBUG:\"\n> + echo \"vacuum\" | postgres -o /dev/null -F -Q -D$PGDATA template1 > /dev/null\n> \n> fi\n> \n\nThere must be something else different on your machine. We all use it\nwithout problem. Are you running csh, or changed it somehow?\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, 22 Mar 1998 15:05:03 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
},
{
"msg_contents": "> \tOdd, since my testing has always been done on FreeBSD-current, and\n> I've already done a couple of builds under FreeBSD 2.2-STABLE :(\n> \n> \tBut, looking at your patches for initdb.sh, I've applied them, as\n> I think that it provides a non-shell dependent solution, which I think is\n> cleaner...\n> \n> \tI'm just about to do a build on my system, to make sure\n> everythingis okay before I bundle...but the new version should be up (and\n> announced) later this evening...\n\nI recommend against applying the patch. \n\nFirst, it is very late in the patch game to be applying an initdb patch.\nSecond, the original code should work fine, and the top of the file\nsays #/bin/sh, so we don't need a shell-independent solution. The rest\nof the file uses /bin/sh commands so it will simply not work with\nanother shell. I suggest the original user has some problem with their\nstderr handling in their OS configuration.\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, 22 Mar 1998 15:11:14 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
},
{
"msg_contents": "> I recommend against applying the patch. \n> \n> First, it is very late in the patch game to be applying an initdb patch.\n> Second, the original code should work fine, and the top of the file\n> says #/bin/sh, so we don't need a shell-independent solution. The rest\n> of the file uses /bin/sh commands so it will simply not work with\n> another shell. I suggest the original user has some problem with their\n> stderr handling in their OS configuration.\n\nMaybe -o is better. Marc will decide.\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, 22 Mar 1998 15:14:36 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
},
{
"msg_contents": "On Sun, Mar 22, 1998 at 04:02:52PM -0400, The Hermit Hacker wrote:\n> \n> \n> I just updated the ftp site such that there is a postgresql-6.3.1.tar.gz\n> file, which corrects alot of post-release bugs that cropped up, including\n> those you listed below...\n> \n> Let me know if this help...it hasn't been officially released yet, but,\n> unless anyone comes up with an critical problems with it, that tar file\n> will be what is announced tomorrow morning...\n\nThanks, you were blindingly fast ;-)\n\nThen maybe I will be commit ready if Satoshi agrees.\n\n\tAndreas ///\n\n-- \nAndreas Klemm http://www.FreeBSD.ORG/~andreas\npowered by ,,symmetric multiprocessor FreeBSD''\n",
"msg_date": "Sun, 22 Mar 1998 21:38:55 +0100",
"msg_from": "Andreas Klemm <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
},
{
"msg_contents": "BTW, what tcl tk version do you recommend ?\n80 or 81 ?\n\n-- \nAndreas Klemm http://www.FreeBSD.ORG/~andreas\npowered by ,,symmetric multiprocessor FreeBSD''\n",
"msg_date": "Sun, 22 Mar 1998 21:40:04 +0100",
"msg_from": "Andreas Klemm <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
},
{
"msg_contents": "On Sun, 22 Mar 1998, Andreas Klemm wrote:\n\n> On Sun, Mar 22, 1998 at 04:02:52PM -0400, The Hermit Hacker wrote:\n> > \n> > \n> > I just updated the ftp site such that there is a postgresql-6.3.1.tar.gz\n> > file, which corrects alot of post-release bugs that cropped up, including\n> > those you listed below...\n> > \n> > Let me know if this help...it hasn't been officially released yet, but,\n> > unless anyone comes up with an critical problems with it, that tar file\n> > will be what is announced tomorrow morning...\n> \n> Thanks, you were blindingly fast ;-)\n\n\tI wish I knew enough about building the ports (And had the time to\nlearn *sigh*)...then I could save you even more time :(\n\n",
"msg_date": "Mon, 23 Mar 1998 08:12:32 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
},
{
"msg_contents": "> > Thanks, you were blindingly fast ;-)\n> \n> \tI wish I knew enough about building the ports (And had the time to\n> learn *sigh*)...then I could save you even more time :(\n\nYeah, but don't desperate ;-)\n\n-- \nAndreas Klemm http://www.FreeBSD.ORG/~andreas\npowered by ,,symmetric multiprocessor FreeBSD''\n",
"msg_date": "Tue, 24 Mar 1998 18:31:56 +0100",
"msg_from": "Andreas Klemm <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] urgent: pgsql 6.3 FreeBSD port update trouble,\n\t--with-tcl doesn't work"
}
] |
[
{
"msg_contents": "Hi,\n\nThe buffer overrun while creating sequences turns out to be \na read access overrun. For the sake of doing things properly \nhere is a patch which fixes it.\n\nThis patch is for the file backend/commands/sequence.c.\n\n87a88\n> \tNameData\tname;\n113c114,115\n< \t\t\t\tvalue[i - 1] = PointerGetDatum(seq->seqname);\n---\n> \t\t\t\tnamestrcpy(&name, seq->seqname);\n> \t\t\t\tvalue[i - 1] = NameGetDatum(&name);\n\n\nWith regards from Maurice.\n",
"msg_date": "Sun, 22 Mar 1998 14:10:06 +0100",
"msg_from": "Maurice Gittens <[email protected]>",
"msg_from_op": true,
"msg_subject": "patch for buffer overrun with sequences"
},
{
"msg_contents": "Applied.\n\n> \n> Hi,\n> \n> The buffer overrun while creating sequences turns out to be \n> a read access overrun. For the sake of doing things properly \n> here is a patch which fixes it.\n> \n> This patch is for the file backend/commands/sequence.c.\n> \n> 87a88\n> > \tNameData\tname;\n> 113c114,115\n> < \t\t\t\tvalue[i - 1] = PointerGetDatum(seq->seqname);\n> ---\n> > \t\t\t\tnamestrcpy(&name, seq->seqname);\n> > \t\t\t\tvalue[i - 1] = NameGetDatum(&name);\n> \n> \n> With regards from Maurice.\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": "Mon, 30 Mar 1998 12:14:31 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] patch for buffer overrun with sequences"
}
] |
[
{
"msg_contents": "Hi !\n\nAm currently testing the port update of postgresql 6.3.\nAfter some trouble I got it managed ....\n\nNow I have trouble getting pgaccess running *sigh* :-)\n\na) if I don't load \n\t/usr/local/pgsql/lib/libpq.so\n\tprior\n\t/usr/local/pgsql/lib/libpgtcl.so\n\nI get the following failure:\n\nError in startup script: couldn't load file \"/usr/local/pgsql/lib/libpgtcl.so\": Undefined symbol \"_PQfinish\" in wish8.0:/usr/local/pgsql/lib/libpgtcl.so\n while executing\n\"load /usr/local/pgsql/lib/libpgtcl.so\"\n (procedure \"main\" line 4)\n invoked from within\n\"main $argc $argv\"\n (file \"/usr/local/bin/pgaccess\" line 4507)\n\nOk, after patching pgaccess to load the other library it looks like\nthis:\n\nandreas@titan{1062} $ pgaccess\nError in startup script: couldn't load file \"/usr/local/pgsql/lib/libpq.so\": Undefined symbol \"_crypt\" in wish8.0:/usr/local/pgsql/lib/libpq.so\n while executing\n\"load /usr/local/pgsql/lib/libpq.so\"\n (procedure \"main\" line 3)\n invoked from within\n\"main $argc $argv\"\n (file \"/usr/local/bin/pgaccess\" line 4507)\n\nWhat the hell is this ? ;-)\n\nSomebody with a quick idea, whats different in postgresql 6.3 compared\nto 6.2.1 ????\n\nSame when using tcl81/tk81...\n\n\tAndreas ///\n\n-- \nAndreas Klemm http://www.FreeBSD.ORG/~andreas\npowered by ,,symmetric multiprocessor FreeBSD''\n",
"msg_date": "Sun, 22 Mar 1998 15:22:33 +0100",
"msg_from": "Andreas Klemm <[email protected]>",
"msg_from_op": true,
"msg_subject": "problems running pgaccess 0.73 and 0.84 with new postgresql 6.3"
},
{
"msg_contents": "On Sun, 22 Mar 1998, Andreas Klemm wrote:\n\n> What the hell is this ? ;-)\n> \n> Somebody with a quick idea, whats different in postgresql 6.3 compared\n> to 6.2.1 ????\n> \n> Same when using tcl81/tk81...\n\nThis appears to be FreeBSD specific...when v6.3 was released, I started to\nplay around with pgaccess under Solaris x86, and it run without a\nhitch...here, under FreeBSD 3.0-current, I hit the same problem :(\n\nAs for what is different compared to 6.2.1...the shorter list would be\nwhat is the same ... \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Sun, 22 Mar 1998 16:05:46 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: problems running pgaccess 0.73 and 0.84 with new postgresql 6.3"
},
{
"msg_contents": "I confirmed, that -current is broken concerning pgaccess\nstartup. The old pgaccess/postgresql port doesn't run\nanymore.\n\nConcerning 6.3.1,\n\n--with-tcl still isn't sufficient, to build the libpgtcl.*\nlibraries.\n\nMy port now again uses USE_TCL=true instead of USE_TCL=yes.\n\n\tAndreas ///\n\n-- \nAndreas Klemm http://www.FreeBSD.ORG/~andreas\npowered by ,,symmetric multiprocessor FreeBSD''\n",
"msg_date": "Sun, 22 Mar 1998 22:43:25 +0100",
"msg_from": "Andreas Klemm <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: problems running pgaccess 0.73 and 0.84 with new postgresql 6.3"
},
{
"msg_contents": "On Sun, 22 Mar 1998, Andreas Klemm wrote:\n\n> I confirmed, that -current is broken concerning pgaccess\n> startup. The old pgaccess/postgresql port doesn't run\n> anymore.\n> \n> Concerning 6.3.1,\n> \n> --with-tcl still isn't sufficient, to build the libpgtcl.*\n> libraries.\n\n\tThen there is a problem on your system...I did the -with-tcl on my\nFreeBSD 3.0-CURRENT machine and it was sufficient to build the libpgtcl.*\nlibraries...\n\n\n",
"msg_date": "Mon, 23 Mar 1998 08:13:38 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: problems running pgaccess 0.73 and 0.84 with new postgresql 6.3"
}
] |
[
{
"msg_contents": "Hi,\n\nI noticed that the C++ interface for postgresql hasn't been modified for a\nwhile.\nDoes it currently have a maintainer?\nThere are some features I would like to see added to it.\n\nThanks,\nWith regards from Maurice.\n\n\n",
"msg_date": "Sun, 22 Mar 1998 18:38:33 +0100",
"msg_from": "\"Maurice Gittens\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "C++ interface"
},
{
"msg_contents": "On Sun, 22 Mar 1998, Maurice Gittens wrote:\n\n> Hi,\n> \n> I noticed that the C++ interface for postgresql hasn't been modified for a\n> while.\n> Does it currently have a maintainer?\n\n\tThere is no maintainer that I'm aware of...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Sun, 22 Mar 1998 14:12:24 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] C++ interface"
}
] |
[
{
"msg_contents": "\nI just put the final patch in place (unannounced) for anyone to try out...\n\nftp://ftp.postgresql.org/pub/postgresql-6.3-6.3.1.gz\n\nUnless there are any outstandign problems with either that patch or the\nfull distribution, I will put out an announcement of those two Monday\nmorning...\n\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Sun, 22 Mar 1998 16:08:31 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "Final Patch?"
}
] |
[
{
"msg_contents": "I have just made a change to allow configure to properly find tcl/tk. \nThe old code never found tk.h if it existed in /usr/include, and their\nwere other problems.\n\nI also removed the version-specific stuff from tcl/tk searching, so they\nwill have to be called tcl.h/tcl.[a/so] and tk.h/tk.[a/so] without the\nversion numbers. You will also have to specify the search location for\nthose directories when prompted by configure.\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, 22 Mar 1998 23:10:48 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "mega-patch"
},
{
"msg_contents": " I also removed the version-specific stuff from tcl/tk searching, so they\n will have to be called tcl.h/tcl.[a/so] and tk.h/tk.[a/so] without the\n version numbers. You will also have to specify the search location for\n those directories when prompted by configure.\n\nThis seems to defeat the purpose of the configure script. Isn't it\nsupposed to find whatever is needed, whereever it is, and let the rest\nof the software do the right thing?\n\nIn that case, it seems to me that configure should be checking all the\nrelevant version-specific locations without requiring any prompts.\nAfter a few iterations, the list of appropriate directories will be\nmore-or-less all inclusive, so all will be well. The --use-tcl or\nwhatever config option can also give an explicit location in case the\ncurrent ones are not appropriate.\n\nBut removing the other directories and requiring either a special\ninstallation or response to a prompt seems to go against the nature\n(and usefulness) of the configure system.\n\nCheers,\nBrook\n",
"msg_date": "Mon, 23 Mar 1998 08:35:35 -0700 (MST)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": "\nIts sad when, to save time, I only skim email that doesn't affect me\ndirectly... :(\n\nOn Mon, 23 Mar 1998, Brook Milligan wrote:\n\n> I also removed the version-specific stuff from tcl/tk searching, so they\n> will have to be called tcl.h/tcl.[a/so] and tk.h/tk.[a/so] without the\n> version numbers. You will also have to specify the search location for\n> those directories when prompted by configure.\n> \n> This seems to defeat the purpose of the configure script. Isn't it\n> supposed to find whatever is needed, whereever it is, and let the rest\n> of the software do the right thing?\n> \n> In that case, it seems to me that configure should be checking all the\n> relevant version-specific locations without requiring any prompts.\n> After a few iterations, the list of appropriate directories will be\n> more-or-less all inclusive, so all will be well. The --use-tcl or\n> whatever config option can also give an explicit location in case the\n> current ones are not appropriate.\n> \n> But removing the other directories and requiring either a special\n> installation or response to a prompt seems to go against the nature\n> (and usefulness) of the configure system.\n\nI have to agree, in part, with this...\n\nUnder FreeBSD, because of how our ports collections work, it is possible\nto have something installed using TCL8.0, but the next one to be installed\nrequires TCL8.1. As such, we installed then as /usr/local/lib/tcl80 and\n/usr/local/lib/tcl81, so they don't get confused...\n\nHaving it search several directories in order of preference, IMHO, is\nproper behaviour for configure...it should remain as version search\ninstead of \"just tcl.h\"...\n\n\n",
"msg_date": "Mon, 23 Mar 1998 10:49:12 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": "> This seems to defeat the purpose of the configure script. Isn't it\n> supposed to find whatever is needed, whereever it is, and let the rest\n> of the software do the right thing?\n> \n> In that case, it seems to me that configure should be checking all the\n> relevant version-specific locations without requiring any prompts.\n> After a few iterations, the list of appropriate directories will be\n> more-or-less all inclusive, so all will be well. The --use-tcl or\n> whatever config option can also give an explicit location in case the\n> current ones are not appropriate.\n> \n> But removing the other directories and requiring either a special\n> installation or response to a prompt seems to go against the nature\n> (and usefulness) of the configure system.\n\nIt seems tcl/tk uses directory names with version numbers, and I have\nno intention of searching for every version they have ever released. \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, 23 Mar 1998 11:01:10 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": "> I have to agree, in part, with this...\n> \n> Under FreeBSD, because of how our ports collections work, it is possible\n> to have something installed using TCL8.0, but the next one to be installed\n> requires TCL8.1. As such, we installed then as /usr/local/lib/tcl80 and\n> /usr/local/lib/tcl81, so they don't get confused...\n> \n> Having it search several directories in order of preference, IMHO, is\n> proper behaviour for configure...it should remain as version search\n> instead of \"just tcl.h\"...\n\nOK, go ahead. I did fix several problem with the tcl/tk stuff also. \nFor example, if it found /usr/include/tk.h, the flag would be -I with no\ndirectory, causing any compile to fail.\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, 23 Mar 1998 11:03:09 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": " It seems tcl/tk uses directory names with version numbers, and I have\n no intention of searching for every version they have ever released. \n\nI agree that it seems like a pain to do so (and perhaps that argues\nagainst the way tcl/tk is installed by the packaging systems), but\ngiven that some common packaging systems do it that way, it seems like\nwe should handle the situation nicely.\n\nWould it be possible to simply generate a list of directories to\nsearch from a regular expression (e.g., tcl[0-9.]*), reverse sort them\nso the later ones come first, then check for the right header file?\nIf too old a version was found we could flag it with a warning from\nconfigure.\n\nThat might be a win over a fixed list, because we wouldn't have to\nmaintain the list, which I think was your objection (which I sort of\nagree with, too, from the point of view of a maintainer).\n\nCheers,\nBrook\n",
"msg_date": "Mon, 23 Mar 1998 09:28:38 -0700 (MST)",
"msg_from": "Brook Milligan <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": "On Mon, 23 Mar 1998, Bruce Momjian wrote:\n\n> > This seems to defeat the purpose of the configure script. Isn't it\n> > supposed to find whatever is needed, whereever it is, and let the rest\n> > of the software do the right thing?\n> > \n> > In that case, it seems to me that configure should be checking all the\n> > relevant version-specific locations without requiring any prompts.\n> > After a few iterations, the list of appropriate directories will be\n> > more-or-less all inclusive, so all will be well. The --use-tcl or\n> > whatever config option can also give an explicit location in case the\n> > current ones are not appropriate.\n> > \n> > But removing the other directories and requiring either a special\n> > installation or response to a prompt seems to go against the nature\n> > (and usefulness) of the configure system.\n> \n> It seems tcl/tk uses directory names with version numbers, and I have\n> no intention of searching for every version they have ever released. \n\n\tWhy not? Or, rather, why not just do the ones that get reported\nto us? It does hurt to, and I think it hurts more not too...\n\n\n",
