Datasets:

wanshenl
/

pgsql-hackers

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 58.4k rows

threads listlengths 1 2.99k
[ { "msg_contents": "\nOops...too late :) Someone close to Tom want to break his fingers for me?\n:)\n\nOn Wed, 28 Oct 1998, Tom Lane wrote:\n\n> Update of /usr/local/cvsroot/pgsql/src/include/catalog\n> In directory hub.org:/tmp/cvs-serv25049\n> \n> Modified Files:\n> \tpg_operator.h \n> Log Message:\n> Fix a veritable boatload of errors in oprcom, oprnegate,\n> oprlsortop and oprrsortop links. There's still a bug involving\n> conflicting definitions for point @ path, but I'm not taking\n> responsibility for deciding which one is right...\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 29 Oct 1998 00:40:28 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [COMMITTERS] 'pgsql/src/include/catalog pg_operator.h'" }, { "msg_contents": "> \n> Oops...too late :) Someone close to Tom want to break his fingers for me?\n> :)\n> \n> On Wed, 28 Oct 1998, Tom Lane wrote:\n> \n> > Update of /usr/local/cvsroot/pgsql/src/include/catalog\n> > In directory hub.org:/tmp/cvs-serv25049\n> > \n> > Modified Files:\n> > \tpg_operator.h \n> > Log Message:\n> > Fix a veritable boatload of errors in oprcom, oprnegate,\n> > oprlsortop and oprrsortop links. There's still a bug involving\n> > conflicting definitions for point @ path, but I'm not taking\n> > responsibility for deciding which one is right...\n> > \n\nWhere is Tom located, anyway.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Oct 1998 00:02:38 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "On Thu, 29 Oct 1998, Bruce Momjian wrote:\n\n> > \n> > Oops...too late :) Someone close to Tom want to break his fingers for me?\n> > :)\n> > \n> > On Wed, 28 Oct 1998, Tom Lane wrote:\n> > \n> > > Update of /usr/local/cvsroot/pgsql/src/include/catalog\n> > > In directory hub.org:/tmp/cvs-serv25049\n> > > \n> > > Modified Files:\n> > > \tpg_operator.h \n> > > Log Message:\n> > > Fix a veritable boatload of errors in oprcom, oprnegate,\n> > > oprlsortop and oprrsortop links. There's still a bug involving\n> > > conflicting definitions for point @ path, but I'm not taking\n> > > responsibility for deciding which one is right...\n> > > \n> \n> Where is Tom located, anyway.\n\n\tNo idea...guess he's safe for now grin\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 29 Oct 1998 01:08:50 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n>> Where is Tom located, anyway.\n\n> \tNo idea...guess he's safe for now grin\n\nI'm in Pittsburgh, PA (as you could've guessed from my email address).\n\nBruce could be on my doorstep in five hours or so, depending on how far\nhe wanted to stretch the speed limit. But I have no idea where the\nrest of yunz are.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Oct 1998 10:46:18 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "> I'm in Pittsburgh, PA (as you could've guessed from my email address).\n> \n> Bruce could be on my doorstep in five hours or so, depending on how far\n> he wanted to stretch the speed limit. But I have no idea where the\n> rest of yunz are.\n> \n> \t\t\tregards, tom lane\n\n Hamburg, Germany.\n\nJan\n\n-- \n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n", "msg_date": "Thu, 29 Oct 1998 17:36:26 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "On Thu, 29 Oct 1998, Tom Lane wrote:\n\n> The Hermit Hacker <[email protected]> writes:\n> >> Where is Tom located, anyway.\n> \n> > \tNo idea...guess he's safe for now grin\n> \n> I'm in Pittsburgh, PA (as you could've guessed from my email address).\n> \n> Bruce could be on my doorstep in five hours or so, depending on how far\n> he wanted to stretch the speed limit. But I have no idea where the\n> rest of yunz are.\n\nI'm 30+ miles north of Detroit so your thumbs are safe from here :)\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n\n", "msg_date": "Thu, 29 Oct 1998 11:48:03 -0500 (EST)", "msg_from": "Vince Vielhaber <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "On Thu, 29 Oct 1998, Tom Lane wrote:\n\n> The Hermit Hacker <[email protected]> writes:\n> >> Where is Tom located, anyway.\n> \n> > \tNo idea...guess he's safe for now grin\n> \n> I'm in Pittsburgh, PA (as you could've guessed from my email address).\n> \n> Bruce could be on my doorstep in five hours or so, depending on how far\n> he wanted to stretch the speed limit. But I have no idea where the\n> rest of yunz are.\n\n\tWolfville, Nova Scotia, Canada...~17hrs from Toronto, driving due\nEast...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 29 Oct 1998 13:32:52 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "> The Hermit Hacker <[email protected]> writes:\n> >> Where is Tom located, anyway.\n> \n> > \tNo idea...guess he's safe for now grin\n> \n> I'm in Pittsburgh, PA (as you could've guessed from my email address).\n> \n> Bruce could be on my doorstep in five hours or so, depending on how far\n> he wanted to stretch the speed limit. But I have no idea where the\n> rest of yunz are.\n\nWe can protect each other, if Marc crosses the border.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Oct 1998 12:50:01 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "Thus spake Tom Lane\n> Bruce could be on my doorstep in five hours or so, depending on how far\n> he wanted to stretch the speed limit. But I have no idea where the\n> rest of yunz are.\n\nToronto, Ontario.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Thu, 29 Oct 1998 13:01:56 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "> But I have no idea where the\n> rest of yunz are.\n\nSouth Pasadena, California...\n\n - Thomas\n", "msg_date": "Thu, 29 Oct 1998 18:23:00 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "> Thus spake Tom Lane\n> > Bruce could be on my doorstep in five hours or so, depending on how far\n> > he wanted to stretch the speed limit. But I have no idea where the\n> > rest of yunz are.\n> \n> Toronto, Ontario.\n> \n\nShould I add locations to the TODO list of developers?\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Oct 1998 13:25:58 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "Then <[email protected]> spoke up and said:\n> Should I add locations to the TODO list of developers?\n\nNo: use them as data for one (or more) of the regression tests...\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=====================================================================", "msg_date": "29 Oct 1998 13:51:05 -0500", "msg_from": "<[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "On Thu, 29 Oct 1998, Bruce Momjian wrote:\n\n> > The Hermit Hacker <[email protected]> writes:\n> > >> Where is Tom located, anyway.\n> > \n> > > \tNo idea...guess he's safe for now grin\n> > \n> > I'm in Pittsburgh, PA (as you could've guessed from my email address).\n> > \n> > Bruce could be on my doorstep in five hours or so, depending on how far\n> > he wanted to stretch the speed limit. But I have no idea where the\n> > rest of yunz are.\n> \n> We can protect each other, if Marc crosses the border.\n\n\tgroan slap forehead First I fear 'rogue cops' on the\nhighways...now I have to fear 'rogue programmers' :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 29 Oct 1998 22:14:04 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "On Thu, 29 Oct 1998, Thomas G. Lockhart wrote:\n\n> > But I have no idea where the\n> > rest of yunz are.\n> \n> South Pasadena, California...\n\n\tEver get cold down there? I need more heat, just haven't figured\nout where to go for it yet :) Wait, you have horses too, don't you?\nHrmmmm.... :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 29 Oct 1998 22:15:01 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" }, { "msg_contents": "On Thu, 29 Oct 1998, Bruce Momjian wrote:\n\n> > Thus spake Tom Lane\n> > > Bruce could be on my doorstep in five hours or so, depending on how far\n> > > he wanted to stretch the speed limit. But I have no idea where the\n> > > rest of yunz are.\n> > \n> > Toronto, Ontario.\n> > \n> \n> Should I add locations to the TODO list of developers?\n\n\tThat would be cool :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 29 Oct 1998 22:15:18 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog\n\tpg_operator.h'" } ]
[ { "msg_contents": "OK, I have done the deed. If I blew it, you can back out the last\nfew commits and then come break my fingers ... but I believe it's\nright.\n\nWhat I found was a boatload of minor typos in operator definitions\nfor lesser-used types. (My faith in the analysis procedure is\nstrengthened by the fact that it didn't gripe about any operators\nfor the thoroughly-shaken-out types like int4 and float8.) The\noperators I had to change were for these types and combinations:\n\tint8\n\tint4 vs. int8, and vice versa\n\tint2 vs. int4, and vice versa\n\tbool < bool, bool > bool\n\ttinterval\n\tpath\n\tpath vs. point\n\tbox vs. point\n\tfloat4 vs. float8, and vice versa\n\tcircle vs. point\n\tlseg\n\nA typical bug is \"int48le claims to be the commutator of int48ge,\nbut it ought to commute with int84ge\". I also saw a lot of confusion\nbetween the commutator and negator roles ... we could do with some\nbetter documentation, I suspect.\n\nThe only regression test that changed results was tinterval, which\nI already knew was broken. This says that the coverage of our\nregression tests for off-the-beaten-path datatypes is not thorough :-(\n\nMost of the checks I used to find bogus entries are embodied in \nthe new regression test opr_sanity --- you can look at that script\nto see whether you agree with the conditions I checked.\n\nOne thing I'm still a little bothered by is that I don't understand\nthe difference between the oprlsortop and oprrsortop columns. In\nall the existing tuples that weren't demonstrably bogus, the two\ncolumns have the same value. What's the point of distinguishing\nthem, and do the conditions expressed in opr_sanity need refinement?\n\nNote that these changes require an initdb to be effective. Worse,\nthere's still one initdb-forcing change to come, namely resolving\nwhich operator really is point @ path. The opr_sanity regression\ntest will keep failing until we fix this ;-). So, the geometric\noperator gurus had better get together and decide which it should\nbe, pronto.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Oct 1998 00:26:47 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Post-mortem: Another bug in pg_operator.h" } ]
[ { "msg_contents": "\n> You might be tempted to run configure with the additional argument\n> --with-CC='cc -W0' to activate the native C-compiler. However, when I\n> did this, compilation stopped with this error message.\n> \n> cc -W0 -I../../../include -I../../../backend -I/home/tools/include\n> -I../.. -c istrat.c -o istrat.o\n> istrat.c 496: [error]: 2324 Undefined: 'F_OIDEQ'\n> 2086 c1: errors: 1, warnings: 15\n\nThis is the symptom, when $(CPP) fails to take it's input from stdin. \n(F_OIDEQ should be defined in fmgr.h)\nLooks like Tom Lane already has a fix, that will be in 6.4 final.\n\nAndreas\n\n", "msg_date": "Thu, 29 Oct 1998 09:35:20 +0100", "msg_from": "Andreas Zeugswetter <[email protected]>", "msg_from_op": true, "msg_subject": "AW: [HACKERS] Last call? (Siemens RM Port)" } ]
[ { "msg_contents": "\nThe network.c from current CVS has:\n\nxlc -I../../../include -I../../../backend -I/usr/local/include \n-qchars=signed -qmaxmem=8192 -qhalt=w -qsrcmsg -qcheck=divzero -qlonglong \n -g -qfullpath -I../.. -c network.c -o network.o\n 400 \| if (inet_cidr_ntop(2, &addr, (((inet_struct \n)(((struct varlena )(ip))->vl_dat))->bits), tmp, sizeof(tmp)) < 0)\n \n ............................................................. \n...............................................................a...\na - 1506-068 (S) Operation between types \"char\" and \"int\" is not allowed.\ngmake[3]: [network.o] Error 1\n\nIt should probably read as follows: (but please check)\n(I sure hope it comes without those stupid wraps my mailer always \ngenerates)\n\n* network.c.ori\tThu Oct 29 09:51:42 1998\n--- network.c\tThu Oct 29 10:35:59 1998\n*************\n* 397,403 ***\n \t\t/ It's an IP V4 address: /\n \t\tint\taddr = htonl(ntohl(ip_v4addr(ip)) & (0xffffffff << (32 - \nip_bits(ip))));\n\n! \t\tif (inet_cidr_ntop(AF_INET, &addr, ip_bits(ip), tmp, sizeof(tmp)) < 0)\n \t\t{\n \t\t\telog(ERROR, \"unable to print network (%s)\", strerror(errno));\n \t\t\treturn (NULL);\n--- 397,403 ----\n \t\t/ It's an IP V4 address: /\n \t\tint\taddr = htonl(ntohl(ip_v4addr(ip)) & (0xffffffff << (32 - \nip_bits(ip))));\n\n! \t\tif (inet_cidr_ntop(AF_INET, &addr, ip_bits(ip), tmp, sizeof(tmp)) == \nNULL)\n \t\t{\n \t\t\telog(ERROR, \"unable to print network (%s)\", strerror(errno));\n \t\t\treturn (NULL);\n\n\nAttached is also a correct inet.out\n\n \n\nbegin 600 inet.out\nM455%4EDZ($123U @5$%\"3$4@24Y%5%]40DP[\"D524D]2.B @4F5L871I;VX@\nM:6YE=%]T8FP@1&]E<R!.;W0@17AI<W0A\"E%515)9.B!#4D5!5$4@5$%\"3$4@\nM24Y%5%]40DP@&,@8VED<BP@:2!I;F5T3L455%4EDZ($E.4T525\"!)3E1/\nM($E.151?5$),(\"AC+\"!I2!604Q515,@\"<Q.3(N,38X+C$G+\" G,3DR+C$V\nM.\"XQ+C(R-B\\R-\"<I.PI1545263H@24Y315)4($E.5$\\@24Y%5%]40DP@&,L\nM(&DI(%9!3%5%4R H)S$Y,BXQ-C@N,2XR+S(T)RP@)S$Y,BXQ-C@N,2XR,C8G\nM3L455%4EDZ($E.4T525\"!)3E1/($E.151?5$),(\"AC+\"!I2!604Q515,@\nM\"<Q,\"<L(\"<Q,\"XQ+C(N,R\\X)RD[\"E%515)9.B!)3E-%4E0@24Y43R!)3D54\nM7U1\"3\" H8RP@:2D@5D%,5453(\"@G,3 N,\"XP+C G+\" G,3 N,2XR+C,O.\"<I\nM.PI1545263H@24Y315)4($E.5$\\@24Y%5%]40DP@&,L(&DI(%9!3%5%4R H\nM)S$P+C$N,BXS)RP@)S$P+C$N,BXS+S,R)RD[\"E%515)9.B!)3E-%4E0@24Y4\nM3R!)3D547U1\"3\" H8RP@:2D@5D%,5453(\"@G,3 N,2XR)RP@)S$P+C$N,BXS\nM+S(T)RD[\"E%515)9.B!)3E-%4E0@24Y43R!)3D547U1\"3\" H8RP@:2D@5D%,\nM5453(\"@G,3 N,2<L(\"<Q,\"XQ+C(N,R\\Q-B<I.PI1545263H@24Y315)4($E.\nM5$\\@24Y%5%]40DP@&,L(&DI(%9!3%5%4R H)S$P)RP@)S$P+C$N,BXS+S@G\nM3L455%4EDZ(%-%3$5#5\" G)R!A<R!E:6=H=\"P@8R!A<R!C:61R+\"!I(&%S\nM(&EN970@1E)/32!)3D547U1\"3#L96EG:'1\\8VED<B @(\" @(\" @?&EN970@\nM(\" @(\" @(\" @(\" +2TM+2TK+2TM+2TM+2TM+2TMRTM+2TM+2TM+2TM+2TM\nM+2T(\" @(\"!\\,3DR+C$V.\"XQ+S(T?#$Y,BXQ-C@N,2XR,C8O,C0(\" @(\"!\\\nM,3DR+C$V.\"XQ+S(T?#$Y,BXQ-C@N,2XR,C8@(\" (\" @(\"!\\,3 O.\" @(\" @\nM(\" @?#$P+C$N,BXS+S@@(\" @(\" (\" @(\"!\\,3 N,\"XP+C O,S(@?#$P+C$N\nM,BXS+S@@(\" @(\" (\" @(\"!\\,3 N,2XR+C,O,S(@?#$P+C$N,BXS(\" @(\" @\nM(\" (\" @(\"!\\,3 N,2XR+S(T(\" @?#$P+C$N,BXS+S(T(\" @(\" (\" @(\"!\\\nM,3 N,2\\Q-B @(\" @?#$P+C$N,BXS+S$V(\" @(\" (\" @(\"!\\,3 O.\" @(\" @\nM(\" @?#$P+C$N,BXS+S@@(\" @(\" *#@@<F]W<RD\"E%515)9.B!314Q%0U0@\nM)R<@87,@96EG:'0L(&D@87,@:6YE=\"P@:&]S=\"AI2!&4D]-($E.151?5$),\nM.PIE:6=H='QI;F5T(\" @(\" @(\" @(\" @?\" @(\" @(\" @(&AO<W0+2TM+2TK\nM+2TM+2TM+2TM+2TM+2TM+2LM+2TM+2TM+2TM+2TM\"B @(\" @?#$Y,BXQ-C@N\nM,2XR,C8O,C1\\,3DR+C$V.\"XQ+C(R-@H@(\" @('PQ.3(N,38X+C$N,C(V(\" @\nM?#$Y,BXQ-C@N,2XR,C8(\" @(\"!\\,3 N,2XR+C,O.\" @(\" @('P@(\" @(#$P\nM+C$N,BXS\"B @(\" @?#$P+C$N,BXS+S@@(\" @(\"!\\(\" @(\" Q,\"XQ+C(N,PH@\nM(\" @('PQ,\"XQ+C(N,R @(\" @(\" @?\" @(\" @,3 N,2XR+C,(\" @(\"!\\,3 N\nM,2XR+C,O,C0@(\" @('P@(\" @(#$P+C$N,BXS\"B @(\" @?#$P+C$N,BXS+S$V\nM(\" @(\"!\\(\" @(\" Q,\"XQ+C(N,PH@(\" @('PQ,\"XQ+C(N,R\\X(\" @(\" @?\" @\nM(\" @,3 N,2XR+C,*#@@<F]W<RD\"E%515)9.B!314Q%0U0@)R<@87,@96EG\nM:'0L(&,@87,@8VED<BP@8G)O861C87-T&,I+ H@(&D@87,@:6YE=\"P@8G)O\nM861C87-T&DI($923TT@24Y%5%]40DP[\"F5I9VAT?&-I9'(@(\" @(\" @('P@\nM(\" @(&)R;V%D8V%S='QI;F5T(\" @(\" @(\" @(\" @?\" @(\" @8G)O861C87-T\nM\"BTM+2TMRTM+2TM+2TM+2TM+2LM+2TM+2TM+2TM+2TM+2LM+2TM+2TM+2TM\nM+2TM+2TMRTM+2TM+2TM+2TM+2TM\"B @(\" @?#$Y,BXQ-C@N,2\\R-'P@,3DR\nM+C$V.\"XQ+C(U-7PQ.3(N,38X+C$N,C(V+S(T?\" Q.3(N,38X+C$N,C4U\"B @\nM(\" @?#$Y,BXQ-C@N,2\\R-'P@,3DR+C$V.\"XQ+C(U-7PQ.3(N,38X+C$N,C(V\nM(\" @?\" Q.3(N,38X+C$N,C(V\"B @(\" @?#$P+S@@(\" @(\" @('PQ,\"XR-34N\nM,C4U+C(U-7PQ,\"XQ+C(N,R\\X(\" @(\" @?#$P+C(U-2XR-34N,C4U\"B @(\" @\nM?#$P+C N,\"XP+S,R('P@(\" @(\" Q,\"XP+C N,'PQ,\"XQ+C(N,R\\X(\" @(\" @\nM?#$P+C(U-2XR-34N,C4U\"B @(\" @?#$P+C$N,BXS+S,R('P@(\" @(\" Q,\"XQ\nM+C(N,WPQ,\"XQ+C(N,R @(\" @(\" @?\" @(\" @(#$P+C$N,BXS\"B @(\" @?#$P\nM+C$N,B\\R-\" @('P@(\" @,3 N,2XR+C(U-7PQ,\"XQ+C(N,R\\R-\" @(\" @?\" @\nM(\" Q,\"XQ+C(N,C4U\"B @(\" @?#$P+C$O,38@(\" @('P@(#$P+C$N,C4U+C(U\nM-7PQ,\"XQ+C(N,R\\Q-B @(\" @?\" @,3 N,2XR-34N,C4U\"B @(\" @?#$P+S@@\nM(\" @(\" @('PQ,\"XR-34N,C4U+C(U-7PQ,\"XQ+C(N,R\\X(\" @(\" @?#$P+C(U\nM-2XR-34N,C4U\"B@X(')O=W,I\"@I1545263H@4T5,14-4(\"<G(&%S(&5I9VAT\nM+\"!C(&%S(&-I9'(L(&YE='=O<FLH8RD@87,@(FYE='=O<FLH8VED<BDB+ H@\nM(&D@87,@:6YE=\"P@;F5T=V]R:RAI2!A<R B;F5T=V]R:RAI;F5T2(@1E)/\nM32!)3D547U1\"3#L96EG:'1\\8VED<B @(\" @(\" @?&YE='=O<FLH8VED<BE\\\nM:6YE=\" @(\" @(\" @(\" @('QN971W;W)K&EN970I(\" @\"BTM+2TMRTM+2TM\nM+2TM+2TM+2LM+2TM+2TM+2TM+2TMRTM+2TM+2TM+2TM+2TM+2TK+2TM+2TM\nM+2TM+2TM+2TM+0H@(\" @('PQ.3(N,38X+C$O,C1\\,3DR+C$V.\"XQ+S(T('PQ\nM.3(N,38X+C$N,C(V+S(T?#$Y,BXQ-C@N,2\\R-\" @(\" (\" @(\"!\\,3DR+C$V\nM.\"XQ+S(T?#$Y,BXQ-C@N,2\\R-\"!\\,3DR+C$V.\"XQ+C(R-B @('PQ.3(N,38X\nM+C$N,C(V+S,R\"B @(\" @?#$P+S@@(\" @(\" @('PQ,\"\\X(\" @(\" @(\" @?#$P\nM+C$N,BXS+S@@(\" @(\"!\\,3 O.\" @(\" @(\" @(\" @( H@(\" @('PQ,\"XP+C N\nM,\"\\S,B!\\,3 N,\"XP+C O,S(@('PQ,\"XQ+C(N,R\\X(\" @(\" @?#$P+S@@(\" @\nM(\" @(\" @(\" (\" @(\"!\\,3 N,2XR+C,O,S(@?#$P+C$N,BXS+S,R(\"!\\,3 N\nM,2XR+C,@(\" @(\" @('PQ,\"XQ+C(N,R\\S,B @(\" @\"B @(\" @?#$P+C$N,B\\R\nM-\" @('PQ,\"XQ+C(O,C0@(\" @?#$P+C$N,BXS+S(T(\" @(\"!\\,3 N,2XR+S(T\nM(\" @(\" @( H@(\" @('PQ,\"XQ+S$V(\" @(\"!\\,3 N,2\\Q-B @(\" @('PQ,\"XQ\nM+C(N,R\\Q-B @(\" @?#$P+C$O,38@(\" @(\" @(\" (\" @(\"!\\,3 O.\" @(\" @\nM(\" @?#$P+S@@(\" @(\" @(\"!\\,3 N,2XR+C,O.\" @(\" @('PQ,\"\\X(\" @(\" @\nM(\" @(\" @\"B@X(')O=W,I\"@I1545263H@4T5,14-4(\"<G(&%S(&5I9VAT+\"!C\nM(&%S(&-I9'(L(&UA<VML96XH8RD@87,@(FUA<VML96XH8VED<BDB+ H@(&D@\nM87,@:6YE=\"P@;6%S:VQE;BAI2!A<R B;6%S:VQE;BAI;F5T2(@1E)/32!)\nM3D547U1\"3#L96EG:'1\\8VED<B @(\" @(\" @?&UA<VML96XH8VED<BE\\:6YE\nM=\" @(\" @(\" @(\" @('QM87-K;&5N&EN970I\"BTM+2TMRTM+2TM+2TM+2TM\nM+2LM+2TM+2TM+2TM+2TMRTM+2TM+2TM+2TM+2TM+2TK+2TM+2TM+2TM+2TM\nM+0H@(\" @('PQ.3(N,38X+C$O,C1\\(\" @(\" @(\" @(\" R-'PQ.3(N,38X+C$N\nM,C(V+S(T?\" @(\" @(\" @(\" @,C0(\" @(\"!\\,3DR+C$V.\"XQ+S(T?\" @(\" @\nM(\" @(\" @,C1\\,3DR+C$V.\"XQ+C(R-B @('P@(\" @(\" @(\" @(#,R\"B @(\" @\nM?#$P+S@@(\" @(\" @('P@(\" @(\" @(\" @(\" X?#$P+C$N,BXS+S@@(\" @(\"!\\\nM(\" @(\" @(\" @(\" @. H@(\" @('PQ,\"XP+C N,\"\\S,B!\\(\" @(\" @(\" @(\" S\nM,GPQ,\"XQ+C(N,R\\X(\" @(\" @?\" @(\" @(\" @(\" @(#@(\" @(\"!\\,3 N,2XR\nM+C,O,S(@?\" @(\" @(\" @(\" @,S)\\,3 N,2XR+C,@(\" @(\" @('P@(\" @(\" @\nM(\" @(#,R\"B @(\" @?#$P+C$N,B\\R-\" @('P@(\" @(\" @(\" @(#(T?#$P+C$N\nM,BXS+S(T(\" @(\"!\\(\" @(\" @(\" @(\" R- H@(\" @('PQ,\"XQ+S$V(\" @(\"!\\\nM(\" @(\" @(\" @(\" Q-GPQ,\"XQ+C(N,R\\Q-B @(\" @?\" @(\" @(\" @(\" @,38\nM(\" @(\"!\\,3 O.\" @(\" @(\" @?\" @(\" @(\" @(\" @(#A\\,3 N,2XR+C,O.\" @\nM(\" @('P@(\" @(\" @(\" @(\" X\"B@X(')O=W,I\"@I1545263H@4T5,14-4(\"<G\nM(&%S('1W;RP@8R!A<R!C:61R+\"!M87-K;&5N&,I(&%S(\")M87-K;&5N&-I\nM9'(I(BP(\"!I(&%S(&EN970L(&UA<VML96XH:2D@87,@(FUA<VML96XH:6YE\nM=\"DB($923TT@24Y%5%]40DP(\"!72$5212!M87-K;&5N&,I(#P](#@[\"G1W\nM;WQC:61R?&UA<VML96XH8VED<BE\\:6YE=\" @(\" @('QM87-K;&5N&EN970I\nM\"BTM+2LM+2TMRTM+2TM+2TM+2TM+2TK+2TM+2TM+2TM+2LM+2TM+2TM+2TM\nM+2TM\"B @('PQ,\"\\X?\" @(\" @(\" @(\" @(#A\\,3 N,2XR+C,O.'P@(\" @(\" @\nM(\" @(\" X\"B @('PQ,\"\\X?\" @(\" @(\" @(\" @(#A\\,3 N,2XR+C,O.'P@(\" @\nM(\" @(\" @(\" X\"B@R(')O=W,I\"@I1545263H@4T5,14-4(\"<G(&%S('-I>\"P@\nM8R!A<R!C:61R+\"!I(&%S(&EN970@1E)/32!)3D547U1\"3 H@(%=(15)%(&,@\nM/2!I.PIS:7A\\8VED<B @(\" @(\" @?&EN970@(\" @(\" @(\" @(\" +2TMRTM\nM+2TM+2TM+2TM+2LM+2TM+2TM+2TM+2TM+2TM\"B @('PQ.3(N,38X+C$O,C1\\\nM,3DR+C$V.\"XQ+C(R-B\\R- H@(\"!\\,3 O.\" @(\" @(\" @?#$P+C$N,BXS+S@@\nM(\" @(\" (\" @?#$P+C$N,BXS+S,R('PQ,\"XQ+C(N,R @(\" @(\" @\"B @('PQ\nM,\"XQ+C(O,C0@(\"!\\,3 N,2XR+C,O,C0@(\" @( H@(\"!\\,3 N,2\\Q-B @(\" @\nM?#$P+C$N,BXS+S$V(\" @(\" (\" @?#$P+S@@(\" @(\" @('PQ,\"XQ+C(N,R\\X\n1(\" @(\" @\"B@V(')O=W,I\"@H\n`\nend\n\n", "msg_date": "Thu, 29 Oct 1998 11:09:05 +0100", "msg_from": "Andreas Zeugswetter <[email protected]>", "msg_from_op": true, "msg_subject": "Bug in current network.c" }, { "msg_contents": "Thus spake Andreas Zeugswetter\n> a - 1506-068 (S) Operation between types \"char\" and \"int\" is not allowed.\n> gmake[3]: ** [network.o] Error 1\n> \n> It should probably read as follows: (but please check)\n> (I sure hope it comes without those stupid wraps my mailer always \n> generates)\n\nFor some reason network.c reverted to an earlier version in the tree. Your\npatch was already submitted by someone (Bruce?) but the changes, like my\nother changes, were lost. I assume that your patch is correct but the\nperson who submitted the previous patch should double check that nothing\nelse is missing.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Thu, 29 Oct 1998 08:15:31 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Bug in current network.c" }, { "msg_contents": "[email protected] (D'Arcy J.M. Cain) writes:\n> For some reason network.c reverted to an earlier version in the tree. Your\n> patch was already submitted by someone (Bruce?) but the changes, like my\n> other changes, were lost.\n\nIt looks like Bruce copied-and-pasted a little too much while applying\nyour last patch. I had fixed the \"< 0\" tests that ought to be \"== NULL\"\non Sunday, but one of them was on a line that you wanted changed for other\nreasons, and it went back to being wrong...\n\nFixed now.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Oct 1998 11:16:30 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Bug in current network.c " }, { "msg_contents": "Thus spake Tom Lane\n> [email protected] (D'Arcy J.M. Cain) writes:\n> > For some reason network.c reverted to an earlier version in the tree. Your\n> > patch was already submitted by someone (Bruce?) but the changes, like my\n> > other changes, were lost.\n> \n> It looks like Bruce copied-and-pasted a little too much while applying\n> your last patch. I had fixed the \"< 0\" tests that ought to be \"== NULL\"\n> on Sunday, but one of them was on a line that you wanted changed for other\n> reasons, and it went back to being wrong...\n\nTrust us. We're professionals. :-)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Thu, 29 Oct 1998 12:59:33 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Bug in current network.c" } ]
[ { "msg_contents": "\n> Would you check that things build on AIX with the attached files\n> inserted?\n\nI could not read your attachment, but I am currently compiling current CVS, \nand see that the change is in there (including configure).\nIt should also fix the Siemens RM problem that needs gcc for configure.\n\nAnd yes, it does work like shrink wrapped :-) Thanks\n\nAndreas\n\n\n", "msg_date": "Thu, 29 Oct 1998 11:10:41 +0100", "msg_from": "Andreas Zeugswetter <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] configure and cpp detection " } ]
[ { "msg_contents": "Hello!\n\nI tried postgresql.v6.4-BETA3 and found some problems especially\nwith FreeBSD.\nThe following is my environment:\n\tFreeBSD 2.2.7-RELEASE\n\tGNU Make version 3.76.1, by Richard Stallman and Roland McGrath.\n\tgcc version 2.7.2.1\n\tGNU Bison version 1.25\n\tflex version 2.5.4\n\n* makefiles/Makefile.freebsd\n\tIt should have \"-Bforcearchive\" option just like as\n\tmakefiles/Makefile.bsd (and it used to be).\n\tWithout it, I got such errors in test/regress/results/misc.out.\n\n========================================================================\nERROR: Can't find function reverse_name in file /usr/local/src/postgresql.v6.4-BETA3/pgsql/src/test/regress/input/../regress.so\n========================================================================\n\n\t# And the original Makefile.freebsd does not seem to work\n\t# as you intended. As `\\' overrides `#', the last 2 lines\n\t# are commented out together.\n\n\n* Makefile.shlib\n\tFreeBSD is not supported.\n\tSo no shared library for plpgsql is installed and regression\n\ttest for it fails.\n\n========================================================================\ngmake -C src install\ngmake[3]: Entering directory `/usr/local/src/postgresql.v6.4-BETA3/pgsql/src/pl/\nplpgsql/src'\n/usr/bin/install -c -m 644 /usr/local/pgsql/lib/plpgsql.so\nusage: install [-CcDps] [-f flags] [-g group] [-m mode] [-o owner] file1 file2\n install [-CcDps] [-f flags] [-g group] [-m mode] [-o owner] file1 ...\n fileN directory\n install -d [-g group] [-m mode] [-o owner] directory ...\ngmake[3]: *** [install] Error 64\n========================================================================\n\n\tAnd no shared libraries for libecpg, libpq++ and libpq.a\n\tare made and installed.\n\n\n* pg_dumpall\n\tPg_dumpall command outputs \"create database\" statement such as:\n========================================================================\ncreate database with encoding='EUC_JP' DBNAME;\n========================================================================\n\tBut it should be:\n========================================================================\ncreate database DBNAME with encoding='EUC_JP';\n========================================================================\n\n\nRegards\n\n-- \nASCII CORPORATION\nTechnical Center\nSHIOZAKI Takehiko\n<[email protected]>\n", "msg_date": "Thu, 29 Oct 1998 19:29:48 +0900 (JST)", "msg_from": "SHIOZAKI Takehiko <[email protected]>", "msg_from_op": true, "msg_subject": "v6.4-BETA3 problem with FreeBSD" } ]
[ { "msg_contents": "Hi, all.\n\nI am testing 6.4 on a DU 4.0d system using DEC's (well, Compaq's) C\ncompiler, gmake 3.75, bison 1.25 and flex 2.5.4, and I've found two\nproblems.\n\n1 - The old \"configure --with-CC=cc\" problem is still there: configure\nfails if you specify --with-CC=cc when it is going to test whether the C\ncompiler needs the \"-traditional\" flag. I think I remember it was due to a\nbug in autoconf itself, so there isn't much to be done, except patch the\nconfigure script by hand before making the release. I can supply a patch,\nif needed.\n\n2 - I get the following error when compiling:\n\ncc -I../../../include -I../../../backend -I/usr/local/include -DNOFIXADE -O4 -Olimit 2000 -I../.. -c istrat.c -o istrat.o\ncc: Error: istrat.c, line 495: In this statement, \"F_OIDEQ\" is not declared.\n F_OIDEQ,\n-----------------------------------------------------------^\ncc: Error: istrat.c, line 561: In this statement, \"F_OIDEQ\" is not declared.\n F_OIDEQ,\n-----------------------------------------------------------^\ncc: Error: istrat.c, line 604: In this statement, \"F_OIDEQ\" is not declared.\n F_OIDEQ,\n-----------------------------------------------------------^\ncc: Error: istrat.c, line 642: In this statement, \"F_OIDEQ\" is not declared.\n F_OIDEQ,\n---------------------------------------------------^\ngmake[3]: * [istrat.o] Error 1\ngmake[3]: Leaving directory\n`/usr/local/pgsql.beta/src/pgsql/src/backend/access/\nindex'\ngmake[2]: * [submake] Error 2\ngmake[2]: Leaving directory\n`/usr/local/pgsql.beta/src/pgsql/src/backend/access'\ngmake[1]: * [access.dir] Error 2\ngmake[1]: Leaving directory `/usr/local/pgsql.beta/src/pgsql/src/backend'\ngmake: * [all] Error 2\n\nIt bombs compiling backend/access/index/istrat.c. I have searched all the\noccurences of \"F_OIDEQ\" in all files (that is all files, not just .c or\n.h), and found it in the following ones:\n\n./backend/access/index/istrat.c\n./backend/catalog/heap.c\n./backend/catalog/index.c\n./backend/catalog/indexing.c\n./backend/catalog/pg_operator.c\n./backend/commands/remove.c\n./backend/commands/trigger.c\n./backend/commands/vacuum.c\n./backend/executor/execUtils.c\n./backend/optimizer/path/indxpath.c\n./backend/optimizer/util/plancat.c\n./backend/parser/parse_func.c\n./backend/rewrite/rewriteRemove.c\n./backend/utils/adt/selfuncs.c\n./backend/utils/cache/catcache.c\n./backend/utils/cache/relcache.c\n\nAs you can see, F_OIDEQ isn't defined in any header file nor in any other\nkind of config file. I've also searched in /usr/include/... and\n/usr/local/include..., just in case.\n\nAny hints?\n\nP.S.: I've also noticed that there is a \"F_OID8EQ\" that's found only in\nbackend/catalog/indexing.c and backend/utils/cache/catcache.c.\n\n-------------------------------------------------------------------\nPedro Josï¿½ Lobo Perea Tel: +34 91 336 78 19\nCentro de Cï¿½lculo Fax: +34 91 331 92 29\nEUIT Telecomunicaciï¿½n - UPM e-mail: [email protected]\n\n", "msg_date": "Thu, 29 Oct 1998 12:30:43 +0100 (MET)", "msg_from": "\"Pedro J. Lobo\" <[email protected]>", "msg_from_op": true, "msg_subject": "6.4 BETA2 fails to compile on Digital Unix 4.0d" }, { "msg_contents": "> I am testing 6.4 on a DU 4.0d system using DEC's (well, Compaq's) C\n> compiler, gmake 3.75, bison 1.25 and flex 2.5.4, and I've found two\n> problems.\n> Any hints?\n\nI'm guessing that you have a problem with having your \"cpp\" preprocessor\npick up input from a pipe (stdin). Just yesterday Tom Lane fixed that by\nno longer requiring cpp or equivalent to do that. Can you look at your\nconfigure results and verify that the conclusions it reaches about CPP\nand CPPSTDIN are correct?\n\nThe two scripts which use it are src/backend/utils/Gen_fmgrtab.sh and\nsrc/backend/catalog/genbki.sh. Go to src/backend/utils and type \"make\"\nand see if files are made as you would expect. If there is a problem,\nthen perhaps Tom can send you new versions, or you can hand-modify them\nyourself, or you can wait for a new beta. In Gen_fmgrtab.sh.in the new\ncode looks like:\n\n42a43\n> CPPTMPFILE=fmgrtmp.c\n65,66c66,68\n< $4 == \"11\" { print; next; }' \| \\\n< @CPP@ @CPPSTDIN@ $BKIOPTS \| \\\n---\n> $4 == \"11\" { print; next; }' > $CPPTMPFILE\n>\n> @CPP@ $BKIOPTS $CPPTMPFILE \| \\\n69a72,73\n> rm -f $CPPTMPFILE\n\nto have cpp read from file fmgrtmp.c rather than from stdin.\n\n - Tom\n", "msg_date": "Thu, 29 Oct 1998 14:35:30 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d" }, { "msg_contents": "\"Pedro J. Lobo\" <[email protected]> writes:\n> 1 - The old \"configure --with-CC=cc\" problem is still there: configure\n> fails if you specify --with-CC=cc when it is going to test whether the C\n> compiler needs the \"-traditional\" flag. I think I remember it was due to a\n> bug in autoconf itself, so there isn't much to be done, except patch the\n> configure script by hand before making the release. I can supply a patch,\n> if needed.\n\nCan you send me the details on this one? I'm not real eager for us to\ntry to apply a hand patch to configure (it would get lost, because we\nchange configure.in and rerun autoconf every few days, seems like).\nBut maybe there is another way. In any case, I can try to make sure\nthat the next release of Autoconf fixes the underlying problem.\n\n> 2 - I get the following error when compiling:\n\n> cc -I../../../include -I../../../backend -I/usr/local/include -DNOFIXADE -O4 -Olimit 2000 -I../.. -c istrat.c -o istrat.o\n> cc: Error: istrat.c, line 495: In this statement, \"F_OIDEQ\" is not declared.\n\nThis is a symptom of incorrect configure guess about how to do\ncpp-from-stdin ... the file that should contain the definition of\nF_OIDEQ is winding up empty :-(\n\nIt should be fixed in the current CVS sources, or BETA4 whenever that's\nout.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Oct 1998 10:58:49 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d " }, { "msg_contents": "On Thu, 29 Oct 1998, Thomas G. Lockhart wrote:\n\n>> I am testing 6.4 on a DU 4.0d system using DEC's (well, Compaq's) C\n>> compiler, gmake 3.75, bison 1.25 and flex 2.5.4, and I've found two\n>> problems.\n>> Any hints?\n>\n>I'm guessing that you have a problem with having your \"cpp\" preprocessor\n>pick up input from a pipe (stdin). Just yesterday Tom Lane fixed that by\n>no longer requiring cpp or equivalent to do that. Can you look at your\n>configure results and verify that the conclusions it reaches about CPP\n>and CPPSTDIN are correct?\n\nIt seems that you won the prize :-) I have patched Gen_fmgrtab.sh.in and\ngenbki.sh.in and executed configure again, and now it is compiling fine.\n\nOne \"minor\" problem is that I am getting many warnings complaining that\nMAP_FILE is being redefined. Indeed, it is defined in\n/usr/include/sys/mman.h (a system header file) for being used with mmap\nand friends, and is redefined in include/libpq/hba.h. Since hba.h is\nparsed after mmap.h and the preprocessor uses the last value seen\neverything is allright, but I think that MAP_FILE should be renamed to\nsomething like PG_MAP_FILE. Maybe it can wait for 6.4.1 or 6.5, but it\nshould be done.\n\nI'll post a complete report if/when the build ends and (hopefully) the\nregression tests are completed.\n\n-------------------------------------------------------------------\nPedro Josï¿½ Lobo Perea Tel: +34 91 336 78 19\nCentro de Cï¿½lculo Fax: +34 91 331 92 29\nEUIT Telecomunicaciï¿½n - UPM e-mail: [email protected]\n\n\n\n", "msg_date": "Thu, 29 Oct 1998 17:40:58 +0100 (MET)", "msg_from": "\"Pedro J. Lobo\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d" }, { "msg_contents": "On Thu, 29 Oct 1998, Tom Lane wrote:\n\n>\"Pedro J. Lobo\" <[email protected]> writes:\n>> 1 - The old \"configure --with-CC=cc\" problem is still there: configure\n>> fails if you specify --with-CC=cc when it is going to test whether the C\n>> compiler needs the \"-traditional\" flag. I think I remember it was due to a\n>> bug in autoconf itself, so there isn't much to be done, except patch the\n>> configure script by hand before making the release. I can supply a patch,\n>> if needed.\n>\n>Can you send me the details on this one? I'm not real eager for us to\n>try to apply a hand patch to configure (it would get lost, because we\n>change configure.in and rerun autoconf every few days, seems like).\n\nYes, it would be a real pain. I was considering it only as the last\noption.\n\n>But maybe there is another way. In any case, I can try to make sure\n>that the next release of Autoconf fixes the underlying problem.\n\nOk, here's what happens: the line \"AC_PROG_GCC_TRADITIONAL\" in\nconfigure.in expands to the following in configure:\n\nif test $ac_cv_prog_gcc = yes; then\n echo $ac_n \"checking whether ${CC-cc} needs -traditional\"\"... $ac_c\" 1>&6\necho \"configure:3961: checking whether ${CC-cc} needs -traditional\" >&5\nif eval \"test \\\"`echo '$''{'ac_cv_prog_gcc_traditional'+set}'`\\\" = set\"; then\n[ more follows ]\n\nThe problem is that $ac_cv_prog_gcc is empty if you specify\n\"--with-CC=\", so the first line expands to \"if test = yes; then\", which\nis clearly incorrect.\n\nMy workaround is to substitute the first line with \"if test\nx$ac_cv_prog_gcc = xyes; then\" in configure itself. Dirty, but it works.\n\n>> 2 - I get the following error when compiling:\n>\n>> cc -I../../../include -I../../../backend -I/usr/local/include -DNOFIXADE -O4 -Olimit 2000 -I../.. -c istrat.c -o istrat.o\n>> cc: Error: istrat.c, line 495: In this statement, \"F_OIDEQ\" is not declared.\n>\n>This is a symptom of incorrect configure guess about how to do\n>cpp-from-stdin ... the file that should contain the definition of\n>F_OIDEQ is winding up empty :-(\n\nYes, that was the problem. I fixed Gen_fmgrtab.sh.in and genbki.in as\nThomas suggested, and the thing compled (mostly) fine.\n\n>It should be fixed in the current CVS sources, or BETA4 whenever that's\n>out.\n\nI don't know if I will be able to test BETA4, because I'm leaving for a\nshort vacation (lucky me :-) on Saturday and won't be back until Thursday.\nHowever, it should work as expected.\n\n-------------------------------------------------------------------\nPedro Josï¿½ Lobo Perea Tel: +34 91 336 78 19\nCentro de Cï¿½lculo Fax: +34 91 331 92 29\nEUIT Telecomunicaciï¿½n - UPM e-mail: [email protected]\n\n", "msg_date": "Thu, 29 Oct 1998 18:24:00 +0100 (MET)", "msg_from": "\"Pedro J. Lobo\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d " }, { "msg_contents": "On Thu, 29 Oct 1998, Tom Lane wrote:\n\n> \"Pedro J. Lobo\" <[email protected]> writes:\n> > 1 - The old \"configure --with-CC=cc\" problem is still there: configure\n> > fails if you specify --with-CC=cc when it is going to test whether the C\n> > compiler needs the \"-traditional\" flag. I think I remember it was due to a\n> > bug in autoconf itself, so there isn't much to be done, except patch the\n> > configure script by hand before making the release. I can supply a patch,\n> > if needed.\n> \n> Can you send me the details on this one? I'm not real eager for us to\n> try to apply a hand patch to configure (it would get lost, because we\n> change configure.in and rerun autoconf every few days, seems like).\n> But maybe there is another way. In any case, I can try to make sure\n> that the next release of Autoconf fixes the underlying problem.\n\n\tIf its a bug in autoconf itself, we can patch the autoconf .m4\nfiles directly on Hub.Org, and just make sure we autoconf with that...I\nhave no problemsm with that...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 29 Oct 1998 13:31:56 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d " }, { "msg_contents": "\"Pedro J. Lobo\" <[email protected]> writes:\n> Ok, here's what happens: the line \"AC_PROG_GCC_TRADITIONAL\" in\n> configure.in expands to the following in configure:\n\n> if test $ac_cv_prog_gcc = yes; then\n\n> The problem is that $ac_cv_prog_gcc is empty if you specify\n> \"--with-CC=\", so the first line expands to \"if test = yes; then\", which\n> is clearly incorrect.\n\nAh. OK, Autoconf is maybe not being too robust here, but it's our bug.\nAC_PROG_GCC_TRADITIONAL assumes you've run AC_PROG_CC, and we don't do\nthat if --with-CC is given.\n\nI will rearrange the configure script so that that's true. As far as I\ncan see, there's no reason to run AC_PROG_GCC_TRADITIONAL at all if the\nuser specifies --with-CC ... if the user wants to override the script's\nidea of what C compiler to use, then it's also the user's responsibility\nto supply any special CFLAGS that might be needed. So we can just move\nthe macro into the \"else\" case where AC_PROG_CC is run.\n\nThe -traditional switch shouldn't be needed anyway in any reasonably\nmodern gcc installation, so we could probably get away with just\nremoving that test entirely. But I'll leave it in.\n\n> I don't know if I will be able to test BETA4, because I'm leaving for a\n> short vacation (lucky me :-) on Saturday and won't be back until Thursday.\n\nHave a good trip!\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Oct 1998 12:45:49 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d " }, { "msg_contents": "> \tIf its a bug in autoconf itself, we can patch the autoconf .m4\n> files directly on Hub.Org, and just make sure we autoconf with that...I\n> have no problemsm with that...\n\n* acspecific.m4.orig Thu Oct 29 12:43:02 1998\n--- acspecific.m4 Thu Oct 29 12:49:39 1998\n***********\n* 81,87 **\n AC_PROG_CC_WORKS\n AC_PROG_CC_GNU\n\n! if test $ac_cv_prog_gcc = yes; then\n GCC=yes\n dnl Check whether -g works, even if CFLAGS is set, in case the package\n dnl plays around with CFLAGS (such as to build both debugging and\n--- 81,87 ----\n AC_PROG_CC_WORKS\n AC_PROG_CC_GNU\n\n! if test \"$ac_cv_prog_gcc\" = yes; then\n GCC=yes\n dnl Check whether -g works, even if CFLAGS is set, in case the package\n dnl plays around with CFLAGS (such as to build both debugging and\n***********\n* 215,221 ***\n AC_DEFUN(AC_PROG_GCC_TRADITIONAL,\n [AC_REQUIRE([AC_PROG_CC])dnl\n AC_REQUIRE([AC_PROG_CPP])dnl\n! if test $ac_cv_prog_gcc = yes; then\n AC_CACHE_CHECK(whether ${CC-cc} needs -traditional,\n ac_cv_prog_gcc_traditional,\n [ ac_pattern=\"Autoconf.'x'\"\n--- 215,221 ----\n AC_DEFUN(AC_PROG_GCC_TRADITIONAL,\n [AC_REQUIRE([AC_PROG_CC])dnl\n AC_REQUIRE([AC_PROG_CPP])dnl\n! if test \"$ac_cv_prog_gcc\" = yes; then\n AC_CACHE_CHECK(whether ${CC-cc} needs -traditional,\n ac_cv_prog_gcc_traditional,\n [ ac_pattern=\"Autoconf.'x'\"\n\nShould work... if the other message is correct about the source of the\nproblem. There really needs to be a sanity check for --with-cc or whatever\nit is.\n\nTaral\n\n", "msg_date": "Thu, 29 Oct 1998 11:51:24 -0600", "msg_from": "\"Taral\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d " }, { "msg_contents": "\"Taral\" <[email protected]> writes:\n> ** acspecific.m4.orig Thu Oct 29 12:43:02 1998\n> --- acspecific.m4 Thu Oct 29 12:49:39 1998\n\nI have just committed a configure.in fix that eliminates the need for\nthis change. It turns out that the shell syntax failure was just the\ntip of the iceberg: the real problem is that AC_PROG_CC must be run\neven if there is a user-supplied CC setting.\n\n> There really needs to be a sanity check for --with-cc or whatever it is.\n\nThere is: AC_PROG_CC makes sure that the supplied CC setting works,\nand incidentally discovers a few other things like whether it is gcc.\nOur problem was we were bypassing that macro if --with-CC was given.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Oct 1998 13:23:29 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d " }, { "msg_contents": "\t If its a bug in autoconf itself, we can patch the autoconf .m4\n files directly on Hub.Org, and just make sure we autoconf with that...I\n have no problemsm with that...\n\nBetter to create an aclocal or whatever and distribute that so that\nanyone running autoconf gets the changes. Otherwise, it will be\nimpossible to maintain the configure system.\n\nCheers,\nBrook\n", "msg_date": "Thu, 29 Oct 1998 17:25:07 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] 6.4 BETA2 fails to compile on Digital Unix 4.0d" } ]
[ { "msg_contents": "This is a diff of several regression tests I run on my Linux box (last cvs):\n\ndv:~/cvs/pgsql/src/test/regress$ diff results/opr_sanity.out expected/opr_sanity.out \n54,58c54,56\n< oid\|oprcode \|oid\|oprcode \n< ---+-----------------+---+-----------------\n< 512\|on_ppath \|754\|pt_contained_path\n< 754\|pt_contained_path\|512\|on_ppath \n< (2 rows)\n---\n> oid\|oprcode\|oid\|oprcode \n> ---+-------+---+-------\n> (0 rows)\n\n\ndv:~/cvs/pgsql/src/test/regress$ diff results/random.out expected/random.out \n22,23c22\n< 124\n< (1 row)\n---\n> (0 rows)\n\ndv:~/cvs/pgsql/src/test/regress$ diff results/geometry.out expected/geometry.out \n115c115\n< \|(-10,0) \|[(-1000000,200),(300000,-40)]\|(-9.99715942258202,15.3864610140472) \n---\n> \|(-10,0) \|[(-1000000,200),(300000,-40)]\|(-9.99715942258202,15.3864610140473) \n\n\n\nOleg\n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n", "msg_date": "Thu, 29 Oct 1998 15:18:15 +0300 (MSK)", "msg_from": "Oleg Bartunov <[email protected]>", "msg_from_op": true, "msg_subject": "regression test results - Linux, cvs" }, { "msg_contents": "> This is a diff of several regression tests I run on my Linux box\n> (last cvs):\n\nAll known problems/features being fixed now. \n\nThe random test fails occasionally (at random, suprisingly enough :)\nsince I'm doing a statistical test and occasionally the result falls\noutside the arbitrary limits I set. But I didn't want the limits to be\ntoo loose otherwise we might miss truly incorrect results. If you rerun,\nrandom is likely to pass.\n\n - Tom\n", "msg_date": "Thu, 29 Oct 1998 14:38:41 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] regression test results - Linux, cvs" }, { "msg_contents": "On Thu, 29 Oct 1998, Thomas G. Lockhart wrote:\n\n> Date: Thu, 29 Oct 1998 14:38:41 +0000\n> From: \"Thomas G. Lockhart\" <[email protected]>\n> To: Oleg Bartunov <[email protected]>\n> Cc: [email protected]\n> Subject: Re: [HACKERS] regression test results - Linux, cvs\n> \n> > This is a diff of several regression tests I run on my Linux box\n> > (last cvs):\n> \n> All known problems/features being fixed now. \n> \n> The random test fails occasionally (at random, suprisingly enough :)\n> since I'm doing a statistical test and occasionally the result falls\n> outside the arbitrary limits I set. But I didn't want the limits to be\n> too loose otherwise we might miss truly incorrect results. If you rerun,\n> random is likely to pass.\n> \n\nYes,\nrandom test passed ok 1 from 3 tries. I have no time to test its\nrandomness :-) \n\nbtw, vacuum crash when postmaster starts with -B 1024 option\nseems fixed now !!!! I tried many times 'vacuum analyze' and never\nget problem ! Probably this is a bonus of last fixes :-)\n\tRegards,\n\t\tOleg\n\n\n\n> - Tom\n> \n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n", "msg_date": "Thu, 29 Oct 1998 17:52:43 +0300 (MSK)", "msg_from": "Oleg Bartunov <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] regression test results - Linux, cvs" }, { "msg_contents": "> On Thu, 29 Oct 1998, Thomas G. Lockhart wrote:\n> \n> > Date: Thu, 29 Oct 1998 14:38:41 +0000\n> > From: \"Thomas G. Lockhart\" <[email protected]>\n> > To: Oleg Bartunov <[email protected]>\n> > Cc: [email protected]\n> > Subject: Re: [HACKERS] regression test results - Linux, cvs\n> > \n> > > This is a diff of several regression tests I run on my Linux box\n> > > (last cvs):\n> > \n> > All known problems/features being fixed now. \n> > \n> > The random test fails occasionally (at random, suprisingly enough :)\n> > since I'm doing a statistical test and occasionally the result falls\n> > outside the arbitrary limits I set. But I didn't want the limits to be\n> > too loose otherwise we might miss truly incorrect results. If you rerun,\n> > random is likely to pass.\n> > \n> \n> Yes,\n> random test passed ok 1 from 3 tries. I have no time to test its\n> randomness :-) \n> \n> btw, vacuum crash when postmaster starts with -B 1024 option\n> seems fixed now !!!! I tried many times 'vacuum analyze' and never\n> get problem ! Probably this is a bonus of last fixes :-)\n\nMy guess is the catalog changes Tom did fixed it. Vacuum analyze uses\nthem quite a bit, and buffer cache size could affect which duplicate was\npicked.\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Thu, 29 Oct 1998 12:43:04 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] regression test results - Linux, cvs" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n>> btw, vacuum crash when postmaster starts with -B 1024 option\n>> seems fixed now !!!! I tried many times 'vacuum analyze' and never\n>> get problem ! Probably this is a bonus of last fixes :-)\n\n> My guess is the catalog changes Tom did fixed it. Vacuum analyze uses\n> them quite a bit, and buffer cache size could affect which duplicate was\n> picked.\n\nHmm, that would be an unexpected side benefit, wouldn't it! It could be\ntrue, if vacuum depends on pg_operator entries. That'd explain why the\nrest of us couldn't duplicate Oleg's problem: I'll bet no one who tried\nhad tables containing the data types that had bogus entries. (In fact,\nI imagine you need to have some indexes on those data types before\nyou'd see such a problem in vacuum, no?)\n\nIt occurs to me that there ought to be a VACUUM ANALYZE somewhere in\nthe regression suite, probably at the end where it has a whole database\nof weird stuff to chew on.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Oct 1998 14:49:36 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] regression test results - Linux, cvs " } ]
[ { "msg_contents": "Hi\nI just compiled beta3 & today snapshot on my linux 2.0.35 /debian slink\n(glibc6)\nall went well\nbut when i start psql I get\nTrace/breakpoint trap\nand it core dump!\nI never seen something like this,\nhow make it work ?\nthanks\n\n\n\n\n", "msg_date": "Thu, 29 Oct 1998 15:31:12 +0100", "msg_from": "Giuseppe Tanzilli <[email protected]>", "msg_from_op": true, "msg_subject": "beta3 compile problem" } ]
[ { "msg_contents": "Hi, all.\n\nI've just compiled 6.4 BETA2 on my DU 4.0d system with DEC C, and have\nalso run the regression tests.\n\nI had the following problems when building:\n\n1 - The DEC C preprocessor doesn't accept input from stdin. configure\ntries to use \"cc -E -\", but it doesn't work. According to Tom Lane, this\nshould be fixed in BETA4. I worked around this problem patching\nGen_fmgrtab.sh.in and genbki.sh.in as suggested by Thomas Lockhart in an\nearlier message.\n\n2 - I got many warnings complaining that MAP_FILE was redefined. Indeed,\nMAP_FILE is defined by many systems in \"sys/mman.h\", as it is used with\nmmap(2) and friends. MAP_FILE in include/libpq/hba.h should be renamed to\nsomething like PG_MAP_FILE.\n\n3 - I had to add the option '-std' to the compiler flags when compiling in\ninterfaces/libpq. This one is platform-specific.\n\n4 - I had to edit Makefile.tcldefs in pl/tcl. Digital Unix uses 'ld\n-shared -expect_unresolved \"*\"' to generate shared libraries, and so it is\nspecified in my system's /usr/local/lib/tclConfig.sh. However, somwhere in\nthe config process the quotes around the asterisk dissappeared, and so the\nshell expanded the asterisk. I had to put the quotes back. I didn't have\nthe time to investigate more.\n\nThe following regression tests failed:\n\n- char, varchar\n\nThey failed because for some obscure reason 'A' is considered greater than\n'a'. I won't be able to investigate on this until tomorrow.\n\n- int2, int4\n\nThey failed because of differences in the expected error message, so they\nare ok.\n\n- int8\n\nThe int8 type is completely brain damaged on my system. Strange, because\nDigital Unix is a true 64 bit system. Again, I'll try to look at it\ntomorrow.\n\n- float8\n\nIt had one error-message-format error (not really an error) and one real\nerror (\"exp() result is out of range\" when no error was expected).\n\n- abstime, tinterval, horology\n\nSeems to be a time zone problem, though I ran the test with TZ=PST8PDT. I\nhaven't had time to investigate deeper.\n\n- select_implicit, select_having, random, rules, plpgsql\n\nI haven't had time to look at these. I'll look at them tomorrow.\n\nGreetings,\n\n\tPedro.\n\n-------------------------------------------------------------------\nPedro Josï¿½ Lobo Perea Tel: +34 91 336 78 19\nCentro de Cï¿½lculo Fax: +34 91 331 92 29\nEUIT Telecomunicaciï¿½n - UPM e-mail: [email protected]\n\n\n", "msg_date": "Thu, 29 Oct 1998 18:49:39 +0100 (MET)", "msg_from": "\"Pedro J. Lobo\" <[email protected]>", "msg_from_op": true, "msg_subject": "Build report for 6.4 BETA2 on Digital Unix 4.0d" } ]
[ { "msg_contents": "I did a CVS update tuesday night, built it yesterday, and ran the\nregression test today... Here are the results of running\n\"grep failed\" on regress.out:\n\ndestroydb: database destroy failed on regression.\nfloat8 .. failed\n\n It appears that there is some confusion whether 1e200 is an\n overflow or underflow here. Something to check in libc is my\n guess. I also got errors with 10e-400 and -10e-400 underflowing.\n\ndatetime .. failed\n\n Here I get my same old problem of: (There are others, but I suspect\n they all have the same root cause.)\n QUERY: SELECT ('now'::datetime - 'current'::datetime) AS \"ZeroSecs\";\n ! ZeroSecs \n ! -----------------------------\n ! @ 428 days 7 hours 8 secs ago\n\nhorology .. failed\n\n I suspect the same stuff that makes datetime fail is making\n this fail too...\n\ninet .. failed\n \n This seems to be caused by:\n + ERROR: type name lookup of cidr failed\n I did a gmake distclean, configure, gmake, gmake install, initdb,\n cd test/regress, gmake all, gmake runtest. Did I miss something?\n Is the beta different from the CVS version I got (Tuesday night)?\n\nsanity_check .. failed\n\n This one looks pretty serious, but perhaps it's not:\n QUERY: VACUUM;\n ! pqReadData() -- backend closed the channel unexpectedly.\n ! This probably means the backend terminated abnormally before\n or while processing the request.\n ! We have lost the connection to the backend, so further processing\n is impossible. Terminating.\n\nmisc .. failed\n\n The results of this query are reversed:\n\n QUERY: SELECT unique1 FROM onek2 WHERE unique1 < 2;\n\n And inet_tbl is missing from another query. (Only looking at\n the 3 line diff context doesn't tell me much, but I suspect\n that it's the same error that caused the inet tests to fail.)\n\n\nThere you have it, NetBSD-current/pc532 (from about a month ago)\nwith Tuesday night's CVS update of PostgreSQL is not quite ready\nfor prime time. It's pretty close though. If someone wants to\nlog into my machine and take a whack at it, I'll try to set something\nup. Otherwise, I'll have to wait a week or two before I can touch\nit.\n\nI don't think I have the ability to submit any diffs this late in\nthe beta cycle safely anyway. Besides I have a big part of a\nhomework assignment to do before Monday, and the prof scheduled\na midterm the same day he wants to collect the homework. 8-( (Too\nbad I can't quit my job for a year and finish this degree work all\nat once.)\n\nJon Buller\n", "msg_date": "Thu, 29 Oct 1998 14:30:02 -0600", "msg_from": "Jon Buller <[email protected]>", "msg_from_op": true, "msg_subject": "NS32K regression test" }, { "msg_contents": "Jon Buller <[email protected]> writes:\n> I did a CVS update tuesday night, built it yesterday, and ran the\n> regression test today... Here are the results of running\n> \"grep failed\" on regress.out:\n> [ lots of problems, including ]\n> This seems to be caused by:\n> + ERROR: type name lookup of cidr failed\n\nI think you must have been unlucky enough to get an inconsistent\nfileset. That's always a risk when grabbing stuff directly from the\nCVS repository --- if someone else is checking in updates at the same\ntime, you might get some old files and some new.\n\nIf you have time, please do another update and run another full\nrebuild from the current fileset. We fixed a couple of problems\nsince Tuesday anyway.\n\n\t\t\tregards, tom lane\n", "msg_date": "Thu, 29 Oct 1998 19:41:46 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] NS32K regression test " }, { "msg_contents": "Tom Lane <[email protected]> writes:\n\n> I think you must have been unlucky enough to get an inconsistent\n> fileset. That's always a risk when grabbing stuff directly from the\n> CVS repository --- if someone else is checking in updates at the same\n> time, you might get some old files and some new.\n\nHmm. Probably for the inet bug, that one suprised me a bit. I'm\nnot too sure my libc and libm deal with the corner cases of IEEE-754\ndouble precision yet, so I wasn't suprised by those. And I was\nexpecting the datetime stuff to fail, since it did a few months\nago, and I haven't figured out what's going on there yet, and a\nlack of debugging time hasn't helped.\n\n> If you have time, please do another update and run another full\n> rebuild from the current fileset. We fixed a couple of problems\n> since Tuesday anyway.\n\nJust finished the update, going to start the build now... The nice\nthing about all this is major portions of it can be done without\nany intervention, and it's not too hard to spare a few minutes\nevery few hours.\n\nI'll let you all know how it this one works out too. I'm not\nexpecting much more than the inet problems to go away, and maybe\nthat backwards select since I saw those messages about every operator\nunder the sun not wor... 8-)\n\nJon\n", "msg_date": "Thu, 29 Oct 1998 20:56:43 -0600", "msg_from": "Jon Buller <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] NS32K regression test " } ]