"msg_date": "Mon, 23 Mar 1998 11:49:19 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": "> > > But removing the other directories and requiring either a special\n> > > installation or response to a prompt seems to go against the nature\n> > > (and usefulness) of the configure system.\n> > \n> > It seems tcl/tk uses directory names with version numbers, and I have\n> > no intention of searching for every version they have ever released. \n> \n> \tWhy not? Or, rather, why not just do the ones that get reported\n> to us? It does hurt to, and I think it hurts more not too...\n\nIf someone wants to play with that, go ahead. I see no reason the\ninstallation person, if they want tcl/tk, should type in those directory\nnames. However, if the actual library and include files have version\nnumbers in them, well, that is a pain. Do they edit their source and\nmakefiles every time they upgrade versions?\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, 23 Mar 1998 12:06:30 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": "Marc wrote:\n\n| Having it search several directories in order of preference, IMHO, is\n| proper behaviour for configure...it should remain as version search\n| instead of \"just tcl.h\"...\n\nBruce answered:\n\n| OK, go ahead. I did fix several problem with the tcl/tk stuff also. \n| For example, if it found /usr/include/tk.h, the flag would be -I with no\n| directory, causing any compile to fail.\n\nWhile you guys are thinking about this, let me remind you of the\nfollowing patch for these problems that I sent you on March 2nd. It\nfixes the problem Bruce mentions, and extends the version checking to\ninclude the last versions of TCL and TK prior to 8.0 -- since I can\nobserve here that they do work.\n\nMarc's argument that version numbers have to stay where TCL/TK is\ninvolved is very important: they are nearly always incompatible with\nthemselves in preceding and following versions, making it necessary\nfor a typical site to have one complete TCL/TK installation for each\nand every application that uses the damn things. (You may recall that\nI expressed amazement at my discovery that I could use 7.6/4.2 with\nboth Scotty and PostgreSQL.)\n\nAnyway, my take on the --with-tcl (and --with-perl) things, relative\nto PostgreSQL 6.3:\n\n*** configure.in.orig\tMon Mar 2 06:33:14 1998\n--- configure.in\tMon Mar 2 21:42:59 1998\n***************\n*** 239,246 ****\n AC_ARG_WITH(\n tcl,\n [ --with-tcl use tcl ],\n! USE_TCL=true AC_MSG_RESULT(enabled),\n! USE_TCL=false AC_MSG_RESULT(disabled)\n )\n export USE_TCL\n USE_X=$USE_TCL\n--- 239,246 ----\n AC_ARG_WITH(\n tcl,\n [ --with-tcl use tcl ],\n! USE_TCL=true; AC_MSG_RESULT(enabled),\n! USE_TCL=false; AC_MSG_RESULT(disabled)\n )\n export USE_TCL\n USE_X=$USE_TCL\n***************\n*** 250,257 ****\n AC_ARG_WITH(\n perl,\n [ --with-perl use perl ],\n! USE_PERL=true AC_MSG_RESULT(enabled),\n! USE_PERL=false AC_MSG_RESULT(disabled)\n )\n export USE_PERL\n \n--- 250,257 ----\n AC_ARG_WITH(\n perl,\n [ --with-perl use perl ],\n! USE_PERL=true; AC_MSG_RESULT(enabled),\n! USE_PERL=false; AC_MSG_RESULT(disabled)\n )\n export USE_PERL\n \n***************\n*** 563,570 ****\n if test \"$USE_TCL\" = \"true\"\n then\n TCL_INCDIR=no\n! AC_CHECK_HEADER(tcl.h, TCL_INCDIR=)\n! for f in /usr/include /usr/include/tcl8.0 /usr/local/include /usr/local/include/tcl8.0; do\n if test \"$TCL_INCDIR\" = \"no\"; then\n AC_CHECK_HEADER($f/tcl.h, TCL_INCDIR=$f)\n fi\n--- 563,569 ----\n if test \"$USE_TCL\" = \"true\"\n then\n TCL_INCDIR=no\n! for f in /usr/include /usr/include/tcl8.0 /usr/include/tcl7.6 /usr/local/include /usr/local/include/tcl8.0 /usr/local/include/tcl7.6; do\n if test \"$TCL_INCDIR\" = \"no\"; then\n AC_CHECK_HEADER($f/tcl.h, TCL_INCDIR=$f)\n fi\n***************\n*** 580,586 ****\n if test \"$USE_TCL\" = \"true\"\n then\n TCL_LIB=\n! for f in tcl8.0 tcl80; do\n if test -z \"$TCL_LIB\"; then\n AC_CHECK_LIB($f, main, TCL_LIB=$f)\n fi\n--- 579,585 ----\n if test \"$USE_TCL\" = \"true\"\n then\n TCL_LIB=\n! for f in tcl8.0 tcl80 tcl7.6 tcl76; do\n if test -z \"$TCL_LIB\"; then\n AC_CHECK_LIB($f, main, TCL_LIB=$f)\n fi\n***************\n*** 606,616 ****\n ice_save_CPPFLAGS=\"$CPPFLAGS\"\n ice_save_LDFLAGS=\"$LDFLAGS\"\n \n CPPFLAGS=\"$CPPFLAGS $X_CFLAGS -I$TCL_INCDIR\"\n \n TK_INCDIR=no\n! AC_CHECK_HEADER(tk.h, TK_INCDIR=)\n! for f in /usr/include /usr/include/tk8.0 /usr/local/include /usr/local/include/tk8.0; do\n if test \"$TK_INCDIR\" = \"no\"; then\n AC_CHECK_HEADER($f/tk.h, TK_INCDIR=$f)\n fi\n--- 605,619 ----\n ice_save_CPPFLAGS=\"$CPPFLAGS\"\n ice_save_LDFLAGS=\"$LDFLAGS\"\n \n+ if test \"$TCL_INCDIR\" = \"\"\n+ then\n+ CPPFLAGS=\"$CPPFLAGS $X_CFLAGS\"\n+ else\n CPPFLAGS=\"$CPPFLAGS $X_CFLAGS -I$TCL_INCDIR\"\n+ fi\n \n TK_INCDIR=no\n! for f in /usr/include /usr/include/tk8.0 /usr/include/tk4.2 /usr/local/include /usr/local/include/tk8.0 /usr/local/include/tk4.2; do\n if test \"$TK_INCDIR\" = \"no\"; then\n AC_CHECK_HEADER($f/tk.h, TK_INCDIR=$f)\n fi\n***************\n*** 631,637 ****\n if test \"$USE_TCL\" = \"true\"\n then\n TK_LIB=\n! for f in tk8.0 tk80; do\n if test -z \"$TK_LIB\"; then\n AC_CHECK_LIB($f, main, TK_LIB=$f)\n fi\n--- 634,640 ----\n if test \"$USE_TCL\" = \"true\"\n then\n TK_LIB=\n! for f in tk8.0 tk80 tk4.2 tk42; do\n if test -z \"$TK_LIB\"; then\n AC_CHECK_LIB($f, main, TK_LIB=$f)\n fi\n*** configure.orig\tMon Mar 2 15:55:42 1998\n--- configure\tMon Mar 2 21:43:39 1998\n***************\n*** 825,833 ****\n # Check whether --with-tcl or --without-tcl was given.\n if test \"${with_tcl+set}\" = set; then\n withval=\"$with_tcl\"\n! USE_TCL=true echo \"$ac_t\"\"enabled\" 1>&6\n else\n! USE_TCL=false echo \"$ac_t\"\"disabled\" 1>&6\n \n fi\n \n--- 825,833 ----\n # Check whether --with-tcl or --without-tcl was given.\n if test \"${with_tcl+set}\" = set; then\n withval=\"$with_tcl\"\n! USE_TCL=true; echo \"$ac_t\"\"enabled\" 1>&6\n else\n! USE_TCL=false; echo \"$ac_t\"\"disabled\" 1>&6\n \n fi\n \n***************\n*** 839,847 ****\n # Check whether --with-perl or --without-perl was given.\n if test \"${with_perl+set}\" = set; then\n withval=\"$with_perl\"\n! USE_PERL=true echo \"$ac_t\"\"enabled\" 1>&6\n else\n! USE_PERL=false echo \"$ac_t\"\"disabled\" 1>&6\n \n fi\n \n--- 839,847 ----\n # Check whether --with-perl or --without-perl was given.\n if test \"${with_perl+set}\" = set; then\n withval=\"$with_perl\"\n! USE_PERL=true; echo \"$ac_t\"\"enabled\" 1>&6\n else\n! USE_PERL=false; echo \"$ac_t\"\"disabled\" 1>&6\n \n fi\n \n***************\n*** 5393,5432 ****\n if test \"$USE_TCL\" = \"true\"\n then\n TCL_INCDIR=no\n! ac_safe=`echo \"tcl.h\" | sed 'y%./+-%__p_%'`\n! echo $ac_n \"checking for tcl.h\"\"... $ac_c\" 1>&6\n! echo \"configure:5399: checking for tcl.h\" >&5\n! if eval \"test \\\"`echo '$''{'ac_cv_header_$ac_safe'+set}'`\\\" = set\"; then\n! echo $ac_n \"(cached) $ac_c\" 1>&6\n! else\n! cat > conftest.$ac_ext <<EOF\n! #line 5404 \"configure\"\n! #include \"confdefs.h\"\n! #include <tcl.h>\n! EOF\n! ac_try=\"$ac_cpp conftest.$ac_ext >/dev/null 2>conftest.out\"\n! { (eval echo configure:5409: \\\"$ac_try\\\") 1>&5; (eval $ac_try) 2>&5; }\n! ac_err=`grep -v '^ *+' conftest.out`\n! if test -z \"$ac_err\"; then\n! rm -rf conftest*\n! eval \"ac_cv_header_$ac_safe=yes\"\n! else\n! echo \"$ac_err\" >&5\n! echo \"configure: failed program was:\" >&5\n! cat conftest.$ac_ext >&5\n! rm -rf conftest*\n! eval \"ac_cv_header_$ac_safe=no\"\n! fi\n! rm -f conftest*\n! fi\n! if eval \"test \\\"`echo '$ac_cv_header_'$ac_safe`\\\" = yes\"; then\n! echo \"$ac_t\"\"yes\" 1>&6\n! TCL_INCDIR=\n! else\n! echo \"$ac_t\"\"no\" 1>&6\n! fi\n! \n! for f in /usr/include /usr/include/tcl8.0 /usr/local/include /usr/local/include/tcl8.0; do\n if test \"$TCL_INCDIR\" = \"no\"; then\n ac_safe=`echo \"$f/tcl.h\" | sed 'y%./+-%__p_%'`\n echo $ac_n \"checking for $f/tcl.h\"\"... $ac_c\" 1>&6\n--- 5393,5399 ----\n if test \"$USE_TCL\" = \"true\"\n then\n TCL_INCDIR=no\n! for f in /usr/include /usr/include/tcl8.0 /usr/include/tcl7.6 /usr/local/include /usr/local/include/tcl8.0 /usr/local/include/tcl7.6; do\n if test \"$TCL_INCDIR\" = \"no\"; then\n ac_safe=`echo \"$f/tcl.h\" | sed 'y%./+-%__p_%'`\n echo $ac_n \"checking for $f/tcl.h\"\"... $ac_c\" 1>&6\n***************\n*** 5473,5479 ****\n if test \"$USE_TCL\" = \"true\"\n then\n TCL_LIB=\n! for f in tcl8.0 tcl80; do\n if test -z \"$TCL_LIB\"; then\n echo $ac_n \"checking for main in -l$f\"\"... $ac_c\" 1>&6\n echo \"configure:5480: checking for main in -l$f\" >&5\n--- 5440,5446 ----\n if test \"$USE_TCL\" = \"true\"\n then\n TCL_LIB=\n! for f in tcl8.0 tcl80 tcl7.6 tcl76; do\n if test -z \"$TCL_LIB\"; then\n echo $ac_n \"checking for main in -l$f\"\"... $ac_c\" 1>&6\n echo \"configure:5480: checking for main in -l$f\" >&5\n***************\n*** 5531,5573 ****\n ice_save_CPPFLAGS=\"$CPPFLAGS\"\n ice_save_LDFLAGS=\"$LDFLAGS\"\n \n! CPPFLAGS=\"$CPPFLAGS $X_CFLAGS -I$TCL_INCDIR\"\n! \n! TK_INCDIR=no\n! ac_safe=`echo \"tk.h\" | sed 'y%./+-%__p_%'`\n! echo $ac_n \"checking for tk.h\"\"... $ac_c\" 1>&6\n! echo \"configure:5540: checking for tk.h\" >&5\n! if eval \"test \\\"`echo '$''{'ac_cv_header_$ac_safe'+set}'`\\\" = set\"; then\n! echo $ac_n \"(cached) $ac_c\" 1>&6\n! else\n! cat > conftest.$ac_ext <<EOF\n! #line 5545 \"configure\"\n! #include \"confdefs.h\"\n! #include <tk.h>\n! EOF\n! ac_try=\"$ac_cpp conftest.$ac_ext >/dev/null 2>conftest.out\"\n! { (eval echo configure:5550: \\\"$ac_try\\\") 1>&5; (eval $ac_try) 2>&5; }\n! ac_err=`grep -v '^ *+' conftest.out`\n! if test -z \"$ac_err\"; then\n! rm -rf conftest*\n! eval \"ac_cv_header_$ac_safe=yes\"\n! else\n! echo \"$ac_err\" >&5\n! echo \"configure: failed program was:\" >&5\n! cat conftest.$ac_ext >&5\n! rm -rf conftest*\n! eval \"ac_cv_header_$ac_safe=no\"\n! fi\n! rm -f conftest*\n! fi\n! if eval \"test \\\"`echo '$ac_cv_header_'$ac_safe`\\\" = yes\"; then\n! echo \"$ac_t\"\"yes\" 1>&6\n! TK_INCDIR=\n else\n! echo \"$ac_t\"\"no\" 1>&6\n fi\n \n! for f in /usr/include /usr/include/tk8.0 /usr/local/include /usr/local/include/tk8.0; do\n if test \"$TK_INCDIR\" = \"no\"; then\n ac_safe=`echo \"$f/tk.h\" | sed 'y%./+-%__p_%'`\n echo $ac_n \"checking for $f/tk.h\"\"... $ac_c\" 1>&6\n--- 5498,5512 ----\n ice_save_CPPFLAGS=\"$CPPFLAGS\"\n ice_save_LDFLAGS=\"$LDFLAGS\"\n \n! if test \"$TCL_INCDIR\" = \"\"\n! then\n! CPPFLAGS=\"$CPPFLAGS $X_CFLAGS\"\n else\n! CPPFLAGS=\"$CPPFLAGS $X_CFLAGS -I$TCL_INCDIR\"\n fi\n \n! TK_INCDIR=no\n! for f in /usr/include /usr/include/tk8.0 /usr/include/tk4.2 /usr/local/include /usr/local/include/tk8.0 /usr/local/include/tk4.2; do\n if test \"$TK_INCDIR\" = \"no\"; then\n ac_safe=`echo \"$f/tk.h\" | sed 'y%./+-%__p_%'`\n echo $ac_n \"checking for $f/tk.h\"\"... $ac_c\" 1>&6\n***************\n*** 5619,5625 ****\n if test \"$USE_TCL\" = \"true\"\n then\n TK_LIB=\n! for f in tk8.0 tk80; do\n if test -z \"$TK_LIB\"; then\n echo $ac_n \"checking for main in -l$f\"\"... $ac_c\" 1>&6\n echo \"configure:5626: checking for main in -l$f\" >&5\n--- 5558,5564 ----\n if test \"$USE_TCL\" = \"true\"\n then\n TK_LIB=\n! for f in tk8.0 tk80 tk4.2 tk42; do\n if test -z \"$TK_LIB\"; then\n echo $ac_n \"checking for main in -l$f\"\"... $ac_c\" 1>&6\n echo \"configure:5626: checking for main in -l$f\" >&5\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n",
"msg_date": "23 Mar 1998 20:59:10 +0100",
"msg_from": "Tom Ivar Helbekkmo <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": "On 23 Mar 1998, Tom Ivar Helbekkmo wrote:\n\n> Marc's argument that version numbers have to stay where TCL/TK is\n> involved is very important: they are nearly always incompatible with\n> themselves in preceding and following versions, making it necessary\n> for a typical site to have one complete TCL/TK installation for each\n> and every application that uses the damn things. (You may recall that\n> I expressed amazement at my discovery that I could use 7.6/4.2 with\n> both Scotty and PostgreSQL.)\n> \n> Anyway, my take on the --with-tcl (and --with-perl) things, relative\n> to PostgreSQL 6.3:\n\n\tCan you take a quick look at the way things are right now and\nsubmit a new patch against the current version for the tcl stuff? I think\nthis \"multiple version\" thing with tcl/tk is an inconvience, but a\nnecessary one based on the way they do do development :(\n\n\tOther then this, I don't think we have any outstanding issues we\nwanted covered...is there? If not, if this can be submitted before 4pm\ntoday, I'll include it...if not, the patch/tar file is created and\n*finally* released at that point :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 24 Mar 1998 00:54:56 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
},
{
"msg_contents": "Marc wrote:\n\n> \tCan you take a quick look at the way things are right now and\n> submit a new patch against the current version for the tcl stuff? I\n> think this \"multiple version\" thing with tcl/tk is an inconvience,\n> but a necessary one based on the way they do do development :(\n\nSorry -- not enough time for that. It would mean changing several\nfiles (after finding out which files to change) in ways that I can't\ntest sufficiently well within the time frame. I'll have to do it\neventually, of course, in order to get PostgreSQL 6.3.1 to work with\nTCL/TK here, and I'll send you the changes it takes. An alternative,\nof course, is to get Bruce to identify and back out the changes that\nhe made in order to remove all version-specific TCL/TK stuff from\nconfigure*, various makefiles and whatever. Then, you could just slip\nmy patch right into configure.in, add support for tcl/tk 8.1 to match\nthe version 8.0 stuff, and generate a new configure...\n\nI don't even like TCL! It's just that pgaccess is _so_ _cute_! :-)\n\n[Meanwhile, I'm hacking on PL/Scheme -- I can already create and use\nfunctions with \"language 'scheme'\", and I'm working on integrating\ncalls back into the backend into Scheme the way I want them. Having a\ngreat time with this stuff. Don't hold your breath waiting for this,\nthough; it's strictly a spare time project, and will take a while.]\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n",
"msg_date": "24 Mar 1998 19:05:52 +0100",
"msg_from": "Tom Ivar Helbekkmo <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] mega-patch--tcl/tk configuration"
}
] |
[
{
"msg_contents": "Found more tcl/tk configure problems. Just fixed them.\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, 23 Mar 1998 01:34:21 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "tcl/tk"
}
] |
[
{
"msg_contents": "On BSDI, 'perl' is perl4 and 'perl5' is perl5. This causes\ninterfaces/Makefile to fail when running perl5/Makefile.PL.\n\nDoes anyone else have such a problem? I don't want to fix it unless it\nis a problem for others.\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, 23 Mar 1998 01:39:31 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "perl/perl5"
},
{
"msg_contents": "\nI would consider that to be a bug, not a feature. Once perl is\ninstalled locally from source, it becomes 'perl' and replaces common\npath locations of the old version with itself (/usr/bin/perl).\n\nOn Mon, 23 March 1998, at 01:39:31, Bruce Momjian wrote:\n\n> On BSDI, 'perl' is perl4 and 'perl5' is perl5. This causes\n> interfaces/Makefile to fail when running perl5/Makefile.PL.\n> \n> Does anyone else have such a problem? I don't want to fix it unless it\n> is a problem for others.\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",
"msg_date": "Sun, 22 Mar 1998 22:54:13 -0800 (PST)",
"msg_from": "Brett McCormick <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] perl/perl5"
},
{
"msg_contents": "> I would consider that to be a bug, not a feature. Once perl is\n> installed locally from source, it becomes 'perl' and replaces common\n> path locations of the old version with itself (/usr/bin/perl).\n> \n> On Mon, 23 March 1998, at 01:39:31, Bruce Momjian wrote:\n> \n> > On BSDI, 'perl' is perl4 and 'perl5' is perl5. This causes\n> > interfaces/Makefile to fail when running perl5/Makefile.PL.\n> > \n> > Does anyone else have such a problem? I don't want to fix it unless it\n> > is a problem for others.\n\nIf you ask a perl4 question on the comp.lang.perl, well, just lets say you\nwon't be \"helped\" much. The offical perl position is the perl4 is an\nobsolete nightmare that they are all trying to forget as rapidly as possible\nand that perl5 is the bright shiny present and future. That is, perl4\nis not supported anymore, please upgrade etc...\n\nSo, given that it is pretty darn easy to get perl5 these days, it has been\nout for a few years with good results, is very compatible with perl4 anyhow,\nand really is better for a number of reasons, I see no reason to continue\nto support perl4.\n\nLets just do it. If for some obsolete platforms like SCO or suchlike this\nmeans that they will have to download perl5 and pgsql, so be it. The DB\nsupport is so much better it will be doing them a favor anyway...\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": "Sun, 22 Mar 1998 23:21:37 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] perl/perl5"
},
{
"msg_contents": "On Mon, 23 Mar 1998, Bruce Momjian wrote:\n\n> On BSDI, 'perl' is perl4 and 'perl5' is perl5. This causes\n> interfaces/Makefile to fail when running perl5/Makefile.PL.\n> \n> Does anyone else have such a problem? I don't want to fix it unless it\n> is a problem for others.\n\n\tNo problem here...nothing to fix, IMHO :)\n\n\n",
"msg_date": "Mon, 23 Mar 1998 08:15:53 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] perl/perl5"
},
{