[ { "msg_contents": "Hello:\n\nI'd like to learn a bit about how concurrency control is managed by\nPostgreSQL, for a project I'm starting. I looked over the documentation,\nFAQs, features lists, etc. on the postgresql web site but they are aimed\nmore toward users and application programmers than tech-heads who want to\nlook under the hood.\n\nI'm just wondering (a) what scheme(s) PostgreSQL uses (eg. two-phase\nlocking, timestamp, etc.), (b) what recovery procedures are available, and\n(c) what source code modules I should concentrate looking at to see how\nthese controls have been implemented.\n\nThanks in advance.\n\n--------------< LINUX: The choice of a GNU generation. >--------------\nSteve Frampton <[email protected]> http://qlink.queensu.ca/~3srf\n\n", "msg_date": "Thu, 29 Oct 1998 20:04:19 -0500 (EST)", "msg_from": "Steve Frampton <[email protected]>", "msg_from_op": true, "msg_subject": "Question on concurrency controls" }, { "msg_contents": "Hello:\n\nIs it true that there is no support for row-level locking at this time?\nIs this something that is being worked on?\n\nThanks...\n\n--------------< LINUX: The choice of a GNU generation. >--------------\nSteve Frampton <[email protected]> http://qlink.queensu.ca/~3srf\n\n", "msg_date": "Sun, 1 Nov 1998 17:01:27 -0500 (EST)", "msg_from": "Steve Frampton <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Question on concurrency controls" }, { "msg_contents": "On Sun, 1 Nov 1998, Steve Frampton wrote:\n\n> Hello:\n> \n> Is it true that there is no support for row-level locking at this time?\n> Is this something that is being worked on?\n\n\tYes, and yes...\n\n\tVadim has it planned for v6.5...v6.4 is just about to be released,\nwith v6.5 being released first third new year :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sun, 1 Nov 1998 18:28:02 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Question on concurrency controls" }, { "msg_contents": "> \n> On Sun, 1 Nov 1998, Steve Frampton wrote:\n> \n> > Is it true that there is no support for row-level locking at this time?\n> > Is this something that is being worked on?\n> \n> \tYes, and yes...\n> \n> \tVadim has it planned for v6.5...v6.4 is just about to be released,\n> with v6.5 being released first third new year :)\n> \n\nInteresting... please let me know what levels of concurrency are available ?\nAlternatively is there documentation describing concurrency?\n\nRegards\nTheo\n\n\n", "msg_date": "Tue, 3 Nov 1998 07:22:52 +0200 (SAST)", "msg_from": "Theo Kramer <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Question on concurrency controls" } ]
[ { "msg_contents": "Hello!\n\nI'm the original reporter of \"vacuum crash\" problem.\nThank you for your quick fix.\n\nNow, I found another problem in nbtsearch.c.\nThis is my patch for it.\n# But, what I want to know is why postgres reaches this FATAL status...\n\n========================================================================\n* nbtsearch.c.orig\tTue Oct 20 16:42:59 1998\n--- nbtsearch.c\tTue Oct 20 16:43:15 1998\n***********\n* 217,223 ***\n \t\t\t\t\t\t\t\t\t\tPageGetItemId(page, P_FIRSTKEY),\n \t\t\t\t\t\t\t\t\t\tBTEqualStrategyNumber))\n \t\t\t\t\t\t\telog(FATAL, \"btree: BTP_CHAIN flag was expected in %s (access = %s)\",\n! \t\t\t\t\t\t\t\t rel->rd_rel->relname, access ? \"bt_write\" : \"bt_read\");\n \t\t\t\t\t\tif (_bt_skeycmp(rel, keysz, scankey, page,\n \t\t\t\t\t\t\t\t\t\tPageGetItemId(page, offmax),\n \t\t\t\t\t\t\t\t\t\tBTEqualStrategyNumber))\n--- 217,223 ----\n \t\t\t\t\t\t\t\t\t\tPageGetItemId(page, P_FIRSTKEY),\n \t\t\t\t\t\t\t\t\t\tBTEqualStrategyNumber))\n \t\t\t\t\t\t\telog(FATAL, \"btree: BTP_CHAIN flag was expected in %s (access = %s)\",\n! \t\t\t\t\t\t\t\t rel->rd_rel->relname.data, access ? \"bt_write\" : \"bt_read\");\n \t\t\t\t\t\tif (_bt_skeycmp(rel, keysz, scankey, page,\n \t\t\t\t\t\t\t\t\t\tPageGetItemId(page, offmax),\n \t\t\t\t\t\t\t\t\t\tBTEqualStrategyNumber))\n========================================================================\n\nAnd with the following script, you can find suspicious lines which\nhave pointer type incompatibility but work fortunately* on most\nsystems.\n\n========================================================================\n#/bin/sh\n\nfind PGSRC -type f -name '*.[ch]' -print \|\nxargs egrep -e 'rd_rel->relname' /dev/null \|\negrep -v 'relname\\)?\\.data'\n========================================================================\n\nRegards\n\n-- \nASCII CORPORATION\nTechnical Center\nSHIOZAKI Takehiko\n<[email protected]>\n", "msg_date": "Fri, 30 Oct 1998 10:08:54 +0900 (JST)", "msg_from": "SHIOZAKI Takehiko <[email protected]>", "msg_from_op": true, "msg_subject": "relname.data" }, { "msg_contents": "Applied.\n\nWhat I also have seen done is changing:\n\n\trel->rd_rel->relname\n\nto:\n\n\t&rel->rd_rel->relname\n\nThat should also fix it.\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Mon, 2 Nov 1998 10:37:08 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] relname.data" } ]
[ { "msg_contents": "\nhi!\n\nis there any idea to make crypted communication between the backend and\nthe frontend ?\nIf I remember correctly somebody wrote something about this issue before\n\nthanks,\n\t\tredax\n\n\n.----------------------------------------------------------.\n\|Zsolt Varga \| tel/fax: +36 36 422811 \|\n\| AgriaComputer LTD \| email: [email protected] \|\n\| System Administrator \| URL: http://www.agria.hu/ \|\n`----------------------------------------------------------'\n\n", "msg_date": "Fri, 30 Oct 1998 15:09:00 +0100 (CET)", "msg_from": "Zsolt Varga <[email protected]>", "msg_from_op": true, "msg_subject": "encoded communincation bw Backend and Libpq frontend?" }, { "msg_contents": "Zsolt Varga <[email protected]> writes:\n> is there any idea to make crypted communication between the backend and\n> the frontend ?\n\nI don't think there's any need for us to add that to Postgres. You\nshould be able to set it up right now using ssh (if you don't have ssh,\nsee http://www.cs.hut.fi/ssh). Free bonus: ssh can apply data\ncompression to the link as well.\n\nI've never actually used ssh to forward non-telnet TCP connections,\nso I'm not sure exactly what you have to type to set it up. If you\nwork out the details, would you make a little writeup to be added to\nthe Postgres documentation? That would save time for other people\nin the future.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 30 Oct 1998 10:34:54 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] encoded communincation bw Backend and Libpq frontend? " } ]
[ { "msg_contents": "OK, I'm about ready to freeze the admin guide and installation info. I'd\nlike to put the TODO list into the release notes, along with a few\nparagraphs highlighting feature improvements/changes which might affect\nusers.\n\nDo you want to write something, or should I go ahead? Also, I'd like to\nclean up/consolidate some of the entries in the TODO list; should I do\nthat and commit it or do it and just send it directly to you?\n\n - Tom\n", "msg_date": "Fri, 30 Oct 1998 15:14:11 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Ready for admin guide and INSTALL" }, { "msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> OK, I'm about ready to freeze the admin guide and installation info.\n\nCan the peanut gallery get a look first?\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 30 Oct 1998 11:18:30 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Ready for admin guide and INSTALL " }, { "msg_contents": "> OK, I'm about ready to freeze the admin guide and installation info. I'd\n> like to put the TODO list into the release notes, along with a few\n> paragraphs highlighting feature improvements/changes which might affect\n> users.\n> \n> Do you want to write something, or should I go ahead? Also, I'd like to\n> clean up/consolidate some of the entries in the TODO list; should I do\n> that and commit it or do it and just send it directly to you?\n\nI have updated the TODO list, and it will probably not change before 6.4\nis released. You can use the copy in the distribution, or on the web\npage. The web page copy has HTML markups.\n\nIf you want to clean up the TODO list, please do. E-mail me an ASCII\ncopy back, and I will update it here. If you want the markups of your\nchanges in html, change the ASCII, and I will run it through to get\nhtml.\n\nAs far as a highlights, I can't think of some good words for it, so\nplease go ahead and write something. In previous releases, certain\nitems clearly came to me, but not with 6.4.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 30 Oct 1998 12:17:49 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: Ready for admin guide and INSTALL" }, { "msg_contents": "> > OK, I'm about ready to freeze the admin guide and installation info.\n> Can the peanut gallery get a look first?\n\nI've posted almost-final installation instructions in html format on the\nweb site at\n\n ftp://postgresql.org/pub/patches/installation.tar.gz\n\nThese will replace INSTALL as plain text for v6.4. The Admin Guide\nwill also have this info in both html and Postscript formats.\n\nIf you have comments let me know; I will need to freeze this stuff this\nevening.\n\n - Tom\n", "msg_date": "Fri, 30 Oct 1998 19:59:44 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Ready for admin guide and INSTALL" }, { "msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> I've posted almost-final installation instructions in html format on the\n> web site at\n> ftp://postgresql.org/pub/patches/installation.tar.gz\n\nIt's looking pretty good. I have a few more suggestions...\n\n\nMajor suggestion: chapter 3 (detailed discussion of configure options)\nprobably ought to be placed after the step-by-step install, not before.\nA person who doesn't already know what configure is would be pretty lost\nupon hitting the first page of chapter 3, I think.\n\nMinor stuff:\n\nInstall step 9 says:\n\n Note: There are two places in this installation procedure where you will\n have an opportunity to specify installation locations for programs,\n libraries, documentation, and other files. Usually it is sufficient to\n specify these at the make install stage of installation.\n\nMake install?? I think it is much safer to tell people to select these\nlocations during configure. I would not want to bet that you can change\nthe install target just by tweaking make parameters at the install step.\nOn HPUX, for example, that will definitely fail, because the path to the\nshared libraries in /usr/local/pgsql/lib (eg) is embedded into the\nbinaries for psql and friends at build time, not install time.\n\n\nIn the lengthier discussion of configure options, I recommend not listing\n --datadir=DIR read-only architecture-independent data in DIR\n [PREFIX/share]\n --sysconfdir=DIR read-only single-machine data in DIR [PREFIX/etc]\n --sharedstatedir=DIR modifiable architecture-independent data in DIR\n [PREFIX/com]\n --srcdir=DIR find the sources in DIR [configure dir or ..]\n --with-x use the X Window System\nThese are generic configure options that are not actually used by our\nsetup, so they're hardly worth listing. Also, I suggest adding\n --without-CXX do not try to build C++ code\n(which is mentioned inline in the install procedure, but it's missing\nin the allegedly more complete discussion of options...)\n\n\nAnd in the trivia category: ports list refers to HPUX \"9.0.x\", but it\nshould say \"9.0x\".\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 30 Oct 1998 17:17:55 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Ready for admin guide and INSTALL " }, { "msg_contents": "Argh! Forgot the most important point --- the install procedure is\nstill missing anything about installing the Perl module. Recommend\nadding the following, or something much like it, just after the\nprimary install step (between items 15 and 16, currently):\n\n\n<Step Performance=\"optional\">\n<Para>\nIf you used the <option>--with-perl</option> option to configure, check\nthe install log to see whether the Perl module was actually installed.\nIf you've followed our advice to make the Postgres files be owned by\nan unprivileged userid, then the Perl module won't have been installed,\nfor lack of write privileges on the Perl library directories. You can\ncomplete its installation, either now or later, by su'ing to the user that\ndoes own the Perl library (often root) and doing\n<ProgramListing>\n$ cd /usr/src/pgsql/src/interfaces/perl5\n$ gmake install\n</ProgramListing>\n</Para>\n</Step>\n\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 30 Oct 1998 17:51:00 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Ready for admin guide and INSTALL " }, { "msg_contents": "> It's looking pretty good. I have a few more suggestions...\n\nOK, but save some for v6.5. As you can see there are some parts of these\ndocs which are just hanging together with baling wire and gum, and could\nreally use a cleanup/update/normalization.\n\n> Major suggestion: chapter 3 (detailed discussion of configure options)\n> probably ought to be placed after the step-by-step install, not \n> before.\n\nYeah. I'll put a mention in beforehand and in the procedure and move the\ntext to afterwards. I already said something about \"plain vanilla\"\ninstallations not needing many/any options.\n\n> Install step 9 says:\n> Note: There are two places in this installation procedure where...\n\n> Make install?? I think it is much safer to tell people to select \n> these locations during configure.\n\nIn testing the standalone ODBC stuff, I found that some parameters had\nto be specified during the make stage for some reason. The stuff was\nstolen from the main configure, so shares a lot of traits.\n\nI can't write a definitive installation doc given that I should have\nfrozen all of this several days ago. So I'm inclined to mention both,\nand we can hash it out _early_ in the v6.5 cycle. The good news is that\nthe installation stuff is more accurate than it used to be. The bad news\nis that it pretty clearly isn't perfect. Thank god for mailing lists,\neh?\n\n> In the lengthier discussion of configure options, I recommend not \n> listing...\n\nOK.\n\n> And in the trivia category: ports list refers to HPUX \"9.0.x\", but it\n> should say \"9.0x\".\n\nFixed. And frozen. I'm going to take Bruce's ToDo/release list, put it\ninto a new file current.sgml, add a few paragraphs of notes, and be done\nwith it.\n\n - Tom^H^H^HThomas\n", "msg_date": "Sat, 31 Oct 1998 03:59:54 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Ready for admin guide and INSTALL" } ]
[ { "msg_contents": "After rebuilding and retesting from a CVS update last night, I can\nreport that all tests passed with the following exceptions:\n\nfloat8 has some problems with very large and very small numbers.\ndatetime and horology have some problems with 'current' and 'epoch'\n which might (or might not) be related to the float8 problems.\n\nI am quite impressed. Because of all of your work, I can do my\nsemester database project at home in PostgreSQL rather than on the\nschools (extremely poorly managed) network and (OK as far as I can\ntell) Oracle system.\n\nThanks much. I'll have more diffs when I figure out the problems.\nMaybe for 6.4.1, maybe for 6.5 kinda depends on the bugs, and my\ntime, and how much that baby due on Jan 2 is going to keep me awake\nat night 8-). (Our first one was really really bad about that for\na few months.)\n\nJon\n", "msg_date": "Fri, 30 Oct 1998 10:59:21 -0600", "msg_from": "Jon Buller <[email protected]>", "msg_from_op": true, "msg_subject": "NetBSD/NS32K regression results" }, { "msg_contents": "Jon Buller <[email protected]> writes:\n> After rebuilding and retesting from a CVS update last night, I can\n> report that all tests passed with the following exceptions:\n\nGreat!\n\n> float8 has some problems with very large and very small numbers.\n> datetime and horology have some problems with 'current' and 'epoch'\n> which might (or might not) be related to the float8 problems.\n\nI think you guessed right. IIRC, datetime has an underlying\nrepresentation of float8, and some of the special values like\n\"current\" are represented as particular very-large or very-small\nnumbers. So it might break if your machine's float math is not\nup to IEEE-float specs. However, you can do an awful lot of stuff\nwith datetime without ever needing the special values, so this\nmight not be much of a problem for you.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 30 Oct 1998 16:43:58 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] NetBSD/NS32K regression results " } ]
[ { "msg_contents": "> On Thu, 29 Oct 1998, Thomas G. Lockhart wrote:\n> \n> > > But I have no idea where the\n> > > rest of yunz are.\n> > \n> > South Pasadena, California...\n> \n> \tEver get cold down there? I need more heat, just haven't\n> figured\n> out where to go for it yet :) Wait, you have horses too, don't you?\n> Hrmmmm.... :)\nHave you been to Vancouver? It's wonderful there. I live in Dallas,\nTexas. But Vancouver is warmer in the winter than it is here. Here it\nusually gets below freezing two or three times in the winter for a total\nof about 10 days. And once ever 4 years or so we actually get some snow\nthat's still around the next day as slush (which is really bad when it\nfreezes again that night because no one knows how to drive on snow and\nice here).\nSo what does this have to do with pg_operator.h?\n\t-DEJ\n\n", "msg_date": "Fri, 30 Oct 1998 11:23:50 -0600", "msg_from": "\"Jackson, DeJuan\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] Re: [COMMITTERS] 'pgsql/src/include/catalog pg_oper\n\tator.h'" } ]
[ { "msg_contents": "> I think we need to add this to INSTALL file after setting\n> environmnet variable TZ.\n> 'If you compile Postgres with --enable-locale option\n> set locale environment to C (or unset all LC_* variables)\n> before starting postmaster.'\n\nHi Oleg. We are rewriting the installation instructions, and I have\nincluded your full description of locale support into those\ninstructions. I had not added in an explicit mention into the\ninstallation procedure also, but will do that.\n\n - Tom\n", "msg_date": "Fri, 30 Oct 1998 18:58:19 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] a little addition to INSTALL (fwd)" } ]
[ { "msg_contents": "Hi.\n\nI just built postgresql 6.4 beta 4 on HPUX 10.01. This was not a clean compile\n+, \nand I had to make minor modifications to get it to compile. First of all,\nhpux 10.01 does not define SA_RESTART at all (10.20 does, but 10.01 does not). \n+\nTo compile under 10.01, you cant define USE_POSIX_SIGNALS in os.h I fixed this\n+\nby just deleting these lines from os.h:\n\n#ifdef HPUX_10\n#define USE_POSIX_SIGNALS\n#endif\n\nSecondly, lex on hpux 10.01 has problems as well. I got the following errors \ndue to lex on 10.01:\n\nmake -C lib all\nmake[3]: Entering directory `/home/schout/pgsql/src/interfaces/ecpg/lib'\nmake[3]: Nothing to be done for `all'.\nmake[3]: Leaving directory `/home/schout/pgsql/src/interfaces/ecpg/lib'\nmake -C preproc all\nmake[3]: Entering directory `/home/schout/pgsql/src/interfaces/ecpg/preproc'\nlex pgc.l\n\"pgc.l\" line 333: warning 45: Non-portable Character Class\n\"pgc.l\" line 333: warning 45: Non-portable Character Class\n\"pgc.l\" line 471: warning 45: Non-portable Character Class\n\"pgc.l\" line 471: warning 45: Non-portable Character Class\n\"pgc.l\" line 471: warning 45: Non-portable Character Class\n\"pgc.l\" line 471: warning 45: Non-portable Character Class\n\"pgc.l\" line 475: warning 45: Non-portable Character Class\n\"pgc.l\" line 475: warning 45: Non-portable Character Class\n\"pgc.l\" line 536: warning 45: Non-portable Character Class\n\"pgc.l\" line 536: warning 45: Non-portable Character Class\n\"pgc.l\" line 585: warning 45: Non-portable Character Class\n\"pgc.l\" line 585: warning 45: Non-portable Character Class\n\"pgc.l\" line 674: warning 48: Undefined start condition <EOF\n\"pgc.l\" line 708: error 14: Too many transitions, Try using %a num\n671/1000 nodes(%e), 1482/2500 positions(%p), 168/500 (%n), 14470 transitions, 89\n+5/1000 packed char classes(%k), 2000/2000 packed transitions(%a), 0/0 output sl\n+ots(%o)\nmake[3]: * [pgc.c] Error 1\nmake[3]: Leaving directory `/home/schout/pgsql/src/interfaces/ecpg/preproc'\nmake[2]: * [all] Error 2\nmake[2]: Leaving directory `/home/schout/pgsql/src/interfaces/ecpg'\nmake[1]: * [all] Error 2\nmake[1]: Leaving directory `/home/schout/pgsql/src/interfaces'\nmake: * [all] Error 2\n\n\n\nI got around this problem by installing flex-2.5.4a from prep.ai.mit.edu, and\nsym linking /usr/bin/lex to /opt/gnu/bin/flex (where I installed flex).\n\nAfter that, everything pretty much went smoothly. I did have to manually commen\n+t\nout the rules to build libpq++ because apparently ecgs wont let you specify the\n+\n\"-c\" and \"-o\" flags together, but that is not a big deal to me since I dont \n+intend\nto use the libpq++ interface on this machine anyways. It would be nice if there\n+ \nwas a --disable-libpq++ rule for configure that would allow you to specify that \n+\nyou did not want to use it. And possibly something similar like \n--with-flex=/opt/gnu/bin/flex to override the choice of /usr/bin/flex (maybe the\n+se\nalready exist??).\n\nAnyways, Some food for thought. Im sure its too late to address these for the \n+\nupcoming 6.4 release, but maybe for the next minor release ;)\n\nMike\n", "msg_date": "Fri, 30 Oct 1998 14:01:35 -0600", "msg_from": "[email protected]", "msg_from_op": true, "msg_subject": "HPUX 10.01 build issues and solutions." }, { "msg_contents": "[email protected] writes:\n> hpux 10.01 does not define SA_RESTART at all (10.20 does, but 10.01\n> does not).\n\nHmm. I wonder if we'd be better off to treat 10.01 as being like HPUX 9?\nI think the choice between posix and non-posix signals is the main\ndifference that we care about...\n\n> Secondly, lex on hpux 10.01 has problems as well.\n\nA lot of systems have inadequate lexes installed. There is supposed to\nbe a test early in the install to see if your lex is up to snuff (and if\nnot it'll tell you to get flex). Either that's not working, or ecpg is\nexercising some new stuff that the test doesn't check for. I don't know\nenough about lex to try to improve the test --- anyone care to work on\nthat?\n\n> It would be nice if there was a --disable-libpq++ rule for configure\n> that would allow you to specify that you did not want to use it.\n\nThere is, it's spelled \"--without-CXX\". This needs to be documented\nbetter, but we're working on the install doc even now.\n\n> And possibly something similar like --with-flex=/opt/gnu/bin/flex to\n> override the choice of /usr/bin/flex (maybe these already exist??).\n\nconfigure will select flex in preference to lex if it finds flex in your\nPATH --- and if you have more than one flex in your PATH, the first one\ngets picked. Perhaps you just needed to re-run configure with an\nappropriate PATH? Alternatively, you can force the choice by setting\nthe LEX and LEXLIB environment variables before running configure (also\nundocumented I fear).\n\nFinally, there's always the option of editing Makefile.global\nafterwards. Right now it can be a bit tedious to look through all the\nfiles that configure emits, but we're going to try to consolidate the\nconfiguration decisions into just a couple of files in future releases.\n\n(BTW, hackers, I think this is an excellent example of why hand-editable\nconfigure output is a good thing. It's not that Mike couldn't have\npersuaded configure to do what he needed, it's that it's not always so\nobvious how to do that ... and editing the resulting Makefile.global\nmight be a lot easier than figuring out how to make configure do the\nright thing.)\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 30 Oct 1998 16:38:14 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] HPUX 10.01 build issues and solutions. " }, { "msg_contents": "> [email protected] writes:\n> > hpux 10.01 does not define SA_RESTART at all (10.20 does, but 10.01\n> > does not).\n> \n> Hmm. I wonder if we'd be better off to treat 10.01 as being like HPUX 9?\n> I think the choice between posix and non-posix signals is the main\n> difference that we care about...\n> \n> > Secondly, lex on hpux 10.01 has problems as well.\n> \n> A lot of systems have inadequate lexes installed. There is supposed to\n> be a test early in the install to see if your lex is up to snuff (and if\n> not it'll tell you to get flex). Either that's not working, or ecpg is\n> exercising some new stuff that the test doesn't check for. I don't know\n> enough about lex to try to improve the test --- anyone care to work on\n> that?\n\nIt just checks for buggy flex 2.5.3, that almost everyone had in early\n6.* releases. Not sure how many people still have that version, but I\nam sure we will require the test for several more years.\n\nThe big problem was that the bug only showed errors once you started to\nuse PostgreSQL, and the errors were very unclear.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 30 Oct 1998 17:08:08 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] HPUX 10.01 build issues and solutions." }, { "msg_contents": "[email protected] writes:\n> hpux 10.01 does not define SA_RESTART at all (10.20 does, but 10.01\n> does not). To compile under 10.01, you cant define USE_POSIX_SIGNALS\n> in os.h I fixed this by just deleting these lines from os.h:\n\n> #ifdef HPUX_10\n> #define USE_POSIX_SIGNALS\n> #endif\n\nI don't think that will work completely --- it may compile, but you will\nnot get the right behavior of signals. You need BSD signal semantics if\nyou don't have POSIX signals, and to get that on HPUX 10 you had better\nmodify makefiles/Makefile.hpux to link libBSD before libc. In the code\n\tifneq ($(HPUX_MAJOR), 09)\n\t LDFLAGS:= -Wl,-E -lc $(LDFLAGS)\n\tendif\nadd -lBSD before -lc.\n\n\nSince it's not clear where between HPUX 10.01 and 10.20 HP added POSIX\nsignal support, I can't easily fix this on the basis of testing the OS\nversion number. That's the wrong approach anyway, really. I am\nthinking that the right approach is to stop depending on the port header\nfiles to tell us whether to use POSIX signals, and instead to have\nconfigure auto-detect it. It would be easy to see whether <signal.h>\ncontains a definition for SA_RESTART, and that probably would do as a\nconfigure test for POSIX signals. (We'd better make sure that the\ntemplate files can override the test, of course, in case the platform\nhas the SA_RESTART macro but POSIX signals don't work right...)\n\nI don't think we should risk a change of that magnitude so close before\na major release, but I will work on it after 6.4 is out the door. For\n6.4 we will have to offer the above workaround to anyone who needs to\nbuild for HPUX 10.01.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 30 Oct 1998 19:08:47 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] HPUX 10.01 build issues and solutions. " }, { "msg_contents": "On Fri, 30 Oct 1998, Tom Lane wrote:\n\n> Finally, there's always the option of editing Makefile.global\n> afterwards. Right now it can be a bit tedious to look through all the\n> files that configure emits, but we're going to try to consolidate the\n> configuration decisions into just a couple of files in future releases.\n\n\tWe are? raised eyebrow\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 30 Oct 1998 23:10:13 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] HPUX 10.01 build issues and solutions. " }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n> On Fri, 30 Oct 1998, Tom Lane wrote:\n>> ... Right now it can be a bit tedious to look through all the\n>> files that configure emits, but we're going to try to consolidate the\n>> configuration decisions into just a couple of files in future releases.\n\n> \tWe are? raised eyebrow\n\nWell, it's on my to-do list...\n\n\t\t\tregards, tom lane\n", "msg_date": "Sat, 31 Oct 1998 12:27:38 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] HPUX 10.01 build issues and solutions. " }, { "msg_contents": "On Sat, 31 Oct 1998, Tom Lane wrote:\n\n> The Hermit Hacker <[email protected]> writes:\n> > On Fri, 30 Oct 1998, Tom Lane wrote:\n> >> ... Right now it can be a bit tedious to look through all the\n> >> files that configure emits, but we're going to try to consolidate the\n> >> configuration decisions into just a couple of files in future releases.\n> \n> > \tWe are? raised eyebrow\n> \n> Well, it's on my to-do list...\n\n\tI do not agree with the requirement to consolidate the\nconfiguration decisions into just a couple of files...we do nothing\ndifferently, as it is now, then any other large package does it, and I do\nnot see a requirement to change from what currently works.\n\n\tIf a particular port doesn't \"configure\" correctly, then fix the\nproblem, but fixing it doesn't mean consolidate what currently exists...\n\n Marc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sat, 31 Oct 1998 14:10:48 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] HPUX 10.01 build issues and solutions. " }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n> \tI do not agree with the requirement to consolidate the\n> configuration decisions into just a couple of files...we do nothing\n> differently, as it is now, then any other large package does it, and I do\n> not see a requirement to change from what currently works.\n\nWhat we have now works OK for run-of-the-mill installations on supported\nplatforms. It's not very friendly for people who have slightly\nnonstandard setups. We've seen at least one actual example of that\nsince the discussion started (not counting Thomas' original complaint,\nwhich I suppose you could write off as pilot error ... but if even\nThomas doesn't understand how to configure Postgres, then it must be a\npretty closely held secret...). I expect a pass through the pgsql-admin\narchives would turn up a bunch more examples.\n\nAlso, while I have seen packages that use configure to rewrite a ton of\nsubdirectory Makefiles (GNU Emacs, for example), I think they have\nsimpler and less failure-prone configuration setups than we do. Emacs,\nfor example, isn't trying to interface to both Tcl/Tk and Perl --- two\npackages that are out there in multiple versions and have no well\ndefined installation location.\n\nAutoconf configure scripts are inherently not 100% reliable. They are\nheuristic tests that maybe will work 95% or 99% or even 99.9% of the\ntime, but trying to make them absolutely bulletproof for any target\nsystem setup is folly. Good practice is to provide a manual override\npath for any decision the script makes. Now, we can do that either by\nadding configure script options or by making it reasonably convenient\nto hand-modify the results of configure. I submit that it is\ncounterproductive to have a list of script options that approaches the\nsize of Makefile.global.in + config.h.in. That will just confuse the\naverage person who has no need for most of those options.\n\nAlso, to bring the discussion back to Thomas' original complaint, what\nof Makefile.custom? If editing Makefile.global stops being a useful\ntechnique because we've put copies of configuration settings in too many\nother places, then Makefile.custom is also a useless way of customizing\nthe system.\n\nIn fact there's an even more direct issue, namely the port files.\nRoughly speaking, Makefile.port is capable of altering decisions\nexpressed in Makefile.global, and os.h is capable of altering decisions\nexpressed in config.h. If we scatter configuration settings everywhere,\nthen those two files no longer have their hands on all the levers, and\nwriting new port files becomes more difficult if not impossible.\nFurthermore, that style of working makes it more likely that developers\nwill break existing ports, by causing configure to write some choice\ninto some file without noticing that the port files for a platform other\nthan their own want to override that choice.\n\n(Yeah, I know that we'd like to reduce dependence on the port files\nby having configure auto-detect more things. But I don't expect to be\nable to eliminate the port files entirely --- there will always be some\nthings that it's not really practical to make configure detect.)\n\nSo I believe that it's good style and good sense to try to centralize\nthe results of configure in as few files as possible, preferably just\nthose two.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 01 Nov 1998 13:26:10 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] HPUX 10.01 build issues and solutions. " } ]
[ { "msg_contents": "We will not be splitting the CVS tree after the 6.4 release.\n\nWe will continue with one source tree, applying patches needed for any\n6.4.* minor releases.\n\nWhen we have finished with 6.4.* releases, we will mark the cvs tree,\nand start on 6.5. We mark it so we can supply patches to fix serious\nproblems that appear after 6.4.* has been closed.\n\nAny risky feature patches that are posted during this period will be\nheld, and applied once 6.4.* is finished. We expect the 6.4.* feature\nfreeze period to last about one month.\n\nCertain _safe_ feature patches will be applied during this period.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 30 Oct 1998 18:50:40 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "CVS trees and post-6.4 release issues" }, { "msg_contents": "On Fri, 30 Oct 1998, Bruce Momjian wrote:\n\n> We will not be splitting the CVS tree after the 6.4 release.\n\n\tActually, the 'splitting' and 'tagging' are one and the same, and\nit will be tagged the moment I roll the release tar ball...that is\nautomatic shrug Its been done for every past release...if you don't use\nthe branch, not a problem, but it is created...\n\n> We will continue with one source tree, applying patches needed for any\n> 6.4.* minor releases.\n> \n> When we have finished with 6.4.* releases, we will mark the cvs tree,\n> and start on 6.5. We mark it so we can supply patches to fix serious\n> problems that appear after 6.4.* has been closed.\n> \n> Any risky feature patches that are posted during this period will be\n> held, and applied once 6.4.* is finished. We expect the 6.4.* feature\n> freeze period to last about one month.\n> \n> Certain _safe_ feature patches will be applied during this period.\n\n\tI don't think so...I heard that one after the BETA freeze, and\nlook at the mess that created. If there is a feature freeze for\npost-release bugs to be fixed, features are frozen, period, end of\nstory, do not pass go, do not collect $200, end of story...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 30 Oct 1998 23:14:35 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CVS trees and post-6.4 release issues" }, { "msg_contents": "> On Fri, 30 Oct 1998, Bruce Momjian wrote:\n> \n> > We will not be splitting the CVS tree after the 6.4 release.\n> \n> \tActually, the 'splitting' and 'tagging' are one and the same, and\n> it will be tagged the moment I roll the release tar ball...that is\n> automatic shrug Its been done for every past release...if you don't use\n> the branch, not a problem, but it is created...\n\nYes.\n\n> \n> > We will continue with one source tree, applying patches needed for any\n> > 6.4.* minor releases.\n> > \n> > When we have finished with 6.4.* releases, we will mark the cvs tree,\n> > and start on 6.5. We mark it so we can supply patches to fix serious\n> > problems that appear after 6.4.* has been closed.\n> > \n> > Any risky feature patches that are posted during this period will be\n> > held, and applied once 6.4.* is finished. We expect the 6.4.* feature\n> > freeze period to last about one month.\n> > \n> > Certain _safe_ feature patches will be applied during this period.\n> \n> \tI don't think so...I heard that one after the BETA freeze, and\n> look at the mess that created. If there is a feature freeze for\n> post-release bugs to be fixed, features are frozen, period, end of\n> story, do not pass go, do not collect $200, end of story...\n\nThe only feature item I think we considered adding was Jan's LIMIT. Do\nyou want that? Let us know.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 30 Oct 1998 22:44:06 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] CVS trees and post-6.4 release issues" }, { "msg_contents": "On Fri, 30 Oct 1998, Bruce Momjian wrote:\n\n> The only feature item I think we considered adding was Jan's LIMIT. Do\n> you want that? Let us know.\n\n\tIf you can personally guarantee that it will in absolutely no way\ncause any problems with the system on any platform...go for it grin\n\n\tThe feature freeze is meant to allow us to cover over any\noperating system related tweaks...adding in a new feature opens the door\nfor new problems to creep in...\n\n Marc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 30 Oct 1998 23:53:05 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CVS trees and post-6.4 release issues" } ]
[ { "msg_contents": "Does anyone understand why configure.in wants to include libcurses in\n$(LIBS) ? There is no program in the PostgreSQL distribution that uses\nthat library (at least, I could not find any reference to curses.h).\n\nOn HPUX 10, including libcurses causes problems because it defines a\nselect() routine that has nothing to do with the select() system call.\nThere is a hack in Makefile.hpux that gets around this (by linking -lc\nahead of the libraries chosen by configure), but that is going to cause\nproblems for supporting HPUX 10.01, because signal considerations demand\nthat -lBSD come first on that platform.\n\nI would like to pull out the configure.in code that adds -lcurses to\nLIBS, but I'm hesitant to do it at this late stage in the release cycle\nwithout knowing why it was there in the first place. Did we use to have\na frontend that used curses, or something like that?\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 30 Oct 1998 18:54:13 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Why is libcurses added to build by configure?" }, { "msg_contents": "On Fri, 30 Oct 1998, Tom Lane wrote:\n\n> Does anyone understand why configure.in wants to include libcurses in\n> $(LIBS) ? There is no program in the PostgreSQL distribution that uses\n> that library (at least, I could not find any reference to curses.h).\n> \n> On HPUX 10, including libcurses causes problems because it defines a\n> select() routine that has nothing to do with the select() system call.\n> There is a hack in Makefile.hpux that gets around this (by linking -lc\n> ahead of the libraries chosen by configure), but that is going to cause\n> problems for supporting HPUX 10.01, because signal considerations demand\n> that -lBSD come first on that platform.\n> \n> I would like to pull out the configure.in code that adds -lcurses to\n> LIBS, but I'm hesitant to do it at this late stage in the release cycle\n> without knowing why it was there in the first place. Did we use to have\n> a frontend that used curses, or something like that?\n\n\tGood stuff, you are learning...do not pull anything out, period.\nDon't add anything in, period. There was a reason for it, but it was\nsooooo long ago that I don't recall what it was. It has worked up until\nnow, and if only HPUX 10.01 is reporting a problem with it...leave it.\nMark it as un-supported at this time, and work at fixing it post-release.\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 30 Oct 1998 23:16:31 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Why is libcurses added to build by configure?" }, { "msg_contents": "\nOn 30-Oct-98 Tom Lane wrote:\n> Does anyone understand why configure.in wants to include libcurses in\n> $(LIBS) ? There is no program in the PostgreSQL distribution that uses\n> that library (at least, I could not find any reference to curses.h).\n\nI'm guessing here, but could it be a replacement for things like readline?\nI know old OS versions could have any number of things stuck to it, and \nafter some of the things I've seen HP-UX do over the years, nothing really\nsurprises me anymore.\n\n> \n> On HPUX 10, including libcurses causes problems because it defines a\n> select() routine that has nothing to do with the select() system call.\n> There is a hack in Makefile.hpux that gets around this (by linking -lc\n> ahead of the libraries chosen by configure), but that is going to cause\n> problems for supporting HPUX 10.01, because signal considerations demand\n> that -lBSD come first on that platform.\n> \n> I would like to pull out the configure.in code that adds -lcurses to\n> LIBS, but I'm hesitant to do it at this late stage in the release cycle\n> without knowing why it was there in the first place. Did we use to have\n> a frontend that used curses, or something like that?\n\nHow wise would that be at this time? That may be something to do at\na time like a minor version release.\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n", "msg_date": "Fri, 30 Oct 1998 22:33:30 -0500 (EST)", "msg_from": "Vince Vielhaber <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] Why is libcurses added to build by configure?" }, { "msg_contents": "> I would like to pull out the configure.in code that adds -lcurses to\n> LIBS, but I'm hesitant to do it at this late stage in the release \n> cycle without knowing why it was there in the first place.\n\nYup. I think we are pretty much frozen by now. Seems like a v6.5 thing\nsince it may have effects on other platforms.\n\n - Tom\n", "msg_date": "Sat, 31 Oct 1998 03:36:36 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Why is libcurses added to build by configure?" }, { "msg_contents": "psql (at least in 6.3.2) uses curses and readline. I know this because I\ntried to compile under a SuSE 5.3 distribution. Configure couldn't find the\ncurses library, so it wasn't added to LIBS, so the linker didn't try to use\nit, so psql had unresolved references, all of which seems to indicate (even\nwithout reading the sources) that something in there wants curses. I.e.,\nit's not just configure.\n\nInterestingly, failing to find curses did NOT stop configure, it went on to\nmake everything. \"make all\" didn't stop when psql failed to build either - it\nwent on to report all well at the end. Not until \"make install\" did the\nfailure become evident (though it was logged of course).\n\nRegards, K.\n\nAm 30-Oct-98 schrieb Tom Lane:\n> Does anyone understand why configure.in wants to include libcurses in\n> $(LIBS) ? There is no program in the PostgreSQL distribution that uses\n> that library (at least, I could not find any reference to curses.h).\n> \n> On HPUX 10, including libcurses causes problems because it defines a\n> select() routine that has nothing to do with the select() system call.\n> There is a hack in Makefile.hpux that gets around this (by linking -lc\n> ahead of the libraries chosen by configure), but that is going to cause\n> problems for supporting HPUX 10.01, because signal considerations demand\n> that -lBSD come first on that platform.\n> \n> I would like to pull out the configure.in code that adds -lcurses to\n> LIBS, but I'm hesitant to do it at this late stage in the release cycle\n> without knowing why it was there in the first place. Did we use to have\n> a frontend that used curses, or something like that?\n> \n> regards, tom lane\n> \n\n---\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\nKarl Auer ([email protected]) Geschaeft/work +41-1-6327531\nKommunikation, ETHZ RZ Privat/home +41-1-4517941\nClausiusstrasse 59 Fax +41-1-6321225\nCH-8092 ZUERICH Switzerland\n", "msg_date": "Sat, 31 Oct 1998 12:30:02 +0100 (MET)", "msg_from": "Karl Auer <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] Why is libcurses added to build by configure?" }, { "msg_contents": "Karl Auer <[email protected]> writes:\n> psql (at least in 6.3.2) uses curses and readline. I know this because\n> I tried to compile under a SuSE 5.3 distribution.\n\nHmm. psql does use readline, but there's no direct dependency on\ncurses. (I verified this by building without curses; it works, here.)\n\nPerhaps there are some implementations of readline that depend on\ncurses? The GNU libreadline I have here appears to use termcap,\nbut not curses. There might be other versions though.\n\nI think it would be possible to extend configure to detect which\nlibrary(s) libreadline depends on, and only include those in the link.\nBut that's not something to risk at this stage...\n\n\t\t\tregards, tom lane\n", "msg_date": "Sat, 31 Oct 1998 12:44:36 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Why is libcurses added to build by configure? " }, { "msg_contents": "Tom Lane wrote:\n> \n> Karl Auer <[email protected]> writes:\n> > psql (at least in 6.3.2) uses curses and readline. I know this because\n> > I tried to compile under a SuSE 5.3 distribution.\n> \n> Hmm. psql does use readline, but there's no direct dependency on\n> curses. (I verified this by building without curses; it works, here.)\n> \n> Perhaps there are some implementations of readline that depend on\n> curses? The GNU libreadline I have here appears to use termcap,\n> but not curses. There might be other versions though.\n> \n> I think it would be possible to extend configure to detect which\n> library(s) libreadline depends on, and only include those in the link.\n> But that's not something to risk at this stage...\n\nreadline needs terminfo/termcap. On many Linux systems terminfo support\nis in the [n]curses library.\n\nPhil\n", "msg_date": "Sun, 01 Nov 1998 10:27:10 +0000", "msg_from": "Phil Thompson <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Why is libcurses added to build by configure?" }, { "msg_contents": "Phil Thompson <[email protected]> writes:\n> readline needs terminfo/termcap. On many Linux systems terminfo support\n> is in the [n]curses library.\n\nIn fact, what I discovered after some further tracking of dependencies\nis that the routines needed by readline are available in either\nlibtermcap or libcurses on HPUX. According to Horton's _Portable C\nSoftware_, having a separate libtermcap is Berkeley practice and folding\nit into libcurses is a SysV-ism.\n\nSo, the safest approach seems to be to leave configure alone and solve\nthe select()-in-libcurses problem with some hackery in Makefile.hpux.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 01 Nov 1998 13:52:06 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Why is libcurses added to build by configure? " }, { "msg_contents": "Doesn't work here :-) I suspect you are correct, because now that I inspect\nthe makefile output more closely, it is libreadline that has the unresolved\nreferences, not any object file in the source tree.\n\nRegards, K.\n\nAm 31-Oct-98 schrieb Tom Lane:\n> Karl Auer <[email protected]> writes:\n>> psql (at least in 6.3.2) uses curses and readline. I know this because\n>> I tried to compile under a SuSE 5.3 distribution.\n> \n> Hmm. psql does use readline, but there's no direct dependency on\n> curses. (I verified this by building without curses; it works, here.)\n> \n> Perhaps there are some implementations of readline that depend on\n> curses? The GNU libreadline I have here appears to use termcap,\n> but not curses. There might be other versions though.\n> \n> I think it would be possible to extend configure to detect which\n> library(s) libreadline depends on, and only include those in the link.\n> But that's not something to risk at this stage...\n> \n> regards, tom lane\n> \n\n---\n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\nKarl Auer ([email protected]) Geschaeft/work +41-1-6327531\nKommunikation, ETHZ RZ Privat/home +41-1-4517941\nClausiusstrasse 59 Fax +41-1-6321225\nCH-8092 ZUERICH Switzerland\n", "msg_date": "Mon, 02 Nov 1998 09:21:53 +0100 (MET)", "msg_from": "Karl Auer <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Why is libcurses added to build by configure?" } ]
[ { "msg_contents": "\n\tOk, here is a run down on the regression tests for\npgsql snapshot (09271998) on Linux/Alpha:\n\nint2 .. failed - Difference in error messages. (ok)\nint4 .. failed - Wrapped number when should have errored. (ok)\nfloat4 .. failed - Overflow and underflow numbers rounded to inf and\n 9.99999E-40 when error message should have resulted.\n\t\t (ok)\nfloat8 .. failed - Same as above. (ok)\nnumerology .. failed - First same as int4, then a backend failure.\n (Problem?)\ngeometry .. failed - Off by one in nth decimal place. (ok)\ndatetime .. failed - Problems with adding time to 'today'. (Problem!)\nreltime .. failed - More time calculation failures. (Problem!)\nabstime .. failed - All dates a long way off, by a few hundred years.\n (PROBLEM!)\ntinterval .. failed - Same as above. (PROBLEM!)\nhorology .. failed - Same as above. (PROBLEM!)\nselect_having .. failed - Backend failure, probably memory violation.\n (Problem)\nunion .. failed - Same as int4. (ok)\nrandom .. failed - Failure on a type and oid, wierd. (Problem?)\nmisc .. failed - More wierd errors. :( (Problem?)\n\n\tIt appears most of the problems are concentrated in the datetime\ncode, and the rest of the problems are cascade failures from trying to use\nthe faulty datetime functions for more extensive tests. There are a few\nother problems (i.e. select_having), but I feel that the datetime stuff\nneeds to be sorted out first.\n\tI have tried to figure out what is up with the datetime code many\ntimes, but always ended in a dead end. Any test programs I write to see if\nI can isolate the problem work fine. Any attempts at debugging are\nfruitless as gdb is not 100% stable on Linux/Alpha and gets lost sometimes\n(returning random values for memory address! :( ).\n\tWhat I have been able to track down that any call to localtime()\nlibrary call is returning a date at some point in the year 2136! The\nsystem clock is correct. localtime() does conversion between time formats\nand somewhere it is getting its feet stepped on. Date time data going in\nis fine, but coming out it is scrambled. Test programs that just run data\nthrough localtime() work just fine. \n\tAlso, this only seems to happen when 'today' (if I\nremember correctly) is used for datetime input. Actual dates seem to work\njust fine though.\n\tThe biggest trouble in finding this bug is how \"complex\" the\ndatetime code is... Actually after hacking at it for a few hours (with no\nsuccess) I have other words to describe it, but I will refrain. :) Might\nsome one explain to me what exactly is happening in the datetime code so I\nmight have some idea what is going on. I am relatively good at C, but even\nthis tests my skills pretty good. \n\tAs I find time, I will work further to resolve these problems. If\nany one else has any ideas or advice on how to solve this problem, please\nsend them my way!\n\n\tPS. If some of this appears vague, it is. I have not had time to\nhack recently on pgsql thanks to too much schoolwork. :( Just with the 6.4\nrelease going out soon, I though I better see what help I could find...\n\n\tPPS. Patch for pgsql snapshots to compile on Linux/Alpha are\ncoming shortly.\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": "Fri, 30 Oct 1998 19:37:00 -0600 (CST)", "msg_from": "Ryan Kirkpatrick <[email protected]>", "msg_from_op": true, "msg_subject": "Linux/Alpha Regression tests...." }, { "msg_contents": "> The biggest trouble in finding this bug is how \"complex\" the\n> datetime code is... Actually after hacking at it for a few hours (with \n> no success) I have other words to describe it, but I will refrain.\n\nI'll assume it would be a compliment :) I was happy with how simple I\nwas able to make the code. Date/time stuff ain't easy.\n\nI'm guessing that the problem is in one of two areas:\n\n1) the configure/build process is confused as to what kind of time\nsupport you have on your system. There are two styles of local time\nconversion interfaces on Unix systems: one which uses global variables\nto return info on time zones, and one which returns that info as part of\nthe tz structure. The third type, which is pretty much no timezone\nsupport, is probably not relevant, though if your system is really\nconfused you should eliminate that as a possibility. The second type is\nreentrant, the first is not. It doesn't really matter which kind your\nsystem uses because each backend is essentially single-threaded, but if\nit guesses wrong on what kind of system you have then the results will\nbe garbage.\n\n2) the Postgres code stomps on the global variables used for timezone\ninfo. I think I got that resolved, but I figured I should mention it\njust in case.\n\nIt is most likely that (1) is the problem. Try compiling with\n\"-DDATEDEBUG\" specified, and see if that helps you track through the\ncode. Though it seems like you already know where the problem is\nlocated, so I'd just pound on an example until you get it figured out.\n\nGood luck.\n\n - Tom\n", "msg_date": "Sat, 31 Oct 1998 03:32:57 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PORTS] Linux/Alpha Regression tests...." } ]
[ { "msg_contents": "\n\tOk, here they are, the long awaited patches to make pgsql compile\nonce again on Linux/Alpha. It turned out that our old friend S_LOCK was\nthe problem. Apparently some one had pulled the S_LOCK assembly code out\nof the static function that I put it in, and attempted to make it inlined\nagain (which is what broke it the first time around). On a whim, I\nsearched the net looking for an Alpha Assembly Programmer's Manual, and\nsee if it would help me any. I was lucky on both accounts. I found the\nmanual, and quickly figured out to make local labels. So... The S_LOCK\ncode for Linux/Alpha is finally safe for inlining! :)\n\tHere is the patch for ./pgsql/src/include/storage/s_lock.h\n\n89c87\n< bne $0, already_set \\n\\\n---\n> bne $0, 3f \\n\\\n91c89\n< bne $0, already_set \\n\\\n---\n> bne $0, 3f \\n\\\n94,95c92,93\n< beq $0, stqc_fail \\n\\\n< success: bis $31, $31, %1 \\n\\\n---\n> beq $0, 2f \\n\\\n> bis $31, $31, %1 \\n\\\n97,100c95,98\n< jmp $31, end\t \\n\\\n< stqc_fail: or $31, 1, $0\t \\n\\\n< already_set: bis $0, $0, %1\t \\n\\\n< end: nop \": \"=m\"(lock), \"=r\"(_res): :\"0\");\n---\n> jmp $31, 4f\t \\n\\\n> 2: or $31, 1, $0\t \\n\\\n> 3: bis $0, $0, %1\t \\n\\\n> 4: nop \": \"=m\"(lock), \"=r\"(_res): :\"0\");\n\n\tThough there are two other catches at the moment in compiling\npgsql on Alpha as well:\n\n\t1. The optimization flag, -O2, that is used, causes some code to\nfail to compile with very strange errors, and it causes s_locks to jam.\nApparently this is due to the (somewhat) unstable nature of gcc/egcs on\nAlphas at the moment. Turning it down to -O1 solves the problem and\neverything works fine.\n\n\t2. C++ is pretty much broken on Linux/Alpha (Debian 2.0 at least),\ndue to missing header files. This means that libpq++ can't be compiled.\nEither tell configure not to detect/use C++, or comment out the line in\ninterfaces/Makefile that causes libpq++ to be built.\n\n\tBeyond that, it should compile, install, and initdb just fine.\nRegressions tests are a mixed bag at this time (see my seperate message\nconcerning this), but look promising.\n\tHere is the specs on my machine for reference:\n\nXL366 Alpha w/192MB of RAM\nDebian 2.0 Alpha (unstable)\nLinux 2.0.34\ngcc 2.91.57-5\nglibc 2.0.7t-1\n\n\tI would be very interested in regression tests/results from\npeople running RedHat on thier Alphas. Would help me to narrow down what\nis causing some of the regression test problems. If you do send them to\nme, make sure to include what kernel, gcc, and glibc you are using.\nThanks.\n\n\tOk, that should pretty much cover it for the 6.4 release on the\nLinux/Alpha end. I would mark it as working, but still in need of further\nwork. TTYAL.\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": "Fri, 30 Oct 1998 19:53:28 -0600 (CST)", "msg_from": "Ryan Kirkpatrick <[email protected]>", "msg_from_op": true, "msg_subject": "Patches for Pgsql on Linux/Alpha (RE: Last Call...)" }, { "msg_contents": "> \n> \tOk, here they are, the long awaited patches to make pgsql compile\n> once again on Linux/Alpha. It turned out that our old friend S_LOCK was\n> the problem. Apparently some one had pulled the S_LOCK assembly code out\n> of the static function that I put it in, and attempted to make it inlined\n> again (which is what broke it the first time around). On a whim, I\n> searched the net looking for an Alpha Assembly Programmer's Manual, and\n> see if it would help me any. I was lucky on both accounts. I found the\n> manual, and quickly figured out to make local labels. So... The S_LOCK\n> code for Linux/Alpha is finally safe for inlining! :)\n> \tHere is the patch for ./pgsql/src/include/storage/s_lock.h\n> \n> 89c87\n> < bne $0, already_set \\n\\\n> ---\n> > bne $0, 3f \\n\\\n> 91c89\n> < bne $0, already_set \\n\\\n> ---\n> > bne $0, 3f \\n\\\n> 94,95c92,93\n> < beq $0, stqc_fail \\n\\\n> < success: bis $31, $31, %1 \\n\\\n> ---\n> > beq $0, 2f \\n\\\n> > bis $31, $31, %1 \\n\\\n> 97,100c95,98\n> < jmp $31, end\t \\n\\\n> < stqc_fail: or $31, 1, $0\t \\n\\\n> < already_set: bis $0, $0, %1\t \\n\\\n> < end: nop \": \"=m\"(lock), \"=r\"(_res): :\"0\");\n> ---\n> > jmp $31, 4f\t \\n\\\n> > 2: or $31, 1, $0\t \\n\\\n> > 3: bis $0, $0, %1\t \\n\\\n> > 4: nop \": \"=m\"(lock), \"=r\"(_res): :\"0\");\n> \n> \tThough there are two other catches at the moment in compiling\n> pgsql on Alpha as well:\n> \n\nI have applied this. However, we normally prefer context diffs. \nFortunately, this patch had old and new rows, so it was almost the same\nas a context diff.\n\nIt also only dealt with the alpha locking code, which he stated was\nclearly broken, so application seemed good. Some of your more agressive\npatches may have to wait a week until 6.4 is released.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 30 Oct 1998 21:15:01 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PATCHES] Patches for Pgsql on Linux/Alpha (RE: Last Call...)" }, { "msg_contents": "On Fri, 30 Oct 1998, Bruce Momjian wrote:\n\n> I have applied this. However, we normally prefer context diffs. \n> Fortunately, this patch had old and new rows, so it was almost the same\n> as a context diff.\n\n\tSorry... I was in a bit of a hurry to get it out, and just grabbed\nthe a patch I had made a while back. It was so small and only covered one\nfile, I didn't think there would be much problem.\n\n> It also only dealt with the alpha locking code, which he stated was\n> clearly broken, so application seemed good. Some of your more agressive\n> patches may have to wait a week until 6.4 is released.\n\n\tBy more aggressive patches, do you mean the optimizer and libpq++\nadvice, or did I miss something? Also, by 'he' and 'you' who do you mean?\nI think you were speaking to more than one person, but I am not sure...\nPlease clarify, thanks. :)\n\n----------------------------------------------------------------------------\n\| \"For to me to live is Christ, and to die is gain.\" \|\n\| --- Philippians 1:21 (KJV) \|\n----------------------------------------------------------------------------\n\| Ryan Kirkpatrick \| Boulder, Colorado \| [email protected] \|\n----------------------------------------------------------------------------\n\| http://www-ugrad.cs.colorado.edu/~rkirkpat/ \|\n----------------------------------------------------------------------------\n\n\n", "msg_date": "Fri, 30 Oct 1998 20:39:38 -0600 (CST)", "msg_from": "Ryan Kirkpatrick <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [PATCHES] Patches for Pgsql on Linux/Alpha (RE: Last Call...)" }, { "msg_contents": "> On Fri, 30 Oct 1998, Bruce Momjian wrote:\n> \n> > I have applied this. However, we normally prefer context diffs. \n> > Fortunately, this patch had old and new rows, so it was almost the same\n> > as a context diff.\n> \n> \tSorry... I was in a bit of a hurry to get it out, and just grabbed\n> the a patch I had made a while back. It was so small and only covered one\n> file, I didn't think there would be much problem.\n\nYes, I am just a little paranoid this close to final release, November\n1.\n\n> \n> > It also only dealt with the alpha locking code, which he stated was\n> > clearly broken, so application seemed good. Some of your more agressive\n> > patches may have to wait a week until 6.4 is released.\n> \n> \tBy more aggressive patches, do you mean the optimizer and libpq++\n> advice, or did I miss something? Also, by 'he' and 'you' who do you mean?\n> I think you were speaking to more than one person, but I am not sure...\n> Please clarify, thanks. :)\n\n'he' and 'your' is you. By more agressive, I mean things that could\naffect other platforms, were we can't get 100% agreement from the other\ndevelopers that this is a safe change.\n\nLet's see what you find, and what other say. I don't want to apply\nsomething at this point that is going to potentailly break another\nplatform, and the date/time routines have historically done this.\n\nWe can put them in a 6.4.* minor release, which I am sure we will have a\nfew weeks after the major.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 30 Oct 1998 21:53:14 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [PATCHES] Patches for Pgsql on Linux/Alpha (RE: Last Call...)" }, { "msg_contents": "On Fri, 30 Oct 1998, Bruce Momjian wrote:\n\n> Yes, I am just a little paranoid this close to final release, November\n> 1.\n\n\tUnderstandable. Next time they will be in a much cleaner format.\n\n> > \tBy more aggressive patches, do you mean the optimizer and libpq++\n> > advice, or did I miss something? Also, by 'he' and 'you' who do you mean?\n> > I think you were speaking to more than one person, but I am not sure...\n> > Please clarify, thanks. :)\n> \n> 'he' and 'your' is you. By more agressive, I mean things that could\n> affect other platforms, were we can't get 100% agreement from the other\n> developers that this is a safe change.\n> \n> Let's see what you find, and what other say. I don't want to apply\n> something at this point that is going to potentailly break another\n> platform, and the date/time routines have historically done this.\n> \n> We can put them in a 6.4.* minor release, which I am sure we will have a\n> few weeks after the major.\n\n\tOk, I follow now, thanks for the clarification. The patch to the\nAlpha S_LOCK code is the only patch I have at this time. I guess I should\nhave been clearer about the datetime problems. I do not have even a\nprospective patch for those problems at this time, or plan to have the\ntime to even begin such a patch for at least a week or two. When I do get\nsuch a patch, I will make sure to have a few other Linux/Alpha people test\nit, as well as test what it does to a Linux/Intel compile of pgsql before\nI post it to be added to the CVS tree. \n\tThe alpha S_LOCK patch only affected a single OS, and thought it\nwould be rather safe. Next time I will try and get it in earlier (last\nweek or two has been hectic with school :( ). Thanks for your\nunderstanding. TTYL.\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": "Sat, 31 Oct 1998 09:10:51 -0600 (CST)", "msg_from": "Ryan Kirkpatrick <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [PATCHES] Patches for Pgsql on Linux/Alpha (RE: Last Call...)" } ]
[ { "msg_contents": "afaik the docs are now finished for v6.4. It's 1:30 in the morning, so I\nmay be forgetting something; send mail or fix it yourself :)\n\nBruce, there is a new file current.sgml which is the current release\nnotes. It is a copy of part of your TODO file, so you can diff it with\nthe TODO to see the changes.\n\nI'm sleeping in a bit and then leaving town for the weekend. Send mail\nand I can make fixes when I return (if it isn't too late for the\nrelease).\n\nThanks everyone for the docs contributions! We have over 500 pages of\nhardcopy docs of increasingly good quality, and lots of new info for\nthis release. I just realized that I short-changed the docs themselves\nin summarizing changes for this release, and will fix that if I have a\nchance before the release. Otherwise we'll write it up and take credit\nfor new stuff in v6.5 too ;)\n\n - Thomas\n", "msg_date": "Sat, 31 Oct 1998 09:43:30 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Docs finished for v6.4?" } ]
[ { "msg_contents": "I've posted the \"final versions\" of the v6.4 docs on the web site at\n\n ftp://postgresql.org/pub/patches/*.{tar,ps}.gz\n\nAlso, a new psqlODBC standalone tarball is available in the same\nlocation.\n\nI'm out of town until Sunday evening, so if things look solid for a v6.4\nrelease before then good luck! (Hmm, that doesn't sound quite right, but\nyou know what I mean :) Also, if someone finds \"release-busting\" flaws\nin these \"final docs\" I can fix them Sunday eve or Monday morning.\n\n - Thomas\n", "msg_date": "Sat, 31 Oct 1998 16:13:32 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "New docs and psqlODBC" } ]
[ { "msg_contents": "\n\nI was hoping someone could shed some light on the \nfollowing problem:\n\nThe subselect under 6.4 ignores indexes built on \ntables. For example, consider the following two\ntables:\n\n\\d sales;\n\nTable = sales\n\nsupplysource varchar() not null 16\nsupply varchar() not null 16\nsupplyunit varchar() not null 2\nquantity float8 not null 8\ntarget varchar() not null 16\ncostcntr varchar() not null 8\nsaletype varchar() not null 16\nsaledate datetime not null 8\n\nIndices: k_sales\n k_sales_saledate\n k_sales_supply\n k_sales_target \n\n\n\\d locations;\n\nTable = locations\nlocation varchar() not null 16\ncostcntr varchar() not null 5\nsupplypath varchar() not null 64\nformpath varchar() not null 64\nengineerpath varchar() not null 64\n\nIndices: k_locations\n k_locations_costcntr\n k_locations_location \n\nwith the following query:\n\nexplain select supply from sales where target in\n(select location from locations);\nNOTICE: QUERY PLAN:\n\nSeq Scan on sales (cost=5738.60 size=116806 width=12)\n SubPlan\n -> Seq Scan on locations (cost=7.49 size=136 width=12)\n\nEXPLAIN\n\nThis results in a full table scan on the outer table\nwhere there are 150K rows. As a result, it is the\nequivalent of a full table scan on 150K (sales) * \n100 (locations) rows = 15,000,000 row table scan.\n\nOracle 8.0.5.00 for Linux Early Adaptor Program \ncorrectly uses the index on the same two tables\nresulting on a sequential access on the subselected\ntable (locations) and an index look-up on the \nselect table (sales), according to EXPLAIN PLAN.\n\nDue to the failure of PostgreSQL 6.4 to correctly\nuse indexes with subselects, it makes subselects\npractically unusable.\n\n(NOTE: This occurs regardless of whether or not it\nis executed immediately following a VACUUM)\n\nAre there any patches available to fix this problem?\nWill this fixed in the 6.4 release?\n\nThanks for any information,\n\nMarcus Mascari\n([email protected]) \n\n\n\n Date:\n Tue, 27 Oct 1998 17:30:44 -0800 (PST)\n From:\n Marcus Mascari <[email protected]>Add to Address Book\n Subject:\n Please help (Someone?) - subselect not using indexes\n To:\n [email protected]\n\n\n\nI was hoping someone could shed some light on the \nfollowing problem:\n\nThe subselect under 6.4 ignores indexes built on \ntables. For example, consider the following two\ntables:\n\n\\d sales;\n\nTable = sales\n\nsupplysource varchar() not null 16\nsupply varchar() not null 16\nsupplyunit varchar() not null 2\nquantity float8 not null 8\ntarget varchar() not null 16\ncostcntr varchar() not null 8\nsaletype varchar() not null 16\nsaledate datetime not null 8\n\nIndices: k_sales\n k_sales_saledate\n k_sales_supply\n k_sales_target \n\n\n\\d locations;\n\nTable = locations\nlocation varchar() not null 16\ncostcntr varchar() not null 5\nsupplypath varchar() not null 64\nformpath varchar() not null 64\nengineerpath varchar() not null 64\n\nIndices: k_locations\n k_locations_costcntr\n k_locations_location \n\nwith the following query:\n\nexplain select supply from sales where target in\n(select location from locations);\nNOTICE: QUERY PLAN:\n\nSeq Scan on sales (cost=5738.60 size=116806 width=12)\n SubPlan\n -> Seq Scan on locations (cost=7.49 size=136 width=12)\n\nEXPLAIN\n\nThis results in a full table scan on the outer table\nwhere there are 150K rows. As a result, it is the\nequivalent of a full table scan on 150K (sales) * \n100 (locations) rows = 15,000,000 row table scan.\n\nOracle 8.0.5.00 for Linux Early Adaptor Program \ncorrectly uses the index on the same two tables\nresulting on a sequential access on the subselected\ntable (locations) and an index look-up on the \nselect table (sales), according to EXPLAIN PLAN.\n\nDue to the failure of PostgreSQL 6.4 to correctly\nuse indexes with subselects, it makes subselects\npractically unusable.\n\n(NOTE: This occurs regardless of whether or not it\nis executed immediately following a VACUUM)\n\nAre there any patches available to fix this problem?\nWill this fixed in the 6.4 release?\n\nThanks for any information,\n\nMarcus Mascari\n([email protected])\n\n\n\n\n_________________________________________________________\nDO YOU YAHOO!?\nGet your free @yahoo.com address at http://mail.yahoo.com\n\n\n Message - Download\n \n\n \n Back\nto Sent \n\n\n\nCopyright �� 1997-98 Yahoo! Inc. All rights reserved. \n\n\n\n\n\n\n\n_________________________________________________________\nDO YOU YAHOO!?\nGet your free @yahoo.com address at http://mail.yahoo.com\n\n", "msg_date": "Sat, 31 Oct 1998 11:16:17 -0800 (PST)", "msg_from": "Marcus Mascari <[email protected]>", "msg_from_op": true, "msg_subject": "PLEASE...SOMEONE COMMENT: PostgreSQL 6.4BETA not using indexes with\n\tsubselects" } ]
[ { "msg_contents": "I have a couple questions before I start a large project using postgres.\n I'm posting this to the hackers list because if the answer to any of\nthese questions is no, if somebody explains why, I would be glad to help\nimplement some features once I get into using it everyday. So, here we go:\n\n1) How much data can postgres handle? If I'm restricted by maximum\nfile size in the OS, can I segment the data file, and if so, how many times?\n\n2) Can I store the indexes for the database on a separate drive from\nthe data itself? This is useful so that I can place the indexes on a\nsolid state disk for fast searching, and leave the data on traditional medium.\n\n3) Is there a maximum number of records per table? i.e., if I can\nsegment the data file, can I still only have 2 gigs of data per table?\n\n-- \nNick Bastin\nRBB Systems, Inc.\n", "msg_date": "Sat, 31 Oct 1998 14:33:49 -0500", "msg_from": "Nick Bastin <[email protected]>", "msg_from_op": true, "msg_subject": "A couple of questions..." }, { "msg_contents": "> 1) How much data can postgres handle? If I'm restricted by maximum\n> file size in the OS, can I segment the data file, and if so, how many times?\n\nI vaguely recall that Postgres has support for segmented files, but that\nit has not been reported to have been tested recently. You probably\ncan't count on it working, but try testing and see.\n\n> 2) Can I store the indexes for the database on a separate drive from\n> the data itself? This is useful so that I can place the indexes on a\n> solid state disk for fast searching, and leave the data on traditional \n> medium.\n\nOfficially, no. If you have a fairly static database, you can create the\nindex, and then move it to another disk and create a soft link to it.\nThe Postgres server doesn't know about this directly, but can probably\nbe fooled.\n\n> 3) Is there a maximum number of records per table? i.e., if I can\n> segment the data file, can I still only have 2 gigs of data per table?\n\nDon't know. Sorry.\n\n - Tom\n", "msg_date": "Mon, 02 Nov 1998 06:15:50 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A couple of questions..." }, { "msg_contents": "On Mon, 2 Nov 1998, Thomas G. Lockhart wrote:\n\n> > 1) How much data can postgres handle? If I'm restricted by maximum\n> > file size in the OS, can I segment the data file, and if so, how many times?\n> \n> I vaguely recall that Postgres has support for segmented files, but that\n> it has not been reported to have been tested recently. You probably\n> can't count on it working, but try testing and see.\n\nI did start to test this about 2 weeks ago, but I hit problems way before\nI got to that point. I'm in the middle of running the test again at the\nmoment.\n\n> > 2) Can I store the indexes for the database on a separate drive from\n> > the data itself? This is useful so that I can place the indexes on a\n> > solid state disk for fast searching, and leave the data on traditional \n> > medium.\n> \n> Officially, no. If you have a fairly static database, you can create the\n> index, and then move it to another disk and create a soft link to it.\n> The Postgres server doesn't know about this directly, but can probably\n> be fooled.\n> \n> > 3) Is there a maximum number of records per table? i.e., if I can\n> > segment the data file, can I still only have 2 gigs of data per table?\n> \n> Don't know. Sorry.\n> \n> - Tom\n> \n\n-- \n Peter T Mount [email protected]\n Main Homepage: http://www.retep.org.uk\nPostgreSQL JDBC Faq: http://www.retep.org.uk/postgres\n Java PDF Generator: http://www.retep.org.uk/pdf\n\n", "msg_date": "Mon, 2 Nov 1998 20:47:43 +0000 (GMT)", "msg_from": "Peter T Mount <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A couple of questions..." } ]
[ { "msg_contents": "Until earlier this week, the various snapshots have been working fine\non my system, NetBSD 1.3.2/i386. As of a couple of days ago the\nbackend started to crash. I hoped this was a temporary glitch with\nrecent patches which would disappear in a day or so as other stuff got\nsorted out. Nothing seems to have changed, though, over the last\nseveral days, and connections fail with the backend crashing.\n\nEverything seems to compile fine; the only warnings during backend\ncompilation are given below. The select warning in s_lock.c requires\nan\n\n #include <unistd.h>\n\nto remove it, but that doesn't fix the crashing problem. I would\nsuggest a patch for that, but I'm not sure what systems have unistd.h\nand what don't, so I'm not sure if the obvious thing of putting that\nline in s_lock.c is the right thing to do.\n\nCan anyone with more experience tracking down crashing backends give\nsome guidance? I hate to see 6.4 shipped with one of the supported\nbackends crashing! Unfortunately, I'm not sure where to look. I\ndidn't notice any suspicious patches coming through in the last few\ndays, but I don't see everything that is committed. I'm also not sure\nif the NetBSD/vax patches could have affected NetBSD/i386 stuff.\n\nAny help is greatly appreciated!\n\nCheers,\nBrook\n\n===========================================================================\nWarnings found during backend compile; directories noted, but lots of\ncommands deleted. In the past I have seen warnings about some of the\nbison/lexer stuff, so I tend to ignore them; I'm not sure if these are\ndifferent than the \"normal\" warnings.\n\n gmake[2]: Entering directory `/usr/pkgsrc-local/databases/postgresql-current/work/pgsql/src/backend/bootstrap'\n gcc -I../../include -I../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I.. -Wno-error -c bootparse.c -o bootparse.o\n /usr/pkg/share/bison.simple: In function `Int_yyparse':\n /usr/pkg/share/bison.simple:327: warning: implicit declaration of function `Int_yyerror'\n /usr/pkg/share/bison.simple:387: warning: implicit declaration of function `Int_yylex'\n gcc -I../../include -I../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I.. -Wno-error -c bootscanner.c -o bootscanner.o\n lex.Int_yy.c:683: warning: no previous prototype for `Int_yylex'\n bootscanner.l:137: warning: no previous prototype for `Int_yyerror'\n\n gmake[2]: Entering directory `/usr/pkgsrc-local/databases/postgresql-current/work/pgsql/src/backend/parser'\n gcc -I../../include -I../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I.. -Wno-error -c gram.c -o gram.o\n /usr/pkg/share/bison.simple: In function `yyparse':\n /usr/pkg/share/bison.simple:327: warning: implicit declaration of function `yyerror'\n /usr/pkg/share/bison.simple:387: warning: implicit declaration of function `yylex'\n gcc -I../../include -I../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I.. -Wno-error -c scan.c -o scan.o\n lex.yy.c:820: warning: no previous prototype for `yylex'\n scan.l:426: warning: no previous prototype for `yyerror'\n lex.yy.c:2174: warning: `yy_flex_realloc' defined but not used\n\n gmake[3]: Entering directory `/usr/pkgsrc-local/databases/postgresql-current/work/pgsql/src/backend/storage/buffer'\n gcc -I../../../include -I../../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I../.. -c s_lock.c -o s_lock.o\n s_lock.c: In function `s_lock':\n s_lock.c:70: warning: implicit declaration of function `select'\n\n gmake[2]: Entering directory `/usr/pkgsrc-local/databases/postgresql-current/work/pgsql/src/backend/utils'\n gcc -I../../../include -I../../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I../.. -c network.c -o network.o\n network.c: In function `network_network':\n network.c:392: warning: unused variable `ptr'\n gcc -I../../../include -I../../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I../.. -c inet_net_ntop.c -o inet_net_ntop.o\n inet_net_ntop.c: In function `inet_net_ntop_ipv4':\n inet_net_ntop.c:192: warning: unused variable `m'\n gcc -I../../../include -I../../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I../.. -c inet_net_pton.c -o inet_net_pton.o\n inet_net_pton.c: In function `inet_cidr_pton_ipv4':\n inet_net_pton.c:104: warning: `tmp' might be used uninitialized in this function\n\n gmake[3]: Entering directory `/usr/pkgsrc-local/databases/postgresql-current/work/pgsql/src/backend/utils/fmgr'\n gcc -I../../../include -I../../../backend -I/usr/pkg/include -I/usr/pkg/include/tcl8.0 -I/usr/pkg/include/tk8.0 -O2 -pipe -Wall -Wmissing-prototypes -I../.. -c dfmgr.c -o dfmgr.o\n dfmgr.c:283: warning: no previous prototype for `trigger_dynamic'\n", "msg_date": "Sun, 1 Nov 1998 09:09:51 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": true, "msg_subject": "backend crashing on NetBSD 1.3.2/i386" }, { "msg_contents": "> Until earlier this week, the various snapshots have been working fine\n> on my system, NetBSD 1.3.2/i386. As of a couple of days ago the\n> backend started to crash. I hoped this was a temporary glitch with\n> recent patches which would disappear in a day or so as other stuff got\n> sorted out. Nothing seems to have changed, though, over the last\n> several days, and connections fail with the backend crashing.\n\nYou have to run the backend using gdb, and get a backtrace of the crash,\nor check the postmaster logs for any information.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sun, 1 Nov 1998 12:12:26 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386" }, { "msg_contents": "Brook Milligan <[email protected]> writes:\n> Until earlier this week, the various snapshots have been working fine\n> on my system, NetBSD 1.3.2/i386. As of a couple of days ago the\n> backend started to crash.\n\nWhen did this start, exactly?\n\n> I hoped this was a temporary glitch with\n> recent patches which would disappear in a day or so as other stuff got\n> sorted out. Nothing seems to have changed, though, over the last\n> several days, and connections fail with the backend crashing.\n\nA reproducible crash at startup ought to be pretty easy to nail down.\nDoes it produce a corefile? If so fire up gdb and get a backtrace so\nwe can see where the crash occurs. (The trace would give more info if\nyou compiled with -g, but even without it would be helpful.)\n\nOne thing that comes to mind quickly is that some of the changes this\nweek required an initdb to be fully effective. If you forgot the initdb\nmaybe a crash at startup would result; I'm not sure.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 01 Nov 1998 12:23:07 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386 " }, { "msg_contents": "I'm running an installation on NetBSD/i386 1.3.2 that I upgraded\n(using 'cvs update') just a couple of hours ago, and it's behaving\njust fine. I've loaded a few megabytes of data into it, and done a\nbit of updating and querying, with no problems.\n\nTom Lane <[email protected]> writes:\n\n> One thing that comes to mind quickly is that some of the changes this\n> week required an initdb to be fully effective. If you forgot the initdb\n> maybe a crash at startup would result; I'm not sure.\n\nThat may be it. I always use pg_dump and initdb when I update the\nsystems I run a current snapshot of PostgreSQL on, just to be sure.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "01 Nov 1998 19:18:10 +0100", "msg_from": "Tom Ivar Helbekkmo <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386" }, { "msg_contents": " You have to run the backend using gdb, and get a backtrace of the crash,\n or check the postmaster logs for any information.\n\nNo core dump that I can find. But, the logs are reporting errors from\nsemget in ipc.c. Specifically, it is complaining of not enough space,\nbut I have no full filesystems. This stuff is part of the locking\ncode isn't it? Did that get tweaked with the NetBSD/vax patches\nrecently?\n\nCheers,\nBrook\n", "msg_date": "Sun, 1 Nov 1998 15:25:58 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386" }, { "msg_contents": " > Until earlier this week, the various snapshots have been working fine\n > on my system, NetBSD 1.3.2/i386. As of a couple of days ago the\n > backend started to crash.\n\n When did this start, exactly?\n\nSometime between Monday and Thursday, but Marc just trashed the old\nBETA* files so I can't go back and check. I did manage to catch the\nBETA3 just before it disappeared and it exhibits the problem.\n\n One thing that comes to mind quickly is that some of the changes this\n week required an initdb to be fully effective. If you forgot the initdb\n maybe a crash at startup would result; I'm not sure.\n\nNope. For testing the new version I've been doing a clean install\nevery time, including initdb and everything.\n\nCheers,\nBrook\n", "msg_date": "Sun, 1 Nov 1998 15:28:38 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386" }, { "msg_contents": " I'm running an installation on NetBSD/i386 1.3.2 that I upgraded\n (using 'cvs update') just a couple of hours ago, and it's behaving\n just fine. I've loaded a few megabytes of data into it, and done a\n bit of updating and querying, with no problems.\n\nDo you see the warnings about select() when compiling s_lock.c?\n\nCheers,\nBrook\n", "msg_date": "Sun, 1 Nov 1998 15:31:11 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386" }, { "msg_contents": "On Sun, 1 Nov 1998, Brook Milligan wrote:\n\n> You have to run the backend using gdb, and get a backtrace of the crash,\n> or check the postmaster logs for any information.\n> \n> No core dump that I can find. But, the logs are reporting errors from\n> semget in ipc.c. Specifically, it is complaining of not enough space,\n> but I have no full filesystems. This stuff is part of the locking\n> code isn't it? Did that get tweaked with the NetBSD/vax patches\n> recently?\n\n\tsemget deals with shared memory, not file systems...under FreeBSD,\nyou do:\n\n%ipcs\nMessage Queues:\nT ID KEY MODE OWNER GROUP\n\nShared Memory:\nT ID KEY MODE OWNER GROUP\nm 131072 2063597841 --rw-rw-rw- scrappy staff\n\nSemaphores:\nT ID KEY MODE OWNER GROUP\n\n\n\tTo see what is being used. ipcrm to remove 'stale'\nhandles...sounds like your system isn't releasing when you kill of\npostgres daemon...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sun, 1 Nov 1998 18:59:34 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386" }, { "msg_contents": "\t semget deals with shared memory, not file systems...under FreeBSD,\n\nThere error message said \"No space on device\" because semget returned\nENOSPC, hence my initial confusion about what was going on.\n\n\t To see what is being used. ipcrm to remove 'stale'\n handles...sounds like your system isn't releasing when you kill of\n postgres daemon...\n\nThat was it! Thanks, Marc. I have no idea where the extra semaphores\ncame from (they weren't owned by pgsql), but everything works again\nafter they were deleted (that is BETA5 passes regression on NetBSD\n1.3.2/i386). Does a kill signal to the postmaster prevent the cleanup\nof these? Should postmasters be killed with HUP?\n\nSorry for the diversion. Now back to your regularly scheduled\nrelease. :)\n\nThanks for the help.\n\nCheers,\nBrook\n", "msg_date": "Sun, 1 Nov 1998 21:51:15 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386" }, { "msg_contents": "Brook Milligan <[email protected]> writes:\n\n> Do you see the warnings about select() when compiling s_lock.c?\n\nYup.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "02 Nov 1998 12:37:44 +0100", "msg_from": "Tom Ivar Helbekkmo <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] backend crashing on NetBSD 1.3.2/i386" } ]
[ { "msg_contents": "\nMorning...\n\n\tJust put up a BETA5 tar file...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Sun, 1 Nov 1998 13:00:18 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "BETA5 ..." } ]
[ { "msg_contents": "It seems that we have a small problem with the operators and functions\nin the inet and cidr types. If the value in a field is null then the\nbackend crashes when you try to apply one of the operators or functions\non it. I have a small patch ready to fix the functions but I'm not\nsure what the behaviour should be for the operators. For the functions\nI return an empty string if the argument is null except for masklen()\nwhich returns 0. Should masklen return 32?\n\nAt Marc's request I am holding my patch until 6.4 is frozen.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Sun, 1 Nov 1998 13:19:11 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "A small problem with the new inet and cidr types" }, { "msg_contents": "> It seems that we have a small problem with the operators and functions\n> in the inet and cidr types. If the value in a field is null then the\n> backend crashes when you try to apply one of the operators or functions\n> on it. I have a small patch ready to fix the functions but I'm not\n> sure what the behaviour should be for the operators. For the functions\n> I return an empty string if the argument is null except for masklen()\n> which returns 0. Should masklen return 32?\n> \n> At Marc's request I am holding my patch until 6.4 is frozen.\n\nHuh. You have a backend crash fix, and we can't apply it until after\n6.4 is out the door?\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sun, 1 Nov 1998 17:44:13 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "On Sun, 1 Nov 1998, Bruce Momjian wrote:\n\n> > It seems that we have a small problem with the operators and functions\n> > in the inet and cidr types. If the value in a field is null then the\n> > backend crashes when you try to apply one of the operators or functions\n> > on it. I have a small patch ready to fix the functions but I'm not\n> > sure what the behaviour should be for the operators. For the functions\n> > I return an empty string if the argument is null except for masklen()\n> > which returns 0. Should masklen return 32?\n> > \n> > At Marc's request I am holding my patch until 6.4 is frozen.\n> \n> Huh. You have a backend crash fix, and we can't apply it until after\n> 6.4 is out the door?\n\n\tD'Arcy is going to release a patch for the fix on the same day as\nthe release, as it is something that only affects a very small aspect of\nthe server, and will affect very few ppl...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 2 Nov 1998 01:12:28 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake Bruce Momjian\n> > At Marc's request I am holding my patch until 6.4 is frozen.\n> \n> Huh. You have a backend crash fix, and we can't apply it until after\n> 6.4 is out the door?\n\nWell, it wasn't complete anyway since I wasn't sure what to do about\nthe operators. However, I have since figured it out, I think. I\nthink that if either argument to an operator function is NULL then\nthe function should return FALSE. My reasoning is that it should be\nconsistent and people should be able to decide the sense of the test\nfor the results they expect. Every test can be expressed as the\nnegation of another test so the user can decide whether they want\nnulls to be included or not. For example;\n\nSELECT c < i FROM inet_tbl;\nSELECT NOT (c >= i) FROM inet_tbl;\n\nWhile both of the above selects are nominally equivalent, in the first\ncase rows where i1 or i2 is null will return 0 while in the second case\nthey return 1. Does this seem right?\n\nNow, after making the above changes to my local copy I tested it and I\nfound that when one of the fields was null, the bool result is blank, not\nt or f. I'm sure it goes to the functions as it crashes without the\nchanges so I'm not sure why it is blank after.\n\nOne more thing, I'm not sure how this gets fixed but the function results\nare still right justified. Weren't we going to left justify them?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 2 Nov 1998 08:40:12 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake The Hermit Hacker\n> > Huh. You have a backend crash fix, and we can't apply it until after\n> > 6.4 is out the door?\n> \n> \tD'Arcy is going to release a patch for the fix on the same day as\n> the release, as it is something that only affects a very small aspect of\n> the server, and will affect very few ppl...\n\nNot so sure about that, Marc. It is, however, easy to work around. You\njust have to set NOT NULL for any inet or cidr type but I already have\na use for null ip numbers. That's how I found the problem.\n\nOf course, it's such a new type that I'm sure it won't be heavily used for\na little while anyway.\n\nI just had another thought. Do any other types exhibit this problem?\nI think I'll go check enhance a few regression tests.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 2 Nov 1998 08:43:47 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake D'Arcy J.M. Cain\n> I just had another thought. Do any other types exhibit this problem?\n> I think I'll go check enhance a few regression tests.\n\nOK, there are more problems. If you apply the following patch to the\nregression tests you will crash the backend in a number of places. I\nalso added some null inserts in places that it didn't crash (the date\nand time functions are pretty clean) but it will, of course, change\nthe expected output.\n\nI won't send this to patches since it crashes the backend. I'll let others\nfix the tests for the types that they fix.\n\n\n* ../sql/abstime.sql\tSat May 31 22:30:19 1997\n--- abstime.sql\tMon Nov 2 09:17:48 1998\n***********\n* 23,28 **\n--- 23,30 ----\n \n INSERT INTO ABSTIME_TBL (f1) VALUES ('May 10, 1947 23:59:12');\n \n+ INSERT INTO ABSTIME_TBL (f1) VALUES (null);\n+ \n \n -- what happens if we specify slightly misformatted abstime? \n INSERT INTO ABSTIME_TBL (f1) VALUES ('Feb 35, 1946 10:00:00');\n* ../sql/char.sql\tSun Nov 30 21:46:01 1997\n--- char.sql\tMon Nov 2 09:24:51 1998\n*************\n* 30,35 **\n--- 30,38 ----\n -- zero-length char \n INSERT INTO CHAR_TBL (f1) VALUES ('');\n \n+ -- null input\n+ INSERT INTO CHAR_TBL (f1) VALUES (null);\n+ \n -- try char's of greater than 1 length \n INSERT INTO CHAR_TBL (f1) VALUES ('cd');\n \n* ../sql/circle.sql\tTue Jul 29 12:22:44 1997\n--- circle.sql\tMon Nov 2 09:26:18 1998\n*************\n* 16,21 **\n--- 16,23 ----\n \n INSERT INTO CIRCLE_TBL VALUES ('<(100,0),100>');\n \n+ INSERT INTO CIRCLE_TBL VALUES (null);\n+ \n -- bad values\n \n INSERT INTO CIRCLE_TBL VALUES ('<(-100,0),-100>');\n* ../sql/datetime.sql\tFri Nov 14 21:55:57 1997\n--- datetime.sql\tMon Nov 2 09:30:37 1998\n*************\n* 18,23 ***\n--- 18,24 ----\n INSERT INTO DATETIME_TBL VALUES ('tomorrow');\n INSERT INTO DATETIME_TBL VALUES ('tomorrow EST');\n INSERT INTO DATETIME_TBL VALUES ('tomorrow zulu');\n+ INSERT INTO DATETIME_TBL VALUES (null);\n \n SELECT count() AS one FROM DATETIME_TBL WHERE d1 = 'today'::datetime;\n SELECT count() AS one FROM DATETIME_TBL WHERE d1 = 'tomorrow'::datetime;\n************\n* 42,47 **\n--- 43,49 ----\n INSERT INTO DATETIME_TBL VALUES ('-infinity');\n INSERT INTO DATETIME_TBL VALUES ('infinity');\n INSERT INTO DATETIME_TBL VALUES ('epoch');\n+ INSERT INTO DATETIME_TBL VALUES (null);\n \n -- Postgres v6.0 standard output format\n INSERT INTO DATETIME_TBL VALUES ('Mon Feb 10 17:32:01 1997 PST');\n* ../sql/horology.sql\tSat Sep 20 12:34:07 1997\n--- horology.sql\tMon Nov 2 09:34:55 1998\n*************\n* 15,20 **\n--- 15,22 ----\n WHERE d1 BETWEEN '13-jun-1957' AND '1-jan-1997'\n OR d1 BETWEEN '1-jan-1999' AND '1-jan-2010';\n \n+ INSERT INTO TEMP_DATETIME (f1) VALUES (null);\n+ \n SELECT '' AS ten, f1 AS datetime\n FROM TEMP_DATETIME\n ORDER BY datetime;\n* ../sql/inet.sql\tThu Oct 29 13:13:03 1998\n--- inet.sql\tMon Nov 2 09:36:06 1998\n*************\n* 15,20 **\n--- 15,21 ----\n INSERT INTO INET_TBL (c, i) VALUES ('10', '10.1.2.3/8');\n INSERT INTO INET_TBL (c, i) VALUES ('10', '11.1.2.3/8');\n INSERT INTO INET_TBL (c, i) VALUES ('10', '9.1.2.3/8');\n+ INSERT INTO INET_TBL (c, i) VALUES (null, null);\n \n SELECT '' AS ten, c AS cidr, i AS inet FROM INET_TBL;\n \n* ../sql/lseg.sql\tTue Jul 29 12:22:46 1997\n--- lseg.sql\tMon Nov 2 09:36:46 1998\n*************\n* 10,15 **\n--- 10,16 ----\n INSERT INTO LSEG_TBL VALUES ('10,-10 ,-3,-4');\n INSERT INTO LSEG_TBL VALUES ('[-1e6,2e2,3e5, -4e1]');\n INSERT INTO LSEG_TBL VALUES ('(11,22,33,44)');\n+ INSERT INTO LSEG_TBL VALUES (null);\n \n -- bad values for parser testing\n INSERT INTO LSEG_TBL VALUES ('(3asdf,2 ,3,4r2)');\n* ../sql/name.sql\tMon Apr 27 09:50:03 1998\n--- name.sql\tMon Nov 2 09:37:57 1998\n*************\n* 28,33 ***\n--- 28,35 ----\n \n INSERT INTO NAME_TBL(f1) VALUES ('1234567890ABCDEFGHIJKLMNOPQRSTUVWXYZ');\n \n+ INSERT INTO NAME_TBL(f1) VALUES (null);\n+ \n \n SELECT '' AS seven, NAME_TBL.;\n \n* ../sql/path.sql\tTue Jun 3 10:23:37 1997\n--- path.sql\tMon Nov 2 09:39:36 1998\n***********\n* 22,27 **\n--- 22,29 ----\n \n INSERT INTO PATH_TBL VALUES ('(11,12,13,14)');\n \n+ INSERT INTO PATH_TBL VALUES (null);\n+ \n -- bad values for parser testing\n INSERT INTO PATH_TBL VALUES ('[(,2),(3,4)]');\n \n* ../sql/point.sql\tWed Sep 24 13:55:38 1997\n--- point.sql\tMon Nov 2 09:40:49 1998\n*************\n* 12,17 **\n--- 12,19 ----\n \n INSERT INTO POINT_TBL(f1) VALUES ('(-5.0,-12.0)');\n \n+ INSERT INTO POINT_TBL(f1) VALUES (null);\n+ \n -- bad format points \n INSERT INTO POINT_TBL(f1) VALUES ('asdfasdf');\n \n* ../sql/polygon.sql\tTue Jul 29 12:22:48 1997\n--- polygon.sql\tMon Nov 2 09:41:50 1998\n*************\n* 25,30 **\n--- 25,32 ----\n \n INSERT INTO POLYGON_TBL(f1) VALUES ('(0.0,1.0),(0.0,1.0)');\n \n+ INSERT INTO POLYGON_TBL(f1) VALUES (null);\n+ \n -- bad polygon input strings \n INSERT INTO POLYGON_TBL(f1) VALUES ('0.0');\n \n* ../sql/reltime.sql\tThu May 8 23:26:51 1997\n--- reltime.sql\tMon Nov 2 09:43:27 1998\n*************\n* 12,17 **\n--- 12,19 ----\n \n INSERT INTO RELTIME_TBL (f1) VALUES ('@ 14 seconds ago');\n \n+ INSERT INTO RELTIME_TBL (f1) VALUES (null);\n+ \n \n -- badly formatted reltimes: \n INSERT INTO RELTIME_TBL (f1) VALUES ('badly formatted reltime');\n* ../sql/timespan.sql\tThu May 8 23:26:56 1997\n--- timespan.sql\tMon Nov 2 09:46:19 1998\n*************\n* 10,15 **\n--- 10,16 ----\n INSERT INTO TIMESPAN_TBL (f1) VALUES ('6 years');\n INSERT INTO TIMESPAN_TBL (f1) VALUES ('5 months');\n INSERT INTO TIMESPAN_TBL (f1) VALUES ('5 months 12 hours');\n+ INSERT INTO TIMESPAN_TBL (f1) VALUES (null);\n \n -- badly formatted timespan: \n INSERT INTO TIMESPAN_TBL (f1) VALUES ('badly formatted timespan');\n* ../sql/tinterval.sql\tSat Sep 20 12:33:24 1997\n--- tinterval.sql\tMon Nov 2 09:47:23 1998\n*************\n* 15,20 ***\n--- 15,22 ----\n INSERT INTO TINTERVAL_TBL (f1)\n VALUES ('[\"Feb 15 1990 12:15:03\" \"current\"]');\n \n+ INSERT INTO TINTERVAL_TBL (f1) VALUES (null);\n+ \n \n -- badly formatted tintervals \n INSERT INTO TINTERVAL_TBL (f1)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 2 Nov 1998 10:00:41 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n>> Huh. You have a backend crash fix, and we can't apply it until after\n>> 6.4 is out the door?\n\n> \tD'Arcy is going to release a patch for the fix on the same day as\n> the release, as it is something that only affects a very small aspect of\n> the server, and will affect very few ppl...\n\nOf course, the other side of that argument is that D'Arcy's fix is a\nvery low-risk affair for the people who won't be using the inet types.\nFor the people who will be using them, it seems rather critical.\n\nI think shipping the release without this fix is a mistake.\n\nIt sounds like D'Arcy could use a little advice/help from someone who\nknows about proper handling of null inputs to functions and operators\n(not me...). But if someone like that will step up to the plate, I\ndon't see why we can't squeeze in a solid fix.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 02 Nov 1998 10:19:36 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types " }, { "msg_contents": "On Mon, 2 Nov 1998, Tom Lane wrote:\n\n> The Hermit Hacker <[email protected]> writes:\n> >> Huh. You have a backend crash fix, and we can't apply it until after\n> >> 6.4 is out the door?\n> \n> > \tD'Arcy is going to release a patch for the fix on the same day as\n> > the release, as it is something that only affects a very small aspect of\n> > the server, and will affect very few ppl...\n> \n> Of course, the other side of that argument is that D'Arcy's fix is a\n> very low-risk affair for the people who won't be using the inet types.\n> For the people who will be using them, it seems rather critical.\n\n\tright, which is why we will be releasing a patch at the same time\nas we release v6.4 ...\n\n> I think shipping the release without this fix is a mistake.\n\n\tand I think incorporating it at this late a date in the release\ncycle would also be a mistake. I'd rather a known bug with a patch then\nan unknown bug created by fixing the known bug...\n\n\t...for 99% of the users, they won't be using the INET/CIDR\nstuff...for the 1% that will, they can easily download and install that\none patch to fix the problem...\n\n> It sounds like D'Arcy could use a little advice/help from someone who\n> knows about proper handling of null inputs to functions and operators\n> (not me...). But if someone like that will step up to the plate, I\n> don't see why we can't squeeze in a solid fix.\n\n\tBecause we are less then 2 days away from a release, and we have\nalready over-shoot our release date by 1 month because* of this feature.\nIn the future, as has been discussed by the core, this will not happen\nagain...we should never have held up the release for the INET/CIDR types.\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 2 Nov 1998 11:44:16 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types " }, { "msg_contents": "[email protected] (D'Arcy J.M. Cain) writes:\n> OK, there are more problems. If you apply the following patch to the\n> regression tests you will crash the backend in a number of places.\n\nYipes!\n\nI must withdraw my prior opinion that we should shoehorn in a repair to\nthe INET datatypes for this case. It's clear that we have a wideranging\nproblem that ought to be fixed more globally. But presumably it's\nbeen there for quite a while, and we didn't know it; therefore it's not\ncritical enough to hold up the release.\n\nMy guess is that maybe this should not be fixed in the individual\ndatatypes at all; instead the generic function and operator code should\nbe modified so that if any input value is NULL, then NULL is returned as\nthe result without ever calling the datatype-specific code.\n\nThere might be specific operators for which this is not the right\nbehavior (although none spring to mind immediately). In that case,\nI think the best bet would be to have a per-operator flag, defaulting\nto OFF, which could be turned on for those specific operators that are\nprepared to cope with null inputs.\n\nThoughts?\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 02 Nov 1998 10:50:38 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types " }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n> ... we have\n> already over-shoot our release date by 1 month because of this feature.\n> In the future, as has been discussed by the core, this will not happen\n> again...we should never have held up the release for the INET/CIDR types.\n\nWell, maybe not, but don't beat yourselves up about it. An awful lot of\nsmall but good things have happened in the last month in the way of bug\nfixes and configure improvements. Not to mention docs. If you'd\nreleased 6.4 a month ago, it would've been of substantially lower\nquality than what will go out the door this week. I don't think the\ndelay was such a bad thing.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 02 Nov 1998 11:07:13 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types " }, { "msg_contents": "> Well, maybe not, but don't beat yourselves up about it. An awful lot \n> of small but good things have happened in the last month in the way of \n> bug fixes and configure improvements. Not to mention docs. If you'd\n> released 6.4 a month ago, it would've been of substantially lower\n> quality than what will go out the door this week. I don't think the\n> delay was such a bad thing.\n\nThat's been pointed out :)\n\nBut every release we try to learn something new (or are forced to\nrelearn something old). In this case, imho we delayed focusing the\ndeveloper's group on release-specific issues while waiting for these\nadditions, and could have/should have covered the same territory\nstarting Sept 1 rather than Oct 15. otoh, you did a lot of these fixes\nand improvements over the last month, and only you can guess if you\nwould have been prepared/willing/in the right frame of mind to do the\nsame in September. Also, I got some great docs contributions since Sept\n15 (about when I had planned on freezing the docs), and those things\nwould have had to wait for v6.5, at least for hardcopy versions.\n\nSo, the only issue really is trying to be a bit more decisive about\nwhich features will go into a release, feature-freezing when we had\nplanned to (more or less), and then doing the right thing for testing\nand packaging once we have frozen.\n\nAnd then move on to the next release cycle with clear heads and great\nplans ;)\n\n - Tom\n", "msg_date": "Mon, 02 Nov 1998 16:46:21 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake Tom Lane\n> My guess is that maybe this should not be fixed in the individual\n> datatypes at all; instead the generic function and operator code should\n> be modified so that if any input value is NULL, then NULL is returned as\n> the result without ever calling the datatype-specific code.\n\nCould it be tied to the return type? IOW, functions or operators\nthat return bool return FALSE, text return \"\", etc.\n\n> There might be specific operators for which this is not the right\n> behavior (although none spring to mind immediately). In that case,\n> I think the best bet would be to have a per-operator flag, defaulting\n> to OFF, which could be turned on for those specific operators that are\n> prepared to cope with null inputs.\n\nObviously that will have to wait for 6.5 since it requires an initdb\nto add the field. Do we want to wait that long?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 2 Nov 1998 13:06:38 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "> My guess is that maybe this should not be fixed in the individual\n> datatypes at all; instead the generic function and operator code should\n> be modified so that if any input value is NULL, then NULL is returned as\n> the result without ever calling the datatype-specific code.\n\nAFAICT, the function code returns blank when the input is NULL, regardless\nof the function definition... this came up before when someone tried to\nextend the functions and found that func(NULL) called func, but disregarded\nthe return value...\n\nTaral\n\n", "msg_date": "Mon, 2 Nov 1998 12:31:23 -0600", "msg_from": "\"Taral\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] A small problem with the new inet and cidr types " }, { "msg_contents": "> Thus spake Tom Lane\n> > My guess is that maybe this should not be fixed in the individual\n> > datatypes at all; instead the generic function and operator code should\n> > be modified so that if any input value is NULL, then NULL is returned as\n> > the result without ever calling the datatype-specific code.\n> \n> Could it be tied to the return type? IOW, functions or operators\n> that return bool return FALSE, text return \"\", etc.\n> \n> > There might be specific operators for which this is not the right\n> > behavior (although none spring to mind immediately). In that case,\n> > I think the best bet would be to have a per-operator flag, defaulting\n> > to OFF, which could be turned on for those specific operators that are\n> > prepared to cope with null inputs.\n> \n> Obviously that will have to wait for 6.5 since it requires an initdb\n> to add the field. Do we want to wait that long?\n\nThe only thing I can add here is to look at the other functions, and do\nwhat they do.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Mon, 2 Nov 1998 13:35:43 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "[email protected] (D'Arcy J.M. Cain) writes:\n>> My guess is that maybe this should not be fixed in the individual\n>> datatypes at all; instead the generic function and operator code should\n>> be modified so that if any input value is NULL, then NULL is returned as\n>> the result without ever calling the datatype-specific code.\n\n> Could it be tied to the return type? IOW, functions or operators\n> that return bool return FALSE, text return \"\", etc.\n\nThat strikes me as really dangerous. What you are basically proposing\nis to overload a perfectly good value of each type as meaning \"maybe\nit's really this data value, and maybe it's the result of an operation\non NULL\". We have a good way to represent an unknown/undefined result,\nand that is to return NULL; we shouldn't replace that with a kluge.\n\nCompare for example the IEEE floating point standards, in which there\nare specific \"exceptional values\" (Not a Numbers, or NaNs). NaNs\npropagate through operations --- for example, NaN + 23 yields NaN, not\n0 or any other ordinary data value. The people who defined it that way\nknew what they were doing. Practically every formal computational model\ninvented in the last twenty years has had a similar concept of an\n\"undefined\" or \"bottom\" value, and they all treat it that way.\n\nNow, any Postgres implementation running on an IEEE-float platform will\nautomatically do the right things in the float4 and float8 operators,\nso NaNs in a database should do the right things. (BTW, do the input\nand output conversion operators for floats support a representation for\nNaNs? If not that should be on the to-do list...) But we need a\nsimilar concept for all the other data types. As far as I can see,\nNULL should be it.\n\n(In fact, I'd be surprised if the SQL standard doesn't mandate that\nbehavior. I don't have a copy to look at, however.)\n\nIt sounds like the function/operator support already gets this almost\nright, but it shouldn't call the datatype-specific routine at all if\nit's not going to use the result (IMHO anyway).\n\n>> There might be specific operators for which this is not the right\n>> behavior (although none spring to mind immediately). In that case,\n>> I think the best bet would be to have a per-operator flag, defaulting\n>> to OFF, which could be turned on for those specific operators that are\n>> prepared to cope with null inputs.\n\n> Obviously that will have to wait for 6.5 since it requires an initdb\n> to add the field. Do we want to wait that long?\n\nI haven't yet heard an example case demonstrating that it's necessary\nat all...\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 02 Nov 1998 14:06:06 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types " }, { "msg_contents": "Thus spake Bruce Momjian\n> The only thing I can add here is to look at the other functions, and do\n> what they do.\n\nUh, they crash the backend. :-)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 2 Nov 1998 14:36:04 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake Taral\n> > My guess is that maybe this should not be fixed in the individual\n> > datatypes at all; instead the generic function and operator code should\n> > be modified so that if any input value is NULL, then NULL is returned as\n> > the result without ever calling the datatype-specific code.\n> \n> AFAICT, the function code returns blank when the input is NULL, regardless\n> of the function definition... this came up before when someone tried to\n> extend the functions and found that func(NULL) called func, but disregarded\n> the return value...\n\nWell that sure fits with my observations. Sure seems wrong though. We\nshould either use the return value or don't call the function in the\nfirst place. I vote for the latter even though I have spent the time\nfixing inet. It seems like the proper method.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 2 Nov 1998 14:41:46 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "> Thus spake Bruce Momjian\n> > The only thing I can add here is to look at the other functions, and do\n> > what they do.