"msg_contents": "> If you ask a perl4 question on the comp.lang.perl, well, just lets say you\n> won't be \"helped\" much. The offical perl position is the perl4 is an\n> obsolete nightmare that they are all trying to forget as rapidly as possible\n> and that perl5 is the bright shiny present and future. That is, perl4\n> is not supported anymore, please upgrade etc...\n> \n> So, given that it is pretty darn easy to get perl5 these days, it has been\n> out for a few years with good results, is very compatible with perl4 anyhow,\n> and really is better for a number of reasons, I see no reason to continue\n> to support perl4.\n> \n> Lets just do it. If for some obsolete platforms like SCO or suchlike this\n> means that they will have to download perl5 and pgsql, so be it. The DB\n> support is so much better it will be doing them a favor anyway...\n\nOK, I will leave our code alone. Our Makefile.PL requires perl5, and\nBSD/OS has perl as pel5, so if other people had problems doing make of\ninterfaces/perl5, I was going to add code to look for perl5 first, and\nuse perl only if a perl5 binary did not exist, but if no one else has\nthis problem, it is not worth changing 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": "Mon, 23 Mar 1998 09:28:18 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] perl/perl5"
},
{
"msg_contents": "> From: [email protected] (David Gould)\n> Date: Sun, 22 Mar 1998 23:21:37 -0800 (PST)\n\n> If you ask a perl4 question on the comp.lang.perl, well, just lets say you\n> won't be \"helped\" much. The offical perl position is the perl4 is an\n> obsolete nightmare that they are all trying to forget as rapidly as possible\n> and that perl5 is the bright shiny present and future. That is, perl4\n> is not supported anymore, please upgrade etc...\n> \n> So, given that it is pretty darn easy to get perl5 these days, it has been\n> out for a few years with good results, is very compatible with perl4 anyhow,\n> and really is better for a number of reasons, I see no reason to continue\n> to support perl4.\n> \n> Lets just do it. If for some obsolete platforms like SCO or suchlike this\n> means that they will have to download perl5 and pgsql, so be it. The DB\n> support is so much better it will be doing them a favor anyway...\n> \n> -dg\n\nI basically agree with this but there is one minor problem with\nFreeBSD in that it still inexplicably ships with /usr/bin/perl which\nis Perl 4 - you have to add Perl 5 which becomes /usr/local/bin/perl.\nIf I had total control of all our servers I would just rename the old\nperl to perl4 (or delete it!) but I don't have that luxury.\n\nSo the FreeBSD pg/Perl install should probably have a tweak to look\nfor /usr/local/bin/perl and not stumble into /usr/bin/[dead]perl.\nDon't know about other BSD's.\n\nHal\n\n",
"msg_date": "Mon, 23 Mar 1998 11:18:20 -0600 (CST)",
"msg_from": "Hal Snyder <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] perl/perl5"
},
{
"msg_contents": "> I basically agree with this but there is one minor problem with\n> FreeBSD in that it still inexplicably ships with /usr/bin/perl which\n> is Perl 4 - you have to add Perl 5 which becomes /usr/local/bin/perl.\n> If I had total control of all our servers I would just rename the old\n> perl to perl4 (or delete it!) but I don't have that luxury.\n> \n> So the FreeBSD pg/Perl install should probably have a tweak to look\n> for /usr/local/bin/perl and not stumble into /usr/bin/[dead]perl.\n> Don't know about other BSD's.\n> \n\nMy idea was to search for a perl5 binary first, then perl, but that\ndoesn't seem to work for anyone else. No one has complained about this\nyet, either.\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, 23 Mar 1998 12:26:43 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] perl/perl5"
}
] |
[
{
"msg_contents": "David Gould wrote:\n>That said, I have a plan in mind to get a nice performance boost for palloc().\n>As a side effect, I will guarantee 8 byte alignment of pointers returned\n>from palloc().\n\nWow, this sounds really good, will you try to prealloc bigger chunks of memory\nto avoid the many system calls ? I was wondering if that could give a speed boost.\n\nAndreas\n\n\n\n\n\n",
"msg_date": "Mon, 23 Mar 1998 10:13:31 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] patch for memory overrun on Linux(i386)"
},
{
"msg_contents": "> David Gould wrote:\n> >That said, I have a plan in mind to get a nice performance boost for palloc().\n> >As a side effect, I will guarantee 8 byte alignment of pointers returned\n> >from palloc().\n> \n> Wow, this sounds really good, will you try to prealloc bigger chunks of memory\n> to avoid the many system calls ? I was wondering if that could give a speed boost.\n\nYes to both points.\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": "Mon, 23 Mar 1998 11:04:26 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] patch for memory overrun on Linux(i386)"
}
] |
[
{
"msg_contents": "I just cvsup'ed the latest snapshot. After compiling and installing I can\nconnect but not display my tables anymore:\n\ngauss:test 170) psql 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\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.\nmm=> \\q\n\nWorking with my tables still works fine, just \\z doesn't.\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, 23 Mar 1998 11:22:25 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "problem with current snapshot"
}
] |
[
{
"msg_contents": "> 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\nIt appears from man pages on our Irix system that Irix6 has it but Irix5\ndoes not.\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": "Mon, 23 Mar 1998 10:23:32 GMT",
"msg_from": "Andrew Martin <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: sched_yield()"
}
] |
[
{
"msg_contents": "> Gee, I just checked and BSDI has it. However, it appears to work only\n> on 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>From the Irix6 man page:\n\n #include <sched.h>\n...\n The sched_yield system call causes the calling process to relinquish the\n processor to a runnable process of higher or equal priority. \n\nLooks like it should work here, but needs a different header file from BDSI.\n\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": "Mon, 23 Mar 1998 10:32:11 GMT",
"msg_from": "Andrew Martin <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Re: sched_yield()"
}
] |
[
{
"msg_contents": " Reply to: RE>\n\nHello,\n\n Given the results of a query, I need to determine the maximum length of a\nvariable length field using the perl interface to postgres. Is there\nsomething in the \"results\" buffer I can use or do I need to query the system\ncatalogs for this information? Also, what field in which class contains\nthis information? I couldn't seem to find a field in the pg_attribute class\nwhich provides this information.\n\n Also, is their a descriptive database design document available for the\nsystem catalogs?\n\nTIA,\nAndy\n\n",
"msg_date": "23 Mar 1998 11:15:10 U",
"msg_from": "\"Andy Farrell\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: "
}
] |
[
{
"msg_contents": "I'm about finished with indicator variables. They seem to work well with\noutput variables as well as with input variables in insert statements.\n\nNow the one question: Informix and Sybase have three different settings for\nindicator variables: \n\n-1 means value is null\n0 means value is okay\n> 0 means value is okay, but string was truncated to n bytes\n\nI tried Oracle but apparently they just use 0 and -1. Does the standard say\nanything about this?\n\nBTW is this SQL standard on-line anywhere? A WWW page maybe?\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, 23 Mar 1998 15:09:23 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "just another standards question"
},
{
"msg_contents": "> Now the one question: Informix and Sybase have three different \n> settings for indicator variables:\n> \n> -1 means value is null\n> 0 means value is okay\n> > 0 means value is okay, but string was truncated to n bytes\n> \n> I tried Oracle but apparently they just use 0 and -1. Does the \n> standard say anything about this?\n> \n> BTW is this SQL standard on-line anywhere? A WWW page maybe?\n\nI had found a draft of the SQL92 standard on-line somewhere many months\nago; don't remember where. It is ~1.6MB uncompressed, probably much\nsmaller compressed; do you want me to send you a copy?\n\nI believe that the official standard is available in print only, since\nit is a cost-covering operation of the standards organization.\n\nAnyway, it looks like Informix and Sybase have it right:\n\n 4.18.3 Indicator parameters\n\n An indicator parameter is an integer parameter that is specified\n immediately following another parameter. Its primary use is to\n indicate whether the value that the other parameter assumes or\n supplies is a null value. An indicator parameter cannot immediately\n follow another indicator parameter.\n\n The other use for indicator parameters is to indicate whether\n string data truncation occurred during a transfer between a host\n program and an SQL-implementation in parameters or host variables.\n If a non-null string value is transferred and the length of the\n target data item is sufficient to accept the entire source data\n item, then the indicator parameter or variable is set to 0 to in-\n dicate that truncation did not occur. However, if the length of\n the target data item is insufficient, then the indicator parame-\n ter or variable is set to the length of the source data item (in\n characters or bits, as appropriate) to indicate that truncation\n occurred and to indicate the original length in characters or bits,\n as appropriate, of the source.\n ...\n\n - Tom\n",
"msg_date": "Mon, 23 Mar 1998 15:52:00 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] just another standards question"
},
{
"msg_contents": "Thomas G. Lockhart writes:\n> I had found a draft of the SQL92 standard on-line somewhere many months\n> ago; don't remember where. It is ~1.6MB uncompressed, probably much\n> smaller compressed; do you want me to send you a copy?\n\nYes, please.\n\n> Anyway, it looks like Informix and Sybase have it right:\n> \n> 4.18.3 Indicator parameters\n> \n> An indicator parameter is an integer parameter that is specified\n> immediately following another parameter. Its primary use is to\n> indicate whether the value that the other parameter assumes or\n> supplies is a null value. An indicator parameter cannot immediately\n> follow another indicator parameter.\n\nOkay, that works completely.\n\n> The other use for indicator parameters is to indicate whether\n> string data truncation occurred during a transfer between a host\n> program and an SQL-implementation in parameters or host variables.\n> If a non-null string value is transferred and the length of the\n> target data item is sufficient to accept the entire source data\n> item, then the indicator parameter or variable is set to 0 to in-\n> dicate that truncation did not occur. However, if the length of\n> the target data item is insufficient, then the indicator parame-\n> ter or variable is set to the length of the source data item (in\n> characters or bits, as appropriate) to indicate that truncation\n> occurred and to indicate the original length in characters or bits,\n> as appropriate, of the source.\n\nSo I have to set the indicator also during writes? That is if I insert\nstring \"foobar\" into char(3), do I have to set the indicator to 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",
"msg_date": "Mon, 23 Mar 1998 16:52:52 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] just another standards question"
}
] |
[
{
"msg_contents": "Argh!!!!\n\nYes, I forgot that make clean.\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:\tMonday, March 23, 1998 3:35 PM\n> To:\[email protected]\n> Subject:\tRe: [HACKERS] problem with current snapshot\n> \n> I assume you did a make clean too?\n> \n> \n",
"msg_date": "Mon, 23 Mar 1998 15:59:41 +0100",
"msg_from": "\"Meskes, Michael\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] problem with current snapshot"
}
] |
[
{
"msg_contents": "Yes, but since it is, why don't we inline it? Is there a possible\ncaveat?\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:\tMonday, March 23, 1998 4:08 PM\n> To:\[email protected]\n> Cc:\[email protected]\n> Subject:\tRe: [HACKERS] Profile of current backend\n> > Isn't this a good solution? A function only called by one other\n> function has\n> > its right to exist only for readability. And this optimization could\n> be done\n> > automatically.\n> \n> Wouldn't be such a big deal if it was not call so many times.\n> \n> -- \n> Bruce Momjian | 830 Blythe Avenue\n> [email protected] | Drexel Hill, Pennsylvania\n> 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, 23 Mar 1998 16:02:57 +0100",
"msg_from": "\"Meskes, Michael\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] Profile of current backend"
},
{
"msg_contents": "> \n> Yes, but since it is, why don't we inline it? Is there a possible\n> caveat?\n\nInline it. I just haven't had time or studied that part of the code. I\nhad hoped the original researcher would do the honors, but I don't\nremember who that was, or if it was you.\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": "Mon, 23 Mar 1998 10:57:37 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Profile of current backend"
}
] |
[
{
"msg_contents": "\nI noticed a couple of tcl/tk improvements from Bruce this morning, and\nfixed the geqo_paths.c problem...unless there are any major disagreements,\nI will announce this by 4:30ADT this afternoon as being the official\n6.3.1...\n\n\n\n",
"msg_date": "Mon, 23 Mar 1998 10:05:14 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "Final one...?"
},
{
"msg_contents": "> \n> \n> I noticed a couple of tcl/tk improvements from Bruce this morning, and\n> fixed the geqo_paths.c problem...unless there are any major disagreements,\n> I will announce this by 4:30ADT this afternoon as being the official\n> 6.3.1...\n> \n\nGood. Doing some manual pages cleanups, and now I see it is being\ncalled 6.3.1 so I have to change the HISTORY, TODO, and other things.\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, 23 Mar 1998 10:09:54 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Final one...?"
},
{
"msg_contents": "On Mon, 23 Mar 1998, Bruce Momjian wrote:\n\n> > \n> > \n> > I noticed a couple of tcl/tk improvements from Bruce this morning, and\n> > fixed the geqo_paths.c problem...unless there are any major disagreements,\n> > I will announce this by 4:30ADT this afternoon as being the official\n> > 6.3.1...\n> > \n> \n> Good. Doing some manual pages cleanups, and now I see it is being\n> called 6.3.1 so I have to change the HISTORY, TODO, and other things.\n\n\tNo no no...no more changes :) What do you have it called\neverywhere else? I'll change my file names...I want to get this out :)\n\n\n",
"msg_date": "Mon, 23 Mar 1998 10:11:45 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Final one...?"
},
{
"msg_contents": "> \n> On Mon, 23 Mar 1998, Bruce Momjian wrote:\n> \n> > > \n> > > \n> > > I noticed a couple of tcl/tk improvements from Bruce this morning, and\n> > > fixed the geqo_paths.c problem...unless there are any major disagreements,\n> > > I will announce this by 4:30ADT this afternoon as being the official\n> > > 6.3.1...\n> > > \n> > \n> > Good. Doing some manual pages cleanups, and now I see it is being\n> > called 6.3.1 so I have to change the HISTORY, TODO, and other things.\n> \n> \tNo no no...no more changes :) What do you have it called\n> everywhere else? I'll change my file names...I want to get this out :)\n\nIf is was just going to be called a patch, we could continue to call it\n6.3, but since there is now a 6.3.1 release, I need to call it 6.3.1.\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, 23 Mar 1998 10:59:10 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Final one...?"
}
] |
[
{
"msg_contents": "> 1. Regression tests make very extensive use of char16. Seems to\n> me that these should be changed to a name. Same for tutorial and\n> any other example references in the code. Seem ok?\n\n** down sinks the mast of the old postquel, there, can you see it blink in the deep green water ? **\n** nostalgia on my face ** I think I forgot to mention my nicest hobby, sailing ** \n\nI think char(16) would be the correct replacement in the regression tests\nand tutorial, since this is what we want people to use. name is more or less\nan internal type for system catalogs and the like. I would not propagate it too much.\nOf course regress.c would need a little patch then, but no big deal.\n\nAndreas\n\n\n",
"msg_date": "Mon, 23 Mar 1998 16:15:17 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] char types gone."
},
{
"msg_contents": "On Mon, 23 Mar 1998, Zeugswetter Andreas wrote:\n\n> > 1. Regression tests make very extensive use of char16. Seems to\n> > me that these should be changed to a name. Same for tutorial and\n> > any other example references in the code. Seem ok?\n> \n> ** down sinks the mast of the old postquel, there, can you see it blink\n> in the deep green water ? ** ** nostalgia on my face ** I think I forgot\n> to mention my nicest hobby, sailing **\n\n\tOkay, this one has me...what benefit is there to char16 over\nchar(16)? Why, in the first place, was it implemented as char16 anyway?\n\n\tAs for \"down sinks the mast\"...so far as I've been able to tell,\nwe have *never* removed anything that wasn't replaced by a better way of\ndoing it (time travel with triggers)...we didn't leave anyone floundering\nfor a solution to it, we provided it...but now, its not forced on\n*everyone*, just on those databases/tables that require it...\n\n\tSame functionality, better performance...\n\n\n\n\n",
"msg_date": "Mon, 23 Mar 1998 10:34:16 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] char types gone."