\n> \n> Uh, they crash the backend. :-)\n\nOh.\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Mon, 2 Nov 1998 14:48:29 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n\n> \tand I think incorporating it at this late a date in the release\n> cycle would also be a mistake.\n\nWhile we're talking of release cycles, I'd like to take the liberty to\npost something that I've been keeping around for a while, which was\nwritten by Dave Burgess (of BSD FAQ fame) some time back. He was\nposting to a NetBSD mailing list, so there is wording in here that's\nspecific to that environment, but there is much good, general truth\nhere. With respect to the CIDR case, note especially his \"get it\nright or get it out\" rule. Translated to PostgreSQL 6.4 terms, we\nget: \"Whoah! This wasn't what we wanted! Decide quick: do we ship\nwhat we have, or do we ditch it and do it right for 6.5?\". We chose\nto do neither, and there's a lesson to be learned from this.\n\nLet me also take this opportunity to submit a plug for \"The Mythical\nMan-Month\" by Frederick P. Brooks. Anyone involved in software\ndevelopment who hasn't yet read it, should. Get the 1995 special\nedition, expanded and edited from the original.\n\nAnyway, here's Dave's writing on release engineering:\n\n\| 1. When the interface to a system changes radically, the major number\n\| should be bumped. The minor number should be reset to 0.\n\| \n\| 2. When more than 50% of the code in a system has been changed, the\n\| major number should be bumped EXCEPT if the minor number is 0. This\n\| takes into account the case where a recent interface change drives a\n\| large code churn.\n\| \n\| 3. There should be a target list of changes for each release. When\n\| those changes are accomplished, the minor number should be bumped unless\n\| the minor number was reset to 0 by 1 or 2 (above). The target list\n\| should also have a target date (six months is a good consensus\n\| timeframe, from what I've heard). \n\| \n\| NOTE: If the list only has one item on it, but a bunch of other\n\| projects have been started, then when that single item is ready to roll,\n\| everyone has to be ready to either get it right or get it out.\n\| \n\| 4. Once the target list is complete, all changes in progress are\n\| brought to closure and the system is Alpha released. The Alpha baseline\n\| is established. This means either backing out the work in progress or\n\| bringing the work to a state where it is working or will not impact\n\| performance.\n\| \n\| 5. The bugs from the Alpha test cycle are documented and analyzed. If\n\| there are no Class A or Class B problems (A being \"causes damage to\n\| equipment or files\" and B being \"causes reproducable system halts and\n\| panics\") the code should be promoted to Beta. If there are, these\n\| problems are addressed. The Alpha is then re-released. All SPR driven \n\| changes are made to the Alpha Baseline code and the -current code.\n\| \n\| 6. Once the Beta has been released, a firm release date is established.\n\| For most of the Govt project I've worked on, that is 28 days out. If no\n\| new Class A or Class B SPRs are received, the code is released on the\n\| release date. If there are Class A or Class B SPRs, the changes are\n\| made. The release date does not slip. For us, a 14 day lead should be\n\| plenty.\n\| \n\| 7. If the release date cannot be met, a new Beta is released and a new\n\| release date is established.\n\| \n\| The advantage of this is that there are a few driving upgrades that\n\| force new releases. As an example, we could have established a short\n\| list of \"2.0 candidate targets.\" which could have included \"bounce\n\| buffer support\" and \"working 100MB/s drivers for all existing network\n\| cards.\"\n\| \n\| The logistics for this are daunting, but are workable. We are already\n\| working using good CM practices, our testing is good, and the source\n\| code baseline could literally happen almost anytime.\n\| \n\| The trick, then, is to find someone willing to do the Release\n\| Engineering. It's a shame we don't know anyone with years of experience\n\| doing this....\n\| \n\| Now for the schedule in real terms:\n\| \n\| Day 0 - The list of existing \"work in progress\" and \"need to fix\"\n\| problems is culled for the next release.\n\| \n\| Day 20 - The list is agreed upon; if anyone disagrees with the\n\| feasability of an item for the releas in 5 months, it is defered.\n\| \n\| 100 days pass for people to work on their projects. This should be\n\| enough to at least get to the point that you know whether you will make\n\| the deadline.\n\| \n\| Day 120 - The release candidate is checked for completion of the list.\n\| The list is reviewed for possible deletions. Work continues.\n\| \n\| Day 150 - The release candidate is checked. This is the \"drop dead\" for\n\| the projects in work. Everything that is working stays, everything else\n\| is either removed or neutered. This is the Alpha release. At this\n\| point, the -current tree is baselined, as is. Work on -current\n\| continues while the Alpha and Beta trees are used strictly for Problem\n\| reports. Yes, this involves making the changes twice; once in the Alpha\n\| tree and once in -current.\n\| \n\| Day 165 - The Beta is released. Changes to the Alpha source tree are\n\| included and the Beta tree is baselined.\n\| \n\| Day 180 - The release is cut and the next version list is started, Day\n\| 0 is today for release +.1.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "02 Nov 1998 21:02:22 +0100", "msg_from": "Tom Ivar Helbekkmo <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake Tom Ivar Helbekkmo\n> While we're talking of release cycles, I'd like to take the liberty to\n> post something that I've been keeping around for a while, which was\n> written by Dave Burgess (of BSD FAQ fame) some time back. He was\n> posting to a NetBSD mailing list, so there is wording in here that's\n> specific to that environment, but there is much good, general truth\n> here. With respect to the CIDR case, note especially his \"get it\n> right or get it out\" rule. Translated to PostgreSQL 6.4 terms, we\n> get: \"Whoah! This wasn't what we wanted! Decide quick: do we ship\n> what we have, or do we ditch it and do it right for 6.5?\". We chose\n> to do neither, and there's a lesson to be learned from this.\n\nPerhaps but let me point out a few things.\n\n - The code wasn't that late. Other things were holding up release\n and the inet/cidr stuff was in by the drop dead date we were given\n at the time.\n\n - We may not have followed the release method you quote but that\n wasn't our release method at the time. Perhaps it should be\n but let's not be blinded by hindsight.\n\n - The latest problem has turned out to be a general problem with\n the system. Rather than being an added bug, having this in now\n just allowed us to find an existing problem sooner than we might\n have otherwise.\n\n - Because of a few late night coding sessions we now have a really\n cool feature and we beat the market. That's something we can use in\n our announcement. Putting this feature in, even if we pushed the\n deadline a little, didn't weaken the product, it strengthened it.\n\nI don't want to sound defensive but people have been sounding negative\nabout having this type in. I just want to make sure that we all see\nthe positive side of this too.\n\n> Let me also take this opportunity to submit a plug for \"The Mythical\n> Man-Month\" by Frederick P. Brooks. Anyone involved in software\n> development who hasn't yet read it, should. Get the 1995 special\n> edition, expanded and edited from the original.\n\nI'll second that.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 2 Nov 1998 16:48:18 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "On Mon, 2 Nov 1998, D'Arcy J.M. Cain wrote:\n\n> Perhaps but let me point out a few things.\n> \n> - The code wasn't that late. Other things were holding up release\n> and the inet/cidr stuff was in by the drop dead date we were given\n> at the time.\n\n\tNo disagreement on that from here...\n\n> - Because of a few late night coding sessions we now have a really\n> cool feature and we beat the market. That's something we can use in\n> our announcement. Putting this feature in, even if we pushed the\n> deadline a little, didn't weaken the product, it strengthened it.\n\n\tNo disagreement here either...to a certain extent. What we shoudl\nhave done was followed our original release schedualed and looked at\ngetting the INET/CIDR stuff into v6.5 instead. But, hindsight is 20/20\nand its in now...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 2 Nov 1998 19:28:24 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": ">\n> Thus spake Taral\n> > > My guess is that maybe this should not be fixed in the individual\n> > > datatypes at all; instead the generic function and operator code should\n> > > be modified so that if any input value is NULL, then NULL is returned as\n> > > the result without ever calling the datatype-specific code.\n> >\n> > AFAICT, the function code returns blank when the input is NULL, regardless\n> > of the function definition... this came up before when someone tried to\n> > extend the functions and found that func(NULL) called func, but disregarded\n> > the return value...\n>\n> Well that sure fits with my observations. Sure seems wrong though. We\n> should either use the return value or don't call the function in the\n> first place. I vote for the latter even though I have spent the time\n> fixing inet. It seems like the proper method.\n\n Not calling a function if one of it's arguments is NULL?\n Isn't NULL a legal value?\n\n I know that the function manager interface is damned stupid\n in the case of NULL's. Some of the interface functions pass\n isNull as in/out value and some do not. And the in value only\n tells if any of the arguments are NULL, not which of them. It\n hit me when building PL/pgSQL and PL/Tcl.\n\n Let's redesign the function call interface and define that\n any function has to handle NULL arguments properly. Yes, I\n know what that means :-).\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, 3 Nov 1998 10:21:06 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Jan Wieck wrote:\n> \n> Not calling a function if one of it's arguments is NULL?\n> Isn't NULL a legal value?\n> \n> I know that the function manager interface is damned stupid\n> in the case of NULL's. Some of the interface functions pass\n> isNull as in/out value and some do not. And the in value only\n> tells if any of the arguments are NULL, not which of them. It\n> hit me when building PL/pgSQL and PL/Tcl.\n> \n> Let's redesign the function call interface and define that\n> any function has to handle NULL arguments properly. Yes, I\n> know what that means :-).\n\nAn easier way would be _not_ to call a function with NULL arguments, \n_unless_ it declares that it can handle them.\n\nIt would probably do the right thing in many (most?) places.\n\nBut we do need to redesign not only the function _call_ interface, but\nmost likely also the function _definition_ mechanisms as well.\n\nOn the whole I would like the functions to be more object-like, \nmeaning that they should have methods (metadata) that could tell \nthings about them. In addition to an method that tells if the \nfunction can take NULL as an argument and in what positions, \nI envision we could also use a method that would tell the max print \nlength and other \"field\" attribute of the function return value given \nthe attributes of function arguments. For example we would be able to \ndetermine that the max length of concatenating varchar(5) and varchar(7) \nis varchar(12) and not a varchar of infinite length as we have to \nassume now.\n\nIs'nt PostgreSQL supposed to be somewhat OO DBMS ? \n\n----------\nHannu\n", "msg_date": "Tue, 03 Nov 1998 13:07:44 +0200", "msg_from": "Hannu Krosing <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake Jan Wieck\n> > Well that sure fits with my observations. Sure seems wrong though. We\n> > should either use the return value or don't call the function in the\n> > first place. I vote for the latter even though I have spent the time\n> > fixing inet. It seems like the proper method.\n> \n> Not calling a function if one of it's arguments is NULL?\n> Isn't NULL a legal value?\n\nSure but what is the reasonable thing to do if we perform a function\non a null? Let's take the inet/cidr functions that started this. If\ni is a null field then these seem reasonable conversions.\n\n host(i) ==> null\n network(i) ==> null\n broadcast(i) ==> null\n\n...etc.\n\nThere may be cases where a function of a null is not null as some people\nhave pointed out but so far no one has come up with a practical example.\n\nI suggested that the actual return value could depend on the return\ntype. The only reason I suggested that was for the case of boolean\nreturns. I see some merit in being able to decide in my select whether\nor not to include rows where one or more operators is null but even\nthere it is probably of marginal utility. If it is absolutely essential\nto display every row then you can always constrain the field to NOT\nNULL and use a marker value for null.\n\n> Let's redesign the function call interface and define that\n> any function has to handle NULL arguments properly. Yes, I\n> know what that means :-).\n\nWell, let's hurry up and decide this so I know whether or not to\nclean out my local patches to inet/cidr. :-)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Tue, 3 Nov 1998 07:24:05 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "D'Arcy J.M. Cain wrote:\n> \n> There may be cases where a function of a null is not null as some people\n> have pointed out but so far no one has come up with a practical example.\n\nisnull(field)\n\nis_any_null(field1,field2,field3)\n\nare_all_nulls(field1,field2,field3)\n\nvalue_or_default(NULL,defaultvalue)\n\n---------\nHannu\n", "msg_date": "Tue, 03 Nov 1998 15:09:12 +0200", "msg_from": "Hannu Krosing <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake Hannu Krosing\n> D'Arcy J.M. Cain wrote:\n> > There may be cases where a function of a null is not null as some people\n> > have pointed out but so far no one has come up with a practical example.\n> \n> isnull(field)\n> \n> is_any_null(field1,field2,field3)\n> \n> are_all_nulls(field1,field2,field3)\n> \n> value_or_default(NULL,defaultvalue)\n\nI meant in the specific type functions. These functions seem like they\ncan easily be handled at a higher level and still never call the type\nfunction code. IOW, if these functions are considered useful, they\nshould be implemented at the function dispatch level.\n\nThat last one seems particularly useful to me and, in fact, could handle\nthe issue of requiring functions to handle nulls all by itself.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Tue, 3 Nov 1998 08:53:16 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "[email protected] (D'Arcy J.M. Cain) writes:\n> Thus spake Hannu Krosing\n>> D'Arcy J.M. Cain wrote:\n>>>> There may be cases where a function of a null is not null as some people\n>>>> have pointed out but so far no one has come up with a practical example.\n>> \n>> isnull(field)\n>> \n>> is_any_null(field1,field2,field3)\n>> \n>> are_all_nulls(field1,field2,field3)\n>> \n>> value_or_default(NULL,defaultvalue)\n\n> I meant in the specific type functions. These functions seem like they\n> can easily be handled at a higher level and still never call the type\n> function code. IOW, if these functions are considered useful, they\n> should be implemented at the function dispatch level.\n\nIn fact they would have to be implemented as generics. If we tried\nto implement them at the type-specific level, then we'd have to assign\nspecific types to the input and output values --- in other words, we'd\nneed a separate implementation for every possible combination of input\ntypes.\n\nThe question at hand is whether any ordinary type-specific operators\nor functions need to be non-strict (ie, capable of returning a non-NULL\noutput given one or more NULL inputs). Before we go to the trouble\nof looking at/fixing every single type-specific operator or function\nroutine in the system, I'd like to see some concrete examples why\njust turning off the spigot at the type-independent function dispatch\nroutine isn't good enough.\n\nNOTE: aggregates definitely must be non-strict; I don't want SUM()\nto come back with a NULL if I happen to include a NULL in its inputs.\nBut there are a lot fewer aggregate functions to look at than plain\nold scalar functions.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 03 Nov 1998 10:35:22 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types " }, { "msg_contents": "> Not calling a function if one of it's arguments is NULL?\n> Isn't NULL a legal value?\n> \n> I know that the function manager interface is damned stupid\n> in the case of NULL's. Some of the interface functions pass\n> isNull as in/out value and some do not. And the in value only\n> tells if any of the arguments are NULL, not which of them. It\n> hit me when building PL/pgSQL and PL/Tcl.\n> \n> Let's redesign the function call interface and define that\n> any function has to handle NULL arguments properly. Yes, I\n> know what that means :-).\n> \n\nAdded to TODO list:\n\n\t* redesign the function call interface to handle NULLs better\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 3 Nov 1998 11:38:03 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "> * redesign the function call interface to handle NULLs better\n\nI was planning on looking at this for v6.5, at least in the context of\ntrying to solve the problem of returning NULL for pass-by-value types.\n\nWe should have some discussion of pass-by-value vs. pass-by-reference\nand whether it is worth having both mechanisms for common data types. As\nit is, functions which return int2 or int4 cannot return NULL because\nthere is no way to represent that with these types. I was thinking of\nimplementing true smallint/integer pass-by-reference types to clean this\nup.\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 02:50:46 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": ">\n> > * redesign the function call interface to handle NULLs better\n>\n> I was planning on looking at this for v6.5, at least in the context of\n> trying to solve the problem of returning NULL for pass-by-value types.\n>\n> We should have some discussion of pass-by-value vs. pass-by-reference\n> and whether it is worth having both mechanisms for common data types. As\n> it is, functions which return int2 or int4 cannot return NULL because\n> there is no way to represent that with these types. I was thinking of\n> implementing true smallint/integer pass-by-reference types to clean this\n> up.\n\n Actually they can - but only if they take exactly one\n argument. fmgr_c() calls those ones with an additional\n isNull bool pointer.\n\n The mess is, that we have different entry points to call a\n function, some of them pass information about NULL and some\n not. What I had in mind was to finally have one single entry\n point that handles NULL for any argument and for the return\n value.\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": "Wed, 4 Nov 1998 11:41:24 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "I am replying to something posted by D'Arcy on November 2nd, so I will\nquote the whole thing:\n\n> Thus spake Tom Ivar Helbekkmo\n> > While we're talking of release cycles, I'd like to take the liberty to\n> > post something that I've been keeping around for a while, which was\n> > written by Dave Burgess (of *BSD FAQ fame) some time back. He was\n> > posting to a NetBSD mailing list, so there is wording in here that's\n> > specific to that environment, but there is much good, general truth\n> > here. With respect to the CIDR case, note especially his \"get it\n> > right or get it out\" rule. Translated to PostgreSQL 6.4 terms, we\n> > get: \"Whoah! This wasn't what we wanted! Decide quick: do we ship\n> > what we have, or do we ditch it and do it right for 6.5?\". We chose\n> > to do neither, and there's a lesson to be learned from this.\n> \n> Perhaps but let me point out a few things.\n> \n> - The code wasn't that late. Other things were holding up release\n> and the inet/cidr stuff was in by the drop dead date we were given\n> at the time.\n> \n> - We may not have followed the release method you quote but that\n> wasn't our release method at the time. Perhaps it should be\n> but let's not be blinded by hindsight.\n> \n> - The latest problem has turned out to be a general problem with\n> the system. Rather than being an added bug, having this in now\n> just allowed us to find an existing problem sooner than we might\n> have otherwise.\n> \n> - Because of a few late night coding sessions we now have a really\n> cool feature and we beat the market. That's something we can use in\n> our announcement. Putting this feature in, even if we pushed the\n> deadline a little, didn't weaken the product, it strengthened it.\n> \n> I don't want to sound defensive but people have been sounding negative\n> about having this type in. I just want to make sure that we all see\n> the positive side of this too.\n> \n> > Let me also take this opportunity to submit a plug for \"The Mythical\n> > Man-Month\" by Frederick P. Brooks. Anyone involved in software\n> > development who hasn't yet read it, should. Get the 1995 special\n> > edition, expanded and edited from the original.\n> \n> I'll second that.\n\nOK. I want to take responsibility for the one-month delay of 6.4.\n\nIf people remember, I posted the mega-patch mid-August, which added\nmulti-key indexes to the system tables, improved the heap API with\nproper locking, and improved system cache usage.\n\nThis patch generated several problems, some of them bugs, some of them\nareas I missed in adding multi-key system indexes. We couldn't go live\nwith beta on September 1 because we had major problems doing initdb on\nsome platforms. We kind of started the beta, but without all platforms\nparticipating, it was pretty useless, and we started to let features in\nduring the month while we struggled to get the mega-patch problems\nresolved.\n\nWe didn't start serious beta until October 1, and D'Arcy got caught with\nPaul Vixie defining/adding the IP type. D'Arcy was just left holding\nthe bag when people started to worry about the delay. It was not his\nfault.\n\nIt was me, putting a 18k line patch in two weeks before beta. Of\ncourse, I started earlier than mid-August, but it takes me a while to\ngenerate 18k lines of diff. :-)\n\nJust wanted to clear that up.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 6 Jan 1999 21:50:04 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Thus spake Bruce Momjian\n> > I don't want to sound defensive but people have been sounding negative\n> > about having this type in. I just want to make sure that we all see\n> > the positive side of this too.\n> > \n> OK. I want to take responsibility for the one-month delay of 6.4.\n\nNo hair shirts, please. :-)\n\n> We didn't start serious beta until October 1, and D'Arcy got caught with\n> Paul Vixie defining/adding the IP type. D'Arcy was just left holding\n> the bag when people started to worry about the delay. It was not his\n> fault.\n\nJust to be clear, I was defending the type, not myself. We can't take\nany minor disagreements personally. If we do we start splintering the\ngroup and that's no good for the project as a whole.\n\nThe bottom line is that there was a little bit of a problem getting\n6.4 out the door but we wound up with a superior product. We also\nhave some ideas for improving the release methodology for next time.\nGiven what we have I don't believe anyone needs to beat themself up\nover some minor problem that we had getting there.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Thu, 7 Jan 1999 07:08:22 -0500 (EST)", "msg_from": "\"D'Arcy\" \"J.M.\" Cain <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n\n> Just wanted to clear that up.\n\nJust for the record, I was _not_ attacking D'Arcy or anyone else! If\nanyone, I'm the one who's to blame for the delays and misunderstanding\nover the IP (CIDR) data type, since what I implemented (based on code\nfrom Paul Vixie), was not actually in accordance with prior consensus\non pgsql-hackers.\n\nMy posting, the one that D'Arcy followed up and you then commented on\nnow, was meant to suggest that when it finally became clear that I'd\nmisunderstood, and implemented something different from what was\nwanted, it was really too close to the release date to try to fix it.\nThe best handling would, in my opinion, have been to quickly decide\nwhether to use it anyway or postpone it until the next release.\n\nAs it turned out, D'Arcy did have time to modify and extend the code\ninto an acceptable solution, so everything actually worked out well.\n\n-tih\n-- \nPopularity is the hallmark of mediocrity. --Niles Crane, \"Frasier\"\n", "msg_date": "07 Jan 1999 13:37:11 +0100", "msg_from": "Tom Ivar Helbekkmo <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr types" } ]
[ { "msg_contents": "Hi,\n\n I'd like to know if there's an operator that converts from float8 to\ntext. It should be because PGSQL can print a float8 number.\n\n\n--------------------------------------------------------------------------------\nSebestyďż˝n Zoltďż˝n AKA Memphisto\t\tIt all seems so stupid,\n\t\t\t\t\tit makes me want to give up. \[email protected]\t\t\tBut why should I give up,\n when it all seems so stupid? \n\nMAKE INSTALL NOT WAR\t\t\tAnd please avoid Necrosoft Widows\n\n\n", "msg_date": "Mon, 2 Nov 1998 16:40:34 +0100 (CET)", "msg_from": "Memphisto <[email protected]>", "msg_from_op": true, "msg_subject": "float8 to text converter" }, { "msg_contents": "Hi Memphisto,\n\nIf you are using v6.4 you can use CAST as:\n\nprova=> select cast('1.7976931348623e+308'::float8 as text);\n?column?\n--------------------\n1.7976931348623e+308\n(1 row)\n\nor:\n\nprova=> select cast(f as text) from a;\ntext\n----------------------\n4714-11--2147483624 BC <--- \n4714-11--2147483624 BC <---\n(2 rows)\n\nOoops! that seems like a bug.\n\nprova=> select f from a;\nf\n--------------------\n1.7976931348623e+308\n1.7976931348623e+308\n(2 rows)\n\nJose'\n\nMemphisto wrote:\n> \n> Hi,\n> \n> I'd like to know if there's an operator that converts from float8 to\n> text. It should be because PGSQL can print a float8 number.\n>\n", "msg_date": "Tue, 03 Nov 1998 10:30:51 +0100", "msg_from": "\"Jose' Soares\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] float8 to text converter" }, { "msg_contents": "On Tue, 3 Nov 1998, Jose' Soares wrote:\n\n> Hi Memphisto,\n> \n> If you are using v6.4 you can use CAST as:\n6.4 is in beta, isn't it(currently i use 6.3.2)? Anyway thanks.\n\n\n--------------------------------------------------------------------------------\nSebestyďż˝n Zoltďż˝n AKA Memphisto\t\tIt all seems so stupid,\n\t\t\t\t\tit makes me want to give up. \[email protected]\t\t\tBut why should I give up,\n when it all seems so stupid? \n\nMAKE INSTALL NOT WAR\t\t\tAnd please avoid Necrosoft Widows\n\n\n", "msg_date": "Tue, 3 Nov 1998 12:35:59 +0100 (CET)", "msg_from": "Memphisto <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [GENERAL] float8 to text converter" } ]
[ { "msg_contents": "1. in psql , \\h create table does not reflect the proper syntax for\nconstraint and check (there is the old syntax there) and there is\nnothing there about primary key !!!\n\n\n\n2. Create a table :\n\ncreate table test(id int4, how char, name text, primary\nkey(\"id\",\"how\"));\n\nAt \\d test, the fields are displayed corectly.\n\nwhen \\d test_pkey , the \"how\" field is displayed as length -4.\n\n---------------------\nConstantin Teodorescu\nFLEX Consulting Braila, ROMANIA\n", "msg_date": "Mon, 02 Nov 1998 18:45:15 +0000", "msg_from": "Constantin Teodorescu <[email protected]>", "msg_from_op": true, "msg_subject": "Small bugs in PostgreSQL 6.4 beta5" } ]
[ { "msg_contents": "I am releasing today the latest PgAccess 0.91 for PostgreSQL 6.4\n\nIt has been checked against the latest beta 5.\n\nSome enhancements into the table design window, allowing multiple\ninheritance, constraints, check and primary keys to be defined.\n\nAlso, a lot of inside changes, due to changes made in libpgtcl.so,\nreturn status codes.\nSome minor bug fixes, a new SQL window to watch commands.\n\nPlease download and include in the final PostgreSQL 6.4 the tar.gz file\nfrom\n\nhttp://www.flex.ro/pgaccess or\n\nftp://ftp.flex.ro/pub/pgaccess\n\n\n\nAll the best,\n\nConstantin Teodorescu\nFLEX Consulting Braila, ROMANIA\n", "msg_date": "Mon, 02 Nov 1998 20:07:23 +0000", "msg_from": "Constantin Teodorescu <[email protected]>", "msg_from_op": true, "msg_subject": "Latest PgAccess 0.91 for PostgreSQL 6.4 has been released" } ]
[ { "msg_contents": "We plan to release PostgreSQL, version 6.4, within the next week.\n\nAn announcement will be made as soon as it is ready. You can see the\nchanges for this release on our web site at the bottom of the TODO list.\n\nWe hope people will be able to upgrade within two weeks after its\nrelease, so we fix any problems and move on to the 6.5 release.\n\n6.4 includes a pg_upgrade utility, that will allow upgrading without a\ndump/reload for 6.3.* users.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Mon, 2 Nov 1998 19:22:18 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Upcoming 6.4 release issues" }, { "msg_contents": "On Mon, 2 Nov 1998, Bruce Momjian wrote:\n\n\tI'm sorry if this question is answered somewhere, but does 6.4\nrequire C++? My main database boy doesn't have a C++ compiler, and I\ndon't plan on getting one for it any time in the near future.\n\n\tI tried a beta yesterday, and it compiled on 0/2 of the machines I\ntried (NetBSD and IRIX). I'm really hoping this new version might fix\nsome of the instability issues I had with 6.3.2 in production.\n\n// We plan to release PostgreSQL, version 6.4, within the next week.\n// \n// An announcement will be made as soon as it is ready. You can see the\n// changes for this release on our web site at the bottom of the TODO list.\n// \n// We hope people will be able to upgrade within two weeks after its\n// release, so we fix any problems and move on to the 6.5 release.\n// \n// 6.4 includes a pg_upgrade utility, that will allow upgrading without a\n// dump/reload for 6.3.* users.\n// \n// -- \n// Bruce Momjian \| http://www.op.net/~candle\n// [email protected] \| (610) 853-3000\n// + If your life is a hard drive, \| 830 Blythe Avenue\n// + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n// \n// \n\n--\nSA, beyond.com The world is watching America,\npub 1024/3CAE01D5 1994/11/03 Dustin Sallings <[email protected]>\n\| Key fingerprint = 87 02 57 08 02 D0 DA D6 C8 0F 3E 65 51 98 D8 BE \nL________________________________________ and America is watching TV. __\n\n", "msg_date": "Mon, 2 Nov 1998 18:49:47 -0800 (PST)", "msg_from": "dustin sallings <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] Upcoming 6.4 release issues" }, { "msg_contents": "On Mon, 2 Nov 1998, dustin sallings wrote:\n\n> On Mon, 2 Nov 1998, Bruce Momjian wrote:\n> \n> \tI'm sorry if this question is answered somewhere, but does 6.4\n> require C++? My main database boy doesn't have a C++ compiler, and I\n> don't plan on getting one for it any time in the near future.\n\n\tThis does not require C++ to compile, nor do I suspect it ever\nwill...I couldn't imagine the effort to go through to actually convert it\nover to C++ :)\n\n> \tI tried a beta yesterday, and it compiled on 0/2 of the machines I\n> tried (NetBSD and IRIX). I'm really hoping this new version might fix\n> some of the instability issues I had with 6.3.2 in production.\n\n\tPlease report any compile errors to pgsql-hackers ... it has been\ntested on NetBSD, but I don't believe Irix has been tested...\n\n\n> \n> // We plan to release PostgreSQL, version 6.4, within the next week.\n> // \n> // An announcement will be made as soon as it is ready. You can see the\n> // changes for this release on our web site at the bottom of the TODO list.\n> // \n> // We hope people will be able to upgrade within two weeks after its\n> // release, so we fix any problems and move on to the 6.5 release.\n> // \n> // 6.4 includes a pg_upgrade utility, that will allow upgrading without a\n> // dump/reload for 6.3.* users.\n> // \n> // -- \n> // Bruce Momjian \| http://www.op.net/~candle\n> // [email protected] \| (610) 853-3000\n> // + If your life is a hard drive, \| 830 Blythe Avenue\n> // + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n> // \n> // \n> \n> --\n> SA, beyond.com The world is watching America,\n> pub 1024/3CAE01D5 1994/11/03 Dustin Sallings <[email protected]>\n> \| Key fingerprint = 87 02 57 08 02 D0 DA D6 C8 0F 3E 65 51 98 D8 BE \n> L________________________________________ and America is watching TV. __\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Mon, 2 Nov 1998 23:27:56 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] Upcoming 6.4 release issues" }, { "msg_contents": "On Mon, 2 Nov 1998, The Hermit Hacker wrote:\n\n\t(experience with 6.4 beta 5) I'm not on this list, so if you have\nany questions, be sure to send directly to me.\n\n// \tThis does not require C++ to compile, nor do I suspect it ever\n// will...I couldn't imagine the effort to go through to actually\n// convert it over to C++ :)\n\n\tThis is what I get when I try to configure on IRIX:\n\nbleu:/tmp/pgsql/src 112> ./configure \ncreating cache ./config.cache\nchecking host system type... mips-sgi-irix6.2\nchecking echo setting...\nchecking setting template to... irix5\nchecking setting USE_LOCALE... disabled\nchecking setting CYR_RECODE... disabled\nchecking setting MULTIBYTE... disabled\nchecking setting DEF_PGPORT... 5432\nchecking setting USE_TCL... disabled\nchecking setting USE_PERL... disabled\nchecking setting USE_ODBC... disabled\nchecking setting ASSERT CHECKING... disabled\nchecking for gcc... no\nchecking for cc... cc\nchecking whether the C compiler (cc ) works... yes\nchecking whether the C compiler (cc ) is a cross-compiler... no\nchecking whether we are using GNU C... no\nchecking how to run the C preprocessor... cc -E\n- setting CPPFLAGS= \n- setting LDFLAGS= \nchecking for c++... no\nchecking for g++... no\nchecking for gcc... no\nchecking for CC... no\nchecking for cxx... no\nchecking for cc++... no\nchecking whether the C++ compiler (gcc ) works... no\nconfigure: error: installation or configuration problem: C++ compiler\ncannot create executables.\n\n\tIf'n I take that out of configure.in, I get a lot further. The\nproblem, though, is that it's treating the lack of a C++ compiler as a\nfatal condition. I just took out the check and set HAVE_Cplusplus=false\nand got to src/pl/plpgsql/src/ which needed -I. and I got it built.\n\n// \tPlease report any compile errors to pgsql-hackers ... it has\n// been tested on NetBSD, but I don't believe Irix has been tested...\n\n\tI don't remember what my NetBSD problem was yet, but I'll try to\nplay with it later.\n\n--\nSA, beyond.com The world is watching America,\npub 1024/3CAE01D5 1994/11/03 Dustin Sallings <[email protected]>\n\| Key fingerprint = 87 02 57 08 02 D0 DA D6 C8 0F 3E 65 51 98 D8 BE \nL________________________________________ and America is watching TV. __\n\n", "msg_date": "Mon, 2 Nov 1998 22:22:44 -0800 (PST)", "msg_from": "dustin sallings <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [GENERAL] Upcoming 6.4 release issues" } ]
[ { "msg_contents": "prova=> select * from a;\nint_2\| int_4\| int_8\n-----+----------+-------------------\n32767\|2147483647\|9223372036620802086\n32767\|2147483647\|9223372036620802086\n(2 rows)\n\nprova=> select avg(int_2), avg(int_4), avg(int_8) from a;\navg\|avg\| avg\n---+---+----------\n -1\| -1\|-233973722\n(1 row)\n\n\nJose'\n", "msg_date": "Tue, 03 Nov 1998 09:16:27 +0100", "msg_from": "\"Jose' Soares\" <[email protected]>", "msg_from_op": true, "msg_subject": "bug on aggregate function AVG() " }, { "msg_contents": "Jose' Soares wrote:\n> \n> prova=> select * from a;\n> int_2\| int_4\| int_8\n> -----+----------+-------------------\n> 32767\|2147483647\|9223372036620802086\n> 32767\|2147483647\|9223372036620802086\n> (2 rows)\n> \n> prova=> select avg(int_2), avg(int_4), avg(int_8) from a;\n> avg\|avg\| avg\n> ---+---+----------\n> -1\| -1\|-233973722\n> (1 row)\n> \n> Jose'\n\nThe same on SUM():\n\nprova=> select sum(int_2), sum(int_4), sum(int_8) from a;\nsum\|sum\| sum\n---+---+----------\n -2\| -2\|-467947444\n(1 row)\n\nbut it works as follow:\n\nprova=> select avg(int_21.0), avg(int_41.0), avg(int_81.0) from a;\n avg\| avg\|avg\n-----+----------+-------------------\n32767\|2147483647\|9.2233720366208e+18\n(1 row)\n\nprova=> select sum(int_21.0), sum(int_41.0), sum(int_81.0) from a;\n sum\| sum\|sum\n-----+----------+--------------------\n65534\|4294967294\|1.84467440732416e+19\n(1 row)\n\nJose'\n", "msg_date": "Tue, 03 Nov 1998 09:21:05 +0100", "msg_from": "\"Jose' Soares\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: bug on aggregate function AVG()" }, { "msg_contents": "> > prova=> select avg(int_2), avg(int_4), avg(int_8) from a;\n> The same on SUM():\n\nSure. For some reason, on most platforms integers are allowed to\noverflow in Postgres. Of course, both SUM() and AVG() take a running\nsum, and once they overflow you are hosed...\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 06:44:55 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: bug on aggregate function AVG()" }, { "msg_contents": "Hi Tom,\n\nmercoledï¿½, 4 novembre 98, you wrote:\n\n>> > prova=> select avg(int_2), avg(int_4), avg(int_8) from a;\n>> The same on SUM():\n\nTGL> Sure. For some reason, on most platforms integers are allowed to\nTGL> overflow in Postgres. Of course, both SUM() and AVG() take a running\nTGL> sum, and once they overflow you are hosed...\n ^^^^^\nI don't know what does the word \"hosed\" mean Tom, I hope you don't want to tell me\nthere's no solution for this problem.\nI see that AVG() and SUM() uses an accumulator not enough big to hold\nthe result of calculation, but the point is: should we consider this\nthing a \"terrible\" bug or an acceptable feature ?\nWhat about to convert every accumulator to float8 ?\n\nselect intero4 from a;\n intero4\n----------\n2147483647\n2147483647\n2147483647\n(3 rows)\n\nselect sum(inter04),sum(intero41.0) from a;\n sum\| sum\n----------+----------\n2147483645\|6442450941\n(1 row)\n\nselect avg(intero4),avg(intero41.0) from a;\n avg\| avg\n---------+----------\n715827881\|2147483647\n(1 row)\n\nAnyway I think we need to work a little bit on aggregates:\n\nMIN() and MAX() doesn't accept a string as parameter.\nSUM() and AVG() gives a wrong result because it goes on overflow.\nand none of them allows the clause DISTINCT.\n\nWhat do you think about ? ;)\n\n Jose'\n\n\n", "msg_date": "Wed, 4 Nov 1998 16:00:39 +0100", "msg_from": "Sferacarta Software <[email protected]>", "msg_from_op": false, "msg_subject": "Re[2]: [HACKERS] Re: bug on aggregate function AVG()" }, { "msg_contents": "> I don't know what does the word \"hosed\" mean Tom, I hope you don't \n> want to tell me there's no solution for this problem.\n\nAs you guessed, \"hosed\" isn't good ;)\n\n> I see that AVG() and SUM() uses an accumulator not enough big to hold\n> the result of calculation, but the point is: should we consider this\n> thing a \"terrible\" bug or an acceptable feature ?\n> What about to convert every accumulator to float8 ?\n\nimho we can't do that because we lose the exact qualities of integers.\nIf you accumulate in float8, and if you take a sum over a very large\ntable, you might start ignoring values. That is, if you have accumulated\n15 or 16 digits worth of number, and then try adding 1 as the next\nnumber, the result will be the same as the input. With integers that is\nnever the case, but we have to deal with overflows better.\n\nI would think we should start signalling overflows rather than silently\noverflowing, but I'm not sure what that entails.\n\n> Anyway I think we need to work a little bit on aggregates:\n> MIN() and MAX() doesn't accept a string as parameter.\n\nYes, at the moment only numeric quantities are supported.\n\n> SUM() and AVG() gives a wrong result because it goes on overflow.\n> and none of them allows the clause DISTINCT.\n\nYes, SELECT SUM(DISTINCT i) FROM t; is not yet supported. That's a\nproject for v6.5.\n\nbtw, I'm also planning on working on your \"NULL problem\" you mentioned\nearlier...\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 15:16:28 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: bug on aggregate function AVG()" }, { "msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n>> I see that AVG() and SUM() uses an accumulator not enough big to hold\n>> the result of calculation, but the point is: should we consider this\n>> thing a \"terrible\" bug or an acceptable feature ?\n>> What about to convert every accumulator to float8 ?\n\n> imho we can't do that because we lose the exact qualities of integers.\n> If you accumulate in float8, and if you take a sum over a very large\n> table, you might start ignoring values.\n\nI think that SUM() on an int column ought to produce an exact result.\nAVG() is a different story --- I think you could make a good case that\nit ought to produce a float result even when the input is integers,\nsince the exact right answer would typically not be integral anyway.\n(A programmer who wants the average rounded to integer should have to\nwrite something like ROUND(AVG(x)), I think.)\n\nOne way you could postpone the overflow problem for SUM() is to\naccumulate the running sum in a \"long\", or even better \"long long\" where\navailable, even if the input datatype is a smaller flavor of int.\nYou might still find that the end result overflows, but if the incoming\nvalues are not all the same sign then this might avoid an unnecessary\nintermediate overflow.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 04 Nov 1998 16:32:32 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: bug on aggregate function AVG() " }, { "msg_contents": "TL> \"Thomas G. Lockhart\" <[email protected]> writes:\n>>> I see that AVG() and SUM() uses an accumulator not enough big to hold\n>>> the result of calculation, but the point is: should we consider this\n>>> thing a \"terrible\" bug or an acceptable feature ?\n>>> What about to convert every accumulator to float8 ?\n\n>> imho we can't do that because we lose the exact qualities of integers.\n>> If you accumulate in float8, and if you take a sum over a very large\n>> table, you might start ignoring values.\n\nTL> I think that SUM() on an int column ought to produce an exact result.\nTL> AVG() is a different story --- I think you could make a good case that\nTL> it ought to produce a float result even when the input is integers,\nTL> since the exact right answer would typically not be integral anyway.\nTL> (A programmer who wants the average rounded to integer should have to\nTL> write something like ROUND(AVG(x)), I think.)\n\nTL> One way you could postpone the overflow problem for SUM() is to\nTL> accumulate the running sum in a \"long\", or even better \"long long\" where\nTL> available, even if the input datatype is a smaller flavor of int.\nTL> You might still find that the end result overflows, but if the incoming\nTL> values are not all the same sign then this might avoid an unnecessary\nTL> intermediate overflow.\nTL> regards, tom lane\n\nI see there are four new built in functions:\nint42pl,int42div, int84pl and int84div.\nI created four new aggregate functions SUM(int2), SUM(int4), AVG(int2) and\nAVG(int4) using these new functions, in this way the accumulator for int2\nis int4 and the accumulator for int4 is int8, this reduce the probabilities\nof overflow cases.\nRemains just the overflow problem for int8.\n\nIs there any reason for not use these functions on SUM() and AVG() on\nofficial release ?\n------------------------------------------------\ndrop aggregate sum2 int2;\nDROP\ncreate aggregate sum2(\n sfunc1 = int42pl,\n basetype = int2,\n stype1 = int4,\n initcond1 = '0'\n);\nCREATE\ndrop aggregate avg2 int2;\nDROP\ncreate aggregate avg2(\n sfunc1 = int42pl,\n basetype = int2,\n stype1 = int4,\n initcond1 = '0',\n sfunc2 = int2inc,\n stype2 = int2,\n finalfunc = int42div,\n initcond2 = '0'\n);\nCREATE\ndrop aggregate sum4 int4;\nDROP\ncreate aggregate sum4(\n sfunc1 = int84pl,\n basetype = int4,\n stype1 = int8,\n initcond1 = '0'\n);\nCREATE\n\ndrop aggregate avg4 int4;\nDROP\ncreate aggregate avg4(\n sfunc1 = int84pl,\n basetype = int4,\n stype1 = int8,\n initcond1 = '0',\n sfunc2 = int4inc,\n stype2 = int4,\n finalfunc = int84div,\n initcond2 = '0'\n);\nCREATE\ndrop table b;\nDROP\ncreate table b(\n i2 int2,\n i4 int4,\n i8 int8\n);\nCREATE\ninsert into b values (32767,2147483647,9223372036620802086);\nNOTICE: Integer input '9223372036620802086' is out of range; promoted to float\n ^^^^^^^^^^^^^^^^^^^^\n what's happen here? seems a test for an int4\n not an int8.\nINSERT 246255 1\ninsert into b values (32767,2147483647,2147483647);\nINSERT 246256 1\nselect * from b;\n i2\| i4\| i8\n-----+----------+-------------------\n32767\|2147483647\|9223372036620802048\n32767\|2147483647\| 2147483647\n(2 rows)\n\nselect sum2(i2) as new_sum2,\n sum(i2) as old_sum2,\n sum4(i4) as new_sum4,\n sum(i4) as old_sum4,\n sum(i8) as only_sum8\nfrom b;\nnew_sum2\|old_sum2\| new_sum4\|old_sum4\| only_sum8\n--------+--------+----------+--------+--------------------\n 65534\| -2\|4294967294\| -2\|-9223372034941265921\n(1 row)\n\nselect avg2(i2) as new_avg2,\n avg(i2) as old_avg2,\n avg4(i4) as new_avg4,\n avg(i4) as old_avg4,\n avg(i8) as only_sum8\nfrom b;\nnew_avg2\|old_avg2\| new_avg4\|old_avg4\| only_sum8\n--------+--------+----------+--------+--------------------\n 32767\| -1\|2147483647\| -1\|-4611686017470632960\n(1 row)\n\nIs there any reason for not use these functions on SUM() and AVG() on\nofficial release ?\n\n - Jose' -\n\n\n", "msg_date": "Thu, 5 Nov 1998 15:23:12 +0100", "msg_from": "Sferacarta Software <[email protected]>", "msg_from_op": false, "msg_subject": "Re[2]: [HACKERS] Re: bug on aggregate function AVG()" }, { "msg_contents": "> Is there any reason for not use these functions on SUM() and AVG() on\n> official release ?\n\nIt sounds like a good idea. The only hesitation I have at the moment is\nthat not all platforms have int8 support, and I'm not certain which\nthese are. Also, accumulating int4 into int8 is probably pretty slow\nsince on 32-bit machines the \"long long\" is usually done in a s/w\nlibrary, not in machine code.\n\nfloat8 might be a better choice for accumulating AVG(), but I'm worried\nabout incorrect results with large tables (> 1M entries) which have\npathological distributions of numbers (e.g. 1M entries with MAXINT and\n1M entries with zero). int4 gives ~9.2 decimal places, float8 gives ~15\ndecimal places, so there is only about ~6 decimal places of headroom.\n\nOf course, why am I worried? That is much better than what we have\ncurrently. And someone reported that at least one commercial system\n(Sybase?) returns float8 for avg() (and sum()?) as I recall.\n\nSo, your suggestion is that for AVG() at least we return something other\nthan the input type; how about returning float8 for any input type?\nDon't know if SUM() could/should behave similarly...\n\n - Tom\n", "msg_date": "Thu, 05 Nov 1998 16:00:46 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: bug on aggregate function AVG()" }, { "msg_contents": "> MIN() and MAX() doesn't accept a string as parameter.\n\nI've added capabilities to do min() and max() on strings. To help with\nthis I've put the new implicit type coersion techniques into the\naggregate function handling. So, for example, the same routine which\nhandles min(text) will also handle min(char()) and min(varchar()).\n\nThese changes are already in the main cvs tree, but are not in the\nv6.4.x branch since they require a small number of catalog changes to\nimplement.\n\nWill post the patches on the patches list in case anyone needs them\nbefore v6.5 comes out...\n\n - Tom\n", "msg_date": "Tue, 08 Dec 1998 06:40:27 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: bug on aggregate function AVG()" } ]
[ { "msg_contents": "Small bugs have been corrected.\n\nProbably theese days, until the final 6.4 release, I will make the final\nchanges.\n\nAs a rule of thumb, before 6.4 final release, please download\npgaccess-0.91.tar.gz package and include it.\n\nI WILL NOT CHANGE THE NUMBER OF RELEASE, it will remain 0.91, but\nchanges will be made inside the tar.gz\n\nAll the best,\nConstantin Teodorescu\nFLEX Consulting Braila\n", "msg_date": "Tue, 03 Nov 1998 09:12:48 +0000", "msg_from": "Constantin Teodorescu <[email protected]>", "msg_from_op": true, "msg_subject": "PgAccess latest version" } ]
[ { "msg_contents": "Let's say I have a table test like this:\n\ni\|c\n-+-\n2\|A\n0\|\n1\|T\n(3 rows)\n\nand I want to use an update to do the following: set the attr c of the\nrecord with number 0 to the value of attr c in the record with number 1.\n\nIn Oracle I'd be able to write:\n\nupdate test set c = (select c from test where i = 1) where i = 0;\n\nIs it correct that we do not allow this? Pardon me if this is already on\ntodo, but I just stumbled across this problem and I'm not sure I can work\naround this by using the from clause (which btw Oracle doesn't even have).\nSince we do have subselects this shouldn't be too difficult.\n\nMichael\n-- \nDr. Michael Meskes \| Th.-Heuss-Str. 61, D-41812 Erkelenz \| Go SF49ers!\nSenior-Consultant \| business: [email protected] \| Go Rhein Fire!\nMummert+Partner \| private: [email protected] \| Use Debian\nUnternehmensberatung AG \| [email protected] \| GNU/Linux!\n", "msg_date": "Tue, 3 Nov 1998 10:27:19 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "update and select" }, { "msg_contents": "> In Oracle I'd be able to write:\n> update test set c = (select c from test where i = 1) where i = 0;\n> Is it correct that we do not allow this?\n\nThat is correct, and it is on the ToDo list as something like \"allow\nsubselects in target expressions\" (though I'm not finding it when I\nlook).\n\nHowever, as you suspect you can rephrase it:\n\ntgl=> update x set c = j.c from x as j where j.i = 1 and x.i = 0;\nUPDATE 1\ntgl=> select * from x;\ni\|c\n-+-\n1\|T\n2\|A\n0\|T\n(3 rows)\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 07:25:24 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] update and select" }, { "msg_contents": "On Wed, Nov 04, 1998 at 07:25:24AM +0000, Thomas G. Lockhart wrote:\n> That is correct, and it is on the ToDo list as something like \"allow\n> subselects in target expressions\" (though I'm not finding it when I\n> look).\n\nWould be nice to have, yes.\n\n> However, as you suspect you can rephrase it:\n> \n> tgl=> update x set c = j.c from x as j where j.i = 1 and x.i = 0;\n\nTrue. But unfortunately that won't help me as I was writing an SQL script to\nbe run on Oracle later on. :-)\n\nMichael\n-- \nDr. Michael Meskes \| Th.-Heuss-Str. 61, D-41812 Erkelenz \| Go SF49ers!\nSenior-Consultant \| business: [email protected] \| Go Rhein Fire!\nMummert+Partner \| private: [email protected] \| Use Debian\nUnternehmensberatung AG \| [email protected] \| GNU/Linux!\n", "msg_date": "Wed, 4 Nov 1998 12:27:34 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] update and select" } ]
[ { "msg_contents": "Try setting the environment variable `CXX' to `\"gcc -O3\"' (If you\nare using `gcc'). For example `CXX=\"gcc -O3\" ./configure'. If you use\nthis you don't nead to have libg++ installed!\n\n-Jason.\n\n\n\n-----Original Message-----\nFrom: dustin sallings [mailto:[email protected]]\nSent: Tuesday, November 03, 1998 1:23 AM\nTo: The Hermit Hacker\nCc: Bruce Momjian; PostgreSQL-general; [email protected]\nSubject: Re: [GENERAL] Upcoming 6.4 release issues\n\n\nOn Mon, 2 Nov 1998, The Hermit Hacker wrote:\n\n\t(experience with 6.4 beta 5) I'm not on this list, so if you\nhave\nany questions, be sure to send directly to me.\n\n// \tThis does not require C++ to compile, nor do I suspect it ever\n// will...I couldn't imagine the effort to go through to actually\n// convert it over to C++ :)\n\n\tThis is what I get when I try to configure on IRIX:\n\nbleu:/tmp/pgsql/src 112> ./configure \ncreating cache ./config.cache\nchecking host system type... mips-sgi-irix6.2\nchecking echo setting...\nchecking setting template to... irix5\nchecking setting USE_LOCALE... disabled\nchecking setting CYR_RECODE... disabled\nchecking setting MULTIBYTE... disabled\nchecking setting DEF_PGPORT... 5432\nchecking setting USE_TCL... disabled\nchecking setting USE_PERL... disabled\nchecking setting USE_ODBC... disabled\nchecking setting ASSERT CHECKING... disabled\nchecking for gcc... no\nchecking for cc... cc\nchecking whether the C compiler (cc ) works... yes\nchecking whether the C compiler (cc ) is a cross-compiler... no\nchecking whether we are using GNU C... no\nchecking how to run the C preprocessor... cc -E\n- setting CPPFLAGS= \n- setting LDFLAGS= \nchecking for c++... no\nchecking for g++... no\nchecking for gcc... no\nchecking for CC... no\nchecking for cxx... no\nchecking for cc++... no\nchecking whether the C++ compiler (gcc ) works... no\nconfigure: error: installation or configuration problem: C++ compiler\ncannot create executables.\n\n\tIf'n I take that out of configure.in, I get a lot further. The\nproblem, though, is that it's treating the lack of a C++ compiler as a\nfatal condition. I just took out the check and set HAVE_Cplusplus=false\nand got to src/pl/plpgsql/src/ which needed -I. and I got it built.\n\n// \tPlease report any compile errors to pgsql-hackers ... it has\n// been tested on NetBSD, but I don't believe Irix has been tested...\n\n\tI don't remember what my NetBSD problem was yet, but I'll try to\nplay with it later.\n\n--\nSA, beyond.com The world is watching America,\npub 1024/3CAE01D5 1994/11/03 Dustin Sallings <[email protected]>\n\| Key fingerprint = 87 02 57 08 02 D0 DA D6 C8 0F 3E 65 51 98 D8 BE\n\nL________________________________________ and America is watching TV. __\n\n", "msg_date": "Tue, 3 Nov 1998 08:06:47 -0500 ", "msg_from": "Jason <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [GENERAL] Upcoming 6.4 release issues" } ]