},
{
"msg_contents": "> As for \"down sinks the mast\"...so far as I've been able to tell,\n> we have *never* removed anything that wasn't replaced by a better way \n> of doing it (time travel with triggers)...we didn't leave anyone \n> floundering for a solution to it, we provided it...but now, its not \n> forced on *everyone*, just on those databases/tables that require \n> it...\n> Same functionality, better performance...\n\nAh, he wasn't complaining, just saluting a retiring soldier :)\n",
"msg_date": "Mon, 23 Mar 1998 16:04:58 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "No, it wasn't me. Sorry.\n\nI believe it was [email protected] (Mattias Kregert). At least his\naddress is on one of the earlier replies.\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\n> -----Original Message-----\n> From:\tBruce Momjian [SMTP:[email protected]]\n> Sent:\tMonday, March 23, 1998 4:58 PM\n> To:\[email protected]\n> Cc:\[email protected]; [email protected]\n> Subject:\tRe: [HACKERS] Profile of current backend\n> \n> > \n> > Yes, but since it is, why don't we inline it? Is there a possible\n> > caveat?\n> \n> Inline it. I just haven't had time or studied that part of the code.\n> I\n> had hoped the original researcher would do the honors, but I don't\n> remember who that was, or if it was you.\n> \n> \n> \n> -- \n> Bruce Momjian | 830 Blythe Avenue\n> [email protected] | Drexel Hill, Pennsylvania\n> 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, 23 Mar 1998 16:53:58 +0100",
"msg_from": "\"Meskes, Michael\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] Profile of current backend"
}
] |
[
{
"msg_contents": "\nThe Hermit Hacker wrote:\n\n>> > 1. Regression tests make very extensive use of char16. Seems to\n>> > me that these should be changed to a name. Same for tutorial and\n>> > any other example references in the code. Seem ok?\n>> \n>> ** down sinks the mast of the old postquel, there, can you see it blink\n>> in the deep green water ? ** ** nostalgia on my face ** I think I forgot\n>> to mention my nicest hobby, sailing **\n>\n>\tOkay, this one has me...what benefit is there to char16 over\n>char(16)? Why, in the first place, was it implemented as char16 anyway?\n\nchar16 was there before varlena's got implemented, the basis for char(16).\nThere is no significant advantage of char16 over char(16) any more.\n\n\tAs for \"down sinks the mast\"...so far as I've been able to tell,\nwe have *never* removed anything that wasn't replaced by a better way of\ndoing it (time travel with triggers)...we didn't leave anyone floundering\nfor a solution to it, we provided it...but now, its not forced on\n*everyone*, just on those databases/tables that require it...\n\n\tSame functionality, better performance...\n\n** up raises the ulralight carbon ultra high speed mast, I like that streamline,\nof course for this the old wooden one had to go ;-) **\nThis was never intended as a critique, I think it is good that char2-16 goes.\nOf course I stick to my real message of this thread:\n\treplace the char16 stuff in the tutorial and the regression tests with char(16)\n\tor even varchar(16) but not with name.\n\nAndreas\n\n",
"msg_date": "Mon, 23 Mar 1998 17:30:42 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "On Mon, 23 Mar 1998, Zeugswetter Andreas wrote:\n\n> \tAs for \"down sinks the mast\"...so far as I've been able to tell,\n> we have *never* removed anything that wasn't replaced by a better way of\n> doing it (time travel with triggers)...we didn't leave anyone floundering\n> for a solution to it, we provided it...but now, its not forced on\n> *everyone*, just on those databases/tables that require it...\n> \n> \tSame functionality, better performance...\n> \n> ** up raises the ulralight carbon ultra high speed mast, I like that streamline,\n> of course for this the old wooden one had to go ;-) **\n\n\nOkay...just checking. I've heard one major complaint about PostgreSQL in\nthe past concerning the fact that we are just gutting it and \"taking the\nPostgres out of PostgreSQL\", which I feel is an unfair assessment...\n\nAs your analogy puts it above, we've replaced key components of the old\n\"wooden, three mast'r\" with a more sleek and streamlined \"carbon fiber\nmast\"...but that new mast holds the same sails, just gets us to the finish\nline faster...\n\nIf anyone feels that we've removed something that was really unique and\nnot replacedit with something better/faster, or provided a work around,\nplease let us know...in most cases, there is alot of discussion both here,\nand in a private list, over whether we go forward with a modification or\nnot, especially something that is part of the *base* structure of the\nsystem.\n\nWe have an *Open* development forum here...if you don't like the route\nsomething is (or appears to be) going, let us know...give us strong\narguements for/against...but don't wait until the discussion is done and\neveryone is in agreement before popping up...:(\n\n\n\n",
"msg_date": "Mon, 23 Mar 1998 11:59:37 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "> > > I have another question about GRANT/REVOKE:\n> > >\n> > > grant <privilege[,privilege,...]>\n> > > on <rel1>[,...<reln>]\n> > > to [public | GROUP <group> | <username>]\n> > > ^^^^^^^^^^^^^\n> > > I don't know how to create a GROUP ?\n> >\n> > I believe that you use \"CREATE USER groupname\", and then can assign\n> > privileges to that pseudo-user/group, and then add users to that \n> > group. Have you tried that?\n> postgres=> create user grupo;\n> CREATE USER\n> postgres=> grant all on tmp to grupo;\n> CHANGE\n> create user joe in group grupo;\n> CREATE USER\n> postgres=> grant select on tmp to group grupo;\n> ERROR: non-existent group \"grupo\"\n\nCan someone tell us how \"groups\" work? I'm not finding enough clues just\nby looking in the parser, and haven't stumbled across it in the docs...\n\n - Tom\n",
"msg_date": "Mon, 23 Mar 1998 16:45:39 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: PostgreSQL reference manual"
}
] |
[
{
"msg_contents": "> > > I have another question about GRANT/REVOKE:\n> > >\n> > > grant <privilege[,privilege,...]>\n> > > on <rel1>[,...<reln>]\n> > > to [public | GROUP <group> | <username>]\n> > > ^^^^^^^^^^^^^\n> > > I don't know how to create a GROUP ?\n> >\n> > I believe that you use \"CREATE USER groupname\", and then can assign\n> > privileges to that pseudo-user/group, and then add users to that \n> > group. Have you tried that?\n> postgres=> create user grupo;\n> CREATE USER\n> postgres=> grant all on tmp to grupo;\n> CHANGE\n> create user joe in group grupo;\n> CREATE USER\n> postgres=> grant select on tmp to group grupo;\n> ERROR: non-existent group \"grupo\"\n\nCan someone tell us how \"groups\" work? I'm not finding enough clues just\nby looking in the parser, and haven't stumbled across it in the docs...\n\nOnce a group is manually created with insert into pg_group values ...\nit can be used by the create user, alter user and grant to group syntax.\nA create group <groupname> is still missing in the grammar,\nit would simply need to do the insert. Groups are very helpful with \nmany users that all need similar grants. You grant rights to the group\nand then grant groups to users. The ANSI92 naming is \"role\",\nif we enhance the group stuff maybe it would be good to convert to the\n\"role\" naming of group code, althougth I like \"group\" more.\n\nAndreas\n\n\n\n\n",
"msg_date": "Mon, 23 Mar 1998 18:10:12 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": "Zeugswetter Andreas wrote:\n> \n> Once a group is manually created with insert into pg_group values ...\n> it can be used by the create user, alter user and grant to group syntax.\n> A create group <groupname> is still missing in the grammar,\n\nI'm working on this.. Should have something working around the end of\nthis week (for ALTER USER and CREATE USER).\n\n> it would simply need to do the insert. Groups are very helpful with\n> many users that all need similar grants. You grant rights to the group\n> and then grant groups to users. The ANSI92 naming is \"role\",\n\nOracle calls it \"role\" and Sybase calls it \"group\".\n\n-- \nStephane Lajeunesse.\nOracle and Sybase DBA\n",
"msg_date": "Mon, 23 Mar 1998 19:23:44 -0500",
"msg_from": "Stephane Lajeunesse <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": "Stephane Lajeunesse writes:\n> > A create group <groupname> is still missing in the grammar,\n> \n> I'm working on this.. Should have something working around the end of\n> this week (for ALTER USER and CREATE USER).\n\nPlease let me use this to tell you all that I would like to get notice of\neach change to gram.y. I am currently modelling ecpg's parser after gram.y\nto get good syntax checking. So I have to make these changes, 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",
"msg_date": "Tue, 24 Mar 1998 09:44:36 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": "On Mon, 23 Mar 1998, Zeugswetter Andreas wrote:\n\n> > > > I have another question about GRANT/REVOKE:\n> > > >\n> > > > grant <privilege[,privilege,...]>\n> > > > on <rel1>[,...<reln>]\n> > > > to [public | GROUP <group> | <username>]\n> > > > ^^^^^^^^^^^^^\n> > > > I don't know how to create a GROUP ?\n> > >\n> > > I believe that you use \"CREATE USER groupname\", and then can assign\n> > > privileges to that pseudo-user/group, and then add users to that \n> > > group. Have you tried that?\n> > postgres=> create user grupo;\n> > CREATE USER\n> > postgres=> grant all on tmp to grupo;\n> > CHANGE\n> > create user joe in group grupo;\n> > CREATE USER\n> > postgres=> grant select on tmp to group grupo;\n> > ERROR: non-existent group \"grupo\"\n> \n> Can someone tell us how \"groups\" work? I'm not finding enough clues just\n> by looking in the parser, and haven't stumbled across it in the docs...\n> \n> Once a group is manually created with insert into pg_group values ...\n> it can be used by the create user, alter user and grant to group syntax.\n> A create group <groupname> is still missing in the grammar,\n> it would simply need to do the insert. Groups are very helpful with \n> many users that all need similar grants. You grant rights to the group\n> and then grant groups to users. The ANSI92 naming is \"role\",\n> if we enhance the group stuff maybe it would be good to convert to the\n> \"role\" naming of group code, althougth I like \"group\" more.\n> \nI created a group as Andreas said but now psql \\z doesn't work anymore\nIt give me always a segmentation fault ... what's wrong...\n\npostgres=> INSERT INTO pg_group VALUES ('tutti',200);\nINSERT 318273 1\n\npostgres=> CREATE USER jose IN GROUP tutti;\nCREATE USER\nSELECT * FROM pg_group;\n\ngroname|grosysid|grolist\n-------+--------+-------\ntutti | 200|\n(1 row)\n \npostgres=> GRANT ALL ON temp TO GROUP tutti;\nCHANGE\n\npostgres=> \\z\n$ Segmentation fault\n jose'\n\n",
"msg_date": "Tue, 24 Mar 1998 11:11:48 +0000 (UTC)",
"msg_from": "\"Jose' Soares Da Silva\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": "> \n> Stephane Lajeunesse writes:\n> > > A create group <groupname> is still missing in the grammar,\n> > \n> > I'm working on this.. Should have something working around the end of\n> > this week (for ALTER USER and CREATE USER).\n> \n> Please let me use this to tell you all that I would like to get notice of\n> each change to gram.y. I am currently modelling ecpg's parser after gram.y\n> to get good syntax checking. So I have to make these changes, too.\n> \n\nGood idea on telling you of each change, but I also recommend that every\ntime you update the ecpg grammer, you save a copy the gram.y that you\nused to do it, so later when you need to get it back in sync, you can do\na diff on the old and new one to see each change so you don't miss any.\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, 24 Mar 1998 20:47:01 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": "Bruce Momjian writes:\n> > Stephane Lajeunesse writes:\n> > > > A create group <groupname> is still missing in the grammar,\n> > > \n> > > I'm working on this.. Should have something working around the end of\n> > > this week (for ALTER USER and CREATE USER).\n> > \n> > Please let me use this to tell you all that I would like to get notice of\n> > each change to gram.y. I am currently modelling ecpg's parser after gram.y\n> > to get good syntax checking. So I have to make these changes, too.\n> \n> Good idea on telling you of each change, but I also recommend that every\n> time you update the ecpg grammer, you save a copy the gram.y that you\n> used to do it, so later when you need to get it back in sync, you can do\n> a diff on the old and new one to see each change so you don't miss any.\n\nConsider also not updateing the grammar. The strength of PostgreSQL is that\nfunctions can be added to work inside the server. These functions can often\ndo whatever is being proposed as new syntax.\n\nSo, instead of cluttering up the grammar with non-standard SQLish stuff\nto handle things like groups, just create an administrative function to\ndo this job.\n\n* return create_group('groupname');\n* return add_user_to_group('groupname', 'username');\n* return drop_group('groupname');\n\nThese can be written in C, in SQL, or what ever far more quickly and with\nmuch less risk of destabilizing the system than the parser can be modified.\nIt also avoids making incompatibility with ecpg.\n\nAnd, in keeping with the recent anti-bloat thread, these can be loadable\nextensions, not part of the core. So if you don't use groups, you don't pay\nfor them.\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": "Tue, 24 Mar 1998 18:34:09 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": "Bruce Momjian writes:\n> Good idea on telling you of each change, but I also recommend that every\n> time you update the ecpg grammer, you save a copy the gram.y that you\n> used to do it, so later when you need to get it back in sync, you can do\n> a diff on the old and new one to see each change so you don't miss any.\n\nTom had the same idea. I wonder if you ever used that before. :-)\n\nAnyway, this is no doubt a very good idea. Thanks.\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, 25 Mar 1998 15:43:06 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": "David Gould writes:\n> Consider also not updateing the grammar. The strength of PostgreSQL is that\n> functions can be added to work inside the server. These functions can often\n> do whatever is being proposed as new syntax.\n\nSo you want me to not check the syntax while parsing the embedded SQL code?\n\n> So, instead of cluttering up the grammar with non-standard SQLish stuff\n> to handle things like groups, just create an administrative function to\n> do this job.\n> \n> * return create_group('groupname');\n> * return add_user_to_group('groupname', 'username');\n> * return drop_group('groupname');\n\nBut this is not embedded SQL.\n\n> These can be written in C, in SQL, or what ever far more quickly and with\n> much less risk of destabilizing the system than the parser can be modified.\n> It also avoids making incompatibility with ecpg.\n\nOkay, but I should at least check for ANSI SQL.\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, 25 Mar 1998 15:45:09 +0100 (CET)",
"msg_from": "Michael Meskes <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": "> \n> Bruce Momjian writes:\n> > Good idea on telling you of each change, but I also recommend that every\n> > time you update the ecpg grammer, you save a copy the gram.y that you\n> > used to do it, so later when you need to get it back in sync, you can do\n> > a diff on the old and new one to see each change so you don't miss any.\n> \n> Tom had the same idea. I wonder if you ever used that before. :-)\n> \n> Anyway, this is no doubt a very good idea. Thanks.\n> \n\nI believe I suggested it to the OpenLink people who mirror our libpq\nchanges, and to the person who keeps the Japanese version of the FAQ.\n\nReally the only way to do it.\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, 25 Mar 1998 09:52:19 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
},
{
"msg_contents": " > \n> David Gould writes:\n> > Consider also not updateing the grammar. The strength of PostgreSQL is that\n> > functions can be added to work inside the server. These functions can often\n> > do whatever is being proposed as new syntax.\n> \n> So you want me to not check the syntax while parsing the embedded SQL code?\n\nWhat I think we was suggesting is that we add non-ANSI functionality as\nfunction calls rather than grammer changes with keywords. The only\ndisadvantage is that it is a little more cumbersom, and less intuitive\nfor users.\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, 25 Mar 1998 09:54:54 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] Re: PostgreSQL reference manual"
}
] |
[
{
"msg_contents": "> >Hmmm. It does raise the question: where should they go? I'm guessing\n> >that they are holdovers from the PostQuel language, but don't know for\n> >sure. We should either document them or remove them from the parser.\n> \n> >How about writing a doc page for them, but basically _only_ show syntax\n> >and then say that they are equivalent to, for example, ROLLBACK, and\n> >refer to that doc for more info?\n\nI have changed all the references to the 'end' manual page to 'commit',\nbut left the 'end' manual page.\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, 23 Mar 1998 12:11:40 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] PostgreSQL Reference Guide"
},
{
"msg_contents": "On Mon, 23 Mar 1998, Thomas G. Lockhart wrote:\n>> Seems that rules doesn't work.\n>> There is an example how to create a rule to update a view on \"man create view\"\n>> --Specify deletion semantics for toyemp\n>> create rule example1 as\n>> on delete to toyemp\n>> do instead delete emp\n>> where emp.oid = current.oid\n>>\n>> Should I document CREATE RULE or ...?\n\n>How about posting a question to the docs list _and_ to the hackers list\n>saying you are writing docs on CREATE RULE and asking for a working\n>example. There are one or two people who have been working on this\n>recently, and they should be able to help.\n\nYou are rigth. Please there's somebody that can HELPPPPPPPPPPP me ???\n\n>> - Do you know what kind of privilege one may have on indexes or sequences ?\n\n>Indices - none, since they are just optimizations for tables.\n>Sequences - not sure. Ask on the lists?\n\n>> btw, how is it going?\n>>\n>> - Very well. We have already the following 12 pages:\n\n>Great! Are you writing with the SGML markup, or are we adding that\n>later?\n\nFor now I'm writing it in text mode, later I, or someone else can convert \nit to SGML. What do you think about ?\n\n>> I have a few questions about some statements:\n>>\n>> * ABORT is the same as COMMIT.\n>> * END is the same as ROLLBACK.\n\n>No, the opposite:\n>ABORT == ROLLBACK\n>END == COMMIT\n\nYes. I know.\n\n>> I suppose that ABORT and END are both deprecated feature and I'm thinking\n>> to skip this pages.\n\n>Hmmm. It does raise the question: where should they go? I'm guessing\n>that they are holdovers from the PostQuel language, but don't know for\n>sure. We should either document them or remove them from the parser.\n\n>How about writing a doc page for them, but basically _only_ show syntax\n>and then say that they are equivalent to, for example, ROLLBACK, and\n>refer to that doc for more info?\n\n - Tom\n\n",
"msg_date": "Mon, 23 Mar 1998 17:33:01 +0000 (UTC)",
"msg_from": "\"Jose' Soares Da Silva\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "PostgreSQL Reference Guide"
},
{
"msg_contents": "On Mon, 23 Mar 1998, Bruce Momjian wrote:\n\n> > >Hmmm. It does raise the question: where should they go? I'm guessing\n> > >that they are holdovers from the PostQuel language, but don't know for\n> > >sure. We should either document them or remove them from the parser.\n> > \n> > >How about writing a doc page for them, but basically _only_ show syntax\n> > >and then say that they are equivalent to, for example, ROLLBACK, and\n> > >refer to that doc for more info?\n> \n> I have changed all the references to the 'end' manual page to 'commit',\n> but left the 'end' manual page.\n> \n...and what about the non-standard ABORT statement this is also an alias\n to ROLLBACK. Do we need it ?\n Ciao, Jose'\n\n",
"msg_date": "Tue, 24 Mar 1998 11:23:34 +0000 (UTC)",
"msg_from": "\"Jose' Soares Da Silva\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [DOCS] Re: [HACKERS] PostgreSQL Reference Guide"
},
{
"msg_contents": "> > I have changed all the references to the 'end' manual page to 'commit',\n> > but left the 'end' manual page.\n> > \n> ...and what about the non-standard ABORT statement this is also an alias\n> to ROLLBACK. Do we need it ?\n\nDone.\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, 24 Mar 1998 20:53:28 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [DOCS] Re: [HACKERS] PostgreSQL Reference Guide"
}
] |
[
{
"msg_contents": "> > \n> > \n> > I noticed a couple of tcl/tk improvements from Bruce this morning, and\n> > fixed the geqo_paths.c problem...unless there are any major disagreements,\n> > I will announce this by 4:30ADT this afternoon as being the official\n> > 6.3.1...\n> > \n> \n> Good. Doing some manual pages cleanups, and now I see it is being\n> called 6.3.1 so I have to change the HISTORY, TODO, and other things.\n> \n\nAny chance of getting the updated machine-specific FAQs in? (It may only be\nthe Irix one that's changed). The Irix one certainly has lots more up-to-date\nand useful info on getting PostgreSQL to compile under Irix6\n\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": "Mon, 23 Mar 1998 17:27:01 GMT",
"msg_from": "Andrew Martin <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Final one...?"
},
{
"msg_contents": "> > Good. Doing some manual pages cleanups, and now I see it is being\n> > called 6.3.1 so I have to change the HISTORY, TODO, and other things.\n> > \n> \n> Any chance of getting the updated machine-specific FAQs in? (It may only be\n> the Irix one that's changed). The Irix one certainly has lots more up-to-date\n> and useful info on getting PostgreSQL to compile under Irix6\n\nDone. Didn't expect any change in those in two weeks.\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, 23 Mar 1998 12:53:34 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Final one...?"
}
] |
[
{
"msg_contents": "You seem to be getting me wrong, right from the start. I appreciate your work on removing \nchar2-char16. I also think it is the correct step. You have the hackers behind you, the discussion was\nabout 2 - 3 weeks ago. I was part of it. There are three things that will need documentation:\n\t1. the replacement for char16 is char(16)\n\t2. char(16) gives and ignores trailing blanks\n\t3. refer to VARDATA(someparameter) to get your (char *) in a C function\n\nThomas wrote:\n> > Same functionality, better performance...\n> Ah, he wasn't complaining, just saluting a retiring soldier :)\n\nThis description makes me happy. It is exactly what I wanted to express.\n\nGosh, I sure did not want to nag, sorry if that is how you interpreted it.\n \nAndreas \n\n\n",
"msg_date": "Mon, 23 Mar 1998 18:34:18 +-100",
"msg_from": "Zeugswetter Andreas <[email protected]>",
"msg_from_op": true,
"msg_subject": "AW: [HACKERS] char types gone."
},
{
"msg_contents": "> \n> You seem to be getting me wrong, right from the start. I appreciate your work on removing \n> char2-char16. I also think it is the correct step. You have the hackers behind you, the discussion was\n> about 2 - 3 weeks ago. I was part of it. There are three things that will need documentation:\n> \t1. the replacement for char16 is char(16)\n> \t2. char(16) gives and ignores trailing blanks\n\nI don't think this is true.\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, 23 Mar 1998 13:00:49 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] char types gone."