[ { "msg_contents": "Thus spake Jan Wieck\n> D'Arcy J.M. Cain wrote:\n> > There may be cases where a function of a null is not null as some people\n> > have pointed out but so far no one has come up with a practical example.\n> \n> CREATE FUNCTION cnt_t1_b(text) RETURNS int4 AS\n> 'SELECT count() FROM t1 WHERE b = $1'\n> LANGUAGE 'sql';\n> \n> BTW: It does not work with NULL argument currently, but IMHO\n> it should count the number of rows where b is NULL.\n\nWell, this is the first example that I have seen that wouldn't work\nunless a function with a null argument actually called the function but\nyou do realize that this wouldn't work anyway, right? The following\nis a parse error.\n\n SELECT count() FROM t1 WHERE b = null;\n\nMind you, I think that's a weakness but I don't know what the issues\nare with respect to the code or the standard.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Tue, 3 Nov 1998 09:06:38 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr typesg" }, { "msg_contents": "[email protected] (D'Arcy J.M. Cain) writes:\n> you do realize that this wouldn't work anyway, right? The following\n> is a parse error.\n\n> SELECT count() FROM t1 WHERE b = null;\n\n> Mind you, I think that's a weakness but I don't know what the issues\n> are with respect to the code or the standard.\n\nI believe the accepted spelling of that query is\n\n SELECT count() FROM t1 WHERE b IS NULL;\n\n(or IS NOT NULL). I don't know either what the SQL standard has to say\nabout the issue --- does it expect \"= NULL\" to be a synonym for \"IS NULL\"?\n\nThe CREATE FUNCTION example does seem to illustrate that it'd be nice\nif \"=\" and \"!=\" worked on NULL values. I'd still object to trying\nto define an order that includes NULL, so \"3 < NULL\" should return NULL,\nbut I can see the reasonableness of defining \"3 != NULL\" as TRUE.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 03 Nov 1998 10:46:14 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr typesg " }, { "msg_contents": "Thus spake Tom Lane\n> [email protected] (D'Arcy J.M. Cain) writes:\n> > you do realize that this wouldn't work anyway, right? The following\n> > is a parse error.\n> \n> > SELECT count() FROM t1 WHERE b = null;\n> \n> I believe the accepted spelling of that query is\n> \n> SELECT count() FROM t1 WHERE b IS NULL;\n\nWell, yes. That's my point. The problem is to specify that syntax if\nthe test is against null and the previous if not. Using PL is one way\nbut it would be nice to have a pure sql way to do it too.\n\n> The CREATE FUNCTION example does seem to illustrate that it'd be nice\n> if \"=\" and \"!=\" worked on NULL values. I'd still object to trying\n> to define an order that includes NULL, so \"3 < NULL\" should return NULL,\n> but I can see the reasonableness of defining \"3 != NULL\" as TRUE.\n\nActually I see it as FALSE. That's what I was suggesting earlier. All\ncomparisons to null should be false no matter what the sense of the\ntest. That way you can always decide in the select statement whether\nyou want a particular comparison to null to be included or not. For\nexample, say I have a table of IP numbers and some are null. If I\nneed to find all IPs that are less than some other IP then I can do\n\n SELECT * FROM t WHERE i1 < i2;\n\nBut let's say that in another case I needed the same test except I\nwanted to include those rows where one or the other was null. Then\nI do this.\n\n SELECT * FROM t WHERE NOT (i1 >= i2);\n\nSee, the \"i1 < i2\" test is nominally the same as the \"NOT (i1 >= i2)\"\none but if operators consistently returned FALSE when given nulls\nthen you can use one or the other depending on what output you needed.\n\nJust a thought.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Tue, 3 Nov 1998 11:57:36 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr typesg" }, { "msg_contents": "[email protected] (D'Arcy J.M. Cain) writes:\n>> but I can see the reasonableness of defining \"3 != NULL\" as TRUE.\n\n> Actually I see it as FALSE. That's what I was suggesting earlier. All\n> comparisons to null should be false no matter what the sense of the\n> test.\n\nHmm. That yields extremely unintuitive results for = and !=. That is,\n\n\tSELECT * FROM t WHERE b = NULL;\n\nwill never return any rows, even if there are some where b is null;\nand\n\n\tSELECT * FROM t WHERE b != NULL;\n\nwill never return any rows, even if there are some where b isn't null.\n\nIf this is the definition then you cannot select rows with null entries\nusing the same syntax as for selecting rows with particular non-null\nvalues, which is what I thought the point of the CREATE FUNCTION example\nwas.\n\n> That way you can always decide in the select statement whether\n> you want a particular comparison to null to be included or not.\n> [D'Arcy proposes that these ops need not give the same result:\n> SELECT * FROM t WHERE i1 < i2;\n> SELECT * FROM t WHERE NOT (i1 >= i2);\n\nUgh. I think it'd be a lot more intuitive to write something like\n\n\tSELECT * FROM t WHERE i1 < 33 OR i1 IS NULL;\n\nBut getting this to work without introducing unintended consequences\nmight be pretty tricky too. If \"NULL < 33\" returns NULL, as I'd prefer,\nthen OR has to be non-strict, and in fact NULL OR 't' has to give 't'.\nThat looks pretty reasonable at first glance, but there are probably\nother examples where it does something undesirable.\n\nMaybe, the boolean-combining operators (AND, OR, NOT) can safely be\nmade non-strict (treating NULL as FALSE), but I'm wary of that.\n\nWe probably ought to go re-read the IEEE float math specs. What I think\nyou are getting at is almost the same as their distinction between\n\"NaN-aware\" and \"non-NaN-aware\" comparison operators, but I've forgotten\nthe details of how those work. (And I have to leave in a minute, so I\ncan't look them up right now...)\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 03 Nov 1998 19:26:16 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Comparisons on NULLs (was Re: A small problem...)" }, { "msg_contents": "\nOn 04-Nov-98 Tom Lane wrote:\n> [email protected] (D'Arcy J.M. Cain) writes:\n>>> but I can see the reasonableness of defining \"3 != NULL\" as TRUE.\n> \n>> Actually I see it as FALSE. That's what I was suggesting earlier. All\n>> comparisons to null should be false no matter what the sense of the\n>> test.\n> \n> Hmm. That yields extremely unintuitive results for = and !=. That is,\n> \n> SELECT * FROM t WHERE b = NULL;\n> \n> will never return any rows, even if there are some where b is null;\n> and\n> \n> SELECT * FROM t WHERE b != NULL;\n> \n> will never return any rows, even if there are some where b isn't null.\n> \n> If this is the definition then you cannot select rows with null entries\n> using the same syntax as for selecting rows with particular non-null\n> values, which is what I thought the point of the CREATE FUNCTION example\n> was.\n> \n>> That way you can always decide in the select statement whether\n>> you want a particular comparison to null to be included or not.\n>> [D'Arcy proposes that these ops need not give the same result:\n>> SELECT * FROM t WHERE i1 < i2;\n>> SELECT * FROM t WHERE NOT (i1 >= i2);\n> \n> Ugh. I think it'd be a lot more intuitive to write something like\n> \n> SELECT * FROM t WHERE i1 < 33 OR i1 IS NULL;\n> \n> But getting this to work without introducing unintended consequences\n> might be pretty tricky too. If \"NULL < 33\" returns NULL, as I'd prefer,\n> then OR has to be non-strict, and in fact NULL OR 't' has to give 't'.\n> That looks pretty reasonable at first glance, but there are probably\n> other examples where it does something undesirable.\n> \n> Maybe, the boolean-combining operators (AND, OR, NOT) can safely be\n> made non-strict (treating NULL as FALSE), but I'm wary of that.\n> \n> We probably ought to go re-read the IEEE float math specs. What I think\n> you are getting at is almost the same as their distinction between\n> \"NaN-aware\" and \"non-NaN-aware\" comparison operators, but I've forgotten\n> the details of how those work. (And I have to leave in a minute, so I\n> can't look them up right now...)\n\nI looked at this earlier, but it was me that had to leave then I forgot\nall about it till now. Now it's confusing.\n\nLooking at this (and please let's not get into IS vs = yet):\n\nSELECT * FROM t WHERE b = NULL;\n\nI first looked at this from within a C program. Consider the input coming\nfrom a form and constructing the select statement from it's submission\nvalues:\n\n sprintf(buf,\"SELECT * FROM t WHERE a = %d AND b = '%s'\",abc,xyz);\n\nIf I understand what you're saying above, if xyz is NULL and b is NULL\nthen it doesn't matter what a is 'cuze it'll never return any results.\n\nI'll shut up now in case I'm misintrepreting this..\n\nVince.\n-- \n==========================================================================\nVince Vielhaber -- KA8CSH email: [email protected] flame-mail: /dev/null\n # include <std/disclaimers.h> TEAM-OS2\n Online Searchable Campground Listings http://www.camping-usa.com\n \"There is no outfit less entitled to lecture me about bloat\n than the federal government\" -- Tony Snow\n==========================================================================\n\n\n", "msg_date": "Tue, 03 Nov 1998 19:58:55 -0500 (EST)", "msg_from": "Vince Vielhaber <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] Comparisons on NULLs (was Re: A small problem...)" }, { "msg_contents": "Thus spake Tom Lane\n> >> but I can see the reasonableness of defining \"3 != NULL\" as TRUE.\n> \n> > Actually I see it as FALSE. That's what I was suggesting earlier. All\n> > comparisons to null should be false no matter what the sense of the\n> > test.\n> \n> Hmm. That yields extremely unintuitive results for = and !=. That is,\n> \n> \tSELECT * FROM t WHERE b = NULL;\n> \n> will never return any rows, even if there are some where b is null;\n\nHmmm. That would be a problem. Of course, we could treat the null\nvalue at the higher level too. I guess that's why we have the \"IS\nNULL\" syntax in the first place. It is different than comparing the\nactual values.\n\nMarc, how long can we hold 6.4 while we work this all out?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Tue, 3 Nov 1998 23:07:49 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: Comparisons on NULLs (was Re: A small problem...)" }, { "msg_contents": "On Tue, 3 Nov 1998, D'Arcy J.M. Cain wrote:\n\n> Thus spake Tom Lane\n> > >> but I can see the reasonableness of defining \"3 != NULL\" as TRUE.\n> > \n> > > Actually I see it as FALSE. That's what I was suggesting earlier. All\n> > > comparisons to null should be false no matter what the sense of the\n> > > test.\n> > \n> > Hmm. That yields extremely unintuitive results for = and !=. That is,\n> > \n> > \tSELECT * FROM t WHERE b = NULL;\n> > \n> > will never return any rows, even if there are some where b is null;\n> \n> Hmmm. That would be a problem. Of course, we could treat the null\n> value at the higher level too. I guess that's why we have the \"IS\n> NULL\" syntax in the first place. It is different than comparing the\n> actual values.\n> \n> Marc, how long can we hold 6.4 while we work this all out?\n\n\tHow long can we hold what? Is this a new bug that didn't exist\nin previous version of PostgreSQL? \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Wed, 4 Nov 1998 01:22:16 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Comparisons on NULLs (was Re: A small problem...)" }, { "msg_contents": "Thus spake The Hermit Hacker\n> > Marc, how long can we hold 6.4 while we work this all out?\n> \n> \tHow long can we hold what? Is this a new bug that didn't exist\n> in previous version of PostgreSQL? \n\nJeez Marc, you must be working too hard. I'll have to remember to\nsprinkle those smileys a little better next time. :-)\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Wed, 4 Nov 1998 00:59:14 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: Comparisons on NULLs (was Re: A small problem...)" }, { "msg_contents": "On Wed, 4 Nov 1998, D'Arcy J.M. Cain wrote:\n\n> Thus spake The Hermit Hacker\n> > > Marc, how long can we hold 6.4 while we work this all out?\n> > \n> > \tHow long can we hold what? Is this a new bug that didn't exist\n> > in previous version of PostgreSQL? \n> \n> Jeez Marc, you must be working too hard. I'll have to remember to\n> sprinkle those smileys a little better next time. :-)\n\n\troll eyes groan wipe brow\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Wed, 4 Nov 1998 02:21:15 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Comparisons on NULLs (was Re: A small problem...)" }, { "msg_contents": "D'Arcy J.M. Cain <darcy@{druid\|vex}.net> writes:\n> Thus spake Tom Lane\n> > >> but I can see the reasonableness of defining \"3 != NULL\" as TRUE.\n> > \n> > > Actually I see it as FALSE. That's what I was suggesting earlier. All\n> > > comparisons to null should be false no matter what the sense of the\n> > > test.\n> > \n> > Hmm. That yields extremely unintuitive results for = and !=. That is,\n> > \n> > \tSELECT * FROM t WHERE b = NULL;\n> > \n> > will never return any rows, even if there are some where b is null;\n> \n> Hmmm. That would be a problem. Of course, we could treat the null\n> value at the higher level too. I guess that's why we have the \"IS\n> NULL\" syntax in the first place. It is different than comparing the\n> actual values.\n\nNot sure how serious this discussion is, so if I have wandered into the\nmiddle of a joke, just kick me ...\n\nThat said,\n\nSELECT * FROM t WHERE b = NULL;\n\n_should not_ return any rows. NULL is not \"=\" to anything, not even another\nNULL. NULL is also not \">\", or \"<\", or \"!=\" to anything either.\n\nSo, \"NULL = NULL\" is false as is \"NULL != NULL\".\n\nThis indeed is why we have \"IS NULL\" and \"IS NOT NULL\".\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\"Samba is a huge win ... ; it enables open-source techies to stealth\n their Linux boxes so they look like Microsoft servers that somehow\n miraculously fail to suck.\" -- Eric Raymond\n", "msg_date": "Tue, 3 Nov 1998 22:40:32 -0800 (PST)", "msg_from": "[email protected] (David Gould)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: Comparisons on NULLs (was Re: A small problem...)" }, { "msg_contents": "> I believe the accepted spelling of that query is\n> SELECT count() FROM t1 WHERE b IS NULL;\n> (or IS NOT NULL). I don't know either what the SQL standard has to \n> say about the issue --- does it expect \"= NULL\" to be a synonym for \n> \"IS NULL\"?\n\nafaik SQL92 does not define/allow \"= NULL\". However, our friends at M$\nuse this syntax in queries generated by M$Access, presumably pulling\ntheir usual BS in altering standards to reduce interoperability.\n\nSomeone very recently brought this up, and the \"= NULL\" will be\nsynonymous with \"IS NULL\" in the next release (and patches are likely to\nbe available beforehand).\n\n> The CREATE FUNCTION example does seem to illustrate that it'd be nice\n> if \"=\" and \"!=\" worked on NULL values. I'd still object to trying\n> to define an order that includes NULL, so \"3 < NULL\" should return \n> NULL, but I can see the reasonableness of defining \"3 != NULL\" as \n> TRUE.\n\nSorry, got to go with D'Arcy on this one. C.J. Date in his recent book\n\"A Guide to the SQL Standard\" points out the inconsistancies within\nSQL92 regarding tri-value booleans and nulls. However, it is the case\nthat one can mostly assume that any comparison involving a NULL will\nreturn false. Null usually means \"don't know\", not \"isn't\", but\nexpressions are unfortunately required to resolve to true or false.\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 07:01:18 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A small problem with the new inet and cidr typesg" }, { "msg_contents": "Thus spake David Gould\n> D'Arcy J.M. Cain <darcy@{druid\|vex}.net> writes:\n> > Hmmm. That would be a problem. Of course, we could treat the null\n> > value at the higher level too. I guess that's why we have the \"IS\n> > NULL\" syntax in the first place. It is different than comparing the\n> > actual values.\n> \n> SELECT FROM t WHERE b = NULL;\n> \n> _should not_ return any rows. NULL is not \"=\" to anything, not even another\n> NULL. NULL is also not \">\", or \"<\", or \"!=\" to anything either.\n> \n> So, \"NULL = NULL\" is false as is \"NULL != NULL\".\n> \n> This indeed is why we have \"IS NULL\" and \"IS NOT NULL\".\n\nBut no one really has a use for a statement that can never return a row.\nIf we need that we always have \"-- select 1;\" (1/2 :-)) so why not let\n\"= NULL\" be a synonym for \"IS NULL\" and \"NOT = NULL\" be a synonym for \"IS\nNOT NULL?\"\n\nWell, other than the fact that M$ does it? (Other 1/2 of that :-))\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Wed, 4 Nov 1998 08:00:12 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Re: Comparisons on NULLs (was Re: A small problem...)" } ]
[ { "msg_contents": "REGARDING Problem with indicies...\n\nI'm having trouble getting indicies to be used when performing various\n\"covered\" queries.\n\nFor example see the following table, index and query.\n\nCREATE TABLE JDSFTRATT\n(\nATT_CODE char(3),\nDS_ID int,\nFTR_ID int,\nATT_VALUE float,\nATT_TEXT varchar(128) )\n;\n\ncreate index jdsftratt_b_ds_id_ftr_id\non jdsftratt\nusing btree (ds_id int4_ops, ftr_id int4_ops, att_code char4_ops, att_value\nfloat8_ops);\n\nselect ftr_id from jdsftratt;\n\nAs per the explain command, a \"Seq Scan\" (see explain command output below)\nis being performed by postgres for the above query. Also, why would\nsequential scan be performed by postgres given a select count(*) on and\nindexed table?\n\nexplain select ftr_id from jdsftratt;\n\nSeq Scan on jdsftratt (cost=90889.78 size=2134660 width=4)\n\nI have verified that the index exists via querying the system tables. I'm\nusing postgres 6.3.2 on Sun Ultra 10 running Solaris. Any help would be\nappreciated on this. \n\nAndy\n\n", "msg_date": "3 Nov 1998 18:20:49 U", "msg_from": "\"Andy Farrell\" <[email protected]>", "msg_from_op": true, "msg_subject": "Problem with indicies..." } ]
[ { "msg_contents": "\nTomorrow afternoon, at ~5pm ADT, I will be tag'ng and releasing v6.4 to\nthe world muhahahaha\n\nAny last minute \"no, don't do it\" naysayers out there? :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Tue, 3 Nov 1998 17:29:10 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Warning!! " }, { "msg_contents": "Thus spake The Hermit Hacker\n> Tomorrow afternoon, at ~5pm ADT, I will be tag'ng and releasing v6.4 to\n> the world muhahahaha\n> \n> Any last minute \"no, don't do it\" naysayers out there? :)\n\nI have this idea for a new feature...\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Tue, 3 Nov 1998 16:49:07 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "On Tue, 3 Nov 1998, D'Arcy J.M. Cain wrote:\n\n> Thus spake The Hermit Hacker\n> > Tomorrow afternoon, at ~5pm ADT, I will be tag'ng and releasing v6.4 to\n> > the world muhahahaha\n> > \n> > Any last minute \"no, don't do it\" naysayers out there? :)\n> \n> I have this idea for a new feature...\n\n\tShhhh...bruce might hear you rofl\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Tue, 3 Nov 1998 18:37:50 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "> On Tue, 3 Nov 1998, D'Arcy J.M. Cain wrote:\n> \n> > Thus spake The Hermit Hacker\n> > > Tomorrow afternoon, at ~5pm ADT, I will be tag'ng and releasing v6.4 to\n> > > the world muhahahaha\n> > > \n> > > Any last minute \"no, don't do it\" naysayers out there? :)\n> > \n> > I have this idea for a new feature...\n> \n> \tShhhh...bruce might hear you rofl\n\nNew feature! Tell me more...\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Tue, 3 Nov 1998 21:40:02 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "\n\nI'm very new to this list, and I have to first say that 6.4beta5 fixes\nevery problem I had encountered with 6.3.2 (A small list..) A bug I didn't\nencounter isn't a bug for me...\n\nBUT...\n\nJust now I saw and posted this...\n\ntestdb=> select cash_mul_flt8('$1', '123.74');\ncash_mul_flt8\n-------------\n$123.73 \n(1 row)\n\nIf the pence is 0-4 it rounds down... surely that's a rounding bug because\nit should be looking at the next significant figure?\n\nIt's really small, and insignificant but I guess I should tell somone...\n\n\nOn Tue, 3 Nov 1998, Bruce Momjian wrote:\n\n> > On Tue, 3 Nov 1998, D'Arcy J.M. Cain wrote:\n> > \n> > > Thus spake The Hermit Hacker\n> > > > Tomorrow afternoon, at ~5pm ADT, I will be tag'ng and releasing v6.4 to\n> > > > the world muhahahaha\n> > > > \n> > > > Any last minute \"no, don't do it\" naysayers out there? :)\n> > > \n> > > I have this idea for a new feature...\n> > \n> > \tShhhh...bruce might hear you rofl\n> \n> New feature! Tell me more...\n> \n> -- \n> Bruce Momjian \| http://www.op.net/~candle\n> [email protected] \| (610) 853-3000\n> + If your life is a hard drive, \| 830 Blythe Avenue\n> + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n> \n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n", "msg_date": "Wed, 4 Nov 1998 03:07:50 +0000 (GMT)", "msg_from": "A James Lewis <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "> > > Any last minute \"no, don't do it\" naysayers out there? :)\n> > I have this idea for a new feature...\n\ngrin Thanks D'Arcy. Glad to see we didn't pound your sense of humor\nout of ya. Will try harder next time ;)\n\nI've been able to reproduce Jose's report of problems when casting\nfloats to text. The problem is that there is no explicit float to text\nconversion function, but for other reasons I had decided to make float\nand datetime equivalent (to allow weird date arithmetic). When Postgres\nstarts fishing around for conversion functions, it matches them up:\n\ntgl=> select text(f) from a where f = 1;\ntext\n----------------------------\nSat Jan 01 00:00:01 2000 GMT\n(1 row)\n\nYuck. Anyway, I don't think it is a show stopper, and can be fixed with\na patch and in the v6.4.1 release. Probably should not make datetime and\nfloat binary equivalent types as they are now.\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 07:34:52 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "> I'm very new to this list, and I have to first say that 6.4beta5 fixes\n> every problem I had encountered with 6.3.2 (A small list..)\n> Just now I saw and posted this...\n> If the pence is 0-4 it rounds down... surely that's a rounding bug \n> because it should be looking at the next significant figure?\n\ntgl=> select cash_mul_flt8('$123.77', '1');\ncash_mul_flt8\n-------------\n$123.77\n(1 row)\n\ntgl=> select cash_mul_flt8('$1', '123.77');\ncash_mul_flt8\n-------------\n$123.76\n(1 row)\n\nThat's annoying; it's non-symmetric too. The money type is stored as an\ninteger, and the float type is an IEEE double; looks like we have an LSB\nrounding problem. Not sure what to do about it other than remove the\nfunction, which isn't desirable I'm sure...\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 07:45:41 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "\nDon't remove it... After all, I'm using this workaround in my software \nfor now and it seems OK....\n\ntestdb=> select cash_mul_flt8('$1', '123.77');\ncash_mul_flt8\n-------------\n$123.76 \n(1 row)\n\ntestdb=> select cash_mul_flt8('$1', '123.77000001');\ncash_mul_flt8\n-------------\n$123.77 \n(1 row)\n\nIt happens in float4 too....\n\n\nOn Wed, 4 Nov 1998, Thomas G. Lockhart wrote:\n\n> > I'm very new to this list, and I have to first say that 6.4beta5 fixes\n> > every problem I had encountered with 6.3.2 (A small list..)\n> > Just now I saw and posted this...\n> > If the pence is 0-4 it rounds down... surely that's a rounding bug \n> > because it should be looking at the next significant figure?\n> \n> tgl=> select cash_mul_flt8('$123.77', '1');\n> cash_mul_flt8\n> -------------\n> $123.77\n> (1 row)\n> \n> tgl=> select cash_mul_flt8('$1', '123.77');\n> cash_mul_flt8\n> -------------\n> $123.76\n> (1 row)\n> \n> That's annoying; it's non-symmetric too. The money type is stored as an\n> integer, and the float type is an IEEE double; looks like we have an LSB\n> rounding problem. Not sure what to do about it other than remove the\n> function, which isn't desirable I'm sure...\n> \n> - Tom\n> \n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n", "msg_date": "Wed, 4 Nov 1998 13:37:10 +0000 (GMT)", "msg_from": "A James Lewis <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "> \n> \n> I'm very new to this list, and I have to first say that 6.4beta5 fixes\n> every problem I had encountered with 6.3.2 (A small list..) A bug I didn't\n> encounter isn't a bug for me...\n> \n> BUT...\n> \n> Just now I saw and posted this...\n> \n> testdb=> select cash_mul_flt8('$1', '123.74');\n> cash_mul_flt8\n> -------------\n> $123.73 \n> (1 row)\n> \n> If the pence is 0-4 it rounds down... surely that's a rounding bug because\n> it should be looking at the next significant figure?\n> \n> It's really small, and insignificant but I guess I should tell somone...\n> \n\nWorks here on BSD/OS 4.0. Looks like an OS bug, perhaps.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 08:57:57 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "> > I'm very new to this list, and I have to first say that 6.4beta5 fixes\n> > every problem I had encountered with 6.3.2 (A small list..)\n> > Just now I saw and posted this...\n> > If the pence is 0-4 it rounds down... surely that's a rounding bug \n> > because it should be looking at the next significant figure?\n> \n> tgl=> select cash_mul_flt8('$123.77', '1');\n> cash_mul_flt8\n> -------------\n> $123.77\n> (1 row)\n> \n> tgl=> select cash_mul_flt8('$1', '123.77');\n> cash_mul_flt8\n> -------------\n> $123.76\n> (1 row)\n> \n> That's annoying; it's non-symmetric too. The money type is stored as an\n> integer, and the float type is an IEEE double; looks like we have an LSB\n> rounding problem. Not sure what to do about it other than remove the\n> function, which isn't desirable I'm sure...\n\nThis is really bad. I can't imagine a 5-digit double is loosing\nrounding on a double. There must be some other problem, but I can't\nreproduce it here on BSD/OS.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 09:02:54 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "\nIt only happens here for some figues, not all.... mostly xx.x[1234]\n\nOn Wed, 4 Nov 1998, Bruce Momjian wrote:\n\n> > > I'm very new to this list, and I have to first say that 6.4beta5 fixes\n> > > every problem I had encountered with 6.3.2 (A small list..)\n> > > Just now I saw and posted this...\n> > > If the pence is 0-4 it rounds down... surely that's a rounding bug \n> > > because it should be looking at the next significant figure?\n> > \n> > tgl=> select cash_mul_flt8('$123.77', '1');\n> > cash_mul_flt8\n> > -------------\n> > $123.77\n> > (1 row)\n> > \n> > tgl=> select cash_mul_flt8('$1', '123.77');\n> > cash_mul_flt8\n> > -------------\n> > $123.76\n> > (1 row)\n> > \n> > That's annoying; it's non-symmetric too. The money type is stored as an\n> > integer, and the float type is an IEEE double; looks like we have an LSB\n> > rounding problem. Not sure what to do about it other than remove the\n> > function, which isn't desirable I'm sure...\n> \n> This is really bad. I can't imagine a 5-digit double is loosing\n> rounding on a double. There must be some other problem, but I can't\n> reproduce it here on BSD/OS.\n> \n> -- \n> Bruce Momjian \| http://www.op.net/~candle\n> [email protected] \| (610) 853-3000\n> + If your life is a hard drive, \| 830 Blythe Avenue\n> + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n> \n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n", "msg_date": "Wed, 4 Nov 1998 14:57:52 +0000 (GMT)", "msg_from": "A James Lewis <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "\nI hope it is! I dont have the skill set to track it down but mabe a\nlibrary?\n\nOn Wed, 4 Nov 1998, Bruce Momjian wrote:\n\n> > \n> > \n> > I'm very new to this list, and I have to first say that 6.4beta5 fixes\n> > every problem I had encountered with 6.3.2 (A small list..) A bug I didn't\n> > encounter isn't a bug for me...\n> > \n> > BUT...\n> > \n> > Just now I saw and posted this...\n> > \n> > testdb=> select cash_mul_flt8('$1', '123.74');\n> > cash_mul_flt8\n> > -------------\n> > $123.73 \n> > (1 row)\n> > \n> > If the pence is 0-4 it rounds down... surely that's a rounding bug because\n> > it should be looking at the next significant figure?\n> > \n> > It's really small, and insignificant but I guess I should tell somone...\n> > \n> \n> Works here on BSD/OS 4.0. Looks like an OS bug, perhaps.\n> \n> -- \n> Bruce Momjian \| http://www.op.net/~candle\n> [email protected] \| (610) 853-3000\n> + If your life is a hard drive, \| 830 Blythe Avenue\n> + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n> \n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n", "msg_date": "Wed, 4 Nov 1998 15:04:02 +0000 (GMT)", "msg_from": "A James Lewis <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "> It only happens here for some figures, not all.... mostly xx.x[1234]\n\nWhat system and compiler are you using?\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 15:19:03 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "\"Thomas G. Lockhart\" <[email protected]> writes:\n> tgl=> select cash_mul_flt8('$1', '123.77');\n> cash_mul_flt8\n> -------------\n> $123.76\n> (1 row)\n> That's annoying; it's non-symmetric too.\n\nAnd hardware-dependent, evidently, because I get the right thing on HP:\n\nplay=> select cash_mul_flt8('$1', '123.77');\ncash_mul_flt8\n-------------\n$123.77\n(1 row)\n\n\n> The money type is stored as an\n> integer, and the float type is an IEEE double; looks like we have an LSB\n> rounding problem. Not sure what to do about it other than remove the\n> function, which isn't desirable I'm sure...\n\nWhat's needed is explicit rounding. Instead of letting the compiler do\nthe rounding during its implicit float-to-int conversion (which\ngenerally truncates towards 0 or towards -infinity), do\n\n* cash.c~\tWed Nov 4 10:11:57 1998\n--- cash.c\tWed Nov 4 10:13:06 1998\n***********\n* 17,22 **\n--- 17,23 ----\n #include <limits.h>\n #include <ctype.h>\n #include <locale.h>\n+ #include <math.h>\n \n #include \"postgres.h\"\n #include \"miscadmin.h\"\n***********\n* 419,425 ***\n \tif (!PointerIsValid(result = palloc(sizeof(Cash))))\n \t\telog(ERROR, \"Memory allocation failed, can't multiply cash\", NULL);\n \n! \tresult = ((f) (c));\n \n \treturn result;\n }\t/ cash_mul_flt8() /\n--- 420,426 ----\n \tif (!PointerIsValid(result = palloc(sizeof(Cash))))\n \t\telog(ERROR, \"Memory allocation failed, can't multiply cash\", NULL);\n \n! \tresult = floor((f) (c) + 0.5);\n \n \treturn result;\n }\t/ cash_mul_flt8() /\n\n\n(Alternatively you could do it with rint(), although I distrust rint()\nbecause it's context-dependent...)\n\nThe other cash functions also need to be looked at for rounding\nsloppiness, so I see no point in committing this fix by its lonesome.\n\nLessee, who's the author of cash.c ... looks like it's D'Arcy ...\nD'Arcy, you wanna put this on your to-do list?\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 04 Nov 1998 10:23:42 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!! " }, { "msg_contents": "> On Wed, 4 Nov 1998, Bruce Momjian wrote:\n> \n> > > > I'm very new to this list, and I have to first say that 6.4beta5 fixes\n> > > > every problem I had encountered with 6.3.2 (A small list..)\n> > > > Just now I saw and posted this...\n> > > > If the pence is 0-4 it rounds down... surely that's a rounding bug \n> > > > because it should be looking at the next significant figure?\n> > > \n> > > tgl=> select cash_mul_flt8('$123.77', '1');\n> > > cash_mul_flt8\n> > > -------------\n> > > $123.77\n> > > (1 row)\n> \n> It only happens here for some figues, not all.... mostly xx.x[1234]\n\nI used your exact example, and it worked here.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 11:15:15 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "\nSorry, should have been the first thing to say... \n\nLinux 2.0.29\nGCC 2.7.2\n/lib/libc.so.5.3.12\n\nFor some reason it won't compile on my newer system GCC 2.8.1/glibc but\nI'll look at that when I have some time....\n\nOn Wed, 4 Nov 1998, Thomas G. Lockhart wrote:\n\n> > It only happens here for some figures, not all.... mostly xx.x[1234]\n> \n> What system and compiler are you using?\n> \n> - Tom\n> \n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n", "msg_date": "Wed, 4 Nov 1998 17:03:57 +0000 (GMT)", "msg_from": "A James Lewis <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "> Linux 2.0.29\n> GCC 2.7.2\n> /lib/libc.so.5.3.12\n\nSimilar to mine (Linux-2.0.30, gcc-2.7.2.1, same libc), which also\nshowed the symptom. I've also tried compiling in the past with egcs-1.1\nand found many more floating point rounding differences.\n\nSo it's a gcc/libc problem. In it's defense, I'm more accustomed to\nseeing more odd rounding behavior with other platforms.\n\nbtw, this is definitely not* an issue to affect our upcoming release.\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 17:37:34 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "Thus spake Tom Lane\n> And hardware-dependent, evidently, because I get the right thing on HP:\n> \n> play=> select cash_mul_flt8('$1', '123.77');\n> cash_mul_flt8\n> -------------\n> $123.77\n> (1 row)\n\nWorks fine on NetBSD -current.\n\n> ! \tresult = floor((f) * (c) + 0.5);\n\nIs this correct? What about \"select cash_mul_flt8('$1', '123.99')?\"\nWouldn't that convert it to 124.00?\n\n> The other cash functions also need to be looked at for rounding\n> sloppiness, so I see no point in committing this fix by its lonesome.\n> \n> Lessee, who's the author of cash.c ... looks like it's D'Arcy ...\n> D'Arcy, you wanna put this on your to-do list?\n\nWell, I didn't do those portions but I can look at it at some point I\nsuppose. I'm just not sure that it is as simple as suggested above.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Wed, 4 Nov 1998 12:42:58 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "> Well, I didn't do those portions but I can look at it at some point I\n> suppose. I'm just not sure that it is as simple as suggested above.\n\nRight. I'd suggest going slow until we understand where this is actually\ncoming from. If you didn't add these in then it's possible I did; don't\nremember though.\n\n - Thomas\n", "msg_date": "Wed, 04 Nov 1998 17:49:22 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "[email protected] (D'Arcy J.M. Cain) writes:\n>> ! \tresult = floor((f) (c) + 0.5);\n\n> Is this correct? What about \"select cash_mul_flt8('$1', '123.99')?\"\n> Wouldn't that convert it to 124.00?\n\nNo, because the cash type is actually \"int\". Presumably it's scaled\nup by a factor of 100 compared to what's printed --- otherwise you'd\nnever be able to represent pennies at all. So the \"round to nearest\ninteger\" logic depicted above is really rounding to nearest penny.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 04 Nov 1998 16:19:15 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!! " }, { "msg_contents": "A James Lewis <[email protected]> writes:\n\n> Sorry, should have been the first thing to say... \n> \n> Linux 2.0.29\n> GCC 2.7.2\n> /lib/libc.so.5.3.12\n> \n> For some reason it won't compile on my newer system GCC 2.8.1/glibc but\n> I'll look at that when I have some time....\n\nI just tried the same example on 6.4beta4, with linux2.0.36pre\nsomething, glibc and egcs1.1. (It compiled fine, for what it's\nworth.)\n\nI also get this erroneous behaviour:\n\npostgres=> select cash_mul_flt8('$1', '123.77');\ncash_mul_flt8\n-------------\n$123.76 \n(1 row)\n\nand\n\npostgres=> select cash_mul_flt8('$1', '123.99');\ncash_mul_flt8\n-------------\n$123.98 \n(1 row)\n\nThis may be sense from a numerical analysis point of view, but it\nsuggests there's a problem somewhere: this is surely behaving in a\nsurprising way.\n", "msg_date": "04 Nov 1998 22:54:43 +0000", "msg_from": "Bruce Stephens <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "Should I add this to the TODO list?\n\n> \"Thomas G. Lockhart\" <[email protected]> writes:\n> > tgl=> select cash_mul_flt8('$1', '123.77');\n> > cash_mul_flt8\n> > -------------\n> > $123.76\n> > (1 row)\n> > That's annoying; it's non-symmetric too.\n> \n> And hardware-dependent, evidently, because I get the right thing on HP:\n> \n> play=> select cash_mul_flt8('$1', '123.77');\n> cash_mul_flt8\n> -------------\n> $123.77\n> (1 row)\n> \n> \n> > The money type is stored as an\n> > integer, and the float type is an IEEE double; looks like we have an LSB\n> > rounding problem. Not sure what to do about it other than remove the\n> > function, which isn't desirable I'm sure...\n> \n> What's needed is explicit rounding. Instead of letting the compiler do\n> the rounding during its implicit float-to-int conversion (which\n> generally truncates towards 0 or towards -infinity), do\n> \n> cash.c~\tWed Nov 4 10:11:57 1998\n> --- cash.c\tWed Nov 4 10:13:06 1998\n> ***********\n> * 17,22 **\n> --- 17,23 ----\n> #include <limits.h>\n> #include <ctype.h>\n> #include <locale.h>\n> + #include <math.h>\n> \n> #include \"postgres.h\"\n> #include \"miscadmin.h\"\n> ***********\n> * 419,425 ***\n> \tif (!PointerIsValid(result = palloc(sizeof(Cash))))\n> \t\telog(ERROR, \"Memory allocation failed, can't multiply cash\", NULL);\n> \n> ! \tresult = ((f) (c));\n> \n> \treturn result;\n> }\t/ cash_mul_flt8() /\n> --- 420,426 ----\n> \tif (!PointerIsValid(result = palloc(sizeof(Cash))))\n> \t\telog(ERROR, \"Memory allocation failed, can't multiply cash\", NULL);\n> \n> ! \tresult = floor((f) (c) + 0.5);\n> \n> \treturn result;\n> }\t/ cash_mul_flt8() /\n> \n> \n> (Alternatively you could do it with rint(), although I distrust rint()\n> because it's context-dependent...)\n> \n> The other cash functions also need to be looked at for rounding\n> sloppiness, so I see no point in committing this fix by its lonesome.\n> \n> Lessee, who's the author of cash.c ... looks like it's D'Arcy ...\n> D'Arcy, you wanna put this on your to-do list?\n> \n> \t\t\tregards, tom lane\n> \n> \n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 11 Dec 1998 22:54:06 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> Should I add this to the TODO list?\n>> \n>> What's needed is explicit rounding. Instead of letting the compiler do\n>> the rounding during its implicit float-to-int conversion (which\n>> generally truncates towards 0 or towards -infinity), do\n>> [snip]\n>> The other cash functions also need to be looked at for rounding\n>> sloppiness, so I see no point in committing this fix by its lonesome.\n\nYes, that still needs to be dealt with. I suggest\n\n\t Fix roundoff problems in \"cash\" datatype.\n\nActually, the whole cash datatype ought to be obsoleted by real\nfixed-point-numeric support, but that's a much bigger project...\n\n\t\t\tregards, tom lane\n", "msg_date": "Sat, 12 Dec 1998 12:31:27 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!! " }, { "msg_contents": "> Bruce Momjian <[email protected]> writes:\n> > Should I add this to the TODO list?\n> >> \n> >> What's needed is explicit rounding. Instead of letting the compiler do\n> >> the rounding during its implicit float-to-int conversion (which\n> >> generally truncates towards 0 or towards -infinity), do\n> >> [snip]\n> >> The other cash functions also need to be looked at for rounding\n> >> sloppiness, so I see no point in committing this fix by its lonesome.\n> \n> Yes, that still needs to be dealt with. I suggest\n> \n> \t* Fix roundoff problems in \"cash\" datatype.\n> \n\nAdded.\n\n> Actually, the whole cash datatype ought to be obsoleted by real\n> fixed-point-numeric support, but that's a much bigger project...\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sat, 12 Dec 1998 15:03:18 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Warning!!" } ]
[ { "msg_contents": "I'm currently experimenting with the Debian packaging of 6.4.\n\nThe question arises of how to handle the change in library versions from\n1.0 to 2.0. It has been suggested that I should keep libpq.so.1 around\nfor the benefit of those who need to communicate with an older database\nserver.\n\nIs it possible to use libpq.so.2.0 to query a remote database whose\nserver is running PostgreSQL 6.3? What would be the consequences of\nusing libpq.so.1 to query a 6.4 server?\n\n-- \nOliver Elphick [email protected]\nIsle of Wight http://www.lfix.co.uk/oliver\n PGP key from public servers; key ID 32B8FAA1\n ========================================\n \"Lo, children are an heritage of the LORD; and the \n fruit of the womb is his reward.\" Psalms 127:3 \n\n\n", "msg_date": "Wed, 04 Nov 1998 00:09:21 +0000", "msg_from": "\"Oliver Elphick\" <[email protected]>", "msg_from_op": true, "msg_subject": "Mixing library versions" }, { "msg_contents": "On Wed, Nov 04, 1998 at 12:09:21AM +0000, Oliver Elphick wrote:\n> I'm currently experimenting with the Debian packaging of 6.4.\n> \n> The question arises of how to handle the change in library versions from\n> 1.0 to 2.0. It has been suggested that I should keep libpq.so.1 around\n> for the benefit of those who need to communicate with an older database\n> server.\n> \n> Is it possible to use libpq.so.2.0 to query a remote database whose\n> server is running PostgreSQL 6.3? What would be the consequences of\n\nNo. Let's just try /usr/local/pgsql/bin/psql(6.4) on my running 6.3 server:\n\nConnection to database 'mm' failed.\nUnsupported frontend protocol.\n\n> using libpq.so.1 to query a 6.4 server?\n\nNever tried that until now. But it appears to work. I hvae no idea though\nwhere it could break things.\n\nMichael\n-- \nDr. Michael Meskes \| Th.-Heuss-Str. 61, D-41812 Erkelenz \| Go SF49ers!\nSenior-Consultant \| business: [email protected] \| Go Rhein Fire!\nMummert+Partner \| private: [email protected] \| Use Debian\nUnternehmensberatung AG \| [email protected] \| GNU/Linux!\n", "msg_date": "Wed, 4 Nov 1998 09:00:20 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Mixing library versions" }, { "msg_contents": "> On Wed, Nov 04, 1998 at 12:09:21AM +0000, Oliver Elphick wrote:\n> > I'm currently experimenting with the Debian packaging of 6.4.\n> > \n> > The question arises of how to handle the change in library versions from\n> > 1.0 to 2.0. It has been suggested that I should keep libpq.so.1 around\n> > for the benefit of those who need to communicate with an older database\n> > server.\n> > \n> > Is it possible to use libpq.so.2.0 to query a remote database whose\n> > server is running PostgreSQL 6.3? What would be the consequences of\n> \n> No. Let's just try /usr/local/pgsql/bin/psql(6.4) on my running 6.3 server:\n> \n> Connection to database 'mm' failed.\n> Unsupported frontend protocol.\n> \n> > using libpq.so.1 to query a 6.4 server?\n> \n> Never tried that until now. But it appears to work. I hvae no idea though\n> where it could break things.\n\n2.0 and 1.0 libpq can connect to 6.4, but 2.0 libpq can not connect to\nolder databases. In other words, the backend can communicate with old\nlibpq, but old backends can not communicate with 2.0 libpq.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 09:04:31 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Mixing library versions" }, { "msg_contents": "\"Oliver Elphick\" <[email protected]> writes:\n> Is it possible to use libpq.so.2.0 to query a remote database whose\n> server is running PostgreSQL 6.3?\n\nNo. That's why we changed the major version number, to make it\neasier to keep an old libpq around if you need it.\n\n> What would be the consequences of\n> using libpq.so.1 to query a 6.4 server?\n\nShould work fine. The new server still handles the old FE/BE protocol.\n(Or at least, it's supposed to ... I haven't tested that lately, but\nI believe Tatsuo-san did.)\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 04 Nov 1998 10:00:03 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Mixing library versions " } ]
[ { "msg_contents": "\ntest_db=> select cash_mul_flt8('$1', 100.72);\ncash_mul_flt8\n-------------\n$100.71 \n(1 row)\n\nDo I have to explain whats wrong here?\n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n", "msg_date": "Wed, 4 Nov 1998 02:35:08 +0000 (GMT)", "msg_from": "A James Lewis <[email protected]>", "msg_from_op": true, "msg_subject": "Bug report for 6.4Beta5 (Is this the right place?)" } ]
[ { "msg_contents": "> 1). config.sub and config.guess need execute permissions\n\nI think I've fixed that. It came from originally transferring the files\nto the cvs server via ftp, which lost the correct file permissions.\n\n> 2). I'm including a patch file for the rest.\n\nOK, will look at them. Thanks for the feedback.\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 06:34:06 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: proposed psqlodbc installation doc" }, { "msg_contents": "> I went through your psqlODBC installation on Solaris 2.6 and have a \n> few minor fixes:\n\nThanks again for the patches. I've applied all of them so it should make\nit into the final v6.4 release.\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 16:55:25 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: proposed psqlodbc installation doc" } ]
[ { "msg_contents": "I brought my snapshot up-to-date this morning, recompiled everything and\ntried to connect to my test database that is about a month old. I can\nï¿½select * from pg_attributeï¿½ without a problem. But hitting ï¿½\\zï¿½ in psql\ncauses the transfer to be closed. Other psql options like ï¿½\\dï¿½ work fine.\n\nMichael\n-- \nDr. Michael Meskes \| Th.-Heuss-Str. 61, D-41812 Erkelenz \| Go SF49ers!\nSenior-Consultant \| business: [email protected] \| Go Rhein Fire!\nMummert+Partner \| private: [email protected] \| Use Debian\nUnternehmensberatung AG \| [email protected] \| GNU/Linux!\n", "msg_date": "Wed, 4 Nov 1998 10:41:54 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "problem with latest snapshot" }, { "msg_contents": "> I brought my snapshot up-to-date this morning, recompiled everything \n> and tried to connect to my test database that is about a month old.\n\nYeah, you need a dump/reload for anything built earlier than a week or\nso ago. Last-minute but wildly helpful cleanups of pg_operator from Tom\nLane...\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 15:24:31 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] problem with latest snapshot" }, { "msg_contents": "Michael Meskes <[email protected]> writes:\n> I brought my snapshot up-to-date this morning, recompiled everything and\n> tried to connect to my test database that is about a month old. I can\n> 'select * from pg_attribute' without a problem. But hitting '\\z' in psql\n> causes the transfer to be closed. Other psql options like '\\d' work\n> fine.\n\nFWIW, \\z works fine for me. Did you do an initdb after updating?\nThere were some catalog changes as late as last week.\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 04 Nov 1998 10:26:29 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] problem with latest snapshot " }, { "msg_contents": "On Wed, Nov 04, 1998 at 03:24:31PM +0000, Thomas G. Lockhart wrote:\n> Yeah, you need a dump/reload for anything built earlier than a week or\n> so ago. Last-minute but wildly helpful cleanups of pg_operator from Tom\n> Lane...\n\nIt works now. Thanks.\n\nMichael\n\n-- \nDr. Michael Meskes, Leiter Niederlassung West, Datenrevision GmbH\nbusiness: Cuxhavener Str. 36, D-21149 Hamburg\n private: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Thu, 5 Nov 1998 10:32:32 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] problem with latest snapshot" } ]
[ { "msg_contents": "Hi all,\n\n----I'm trying to understand this very interesting language PL/pgSQL\nthanks to Jan Wieck,\n----finally we can trap NULL values as in:\n\ncreate function len(text) returns text as '\ndeclare\n nonullo alias for $1;\nbegin\n if nonullo then\n return length(nonullo);\n else\n return ''<NULL>'';\n end if;\nend;' language 'plpgsql';\nCREATE\n\nselect len(c) from a;\nlen\n------\n12\n<NULL>\n<NULL>\n<NULL>\n12\n(5 rows)\n\n\n------I'm trying to create a function that returns a value instead of\nNULL.\n------The following one works well:\n\ncreate function nvl(integer) returns integer as '\ndeclare\n nullo integer := 0;\n nonullo alias for $1;\nbegin\n if NONULLO then\n return NONULLO;\n else\n return NULLO;\n end if;\nend;' language 'plpgsql';\nCREATE\n\nselect nvl(i) from a;\n nvl\n-------\n 0\n 0\n2232767\n 0\n(4 rows)\n\n\n------I would like to report this little thing, seems PL/pgSQL isn't\ncase insensitive:\ndrop function nvl(integer,integer);\nDROP\ncreate function nvl(integer) returns integer as '\ndeclare\n NULLO integer := 0;\n NONULLO alias for $1;\nbegin\n if NONULLO then\n return NONULLO;\n else\n return NULLO;\n end if;\nend;' language 'plpgsql';\nCREATE\n\nselect nvl(i) from a;\nERROR: attribute 'nonullo' not found\n\n-------and seems it returns only constant values:\n\ndrop function nvl(integer,integer);\nDROP\ncreate function nvl(integer,integer) returns integer as '\ndeclare\n nonullo alias for $1;\n nullo ALIAS FOR $2;\nbegin\n if NONULLO then\n return NONULLO;\n else\n return NULLO;\n end if;\nend;' language 'plpgsql';\nCREATE\n\nselect nvl(i,0) from a;\n nvl\n-------\n\n\n2232767\n\n(4 rows)\n\n Jose'\n", "msg_date": "Wed, 04 Nov 1998 12:20:53 +0100", "msg_from": "\"Jose' Soares\" <[email protected]>", "msg_from_op": true, "msg_subject": "PL/pgSQL a great procedural language for PostgreSQL" } ]
[ { "msg_contents": "The attached commands crash the backend, which exits with status 11. The\nbackend log (debug level 255) does not show anything that is meaningful to\nme. It shows a query, which seems to relate to the final constraint, and\nthen terminates without further explanation:\n\n...\nquery: select 1 from individual where NOT ( surname IS NULL AND forenames IS \nNUL\nL ) \nparser outputs: \n{ QUERY \n...\n } \nafter rewriting: \n{ QUERY \n... \n } \n/usr/lib/postgresql/bin/postmaster: reaping dead processes...\n/usr/lib/postgresql/bin/postmaster: CleanupProc: pid 8970 exited with status 11\n...\n\n\nI don't know where to start looking. I can trace the backend with the\ndebugger if you will tell me which routine to break at.\n\n\nOliver Elphick [email protected]\nIsle of Wight http://www.lfix.co.uk/oliver\n PGP key from public servers; key ID 32B8FAA1\n ========================================\n \"Lo, children are an heritage of the LORD; and the \n fruit of the womb is his reward.\" Psalms 127:3", "msg_date": "Wed, 04 Nov 1998 13:09:03 +0000", "msg_from": "\"Oliver Elphick\" <[email protected]>", "msg_from_op": true, "msg_subject": "latest snapshot crashes backend" }, { "msg_contents": "> The attached commands crash the backend, which exits with status 11.\n\ntgl=> select * from x where not (i is null or c is null);\ni\|c\n-+-\n1\|T\n2\|A\n0\|T\n(3 rows)\n\ntgl=> select * from x where not (i is null and c is null);\npqReadData() -- backend closed the channel unexpectedly.\n\nSo, let's try rewriting it as a workaround:\n\ntgl=> select * from x where (not i is null) or (not c is null);\npqReadData() -- backend closed the channel unexpectedly.\n\nOops. For some reason the NOT/AND is fatal, while NOT/OR is OK. That's\nnot so good. The good news is that it will certainly be repairable with\npatches.\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 15:35:30 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] latest snapshot crashes backend" }, { "msg_contents": "> > The attached commands crash the backend, which exits with status 11.\n> \n> tgl=> select * from x where not (i is null or c is null);\n> i\|c\n> -+-\n> 1\|T\n> 2\|A\n> 0\|T\n> (3 rows)\n> \n> tgl=> select * from x where not (i is null and c is null);\n> pqReadData() -- backend closed the channel unexpectedly.\n> \n> So, let's try rewriting it as a workaround:\n> \n> tgl=> select * from x where (not i is null) or (not c is null);\n> pqReadData() -- backend closed the channel unexpectedly.\n> \n> Oops. For some reason the NOT/AND is fatal, while NOT/OR is OK. That's\n> not so good. The good news is that it will certainly be repairable with\n> patches.\n> \n\nCare to give us a table:\n\n\tselect * from pg_shadow where (usesysid is null and oid is null)\\g\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 12:10:25 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] latest snapshot crashes backend" }, { "msg_contents": "> Care to give us a table:\n> select * from pg_shadow where (usesysid is null and oid is null);\n\nSure, that's easy. Just negate your where clause:\n\ntgl=> select * from pg_shadow\ntgl-> where not (usesysid is null and oid is null);\npqReadData() -- backend closed the channel unexpectedly.\n\nIf you can reproduce this, will you have a chance to look at it? I don't\nknow where the problem is, but suspect that it is farther back than the\nparser transformations. Rewrite system, optimizer, or executor??\n\n - Tom\n", "msg_date": "Wed, 04 Nov 1998 17:40:53 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] latest snapshot crashes backend" }, { "msg_contents": "> > Care to give us a table:\n> > select * from pg_shadow where (usesysid is null and oid is null);\n> \n> Sure, that's easy. Just negate your where clause:\n> \n> tgl=> select * from pg_shadow\n> tgl-> where not (usesysid is null and oid is null);\n> pqReadData() -- backend closed the channel unexpectedly.\n> \n> If you can reproduce this, will you have a chance to look at it? I don't\n> know where the problem is, but suspect that it is farther back than the\n> parser transformations. Rewrite system, optimizer, or executor??\n\nI will look at it. I can reproduce it here.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 12:41:50 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] latest snapshot crashes backend" }, { "msg_contents": "> > tgl=> select * from pg_shadow\n> > tgl-> where not (usesysid is null and oid is null);\n> > pqReadData() -- backend closed the channel unexpectedly.\n\nbtw, you may already have noticed that this doesn't crash if the same\nclause is changed to be a target:\n\ntgl=> select not (usesysid is null and oid is null) from pg_shadow;\n?column?\n--------\nt\nt\n(2 rows)\n\n - Thomas\n", "msg_date": "Wed, 04 Nov 1998 17:54:11 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] latest snapshot crashes backend" }, { "msg_contents": "> > Care to give us a table:\n> > select * from pg_shadow where (usesysid is null and oid is null);\n> \n> Sure, that's easy. Just negate your where clause:\n> \n> tgl=> select * from pg_shadow\n> tgl-> where not (usesysid is null and oid is null);\n> pqReadData() -- backend closed the channel unexpectedly.\n> \n> If you can reproduce this, will you have a chance to look at it? I don't\n> know where the problem is, but suspect that it is farther back than the\n> parser transformations. Rewrite system, optimizer, or executor??\n\nHere is the problem:\n\t\n\t(gdb) print ((Expr ) clause)->oper\n\t$1 = (Node ) 0x0\n\nWe have dealt with this before in relation to NOT. The system assumes\nNOT is an OP_EXPR oper, while it really isn't.\n\nI fixed another query about a month ago relating to this. In that case,\nflatten_tlistentry() was loosing information about EXPR nodes.\n\nIn this case, the code:\n\n Oid opno = ((Oper ) ((Expr ) clause)->oper)->opno;\n\nclearly is making an assumption it should not be making.\n\nIn this case, clause is:\n\n\t(gdb) print ((Expr )clause)[0]\n\t$3 = {type = T_Expr, typeOid = 0, opType = NOT_EXPR, oper = 0x0, \n\t args = 0x8307f90}\n\nwhere opType is not OP_EXPR, but NOT_EXPR.\n\nI am inclined to check for NOT_EXPR, let the selectivity be computed on\nthe subclause, and negate(NOT) the resulting selectivity. Not sure on\nan exact fix yet, but this is a 6.4.1 issue, anyway.\n\n\nProblem is in the optimizer:\n\n---------------------------------------------------------------------------\n\n\n#0 0x80bb1c4 in compute_selec (root=0x82d8590, clauses=0x8307fb0, \n or_selectivities=0x0) at clausesel.c:274\n#1 0x80bb0cb in compute_clause_selec (root=0x82d8590, clause=0x830ea50, \n or_selectivities=0x0) at clausesel.c:182\n#2 0x80bfc9c in add_clause_to_rels (root=0x82d8590, clause=0x830ea50)\n at initsplan.c:211\n#3 0x80bfbea in init_base_rels_qual (root=0x82d8590, clauses=0x830f050)\n at initsplan.c:155\n#4 0x80c01bb in subplanner (root=0x82d8590, flat_tlist=0x830f170, \n qual=0x830f050) at planmain.c:262\n#5 0x80c00f8 in query_planner (root=0x82d8590, command_type=1, \n tlist=0x83077d0, qual=0x83063d0) at planmain.c:176\n#6 0x80c085d in union_planner (parse=0x82d8590) at planner.c:151\n#7 0x80c06d5 in planner (parse=0x82d8590) at planner.c:72\n#8 0x80f56b4 in pg_parse_and_plan (\n query_string=0x80455e4 \"select from pg_shadow where not (usesysid is null and oid is null);\\n\", typev=0x0, nargs=0, queryListP=0x8045594, dest=Debug, \n aclOverride=0 '\\000') at postgres.c:628\n#9 0x80f57f3 in pg_exec_query_dest (\n query_string=0x80455e4 \"select * from pg_shadow where not (usesysid is null and oid is null);\\n\", dest=Debug, aclOverride=0) at postgres.c:722\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 15:01:40 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] latest snapshot crashes backend" }, { "msg_contents": "> > Care to give us a table:\n> > select * from pg_shadow where (usesysid is null and oid is null);\n> \n> Sure, that's easy. Just negate your where clause:\n> \n> tgl=> select * from pg_shadow\n> tgl-> where not (usesysid is null and oid is null);\n> pqReadData() -- backend closed the channel unexpectedly.\n> \n> If you can reproduce this, will you have a chance to look at it? I don't\n> know where the problem is, but suspect that it is farther back than the\n> parser transformations. Rewrite system, optimizer, or executor??\n> \n\nOK, here is the fix. I am not applying it yet until 6.4 is released by\nMarc, OK? In fact, I am not sure how we are going to do patch\napplication after 6.4, so would someone else please apply this?\n\n---------------------------------------------------------------------------\n\n* ./backend/optimizer/path/clausesel.c.orig\tWed Nov 4 16:49:35 1998\n--- ./backend/optimizer/path/clausesel.c\tWed Nov 4 17:11:02 1998\n***********\n* 254,259 ***\n--- 254,264 ----\n \t\t /\n \t\ts1 = 0.1;\n \t}\n+ \telse if (not_clause((Node ) clause))\n+ \t{\n+ \t\t/ negate this baby /\n+ \t\treturn 1 - compute_selec(root, ((Expr )clause)->args, or_selectivities);\n+ \t}\n \telse if (is_subplan((Node *) clause))\n \t{\n \n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 17:13:02 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] latest snapshot crashes backend" } ]