},
{
"msg_contents": "On Mon, 23 Mar 1998, Zeugswetter Andreas wrote:\n\n> You seem to be getting me wrong, right from the start. I appreciate your work on removing \n> char2-char16. I also think it is the correct step. You have the hackers behind you, the discussion was\n> about 2 - 3 weeks ago. I was part of it. There are three things that will need documentation:\n> \t1. the replacement for char16 is char(16)\n> \t2. char(16) gives and ignores trailing blanks\n> \t3. refer to VARDATA(someparameter) to get your (char *) in a C function\n> \n> Thomas wrote:\n> > > Same functionality, better performance...\n> > Ah, he wasn't complaining, just saluting a retiring soldier :)\n> \n> This description makes me happy. It is exactly what I wanted to express.\n> \n> Gosh, I sure did not want to nag, sorry if that is how you interpreted it.\n\n\t*grin* No, I understood that from the second one you sent, I just\nwant to make sure that anyone out there that might feel we've over-gut'd\nisn't afraid to speak up...\n\n\t*both feet buried in mouth*? :)\n\n\n",
"msg_date": "Mon, 23 Mar 1998 13:04:51 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: AW: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "\nThe patch to configure contains the following:\n\n*** 3316,3322 ****\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:3322: checking whether ${CC-cc} needs -traditional\" >&5\n if eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\"; then\n--- 3316,3322 ----\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:3322: checking whether ${CC-cc} needs -traditional\" >&5\n if eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\"; then\n\n\nI am worried that this has undone the patch which was required to get \nconfigure to run properly with anything other than GCC. (I haven't\nactually tried 6.3.1, but this change affects that area of the code and\ndoesn't look healthy...)\n\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": "Mon, 23 Mar 1998 17:43:41 GMT",
"msg_from": "Andrew Martin <[email protected]>",
"msg_from_op": true,
"msg_subject": "URGENT PROBLEM with the 6.3->6.3.1 patch"
},
{
"msg_contents": "Marc, can you comment on this.\n\n> \n> \n> The patch to configure contains the following:\n> \n> *** 3316,3322 ****\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:3322: checking whether ${CC-cc} needs -traditional\" >&5\n> if eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\"; then\n> --- 3316,3322 ----\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:3322: checking whether ${CC-cc} needs -traditional\" >&5\n> if eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\"; then\n> \n> \n> I am worried that this has undone the patch which was required to get \n> configure to run properly with anything other than GCC. (I haven't\n> actually tried 6.3.1, but this change affects that area of the code and\n> doesn't look healthy...)\n> \n> \n> Andrew\n> \n> ----------------------------------------------------------------------------\n> Dr. Andrew C.R. Martin University College London\n> EMAIL: (Work) [email protected] (Home) [email protected]\n> URL: http://www.biochem.ucl.ac.uk/~martin\n> Tel: (Work) +44(0)171 419 3890 (Home) +44(0)1372 275775\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": "Mon, 23 Mar 1998 12:54:52 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] URGENT PROBLEM with the 6.3->6.3.1 patch"
},
{
"msg_contents": "On Mon, 23 Mar 1998, Bruce Momjian wrote:\n\n> Marc, can you comment on this.\n\n\tIt kind of bothers me that something that is standard in autoconf\nbreaks on some systems, while, as far as I can tell, its always been this\nway ;(\n\n\tAndrew, have you asked the developers of autoconf about this?\n\n\n> > The patch to configure contains the following:\n> > \n> > *** 3316,3322 ****\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:3322: checking whether ${CC-cc} needs -traditional\" >&5\n> > if eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\"; then\n> > --- 3316,3322 ----\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:3322: checking whether ${CC-cc} needs -traditional\" >&5\n> > if eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\"; then\n> > \n> > \n> > I am worried that this has undone the patch which was required to get \n> > configure to run properly with anything other than GCC. (I haven't\n> > actually tried 6.3.1, but this change affects that area of the code and\n> > doesn't look healthy...)\n> > \n> > \n> > Andrew\n> > \n> > ----------------------------------------------------------------------------\n> > Dr. Andrew C.R. Martin University College London\n> > EMAIL: (Work) [email protected] (Home) [email protected]\n> > URL: http://www.biochem.ucl.ac.uk/~martin\n> > Tel: (Work) +44(0)171 419 3890 (Home) +44(0)1372 275775\n> > \n> > \n> \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\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Mon, 23 Mar 1998 17:17:08 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] URGENT PROBLEM with the 6.3->6.3.1 patch"
}
] |
[
{
"msg_contents": ">>\n>> Wow, this sounds really good, will you try to prealloc bigger chunks of\nmemory\n>> to avoid the many system calls ? I was wondering if that could give a\nspeed boost.\n>\n>Yes to both points.\n>\n>-dg\n>\n>David Gould [email protected] 510.628.3783 or\n510.305.9468\n>Informix Software (No, really) 300 Lakeside Drive Oakland, CA\n94612\n> - Linux. Not because it is free. Because it is better.\n>\n\nGreat, I'm looking forward to this.\n\nWith regards from Maurice.\n\n",
"msg_date": "Mon, 23 Mar 1998 20:24:38 +0100",
"msg_from": "\"Maurice Gittens\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: AW: [HACKERS] patch for memory overrun on Linux(i386)"
}
] |
[
{
"msg_contents": "With regards to the c++ interface (libpq++),\n\nWhat types of changes should take place? What has changed with\nPostgreSQL since the lib was last worked on? How does its current\nversion relate to the current version of libpq?\n\n-b\n\n --\n================================================================================\nbrian paul rickabaugh\nroanoke college student\ncomputer information systems major\ncomputer science / theatre arts minor\n________________________________________________________________________________\nemail -- [email protected]\nwww -- http://peanuts.roanoke.edu/~brickaba\n\n \"if all the world's a stage, the sun is the ultimate in lighting\ninstruments\"\n================================================================================\n\n\n\n",
"msg_date": "Mon, 23 Mar 1998 14:50:56 -0500",
"msg_from": "brian paul rickabaugh <[email protected]>",
"msg_from_op": true,
"msg_subject": "C++ Interface"
}
] |
[
{
"msg_contents": "> > Have managed to delete char2,4,8 and 16.\n> > Three points now...regression tests, docs and interfaces.\n> > \n> > 1. Regression tests make very extensive use of char16. Seems to\n> > me that these should be changed to a name. Same for tutorial and\n> > any other example references in the code. Seem ok?\n> \n> Any reason not to use \"text\" for these cases? Or even varchar()? \"name\"\n> is probably really only an internal Postgres type. We should have a\n> regression test for it if we don't already. Let me know if you want me\n> to help...\n\nI think the 'name' type should be considered an internal type only and\nmentioned as such wherever it is documented. It's only purpose seems\nto be for a consistent length to internal names and it does that rather\nwell. User's would be better served with the other, varlena-based, types.\n\nAny examples I was going to convert from char16 to char(16). There are a\ncouple of mentions of the char16 struct that will have to be integrated\nor changed to reference the varlena struct.\n\nFinally, I'm redoing the char16 regression test for name.\n\n> > 2. I'll take my best shot at the docs.\n> \n> Where? The man pages or the sgml sources? I'm shooting for getting the\n> sgml sources right, and obsoleting the man page _sources_, regenerating\n> them from the sgml instead.\n\nDo you want patches against doc/src/sgml/*? The man pages references were\nfew, so they were easy to change for consistency sake then til they drop\ninto the bit bucket. Changed a few references in the userguide too.\n\nOnly have to tidy up the docs and I'll submit the patch.\n\ndarrenk\n",
"msg_date": "Mon, 23 Mar 1998 15:39:11 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "\n\n\n\nBruce,\n\nI will try and start taking a look at the libpq++ over the weekend to\nfamilarize myself with\nthe way that it is structured. Then, if someone will promise to help guide\nme through the\nchanges required, I will see if I can work on them.\n\nThanks,\n\nMatt\n\n\n\n\[email protected] on 03/23/98 01:39:55 PM\n\nTo: [email protected]\ncc: [email protected] (bcc: Matt Aycock/IntelliQuest)\nSubject: Re: [QUESTIONS] Re: [HACKERS] C++ interface\n\n\n\n\n>\n> howdy,\n>\n> we're using the libpq++ c++ interface for a medium-sized network,\n> multi-user scheduling application that uses a java front end, a c++\n> middleware server and the postgres backend. the c++ interface (a matter\n> of preference, obvious) is much cleaner than the c interface (especially\n> for True Believers in OO), but it does seem to be suffering from lack of\n> update. it works fine for basic stuff with 6.3, though.\n\nI agree it needs updating, but I didn't have the time to do 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--\nOfficial WWW Site: http://www.postgresql.org\nOnline Docs & FAQ: http://www.postgresql.org/docs\nSearchable Lists: http://www.postgresql.org/mhonarc\n\n\n\n\n\n",
"msg_date": "Mon, 23 Mar 1998 17:00:03 -0400",
"msg_from": "[email protected]",
"msg_from_op": true,
"msg_subject": "Re: [QUESTIONS] Re: [HACKERS] C++ interface"
}
] |
[
{
"msg_contents": "> Okay...just checking. I've heard one major complaint about PostgreSQL in\n> the past concerning the fact that we are just gutting it and \"taking the\n> Postgres out of PostgreSQL\", which I feel is an unfair assessment...\n> \n> As your analogy puts it above, we've replaced key components of the old\n> \"wooden, three mast'r\" with a more sleek and streamlined \"carbon fiber\n> mast\"...but that new mast holds the same sails, just gets us to the finish\n> line faster...\n> \n> If anyone feels that we've removed something that was really unique and\n> not replacedit with something better/faster, or provided a work around,\n> please let us know...in most cases, there is alot of discussion both here,\n> and in a private list, over whether we go forward with a modification or\n> not, especially something that is part of the *base* structure of the\n> system.\n> \n> We have an *Open* development forum here...if you don't like the route\n> something is (or appears to be) going, let us know...give us strong\n> arguements for/against...but don't wait until the discussion is done and\n> everyone is in agreement before popping up...:(\n\nIf for posterity alone, I'm willing to move the char[24816] code to contrib.\nI already have the create and load statements in an sql file, just have to\npolish the Makefile a little. If not with the distribution, then perhaps\nthis could be at the web site for downloading.\n\nHaving the code available for loading might be easier for persons such as\nTatsuo instead of having their users complain of something disappearing.\n\nMy only intention in doing this is to lighten up the postgres core, not to\nforce others to change their existing code.\n\ndarrenk\n",
"msg_date": "Mon, 23 Mar 1998 16:17:28 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "On Mon, 23 Mar 1998, Darren King wrote:\n\n> If for posterity alone, I'm willing to move the char[24816] code to contrib.\n> I already have the create and load statements in an sql file, just have to\n> polish the Makefile a little. If not with the distribution, then perhaps\n> this could be at the web site for downloading.\n> \n> Having the code available for loading might be easier for persons such as\n> Tatsuo instead of having their users complain of something disappearing.\n> \n> My only intention in doing this is to lighten up the postgres core, not to\n> force others to change their existing code.\n\n\tUnless there is a proven performance boost to removing things from\nthe core, which, at this point in time, hasn't been demonstrated,\n*nothing* should be removed from the core...\n\n\t...in fact, I want to bring in the IP code from contrib as part of\nthe core yet...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Mon, 23 Mar 1998 18:37:29 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "bruce,\n\ni too am interested in possibly doing a bit of work on libpq++. i'm\ncurrently working on a project that will be using at very least libpq\n(probably libpq++) and i may be able to dedicate some time to working\non the lib development.\n\n-brian\n\n--\n================================================================================\nbrian paul rickabaugh\nroanoke college student\ncomputer information systems major\ncomputer science / theatre arts minor\n________________________________________________________________________________\nemail -- [email protected]\nwww -- http://peanuts.roanoke.edu/~brickaba\n\n \"if all the world's a stage, the sun is the ultimate in lighting instruments\"\n================================================================================\n\n\n\n",
"msg_date": "Mon, 23 Mar 1998 17:35:46 -0500",
"msg_from": "brian paul rickabaugh <[email protected]>",
"msg_from_op": true,
"msg_subject": "c++ interface"
}
] |
[
{
"msg_contents": "> > If for posterity alone, I'm willing to move the char[24816] code to contrib.\n> > I already have the create and load statements in an sql file, just have to\n> > polish the Makefile a little. If not with the distribution, then perhaps\n> > this could be at the web site for downloading.\n> > \n> > Having the code available for loading might be easier for persons such as\n> > Tatsuo instead of having their users complain of something disappearing.\n> > \n> > My only intention in doing this is to lighten up the postgres core, not to\n> > force others to change their existing code.\n> \n> \tUnless there is a proven performance boost to removing things from\n> the core, which, at this point in time, hasn't been demonstrated,\n> *nothing* should be removed from the core...\n> \n> \t...in fact, I want to bring in the IP code from contrib as part of\n> the core yet...\n\nWould be nice if stuff like that (and the money and geometrics) could be\ncompiled in in the same manner that Apache is a basic web server that lets\nyou pick and choose modules/features you want to build into it.\n\nI don't know how big of a performance boost it provides in the cache, but\nremoving the functions associated with the char types shrank the pg_proc\ntable from 906 to 842 entries or a bit over 7%.\n\nWant to shrink it further? Of those remaining 842, _230_ are for the geometric\ntypes! Throw in 25 more for the cash/money functions. Bloat city if you\nnever use these things. Thirty percent could be moved out to contrib and\nnot missed by most postgres users.\n\nI think a reasonable goal is to provide an sql server that meets the current sql\nstandard with the possibility of having other types and functionality available\nas modules. These modules could be made into the template1 at compile time\nor made and installed on a per-database basis. Make your server and database\nwith whatever you want in it.\n\nWhy force others to have functionality 'x', 'y' and 'z' when they only want 'y'?\n\ndarrenk\n",
"msg_date": "Mon, 23 Mar 1998 18:14:58 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "> Would be nice if stuff like that (and the money and geometrics) could be\n> compiled in in the same manner that Apache is a basic web server that lets\n> you pick and choose modules/features you want to build into it.\n> \n> I don't know how big of a performance boost it provides in the cache, but\n> removing the functions associated with the char types shrank the pg_proc\n> table from 906 to 842 entries or a bit over 7%.\n> \n> Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> types! Throw in 25 more for the cash/money functions. Bloat city if you\n> never use these things. Thirty percent could be moved out to contrib and\n> not missed by most postgres users.\n> \n\nYes, but if they are never referenced, the cache is empty for those\ntypes. Unless there is some performance change with their removal, why\nremove them? Disk space of binary?\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, 23 Mar 1998 22:27:50 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "On Mon, 23 Mar 1998, Bruce Momjian wrote:\n\n> > Would be nice if stuff like that (and the money and geometrics) could be\n> > compiled in in the same manner that Apache is a basic web server that lets\n> > you pick and choose modules/features you want to build into it.\n> > \n> > I don't know how big of a performance boost it provides in the cache, but\n> > removing the functions associated with the char types shrank the pg_proc\n> > table from 906 to 842 entries or a bit over 7%.\n> > \n> > Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> > types! Throw in 25 more for the cash/money functions. Bloat city if you\n> > never use these things. Thirty percent could be moved out to contrib and\n> > not missed by most postgres users.\n> > \n> \n> Yes, but if they are never referenced, the cache is empty for those\n> types. Unless there is some performance change with their removal, why\n> remove them? Disk space of binary?\n\n\tI have to agree here...unless a noticeable performance improvement\ncan be demonstrated, I really don't like a trend that moves towards\nremoving core features that make us unique, just cause it makes a table\njust a little smaller...*shrug*\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Mon, 23 Mar 1998 23:45:01 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "On Mon, 23 Mar 1998, The Hermit Hacker wrote:\n\n> On Mon, 23 Mar 1998, Bruce Momjian wrote:\n> \n> > > Would be nice if stuff like that (and the money and geometrics) could be\n> > > compiled in in the same manner that Apache is a basic web server that lets\n> > > you pick and choose modules/features you want to build into it.\n> > > \n> > > I don't know how big of a performance boost it provides in the cache, but\n> > > removing the functions associated with the char types shrank the pg_proc\n> > > table from 906 to 842 entries or a bit over 7%.\n> > > \n> > > Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> > > types! Throw in 25 more for the cash/money functions. Bloat city if you\n> > > never use these things. Thirty percent could be moved out to contrib and\n> > > not missed by most postgres users.\n> > > \n> > \n> > Yes, but if they are never referenced, the cache is empty for those\n> > types. Unless there is some performance change with their removal, why\n> > remove them? Disk space of binary?\n> \n> \tI have to agree here...unless a noticeable performance improvement\n> can be demonstrated, I really don't like a trend that moves towards\n> removing core features that make us unique, just cause it makes a table\n> just a little smaller...*shrug*\n\nI'm also thinking about JDBC. If you remove some features out of the core,\nand reorganise what OID's are associated with the types, then we would\nhave a big performance hit when the driver starts up, and possibly make\nthe driver even larger than it is now.\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, 24 Mar 1998 06:58:36 +0000 (GMT)",
"msg_from": "Peter T Mount <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "On Tue, 24 Mar 1998, Peter T Mount wrote:\n\n> I'm also thinking about JDBC. If you remove some features out of the core,\n> and reorganise what OID's are associated with the types, then we would\n> have a big performance hit when the driver starts up, and possibly make\n> the driver even larger than it is now.\n\n\tFrom discussions that the core have had on this topic over the\npast couple of weeks, I think it will take *alot* to convince us to remove\ncore features.\n\n\tchar2/4/8/16 disappears *only* because they were redundant...it\nhad nothing to do with reducing the sizes of tables...char2 maps to\nchar(2), so why have both. Stuff like the geometric types aren't\nredundant...they are themselves the \"base\" types...\n\n\tOnce v6.3.1 is finished up, I'm going to be looking at merging in\nthe ip_and_mac contrib stuff, and I think that Thomas is considering\nmoving the 64bit stuff into the core from contrib...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 24 Mar 1998 03:12:02 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "I think Darren brings up very good points here. This seems to be the model\nthat many people are going to when developing large systems. This would\nalso help in maintaning the the system.\nWayne\n\nOn Mon, 23 Mar 1998, Darren King wrote:\n\n> > \t...in fact, I want to bring in the IP code from contrib as part of\n> > the core yet...\n> \n> Would be nice if stuff like that (and the money and geometrics) could be\n> compiled in in the same manner that Apache is a basic web server that lets\n> you pick and choose modules/features you want to build into it.\n> \n> I don't know how big of a performance boost it provides in the cache, but\n> removing the functions associated with the char types shrank the pg_proc\n> table from 906 to 842 entries or a bit over 7%.\n> \n> Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> types! Throw in 25 more for the cash/money functions. Bloat city if you\n> never use these things. Thirty percent could be moved out to contrib and\n> not missed by most postgres users.\n> \n> I think a reasonable goal is to provide an sql server that meets the current sql\n> standard with the possibility of having other types and functionality available\n> as modules. These modules could be made into the template1 at compile time\n> or made and installed on a per-database basis. Make your server and database\n> with whatever you want in it.\n> \n> Why force others to have functionality 'x', 'y' and 'z' when they only want 'y'?\n> \n> darrenk\n> \n\n",
"msg_date": "Tue, 24 Mar 1998 06:01:04 -0500 (EST)",
"msg_from": "wward <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "Darren King wrote:\n> \n> Would be nice if stuff like that (and the money and geometrics) could be\n> compiled in in the same manner that Apache is a basic web server that lets\n> you pick and choose modules/features you want to build into it.\n> \n> I don't know how big of a performance boost it provides in the cache, but\n> removing the functions associated with the char types shrank the pg_proc\n> table from 906 to 842 entries or a bit over 7%.\n> \n> Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> types! Throw in 25 more for the cash/money functions. Bloat city if you\n> never use these things. Thirty percent could be moved out to contrib and\n> not missed by most postgres users.\n> \n> I think a reasonable goal is to provide an sql server that meets the current sql\n> standard with the possibility of having other types and functionality available\n> as modules. These modules could be made into the template1 at compile time\n> or made and installed on a per-database basis. Make your server and database\n> with whatever you want in it.\n\nThe problem with the contrib stuff is that most users find it too\ndifficult to dig around in that dir and make and install stuff. The\n\"contrib\" name also give me the feeling that this is unsupported\nstuff/beta/not-to-be-used-in-production/to-be-removed etc.\n\nThe nicest thing would be if contrib was changed to \"modules\", and\neverything was compiled and installed in /usr/local/pgsql/modules/.\nThen the user could install them by typing \"INSMOD ORACLE\" and then\n/usr/local/pgsql/modules/oracle/insmod.sql would be executed and\ninstall shared libraries and sql commands. \"RMMOD ORACLE\" would\ndo the opposite. Just like Linux modules.\nPerhaps it would even be possible to autoload modules if the modules\ndirectories contains a file with commands/functions listed, which\nwould be searched automagically...\n\nThis is the idea:\n\npsql-> SET MODULES TO AUTOLOAD;\nNOTICE: 100 Modules will now load automagically.\npsql-> SELECT MYFUNKYFUNC('hello');\nNOTICE: 202 Unknown function 'MYFUNKYFUNC(text)'\nNOTICE: 100 Module 'MYFUNCTIONS' loaded.\nmyfunkyfunc\n-------------------\nthe word is 'hello'\n\n(1 row)\n\npsql-> INSMOD ORACLE;\nNOTICE: 100 Module 'ORACLE' loaded.\nNOTICE: 110 Module description: Oracle(tm) compatability module. Version\n1.0.\npsql-> LSMOD;\nNOTICE: 100 MYFUNCTIONS ORACLE\npsql-> RMMOD VERBOSE MYFUNCTIONS;\nNOTICE: 100 Module 'MYFUNCTIONS' removed.\nNOTICE: 120 Functions removed: Myfunkyfunc(text) Myotherfunc(int)\nMylastfunc(float)\nNOTICE: 120 Commands removed: Myfunkycommand\n\n\nIs it possible to register commands dynamically in the parser?\n\n/* m */\n",
"msg_date": "Tue, 24 Mar 1998 15:35:31 +0100",