[ { "msg_contents": "Hi,\n\nI'm running PostgreSQL 6.3.2 on Linux 4.1. I get the following error in\npsql when I attempt to update two of the tables in my database (similiar\nupdate on other tables work fine). Can anyone tell me what's going\nwrong, or anything I can do to help diagnose the problem?\n\nsplat=> update status set version=0 where entry_id=402653541;\nERROR: fmgr_info: function 186528: cache lookup failed\n\nPS A select on this same table works fine:\n\nsplat=> select version from status where entry_id=402653541;\nversion\n-------\n -1\n(1 row)\n\n'version' is defined as an int4\n\n\nThanks,\nLyn Finman\n\n\n\n", "msg_date": "Wed, 04 Nov 1998 09:03:50 -0500", "msg_from": "\"Lyn K. Finman\" <[email protected]>", "msg_from_op": true, "msg_subject": "cache lookup failed" } ]
[ { "msg_contents": "> Log Message:\n> Test for and handle NULL arguments to mylog() print statements.\n> Include definition for _IN_ADDR_T if INADDR_NONE is not defined.\n\nI've applied some patches from Frank Delahoyde at Scripps to help with\npsqlODBC compilation and running under Solaris.\n\nIt still compiles and runs under Linux (the only other platform actually\ntested for Unix-native ODBC afaik), so I think this is OK to ship.\n", "msg_date": "Wed, 04 Nov 1998 16:50:20 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [COMMITTERS] 'pgsql/src/interfaces/odbc convert.c qresult.c\n\tresults.c socket.h'" } ]
[ { "msg_contents": "On a RH5.1 box.\nI'm getting dual printing of every error message. \nAlmost panicked when I ran the regression test, but the only fails were\nfor different error messages in int, some float handling function not\noverflowing (exp pow), and some differences in float8 as far as unit in\nthe last place. \n\t-DEJ\n", "msg_date": "Wed, 4 Nov 1998 12:50:12 -0600", "msg_from": "\"Jackson, DeJuan\" <[email protected]>", "msg_from_op": true, "msg_subject": "11/4/98 Snapshot" } ]
[ { "msg_contents": "I was hoping someone could shed some light on the \nfollowing problem:\n\nThe subselect under 6.4 ignores indexes built on \ntables. For example, consider the following two\ntables:\n\n\\d sales;\n\nTable = sales\n\nsupplysource varchar() not null 16\nsupply varchar() not null 16\nsupplyunit varchar() not null 2\nquantity float8 not null 8\ntarget varchar() not null 16\ncostcntr varchar() not null 8\nsaletype varchar() not null 16\nsaledate datetime not null 8\n\nIndices: k_sales\n k_sales_saledate\n k_sales_supply\n k_sales_target \n\n\n\\d locations;\n\nTable = locations\nlocation varchar() not null 16\ncostcntr varchar() not null 5\nsupplypath varchar() not null 64\nformpath varchar() not null 64\nengineerpath varchar() not null 64\n\nIndices: k_locations\n k_locations_costcntr\n k_locations_location \n\nwith the following query:\n\nexplain select supply from sales where target in\n(select location from locations);\nNOTICE: QUERY PLAN:\n\nSeq Scan on sales (cost=5738.60 size=116806 width=12)\n SubPlan\n -> Seq Scan on locations (cost=7.49 size=136 width=12)\n\nEXPLAIN\n\nThis results in a full table scan on the outer table\nwhere there are 150K rows. As a result, it is the\nequivalent of a full table scan on 150K (sales) * \n100 (locations) rows = 15,000,000 row table scan.\n\nOracle 8.0.5.00 for Linux Early Adaptor Program \ncorrectly uses the index on the same two tables\nresulting on a sequential access on the subselected\ntable (locations) and an index look-up on the \nselect table (sales), according to EXPLAIN PLAN.\n\nDue to the failure of PostgreSQL 6.4 to correctly\nuse indexes with subselects, it makes subselects\npractically unusable.\n\n(NOTE: This occurs regardless of whether or not it\nis executed immediately following a VACUUM)\n\nAre there any patches available to fix this problem?\nWill this fixed in the 6.4 release?\n\nThanks for any information,\n\nMarcus Mascari\n([email protected]) \n\n\n\n Date:\n Tue, 27 Oct 1998 17:30:44 -0800 (PST)\n From:\n Marcus Mascari <[email protected]>Add to Address Book\n Subject:\n Please help (Someone?) - subselect not using indexes\n To:\n [email protected]\n\n\n\nI was hoping someone could shed some light on the \nfollowing problem:\n\nThe subselect under 6.4 ignores indexes built on \ntables. For example, consider the following two\ntables:\n\n\\d sales;\n\nTable = sales\n\nsupplysource varchar() not null 16\nsupply varchar() not null 16\nsupplyunit varchar() not null 2\nquantity float8 not null 8\ntarget varchar() not null 16\ncostcntr varchar() not null 8\nsaletype varchar() not null 16\nsaledate datetime not null 8\n\nIndices: k_sales\n k_sales_saledate\n k_sales_supply\n k_sales_target \n\n\n\\d locations;\n\nTable = locations\nlocation varchar() not null 16\ncostcntr varchar() not null 5\nsupplypath varchar() not null 64\nformpath varchar() not null 64\nengineerpath varchar() not null 64\n\nIndices: k_locations\n k_locations_costcntr\n k_locations_location \n\nwith the following query:\n\nexplain select supply from sales where target in\n(select location from locations);\nNOTICE: QUERY PLAN:\n\nSeq Scan on sales (cost=5738.60 size=116806 width=12)\n SubPlan\n -> Seq Scan on locations (cost=7.49 size=136 width=12)\n\nEXPLAIN\n\nThis results in a full table scan on the outer table\nwhere there are 150K rows. As a result, it is the\nequivalent of a full table scan on 150K (sales) * \n100 (locations) rows = 15,000,000 row table scan.\n\nOracle 8.0.5.00 for Linux Early Adaptor Program \ncorrectly uses the index on the same two tables\nresulting on a sequential access on the subselected\ntable (locations) and an index look-up on the \nselect table (sales), according to EXPLAIN PLAN.\n\nDue to the failure of PostgreSQL 6.4 to correctly\nuse indexes with subselects, it makes subselects\npractically unusable.\n\n(NOTE: This occurs regardless of whether or not it\nis executed immediately following a VACUUM)\n\nAre there any patches available to fix this problem?\nWill this fixed in the 6.4 release?\n\nThanks for any information,\n\nMarcus Mascari\n([email protected])\n\n\n\n\n\n_________________________________________________________\nDO YOU YAHOO!?\nGet your free @yahoo.com address at http://mail.yahoo.com\n\n", "msg_date": "Wed, 4 Nov 1998 11:33:02 -0800 (PST)", "msg_from": "Marcus Mascari <[email protected]>", "msg_from_op": true, "msg_subject": "PLEASE...SOMEONE COMMENT: PostgreSQL 6.4BETA not using indexes with\n\tsubselects" } ]
[ { "msg_contents": "It appears that when a function is called if any of the paramaters are\nNULL all of the parameters are NULL.\ntry:\n drop function nvl(int, int);\n create function nvl(int, int) returns boolean as '\n declare\n nonullo alias as $1;\n nullo alias as $2;\n begin\n return (nonullo IS NULL) AND (nullo IS NULL);\n end;' language 'plpgsql';\n select nvl(i,0) from a;\nyou should get:\nnvl\n---\nt\nt\nf\nt\n(4 rows)\n\n\n\n> Hi all,\n> \n> ----I'm trying to understand this very interesting language PL/pgSQL\n> thanks to Jan Wieck,\n> ----finally we can trap NULL values as in:\n> \n> -------and seems it returns only constant values:\n> \n> drop function nvl(integer,integer);\n> DROP\n> create function nvl(integer,integer) returns integer as '\n> declare\n> nonullo alias for $1;\n> nullo ALIAS FOR $2;\n> begin\n> if NONULLO then\n> return NONULLO;\n> else\n> return NULLO;\n> end if;\n> end;' language 'plpgsql';\n> CREATE\n> \n> select nvl(i,0) from a;\n> nvl\n> -------\n> \n> \n> 2232767\n> \n> (4 rows)\n> \n> Jose'\n", "msg_date": "Wed, 4 Nov 1998 13:52:39 -0600", "msg_from": "\"Jackson, DeJuan\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] PL/pgSQL a great procedural language for PostgreSQL" }, { "msg_contents": ">\n> It appears that when a function is called if any of the paramaters are\n> NULL all of the parameters are NULL.\n> try:\n> drop function nvl(int, int);\n> create function nvl(int, int) returns boolean as '\n> declare\n> nonullo alias as $1;\n> nullo alias as $2;\n> begin\n> return (nonullo IS NULL) AND (nullo IS NULL);\n> end;' language 'plpgsql';\n> select nvl(i,0) from a;\n> you should get:\n> nvl\n> ---\n> t\n> t\n> f\n> t\n> (4 rows)\n\n Don't blame PL/pgSQL for that. There is only one bool isNull\n pointer given to PL handlers. How should the PL handler know,\n which of the arguments are null then? As I said on another\n thread, the function call interface needs to get redesigned.\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, 5 Nov 1998 12:37:44 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/pgSQL a great procedural language for PostgreSQL" } ]
[ { "msg_contents": "Nevermind, I forgot to capture stderr from the backend.\n\t-DEJ\n\n> -----Original Message-----\n> From:\tJackson, DeJuan [SMTP:[email protected]]\n> Sent:\tWednesday, November 04, 1998 12:50 PM\n> To:\tPostgreSQL Hackers Mailing List\n> Subject:\t[HACKERS] 11/4/98 Snapshot\n> \n> On a RH5.1 box.\n> I'm getting dual printing of every error message. \n> Almost panicked when I ran the regression test, but the only fails\n> were\n> for different error messages in int, some float handling function not\n> overflowing (exp pow), and some differences in float8 as far as unit\n> in\n> the last place. \n> \t-DEJ\n", "msg_date": "Wed, 4 Nov 1998 15:32:28 -0600", "msg_from": "\"Jackson, DeJuan\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] 11/4/98 Snapshot" } ]
[ { "msg_contents": "\nMorning all...\n\n\tWell, v6.4 is branched and in a \"released state\"...those wanting\nto start moving on, go for it. Those with committer's access to the\nsource tree, Thomas and I are working up docs for how to access the REL6_4\nbranch...\n\n\tFor those wishing to move forward, nothing has changed...except\nthere is no longer a feature freeze.\n\n\tFor those submitting patches for REL6_4, please tag your subject\nlines [REL6_4]...I will be watching for those in pgsql-patches, but I\ndon't want to have to determine if its only a 6.4 vs 6.5 related patch...\n\n\tIf a patch marked [REL6_4] isn't apppllied in a reasonable amount\nof time (ie. I disappear pretty much from Friday to Monday), please feel\nfree to repost it...but as long as that tag is there, I shouldn't miss\nanything...\n\n\tI will be wrapping up the full release later this evening, after I\neat something...\n\n\tGreat work everyone...even if the last bit was a little ...\n'stressful' :)\n\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Wed, 4 Nov 1998 19:50:04 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Open the flood gates...v6.4 is tag'd..." }, { "msg_contents": "Thus spake The Hermit Hacker\n> \tWell, v6.4 is branched and in a \"released state\"...those wanting\n> to start moving on, go for it. Those with committer's access to the\n> source tree, Thomas and I are working up docs for how to access the REL6_4\n> branch...\n\nSo do I send in my patch or have we decided to fix the null arg problem\nin the function dispatcher?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n\n", "msg_date": "Wed, 4 Nov 1998 19:46:13 -0500 (EST)", "msg_from": "[email protected]", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd..." }, { "msg_contents": "> \tWell, v6.4 is branched and in a \"released state\"...those wanting\n> to start moving on, go for it.\n\nI did a fresh 'cvs checkout' just now to make sure I had a good set\nof files, and was dismayed to discover that a whole bunch of dead\nsubdirectories have reappeared in the CVS tree. For example,\n\tpgsql/contrib/ip_and_mac\n\tpgsql/contrib/multikey\n\tpgsql/contrib/plpgsql\n\tpgsql/contrib/sequence\n\tpgsql/contrib/zap_ltv\nwhich have nothing in them except CVS overhead.\n\nI suppose this is a side effect of marking things \"branched\" ...\nWhen you have time to spare on prettification, can you make those\ndirectories go away again?\n\n> Those with committer's access to the source tree, Thomas and I are\n> working up docs for how to access the REL6_4 branch...\n\nI assume if I just check in without paying attention, I will be\nchecking into the 6.5-to-be branch?\n\n\t\t\tregards, tom lane\n", "msg_date": "Wed, 04 Nov 1998 20:14:03 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd... " }, { "msg_contents": "> > \tWell, v6.4 is branched and in a \"released state\"...those wanting\n> > to start moving on, go for it.\n> \n> I did a fresh 'cvs checkout' just now to make sure I had a good set\n> of files, and was dismayed to discover that a whole bunch of dead\n> subdirectories have reappeared in the CVS tree. For example,\n> \tpgsql/contrib/ip_and_mac\n> \tpgsql/contrib/multikey\n> \tpgsql/contrib/plpgsql\n> \tpgsql/contrib/sequence\n> \tpgsql/contrib/zap_ltv\n> which have nothing in them except CVS overhead.\n> \n> I suppose this is a side effect of marking things \"branched\" ...\n> When you have time to spare on prettification, can you make those\n> directories go away again?\n> \n> > Those with committer's access to the source tree, Thomas and I are\n> > working up docs for how to access the REL6_4 branch...\n> \n> I assume if I just check in without paying attention, I will be\n> checking into the 6.5-to-be branch?\n\n-P flag gets rid of those.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Wed, 4 Nov 1998 20:43:36 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd..." }, { "msg_contents": "On Wed, 4 Nov 1998, Tom Lane wrote:\n\n> > \tWell, v6.4 is branched and in a \"released state\"...those wanting\n> > to start moving on, go for it.\n> \n> I did a fresh 'cvs checkout' just now to make sure I had a good set\n> of files, and was dismayed to discover that a whole bunch of dead\n> subdirectories have reappeared in the CVS tree. For example,\n> \tpgsql/contrib/ip_and_mac\n> \tpgsql/contrib/multikey\n> \tpgsql/contrib/plpgsql\n> \tpgsql/contrib/sequence\n> \tpgsql/contrib/zap_ltv\n> which have nothing in them except CVS overhead.\n> \n> I suppose this is a side effect of marking things \"branched\" ...\n> When you have time to spare on prettification, can you make those\n> directories go away again?\n\n\tDid you use the -P option? when you checkout, you should do:\n\n\tcvs checkout -P pgsql\n\n\tIt gets rid of the 'dead' directories tha way...\n\n> > Those with committer's access to the source tree, Thomas and I are\n> > working up docs for how to access the REL6_4 branch...\n> \n> I assume if I just check in without paying attention, I will be\n> checking into the 6.5-to-be branch?\n\n\tIf you didn't check out the v6.4 release explicitly (using -r\nREL6_4), you wont' check into it either...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Wed, 4 Nov 1998 22:35:09 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd... " }, { "msg_contents": "> So do I send in my patch or have we decided to fix the null arg \n> problem in the function dispatcher?\n\nI deleted one or a few messages from this thread, but at least one\nmessage which I still have (kept them once I realized this issue was\nblowing up) mentions:\n\n>> OK, there are more problems. If you apply the following patch to \n>> the regression tests you will crash the backend in a number of \n>> places.\n\nbut there was no patch enclosed or following. \n\nAnyway, I would suggest that, until we plan out what new behaviors we\nreally want in the system that we don't submit a patch which breaks many\nof the data types. That kind of change should, imho, be done offline and\ncommitted as an integrated solution. If you'd like, I'd be happy to\ntrade patches with you as we work out the issues and once we (and anyone\nelse with an interest) are happy with the new behavior and have\nregression tests which pass then we should commit to the tree.\n\nAs you mentioned, the date and time functions check for null inputs and\nbehave pretty well, and an interim solution might include using similar\ntechniques on the inet/cidr types.\n\n - Tom\n", "msg_date": "Thu, 05 Nov 1998 06:20:58 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd..." }, { "msg_contents": "Thus spake Thomas G. Lockhart\n> I deleted one or a few messages from this thread, but at least one\n> message which I still have (kept them once I realized this issue was\n> blowing up) mentions:\n\nAw! And I thought everyone was keeping all my messages just because\nof the wonderful prose.\n\n> >> OK, there are more problems. If you apply the following patch to \n> >> the regression tests you will crash the backend in a number of \n> >> places.\n> \n> but there was no patch enclosed or following. \n\nMy copy of the message has the patch by here it is again.\n\n\n* ../sql/abstime.sql\tSat May 31 22:30:19 1997\n--- abstime.sql\tMon Nov 2 09:17:48 1998\n***********\n* 23,28 **\n--- 23,30 ----\n \n INSERT INTO ABSTIME_TBL (f1) VALUES ('May 10, 1947 23:59:12');\n \n+ INSERT INTO ABSTIME_TBL (f1) VALUES (null);\n+ \n \n -- what happens if we specify slightly misformatted abstime? \n INSERT INTO ABSTIME_TBL (f1) VALUES ('Feb 35, 1946 10:00:00');\n* ../sql/char.sql\tSun Nov 30 21:46:01 1997\n--- char.sql\tMon Nov 2 09:24:51 1998\n*************\n* 30,35 **\n--- 30,38 ----\n -- zero-length char \n INSERT INTO CHAR_TBL (f1) VALUES ('');\n \n+ -- null input\n+ INSERT INTO CHAR_TBL (f1) VALUES (null);\n+ \n -- try char's of greater than 1 length \n INSERT INTO CHAR_TBL (f1) VALUES ('cd');\n \n* ../sql/circle.sql\tTue Jul 29 12:22:44 1997\n--- circle.sql\tMon Nov 2 09:26:18 1998\n*************\n* 16,21 **\n--- 16,23 ----\n \n INSERT INTO CIRCLE_TBL VALUES ('<(100,0),100>');\n \n+ INSERT INTO CIRCLE_TBL VALUES (null);\n+ \n -- bad values\n \n INSERT INTO CIRCLE_TBL VALUES ('<(-100,0),-100>');\n* ../sql/datetime.sql\tFri Nov 14 21:55:57 1997\n--- datetime.sql\tMon Nov 2 09:30:37 1998\n*************\n* 18,23 ***\n--- 18,24 ----\n INSERT INTO DATETIME_TBL VALUES ('tomorrow');\n INSERT INTO DATETIME_TBL VALUES ('tomorrow EST');\n INSERT INTO DATETIME_TBL VALUES ('tomorrow zulu');\n+ INSERT INTO DATETIME_TBL VALUES (null);\n \n SELECT count() AS one FROM DATETIME_TBL WHERE d1 = 'today'::datetime;\n SELECT count() AS one FROM DATETIME_TBL WHERE d1 = 'tomorrow'::datetime;\n************\n* 42,47 **\n--- 43,49 ----\n INSERT INTO DATETIME_TBL VALUES ('-infinity');\n INSERT INTO DATETIME_TBL VALUES ('infinity');\n INSERT INTO DATETIME_TBL VALUES ('epoch');\n+ INSERT INTO DATETIME_TBL VALUES (null);\n \n -- Postgres v6.0 standard output format\n INSERT INTO DATETIME_TBL VALUES ('Mon Feb 10 17:32:01 1997 PST');\n* ../sql/horology.sql\tSat Sep 20 12:34:07 1997\n--- horology.sql\tMon Nov 2 09:34:55 1998\n*************\n* 15,20 **\n--- 15,22 ----\n WHERE d1 BETWEEN '13-jun-1957' AND '1-jan-1997'\n OR d1 BETWEEN '1-jan-1999' AND '1-jan-2010';\n \n+ INSERT INTO TEMP_DATETIME (f1) VALUES (null);\n+ \n SELECT '' AS ten, f1 AS datetime\n FROM TEMP_DATETIME\n ORDER BY datetime;\n* ../sql/inet.sql\tThu Oct 29 13:13:03 1998\n--- inet.sql\tMon Nov 2 09:36:06 1998\n*************\n* 15,20 **\n--- 15,21 ----\n INSERT INTO INET_TBL (c, i) VALUES ('10', '10.1.2.3/8');\n INSERT INTO INET_TBL (c, i) VALUES ('10', '11.1.2.3/8');\n INSERT INTO INET_TBL (c, i) VALUES ('10', '9.1.2.3/8');\n+ INSERT INTO INET_TBL (c, i) VALUES (null, null);\n \n SELECT '' AS ten, c AS cidr, i AS inet FROM INET_TBL;\n \n* ../sql/lseg.sql\tTue Jul 29 12:22:46 1997\n--- lseg.sql\tMon Nov 2 09:36:46 1998\n*************\n* 10,15 **\n--- 10,16 ----\n INSERT INTO LSEG_TBL VALUES ('10,-10 ,-3,-4');\n INSERT INTO LSEG_TBL VALUES ('[-1e6,2e2,3e5, -4e1]');\n INSERT INTO LSEG_TBL VALUES ('(11,22,33,44)');\n+ INSERT INTO LSEG_TBL VALUES (null);\n \n -- bad values for parser testing\n INSERT INTO LSEG_TBL VALUES ('(3asdf,2 ,3,4r2)');\n* ../sql/name.sql\tMon Apr 27 09:50:03 1998\n--- name.sql\tMon Nov 2 09:37:57 1998\n*************\n* 28,33 ***\n--- 28,35 ----\n \n INSERT INTO NAME_TBL(f1) VALUES ('1234567890ABCDEFGHIJKLMNOPQRSTUVWXYZ');\n \n+ INSERT INTO NAME_TBL(f1) VALUES (null);\n+ \n \n SELECT '' AS seven, NAME_TBL.;\n \n* ../sql/path.sql\tTue Jun 3 10:23:37 1997\n--- path.sql\tMon Nov 2 09:39:36 1998\n***********\n* 22,27 **\n--- 22,29 ----\n \n INSERT INTO PATH_TBL VALUES ('(11,12,13,14)');\n \n+ INSERT INTO PATH_TBL VALUES (null);\n+ \n -- bad values for parser testing\n INSERT INTO PATH_TBL VALUES ('[(,2),(3,4)]');\n \n* ../sql/point.sql\tWed Sep 24 13:55:38 1997\n--- point.sql\tMon Nov 2 09:40:49 1998\n*************\n* 12,17 **\n--- 12,19 ----\n \n INSERT INTO POINT_TBL(f1) VALUES ('(-5.0,-12.0)');\n \n+ INSERT INTO POINT_TBL(f1) VALUES (null);\n+ \n -- bad format points \n INSERT INTO POINT_TBL(f1) VALUES ('asdfasdf');\n \n* ../sql/polygon.sql\tTue Jul 29 12:22:48 1997\n--- polygon.sql\tMon Nov 2 09:41:50 1998\n*************\n* 25,30 **\n--- 25,32 ----\n \n INSERT INTO POLYGON_TBL(f1) VALUES ('(0.0,1.0),(0.0,1.0)');\n \n+ INSERT INTO POLYGON_TBL(f1) VALUES (null);\n+ \n -- bad polygon input strings \n INSERT INTO POLYGON_TBL(f1) VALUES ('0.0');\n \n* ../sql/reltime.sql\tThu May 8 23:26:51 1997\n--- reltime.sql\tMon Nov 2 09:43:27 1998\n*************\n* 12,17 **\n--- 12,19 ----\n \n INSERT INTO RELTIME_TBL (f1) VALUES ('@ 14 seconds ago');\n \n+ INSERT INTO RELTIME_TBL (f1) VALUES (null);\n+ \n \n -- badly formatted reltimes: \n INSERT INTO RELTIME_TBL (f1) VALUES ('badly formatted reltime');\n* ../sql/timespan.sql\tThu May 8 23:26:56 1997\n--- timespan.sql\tMon Nov 2 09:46:19 1998\n*************\n* 10,15 **\n--- 10,16 ----\n INSERT INTO TIMESPAN_TBL (f1) VALUES ('6 years');\n INSERT INTO TIMESPAN_TBL (f1) VALUES ('5 months');\n INSERT INTO TIMESPAN_TBL (f1) VALUES ('5 months 12 hours');\n+ INSERT INTO TIMESPAN_TBL (f1) VALUES (null);\n \n -- badly formatted timespan: \n INSERT INTO TIMESPAN_TBL (f1) VALUES ('badly formatted timespan');\n* ../sql/tinterval.sql\tSat Sep 20 12:33:24 1997\n--- tinterval.sql\tMon Nov 2 09:47:23 1998\n*************\n* 15,20 ****\n--- 15,22 ----\n INSERT INTO TINTERVAL_TBL (f1)\n VALUES ('[\"Feb 15 1990 12:15:03\" \"current\"]');\n \n+ INSERT INTO TINTERVAL_TBL (f1) VALUES (null);\n+ \n \n -- badly formatted tintervals \n INSERT INTO TINTERVAL_TBL (f1)\n\n\n> Anyway, I would suggest that, until we plan out what new behaviors we\n> really want in the system that we don't submit a patch which breaks many\n> of the data types. That kind of change should, imho, be done offline and\n> committed as an integrated solution. If you'd like, I'd be happy to\n> trade patches with you as we work out the issues and once we (and anyone\n> else with an interest) are happy with the new behavior and have\n> regression tests which pass then we should commit to the tree.\n\nSounds good to me but why not discuss it on hackers?\n\n> As you mentioned, the date and time functions check for null inputs and\n> behave pretty well, and an interim solution might include using similar\n> techniques on the inet/cidr types.\n\nThat's what my patch that I haven't submitted does.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Thu, 5 Nov 1998 06:24:08 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd..." }, { "msg_contents": "> > but there was no patch enclosed or following.\n> My copy of the message has the patch by here it is again.\n\nOh! I have that patch. I am gathering from the rest of your message that\nthere are two patches being discussed, but had expected a patch having\nto do with source code fixups in that message. Sorry I am befuddled and\nconfused...\n\n> Sounds good to me but why not discuss it on hackers?\n\nSure.\n\n> > ... an interim solution might include using similar\n> > techniques on the inet/cidr types.\n> That's what my patch that I haven't submitted does.\n\nOK, sorry, I was confused about which patches do what. How about\nsubmitting your inet/cidr NULL fixup patch for both v6.4.1 and the\ndevelopment tree? \n\nThen I would propose that we renew this discussion on general\nimprovements in a short while (~ 3 weeks?) to give v6.4 time to settle\nin and me/us time to recover. Pretty tapped out from the last month of\nPostgres. But scrubbing the NULL mechanisms is on my list of interests\nfor v6.5, so would like us to have a solid couple of months to work on\nit, and it sounds like Jan and others will want to contribute.\n\n - Tom\n", "msg_date": "Thu, 05 Nov 1998 14:53:40 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd..." }, { "msg_contents": "The Hermit Hacker <[email protected]> writes:\n> On Wed, 4 Nov 1998, Tom Lane wrote:\n>> I did a fresh 'cvs checkout' just now to make sure I had a good set\n>> of files, and was dismayed to discover that a whole bunch of dead\n>> subdirectories have reappeared in the CVS tree.\n\n> \tDid you use the -P option? when you checkout, you should do:\n> \tcvs checkout -P pgsql\n> \tIt gets rid of the 'dead' directories tha way...\n\nOK, but I hadn't been doing that before ... ah, wait, I see it.\nIn my ~/.cvsrc file I have\n\tcvs -z3\n\tupdate -d -P\nwhich means that cvs update gets the -P flag automatically. So my\nold tree didn't have the deadwood because I'd run update against it.\n\nGuess I should add \"checkout -P\" to .cvsrc.\n\n\t\t\tthanks, tom lane\n", "msg_date": "Thu, 05 Nov 1998 10:13:42 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd... " }, { "msg_contents": "Thus spake Thomas G. Lockhart\n> OK, sorry, I was confused about which patches do what. How about\n> submitting your inet/cidr NULL fixup patch for both v6.4.1 and the\n> development tree? \n\nI just submitted it to patches. I already forget how to submit it\nto 6.4.1.\n\n> Then I would propose that we renew this discussion on general\n> improvements in a short while (~ 3 weeks?) to give v6.4 time to settle\n> in and me/us time to recover. Pretty tapped out from the last month of\n> Postgres. But scrubbing the NULL mechanisms is on my list of interests\n> for v6.5, so would like us to have a solid couple of months to work on\n> it, and it sounds like Jan and others will want to contribute.\n\nOK. I hope we remember to go and clean out the function handlers though.\nNo need for the code bloat if they will never have to handle NULLs.\nPerhaps they should be asserts after that for a while.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Fri, 6 Nov 1998 07:09:00 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Open the flood gates...v6.4 is tag'd..." } ]
[ { "msg_contents": ">> By the way, I thought the file name of the v6.4 would be\n>> \"postgresql-6.4.tar.gz\" not \"postgresql.v6.4.tar.gz\" since previous\n>> versions did not have a name such as \"postgresql.v.\".\n>> Have you changed naming policy?\n>\n>\tI never think about it when I do it...I think I named it v6.3.2\n>last time too, but ended up renaming it for some reason :(\n>\n>\tJust curious, but is there a reason why it matters? If there is,\n>I\"ll gladly change it back again...I'm quite easy to get along with as far\n>as taht is concered :)\n\nNo, I just felt that the file name is unexpected.\n\nHere are file names since 6.1 (I don't remember about 6.0):\n\nppostgresql-v6.1.tar.gz\npostgresql-v6.1.1.tar.gz\npostgresql-6.2.tar.gz\npostgresql-6.2.1.tar.gz\npostgresql-6.3.tar.gz\npostgresql-6.3.1.tar.gz\npostgresql-6.3.2.tar.gz\n\nSo seems that \"postgresql-v6\" had been used only for 6.1 series.\n--\nTatsuo Ishii\[email protected]\n", "msg_date": "Thu, 05 Nov 1998 13:26:01 +0900", "msg_from": "Tatsuo Ishii <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [ANNOUNCE] November 4th, 1998: PostgreSQL v6.4 Released " }, { "msg_contents": "\nSince I published it as v6.4, I just created a symlink to the other name,\njust in case maybe someone is relying on it staying the same (ie. 6.4 vs\nv6.4)...\n\nI advertised it as v6.4, and to avoid having a slew of questions stating\nppl can't find it... :(\n\nOn Thu, 5 Nov 1998, Tatsuo Ishii wrote:\n\n> >> By the way, I thought the file name of the v6.4 would be\n> >> \"postgresql-6.4.tar.gz\" not \"postgresql.v6.4.tar.gz\" since previous\n> >> versions did not have a name such as \"postgresql.v.\".\n> >> Have you changed naming policy?\n> >\n> >\tI never think about it when I do it...I think I named it v6.3.2\n> >last time too, but ended up renaming it for some reason :(\n> >\n> >\tJust curious, but is there a reason why it matters? If there is,\n> >I\"ll gladly change it back again...I'm quite easy to get along with as far\n> >as taht is concered :)\n> \n> No, I just felt that the file name is unexpected.\n> \n> Here are file names since 6.1 (I don't remember about 6.0):\n> \n> ppostgresql-v6.1.tar.gz\n> postgresql-v6.1.1.tar.gz\n> postgresql-6.2.tar.gz\n> postgresql-6.2.1.tar.gz\n> postgresql-6.3.tar.gz\n> postgresql-6.3.1.tar.gz\n> postgresql-6.3.2.tar.gz\n> \n> So seems that \"postgresql-v6\" had been used only for 6.1 series.\n> --\n> Tatsuo Ishii\n> [email protected]\n> \n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 5 Nov 1998 00:33:03 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [ANNOUNCE] November 4th, 1998: PostgreSQL v6.4 Released " } ]
[ { "msg_contents": "Hi all\n\nI am running an older snapshoot of 6.4, so this may not be relavent now.\n\nI did a 'pg_dumpall -z > file.dump', then on another box destroyed all the\ndatabases and did 'psql templet1 < file.dump' and got this:\n\nREVOKE ALL on 'board' from PUBLIC;\nERROR: parser: parse error at or near \"'\"\nGRANT ALL on \"board\" to \"nobody\";\nCHANGE\n\nIt would seem that the single quotes should be double quotes?\nJan, is this your area?\n\nI have notice that most all names now are quoted in dumps, but will work\njust fine with out any quotes, so why the quotes?\n\nThanks, and have a great day\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Thu, 5 Nov 1998 09:26:50 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": true, "msg_subject": "quoting problem?" }, { "msg_contents": "> REVOKE ALL on 'board' from PUBLIC;\n> ERROR: parser: parse error at or near \"'\"\n> GRANT ALL on \"board\" to \"nobody\";\n> CHANGE\n> It would seem that the single quotes should be double quotes?\n\nIt looks like that is now correct in the v6.4beta and v6.4final trees.\nThe program uses a single function, FmtId(), to evaluate strings and to\ndo the quote/noquote thang.\n\n> I have notice that most all names now are quoted in dumps, but will \n> work just fine with out any quotes, so why the quotes?\n\nThe primary reason that is that you are allowed on table creation to use\nreserved keywords if you surround them with double quotes. pg_dump\ndoesn't know about reserved words, so would not be able to selectively\napply double quotes.\n\nYou can suppress most of the double quotes by using the new \"-n\"\nargument to pg_dump (n == no quotes). This gives you essentially the old\nbehavior of applying double quotes only if there is upper case or\nwhitespace in the identifier.\n\n - Tom\n", "msg_date": "Thu, 05 Nov 1998 15:07:07 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] quoting problem?" }, { "msg_contents": ">\n> Hi all\n>\n> I am running an older snapshoot of 6.4, so this may not be relavent now.\n>\n> I did a 'pg_dumpall -z > file.dump', then on another box destroyed all the\n> databases and did 'psql templet1 < file.dump' and got this:\n>\n> REVOKE ALL on 'board' from PUBLIC;\n> ERROR: parser: parse error at or near \"'\"\n> GRANT ALL on \"board\" to \"nobody\";\n> CHANGE\n>\n> It would seem that the single quotes should be double quotes?\n> Jan, is this your area?\n\n Not my area. But I'll take care that it get's fixed in\n v6.4.1.\n\n>\n> I have notice that most all names now are quoted in dumps, but will work\n> just fine with out any quotes, so why the quotes?\n\n Just for the case that some of the identifiers could be\n reserved words or contain upper case letters. By default\n (without quotes), the parser changes anything to lower case\n and does keyword lookup.\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, 5 Nov 1998 18:24:33 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] quoting problem?" } ]
[ { "msg_contents": "Hi all,\n\nI'm trying to guess the size of int8...\n\nhygea=> insert into b values (9223372036854775296);\nNOTICE: Integer input '9223372036854775296' is out of range; promoted to float\nERROR: Floating point conversion to int64 is out of range\nhygea=> select * from b;\n i8\n-------------------\n9223372036854775807\n9223372036854774784\n(2 rows)\n\nhygea=> update b set i8=i8+1;\nUPDATE 2\nhygea=> select * from b;\n i8\n--------------------\n-9223372036854775808\n 9223372036854774785\n(2 rows)\n\n\n-Jose'-\n\n\n", "msg_date": "Thu, 5 Nov 1998 16:11:35 +0100", "msg_from": "Sferacarta Software <[email protected]>", "msg_from_op": true, "msg_subject": "int8 size" }, { "msg_contents": "> I'm trying to guess the size of int8...\n> \n> hygea=> insert into b values (9223372036854775296);\n> NOTICE: Integer input '9223372036854775296' is out of range; promoted to float\n> ERROR: Floating point conversion to int64 is out of range\n\nTry\n insert into b values ('9223372036854775296'::int8);\n\nOtherwise the Postgres parser tries to read it as a number first, can't\ndo it as an int4, so forces it to float8, and then fails to convert back\nto int64. The User's Guide talks about int8 and mentions that it allows\nmore than 18 decimal places, but isn't as specific as you might like.\nThe actual limit is probably +/- (2^63)-1\n\nHmm. On my linux/libc5 box, the number seems to peg at the top no matter\nhow big the input:\n\ntgl=> select '9223372036854775296'::int8;\n-------------------\n9223372036854775296\n\ntgl=> select '10000000000000000000'::int8;\n-------------------\n9223372036854775807\n\ntgl=> select '100000000000000000000'::int8;\n-------------------\n9223372036854775807\n\nNot so good...\n\n - Tom\n", "msg_date": "Thu, 05 Nov 1998 16:25:18 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] int8 size" } ]
[ { "msg_contents": "Hi all\n\nA little off topic, but has/is any one working on a new power by logo for\n6.4?\n\nIf not, then I can give it a go, I'm starting to understand Gimp.)\n\nThanks\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Thu, 5 Nov 1998 13:07:04 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": true, "msg_subject": "6.4 Power by Logo?" }, { "msg_contents": "On Thu, 5 Nov 1998, Terry Mackintosh wrote:\n\n> Hi all\n> \n> A little off topic, but has/is any one working on a new power by logo for\n> 6.4?\n> \n> If not, then I can give it a go, I'm starting to understand Gimp.)\n\nGo for it...if you want to come up with a new and improve logo for the web\nsite, go for it too...\n\nHell...I need a good CD label for the CDs...wanna try that? :) I have\none now, but its...weak :)\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 5 Nov 1998 22:48:16 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] 6.4 Power by Logo?" }, { "msg_contents": "Hi all\n\nOK, I'll give it a go and see what I can come up with, any ideas are\nwelcome, please email directly to me.\n\nHave a great day\nTerry\n\nOn Thu, 5 Nov 1998, The Hermit Hacker wrote:\n\n> On Thu, 5 Nov 1998, Terry Mackintosh wrote:\n> \n> > Hi all\n> > \n> > A little off topic, but has/is any one working on a new power by logo for\n> > 6.4?\n> > \n> > If not, then I can give it a go, I'm starting to understand Gimp.)\n> \n> Go for it...if you want to come up with a new and improve logo for the web\n> site, go for it too...\n> \n> Hell...I need a good CD label for the CDs...wanna try that? :) I have\n> one now, but its...weak :)\n> \n> Marc G. Fournier \n> Systems Administrator @ hub.org \n> primary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n> \n\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Thu, 5 Nov 1998 22:44:14 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] 6.4 Power by Logo?" } ]
[ { "msg_contents": "Dear Y'all,\n\nDo you know why nothing shows up when\nI try to fill in the \"Open database\" popup?\nLooks like I might be typing white on white.\n\nSun ULTRA 60, Solaris 2.6, CDE\n\nDavid McDavid\nLimber Observatory\nPipe Creek TX\nwww.limber.org\n\n", "msg_date": "Tue, 5 Nov 2030 12:36:35 -0600 (CST)", "msg_from": "David McDavid <[email protected]>", "msg_from_op": true, "msg_subject": "pgaccess disappearing ink" } ]
[ { "msg_contents": "> > It appears that when a function is called if any of the paramaters\n> are\n> > NULL all of the parameters are NULL.\n> > try:\n> > drop function nvl(int, int);\n> > create function nvl(int, int) returns boolean as '\n> > declare\n> > nonullo alias as $1;\n> > nullo alias as $2;\n> > begin\n> > return (nonullo IS NULL) AND (nullo IS NULL);\n> > end;' language 'plpgsql';\n> > select nvl(i,0) from a;\n> > you should get:\n> > nvl\n> > ---\n> > t\n> > t\n> > f\n> > t\n> > (4 rows)\n> \n> Don't blame PL/pgSQL for that. There is only one bool isNull\n> pointer given to PL handlers. How should the PL handler know,\n> which of the arguments are null then? As I said on another\n> thread, the function call interface needs to get redesigned.\nWell, Jan, don't get sensitive. I love PL/pgSQL. And I had no illusions\nthat it was your HANDLER causing the problem. I feel that a function\ncall interface redesign is also needed. But, I do have a quick\nquestion, why does it matter which one is NULL if you can still obtain\nthe parameters in the order they were passed why would one become NULL\nthat wasn't before? I'm asking totally from ignorance here.\n\t-DEJ\n\n> Jan\n> \n", "msg_date": "Thu, 5 Nov 1998 13:50:35 -0600", "msg_from": "\"Jackson, DeJuan\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] PL/pgSQL a great procedural language for PostgreSQL" }, { "msg_contents": "> > Don't blame PL/pgSQL for that. There is only one bool isNull\n> > pointer given to PL handlers. How should the PL handler know,\n> > which of the arguments are null then? As I said on another\n> > thread, the function call interface needs to get redesigned.\n> Well, Jan, don't get sensitive. I love PL/pgSQL. And I had no illusions\n> that it was your HANDLER causing the problem. I feel that a function\n> call interface redesign is also needed. But, I do have a quick\n> question, why does it matter which one is NULL if you can still obtain\n> the parameters in the order they were passed why would one become NULL\n> that wasn't before? I'm asking totally from ignorance here.\n\n It might be possible, that even if isNull is true to look at\n the actual arguments given to the PL handler. For datatypes\n passed by reference, a NULL value has to get passed as null\n pointer. I'm not 100% sure if that is really true in all\n cases where PL functions can get called, and we all know what\n happens when accessing a pointer that points to something\n else than a memory location. For arguments passed by value it\n is totally impossible to know if it's a NULL by looking at\n the value itself.\n\n Summary is, that the PL handler cannot be sure which of the\n arguments the function caller meant when calling with isNull\n = TRUE. And I decided for now to be safe and assume he meant\n all.\n\n When accessing data from a specific table, it is possible to\n call the function with a complex type. This time, the\n PL/pgSQL function gets the complete tuple and can look at the\n information there which attributes are NULLs.\n\n CREATE TABLE a (k integer, i integer);\n\n CREATE FUNCTION a_i_checknull(a) RETURNS bool AS '\n DECLARE\n row_a ALIAS FOR $1; -- The dot-notation $1.i does not work!\n BEGIN\n IF row_a.k ISNULL THEN\n RAISE NOTICE ''attribute k has NULL value'';\n END IF;\n IF row_a.i ISNULL THEN\n RAISE NOTICE ''attribute i has NULL value'';\n END IF;\n IF row_a.k ISNULL OR row_a.i ISNULL THEN\n RETURN ''t'';\n END IF;\n RETURN ''f'';\n END;\n ' LANGUAGE 'plpgsql';\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, 5 Nov 1998 21:31:02 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/pgSQL a great procedural language for PostgreSQL" }, { "msg_contents": "> For arguments passed by value it\n> is totally impossible to know if it's a NULL by looking at\n> the value itself.\n\nIs this a sufficient reason to move toward having all user-oriented data\ntypes be pass-by-reference? If we don't do that, then we would need to\npass a null flag for every parameter, or an array of flags, or have some\nglobal array which contains the null flags (which we might get away with\nsince Postgres is pretty much single threaded and non-reentrant in the\nbackend code). What other options might there be?\n\n - Tom\n", "msg_date": "Fri, 06 Nov 1998 07:08:28 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/pgSQL a great procedural language for PostgreSQL" }, { "msg_contents": "Thomas G. Lockhart wrote:\n> \n> > For arguments passed by value it\n> > is totally impossible to know if it's a NULL by looking at\n> > the value itself.\n> \n> Is this a sufficient reason to move toward having all user-oriented data\n> types be pass-by-reference? If we don't do that, then we would need to\n> pass a null flag for every parameter, or an array of flags, or have some\n> global array which contains the null flags (which we might get away with\n> since Postgres is pretty much single threaded and non-reentrant in the\n> backend code). What other options might there be?\n\nPerhaps make the isNull flag a bitmap ?\n\nIn case it is an int (I haven't looked), we get room for flagging 32\narguments\n\nAlso I hope that most current code will run unchanged, in case it tests\nfor \n0 / not 0\n\n----------------\nHannu\n", "msg_date": "Fri, 06 Nov 1998 11:20:23 +0200", "msg_from": "Hannu Krosing <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/pgSQL a great procedural language for PostgreSQL" }, { "msg_contents": "\nhave anyone tested 6.4 on a libc6 based linux (for example debian hamm) ?\n\nbecause I tested it on an older system (libc5 based) and it's working\nwell.\n\nbut on libc6 it's compiled without errors, \nI can make the initdb, and start the postmaster,\nbut psql -l for example not working\nevery postgres backend exits after postmaster forked..\n\nany idea?\n\t\tredax\n\n.----------------------------------------------------------.\n\|Zsolt Varga \| tel/fax: +36 36 422811 \|\n\| AgriaComputer LTD \| email: [email protected] \|\n\| System Administrator \| URL: http://www.agria.hu/ \|\n`----------------------------------------------------------'\n\n", "msg_date": "Fri, 6 Nov 1998 10:33:48 +0100 (CET)", "msg_from": "Zsolt Varga <[email protected]>", "msg_from_op": false, "msg_subject": "linux libc6 & pgsql 6.4" }, { "msg_contents": "Zsolt Varga wrote:\n> \n> have anyone tested 6.4 on a libc6 based linux (for example debian hamm) ?\n> \n> because I tested it on an older system (libc5 based) and it's working\n> well.\n> \n> but on libc6 it's compiled without errors,\n> I can make the initdb, and start the postmaster,\n> but psql -l for example not working\n> every postgres backend exits after postmaster forked..\n> \n> any idea?\n\nDoes createdb work ?\n\n-----------\nHannu\n", "msg_date": "Fri, 06 Nov 1998 15:45:54 +0200", "msg_from": "Hannu Krosing <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] linux libc6 & pgsql 6.4" }, { "msg_contents": "> Perhaps make the isNull flag a bitmap ?\n> In case it is an int (I haven't looked), we get room for flagging 32\n> arguments\n\nHmm. Well, how about if we look at what we would prefer if we were\nallowed to start from the beginning, and then consider this as an\nalternative. Doing this with a bitmap might be a bit arcane.\n\n> Also I hope that most current code will run unchanged, in case it \n> tests for 0 / not 0\n\nGood point.\n\n - Thomas\n", "msg_date": "Fri, 06 Nov 1998 15:01:07 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/pgSQL a great procedural language for PostgreSQL" }, { "msg_contents": "[Since my work email didn't go through on the list here's a short answer\nagain. ]\n\nOn Fri, Nov 06, 1998 at 10:33:48AM +0100, Zsolt Varga wrote:\n> have anyone tested 6.4 on a libc6 based linux (for example debian hamm) ?\n\nYes, me. That is my debian system is newer than hamm.\n\n> but on libc6 it's compiled without errors, \n> I can make the initdb, and start the postmaster,\n> but psql -l for example not working\n> every postgres backend exits after postmaster forked..\n\nWorks fine for me. No problem at all.\n\nMichael\n-- \nDr. Michael Meskes, Leiter Niederlassung West, Datenrevision GmbH\nbusiness: Cuxhavener Str. 36, D-21149 Hamburg\n private: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Fri, 6 Nov 1998 20:48:09 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] linux libc6 & pgsql 6.4" }, { "msg_contents": "Hi all\n\nOn Fri, 6 Nov 1998, Michael Meskes wrote:\n\n> On Fri, Nov 06, 1998 at 10:33:48AM +0100, Zsolt Varga wrote:\n> > have anyone tested 6.4 on a libc6 based linux (for example debian hamm) ?\n> \n> Yes, me. That is my debian system is newer than hamm.\n> \n> > but on libc6 it's compiled without errors, \n> > I can make the initdb, and start the postmaster,\n> > but psql -l for example not working\n> > every postgres backend exits after postmaster forked..\n> \n> Works fine for me. No problem at all.\n\nThis may or may not be related to your problem, but...\nOn my server I did only the min. install of Red Hat 4.2 that I needed to\nrun a server, then custom built most of the stuff such as PHP, Apache,\nPostgreSQL ... etc. I also change the permissions on many directories and\nfiles.\n\nAnd now I find that a normal user can not do some things that involve\n'fork()'. But all works fine as root, so it is a permission problem,\nwhich I still have not pinned down, not even shure how to find it.:(\n\nExample: Pine/ispell or Pine/vim\nAny attemp to run an external program from pine, such as ispell, will not\nwork, pine will say that the program returned 255. So I did some debuging\nstuff and found the problem happens at the fork() call.\n\nDoes this sound like your problem?\nIf so, then start looking very closely at permissions.\n\nHave a great day, and hope that helped.\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Sat, 7 Nov 1998 08:37:51 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] linux libc6 & pgsql 6.4" }, { "msg_contents": "On Sat, Nov 07, 1998 at 08:37:51AM -0500, Terry Mackintosh wrote:\n> > > but on libc6 it's compiled without errors, \n> > > I can make the initdb, and start the postmaster,\n> > > but psql -l for example not working\n> > > every postgres backend exits after postmaster forked..\n> > \n> > Works fine for me. No problem at all.\n> \n> This may or may not be related to your problem, but...\n\nEhem, I do not have that problem. I was the one answering that my\ninstallation is okay. :-9\n\n> And now I find that a normal user can not do some things that involve\n> 'fork()'. But all works fine as root, so it is a permission problem,\n> which I still have not pinned down, not even shure how to find it.:(\n\nI wonder which permissions have to do with fork().\n\nMichael\n-- \nDr. Michael Meskes, Leiter Niederlassung West, Datenrevision GmbH\nbusiness: Cuxhavener Str. 36, D-21149 Hamburg\n private: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Sun, 8 Nov 1998 12:16:40 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] linux libc6 & pgsql 6.4" }, { "msg_contents": "On Fri, 6 Nov 1998, Hannu Krosing wrote:\n\n\|Does createdb work ?\nyeah worked, \nsomebody sent me the solution from the list,\nI have to remove termcap-compat from debian/hamm\nand recompile... after it works well\n\ngreat work, it's 3-4 times faster than 6.3.2 was\n\n\tthanks\t\n\t\tredax\n\n.----------------------------------------------------------.\n\|Zsolt Varga \| tel/fax: +36 36 422811 \|\n\| AgriaComputer LTD \| email: [email protected] \|\n\| System Administrator \| URL: http://www.agria.hu/ \|\n`----------------------------------------------------------'\n\n", "msg_date": "Wed, 11 Nov 1998 10:37:54 +0100 (CET)", "msg_from": "Zsolt Varga <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] linux libc6 & pgsql 6.4" } ]
[ { "msg_contents": "> > Is there any reason for not use these functions on SUM() and AVG()\n> on\n> > official release ?\n> \n> It sounds like a good idea. The only hesitation I have at the moment\n> is\n> that not all platforms have int8 support, and I'm not certain which\n> these are. Also, accumulating int4 into int8 is probably pretty slow\n> since on 32-bit machines the \"long long\" is usually done in a s/w\n> library, not in machine code.\n> \n> float8 might be a better choice for accumulating AVG(), but I'm\n> worried\n> about incorrect results with large tables (> 1M entries) which have\n> pathological distributions of numbers (e.g. 1M entries with MAXINT and\n> 1M entries with zero). int4 gives ~9.2 decimal places, float8 gives\n> ~15\n> decimal places, so there is only about ~6 decimal places of headroom.\n> \n> Of course, why am I worried? That is much better than what we have\n> currently. And someone reported that at least one commercial system\n> (Sybase?) returns float8 for avg() (and sum()?) as I recall.\n> \n> So, your suggestion is that for AVG() at least we return something\n> other\n> than the input type; how about returning float8 for any input type?\n> Don't know if SUM() could/should behave similarly...\n> \n> - Tom\n> \nI think the issue could be address if a float8 sum would be affected by\nan ORDER BY. If so just make a not in the DOCS and FAQ about\nsignificant digits in SUM and AVG. And maybe an example to get the most\nexact SUM and AVG from a table. If SUM and AVG aren't affected by an\nORDER BY I'd say stick with the highest range integer type implemented\nby the system (in hardware if possible).\n\tJust my $0.02,\n\t-DEJ\n", "msg_date": "Thu, 5 Nov 1998 14:04:37 -0600", "msg_from": "\"Jackson, DeJuan\" <[email protected]>", "msg_from_op": true, "msg_subject": "RE: [HACKERS] Re: bug on aggregate function AVG()" } ]
[ { "msg_contents": "Apparently the \"INV_ARCHIVE\" define went away somewhere between 6.3.2 and 6.4,\nbut this was not reflected in the Python interface.\n\nI had to do this:\n\n #ifdef PRE_PG_6_4\n PyDict_SetItemString(dict, \"INV_ARCHIVE\", PyInt_FromLong(INV_ARCHIVE));\n #endif\n\nin order to get the Python interface to compile.\n\n\t-Michael Robinson\n\n", "msg_date": "Fri, 6 Nov 1998 07:53:12 +0800 (GMT)", "msg_from": "Michael Robinson <[email protected]>", "msg_from_op": true, "msg_subject": "Python interface broken in 6.4" } ]
[ { "msg_contents": "I am planning to enhance PgAccess in order to handle plpgsql functions\n(to define, delete , inspect and modify them).\n\nIn this respect, I would like to know :\n\n1. what is the value of prolang field in pg_proc table for plpgsql\nfunctions (on my system is 43009, but it this value stable?)\n\n2. are plpgsql functions interpreted at execution time ? If not (just\nfor the record) is this a big performance penalty ? Should I reconsider\nwriting them in \"C\" for triggers for example , or for the most often\ncalled functions ?\n\n3. in order to modify a function's source, to change to field procsrc in\npg_proc table ? Or should I \"drop function ...\" and \"create function\n...\" again ?\n\n\nPlease CC: me directly to [email protected]\n\nThanks in advance,\nConstantin Teodorescu\nFLEX Consulting Braila, ROMANIA\n", "msg_date": "Fri, 06 Nov 1998 07:32:10 +0000", "msg_from": "Constantin Teodorescu <[email protected]>", "msg_from_op": true, "msg_subject": "A plpgsql question !" } ]