"msg_from": "Mattias Kregert <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "Mattias Kregert wrote:\n> \n> Darren King wrote:\n> > \n> > Would be nice if stuff like that (and the money and geometrics) could be\n> > compiled in in the same manner that Apache is a basic web server that lets\n> > you pick and choose modules/features you want to build into it.\n> > \n> > I don't know how big of a performance boost it provides in the cache, but\n> > removing the functions associated with the char types shrank the pg_proc\n> > table from 906 to 842 entries or a bit over 7%.\n> > \n> > Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> > types! Throw in 25 more for the cash/money functions. Bloat city if you\n> > never use these things. Thirty percent could be moved out to contrib and\n> > not missed by most postgres users.\n> > \n> > I think a reasonable goal is to provide an sql server that meets the current sql\n> > standard with the possibility of having other types and functionality available\n> > as modules. These modules could be made into the template1 at compile time\n> > or made and installed on a per-database basis. Make your server and database\n> > with whatever you want in it.\n> \n> The problem with the contrib stuff is that most users find it too\n> difficult to dig around in that dir and make and install stuff. The\n> \"contrib\" name also give me the feeling that this is unsupported\n> stuff/beta/not-to-be-used-in-production/to-be-removed etc.\n> \n> The nicest thing would be if contrib was changed to \"modules\", and\n> everything was compiled and installed in /usr/local/pgsql/modules/.\n> Then the user could install them by typing \"INSMOD ORACLE\" and then\n> /usr/local/pgsql/modules/oracle/insmod.sql would be executed and\n> install shared libraries and sql commands. \"RMMOD ORACLE\" would\n> do the opposite. Just like Linux modules.\n\nThe difference here is that the entire Linux kernel (modules,\n\"standard\" code, etc)must be memory resident at all times. This is\nnot the case with the majority of unix systems out there. Most\nsystems have copy on write pages, filesystem caching and memory\npaging. The combined effect of these is that unused features do not\nstay resident in memory, they are sent to the swap, or even discarded\nand read in from the executable.\n\nSo this said, I believe there are things that can be done to help this\nprocess along. If every feature is compiled to its respective object\nfiles and then these files are linked together into one executable, a\npage containing a desired feature might also contain unneeded code.\nIf instead, the different features are compiled to libraries and then\nthese libraries are linked to the \"main\" application to create the\nexecutable, then each feature will start on a fresh page, which can\nhelp keep the minimum number of pages resident, even though keeping\nthe whole program resident now takes more pages.\n\nOcie Mitchell\n",
"msg_date": "Tue, 24 Mar 1998 11:01:56 -0800 (PST)",
"msg_from": "[email protected]",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "> > cvsup is updated now.\n> \n> Almost exactly the same result: The library will not install in ~pgsql/lib.\n> I installed by hand; cd ~/src/pgsql/src/interfaces/libpgtcl;make install.\n> I specified in configure --with-tcl, which to my narrow mind, heavily\n> influenced with the 6.2 behavior, it sounds like a bug.\n\nOK, did you remove config.cache before running configure, and supply the\ninclude and library directories when prompted by configure, and do you\nhave tcl.h, tk.h, libtcl.[a,so] and libtk.[a,so] in one of those\ndirectories. If you do all those things, it should find it. You may\nwant to add a set -x to configure just before the tcl/tk probing to see\nwhat it is doing.\n\n> > I think that is all it takes.\n> \n> Nope. I also did (as root :-), ldconfig -m ~pgsql/lib, and added\n> ~pgsql/lib to LD_LIBRARY_PATH.\n\nUnder BSDI, we don't have dynamic shared libraries for performance\nreasons, but have a more SVr3 shared library that is created using a\nspecial shlicc command. You specify that result library on the link\nline, and it works.\n\nThe shame of it is that every OS seems to have different way to load\ndynamic libraries, and with/pgaccess need those types of libraries, so\nthere is no boilerplate solution to this.\n\n> > The executable is src/bin/pgaccess/pgaccess.tcl. I have to find the\n> > libpqtcl.so line, and edit it.\n> \n> Not so simple. first, I added both known Unix methods of broadcasting a\n> library (LD_LIBRARY_PATH, and ldconfig), to no avail.\n> Then, I edit pgaccess.tcl, to have the library explicitly specified.\n> \n> Now I get:\n> \n> $ /usr/local/bin/wish8.1 src/pgsql/src/bin/pgaccess/pgaccess.tcl\n> Error in startup script: couldn't load file\n> \"/usr/local/pgsql/lib/libpgtcl.so\": Undefined symbol \"_PQnotifies\" in\n> wish8.1:/usr/local/pgsql/lib/libpgtcl.so\n> while executing\n> \"load /usr/local/pgsql/lib/libpgtcl.so\"\n> (procedure \"main\" line 3)\n> invoked from within\n> \"main $argc $argv\n> \"\n\nIt actually looks like it found libpgtcl.so, but not cannot load libpq,\nwhich is referenced by libpgtcl.so. I recommend you try that and see if\nit works. I don't do that step here, but it sounds like you may need to\ndo it.\n\n> This holds true for wish 8.1, 8.0, 4.2, 4.1\n> \n> This is why I love TCL so much. The concept of backwards compatability, or\n> inter-port compatability has not been discovered yet. I have a customer\n> who needs/likes pgaccess. Any help will be appreciated.\n\nYea, this tcl/tk version thing is pretty pathetic.\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, 23 Mar 1998 18:26:45 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [PORTS] RE: pgaccess doesn't run on -current anymore, possib"
},
{
"msg_contents": "\nOn 23-Mar-98 Bruce Momjian wrote:\n>> > cvsup is updated now.\n>> \n>> Almost exactly the same result: The library will not install in\n>> ~pgsql/lib.\n>> I installed by hand; cd ~/src/pgsql/src/interfaces/libpgtcl;make\n>> install.\n>> I specified in configure --with-tcl, which to my narrow mind, heavily\n>> influenced with the 6.2 behavior, it sounds like a bug.\n> \n> OK, did you remove config.cache before running configure, and supply the\n> include and library directories when prompted by configure, and do you\n> have tcl.h, tk.h, libtcl.[a,so] and libtk.[a,so] in one of those\n> directories. If you do all those things, it should find it. You may\n> want to add a set -x to configure just before the tcl/tk probing to see\n> what it is doing.\n\nI did rm -rf ~pgsql/src/pgsql which should be almost as good.\nconfig did not complain about anything missing. I have now explicitly\nincluded the paths (the problem is most likely FreeBSD removing things like\n/usr/local/include/{tcl,tk}.h). Configure did NOT complain loudly\nenough about not finding things. IMHO, if I specify --with-tcl and tcl.h\nis not found, this should be a fatal error in configure, or at least in\nmake all, or in make install.\n\n\n>> > I think that is all it takes.\n>> \n>> Nope. I also did (as root :-), ldconfig -m ~pgsql/lib, and added\n>> ~pgsql/lib to LD_LIBRARY_PATH.\n> \n> Under BSDI, we don't have dynamic shared libraries for performance\n> reasons, but have a more SVr3 shared library that is created using a\n> special shlicc command. You specify that result library on the link\n> line, and it works.\n> \n> The shame of it is that every OS seems to have different way to load\n> dynamic libraries, and with/pgaccess need those types of libraries, so\n> there is no boilerplate solution to this.\n\nSomehow, 6.2.1 builds and compiles correctly, so something can be done.\nThe only thing I worry about is repeatability; The next build will go all\nover this mess again, unless we document it here. I am naive enough to\nbelive that Makefiles are intended for this purpose :-)\n\n ...\n\n> It actually looks like it found libpgtcl.so, but not cannot load libpq,\n> which is referenced by libpgtcl.so. I recommend you try that and see if\n> it works. I don't do that step here, but it sounds like you may need to\n> do it.\n\nThis is odd, as all the other executables find the library. Postgres works.\n\nI modified pgaccess.tcl to load the library explicitly. Then it fails on\n_PQnotifies. I add an explicit load for libpq. and it fails on _crypt. \nAll this with wish{4.1,4.2,8.0,8.1}.\nSo, smart simon tries pgtclsh, that simply gives me a % prompt.\nOK, how about pgtksh? \n\nError in startup script: couldn't find procedure Pq_Init\n while executing\n\"load /usr/local/pgsql/lib/libpq.so\"\n (procedure \"main\" line 3)\n invoked from within\n\"main $argc $argv\n\"\n (file \"bin/pgaccess/pgaccess.tcl\" line 4506)\n\nI'd say something is sick here.\n\n\n> Yea, this tcl/tk version thing is pretty pathetic.\n\nI am relived to hear this is not the ignorant observation of the clueless\n(me).\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\nSincerely Yours, \n\nSimon Shapiro\[email protected] Voice: 503.799.2313\n",
"msg_date": "Mon, 23 Mar 1998 16:45:44 -0800 (PST)",
"msg_from": "Simon Shapiro <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [PORTS] RE: pgaccess doesn't run on -current anymore, possib"
}
] |
[
{
"msg_contents": "Hello,\n\nHave foreign keys been implemented yet?\n\n-brian rickabaugh\n\n",
"msg_date": "Mon, 23 Mar 1998 23:04:04 -0500",
"msg_from": "\"PostgreSQL On Peanuts.roanoke.edu\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Foreign Keys"
}
] |
[
{
"msg_contents": "> I've seen this myself on Solaris. If your default value as specified\n> in the CREATE TABLE statement is not exactly as long as the declared\n> width of the field, the backend dies with an assertion as soon as you\n> INSERT a record that needs to use the default.\n\nHmm. I see this on Linux also. It is only a problem for the char() type;\nboth varchar() and text behave properly.\n\nYou might want to post something to the bugs list...\n\n - Tom\n",
"msg_date": "Tue, 24 Mar 1998 06:12:22 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [QUESTIONS] backend dying on insert on 6.3, not dying on 6.2.1 p7"
},
{
"msg_contents": "> I've seen this myself on Solaris. If your default value as specified\n> in the CREATE TABLE statement is not exactly as long as the declared\n> width of the field, the backend dies with an assertion as soon as you\n> INSERT a record that needs to use the default.\n> \n> CREATE TABLE foo\n> (\n> i INT,\n> a CHAR(4) DEFAULT 'x'\n> );\n> \n> INSERT INTO a (i) VALUES (23);\n> \n> Will blow up the backend.\n> \n> CREATE TABLE foo\n> (\n> i INT,\n> a CHAR(4) DEFAULT 'x '\n> );\n> \n\nThis does not surprise me entirely, especially when char() fails but\nvarchar() and text work. It has to do with the way we handle char(). \nCan someone familiar with default handling checking into this.\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, 24 Mar 1998 20:28:49 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [QUESTIONS] backend dying on insert on 6.3, not dying on 6.2.1 p7"
}
] |
[
{
"msg_contents": " I, for one, am in favor of converting some of the type packages to\nloadable modules. Having those in the backend when they aren't being\nused is much like compiling extra modules into my Apache web server\nbecause they're \"kinda neat\", even though they won't be used. Also,\nif we follow the idea that we should have as many unique features in\nthe backend as possible, we could end up with all sorts of features\nthat are only used by a subset of users. For instance, I don't use\nthe geometric types, but I do use a soundex type which I created. Why\nisn't the soundex type a standard part of the backend? I, personally,\nam glad it's not, because I like the version of this type that I\ncreated a lot better than the one that's in contrib.\n As far as the whole performance-improvement issue, I can say that\nif the backend is, say, 50K smaller due to the removal of those\nfunctions, that's just that much less swapping and that much more\nmemory that's available for the OS buffer cache. Isn't that an\nimprovement worth considering?\n How about this as a compromise. We make these packages available\nin the contrib or other such area as loadable modules as well as\nmaking them available right in the main backend code, but setup\nconfigure options to enable/disable them, so when I compile, I can say\n\"--without-geometry\" to compile without those types and functions. If\nI want to add them back in later, I can compile the loadable module\nversion and add them in.\n\n-Brandon :)\n",
"msg_date": "Tue, 24 Mar 1998 06:21:48 -0600 (CST)",
"msg_from": "Brandon Ibach <[email protected]>",
"msg_from_op": true,
"msg_subject": "Data type removal"
},
{
"msg_contents": "On Tue, 24 Mar 1998, Brandon Ibach wrote:\n\n> I, for one, am in favor of converting some of the type packages to\n> loadable modules. Having those in the backend when they aren't being\n> used is much like compiling extra modules into my Apache web server\n> because they're \"kinda neat\", even though they won't be used. Also,\n\n\tI don't know about Apache, but is there any noticeable performance\ndifference between having extra modules installed or not installed? It\nmakes the binary slightly larger, but does it change performance?\n\n> if we follow the idea that we should have as many unique features in the\n> backend as possible, we could end up with all sorts of features that are\n> only used by a subset of users. For instance, I don't use the geometric\n> types, but I do use a soundex type which I created. Why isn't the\n> soundex type a standard part of the backend? I, personally, am glad\n> it's not, because I like the version of this type that I created a lot\n> better than the one that's in contrib. \n\n\tIf yours is an improvement over what we have in contrib, why not\nsubmit it? \n\n> As far as the whole performance-improvement issue, I can say that\n> if the backend is, say, 50K smaller due to the removal of those\n> functions, that's just that much less swapping and that much more\n> memory that's available for the OS buffer cache. Isn't that an\n> improvement worth considering?\n\n\tNot if it removes the Postgres from PostgreSQL...I don't have the\nip_and_mac contrib stuff loaded, because I never think of it. I know I\ncan use it, mind you, just never think of adding it in...\n\n> How about this as a compromise. We make these packages available in\n> the contrib or other such area as loadable modules as well as making\n> them available right in the main backend code, but setup configure\n> options to enable/disable them, so when I compile, I can say\n> \"--without-geometry\" to compile without those types and functions. If I\n> want to add them back in later, I can compile the loadable module\n> version and add them in. \n\n\tAs I stated earlier, if someone wants to add a\n'--without-geometry' option to configure that removes it, I have no\nproblem with that...but it will only be to remove the feature, not add it\nin. Hell, I'm probably one that would even make use of it, since, right\nnow, I don't use the geometric types either...but the default is to have\neverything included. I don't want to have to think about it someday when\nI decide I want to use those geometric types...\n\n\n",
"msg_date": "Tue, 24 Mar 1998 07:53:18 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\nSpeaking of data type removal, I was wondering if there were a better\nway to handle arrays of types. From looking in the catalog, it\nappears that for each type, there is also declared a similar type,\nwhich is the array version. It seems that arrays should be considered\nmore flags on a field, than a field type in themselves. Does this\nmake sense to anybody else?\n\n- -- \n=====================================================================\n| JAVA must have been developed in the wilds of West Virginia. |\n| After all, why else would it support only single inheritance?? |\n=====================================================================\n| Finger [email protected] for my public key. |\n=====================================================================\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQBVAwUBNRfhLIdzVnzma+gdAQGUvwH8CWMmMo633do81jgZd+pPPJPW481nfwB9\nawec8H9PjZ3QsShK4cSIJmC9Yg+IMBp3E+goHYssAO4X42Nf15+0EA==\n=8YSU\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "24 Mar 1998 11:37:00 -0500",
"msg_from": "<[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "The Hermit Hacker said:\n> \n> On Tue, 24 Mar 1998, Brandon Ibach wrote:\n> > I, for one, am in favor of converting some of the type packages to\n> > loadable modules. Having those in the backend when they aren't being\n> > used is much like compiling extra modules into my Apache web server\n> > because they're \"kinda neat\", even though they won't be used. Also,\n> \n> \tI don't know about Apache, but is there any noticeable performance\n> difference between having extra modules installed or not installed? It\n> makes the binary slightly larger, but does it change performance?\n> \n I can't say for sure what effect this would have on performance.\nHowever, what example can you give of a monolithic piece of software\nthat is superior to a slim, well-designed core with an architecture\nfor expansion? In fact, if you want to talk about what puts the\nPostgres in PostgreSQL, I think the ability to dynamically add types\nand functions is a big part of that. I would say that a large part of\nthe reason that packages like MySQL exist is so that people have an\noption of a light-weight, simple SQL database. Why can't we serve\nthat same purpose? Never consider unneeded resource consumption\nlightly, as it can be a very important factor in someone's choice of a\ndatabase package.\n\n> > if we follow the idea that we should have as many unique features in the\n> > backend as possible, we could end up with all sorts of features that are\n> > only used by a subset of users. For instance, I don't use the geometric\n> > types, but I do use a soundex type which I created. Why isn't the\n> > soundex type a standard part of the backend? I, personally, am glad\n> > it's not, because I like the version of this type that I created a lot\n> > better than the one that's in contrib. \n> \n> \tIf yours is an improvement over what we have in contrib, why not\n> submit it? \n> \n Perhaps I will, but my point is that I had a *choice* to do my own\nimplementation, rather than being forced to use one that didn't suit\nmy needs as well.\n\n> > As far as the whole performance-improvement issue, I can say that\n> > if the backend is, say, 50K smaller due to the removal of those\n> > functions, that's just that much less swapping and that much more\n> > memory that's available for the OS buffer cache. Isn't that an\n> > improvement worth considering?\n> \n> \tNot if it removes the Postgres from PostgreSQL...I don't have the\n> ip_and_mac contrib stuff loaded, because I never think of it. I know I\n> can use it, mind you, just never think of adding it in...\n> \n Since when do geometric types put the Postgres in PostgreSQL? Or\nIP and MAC types, for that matter? I can use the improvements in 6.3,\nbut I haven't upgraded yet because it will be a bit of a job. Is\nthere something we can throw in to solve that, or is it maybe\nsomething I have to do if I want the benefits?\n\n> > How about this as a compromise. We make these packages available in\n> > the contrib or other such area as loadable modules as well as making\n> > them available right in the main backend code, but setup configure\n> > options to enable/disable them, so when I compile, I can say\n> > \"--without-geometry\" to compile without those types and functions. If I\n> > want to add them back in later, I can compile the loadable module\n> > version and add them in. \n> \n> \tAs I stated earlier, if someone wants to add a\n> '--without-geometry' option to configure that removes it, I have no\n> problem with that...but it will only be to remove the feature, not add it\n> in. Hell, I'm probably one that would even make use of it, since, right\n> now, I don't use the geometric types either...but the default is to have\n> everything included. I don't want to have to think about it someday when\n> I decide I want to use those geometric types...\n> \n You may notice that the option I suggested was one that could be\ninvoked to remove the feature. In other words, the default is to have\nthe feature included unless the user asks for it to not be. All I'm\nsaying is, don't force the users to run a Postgres which has unneeded\nbaggage.\n\n-Brandon :)\n",
"msg_date": "Tue, 24 Mar 1998 12:02:22 -0600 (CST)",
"msg_from": "Brandon Ibach <[email protected]>",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "Brandon writes:\n> I, for one, am in favor of converting some of the type packages to\n> loadable modules. Having those in the backend when they aren't being\n...\n> As far as the whole performance-improvement issue, I can say that\n> if the backend is, say, 50K smaller due to the removal of those\n> functions, that's just that much less swapping and that much more\n> memory that's available for the OS buffer cache. Isn't that an\n> improvement worth considering?\n\nNo. If the functions are not used, the pages of the executable\nthat contain them are never loaded. So there is no effective memory impact\nfrom having these where they are.\n\nAs far as disk space, disk costs $0.03 per megabyte these days so saving\n50Kb is completely insignificant.\n\nIt may be a good idea to make the system more modular, but there are\nthousands of good ideas for improving it. I think we would be better\nserved by spending our time on something that makes a difference, not\njust mindlessly pushing code around for no particular benefit.\n\n> How about this as a compromise. We make these packages available\n> in the contrib or other such area as loadable modules as well as\n> making them available right in the main backend code, but setup\n> configure options to enable/disable them, so when I compile, I can say\n> \"--without-geometry\" to compile without those types and functions. If\n> I want to add them back in later, I can compile the loadable module\n> version and add them in.\n\nI don't particularly like this model of adding extensions either. It implies\nthat you have to rerun configure and build the whole system to add a module.\nIn fact, all that is needed is to build the module and run the setup\nscripts and the running database picks up the new functionality. \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": "Tue, 24 Mar 1998 11:10:10 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "> \n> Speaking of data type removal, I was wondering if there were a better\n> way to handle arrays of types. From looking in the catalog, it\n> appears that for each type, there is also declared a similar type,\n> which is the array version. It seems that arrays should be considered\n> more flags on a field, than a field type in themselves. Does this\n> make sense to anybody else?\n\nIs an array the same thing as a scalar?\n\nWhat would be the benefit of making arrays \"flags on a field\" instead of\na \"field type in themselves\". Seriously, how would this improve _anything_?\n\nThe postgres type system is very flexible and powerful as is. What is the\nproblem this is trying to solve?\n\nWhat is the motivation for data type removal?\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": "Tue, 24 Mar 1998 11:27:52 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\n\n> From: [email protected] (David Gould)\n> > Speaking of data type removal, I was wondering if there were a better\n> > way to handle arrays of types. From looking in the catalog, it\n> > appears that for each type, there is also declared a similar type,\n> > which is the array version. It seems that arrays should be considered\n> > more flags on a field, than a field type in themselves. Does this\n> > make sense to anybody else?\n> \n> Is an array the same thing as a scalar?\nUhm, I don't believe so.\n\n> What would be the benefit of making arrays \"flags on a field\" instead of\n> a \"field type in themselves\". Seriously, how would this improve _anything_?\n\nIn particular, I was thinking of the PostgreSQL module for Python. It\nhas a nice interface, but needs intimate knowledge of the data types.\nSomehow, it seems to be excessively kludgy to have to have e.g. an\nint4 type, and an int4[] type. When I query a table, if one of the\nfields is an array of integers, then I either want to know that it's\nan array, or it's integer. If the returned type is \"array,\" then I\nhave to magically know that the array is filled with integers. If\nit's an integer, then I just have to recognize that it's actually a\nseries of integers. With the current setup, there has to be separate\ntype handlers for char(2), char(2)[], int4, int4[], int2, int2[],\nfloat8, float8[], etcetera. I'd rather have handlers for the base\ntype, and an iterator that's used for arrays.\n\n> What is the motivation for data type removal?\n\nIn this case, I don't want to *remove* the data type, just change the\nidentification method. An array of int4 should really be recognized\nas both (type)int4 and (type)array. Urhm, since this is PostgreSQL, I\nguess I'm arguing for type composition (how appropriate for an\nobject-relational database).\n\n- -- \n=====================================================================\n| JAVA must have been developed in the wilds of West Virginia. |\n| After all, why else would it support only single inheritance?? |\n=====================================================================\n| Finger [email protected] for my public key. |\n=====================================================================\n\n-----BEGIN PGP SIGNATURE-----\nVersion: 2.6.2\n\niQBVAwUBNRgR1YdzVnzma+gdAQFGVgH+PX4jvMZfhFK1ZzVdYuL844gCrg5ZleDt\nT4x2b6TbOOqNGn7SZVQtpuCZXG7lSDFMNDRan2ft0OmE9ZlJ69ZpkA==\n=q92I\n-----END PGP SIGNATURE-----\n\n",
"msg_date": "24 Mar 1998 15:04:38 -0500",