[ { "msg_contents": "Hello!\n\n I found duplicate entries on http://news.freshmeat.net/readmore?f=postgresql-6.4,\nhaven't checked other sources...\n\n\n In section Enhancements:\n\n\nNew backend programming language PL/pgSQL in backend/pl(Jan)\n and\nPL/pgSQL backend programming language(Jan)\n\n\nNew 8-byte integer type, checked by configure for OS support(Thomas)\n and\nCode for 64-bit integer supported added, configure tested, int8 type(Thomas)\n\n\nLibpq can now be compiled on win32(Magnus)\n and\nPsql and libpq now compile under win32 using win32.mak(Magnus)\n\n\nOleg.\n---- \n Oleg Broytmann http://members.tripod.com/~phd2/ [email protected]\n Programmers don't die, they just GOSUB without RETURN.\n\n", "msg_date": "Fri, 6 Nov 1998 12:17:45 +0300 (MSK)", "msg_from": "Oleg Broytmann <[email protected]>", "msg_from_op": true, "msg_subject": "v6.4 - duplicate entries in announce" }, { "msg_contents": "I have made these changes. Thanks.\n\n> Hello!\n> \n> I found duplicate entries on http://news.freshmeat.net/readmore?f=postgresql-6.4,\n> haven't checked other sources...\n> \n> \n> In section Enhancements:\n> \n> \n> New backend programming language PL/pgSQL in backend/pl(Jan)\n> and\n> PL/pgSQL backend programming language(Jan)\n> \n> \n> New 8-byte integer type, checked by configure for OS support(Thomas)\n> and\n> Code for 64-bit integer supported added, configure tested, int8 type(Thomas)\n> \n> \n> Libpq can now be compiled on win32(Magnus)\n> and\n> Psql and libpq now compile under win32 using win32.mak(Magnus)\n> \n> \n> Oleg.\n> ---- \n> Oleg Broytmann http://members.tripod.com/~phd2/ [email protected]\n> Programmers don't die, they just GOSUB without RETURN.\n> \n> \n> \n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 11 Dec 1998 22:56:53 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] v6.4 - duplicate entries in announce" } ]
[ { "msg_contents": "\n\n\n\n\n\n>have anyone tested 6.4 on a libc6 based linux (for example debian hamm) ?\n\nI always run the latest PostgreSQL snapshot on my debian machine that I try\nto keep up-to-date with unstable. That is I run a potato distribution now.\nAnd PostgreSQL works fine for me.\n\n>but psql -l for example not working\n>every postgres backend exits after postmaster forked..\nNever tried this. But will of course.\n\nMichael\n\nDr. Michael Meskes, Senior-Consultant\nMummert+Partner Unternehmensberatung AG\nTel.: +49211 826 4616\n\n\n", "msg_date": "Fri, 6 Nov 1998 10:57:19 +0100", "msg_from": "mummert&[email protected]", "msg_from_op": true, "msg_subject": "Antwort: [HACKERS] linux libc6 & pgsql 6.4" } ]
[ { "msg_contents": "I have discovered today an interesting bug , when creating table name\ncontaining spaces and one of the fields are auto_increment-able.\nSee the example below :\n\n\nshowroom=> create sequence student_id;\nCREATE\nshowroom=> create table students (id int4 default\nnextval('student_id'),name text);\nCREATE\nshowroom=> drop table students;\nDROP\nshowroom=> create table \"my students\" (id int4 default\nnextval('student_id'), name text);\nERROR: my: Table does not exist.\n\nPlease CC: me directly the answer.\n\n\nConstantin Teodorescu\nFLEX Consulting Braila, ROMANIA\n", "msg_date": "Fri, 06 Nov 1998 14:17:47 +0000", "msg_from": "Constantin Teodorescu <[email protected]>", "msg_from_op": true, "msg_subject": "Bug when special table names meet autoincrement field !!!!" } ]
[ { "msg_contents": "Hi,\n\nWhere could I find the information about setting the amount\nof shared memory on various operating systems ?\n\nIs it available somewhere in PG docs ?\n\nMy interest is in Linux and Solaris.\n\n----------------\nHannu\n", "msg_date": "Fri, 06 Nov 1998 19:43:35 +0200", "msg_from": "Hannu Krosing <[email protected]>", "msg_from_op": true, "msg_subject": "Doc's about setting the amount of shared memory " }, { "msg_contents": "Hi!\n\nOn Fri, 6 Nov 1998, Hannu Krosing wrote:\n> Where could I find the information about setting the amount\n> of shared memory on various operating systems ?\n> \n> My interest is in Linux and Solaris.\n\n Instruction for Solaris.\n\n 1. Read man -s 4 system\n\n 2. Note - /etc/system is kernel config file that have been read once on\nevery boot.\n\n 3. Log in as root\n\n 4. MAKE BACKUP of your /etc/system\n\n 5. Edit /etc/system. Add or modify:\nset shmsys:shminfo_shmmin=100\nset shmsys:shminfo_shmmni=500\nset shmsys:shminfo_shmseg=100\nset shmsys:shminfo_shmmax=2147483648\n\n 6. Save the file\n\n 7. Reboot.\n\n 8 In case of boot failure, use backup file (more info in man -s 4 system).\n\nOleg.\n---- \n Oleg Broytmann National Research Surgery Centre http://sun.med.ru/~phd/\n Programmers don't die, they just GOSUB without RETURN.\n\n", "msg_date": "Fri, 6 Nov 1998 20:52:38 +0300 (MSK)", "msg_from": "Oleg Broytmann <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Doc's about setting the amount of shared memory " } ]
[ { "msg_contents": "Hi all\n\nI started with the contrib/spi/.c stuff and am working toward a function\nto be used by a trigger to update a datetime field upon doing an UPDATE to\na record.\n\nIt is about ready to test, except one thing :)\n\nIs there a function like GetCurrentAbsoluteTime() that will get the\ncurrent time in a form suitable for inserting into a datetime field?\n\nGetCurrentDatetimeTime() maybe???\nor a convertion function?\n\nThanks\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Fri, 6 Nov 1998 14:39:30 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": true, "msg_subject": "SPI hacking" }, { "msg_contents": "> \n> Hi all\n> \n> I started with the contrib/spi/.c stuff and am working toward a function\n> to be used by a trigger to update a datetime field upon doing an UPDATE to\n> a record.\n> \n> It is about ready to test, except one thing :)\n> \n> Is there a function like GetCurrentAbsoluteTime() that will get the\n> current time in a form suitable for inserting into a datetime field?\n> \n> GetCurrentDatetimeTime() maybe???\n> or a convertion function?\n\n What about datetime_in(\"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", "msg_date": "Fri, 6 Nov 1998 21:41:17 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] SPI hacking" }, { "msg_contents": "On Fri, 6 Nov 1998, Jan Wieck wrote:\n\n> > \n> > Hi all\n> > \n> > I started with the contrib/spi/.c stuff and am working toward a function\n> > to be used by a trigger to update a datetime field upon doing an UPDATE to\n> > a record.\n...\n> What about datetime_in(\"now\") ?\n> Jan\n\nYes, that's it! \n(Where is this documented at?)\n\nWell, here it is, seems to work.\nThe attached file is intended to be uppacked in\n/usr/src/pgsql/contrib/spi/ (on my box any way)\nand has a diff file for Makefile and the rest are new files.\n\nI'm in a hurry, got to go, so have not tested it but a little bit.\n\nHope it is usefull.\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!", "msg_date": "Fri, 6 Nov 1998 16:27:53 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": true, "msg_subject": "New SPI contrib stuff, was Re: [HACKERS] SPI hacking" }, { "msg_contents": "Patch applied to developers tree.\n\n> On Fri, 6 Nov 1998, Jan Wieck wrote:\n> \n> > > \n> > > Hi all\n> > > \n> > > I started with the contrib/spi/.c stuff and am working toward a function\n> > > to be used by a trigger to update a datetime field upon doing an UPDATE to\n> > > a record.\n> ...\n> > What about datetime_in(\"now\") ?\n> > Jan\n> \n> Yes, that's it! \n> (Where is this documented at?)\n> \n> Well, here it is, seems to work.\n> The attached file is intended to be uppacked in\n> /usr/src/pgsql/contrib/spi/ (on my box any way)\n> and has a diff file for Makefile and the rest are new files.\n> \n> I'm in a hurry, got to go, so have not tested it but a little bit.\n> \n> Hope it is usefull.\n> Terry Mackintosh <[email protected]> http://www.terrym.com\n> sysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n> \n> Proudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n> -------------------------------------------------------------------\n> Success Is A Choice ... book by Rick Patino, get it, read it!\nContent-Description: \n\n[application/octet-stream is not supported, skipping...]\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sat, 12 Dec 1998 15:21:15 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: New SPI contrib stuff, was Re: [HACKERS] SPI hacking" } ]
[ { "msg_contents": "I see from the 6.4 INSTALL file that Ultrix has made it onto the\nunsupported list. If this is simply an issue of developer time, I\nunderstand (and am not complaining). However, if this is simply \ndue to the lack of a machine, I am more than willing to put one on\nthe net for that purpose.\n\nBTW, I haven't tried 6.4 yet on Ultrix (I'm running 6.3.2 and have\nhad no problems). It may continue to work just fine.\n\nOn the other hand, this may be the final straw needed to get me to\nreplace my poor, pitiful Personal DECstation 5000 with something more\nuseful (like a 486 ;-)\n\n-- \nJeff Aitken\[email protected]\n", "msg_date": "Fri, 6 Nov 1998 16:15:23 -0500 (EST)", "msg_from": "Jeff Aitken <[email protected]>", "msg_from_op": true, "msg_subject": "Ultrix 4.4" } ]
[ { "msg_contents": "> Attempted a build of v6.4 on the Motorola 88k SVR4---\n> used \"./configure --with-template=svr4\".\n> config.guess returns \"m88k-motorola-sysv4\".\n> s_lock.c compile fails because of no slock_t and TAS definition.\n> I believe the 88k has a tas instruction, but I don't have a 88k user's\n> manual :-(\n> I'm stuck at this point unless I can find someone to get me thru this\n> one.\n\nI think that there are some folks on the list running 88k machines, or\nwho have had them in the past. Anyone have some suggestions to help\nDoug??\n\n - Tom\n", "msg_date": "Sat, 07 Nov 1998 05:54:29 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: postgresql-v6.4 - m88k svr4" } ]
[ { "msg_contents": "The bug is simple to reproduce :\n\nshowroom=> create view \"unu doi\" as select * from departamente;\nCREATE\nshowroom=> select * from \"unu doi\";\nERROR: nodeRead: Bad type 0 \n\n\nthe same error for : create view single_name as select * from \"double\nname\";\n\nAlso : dropping the bad view is not possible :\n\nshowroom=> drop view \"unu doi\";\nERROR: nodeRead: Bad type 0\n\n\nI am coming again with the description of the same type bug concerning\nspaces in object names :\n\nshowroom=> create sequence student_id;\nCREATE\nshowroom=> create table students (id int4 default\nnextval('student_id'),name text);\nCREATE\nshowroom=> drop table students;\nDROP\nshowroom=> create table \"my students\" (id int4 default\nnextval('student_id'), name text);\nERROR: my: Table does not exist.\n\nAny patches for them ?\n\nAll the best,\n\nConstantin Teodorescu\nFLEX Consulting Braila, ROMANIA\n", "msg_date": "Sat, 07 Nov 1998 20:20:05 +0000", "msg_from": "Constantin Teodorescu <[email protected]>", "msg_from_op": true, "msg_subject": "Bugs on opening views defined with spaces or upon table names with\n\tspaces !!!" } ]
[ { "msg_contents": "6.4 allows to use ORDER BY with a column that does not appear \nin a target list. This is great. However, following result seems\nstrange to me. Sometimes DISTINCT does not remove duplication.\nIs this normal behavior?\n\ncreate table dtest (i text, j date, k int);\nCREATE\ninsert into dtest values ('a', '1998/12/1', 1);\nINSERT 215243 1\ninsert into dtest values ('a', '1998/12/2', 2);\nINSERT 215244 1\ninsert into dtest values ('a', '1998/12/3', 3);\nINSERT 215245 1\nselect * from dtest;\ni\| j\|k\n-+----------+-\na\|12-01-1998\|1\na\|12-02-1998\|2\na\|12-03-1998\|3\n(3 rows)\n\nselect distinct i from dtest;\ni\n-\na\n(1 row)\n\nselect distinct i from dtest order by j;\ni\n-\na\na\na\n(3 rows)\n\nselect distinct i from dtest order by k;\ni\n-\na\na\na\n(3 rows)\n\n--\nTatsuo Ishii\[email protected]\n\n", "msg_date": "Sun, 8 Nov 1998 08:51:22 +0900", "msg_from": "[email protected] (Tatsuo Ishii)", "msg_from_op": true, "msg_subject": "distinct + order by" }, { "msg_contents": "[email protected] (Tatsuo Ishii) writes:\n> 6.4 allows to use ORDER BY with a column that does not appear \n> in a target list. This is great. However, following result seems\n> strange to me. Sometimes DISTINCT does not remove duplication.\n> Is this normal behavior?\n\nI think there's something funny going on, but it's not clear that\nthis kind of query is sensible anyway.\n\nIt's easy to see what's happening. DISTINCT relies on sorting by the\n\"distinct\" column first, so that a pass of a uniq(1)-like duplicate\ndetector can get rid of the duplicates:\n\ntree=> explain select distinct i from t;\nNOTICE: QUERY PLAN:\n\nUnique (cost=545.65 size=0 width=0)\n -> Sort (cost=545.65 size=0 width=0)\n -> Seq Scan on t (cost=545.65 size=11262 width=4)\n\nEXPLAIN\n\nNow, if we add an ORDER BY:\n\ntree=> explain select distinct i from t order by j;\nNOTICE: QUERY PLAN:\n\nUnique (cost=545.65 size=0 width=0)\n -> Sort (cost=545.65 size=0 width=0)\n -> Seq Scan on t (cost=545.65 size=11262 width=12)\n\nEXPLAIN\n\nOops. The sort is evidently being done on the order-by column,\nthus there's no guarantee that the duplicate i values will be\nbrought together.\n\nStill, I'd expect the Unique pass to get rid of duplicates that did\nmanage to be adjacent. Your example shows that that's not happening.\nSo Unique has some unexpected dependency on how the prior sort pass\nis done. (Maybe it assumes that the sort pass is on exactly the\ncolumns that were requested to be distinct, and re-uses the sort's\ncomparison results somehow?)\n\nIt seems clear that the query tree builder and the Unique executor\nhave different assumptions about how this is supposed to work. That\nprobably will yield bugs, so it needs to be looked at.\n\nBut I'm not convinced that the example you give is a sensible query\nat all. If you have done a SELECT DISTINCT on column A, what does it\nmean to specify that the results are sorted by column B? A single row\nof the result might represent many table rows with different values of\ncolumn B --- how should the system know which of those B values to use\nfor sorting the result?\n\nMy guess is that if you look into the SQL standard, this combination\nis forbidden. Thus the real bug is that Postgres isn't rejecting it.\n\nI'd still recommend that someone take a close look at how this example\nis behaving now, because even if it should be kicked out as a user\nerror, there might be related legal cases that are also having trouble.\n\nIf we did want to make this example behave in a rational way, then\nprobably the right implementation is something like\n\n\t* sort by i,j\n\t* distinct-filter on i only, being careful to keep first row\n\t\tin each set of duplicates\n\t* sort by j\n\nThis would ensure that the final sort by j uses, for each distinct i,\nthe lowest of the j-values associated with that i. This is a totally\narbitrary decision, but at least it will give reproducible results.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 08 Nov 1998 11:16:37 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] distinct + order by " }, { "msg_contents": "I said:\n> If we did want to make this example behave in a rational way, then\n> probably the right implementation is something like\n> \t* sort by i,j\n> \t* distinct-filter on i only, being careful to keep first row\n> \t\tin each set of duplicates\n> \t* sort by j\n> This would ensure that the final sort by j uses, for each distinct i,\n> the lowest of the j-values associated with that i. This is a totally\n> arbitrary decision, but at least it will give reproducible results.\n\nSome closer probing with \"explain verbose\" shows that\n\"SELECT DISTINCT i FROM dtest ORDER BY j\" is actually transformed\ninto this:\n\nUnique on i,j (cost=1.10 size=0 width=0)\n -> Sort by i,j (cost=1.10 size=0 width=0)\n -> Seq Scan on dtest selecting i,j (cost=1.10 size=3 width=16)\n\nThis explains why you get the apparently duplicate i values --- they're\nnot duplicate when both i and j are considered.\n\nIt looks to me like someone tried to make the query tree builder deal\nwith this case in the way I suggest above, but didn't finish the job.\nThe \"Unique\" pass is being done on the wrong targets, and there's no\nfinal sort.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 08 Nov 1998 12:06:59 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] distinct + order by " }, { "msg_contents": "> I said:\n> > If we did want to make this example behave in a rational way, then\n> > probably the right implementation is something like\n> > \t* sort by i,j\n> > \t* distinct-filter on i only, being careful to keep first row\n> > \t\tin each set of duplicates\n> > \t* sort by j\n> > This would ensure that the final sort by j uses, for each distinct i,\n> > the lowest of the j-values associated with that i. This is a totally\n> > arbitrary decision, but at least it will give reproducible results.\n> \n> Some closer probing with \"explain verbose\" shows that\n> \"SELECT DISTINCT i FROM dtest ORDER BY j\" is actually transformed\n> into this:\n> \n> Unique on i,j (cost=1.10 size=0 width=0)\n> -> Sort by i,j (cost=1.10 size=0 width=0)\n> -> Seq Scan on dtest selecting i,j (cost=1.10 size=3 width=16)\n> \n> This explains why you get the apparently duplicate i values --- they're\n> not duplicate when both i and j are considered.\n> \n> It looks to me like someone tried to make the query tree builder deal\n> with this case in the way I suggest above, but didn't finish the job.\n> The \"Unique\" pass is being done on the wrong targets, and there's no\n> final sort.\n\nI have added this to TODO:\n\n\t* SELECT DISTINCT i FROM dtest ORDER BY j generates strange output\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sat, 12 Dec 1998 15:33:13 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] distinct + order by" } ]
[ { "msg_contents": "Hi all\n\nOK, just built 6.4, a couple of things ... :)\n\nWhen I did make all runtest for the regression tests, I guess stderr used\nto be sent to /dev/null? but now it goes to the screen with LOTS of\nERROR this and ERROR that .... but when all was done I looked in the\noutput file and the only thing that failed was triggers, and a diff of\nexpected/results follows:\n\n[postgres@laptop /usr/src/pgsql/src/test/regress]$ diff\nexpected/triggers.out results/triggers.out\n53d52\n< NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n56,57d54\n< NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n< NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted\n59d55\n< NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n62,63d57\n< NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n< NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted\n[postgres@laptop /usr/src/pgsql/src/test/regress]$\n\nSo ... is this OK?\n\nHave a great night, and thanks for the great software.\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Sat, 7 Nov 1998 20:30:19 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": true, "msg_subject": "regression tests" }, { "msg_contents": "Terry Mackintosh <[email protected]> writes:\n> When I did make all runtest for the regression tests, I guess stderr used\n> to be sent to /dev/null? but now it goes to the screen with LOTS of\n> ERROR this and ERROR that ....\n\nDoesn't act that way for me ... all I see is one line per test with\n\"ok\" or \"failed\", same as it ever was.\n\n> but when all was done I looked in the\n> output file and the only thing that failed was triggers, and a diff of\n> expected/results follows:\n\n> [postgres@laptop /usr/src/pgsql/src/test/regress]$ diff\n> expected/triggers.out results/triggers.out\n> 53d52\n> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> 56,57d54\n> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted\n> 59d55\n> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> 62,63d57\n> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted\n> [postgres@laptop /usr/src/pgsql/src/test/regress]$\n\n> So ... is this OK?\n\nNo. Something's wrong.\n\nAs far as the error messages to screen go, maybe someone changed the\nregression driver shell script in a way that doesn't work on your shell.\nCan you check against the last driver script that behaved properly for\nyou?\n\nDunno about the triggers diffs, but most other people have reported\nan exact match for that one.\n\n\t\t\tregards, tom lane\n", "msg_date": "Sun, 08 Nov 1998 11:21:49 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] regression tests " }, { "msg_contents": "Hi Tom\n\nOK, I will do it all agian, maybe I botched some thing, I did do it while\nI was tiered, and I should know better:)\n\nOn Sun, 8 Nov 1998, Tom Lane wrote:\n\n> Terry Mackintosh <[email protected]> writes:\n> > When I did make all runtest for the regression tests, I guess stderr used\n> > to be sent to /dev/null? but now it goes to the screen with LOTS of\n> > ERROR this and ERROR that ....\n> \n> Doesn't act that way for me ... all I see is one line per test with\n> \"ok\" or \"failed\", same as it ever was.\n> \n> > but when all was done I looked in the\n> > output file and the only thing that failed was triggers, and a diff of\n> > expected/results follows:\n> \n> > [postgres@laptop /usr/src/pgsql/src/test/regress]$ diff\n> > expected/triggers.out results/triggers.out\n> > 53d52\n> > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> > 56,57d54\n> > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted\n> > 59d55\n> > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> > 62,63d57\n> > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted\n> > [postgres@laptop /usr/src/pgsql/src/test/regress]$\n> \n> > So ... is this OK?\n> \n> No. Something's wrong.\n> \n> As far as the error messages to screen go, maybe someone changed the\n> regression driver shell script in a way that doesn't work on your shell.\n> Can you check against the last driver script that behaved properly for\n> you?\n> \n> Dunno about the triggers diffs, but most other people have reported\n> an exact match for that one.\n> \n> \t\t\tregards, tom lane\n> \n\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Sun, 8 Nov 1998 13:56:18 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] regression tests " }, { "msg_contents": "Hi Tom and all\n\nOk, started from scratch, all works great. I did however read the INSTALL\nfile more closely, and notice some inconsistancies:\n\n 12. Install the HTML documentation. Type\n $ cd /usr/src/pgsql/doc\n $ gmake install\n The documentation is also available in Postscript\n format. Look for files ending with .ps.gz in the\n same directory.\n\nDoes not 'install' imply 'all install' if 'all' has not already been\ndone?\nAnd if so, would 'all' not include 'man'? Thus ...\n\n 13. Install the man page documentation. Type\n $ cd /usr/src/pgsql/doc\n $ gmake man\n\nThis complains that several files already exist.\n\nThats all, thanks for the great software, you people are a great bunch of\nhackers !-), if only some day I can be as skilled.\nHave a great night\nTerry\n\nOn Sun, 8 Nov 1998, Terry Mackintosh wrote:\n\n> Hi Tom\n> \n> OK, I will do it all agian, maybe I botched some thing, I did do it while\n> I was tiered, and I should know better:)\n> \n> On Sun, 8 Nov 1998, Tom Lane wrote:\n> \n> > Terry Mackintosh <[email protected]> writes:\n> > > When I did make all runtest for the regression tests, I guess stderr used\n> > > to be sent to /dev/null? but now it goes to the screen with LOTS of\n> > > ERROR this and ERROR that ....\n> > \n> > Doesn't act that way for me ... all I see is one line per test with\n> > \"ok\" or \"failed\", same as it ever was.\n> > \n> > > but when all was done I looked in the\n> > > output file and the only thing that failed was triggers, and a diff of\n> > > expected/results follows:\n> > \n> > > [postgres@laptop /usr/src/pgsql/src/test/regress]$ diff\n> > > expected/triggers.out results/triggers.out\n> > > 53d52\n> > > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> > > 56,57d54\n> > > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> > > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted\n> > > 59d55\n> > > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> > > 62,63d57\n> > > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys are deleted\n> > > < NOTICE: check_pkeys_fkey_cascade: 1 tuple(s) of fkeys2 are deleted\n> > > [postgres@laptop /usr/src/pgsql/src/test/regress]$\n> > \n> > > So ... is this OK?\n> > \n> > No. Something's wrong.\n> > \n> > As far as the error messages to screen go, maybe someone changed the\n> > regression driver shell script in a way that doesn't work on your shell.\n> > Can you check against the last driver script that behaved properly for\n> > you?\n> > \n> > Dunno about the triggers diffs, but most other people have reported\n> > an exact match for that one.\n> > \n> > \t\t\tregards, tom lane\n> > \n> \n> Terry Mackintosh <[email protected]> http://www.terrym.com\n> sysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n> \n> Proudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n> -------------------------------------------------------------------\n> Success Is A Choice ... book by Rick Patino, get it, read it!\n> \n> \n\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Sun, 8 Nov 1998 21:42:12 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": true, "msg_subject": "OK now :-) was Re: [HACKERS] regression tests " }, { "msg_contents": "> Ok, started from scratch, all works great. I did however read the \n> INSTALL file more closely, and notice some inconsistancies:\n> 12. Install the HTML documentation. Type\n> $ cd /usr/src/pgsql/doc\n> $ gmake install\n> The documentation is also available in Postscript\n> format. Look for files ending with .ps.gz in the\n> same directory.\n> \n> Does not 'install' imply 'all install' if 'all' has not already been\n> done?\n\nThe html docs are packaged as a tar file inside the main Postgres\ntarball. The only thing the Makefile does is \"install\". Should we put in\nan \"all\" to be consistant with the other makefiles? Perhaps so...\n\n> And if so, would 'all' not include 'man'? Thus ...\n> 13. Install the man page documentation. Type\n> $ cd /usr/src/pgsql/doc\n> $ gmake man\n> This complains that several files already exist.\n\nAs it always has. The man page installation process tries to make\ndirectories without checking to see if they already exist, so you see\nwarnings if so. I did update these instructions with a bias toward the\nhtml and postscript docs, I have to admit. But some people will still\nwant the man pages too so I included that as a separate step.\n\n - Tom\n", "msg_date": "Thu, 12 Nov 1998 04:56:24 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: OK now :-) was Re: [HACKERS] regression tests" }, { "msg_contents": "> Hi Tom and all\n> \n> Ok, started from scratch, all works great. I did however read the INSTALL\n> file more closely, and notice some inconsistancies:\n> \n> 12. Install the HTML documentation. Type\n> $ cd /usr/src/pgsql/doc\n> $ gmake install\n> The documentation is also available in Postscript\n> format. Look for files ending with .ps.gz in the\n> same directory.\n> \n> Does not 'install' imply 'all install' if 'all' has not already been\n> done?\n\nI am confused. What does the 'all' mean?\n\n> And if so, would 'all' not include 'man'? Thus ...\n> \n> 13. Install the man page documentation. Type\n> $ cd /usr/src/pgsql/doc\n> $ gmake man\n> \n> This complains that several files already exist.\n\nI see now. 'gmake install' installs everything, including the man\npages, so no need to do 'gmake man'.\n\nDo we want to split out gmake man and gmake html, or keep just gmake\ninstall and remove mention of gmake man?\n\nCan someone who understands how the INSTALL file and doc/Makefile are\nused please make the changes. Thanks.\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sat, 12 Dec 1998 15:57:09 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: OK now :-) was Re: [HACKERS] regression tests" } ]
[ { "msg_contents": "Hi,\n\nDoes anybody hear about InterSystems Cache' DBMS ?\nDoes it really supports caching from the backend side ?\n\n\n\tRegards,\n\t\t\n\t\tOleg\n\n_____________________________________________________________\nOleg Bartunov, sci.researcher, hostmaster of AstroNet,\nSternberg Astronomical Institute, Moscow University (Russia)\nInternet: [email protected], http://www.sai.msu.su/~megera/\nphone: +007(095)939-16-83, +007(095)939-23-83\n\n", "msg_date": "Sun, 8 Nov 1998 21:50:43 +0300 (MSK)", "msg_from": "Oleg Bartunov <[email protected]>", "msg_from_op": true, "msg_subject": "InterSystems CACHE' DB" } ]
[ { "msg_contents": "> Update of /usr/local/cvsroot/pgsql/src/backend/lib\n> In directory hub.org:/tmp/cvs-serv21717\n> \n> Modified Files:\n> \tstringinfo.c \n> Log Message:\n> Fix a potential infinite loop in appendStringInfo: would lock\n> up if first string to be appended to an empty StringInfo was longer\n> than the initial space allocation.\n> Also speed it up slightly.\n\nDoes this remove the need for vsnprintf?\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sun, 8 Nov 1998 16:44:17 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [COMMITTERS] 'pgsql/src/backend/lib stringinfo.c'" }, { "msg_contents": "Bruce Momjian wrote:\n> \n> > Update of /usr/local/cvsroot/pgsql/src/backend/lib\n> > In directory hub.org:/tmp/cvs-serv21717\n> >\n> > Modified Files:\n> > stringinfo.c\n> > Log Message:\n> > Fix a potential infinite loop in appendStringInfo: would lock\n> > up if first string to be appended to an empty StringInfo was longer\n> > than the initial space allocation.\n> > Also speed it up slightly.\n> \n> Does this remove the need for vsnprintf?\n\nI don't think so,\nvsprintf is still used if 6 places in to src tree, 5 of them is in\nthe backend. Each of these should be examined to determent wheater\nthose can be rewritten or if vsnprintf is needed.\n\nTo make matter worse:\n\nguevara-goran# pwd\n/usr/local/src/cvs/pgsql/src\nguevara-goran# grep -n sprintf `find .` \| wc -l\n 875\nguevara-goran# cd backend/\nguevara-goran# grep -n sprintf `find .` \| wc -l\n 474\n\nTheir is lot of potential overruns in there,\nand since pgsql is a net(-able) server we\nshould take that seriously.\n\nI will look closer at these issues as time permits. \n\n\tmvh,\n-- \n---------------------------------------------\nGï¿½ran Thyni, sysadm, JMS Bildbasen, Kiruna", "msg_date": "Mon, 09 Nov 1998 09:40:54 +0100", "msg_from": "Goran Thyni <[email protected]>", "msg_from_op": false, "msg_subject": "Overruns (was: 'pgsql/src/backend/lib stringinfo.c')" }, { "msg_contents": "Bruce Momjian <[email protected]> writes:\n> Does [stringinfo.c] remove the need for vsnprintf?\n\nNot directly --- the functions it currently provides only know how to\nappend given strings onto a StringInfo object. There's no formatting\ncontrol.\n\nI'm sure we could program around vsnprintf if we were determined\nenough, but there isn't any other equally clean way to do what's\nneeded in tracing. Probably better to spend our effort on providing\na reliable emulation of it for platforms that haven't got it.\n(BTW, I have no reason to think that the emulation we have is broken;\nI was just objecting to starting to depend on it only a week or so\nbefore a major release.)\n\nActually, what would be really whizzy is some way of sprintf'ing\ninto a StringInfo, with the ability to auto-expand the StringInfo as\nneeded. But that requires hooking into the low-level guts of printf,\nand AFAIK there's no portable way to do it short of providing your\nown complete printf implementation. Not worth it.\n\n\t\t\tregards, tom lane\n", "msg_date": "Mon, 09 Nov 1998 10:27:27 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/backend/lib stringinfo.c' " }, { "msg_contents": "Tom Lane wrote:\n> Actually, what would be really whizzy is some way of sprintf'ing\n> into a StringInfo, with the ability to auto-expand the StringInfo as\n> needed. But that requires hooking into the low-level guts of printf,\n> and AFAIK there's no portable way to do it short of providing your\n> own complete printf implementation. Not worth it.\n\nPerl does it transparently, \nwe could embed a perl-engine into the backend. :-)\n\nI am joking but it would be more useful and probably\nless work than writing a dynamic sprintf.\n\nBut their might be someone out-there on the 'Net that\nhas done a open-source dynamic sprintf already.\n \n\tmvh,\n-- \n---------------------------------------------\nGï¿½ran Thyni, sysadm, JMS Bildbasen, Kiruna\n", "msg_date": "Mon, 09 Nov 1998 17:27:03 +0100", "msg_from": "Goran Thyni <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/backend/lib stringinfo.c'" }, { "msg_contents": "Thus spake Tom Lane\n> Actually, what would be really whizzy is some way of sprintf'ing\n> into a StringInfo, with the ability to auto-expand the StringInfo as\n> needed. But that requires hooking into the low-level guts of printf,\n> and AFAIK there's no portable way to do it short of providing your\n> own complete printf implementation. Not worth it.\n\nYah, C really missed the boat there. I remember the Aztec C compiler\nfor CP/M had this neato function called format() which was used to\nimplement all the printf functions. You could write your own printf\ntype function by calling format with the function that handled the\noutput characters. I was real dissapointed to find out that it wasn't\na standard function.\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 9 Nov 1998 12:52:45 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/backend/lib stringinfo.c'" }, { "msg_contents": "D'Arcy J.M. Cain wrote:\n>\n> Yah, C really missed the boat there. I remember the Aztec C compiler\n> for CP/M had this neato function called format() which was used to\n\n Sniff - I remember the good old DR days of CP/M too. Where\n have all the good concepts gone? Someone remembers rasm and\n the rsx constructs for CP/M+? They worked before M$ created\n that damned, conflicting resident device driver hell.\n\n I still miss pip :-(\n\n Pardon to be totally off-topic here - sneeef.\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n", "msg_date": "Mon, 9 Nov 1998 20:01:43 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/backend/lib stringinfo.c'" }, { "msg_contents": "Thus spake Jan Wieck\n> Sniff - I remember the good old DR days of CP/M too. Where\n> have all the good concepts gone? Someone remembers rasm and\n> the rsx constructs for CP/M+? They worked before M$ created\n> that damned, conflicting resident device driver hell.\n> \n> I still miss pip :-(\n\nThat's PIP, buddy. Those were the days of CAPS and, by gum, we liked\nit that way. :-)\n\nSo, you want a copy of my CP/M emulator for Unix?\n\n-- \nD'Arcy J.M. Cain <darcy@{druid\|vex}.net> \| Democracy is three wolves\nhttp://www.druid.net/darcy/ \| and a sheep voting on\n+1 416 424 2871 (DoD#0082) (eNTP) \| what's for dinner.\n", "msg_date": "Mon, 9 Nov 1998 16:53:47 -0500 (EST)", "msg_from": "[email protected] (D'Arcy J.M. Cain)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [COMMITTERS] 'pgsql/src/backend/lib stringinfo.c'" }, { "msg_contents": "> > Does this remove the need for vsnprintf?\n> \n> I don't think so,\n> vsprintf is still used if 6 places in to src tree, 5 of them is in\n> the backend. Each of these should be examined to determent wheater\n> those can be rewritten or if vsnprintf is needed.\n> \n> To make matter worse:\n> \n> guevara-goran# pwd\n> /usr/local/src/cvs/pgsql/src\n> guevara-goran# grep -n sprintf `find .` \| wc -l\n> 875\n> guevara-goran# cd backend/\n> guevara-goran# grep -n sprintf `find .` \| wc -l\n> 474\n> \n> Their is lot of potential overruns in there,\n> and since pgsql is a net(-able) server we\n> should take that seriously.\n> \n> I will look closer at these issues as time permits. \n\n\nAdded to TODO:\n\n\t* fix any sprintf() overruns\n\t* add portable vsnprintf()\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sat, 12 Dec 1998 16:03:12 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": true, "msg_subject": "Re: Overruns (was: 'pgsql/src/backend/lib stringinfo.c')" } ]
[ { "msg_contents": "--postgresql-v6.4/src/pl/plpgsql/src/INSTALL says:\n...\n If you declare the language in the template1 database,\n any subsequently created database will have PL/pgSQL\n support installed automatically.\n\n\n--I declared PL/pgSQL in the template1 database but I have an ERROR\nmessage when I try to\n--use PL/pgSQL language in another database.\n\n\ncreate function nvl(integer) returns integer as\n'declare\n nonullo alias for $1;\nbegin\n if nonullo then\n return nonullo;\n else\n return 0;\n end if;\n end;\n' language 'plpgsql';\nERROR: Unrecognized language specified in a CREATE FUNCTION:\n'plpgsql'. Recogn\nized languages are sql, C, internal and the created procedural\nlanguages.\n\n--Seems that I must to declare PL/pgSQL in every database.\n--Any ideas?\n\n--Please relply to [email protected]\n\n--Jose-\n", "msg_date": "Tue, 10 Nov 1998 11:12:03 +0100", "msg_from": "\"Jose' Soares\" <[email protected]>", "msg_from_op": true, "msg_subject": "PL/pgSQL definition" }, { "msg_contents": ">\n> --postgresql-v6.4/src/pl/plpgsql/src/INSTALL says:\n> ...\n> If you declare the language in the template1 database,\n> any subsequently created database will have PL/pgSQL\n> support installed automatically.\n>\n>\n> --I declared PL/pgSQL in the template1 database but I have an ERROR\n> message when I try to\n> --use PL/pgSQL language in another database.\n>\n>\n> ERROR: Unrecognized language specified in a CREATE FUNCTION:\n> 'plpgsql'. Recogn\n> ized languages are sql, C, internal and the created procedural\n> languages.\n>\n> --Seems that I must to declare PL/pgSQL in every database.\n> --Any ideas?\n>\n\n You must declare PL/pgSQL in template1 BEFORE you create the\n other databases. At createdb time, the template1 database is\n copied into the new created database. All definitions, you\n made until then are automatically in the new db. So you could\n also create tables, functions and anything else you want to\n have by default in any new db.\n\n> --Please relply to [email protected]\n>\n> --Jose-\n>\n>\n\n\nJan\n\n--\n\n#======================================================================#\n# It's easier to get forgiveness for being wrong than for being right. #\n# Let's break this rule - forgive me. #\n#======================================== [email protected] (Jan Wieck) #\n\n\n", "msg_date": "Tue, 10 Nov 1998 12:36:20 +0100 (MET)", "msg_from": "[email protected] (Jan Wieck)", "msg_from_op": false, "msg_subject": "Re: [HACKERS] PL/pgSQL definition" } ]
[ { "msg_contents": "I noticed that when installing postgres the following directory is\nformed: ~pgsql/include/port/bsd (presumably the bsd changes with the\nport upon which it's compiled). This is created as part of the\nfollowing target in src/interfaces/libpq:\n\n beforeinstall-headers:\n\t @if [ ! -d $(HEADERDIR) ]; then mkdir $(HEADERDIR); fi\n\t @if [ ! -d $(HEADERDIR)/port ]; then mkdir $(HEADERDIR)/port; fi\n\t @if [ ! -d $(HEADERDIR)/port/$(PORTNAME) ]; \\\n\t\t then mkdir $(HEADERDIR)/port/$(PORTNAME); fi\n\nHowever, nothing ever seems to be put into this directory and as far\nas I can tell these are the only Makefile lines that reference this\ndirectory in any way.\n\nWhat is the point if creating it in the first place? Should stuff be\nput in there or should the directory disappear?\n\nCheers,\nBrook\n", "msg_date": "Tue, 10 Nov 1998 08:34:52 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": true, "msg_subject": "include/port directory?" }, { "msg_contents": "Brook Milligan <[email protected]> writes:\n> \t @if [ ! -d $(HEADERDIR)/port/$(PORTNAME) ]; \\\n> \t\t then mkdir $(HEADERDIR)/port/$(PORTNAME); fi\n\n> However, nothing ever seems to be put into this directory and as far\n> as I can tell these are the only Makefile lines that reference this\n> directory in any way.\n\nSome of the ports put stuff into that directory, some don't.\n(HPUX does, as you would've noticed if you looked about seven lines\nabove this line.)\n\n> What is the point if creating it in the first place? Should stuff be\n> put in there or should the directory disappear?\n\nI'd say leave it alone. Even if we don't have a lot of use for it\nat the moment, what's it hurt?\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Nov 1998 11:47:43 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] include/port directory? " }, { "msg_contents": " > \t @if [ ! -d $(HEADERDIR)/port/$(PORTNAME) ]; \\\n > \t\t then mkdir $(HEADERDIR)/port/$(PORTNAME); fi\n\n Some of the ports put stuff into that directory, some don't.\n (HPUX does, as you would've noticed if you looked about seven lines\n above this line.)\n\nOops. I missed that.\n\n > What is the point if creating it in the first place? Should stuff be\n > put in there or should the directory disappear?\n\n I'd say leave it alone. Even if we don't have a lot of use for it\n at the moment, what's it hurt?\n\nIt doesn't really hurt anything, except for two things. It causes a\nbit of confusion and it presents a small problem when trying to\nconstruct postgresql packages that contain the entire distribution\n(empty directories need special casing and when figuring out if that\nspecial casing is necessary the confusion arises; or at least it did\nfor me when I was doing exactly that). May I suggest the following\npatch that makes it clear exactly what is going on (when one looks at\nthe installed code) and makes construction of packages\nstraightforward.\n\nCheers,\nBrook\n\n===========================================================================\n--- interfaces/libpq/Makefile.in.orig\tSun Oct 18 18:00:43 1998\n+++ interfaces/libpq/Makefile.in\tTue Nov 10 11:18:17 1998\n@@ -107,6 +107,7 @@\n $(HEADERDIR)/commands/trigger.h\n \t$(INSTALL) $(INSTLOPTS) libpq-fe.h $(HEADERDIR)/libpq-fe.h\n \t$(INSTALL) $(INSTLOPTS) libpq-int.h $(HEADERDIR)/libpq-int.h\n+\t$(INSTALL) $(INSTLOPTS) .keep_me $(HEADERDIR)/port/$(PORTNAME)/.keep_me\n ifeq ($(PORTNAME), hpux)\n \t$(INSTALL) $(INSTLOPTS) $(SRCDIR)/backend/port/hpux/fixade.h \\\n $(HEADERDIR)/port/hpux/fixade.h\n===========================================================================\n--- interfaces/libpq/.keep_me.orig\tTue Nov 10 11:20:28 1998\n+++ interfaces/libpq/.keep_me\tTue Nov 10 11:20:14 1998\n@@ -0,0 +1,2 @@\n+This normally empty directory needs to be kept in the distribution for\n+port-specific header files.\n", "msg_date": "Tue, 10 Nov 1998 12:04:46 -0700 (MST)", "msg_from": "Brook Milligan <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] include/port directory?" }, { "msg_contents": "Brook Milligan <[email protected]> writes:\n> (empty directories need special casing\n\nWhy? And what makes you think there will never be any other empty\nsubdirectories? This \".keep_me\" looks like a much uglier hack than\nfixing whatever it is that can't deal with empty directories.\n\n\t\t\tregards, tom lane\n", "msg_date": "Tue, 10 Nov 1998 16:13:05 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] include/port directory? " } ]
[ { "msg_contents": "Hi,\n\nafter posting this problem to the admin and general group i hope it's ok\nto summarize the problem to this list.\n\nI have build postgres 6.4 under irix 6.5.1 with the irix cc 7.2.1\nThe regression test works without problem but many things still seem to be\nbroken:\n\n1) destroying a database fails with the following error\nERROR: typeidTypeRelid: Invalid type - oid = 0\ndestroydb: database destroy failed on test.\n\n2) pg_dump fails \npg_dump error in finding the template1 database\n\nthe debug info from the server shows\nStartTransactionCommand\nquery: SELECT oid from pg_database where datname = 'template1'\nProcessQuery\n/usr/local/pgsql/bin/postmaster: reaping dead processes...\n/usr/local/pgsql/bin/postmaster: CleanupProc: pid 422466 exited with\nstatus 139\n/usr/local/pgsql/bin/postmaster: CleanupProc: reinitializing shared memory\nand semaphores\nshmem_exit(0) [#0]\n\n3) vacuum a database reports something unusual\nNOTICE: Index pg_attribute_relid_attnum_index: NUMBER OF INDEX' TUPLES\n(915) IS NOT THE SAME AS HEAP' (937)\n\n4) selecting psql \\dd command fails\nERROR: typeidTypeRelid: Invalid type - oid = 0\n\nOverall, i guess the handling of the oid type does not seem to work,\nselecting oid from any table \ncrashes the server :-(\n\nI have checked the versions of bison and flex \nbison --version\nGNU Bison version 1.25\nflex --version\nflex version 2.5.4\n\nso this is probably not the reason. \n\nThe same problem has been reported for DEC OS 4.0d (if i remember it\ncorrectly, using cc -std1).\n\nI would be _very_ pleased if somebody could check this error. If it is\nrequired, a developer can get access to the postgres account on the\nmachine i used to build postgres.\n\nBest Regards,\nFuad\n\n\n--\nFuad Abdallah\nMax-Planck-Institut fuer Zuechtungsforschung / ZWDV\nTel.: 0221/5062-739 / Priv: 0221/584563\n\n", "msg_date": "Tue, 10 Nov 1998 21:25:42 +0100", "msg_from": "Fuad Abdallah <[email protected]>", "msg_from_op": true, "msg_subject": "6.4 error with oid type" } ]
[ { "msg_contents": "\nI am by no means a programmer, but exactly what is this referring to... I\nget ...\n\nIn file included from pgenv.cc:23:\npgenv.h:21: string: No such file or directory\n\nWhich seems fair enough... but I cant figure out what it's supposed to\nsay.... it compiles on my Linux box too... which I don't really get...\nunless it's not reading this file....\n\n/-------------------------------------------------------------------------\n \n * pgenv.h\n * \n \n DESCRIPTION\n * Postgres Environment Class: manages and stores all the\nrequired\n * connection variables.\n \n NOTES\n * Currently under construction.\n \n Copyright (c) 1994, Regents of the University of California\n \n\n-------------------------------------------------------------------------\n */\n \n#ifndef PGENV_H\n#define PGENV_H\n\n#include <string>\n\n\n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n", "msg_date": "Tue, 10 Nov 1998 22:06:54 +0000 (GMT)", "msg_from": "A James Lewis <[email protected]>", "msg_from_op": true, "msg_subject": "#include <string> ??" } ]
[ { "msg_contents": "\nMy fault,\n\nOn solaris there is no libg++ included in the GCC packages... I had to\ninclude the /opt package directory manually.......\n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n---------- Forwarded message ----------\nDate: Tue, 10 Nov 1998 22:06:54 +0000 (GMT)\nFrom: A James Lewis <[email protected]>\nTo: [email protected]\nSubject: #include <string> ??\n\n\nI am by no means a programmer, but exactly what is this referring to... I\nget ...\n\nIn file included from pgenv.cc:23:\npgenv.h:21: string: No such file or directory\n\nWhich seems fair enough... but I cant figure out what it's supposed to\nsay.... it compiles on my Linux box too... which I don't really get...\nunless it's not reading this file....\n\n/-------------------------------------------------------------------------\n \n * pgenv.h\n * \n \n DESCRIPTION\n * Postgres Environment Class: manages and stores all the\nrequired\n * connection variables.\n \n NOTES\n * Currently under construction.\n \n Copyright (c) 1994, Regents of the University of California\n \n\n-------------------------------------------------------------------------\n */\n \n#ifndef PGENV_H\n#define PGENV_H\n\n#include <string>\n\n\n\nJames ([email protected])\nVortex Internet\nMy Windows unders~1 long filena~1, and yours?\n\n\n", "msg_date": "Tue, 10 Nov 1998 22:17:23 +0000 (GMT)", "msg_from": "A James Lewis <[email protected]>", "msg_from_op": true, "msg_subject": "#include <string> ?? (fwd)" } ]
[ { "msg_contents": "I've been banging and banging my head against the specific problem of how\nto tightly couple the object-relational part of postgres with the object-\nlist processing part of Python. In the midst of this head-banging, I thought\nof this possible general solution:\n\nPut an ORB in the backend.\n\nNot only would this, in theory, give fairly reasonable performance for \npersistent object store applications in Python, it would open up PostgreSQL\nfor interoperability with any CORBA-interfaced application. Now that\nPostgreSQL has something of a stored procedural language, that makes it\neven more attractive.\n\nWhat I envision is a \"dumpIDL\" command that, for certain input parameters\n(table/class) would dump the appropriate IDL definition for instances(rows)\nof that class and related methods(functions). And then hack an existing \nORB onto the backend to do all the translation, communications and\nbookkeeping.\n\nSo, I'm soliciting opinions: is this brain fart worth pursuing, or should\nI just forget it ever happened?\n\n\t-Michael Robinson\n\n", "msg_date": "Wed, 11 Nov 1998 13:33:52 +0800 (GMT)", "msg_from": "Michael Robinson <[email protected]>", "msg_from_op": true, "msg_subject": "CORBA interface in backend?" }, { "msg_contents": "On Wed, 11 Nov 1998, Michael Robinson wrote:\n\n> I've been banging and banging my head against the specific problem of how\n> to tightly couple the object-relational part of postgres with the object-\n> list processing part of Python. In the midst of this head-banging, I thought\n> of this possible general solution:\n> \n> Put an ORB in the backend.\n> \n> Not only would this, in theory, give fairly reasonable performance for \n> persistent object store applications in Python, it would open up PostgreSQL\n> for interoperability with any CORBA-interfaced application. Now that\n> PostgreSQL has something of a stored procedural language, that makes it\n> even more attractive.\n> \n> What I envision is a \"dumpIDL\" command that, for certain input parameters\n> (table/class) would dump the appropriate IDL definition for instances(rows)\n> of that class and related methods(functions). And then hack an existing \n> ORB onto the backend to do all the translation, communications and\n> bookkeeping.\n> \n> So, I'm soliciting opinions: is this brain fart worth pursuing, or should\n> I just forget it ever happened?\n\nHow would this tie into the existing system? Does this become an\nextension of it, or replace parts of it?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Wed, 11 Nov 1998 01:54:12 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CORBA interface in backend?" }, { "msg_contents": "> > Put an ORB in the backend.\n>\n> How would this tie into the existing system? Does this become an\n> extension of it, or replace parts of it?\n\nProbably neither. It looks to me like it would be an alternate to the\ncurrent system. So the parser, etc. all stay in place, and the storage\nsystem can be changed out for the ORB.\n\nOr do I have the function of an ORB wrong? (It would help if I could\nremember exactly what it stands for...)\n\nTaral\n\n", "msg_date": "Wed, 11 Nov 1998 00:52:28 -0600", "msg_from": "\"Taral\" <[email protected]>", "msg_from_op": false, "msg_subject": "RE: [HACKERS] CORBA interface in backend?" }, { "msg_contents": "On Wed, Nov 11, 1998 at 01:33:52PM +0800, Michael Robinson wrote:\n> Put an ORB in the backend.\n\nSounds good IMO. Although I have to admit I'm not really well informed about\nCorba.\n\n> Not only would this, in theory, give fairly reasonable performance for \n> persistent object store applications in Python, it would open up PostgreSQL\n> for interoperability with any CORBA-interfaced application. Now that\n> PostgreSQL has something of a stored procedural language, that makes it\n> even more attractive.\n\nOnce again agreed.\n\n> What I envision is a \"dumpIDL\" command that, for certain input parameters\n> (table/class) would dump the appropriate IDL definition for instances(rows)\n> of that class and related methods(functions). And then hack an existing \n> ORB onto the backend to do all the translation, communications and\n> bookkeeping.\n\nThis is where it gets tricky. Which existing ORB are you thinking about? I\nknow of only two free ORBs: The Gnome one (orbit) and mico. Orbit is GPLed\nso there's a problem with PostgreSQL becoming GPLed. However, as long as\nPostgreSQL remain free under the BSD license that should be okay. But we\ndefintely lose the possibility to make this a commercial product. Not that\nwe want to do so.\n\nMico on the other hand is said to be pretty slow. I haven't tested it myself\nthough. And I have no idea what the license situation is.\n\n> So, I'm soliciting opinions: is this brain fart worth pursuing, or should\n> I just forget it ever happened?\n\nI'd say keep on thinking.\n\nMichael\n-- \nDr. Michael Meskes, Leiter Niederlassung West, Datenrevision GmbH\nbusiness: Cuxhavener Str. 36, D-21149 Hamburg\n private: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Wed, 11 Nov 1998 10:01:46 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CORBA interface in backend?" }, { "msg_contents": "On Wed, 11 Nov 1998, Michael Meskes wrote:\n\n> This is where it gets tricky. Which existing ORB are you thinking about? I\n> know of only two free ORBs: The Gnome one (orbit) and mico. Orbit is GPLed\n> so there's a problem with PostgreSQL becoming GPLed. However, as long as\n> PostgreSQL remain free under the BSD license that should be okay. But we\n> defintely lose the possibility to make this a commercial product. Not that\n> we want to do so.\n\n\tBoth are GPLd then, as I'm doing work with MICO and FreeBSD, and\njust checked their license files. \n\n\tNow, here is where I get confused. MICO uses LGPL for its\nlibraries, so does this mean that the ORB hooks could be added where\napplicable, still under the BSD license, but in order to compile it in,\nmico's libraries would have to be installed first? Its something we\ncheck for/do with the tcl stuff...\n\n\tSo, could the ORB hooks be added with a config option of\n'--enable-orb' without the requirement for adding in any GPLd code?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Wed, 11 Nov 1998 15:30:06 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CORBA interface in backend?" }, { "msg_contents": "On Wed, Nov 11, 1998 at 03:30:06PM -0400, The Hermit Hacker wrote:\n> \tBoth are GPLd then, as I'm doing work with MICO and FreeBSD, and\n> just checked their license files. \n> \n> \tNow, here is where I get confused. MICO uses LGPL for its\n> libraries, so does this mean that the ORB hooks could be added where\n> applicable, still under the BSD license, but in order to compile it in,\n> mico's libraries would have to be installed first? Its something we\n> check for/do with the tcl stuff...\n\nYes. As long as we add the ORB hooks ourselves (or use some under the BSD\ncopyright) this is fine. Just linking a program against an LGPLed library\ndoes not force you to (L)GPL the program.\n\n> \tSo, could the ORB hooks be added with a config option of\n> '--enable-orb' without the requirement for adding in any GPLd code?\n\nYes, I think so.\n\nMichael\n-- \nDr. Michael Meskes, Manager of the Western Branch Office, Datenrevision GmbH\nwork: Cuxhavener Str. 36, D-21149 Hamburg\nhome: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Thu, 12 Nov 1998 09:00:48 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CORBA interface in backend?" } ]