"msg_from": "<[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "On Tue, 24 Mar 1998, Brandon Ibach wrote:\n\n> The Hermit Hacker said:\n> > \n> > On Tue, 24 Mar 1998, Brandon Ibach wrote:\n> > > I, for one, am in favor of converting some of the type packages to\n> > > loadable modules. Having those in the backend when they aren't being\n> > > used is much like compiling extra modules into my Apache web server\n> > > because they're \"kinda neat\", even though they won't be used. Also,\n> > \n> > \tI don't know about Apache, but is there any noticeable performance\n> > difference between having extra modules installed or not installed? It\n> > makes the binary slightly larger, but does it change performance?\n> > \n> I can't say for sure what effect this would have on performance.\n> However, what example can you give of a monolithic piece of software\n> that is superior to a slim, well-designed core with an architecture\n> for expansion? \n\n\tPostgreSQL?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 24 Mar 1998 17:22:51 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "On Tue, 24 Mar 1998, David Gould wrote:\n\n> What is the motivation for data type removal?\n\n\tThere appear to be two \"motivations\" from what I'm hearing:\n\n\t1. reduced memory footprint\n\t2. reduced \"disk\" footprint\n\n\tFrom what I've been able to gather, I believe that ppl are\nexpecting that the result is a faster server, based on less \"perceived\"\nfat...nobody has yet been able to prove that though, which I believe\nDarren is working on.\n\n\tPart of this does have me thinking though, in that one of the\n*big* features that I perceived PostgreSQL to have was the fact that it\nwas/is very easy to extend our datatypes and functions without having to\nrecompile the server. \n\n\tSo, right now, I'm kinda sitting on the fence in that I'm curious\nif we *are* taking anything away by moving the geometrics from inside of\nthe server to a loadable module, given that the fact that we *are* able to\ndo that is what appears to be one of our major points of \"uniqueness\"...\n\n\tRight now, there has been alot of speculation as to whether or not\nmoving the geometrics into a loadable module or not would gain (or lose,\nfor that matter) us anything...\n\n\tLet's say we have a $prefix/modules directory, where we move the\ngeometrics stuff into. And, let's say we have an 'loadmod' command added\nthat, if someone did 'loadmod geometrics;' from psql, it loaded up the\ngeometrics module...would we have the *same* functionality as if it were\ncompiled in, like it is now? Or we would lose something?\n\n\tTo me, this is one of the *biggest* features of PostgreSQL, the\nfact that we can easily add in new features without recompiling...but\nwithout seeing actual numbers (performance as well as \"disk\"), we can\ndiscuss this till we are blue in the face.\n\n\tDarren...if you want to provide us with a working model of this,\nusing the geometrics, so be it...I won't say that it will or won't be\nintegrated without seeing this in operation, and without seeing some\n*actual* benefit to it...as the saying goes \"Talk is Cheap...\"\n\n\tSo, I say, let's leave this as a \"to be researched\" and wait for\nsomething concrete to look at...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd|postgresql}.org \n\n",
"msg_date": "Tue, 24 Mar 1998 18:06:17 -0400 (AST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "> As far as the whole performance-improvement issue, I can say that\n> if the backend is, say, 50K smaller due to the removal of those\n> functions, that's just that much less swapping and that much more\n> memory that's available for the OS buffer cache. Isn't that an\n> improvement worth considering?\n\nWell, it is really only 50k per installation, not 50k per backend\nbecause they all share the same executable in the buffer cache.\n\nThe problem with breaking it up, as I see it, is that all of a sudden\nthe regression tests and the nice descriptions of \\do, \\df, etc are much\nharder to implement.\n\nMaybe in six months, when we will have resolved all of the\nperformance/SQL92/feature issues, this will be a good use of our time,\nbut at this point, trying to keep all that straight while working on\nmuch more important issues to end-users is hard to justify.\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, 24 Mar 1998 20:57:59 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "> The postgres type system is very flexible and powerful as is. What is \n> the problem this is trying to solve?\n> \n> What is the motivation for data type removal?\n\nThere are many motivations involved here. I brought it up originally\nbecause the char2-16 types are not supported and do not provide any\nfunctionality over the char(),varchar(),text string types.\n\nOthers suggested that since they do not care about the geometric types\nthat those should be removed too.\n\nI regret bringing it up. Postgres has many unique features, and\nstripping it to become a plain vanilla SQL92 machine is a waste of time\nimo. \n\nIf any restructuring happens which removes, or makes optional, some of\nthe fundamental types, it should be accomplished so that the types can\nbe added in transparently, from a single set of source code, during\nbuild time or after. OIDs would have to be assigned, presumably, and the\nhardcoding of the function lookups for builtin types must somehow be\ndone incrementally. Probably needs more than this to be done right, and\nwithout careful planning and implementation we will be taking a big step\nbackwards.\n\nWith the amount of time being spent in discussion, _still without any\nquantitative estimates for performance improvement_, it seems like\nsomeone should do some measurements. Even without them though it is\npretty clear that it won't benefit a large database, since the fraction\nof time spent constructing a query will be small compared to the time it\ntakes to traverse the tables in the query.\n\nSeems to me that Postgres' niche is at the high end of size and\ncapability, not at the lightweight end competing for design wins against\nsystems which don't even have transactions.\n\n - Tom\n",
"msg_date": "Wed, 25 Mar 1998 07:41:24 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "> If any restructuring happens which removes, or makes optional, some of\n> the fundamental types, it should be accomplished so that the types can\n> be added in transparently, from a single set of source code, during\n> build time or after. OIDs would have to be assigned, presumably, and the\n> hardcoding of the function lookups for builtin types must somehow be\n> done incrementally. Probably needs more than this to be done right, and\n> without careful planning and implementation we will be taking a big step\n> backwards.\n\n\nThis whole discussion reminds me of someone who thinks he is getting low\ngas milage because his tire pressure is low, when in fact, he has a\nwhole in his gas tank. We have some major holes to plug, but they are\nhard to see. People see the tire pressure/pg_class table, and think the\ndatabase can be improved. You could double the size of the system\ntables, or the binary, and see no performance change. Make them\n10-times bigger, and see if you find a difference.\n\nAren't there enough serious issues on the TODO list for people? These\nare actual complaints/bugs or requests from users, not pie-in-the-sky\nwouldn't-it-be-nice-if-we-had ideas.\n\nI had a private e-mail conversation with someone, and they mentioned\nthat we seem to be mired in micro-optimizations, that really are not\ngoing to do us any significant good. I see his point.\n\nLook at the TCL/TK mess I got into just trying to get that to work, and\nremoving the stuff that checked for specific tck/tk version numbers. \nCan you imagine the effort we will go through ripping out types?\n\nAnother thing. It is the type extensibility that makes us more complex,\nnot the types themselves.\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, 25 Mar 1998 09:43:37 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "David Gould wrote:\n> \n> > How about this as a compromise. We make these packages available\n> > in the contrib or other such area as loadable modules as well as\n> > making them available right in the main backend code, but setup\n> > configure options to enable/disable them, so when I compile, I can say\n> > \"--without-geometry\" to compile without those types and functions. If\n> > I want to add them back in later, I can compile the loadable module\n> > version and add them in.\n> \n> I don't particularly like this model of adding extensions either. It implies\n> that you have to rerun configure and build the whole system to add a module.\n> In fact, all that is needed is to build the module and run the setup\n> scripts and the running database picks up the new functionality.\n\n\nThat's the nice thing with Linux. You can throw out the sound driver\nany time and insert a new one whenever you want to. If I decide to\nuse the MIDI port, I can rmmod the old driver and insmod the new\none with midi support...\nIf I connect a Wingdogs PC to my network, I can insmod smbfs and\nthen I can mount the windows file resources from my Linux box. No need\nto upgrade kernel or reboot.\nIf someone comes up with a new super-NFS, I just unmount my nfs's,\ninsmod super-nfs, and mount -a -t nfs...\n\nWith loadable modules in PostgreSQL, you could upgrade the ORACLE\ncompatability module by typing \"make modules modules-install\".\n\nNo need to kill all backends and postmaster, deal with angry users,\nmake backups, install new binaries, dump-and-restore, track down bugs,\nand so on.\n\nIt would also be easier to contribute to PostgreSQL by creating\nmodules, than having to submit patches to the server core...\nLet's say I have made a storage manager for tapes. It would be easier\nto make a module which used a standard set of \"modules\" functions\nto register the new manager in the core, than submitting patches\nto be included in the core...\nAnd if something causes troubles, just remove that module! Then it\nwould not be a big problem if a new feature was buggy. Perhaps it\nwould be easier to find bugs then (binary search - remove module\nby module).\n\nJust imagine Linux if people had to submit patches to the kernel\ninstead of just writing their drivers as modules!!!\n\n\n/* m */\n",
"msg_date": "Wed, 25 Mar 1998 17:36:56 +0100",
"msg_from": "Mattias Kregert <[email protected]>",
"msg_from_op": false,
"msg_subject": "Dynamically loadable modules"
},
{
"msg_contents": "Thomas G. Lockhart wrote:\n> \n> With the amount of time being spent in discussion, _still without any\n> quantitative estimates for performance improvement_, it seems like\n> someone should do some measurements. Even without them though it is\n> pretty clear that it won't benefit a large database, since the fraction\n> of time spent constructing a query will be small compared to the time it\n> takes to traverse the tables in the query.\n\nI don't think the main reason for moving things out from the\ncore would be performance, but for making it easier to manage\nextensions.\n\n/* m */\n",
"msg_date": "Wed, 25 Mar 1998 17:38:54 +0100",
"msg_from": "Mattias Kregert <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "On Wed, 25 Mar 1998, Mattias Kregert wrote:\n\n> Thomas G. Lockhart wrote:\n> > \n> > With the amount of time being spent in discussion, _still without any\n> > quantitative estimates for performance improvement_, it seems like\n> > someone should do some measurements. Even without them though it is\n> > pretty clear that it won't benefit a large database, since the fraction\n> > of time spent constructing a query will be small compared to the time it\n> > takes to traverse the tables in the query.\n> \n> I don't think the main reason for moving things out from the\n> core would be performance, but for making it easier to manage\n> extensions.\n\n\t\"Easier to manage extensions\"??\n\n\n",
"msg_date": "Wed, 25 Mar 1998 11:55:14 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "> I don't think the main reason for moving things out from the\n> core would be performance, but for making it easier to manage\n> extensions.\n\n??\n",
"msg_date": "Wed, 25 Mar 1998 17:01:44 +0000",
"msg_from": "\"Thomas G. Lockhart\" <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "> \n> > The postgres type system is very flexible and powerful as is. What is \n> > the problem this is trying to solve?\n> > \n> > What is the motivation for data type removal?\n> \n> There are many motivations involved here. I brought it up originally\n> because the char2-16 types are not supported and do not provide any\n> functionality over the char(),varchar(),text string types.\n> \n> Others suggested that since they do not care about the geometric types\n> that those should be removed too.\n> \n> I regret bringing it up. Postgres has many unique features, and\n> stripping it to become a plain vanilla SQL92 machine is a waste of time\n> imo. \n\nI agree completely. This was the point I was trying to make by asking \nfor the motivation. If there was a clearcut proven performance gain to be\nhad, I would support it. But as it is just speculation, it seems kinda\npointless to push a bunch of code around (risking breaking it) for no\nvery good reason.\n \n> If any restructuring happens which removes, or makes optional, some of\n> the fundamental types, it should be accomplished so that the types can\n> be added in transparently, from a single set of source code, during\n> build time or after. OIDs would have to be assigned, presumably, and the\n> hardcoding of the function lookups for builtin types must somehow be\n> done incrementally. Probably needs more than this to be done right, and\n> without careful planning and implementation we will be taking a big step\n> backwards.\n\nExactly. Right now modules get installed by building the .so files and\nthen creating all the types, functions, rules, tables, indexes etc. This\nis a bit more complicated than the Linux kernal 'insmod' operation. We could\neasily make the situation worse through careless \"whacking\".\n \n> Seems to me that Postgres' niche is at the high end of size and\n> capability, not at the lightweight end competing for design wins against\n> systems which don't even have transactions.\n\nAnd, there are already a couple of perfectly good 'toy' database systems.\nWhat is the point of having another one? Postgres should move toward\nbecoming an \"industrial strength\" solution.\n\n> - Tom\n\nThank you for bringing some sense to this discussion.\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": "Thu, 26 Mar 1998 16:51:05 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "Bruce Momjian:\n> Aren't there enough serious issues on the TODO list for people? These\n> are actual complaints/bugs or requests from users, not pie-in-the-sky\n> wouldn't-it-be-nice-if-we-had ideas.\n> \n> I had a private e-mail conversation with someone, and they mentioned\n> that we seem to be mired in micro-optimizations, that really are not\n> going to do us any significant good. I see his point.\n\nPlease feel free to quote me if I am the \"someone\" (I suspect I was). \n \n> Another thing. It is the type extensibility that makes us more complex,\n> not the types themselves.\n\nIt is also the type extensibility that makes us more valuable. There are\nlots of \"plain old SQL\" systems out there. There are very few extendable\nsystems.\n\nPerhaps we might want to start encourageing people to use the\nextensibility and write a few useful application modules. I know of a\nfew that have been pretty sucessful in the commercial world:\n\n - Html generation from inside the database. Store your pages in the db\n with embedded <SQL ... /SQL> tags to dynamically generate web content\n based on queries.\n\n - Integrated full text search engines. Store documents and be able to\n query \" select abstract from papers\n where contains(fulltext, 'database', 'transaction', 'recovery')\".\n\nAnd so on.\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": "Thu, 26 Mar 1998 17:01:27 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
}
] |
[
{
"msg_contents": "REGARDING varchar size\n\nGiven the results of a query using the perl interface to postgres, how can I\nfind the maximum length of the data returned for the varchar fields in the\nquery?\n\nIs there something in the \"results\" buffer or can I query the system catalogs\nfor this information? If you need to query the system catalogs, what field\nactually contains the maximum length of a varchar field? In other words, how\ncan I find what the field was defined as in the create table statement (e.g.,\n...attribute1 varchar(12)...) using the system catalogs?\n\nTIA,\nAndy\n\n",
"msg_date": "24 Mar 1998 13:23:08 U",
"msg_from": "\"Andy Farrell\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "varchar size"
}
] |
[
{
"msg_contents": "REGARDING Postgres \"in the field\"\n\nWe are contemplating using postgres in an internal project currently in\nprogress. I have to give a presentation on rdbms that I've evaluated. I'd\nlike to include a slide on what postgres is being used for \"in the field\". \n\nI'd like to get the warm fuzzy that postgres is not just being used for\npost-graduate thesis work. \n\nSo, what's everyone using postgres for?\n\nTIA,\nAndy\n\n",
"msg_date": "24 Mar 1998 13:31:14 U",
"msg_from": "\"Andy Farrell\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "Postgres \"in the field\""
},
{
"msg_contents": "> Date: 24 Mar 1998 13:31:14 U\n> From: \"Andy Farrell\" <[email protected]>\n\n> We are contemplating using postgres in an internal project currently in\n> progress. I have to give a presentation on rdbms that I've evaluated. I'd\n> like to include a slide on what postgres is being used for \"in the field\". \n> \n> I'd like to get the warm fuzzy that postgres is not just being used for\n> post-graduate thesis work. \n> \n> So, what's everyone using postgres for?\n\n1. Network administration. Tables keep track of hostname, function,\n location, and interfaces on our WAN. Automated scripts scan these\n tables to monitor active services including NTP and DNS and report\n possible problems. Hosts are classified by OS and service and\n divided into differing populations. We've enhanced the \"config/Rdist\"\n tools reported in Lisa '94 (Rouillard & Martin) to work from\n a pgsql backend, allowing reconfiguration of some or all host\n populations to be pushed from a single system.\n\n2. Query logging. Our business is computer assisted telephony. When\n someone calls a customer of ours, we run queries for the related\n service. Some results of this process (time of the query, type of\n data returned, success/failure status) are logged and reviewed\n using pgsql.\n\n3. Vendor location. Our newest pgsql app uses location information\n based on the caller's phone number to route the call to one of\n several nearest vendor sites. An rtree index is used for the\n location process.\n\nWe use a fairly pragmatic mix of operating systems and tools, both\ncommercial and free software. So far, PostgreSQL projects are limited\nto internal use and fairly low traffic customer applications. We\nexpect to increase our use of PostgreSQL as our confidence in its\nstability and performance grow.\n\nHal Snyder, M.D.\nV.P. Networking\nVail Systems, Inc.\n\n",
"msg_date": "Tue, 24 Mar 1998 13:23:30 -0600 (CST)",
"msg_from": "Hal Snyder <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Postgres \"in the field\""
},
{
"msg_contents": "On 24 Mar 1998, Andy Farrell wrote:\n\n> So, what's everyone using postgres for?\n\nAlthough we've got a few other things working, (mostly billing systems,\nand suchlike) the database behind http://www.hnba.org/ (The Houston\nNorthWest Bar Association) is postgres 6.2\n\n-- \nJonathan Guthrie ([email protected])\nInformation Broker Systems +281-895-8101 http://www.brokersys.com/\n12703 Veterans Memorial #106, Houston, TX 77014, USA\n\nWe sell Internet access and commercial Web space. We also are general\nnetwork consultants in the greater Houston area.\n\n\n",
"msg_date": "Tue, 24 Mar 1998 20:58:06 -0600 (CST)",
"msg_from": "Jonathan Guthrie <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Postgres \"in the field\""
}
] |
[
{
"msg_contents": "> > I don't know how big of a performance boost it provides in the cache, but\n> > removing the functions associated with the char types shrank the pg_proc\n> > table from 906 to 842 entries or a bit over 7%.\n> > \n> > Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> > types! Throw in 25 more for the cash/money functions. Bloat city if you\n> > never use these things. Thirty percent could be moved out to contrib and\n> > not missed by most postgres users.\n> > \n> \n> Yes, but if they are never referenced, the cache is empty for those\n> types. Unless there is some performance change with their removal, why\n> remove them? Disk space of binary?\n\nHow does the cache really work then? Does one pg disk block map to one buffer?\n\nWhen you say \"the cache is empty for those types.\", what do you mean?\n\ndarrenk\n",
"msg_date": "Tue, 24 Mar 1998 10:10:05 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "On Tue, 24 Mar 1998, Darren King wrote:\n\n> > > I don't know how big of a performance boost it provides in the cache, but\n> > > removing the functions associated with the char types shrank the pg_proc\n> > > table from 906 to 842 entries or a bit over 7%.\n> > > \n> > > Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> > > types! Throw in 25 more for the cash/money functions. Bloat city if you\n> > > never use these things. Thirty percent could be moved out to contrib and\n> > > not missed by most postgres users.\n> > > \n> > \n> > Yes, but if they are never referenced, the cache is empty for those\n> > types. Unless there is some performance change with their removal, why\n> > remove them? Disk space of binary?\n> \n> How does the cache really work then? Does one pg disk block map to one buffer?\n> \n> When you say \"the cache is empty for those types.\", what do you mean?\n\nI think there's some confusion about which cache is meant. I think the \nprevious poster (ie. the one before darren) is thinking of processor \ncache, be it first level or second level. Darren is probably meaning a \npostgresql internal cache....\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": "Tue, 24 Mar 1998 18:02:46 +0100 (MET)",
"msg_from": "Maarten Boekhold <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "> > > I don't know how big of a performance boost it provides in the cache, but\n> > > removing the functions associated with the char types shrank the pg_proc\n> > > table from 906 to 842 entries or a bit over 7%.\n\nNo performance impact.\n\n> > > Want to shrink it further? Of those remaining 842, _230_ are for the geometric\n> > > types! Throw in 25 more for the cash/money functions. Bloat city if you\n> > > never use these things. Thirty percent could be moved out to contrib and\n> > > not missed by most postgres users.\n> > \n> > Yes, but if they are never referenced, the cache is empty for those\n> > types. Unless there is some performance change with their removal, why\n> > remove them? Disk space of binary?\n> \n> How does the cache really work then? Does one pg disk block map to one buffer?\n> \n> When you say \"the cache is empty for those types.\", what do you mean?\n\nThe function cache has one entry for each function in use. If a function has\nnever been called, then no cache entry is ever created for it. The cache\nis organized per function, it is not part of the buffer / page cache so\neach entry only takes a few bytes.\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": "Tue, 24 Mar 1998 11:17:40 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "> > Yes, but if they are never referenced, the cache is empty for those\n> > types. Unless there is some performance change with their removal, why\n> > remove them? Disk space of binary?\n> \n> How does the cache really work then? Does one pg disk block map to one buffer?\n> \n> When you say \"the cache is empty for those types.\", what do you mean?\n\nI believe the buffer page cache does load in those types if they are in\nthe same page as a requsted type, but the relation/function caches only\nstore in requested relations/functions, as far as I know.\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, 24 Mar 1998 21:38:35 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "> \n> The nicest thing would be if contrib was changed to \"modules\", and\n> everything was compiled and installed in /usr/local/pgsql/modules/.\n> Then the user could install them by typing \"INSMOD ORACLE\" and then\n> /usr/local/pgsql/modules/oracle/insmod.sql would be executed and\n> install shared libraries and sql commands. \"RMMOD ORACLE\" would\n> do the opposite. Just like Linux modules.\n\nVadim has something very similar to this in the contrib/spi Makefile.\n\nWould be a nice model to follow. I agree with your other comment about\nthe impression given by \"contrib.\"\n\nContrib could stay the way it is, for independent contributions, then\nhave a new directory, \"modules\", for contributions that are integrated\ninto the postgres make.\n\nTonight I will finalize my patch for the char types removal and then\nlater in the week think more about this idea.\n\n> Perhaps it would even be possible to autoload modules if the modules\n> directories contains a file with commands/functions listed, which\n> would be searched automagically...\n> \n> This is the idea:\n> \n> Is it possible to register commands dynamically in the parser?\n\nCREATE COMMAND would be a new one. Talk about extensibility! :)\n\ndarrenk\n\n",
"msg_date": "Tue, 24 Mar 1998 10:52:26 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] char types gone."