[ { "msg_contents": "There are many cases when they are small lookup tables with few records\nand detail tables with a lot of records.\nIn that cases it would be better to have an int2 autoincrement fields\nfor the lookup tables, not an int4.\n\ncreate table goods(id int2 default nextval('goodies_seq'), name text);\n\ndoes not work. Also,\n\ncreate table goods(create table goods(id int2 default\nnextval('goodies_seq'), name text);\n name text);\n\ndoesn't work either.\n\nThere is a chance to define a sql function next_goods_id returning int4\nas 'select nextval('goodies_seq')::int2' but it isn't so nice.\n\nBTW, shouldn't id int2 default nextval('goodies_seq')::int2 work ?\n\nConstantin Teodorescu\nFLEX Consulting Braila, ROMANIA\n", "msg_date": "Wed, 11 Nov 1998 14:44:58 +0000", "msg_from": "Constantin Teodorescu <[email protected]>", "msg_from_op": true, "msg_subject": "Small int autoincrement columns ?" } ]
[ { "msg_contents": ">On Wed, 11 Nov 1998, Michael Meskes wrote:\n>\n>> This is where it gets tricky. Which existing ORB are you thinking about?\nI\n>> know of only two free ORBs: The Gnome one (orbit) and mico. Orbit is\nGPLed\n>> so there's a problem with PostgreSQL becoming GPLed. However, as long as\n>> PostgreSQL remain free under the BSD license that should be okay. But we\n>> defintely lose the possibility to make this a commercial product. Not\nthat\n>> we want to do so.\n>\n\nthere's also omniORB (http://www.orl.co.uk) which i found when i was looking\nfor vnc. haven't used it, don't know much about it, but i believe it's also\nLGPL'd (maybe GPL'd)\n\n> Both are GPLd then, as I'm doing work with MICO and FreeBSD, and\n>just checked their license files.\n>\n> Now, here is where I get confused. MICO uses LGPL for its\n>libraries, so does this mean that the ORB hooks could be added where\n>applicable, still under the BSD license, but in order to compile it in,\n>mico's libraries would have to be installed first? Its something we\n>check for/do with the tcl stuff...\n>\n> So, could the ORB hooks be added with a config option of\n>'--enable-orb' without the requirement for adding in any GPLd code?\n\nan abstract orb interface? sounds good as long as someone else is doing the\nwork.\n\n\n", "msg_date": "Wed, 11 Nov 1998 14:43:25 -0600", "msg_from": "\"Jeff Hoffmann\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] CORBA interface in backend?" }, { "msg_contents": "> >> know of only two free ORBs: The Gnome one (orbit) and mico.\n> there's also omniORB (http://www.orl.co.uk)\n\nILU is also available from Xerox/parc, and has a license very close to\nthe Berkeley license for Postgres. The only extra is a copyright term\nsaying that the s/w must be distributed in a way which conforms to U.S.\nexport laws. afaik that is just to protect Xerox' bank account from\ntrouble caused by others.\n\n - Tom\n", "msg_date": "Thu, 12 Nov 1998 05:22:37 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CORBA interface in backend?" }, { "msg_contents": "On Thu, Nov 12, 1998 at 05:22:37AM +0000, Thomas G. Lockhart wrote:\n> ILU is also available from Xerox/parc, and has a license very close to\n> the Berkeley license for Postgres. The only extra is a copyright term\n> saying that the s/w must be distributed in a way which conforms to U.S.\n> export laws. afaik that is just to protect Xerox' bank account from\n> trouble caused by others.\n\nThen we're out of trouble. :-) Let's just hope this ORB is good enough. :-)\n\nMichael\n-- \nDr. Michael Meskes, Leiter Niederlassung West, Datenrevision GmbH\nbusiness: Cuxhavener Str. 36, D-21149 Hamburg\n private: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Thu, 12 Nov 1998 08:41:00 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CORBA interface in backend?" } ]
[ { "msg_contents": "\nDear Sirs,\n\nI'm a student of DSc Computer Science at PUC-Rio (Brazil) \nand I would like to know if there is a public version of\nPostgreSQL database in C++ language and how can I get the\nsource files.\n\nCertainly it will be very important to our group of research in \ndatabases.\n\nThanks for yours attention.\n\nAlvaro.\n\n-------------------------------------------------------------\nAlvaro Barbosa\n\nComputer Science Department (Departamento de Informï¿½tica)\nPontifï¿½cia Universidade Catï¿½lica do Rio de Janeiro (PUC-Rio)\nRua Marquï¿½s de Sï¿½o Vicente, 255 - Rio de Janeiro - RJ\n22453-900 - Brasil\n\nE-mail: [email protected]\n\n\n\n\n\n\n", "msg_date": "Wed, 11 Nov 1998 22:09:42 -0300", "msg_from": "Alvaro C P Barbosa <[email protected]>", "msg_from_op": true, "msg_subject": "Questions about PostgreSQL" } ]
[ { "msg_contents": "-----BEGIN PGP SIGNED MESSAGE-----\nHash: SHA1\n\n\nThe problem:\n\n\tCreating a table with a name containing mixed case that has a\n\tconstraint, the creation will fail. For example:\n\n\t mug=> create table \"MyTest\" ( a int4 default 10 );\n\t ERROR: mytest: Table does not exist.\n\t mug=> \n\nThe Solution:\n\n\tI haven't the faintest idea!\n\nThis problem bit me when I was trying to move an MS-Access database to \npostgreSQL while keeping MS-Access as the front end for the database.\n- -- \n____ \| Billy G. Allie \| Domain....: [email protected]\n\| /\| \| 7436 Hartwell \| Compuserve: 76337,2061\n\|-/-\|----- \| Dearborn, MI 48126\| MSN.......: [email protected]\n\|/ \|LLIE \| (313) 582-1540 \| \n\n\n\n-----BEGIN PGP SIGNATURE-----\nVersion: PGPfreeware 5.0i for non-commercial use\nCharset: noconv\n\niQA/AwUBNkpoxaFebSRz8o+3EQI+rwCdH1luSaOd78t0fKax+G4JnDeNndcAoKFV\nQfKhqQuqVhRYPYFUTBpCDwBD\n=KaAB\n-----END PGP SIGNATURE-----\n\n", "msg_date": "Wed, 11 Nov 1998 23:49:10 -0500", "msg_from": "\"Billy G. Allie\" <[email protected]>", "msg_from_op": true, "msg_subject": "A problem with the constraint system." }, { "msg_contents": "Hello Billy,\n\ngiovedï¿½, 12 novembre 98, you wrote:\n\nBGA> -----BEGIN PGP SIGNED MESSAGE-----\nBGA> Hash: SHA1\n\n\nBGA> The problem:\n\nBGA> Creating a table with a name containing mixed case that has a\nBGA> constraint, the creation will fail. For example:\n\nBGA> mug=> create table \"MyTest\" ( a int4 default 10 );\nBGA> ERROR: mytest: Table does not exist.\nBGA> mug=> \nBGA> The Solution:\nBGA> I haven't the faintest idea!\n\n\nI have this error also using CHECK... but the error message goes away\nif I create, drop and re-create the table...\n\nprova=> create table \"MyTest\" ( a int4 check(a>0) );\nERROR: mytest: Table does not exist.\nprova=> create table \"MyTest\" ( a int4 );\nCREATE\nprova=> drop table \"MyTest\";\nDROP\nprova=> create table \"MyTest\" ( a int4 check(a>0) );\nCREATE\n\n-Jose'-\n\n\n", "msg_date": "Thu, 12 Nov 1998 12:44:28 +0100", "msg_from": "Sferacarta Software <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A problem with the constraint system." }, { "msg_contents": "> BGA> Creating a table with a name containing mixed case that has a\n> BGA> constraint, the creation will fail. For example:\n> BGA> mug=> create table \"MyTest\" ( a int4 default 10 );\n> BGA> ERROR: mytest: Table does not exist.\n\nI have a fix for this, though I've only tested with your example. Here\nis the patch...\n\n - Tom", "msg_date": "Thu, 12 Nov 1998 15:17:24 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] A problem with the constraint system." } ]
[ { "msg_contents": "I've gotten part-way toward implementing a CASE construct in Postgres:\n\n select\n case when i = 1 then 10\n when i = 2 then 20\n end\n from t;\n\n(the other parts of the case clause are supported also). But I'm running\ninto problems in the executor, in that although the query tree seems to\nbe well-formed, there is not a corresponding plan or plan tree (not sure\nof the terminology). Any suggestions on where to start, or interest in\nworking on it? (hint, hint)\n\nbtw, I'm hoping to have a superset of the SQL92 definition; the standard\nallows only constants in the \"then\" clauses afaik, but I'm shooting for\nallowing full expressions...\n\n - Tom\n", "msg_date": "Thu, 12 Nov 1998 05:48:38 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "CASE construct" }, { "msg_contents": "> I've gotten part-way toward implementing a CASE construct...\n> ... (but) there is not a corresponding plan or plan tree (not sure\n> of the terminology). Any suggestions on where to start, or interest in\n> working on it? (hint, hint)\n\nAh, I've found Bruce's writeup on the query processing dataflow in\ntools/backend/index.html.\n\nLooks like a good place to start.\n\n - Tom\n", "msg_date": "Thu, 12 Nov 1998 23:32:41 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] CASE construct" }, { "msg_contents": "> > I've gotten part-way toward implementing a CASE construct...\n> > ... (but) there is not a corresponding plan or plan tree (not sure\n> > of the terminology). Any suggestions on where to start, or interest in\n> > working on it? (hint, hint)\n> \n> Ah, I've found Bruce's writeup on the query processing dataflow in\n> tools/backend/index.html.\n> \n> Looks like a good place to start.\n\nI was meaning to reply to your first post, but was sick. The way I\nusually do it is to look at some feature that already uses it, and copy\nthat.\n\nFor the executor, they are usually spread out in many files.\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 13 Nov 1998 01:30:19 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CASE construct" }, { "msg_contents": "> > > I've gotten part-way toward implementing a CASE construct...\n> I was meaning to reply to your first post, but was sick. The way I\n> usually do it is to look at some feature that already uses it, and \n> copy that.\n> For the executor, they are usually spread out in many files.\n\nHope you're feeling better. How scary was the \"Bruce-cam\" the last few\ndays? :)\n\nI've concluded that I need to make some fixups in the optimizer to get a\nplan which knows about column references inside the CASE clause. In the\nmeantime, I've got a rudimentary capability with constants:\n\npostgres=> select case\npostgres-> when 1 > 2 then 10\npostgres-> when 2 > 1 then 20\npostgres-> else 30\npostgres-> end;\nNOTICE: CASE/WHEN not yet implemented\n?column?\n--------\n 20\n(1 row)\n\nI'm working on matching up types between THEN/ELSE clauses at the parser\nstage, but will soon need to get back to trying to figure out the\nplanner. If I stall out for too long I may just commit the changes to\ngive someone else a chance to help out (existing features are not\ndamaged, so it shouldn't hurt).\n\n - Tom\n", "msg_date": "Fri, 13 Nov 1998 06:52:50 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [HACKERS] CASE construct" }, { "msg_contents": "> > > > I've gotten part-way toward implementing a CASE construct...\n> > I was meaning to reply to your first post, but was sick. The way I\n> > usually do it is to look at some feature that already uses it, and \n> > copy that.\n> > For the executor, they are usually spread out in many files.\n> \n> Hope you're feeling better. How scary was the \"Bruce-cam\" the last few\n> days? :)\n\nScary. I avoided it. It pops up a window when it is active, so no one\nsaw the scary-ness.\n\n> I've concluded that I need to make some fixups in the optimizer to get a\n> plan which knows about column references inside the CASE clause. In the\n> meantime, I've got a rudimentary capability with constants:\n> \n> postgres=> select case\n> postgres-> when 1 > 2 then 10\n> postgres-> when 2 > 1 then 20\n> postgres-> else 30\n> postgres-> end;\n> NOTICE: CASE/WHEN not yet implemented\n> ?column?\n> --------\n> 20\n> (1 row)\n\nThat is interesting.\n\n> \n> I'm working on matching up types between THEN/ELSE clauses at the parser\n> stage, but will soon need to get back to trying to figure out the\n> planner. If I stall out for too long I may just commit the changes to\n> give someone else a chance to help out (existing features are not\n> damaged, so it shouldn't hurt).\n\nSure. No problem. Advantage of two trees. (See Marc, I am learning.)\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Fri, 13 Nov 1998 05:50:34 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] CASE construct" } ]
[ { "msg_contents": "This is stl version of GNU Strings\n\nlibpq++ intensively use this package.\n\nIMHO,\n It's mutch safely use non-stl version of GNU Strings\n i.e change all \n string s to String h\n and \n #include <string> to #include <String.h>\n \n \n\n-- \nDmitry Samersoff\n DM\\S, [email protected]\n http://members.xoom.com/Samersoff\n\n", "msg_date": "Thu, 12 Nov 1998 10:24:22 +0300 (MSK)", "msg_from": "Dmitry Samersoff <[email protected]>", "msg_from_op": true, "msg_subject": "Re: <string>" } ]
[ { "msg_contents": ">here is a bug which is reproducible on FreeBSD-2.2.6\n>as well as on FreeBSD-2.2.7:\n>\n>\n>nadia:mergl> echo -n \"testing large objects using blob_read\" >/tmp/gaga\n>nadia:mergl> createdb pgtest\n>nadia:mergl> psql pgtest\n>...\n>pgtest=> CREATE TABLE lobject ( id int4, loid oid );\n>CREATE\n>pgtest=> INSERT INTO lobject (id, loid) VALUES (1, lo_import('/tmp/gaga'));\n>pqReadData() -- backend closed the channel unexpectedly.\n> This probably means the backend terminated abnormally before or while processing the request.\n>We have lost the connection to the backend, so further processing is impossible. Terminating.\n\nReproduced here too. Please try included patches...\n\nTatsuo Ishii\[email protected]\n---------------------------------------------------------------------------\n* postgresql-v6.4/src/backend/storage/large_object/inv_api.c~\tThu Nov 5 17:48:26 1998\n--- postgresql-v6.4/src/backend/storage/large_object/inv_api.c\tThu Nov 12 15:52:59 1998\n***********\n* 549,556 ***\n \t\t\t\ttuplen = inv_wrnew(obj_desc, buf, nbytes - nwritten);\n \t\t\telse\n \t\t\t\ttuplen = inv_wrold(obj_desc, buf, nbytes - nwritten, tuple, buffer);\n \t\t}\n! \t\tReleaseBuffer(buffer);\n \n \t\t/ move pointers past the amount we just wrote /\n \t\tbuf += tuplen;\n--- 549,557 ----\n \t\t\t\ttuplen = inv_wrnew(obj_desc, buf, nbytes - nwritten);\n \t\t\telse\n \t\t\t\ttuplen = inv_wrold(obj_desc, buf, nbytes - nwritten, tuple, buffer);\n+ \t\t\tReleaseBuffer(buffer);\n \t\t}\n! \n \n \t\t/ move pointers past the amount we just wrote /\n \t\tbuf += tuplen;\n", "msg_date": "Thu, 12 Nov 1998 16:29:43 +0900", "msg_from": "Tatsuo Ishii <[email protected]>", "msg_from_op": true, "msg_subject": "Re: [BUGS] bug in postgresql-v6.4 on FreeBSD " }, { "msg_contents": "Tatsuo Ishii wrote:\n> \n> >here is a bug which is reproducible on FreeBSD-2.2.6\n> >as well as on FreeBSD-2.2.7:\n> >\n> >\n> >nadia:mergl> echo -n \"testing large objects using blob_read\" >/tmp/gaga\n> >nadia:mergl> createdb pgtest\n> >nadia:mergl> psql pgtest\n> >...\n> >pgtest=> CREATE TABLE lobject ( id int4, loid oid );\n> >CREATE\n> >pgtest=> INSERT INTO lobject (id, loid) VALUES (1, lo_import('/tmp/gaga'));\n> >pqReadData() -- backend closed the channel unexpectedly.\n> > This probably means the backend terminated abnormally before or while processing the request.\n> >We have lost the connection to the backend, so further processing is impossible. Terminating.\n> \n> Reproduced here too. Please try included patches...\n> \n> Tatsuo Ishii\n> [email protected]\n> ---------------------------------------------------------------------------\n> postgresql-v6.4/src/backend/storage/large_object/inv_api.c~ Thu Nov 5 17:48:26 1998\n> --- postgresql-v6.4/src/backend/storage/large_object/inv_api.c Thu Nov 12 15:52:59 1998\n> ***********\n> * 549,556 ***\n> tuplen = inv_wrnew(obj_desc, buf, nbytes - nwritten);\n> else\n> tuplen = inv_wrold(obj_desc, buf, nbytes - nwritten, tuple, buffer);\n> }\n> ! ReleaseBuffer(buffer);\n> \n> / move pointers past the amount we just wrote /\n> buf += tuplen;\n> --- 549,557 ----\n> tuplen = inv_wrnew(obj_desc, buf, nbytes - nwritten);\n> else\n> tuplen = inv_wrold(obj_desc, buf, nbytes - nwritten, tuple, buffer);\n> + ReleaseBuffer(buffer);\n> }\n> !\n> \n> / move pointers past the amount we just wrote */\n> buf += tuplen;\n\n\n\nyes, this patch works also for me.\nIt is needed to pass 'make test' for DBD-Pg-0.89.\n\n\nthanks\nEdmund\n-- \nEdmund Mergl mailto:[email protected]\nIm Haldenhau 9 http://www.bawue.de/~mergl\n70565 Stuttgart fon: +49 711 747503\nGermany\n", "msg_date": "Fri, 13 Nov 1998 22:43:28 +0000", "msg_from": "Edmund Mergl <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] Re: [BUGS] bug in postgresql-v6.4 on FreeBSD" } ]
[ { "msg_contents": "\n1) A corba interface sounds interesting. I wish I had time\n to investigate corba further. \n\n2) Applixware proports to be able to access corba objects. I guess\n this means it can import/export corba IDLs, and talk over IIOP?\n I haven't tried it out.\n\n3) I'm not sure that the interface needs to be part of the backend.\n Most of the structure information can be obtained from the pg_ \n tables. Having a separate process would be safer for the overall\n stability of the backend. Am I missing something?\n\nIn any case, it seems very worth investigating.\n\n-- cary\n\n", "msg_date": "Thu, 12 Nov 1998 10:00:09 -0500 (EST)", "msg_from": "\"Cary B. O'Brien\" <[email protected]>", "msg_from_op": true, "msg_subject": "Re: CORBA Interface" } ]
[ { "msg_contents": "\n\n-----Original Message-----\nFrom: Dan Gowin \nSent: Thursday, November 12, 1998 10:46 AM\nTo: 'Cary B. O'Brien'\nSubject: RE: [HACKERS] Re: CORBA Interface\n\n\nAll,\n\tI agree with below. The ORB should be a seperate\ndaemon from the backend. Not only will the stability\nof the backend be affected if the ORB is placed in the\nbackend, but performance will also suffer. The ORB should\nbe developed as a seperate daemon which -could- be placed\non a seperate server to increase performance. Also remember,\nhowever the ORB is designed, it will consume alot of CPU\nand memory resources within the server.\n\nD.\n\n-----Original Message-----\nFrom: Cary B. O'Brien [mailto:[email protected]]\nSent: Thursday, November 12, 1998 10:00 AM\nTo: [email protected]\nSubject: [HACKERS] Re: CORBA Interface\n\n\n\n1) A corba interface sounds interesting. I wish I had time\n to investigate corba further. \n\n2) Applixware proports to be able to access corba objects. I guess\n this means it can import/export corba IDLs, and talk over IIOP?\n I haven't tried it out.\n\n3) I'm not sure that the interface needs to be part of the backend.\n Most of the structure information can be obtained from the pg_ \n tables. Having a separate process would be safer for the overall\n stability of the backend. Am I missing something?\n\nIn any case, it seems very worth investigating.\n\n-- cary\n\n", "msg_date": "Thu, 12 Nov 1998 10:57:15 -0500", "msg_from": "Dan Gowin <[email protected]>", "msg_from_op": true, "msg_subject": "FW: [HACKERS] Re: CORBA Interface" }, { "msg_contents": "> The ORB should be a seperate\n> daemon from the backend. Not only will the stability\n> of the backend be affected if the ORB is placed in the\n> backend, but performance will also suffer. The ORB should\n> be developed as a seperate daemon which -could- be placed\n> on a seperate server to increase performance. Also remember,\n> however the ORB is designed, it will consume alot of CPU\n> and memory resources within the server.\n>> 3) I'm not sure that the interface needs to be part of the backend.\n>> Most of the structure information can be obtained from the pg_\n>> tables. Having a separate process would be safer for the overall\n>> stability of the backend. Am I missing something?\n\nI'm not sure I follow this. We would be looking to the ORB to replace\nall on-the-wire communications between frontends and backends, to take\nadvantage of the data marshalling, security and directory layers of\nCorba. So we would want the ORB to infest at least parts of the current\nPostgres backend to eliminate the current socket/IP interface, right?\n\nOr would we only be using the ORB to present a sort-of-thin client\nlayer, to replace libpq for an app? Are there other RDBMSes which claim\nto have a Corba interface? What does the API look like?\n\nI've been looking at Corba for other projects, and have worked with DCE\na bit in the past. I would guess that the interface between, say, psql\nand a server would have psql asking for a postgres server through the\ncorba API, and the server side would provide a fresh backend as a\nresult. The API would include methods to allow queries, etc, and methods\nto return results...\n\n - Tom\n", "msg_date": "Thu, 12 Nov 1998 16:30:00 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: FW: [HACKERS] Re: CORBA Interface" } ]
[ { "msg_contents": "Version : Postgresql 6.4\n\n\nSituation :\n\nwhenever I try to update or delete a row int the users table I get the error message\n\ninfo fmgr : cache lookup failure.\n\nupdate users set username = 'new' where username = 'old';\nor\ndelete from users where username = 'old';\n\nI can do selects with no problems at all.\n\nTABLES:\ncreate table users(username varchar(20) primary key,\n last_name varchar(50),\n first_name varchar(50),\n password varchar(20),\n email_on int4,\n fs_on int4,\n web_access_on int4,\n comments text,\n email_alias varchar(50),\n email_forward varchar(50),\n qouta int4,\n date_created datetime);\n\ncreate table user_groups(username varchar(20),\n group_name varchar(20));\n\nTRIGGERS:\nCREATE TRIGGER users_for_key\n BEFORE DELETE OR UPDATE ON users\n FOR EACH ROW\n EXECUTE PROCEDURE check_foreign_key(1, 'cascade', 'username', 'user_groups', 'username');\n\nCREATE TRIGGER users_user_groups_pri_key\n BEFORE INSERT OR UPDATE ON user_groups\n FOR EACH ROW\n EXECUTE PROCEDURE check_primary_key('username', 'users', 'username');\n\nI am using the supplied refint.so stored procedure module.\n\nAfter I rebuild the whole database this problem seems to disappear, but, will eventually reappear after a few hours. I will then have to rebuild again.\n\nRebuilding the database is not an acceptable work around nor fix.\n\nCould someone please lead me in the right direction ?\n\nThanks,\nBrian Bradley\n\n\n\n\n\n\n\n\n\n\n\n\n\nVersion : Postgresql 6.4\n \n \nSituation :\n \nwhenever I try to update or delete a row int the \nusers table I get the error message\n \ninfo fmgr : cache lookup failure.\n \nupdate users set username = 'new' where username \n= 'old';\nor\ndelete from users where username = 'old';\n \nI can do selects with no problems at all.\n \nTABLES:\ncreate table users(username \nvarchar(20) primary \nkey, \nlast_name \nvarchar(50), \nfirst_name \nvarchar(50), \npassword \nvarchar(20), \nemail_on \nint4, \nfs_on \nint4, \nweb_access_on \nint4, \ncomments \ntext, \nemail_alias \nvarchar(50), \nemail_forward \nvarchar(50), \nqouta \nint4, \ndate_created datetime);\n \ncreate table user_groups(username \nvarchar(20), \ngroup_name varchar(20));\n \nTRIGGERS:\nCREATE TRIGGER \nusers_for_key BEFORE DELETE OR \nUPDATE ON users FOR EACH \nROW EXECUTE PROCEDURE \ncheck_foreign_key(1, 'cascade', 'username', 'user_groups', \n'username');\n \nCREATE TRIGGER \nusers_user_groups_pri_key BEFORE \nINSERT OR UPDATE ON user_groups \nFOR EACH ROW EXECUTE PROCEDURE \ncheck_primary_key('username', 'users', 'username');\n \nI am using the supplied refint.so stored procedure \nmodule.\n \nAfter I rebuild the whole database this problem seems to \ndisappear, but, will eventually reappear after a few hours. I will then have to \nrebuild again.\n \nRebuilding the database is not an acceptable work around nor \nfix.\n \nCould someone please lead me in the right direction \n?\n \nThanks,\nBrian Bradley", "msg_date": "Thu, 12 Nov 1998 11:42:49 -0500", "msg_from": "\"Brian Bradley\" <[email protected]>", "msg_from_op": true, "msg_subject": "cache lookup failure" } ]
[ { "msg_contents": "subscribe\n\n", "msg_date": "Fri, 13 Nov 1998 09:23:16 +0900", "msg_from": "\"=?euc-kr?B?vNux4r/4?=\" <[email protected]>", "msg_from_op": true, "msg_subject": "subscribe" } ]
[ { "msg_contents": "Hi!\ni want to know that how to restart postgresql without using \"Kill -9 xxxx\"\n....\n\nPlease, Help me!!\n\n", "msg_date": "Fri, 13 Nov 1998 09:37:40 +0900", "msg_from": "\"=?euc-kr?B?vNux4r/4?=\" <[email protected]>", "msg_from_op": true, "msg_subject": "How to restart postgresql??" } ]
[ { "msg_contents": "Hello!\n\nA background postgres porocess gets larger with Abort Transactions.\nMemory seems to leak.\nThis occurs also on 6.3.2.\n\nYou can see it with user aborts:\n\n========================================================================\n#!/bin/sh\n\nyes 'begin;\nabort;' \| psql regression\n========================================================================\n\nAnd with internal aborts:\n\n========================================================================\n#!/bin/sh\n\nyes \"insert into Room (roomno) values ('000');\" \| psql regression\n========================================================================\n\nRegards\n\n-- \nASCII CORPORATION\nTechnical Center\nSHIOZAKI Takehiko\n<[email protected]>\n", "msg_date": "Fri, 13 Nov 1998 10:39:08 +0900 (JST)", "msg_from": "SHIOZAKI Takehiko <[email protected]>", "msg_from_op": true, "msg_subject": "memory leak with Abort Transaction" }, { "msg_contents": "> Hello!\n> \n> A background postgres porocess gets larger with Abort Transactions.\n> Memory seems to leak.\n> This occurs also on 6.3.2.\n> \n> You can see it with user aborts:\n> \n> ========================================================================\n> #!/bin/sh\n> \n> yes 'begin;\n> abort;' \| psql regression\n> ========================================================================\n> \n> And with internal aborts:\n> \n> ========================================================================\n> #!/bin/sh\n> \n> yes \"insert into Room (roomno) values ('000');\" \| psql regression\n> ========================================================================\n\nAddd to TODO:\n\n\t* fix memory leak in aborted transactions\n\n\n-- \n Bruce Momjian \| http://www.op.net/~candle\n [email protected] \| (610) 853-3000\n + If your life is a hard drive, \| 830 Blythe Avenue\n + Christ can be your backup. \| Drexel Hill, Pennsylvania 19026\n", "msg_date": "Sat, 12 Dec 1998 17:08:31 -0500 (EST)", "msg_from": "Bruce Momjian <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] memory leak with Abort Transaction" } ]
[ { "msg_contents": "Hi!\ni want to know that how to restart postgresql without using \"Kill -9 xxxx\"\n....\n\n[My system SPEC]\nOS : Sun OS 2.6\npostgresql : 6.3 version\n\nPlease, Help me!!\n\n\n", "msg_date": "Fri, 13 Nov 1998 11:16:32 +0900", "msg_from": "\"=?euc-kr?B?vNux4r/4?=\" <[email protected]>", "msg_from_op": true, "msg_subject": "How to restarting postmaster ??" } ]
[ { "msg_contents": "System Spec\n SUN OS 2.6\n PostgreSQL 6.3 Version\n\n[Error Message]\nsol% postmaster -B 512 -S -D /usr/local/pgsql/data\nIpcMemoryCreate: memKey=5432001 , size=4501144 ,\npermission=384IpcMemoryCreate: shmget(..., create, ...) failed: Invalid\nargument\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: ShmemCreate: cannot create region\n\nI don't know why error occured....\nPlease.. Help me.....\n\n\n\n", "msg_date": "Fri, 13 Nov 1998 12:01:40 +0900", "msg_from": "\"=?euc-kr?B?vNux4r/4?=\" <[email protected]>", "msg_from_op": true, "msg_subject": "Can't starting postmaster with -B 512 Option" }, { "msg_contents": "On Fri, 13 Nov 1998, [euc-kr] �۱�� wrote:\n\n> System Spec\n> SUN OS 2.6\n> PostgreSQL 6.3 Version\n> \n> [Error Message]\n> sol% postmaster -B 512 -S -D /usr/local/pgsql/data\n> IpcMemoryCreate: memKey=5432001 , size=4501144 ,\n> permission=384IpcMemoryCreate: shmget(..., create, ...) failed: Invalid\n> argument\n> This type of error is usually caused by improper\n> shared memory or System V IPC semaphore configuration.\n> See the FAQ for more detailed information\n> FATAL 1: ShmemCreate: cannot create region\n> \n> I don't know why error occured....\n> Please.. Help me.....\n\nFirst off...pick a list, don't send to them all...have narrowed it down to\nthe only one that was appropriate...\n\nSecond of all...Under Solaris 2.6, you have to raise the shared memory\nsegments available via the /etc/systems file. I don' have the appropriate\ndocument sittin gin front of me, or else I'd try and give you a more\ndetailed answer, but, for example, one of our Solaris machines at the\noffice as the following in its /etc/system file:\n\nset shmsys:shminfo_shmmax=16777216\nset shmsys:shminfo_shmmin=1\nset shmsys:shminfo_shmmni=100\nset shmsys:shminfo_shmseg=51\n\nEach of the values tend to \"inter-relate\"...so you'll want to read up on\nthe appropriate docs for this, available through Sun...\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Thu, 12 Nov 1998 23:49:21 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [SQL] Can't starting postmaster with -B 512 Option" } ]
[ { "msg_contents": "With the initial positive response, I dived into the CORBA spec, the\nORBit source, and the PostgreSQL source. I have good news to report.\n\n- CORBA and PostgreSQL were made for each other. The designers of the CORBA\narchitecture had object oriented databases (OODB) in mind as one of the\nstandard implementation of objects. E.g.:\n\n \"An ORB may have access to multiple Interface Repositories. This may occur\n because...an object implementation (such as an OODB) prefers to provide\n its own type information...\"\n\nSo, implementing CORBA for PostgreSQL is simply a matter of following the\nvery clearly designed architecture.\n\n- CORBA is extremely modular with well-defined interfaces between components.\nThis means that it should be possible to write generic CORBA glue for\nPostgreSQL that could work with any ORB.\n\n- The ORBit sources appear to be LGPL'ed, which means they can be linked to\nPostgreSQL without poisoning the BSD license.\n\nI also have bad news to report.\n\n- Most of the CORBA functionality that PostgreSQL would rely on is currently\nunimplemented in ORBit.\n\n- While CORBA provides a very disciplined interface for allowing different\nobject implementations (e.g. Python and PostgreSQL) to share the same address\nspace and execution context safely and efficiently, the PostgreSQL backend\ndoesn't seem ready for it. In particular, it doesn't appear to be thread\nsafe. It may not even be reentrant, from what I can tell. And, if a backend\nprocess is not punctual about reading cache synchronization messages out of\nthe IPC queue, it appears that excessive cache invalidation would hurt\nperformance.\n\n -Michael Robinson\n\n", "msg_date": "Fri, 13 Nov 1998 13:36:05 +0800 (GMT)", "msg_from": "Michael Robinson <[email protected]>", "msg_from_op": true, "msg_subject": "More CORBA and PostgreSQL" }, { "msg_contents": "\nJust to add my bit to this:\n\nAlthough I haven't looked into it yet, but I've come across a book\ndescribing how to use CORBA with Java, and JDBC. I'm going to try to get\nthis book, and see how this could work.\n\nOn Fri, 13 Nov 1998, Michael Robinson wrote:\n\n> With the initial positive response, I dived into the CORBA spec, the\n> ORBit source, and the PostgreSQL source. I have good news to report.\n> \n> - CORBA and PostgreSQL were made for each other. The designers of the CORBA\n> architecture had object oriented databases (OODB) in mind as one of the\n> standard implementation of objects. E.g.:\n> \n> \"An ORB may have access to multiple Interface Repositories. This may occur\n> because...an object implementation (such as an OODB) prefers to provide\n> its own type information...\"\n> \n> So, implementing CORBA for PostgreSQL is simply a matter of following the\n> very clearly designed architecture.\n> \n> - CORBA is extremely modular with well-defined interfaces between components.\n> This means that it should be possible to write generic CORBA glue for\n> PostgreSQL that could work with any ORB.\n> \n> - The ORBit sources appear to be LGPL'ed, which means they can be linked to\n> PostgreSQL without poisoning the BSD license.\n> \n> I also have bad news to report.\n> \n> - Most of the CORBA functionality that PostgreSQL would rely on is currently\n> unimplemented in ORBit.\n> \n> - While CORBA provides a very disciplined interface for allowing different\n> object implementations (e.g. Python and PostgreSQL) to share the same address\n> space and execution context safely and efficiently, the PostgreSQL backend\n> doesn't seem ready for it. In particular, it doesn't appear to be thread\n> safe. It may not even be reentrant, from what I can tell. And, if a backend\n> process is not punctual about reading cache synchronization messages out of\n> the IPC queue, it appears that excessive cache invalidation would hurt\n> performance.\n> \n> -Michael Robinson\n> \n> \n\n-- \nPeter Mount, IT Section\[email protected]\nAnything I write here are my own views, and cannot be taken as being the\nofficial words of Maidstone Borough Council\n\n\n", "msg_date": "Fri, 13 Nov 1998 08:12:20 +0000 (GMT)", "msg_from": "Peter T Mount <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "On Fri, 13 Nov 1998, Michael Robinson wrote:\n\n> - The ORBit sources appear to be LGPL'ed, which means they can be linked to\n> PostgreSQL without poisoning the BSD license.\n\n\tMico is also LGPL'd for the libraries...\n\n> I also have bad news to report.\n> \n> - Most of the CORBA functionality that PostgreSQL would rely on is currently\n> unimplemented in ORBit.\n\n\tI don't know what is implemented, but check out:\n\n\thttp://www.vsb.cs.uni-frankfurt.de/~mico\n\n\tThey \"claim\" a completely 2.2 Corba implementation...\n\n> - While CORBA provides a very disciplined interface for allowing different\n> object implementations (e.g. Python and PostgreSQL) to share the same address\n> space and execution context safely and efficiently, the PostgreSQL backend\n> doesn't seem ready for it. In particular, it doesn't appear to be thread\n> safe. It may not even be reentrant, from what I can tell. And, if a backend\n> process is not punctual about reading cache synchronization messages out of\n> the IPC queue, it appears that excessive cache invalidation would hurt\n> performance.\n\n\tHrmmm...does this mean that we are going to have to move towards a\nthreaded model vs forked? Or is it just going to require some major code\ncleanups for the 'thread safe/reentrant' as aspect?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 13 Nov 1998 09:00:14 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "> I don't know what is implemented, but check out:\n> http://www.vsb.cs.uni-frankfurt.de/~mico\n> They \"claim\" a completely 2.2 Corba implementation...\n\nIf someone were choosing an ORB, they perhaps could look at whatever the\nGnome project chose (either mico or ORBit, can't remember which). Also,\nthey didn't consider ILU because of licensing considerations, but the\nlicense changed very recently and I think would now be a strong\ncandidate...\n\n> > ... the PostgreSQL backend\n> > doesn't seem ready for it. In particular, it doesn't appear to be \n> > thread safe. It may not even be reentrant, from what I can tell. \n> > And, if a backend process is not punctual about reading cache \n> > synchronization messages out of the IPC queue, it appears that \n> > excessive cache invalidation would hurt performance.\n\nThe PG backend is neither reentrant nor threadsafe, and isn't likely to\nbecome so soon (several/many places where global variables are used,\netc).\n\nHowever, with the existing \"forked model\", there is a separate backend\nfor each client, so (if I understand things a bit) the trick will be\nfiguring out how to call a single routine which will give access to a\nclient (as happens now) but without handing off through a socket/IP\nconnection.\n\n - Tom\n", "msg_date": "Fri, 13 Nov 1998 14:49:53 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "On Fri, 13 Nov 1998, Thomas G. Lockhart wrote:\n\n> > I don't know what is implemented, but check out:\n> > http://www.vsb.cs.uni-frankfurt.de/~mico\n> > They \"claim\" a completely 2.2 Corba implementation...\n> \n> If someone were choosing an ORB, they perhaps could look at whatever the\n> Gnome project chose (either mico or ORBit, can't remember which). Also,\n> they didn't consider ILU because of licensing considerations, but the\n> license changed very recently and I think would now be a strong\n> candidate...\n\n\tWhy what Gnome chose vs what KDE is going with? braces for\npolitical discussions grin Then again, isn't the point of ORBs that\nyou don't have to pick one...?\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 13 Nov 1998 10:54:24 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "> Why what Gnome chose vs what KDE is going with? braces for\n> political discussions grin Then again, isn't the point of ORBs \n> that you don't have to pick one...?\n\nI lurk on the gnome mailing list because they are using SGML/DocBook for\ntheir docs efforts, as we are. Don't know if what they chose is the same\nas what KDE is using (assuming that KDE is using Corba stuff). afaik the\nGnome vs KDE discussion has been centered around the QT (graphics?)\nlicensing issue.\n\nYou do have to choose an ORB to work with (though perhaps it can be\nfairly straight-forward to switch to another). However, the ORB you\nchoose should be able to interoperate with other ORBs on the wire, so\nyou can mix and match clients and servers developed separately with\ndifferent ORBs.\n\n - Tom\n", "msg_date": "Fri, 13 Nov 1998 19:35:00 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "On Fri, 13 Nov 1998, The Hermit Hacker wrote:\n\n> political discussions* grin Then again, isn't the point of ORBs that\n> you don't have to pick one...?\n\nOne what?\nAs yet no one has said that an ORB is/does or what the letters stand for,\nor what Corba is/does or what that stands for.\n\nIt might be nice if some one mentioned in simple terms just what is all\nthe about, and what does it do?\n\nThanks\nTotally clueless\n\nTerry Mackintosh <[email protected]> http://www.terrym.com\nsysadmin/owner Please! No MIME encoded or HTML mail, unless needed.\n\nProudly powered by R H Linux 4.2, Apache 1.3, PHP 3, PostgreSQL 6.3\n-------------------------------------------------------------------\nSuccess Is A Choice ... book by Rick Patino, get it, read it!\n\n", "msg_date": "Fri, 13 Nov 1998 15:22:22 -0500 (EST)", "msg_from": "Terry Mackintosh <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "\n\nTerry Mackintosh wrote:\n\n> On Fri, 13 Nov 1998, The Hermit Hacker wrote:\n>\n> > political discussions* grin Then again, isn't the point of ORBs that\n> > you don't have to pick one...?\n>\n> One what?\n> As yet no one has said that an ORB is/does or what the letters stand for,\n> or what Corba is/does or what that stands for.\n\nCommon Object Request Broker Architecture\n\n> It might be nice if some one mentioned in simple terms just what is all\n> the about, and what does it do?\n\nBasically it is a standard way for objects written in different languages to\ncommunicate with each other.\n\n\nMark Bedish\[email protected]\n\n\n", "msg_date": "Fri, 13 Nov 1998 13:44:57 -0700", "msg_from": "Mark Bedish <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "On Fri, 13 Nov 1998, Mark Bedish wrote:\n\n> Basically it is a standard way for objects written in different languages to\n> communicate with each other.\n\n\tMy understanding is that it goes one better then that\neven...objects written in different languages on different platforms and\noperating systems..\n\nMarc G. Fournier \nSystems Administrator @ hub.org \nprimary: [email protected] secondary: scrappy@{freebsd\|postgresql}.org \n\n", "msg_date": "Fri, 13 Nov 1998 17:45:43 -0400 (AST)", "msg_from": "The Hermit Hacker <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "On Fri, Nov 13, 1998 at 02:49:53PM +0000, Thomas G. Lockhart wrote:\n> If someone were choosing an ORB, they perhaps could look at whatever the\n> Gnome project chose (either mico or ORBit, can't remember which). Also,\n\nGnome switched from mico to ORBit for performance reasons.\n\nMichael\n-- \nDr. Michael Meskes, Manager of the Western Branch Office, Datenrevision GmbH\nwork: Cuxhavener Str. 36, D-21149 Hamburg\nhome: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Sat, 14 Nov 1998 14:16:33 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" }, { "msg_contents": "On Fri, Nov 13, 1998 at 07:35:00PM +0000, Thomas G. Lockhart wrote:\n> I lurk on the gnome mailing list because they are using SGML/DocBook for\n> their docs efforts, as we are. Don't know if what they chose is the same\n> as what KDE is using (assuming that KDE is using Corba stuff). afaik the\n> Gnome vs KDE discussion has been centered around the QT (graphics?)\n> licensing issue.\n\nYes. There are ugly discussions going on. The problem with KDE is that they\nhappily incorporated GPLed code and linked it against qt, which RMS claims\nis illegal.\n\nMichael\n-- \nDr. Michael Meskes, Manager of the Western Branch Office, Datenrevision GmbH\nwork: Cuxhavener Str. 36, D-21149 Hamburg\nhome: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Sat, 14 Nov 1998 14:17:51 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] More CORBA and PostgreSQL" } ]
[ { "msg_contents": "The following set of patches will enable the dynamically loaded modules\nin PostgreSQL 6.4 to work properly.\n\nThese patches are actually the concatenated patches I've collected as a\ntentative update to the FreeBSD port collection. Those patches unrelated to\nthe ELF modules are from the previous port of postgresql. The help of anyone\nfamiliar enough with postgresql to inspect them and provide me with\nsuggestions for improvements would be welcomed.\n\nWith the patches in place, the regression tests pass (except for minor\ntextual differences in the depiction of -0.0, and some precision loss\nin floating point numbers). The tests in the plpgsql directory also run\nsuccessfully. PL/Tcl works sometimes, but as often as not the backend\ndies mysteriously - but in any case, not until after the dynamically loaded\nmodules are successfully loaded.\n\nCyrus\n\n\n--- pl/tcl/Makefile.orig\tFri Nov 13 00:08:57 1998\n+++ pl/tcl/Makefile\tFri Nov 13 00:14:50 1998\n@@ -44,8 +44,16 @@\n # they should work if the shared build of tcl was successful\n # on this system.\n #\n-%$(TCL_SHLIB_SUFFIX):\t%.o\n-\t$(TCL_SHLIB_LD) -o $@ $< $(TCL_SHLIB_LD_LIBS) $(TCL_LIB_SPEC) $(TCL_LIBS)\n+# XXX - This rule is already present - and this technique won't work anyway,\n+#\tsince the twice-eval'd value of TCL_SHLIB_LD in FreeBSD ELF:\n+#\n+#\t\tTCL_SHLIB_LD=ld -shared -x -soname $@\n+#\n+#\twill lose the trailing '$@', setting the soname to '-o' below and\n+#\tgenerally messing things up.\n+#\n+#%$(TCL_SHLIB_SUFFIX):\t%.o\n+#\t$(TCL_SHLIB_LD) -o $@ $< $(TCL_SHLIB_LD_LIBS) $(TCL_LIB_SPEC) $(TCL_LIBS)\n \n \n #\n--- makefiles/Makefile.freebsd.orig\tMon Nov 2 00:30:04 1998\n+++ makefiles/Makefile.freebsd\tTue Nov 10 19:28:17 1998\n@@ -1,13 +1,17 @@\n+ifdef ELF_SYSTEM\n+LDFLAGS+= -export-dynamic\n+endif\n+\n %.so: %.o\n+ifdef ELF_SYSTEM\n+\t$(LD) -x -shared -o $@ $<\n+else\n \t$(LD) -x -r -o $<.obj $<\n \t@echo building shared object $@\n \t@rm -f [email protected]\n \t@${AR} cq [email protected] `lorder $<.obj \| tsort`\n \t${RANLIB} [email protected]\n \t@rm -f $@\n-ifdef ELF_SYSTEM\n-\t$(LD) -x -Bshareable -o $@ [email protected]\n-else\n \t$(LD) -x -Bshareable -Bforcearchive -o $@ [email protected]\n endif\n \n--- backend/port/dynloader/freebsd.c.orig\tMon Oct 26 23:41:29 1998\n+++ backend/port/dynloader/freebsd.c\tThu Nov 12 23:39:36 1998\n@@ -83,11 +83,13 @@\n \tvoid\t vp;\n \tchar\t\tbuf[BUFSIZ];\n \n+#ifndef __ELF__\n \tif (name != '_')\n \t{\n \t\tsprintf(buf, \"_%s\", name);\n \t\tname = buf;\n \t}\n+#endif\n \tif ((vp = dlsym(handle, (char ) name)) == (void ) NULL)\n \t\tsprintf(error_message, \"dlsym (%s) failed\", name);\n \treturn vp;\n--- template/freebsd.orig\tSat Nov 7 05:45:29 1998\n+++ template/freebsd\tSat Nov 7 05:52:31 1998\n@@ -1,9 +1,9 @@\n AROPT:cq\n SHARED_LIB:-fpic -DPIC\n-CFLAGS:-O2 -m486 -pipe\n+CFLAGS:-pipe -O2\n SRCH_INC:\n SRCH_LIB:\n-USE_LOCALE:no\n+USE_LOCALE:yes\n DLSUFFIX:.so\n YFLAGS:-d\n YACC:bison -y\n--- bin/pg_passwd/pg_passwd.c.orig\tSat Jan 31 19:09:26 1998\n+++ bin/pg_passwd/pg_passwd.c\tSat Jan 31 19:15:43 1998\n@@ -23,12 +23,16 @@\n \n #endif\n \n+#ifndef _POSIX_SOURCE\n+# define\t_PASSWORD_LEN\t128\t/* max length, not containing NULL /\n+#endif\n+\n char\t comname;\n void\t\tusage(FILE stream);\n void\t\tread_pwd_file(char filename);\n void\t\twrite_pwd_file(char filename, char bkname);\n-void\t\tencrypt_pwd(char key[9], char salt[3], char passwd[14]);\n-int\t\t\tcheck_pwd(char key[9], char passwd[14]);\n+void\t\tencrypt_pwd(char key[9], char salt[3], char passwd[_PASSWORD_LEN+1]);\n+int\t\t\tcheck_pwd(char key[9], char passwd[_PASSWORD_LEN+1]);\n void\t\tprompt_for_username(char username);\n void\t\tprompt_for_password(char prompt, char password);\n \n@@ -148,7 +152,7 @@\n \n \t\tif (q != NULL)\n \t\t\t(q++) = '\\0';\n-\t\tif (strlen(p) != 13)\n+\t\tif (strlen(p) > _PASSWORD_LEN)\n \t\t{\n \t\t\tfprintf(stderr, \"WARNING: %s: line %d: illegal password length.\\n\",\n \t\t\t\t\tfilename, npwds + 1);\n@@ -208,7 +212,7 @@\n }\n \n void\n-encrypt_pwd(char key[9], char salt[3], char passwd[14])\n+encrypt_pwd(char key[9], char salt[3], char passwd[_PASSWORD_LEN+1])\n {\n \tint\t\t\tn;\n \n@@ -242,9 +246,9 @@\n }\n \n int\n-check_pwd(char key[9], char passwd[14])\n+check_pwd(char key[9], char passwd[_PASSWORD_LEN+1])\n {\n-\tchar\t\tshouldbe[14];\n+\tchar\t\tshouldbe[_PASSWORD_LEN+1];\n \tchar\t\tsalt[3];\n \n \tsalt[0] = passwd[0];\n@@ -252,7 +256,7 @@\n \tsalt[2] = '\\0';\n \tencrypt_pwd(key, salt, shouldbe);\n \n-\treturn strncmp(shouldbe, passwd, 13) == 0 ? 1 : 0;\n+\treturn strncmp(shouldbe, passwd, _PASSWORD_LEN) == 0 ? 1 : 0;\n }\n \n void\n@@ -326,7 +330,7 @@\n \tchar\t\tsalt[3];\n \tchar\t\tkey[9],\n \t\t\t\tkey2[9];\n-\tchar\t\te_passwd[14];\n+\tchar\t\te_passwd[_PASSWORD_LEN+1];\n \tint\t\t\ti;\n \n \tcomname = argv[0];\n--- interfaces/libpgtcl/Makefile.in.orig\tSun Oct 18 20:00:41 1998\n+++ interfaces/libpgtcl/Makefile.in\tSat Nov 7 05:12:43 1998\n@@ -27,7 +27,7 @@\n \n OBJS= pgtcl.o pgtclCmds.o pgtclId.o\n \n-SHLIB_LINK= -L../libpq -lpq\n+SHLIB_LINK= -L../libpq -lpq -lcrypt\n \n # Shared library stuff, also default 'all' target\n include $(SRCDIR)/Makefile.shlib\n", "msg_date": "Fri, 13 Nov 98 02:00:39 -0500", "msg_from": "[email protected] (Cyrus Rahman)", "msg_from_op": true, "msg_subject": "Patches for FreeBSD/ELF" } ]
[ { "msg_contents": "\nhi!\n\nI just checked postgres 6.4 and I have to tell you it was a great work!!!\nit's 3-4 times faster than 6.3.2 was. more stable. \nI have a little question:\nthe TEXT fieldtype how many data can store?\n8192 bytes?\n\nI also wonder what's the reason to limit a query to 8192bytes maximum..\nis there any serious problem to have longer queries?\n\nit could be good to have more than 8K fields (without lo_import and so on)\nfor example 64K...\n\nthis is would be good sometimes to store texts...\n\n\tregards,\t\n\t\tredax\n\n.----------------------------------------------------------.\n\|Zsolt Varga \| tel/fax: +36 36 422811 \|\n\| AgriaComputer LTD \| email: [email protected] \|\n\| System Administrator \| URL: http://www.agria.hu/ \|\n`----------------------------------------------------------'\n\n", "msg_date": "Fri, 13 Nov 1998 09:29:47 +0100 (CET)", "msg_from": "Zsolt Varga <[email protected]>", "msg_from_op": true, "msg_subject": "still Query Limits to 8K ?" }, { "msg_contents": "> I also wonder what's the reason to limit a query to 8192bytes \n> maximum..\n> is there any serious problem to have longer queries?\n\nI would think you could find and increase the size of string buffers\nused in psql, libpq, and the backend.\n\n> it could be good to have more than 8K fields (without lo_import and so \n> on) for example 64K...\n\nAnd Darren K (?) worked on parameterizing the maximum allowed tuple\nsize. You may just be able to increase it, rebuild, and reinstall.\nPerhaps you could check the mail archives to see where it was left.\n\n - Tom\n", "msg_date": "Fri, 13 Nov 1998 14:54:49 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] still Query Limits to 8K ?" }, { "msg_contents": "Zsolt Varga <[email protected]> writes:\n> the TEXT fieldtype how many data can store?\n> 8192 bytes?\n\nYup, same as it ever was. (Actually a bit less than 8K, since you have\nto subtract the size of the tuple overhead, not to mention any other\nfields that might be in the same tuple.) In reality you probably don't\nwant to design a database in which tuples are bigger than a K or so.\nIf your tuples run 4K or more then you can only fit one to a disk page,\nwhich means you waste a lot of disk space.\n\nThere has been talk of allowing tuples to span multiple disk pages,\nbut it's not a simple change.\n\n> I also wonder what's the reason to limit a query to 8192bytes maximum..\n> is there any serious problem to have longer queries?\n\nThat's a completely separate issue from the maximum tuple size on disk.\nIt's driven by the size of the text buffers used to hold the current\nquery.\n\nOffhand I think it would not be hard to fix libpq and the backend to\nnot have any particular limit on the textual length of a query. They'd\nstill need to have buffers holding the query, but they could realloc()\nthe buffers bigger as needed. (So, for example, libpq's buffer size\nwould be proportional to the longest query you'd asked in a given\nsession. But the initial buffer size could be much smaller than 8K,\nso for the average frontend app this would actually save memory.)\n\nOne thing I do not know is whether we'd start to run into any hard-wired\nlimits in the parser/planner/executor with very large (complex) queries\n... Thomas, would the parser go down if you handed it a SELECT with a\nfew thousand OR clauses?\n\nI'd be willing to fix libpq and the backend as far as the parser input\nbuffer, if someone else will take responsibility for looking at any\nhard-wired limits that might exist downstream of that.\n\n\t\t\tregards, tom lane\n", "msg_date": "Fri, 13 Nov 1998 10:43:23 -0500", "msg_from": "Tom Lane <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] still Query Limits to 8K ? " }, { "msg_contents": "> One thing I do not know is whether we'd start to run into any \n> hard-wired limits in the parser/planner/executor with very large \n> (complex) queries ... Thomas, would the parser go down if you handed \n> it a SELECT with a few thousand OR clauses?\n\nI'm not recalling any specific limitation here. And if there is one,\nthen we should parameterize it so it is easy to find and easy to pump\nup.\n\n> I'd be willing to fix libpq and the backend as far as the parser input\n> buffer, if someone else will take responsibility for looking at any\n> hard-wired limits that might exist downstream of that.\n\nSure, parameterize everything (if it isn't done already) and then we can\ntry pumping it up and see what breaks. I'll push on any problems we find\ndeeper inside the backend.\n\n - Tom\n", "msg_date": "Fri, 13 Nov 1998 19:41:06 +0000", "msg_from": "\"Thomas G. Lockhart\" <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] still Query Limits to 8K ?" }, { "msg_contents": "Tom Lane wrote:\n> One thing I do not know is whether we'd start to run into any hard-wired\n> limits in the parser/planner/executor with very large (complex) queries\n> ... Thomas, would the parser go down if you handed it a SELECT with a\n> few thousand OR clauses?\n\nYep, I know one:\n\ninclude/utils/elog.h:32:#define ELOG_MAXLEN 4096\n\nI made a patch before 6.4, which was backed out because\nI used vsnprintf which wasn't available everywhere.\n\nThis is still to short, suggests we do this for 6.4.1:\n\n#define ELOG_MAXLEN 8192\n\nAnd use my patch (with fixes) for 6.5.\n\n\tregards,\n-- \n-----------------\nGï¿½ran Thyni\nhttp://kirra.net/\n", "msg_date": "Sat, 21 Nov 1998 12:14:22 +0100", "msg_from": "Goran Thyni <[email protected]>", "msg_from_op": false, "msg_subject": "Re: [HACKERS] still Query Limits to 8K ?" } ]
[ { "msg_contents": "I've just been send the following problem report:\n\n> >1. To store the data file I am using lo_import function of large\n> >object interface. As I am storing(importing into database) the file\n> >even 100 bytes only. During the importing process it creates two files\n> >into the database area xinv,xinx .I don't know why it is taking the\n> >space of 16384 Bytes for each xinv* xinx* file in the database.\n> > This much space consumption is unmanageable. Please help me.\n> >Actually how this import the file I don't know.\n> >Thanks alot\n> ...\n\nI am sending a sample source code.\n 1 #include <stdio.h>\n 2 #include \"libpq-fe.h\"\n 3 #include <stdlib.h>\n 4 char shotno[10];\n 5 main()\n 6 {\n 7 char filename[35];\n 8 char pghost,pgport,pgoptions,pgtty,query2[100];\n 9 char dbName, shotno[10];\n 10 PGconn conn;\n 11 PGresult res2;\n 12 int shot,gchno;\n 13 pghost = \"sun9\"; / host name of the backend server /\n 14 pgport = \"5432\"; / port of the backend server /\n 15 pgoptions = NULL; / special options to start up the backend /\n 16 pgtty = NULL; / debugging tty for the backend server /\n 17 dbName = \"aditya\";\n 18 / make a connection to the database /\n 19 conn = PQsetdb(pghost, pgport, pgoptions, pgtty, dbName);\n 20 / check to see that the backend connection was successfully made /\n 21 if (PQstatus(conn) == CONNECTION_BAD)\n 22 {\n 23 fprintf(stderr,\"Connection to database '%s' failed.\\n\", dbName);\n 24 fprintf(stderr,\"%s\", PQerrorMessage(conn));\n 25 exit(0);\n 26 }\n 27 printf(\"\\nEnter the shot no... \");\n 28 scanf(\"%s\",shotno);\n 29 shot=atoi(shotno);\n 30 gchno=200;\n 31 filename=\"/users13/aditya/vax/a05966.lgg\";\n 32 sprintf(query2,\"insert into ch_data values ( %d ,%d , lo_import('%s')) ; \" , shot ,gchno , filename);\n 33 res2 = PQexec(conn,query2);\n 34 }\n 35 printf(\"\\n The Data File For This Shot has transferred 'OK' \\n\");\n 36 }\n--\nIn this code the file I am importing \"/users13/aditya/vax/a05966.lgg\"\nhas a size of 3328 bytes only. This file consumes the space in database\nis of 216384 bytes. It creates two files xinv* and xinx* in database of\nsize 16384 bytes each.\n\nsun9:/work/postgres/local/data/base/aditya>$ ls -laF\n-rw------- 1 postgres guest 16384 Nov 12 10:37 xinx154230\n-rw------- 1 postgres guest 16384 Nov 12 10:37 xinv154230\nHow can I reduce the size of the file in database.\n\t\t\t\t\tH.K.Gulati\t\t\t\t\t\t\t\t\[email protected]\n\t\t\t\t\tEngineer - SC\n\t\t\t\t\tInstitute For Plasma Research\n\t\t\t\t\tBhat Gandhinagar(GUJARAT)\n\nMichael\n\n-- \nDr. Michael Meskes, Manager of the Western Branch Office, Datenrevision GmbH\nwork: Cuxhavener Str. 36, D-21149 Hamburg\nhome: Th.-Heuss-Str. 61, D-41812 Erkelenz, [email protected]\nGo SF49ers! Go Rhein Fire! Use Debian GNU/Linux!\n\n\n", "msg_date": "Fri, 13 Nov 1998 11:32:38 +0100", "msg_from": "Michael Meskes <[email protected]>", "msg_from_op": true, "msg_subject": "lo_import problem" } ]

Subsets and Splits

No community queries yet

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