},
{
"msg_contents": "On Tue, 24 Mar 1998, Darren King wrote:\n\n> > \n> > The nicest thing would be if contrib was changed to \"modules\", and\n> > everything was compiled and installed in /usr/local/pgsql/modules/.\n> > Then the user could install them by typing \"INSMOD ORACLE\" and then\n> > /usr/local/pgsql/modules/oracle/insmod.sql would be executed and\n> > install shared libraries and sql commands. \"RMMOD ORACLE\" would\n> > do the opposite. Just like Linux modules.\n> \n> Vadim has something very similar to this in the contrib/spi Makefile.\n> \n> Would be a nice model to follow. I agree with your other comment about\n> the impression given by \"contrib.\"\n> \n> Contrib could stay the way it is, for independent contributions, then\n> have a new directory, \"modules\", for contributions that are integrated\n> into the postgres make.\n\n\tDoing something like an LKM (loadable kernel module) would be,\nIMHO, neat, if it could be done seamlessly...we have the loading\ncapabilities now...but seamlessly would mean that I didn't have to think\nabout adding it each time I upgraded my server. If you can pull *that*\noff, then we might have something to consider. \n\n\tAs long as JoeUser that wants to use the geometric types doesn't\nhave to think of adding them in to use them, great...I don't want, in 6\nmonths, to realize that I don't have those geometric types installed when\nI happen to require them :(\n\n\n\n",
"msg_date": "Tue, 24 Mar 1998 11:38:54 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] char types gone."
}
] |
[
{
"msg_contents": "> > How about this as a compromise. We make these packages available in\n> > the contrib or other such area as loadable modules as well as making\n> > them available right in the main backend code, but setup configure\n> > options to enable/disable them, so when I compile, I can say\n> > \"--without-geometry\" to compile without those types and functions. If I\n> > want to add them back in later, I can compile the loadable module\n> > version and add them in. \n> \n> \tAs I stated earlier, if someone wants to add a\n> '--without-geometry' option to configure that removes it, I have no problem\n> with that...but it will only be to remove the feature, not add it in.\n\nI can live with this. Everything is \"--with-xxx\" by default, but can not\nbe built in by using \"--without-xxx\".\n\nWould it be acceptable to move the code for these to a new directory, say,\nsrc/modules? Something along the lines of...\n\nsrc/modules/geometric\nsrc/modules/ip_and_mac\n\nThis would allow for each type to have a pg_proc.h, pg_type.h, etc. Much\ncleaner than #define'ing the heck out of the existing include files. The\ngeometric/pg_proc.h would contain the entries from pg_proc.h. Then there\nwould also be a .sql file that contains the necessary commands to load\nthe module if it was not compiled in or was just needed in one database.\n\nForcing someone to re-compile to use module would seem to go completely\nagainst the extensibility side of postgres.\n\nI think once there is one thing there as a module, it will serve as an\nexample for others. A simple example would be the cash/money code. Add\nan indexing method to it and it would be a brief but complete example.\n\ndarrenk\n",
"msg_date": "Tue, 24 Mar 1998 11:11:02 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Data type removal"
},
{
"msg_contents": "On Tue, 24 Mar 1998, Darren King wrote:\n\n> > > How about this as a compromise. We make these packages available in\n> > > the contrib or other such area as loadable modules as well as making\n> > > them available right in the main backend code, but setup configure\n> > > options to enable/disable them, so when I compile, I can say\n> > > \"--without-geometry\" to compile without those types and functions. If I\n> > > want to add them back in later, I can compile the loadable module\n> > > version and add them in. \n> > \n> > \tAs I stated earlier, if someone wants to add a\n> > '--without-geometry' option to configure that removes it, I have no problem\n> > with that...but it will only be to remove the feature, not add it in.\n> \n> I can live with this. Everything is \"--with-xxx\" by default, but can not\n> be built in by using \"--without-xxx\".\n> \n> Would it be acceptable to move the code for these to a new directory, say,\n> src/modules? Something along the lines of...\n> \n> src/modules/geometric\n> src/modules/ip_and_mac\n> \n> This would allow for each type to have a pg_proc.h, pg_type.h, etc. Much\n> cleaner than #define'ing the heck out of the existing include files. The\n> geometric/pg_proc.h would contain the entries from pg_proc.h. Then there\n> would also be a .sql file that contains the necessary commands to load\n> the module if it was not compiled in or was just needed in one database.\n> \n> Forcing someone to re-compile to use module would seem to go completely\n> against the extensibility side of postgres.\n> \n> I think once there is one thing there as a module, it will serve as an\n> example for others. A simple example would be the cash/money code. Add\n> an indexing method to it and it would be a brief but complete example.\n\n\tWhy don't we start with the ip_and_mac stuff...take that,\nintegrate it into the \"core\", build it so that --without-ip_and_mac\ndisables it, and let's see how that one works.\n\n\tWhat I'm curious about right now is what different its going to\nmake. Does having ip_and_mac in the core, even if I don't use it, reduce\nperformance, or makes no difference? How much does it increase the\nfootprint of the binary? If it makes negligble difference, then it isn't\nworth doing, you may as well just leave everything in there...\n\n\tBut, build us a sample with the ip_and_mac stuff, as to what you\nare thinking and let's go from that...but ignore the *core* stuff for\nnow...\n\n",
"msg_date": "Tue, 24 Mar 1998 11:51:55 -0500 (EST)",
"msg_from": "The Hermit Hacker <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Data type removal"
}
] |
[
{
"msg_contents": "> > What would be the benefit of making arrays \"flags on a field\" instead of\n> > a \"field type in themselves\". Seriously, how would this improve _anything_?\n> \n> In particular, I was thinking of the PostgreSQL module for Python. It\n> has a nice interface, but needs intimate knowledge of the data types.\n> Somehow, it seems to be excessively kludgy to have to have e.g. an\n> int4 type, and an int4[] type. When I query a table, if one of the\n> fields is an array of integers, then I either want to know that it's\n> an array, or it's integer. If the returned type is \"array,\" then I\n> have to magically know that the array is filled with integers. If\n> it's an integer, then I just have to recognize that it's actually a\n> series of integers. With the current setup, there has to be separate\n> type handlers for char(2), char(2)[], int4, int4[], int2, int2[],\n> float8, float8[], etcetera. I'd rather have handlers for the base\n> type, and an iterator that's used for arrays.\n> \n\nFWIW,\n\nHaven't followed this too closely, but there is some array_iterator\ncode in the contrib directory coutesy of Massimo.\n\ndarrenk\n",
"msg_date": "Tue, 24 Mar 1998 15:16:21 -0500",
"msg_from": "[email protected] (Darren King)",
"msg_from_op": true,
"msg_subject": "Re: [HACKERS] Data type removal"
}
] |
[
{
"msg_contents": "\n> On Tue, 24 Mar 1998, David Gould wrote:\n> \n> > What is the motivation for data type removal?\n> \n> \tThere appear to be two \"motivations\" from what I'm hearing:\n> \n> \t1. reduced memory footprint\n> \t2. reduced \"disk\" footprint\n\t3. modularization (In my experience, that in and of itself is a\nbenefit.)\n\t3a. Encapsulation \n\t3b. Re-Organization (isn't always a bad thing)\n\nI haven't looked at enough of the postgres code to be anything even, in\nany small way, an expert, but I have found that having someone approach\na problem from a new perspective, when viewed with an open mind is never\ndetrimental. So I for one would like to see what Darren comes up with,\ncould be very useful even if it isn't for what he originally intends.\n\n\tJust my $0.01 ( I can no longer afford a full $0.02)\n\t\t-DEJ\n\n",
"msg_date": "Tue, 24 Mar 1998 16:56:51 -0600",
"msg_from": "\"Jackson, DeJuan\" <[email protected]>",
"msg_from_op": true,
"msg_subject": "RE: [HACKERS] Data type removal"
},
{
"msg_contents": "\n\n> > \t1. reduced memory footprint\n> > \t2. reduced \"disk\" footprint\n> \t3. modularization (In my experience, that in and of itself is a\n> benefit.)\nTHIS IS RIGHT ON !!!\n\nWayne\n\n> \t3a. Encapsulation \n> \t3b. Re-Organization (isn't always a bad thing)\n> \n> I haven't looked at enough of the postgres code to be anything even, in\n> any small way, an expert, but I have found that having someone approach\n> a problem from a new perspective, when viewed with an open mind is never\n> detrimental. So I for one would like to see what Darren comes up with,\n> could be very useful even if it isn't for what he originally intends.\n> \n> \tJust my $0.01 ( I can no longer afford a full $0.02)\n> \t\t-DEJ\n> \n> \n\n",
"msg_date": "Tue, 24 Mar 1998 21:07:23 -0500 (EST)",
"msg_from": "wward <[email protected]>",
"msg_from_op": false,
"msg_subject": "RE: [HACKERS] Data type removal"
}
] |
[
{
"msg_contents": "Seems like there is something wrong with this patch...\n\nPatching file src/bin/pgaccess/formdemo.sql using Plan A...\npatch: **** malformed patch at line 22987: box} {}}\n[root@mimo postgresql-6.3]# \n\nMike\n\n-- \nWWW: http://www.lodz.pdi.net/~mimo tel: Int. Acc. Code + 48 42 148340\nadd: Michal Mosiewicz * Bugaj 66 m.54 * 95-200 Pabianice * POLAND\n",
"msg_date": "Wed, 25 Mar 1998 00:01:14 +0100",
"msg_from": "Michal Mosiewicz <[email protected]>",
"msg_from_op": true,
"msg_subject": "Patch 6.3-6.3.1 wrong?"
}
] |
[
{
"msg_contents": "xxxx=> \\d logdt\n\nTable = logdt\n+----------------------------------+----------------------------------+-------+\n| Field | Type |\nLength|\n+----------------------------------+----------------------------------+-------+\n| dt | timestamp \n| 4 |\n+----------------------------------+----------------------------------+-------+\nxxxx=> explain select * from log where dt < '19980203';\nNOTICE: QUERY PLAN:\n\nSeq Scan on log (cost=105832.15 size=699588 width=62)\n\nThere is an index on log table, dt field. The index is b-tree.\nHowever it doesn't seem to be used. (Of course I have vacuumed it). The\ntable has about 2M records. I don't think that Seq Scan is a good idea.\n\nAlso, what if I agregate it on dt field to count(*) or sum some values.\nIt would be sequentially scanned, then sorted, then grouped and finally\nagregated, right?\n\nWell, sometimes it may be good enough. But if this table is big enough,\nit would be wiser to use index to select ranges from the table and then\nagregate those values without sorting. \n\nOnce I saw index based agregates in the TODO list, but somehow it\ndisappeared.\n\nRegards,\nMike\n\n-- \nWWW: http://www.lodz.pdi.net/~mimo tel: Int. Acc. Code + 48 42 148340\nadd: Michal Mosiewicz * Bugaj 66 m.54 * 95-200 Pabianice * POLAND\n",
"msg_date": "Wed, 25 Mar 1998 04:17:18 +0100",
"msg_from": "Michal Mosiewicz <[email protected]>",
"msg_from_op": true,
"msg_subject": "Optimizer fails?"
},
{
"msg_contents": "> Seq Scan on log (cost=105832.15 size=699588 width=62)\n> \n> There is an index on log table, dt field. The index is b-tree.\n> However it doesn't seem to be used. (Of course I have vacuumed it). The\n> table has about 2M records. I don't think that Seq Scan is a good idea.\n> \n> Also, what if I agregate it on dt field to count(*) or sum some values.\n> It would be sequentially scanned, then sorted, then grouped and finally\n> agregated, right?\n\nI assume you have vacuum analyze too? It may help.\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, 25 Mar 1998 00:33:15 -0500 (EST)",
"msg_from": "Bruce Momjian <[email protected]>",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Optimizer fails?"
},
{
"msg_contents": "Michal Mosiewicz asks:\n> xxxx=> \\d logdt\n> \n> Table = logdt\n> +----------------------------------+----------------------------------+-------+\n> | Field | Type |\n> Length|\n> +----------------------------------+----------------------------------+-------+\n> | dt | timestamp \n> | 4 |\n> +----------------------------------+----------------------------------+-------+\n> xxxx=> explain select * from log where dt < '19980203';\n> NOTICE: QUERY PLAN:\n> \n> Seq Scan on log (cost=105832.15 size=699588 width=62)\n> \n> There is an index on log table, dt field. The index is b-tree.\n> However it doesn't seem to be used. (Of course I have vacuumed it). The\n> table has about 2M records. I don't think that Seq Scan is a good idea.\n\nThis is almost certainly correct optimizer behavior, unless there are very few\nrows that would be qualified by \"< '19980203'\". Do you know how many rows\nwere returned?\n\nThe key to all this is that scanning sequentially is very cheap. Disks like\nit, the OS filesystem is optimized for it, striping across multiple devices\nis a big win... It is easy to scan 10 Mb per second on PC class hardward.\n\nUsing an index means using random I/O for each row. Unless the target table\nfits in the buffer cache this means you are limited to examining a few\ndozen rows per second (good disks can sustain about 80 to 100 random IOs\nper second).\n\nBack of the envelope calculation:\n\n Assume\n\n rowsize = 100 bytes.\n Sequential I/O rate = 1Mb/second.\n Random I/O rate = 50 IO/second.\n\n Then\n\n 2M rows @ 10,000 rows per Mb = 200 MB = 200 seconds.\n\n 200 seconds * 50 IO per second = 10,000 IOs\n\n So\n\n If less than 10,000 rows (0.5 % of the table) would qualify, the index\n scan might be faster. Otherwise the table scan is optimal.\n\nThis calculation ignores the overhead of actually scanning the index and\nprobably underestimates the sequential I/O rate a so in real life, the table\nscan is may be even better than this.\n\n\n> Also, what if I agregate it on dt field to count(*) or sum some values.\n> It would be sequentially scanned, then sorted, then grouped and finally\n> agregated, right?\n>\n> Well, sometimes it may be good enough. But if this table is big enough,\n> it would be wiser to use index to select ranges from the table and then\n> agregate those values without sorting. \n\nAssuming a table and query like:\n\n create table foo (d date, m money); - 2 M rows of this, index on d.\n\n select d, sum(m) from foo\n group by d;\n\nI would be very surprised if there existed a better plan than to sort the\ntable and then scan the sorted table (once) to do the grouping and\naggregation.\n\nIf you knew the number of 'd' values was smallish, you might try scanning\nthe table and building a hash containing the d values and aggregates.\n\nI don't know if postgreSQL has this strategy. It is fairly uncommon to\nactually find it implemented in real systems.\n\n> Once I saw index based agregates in the TODO list, but somehow it\n> disappeared.\n\nDo you mean using covering indexes? This would be very worthwhile, but might\nbe a bit of a job.\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": "Tue, 24 Mar 1998 21:48:59 -0800 (PST)",
"msg_from": "[email protected] (David Gould)",
"msg_from_op": false,
"msg_subject": "Re: [HACKERS] Optimizer fails?"
}
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.