postgresql.git
24 years agoBack-patch nodeMaterial to honor chgParam by recomputing its output.
Tom Lane [Fri, 6 Oct 2000 01:28:47 +0000 (01:28 +0000)]
Back-patch nodeMaterial to honor chgParam by recomputing its output.

24 years agoavoid database-wide restart on write error
Hiroshi Inoue [Wed, 4 Oct 2000 07:50:00 +0000 (07:50 +0000)]
avoid database-wide restart on write error

24 years agoBack-patch fix for TRUNCATE failure on relations with indexes.
Tom Lane [Sat, 30 Sep 2000 18:47:07 +0000 (18:47 +0000)]
Back-patch fix for TRUNCATE failure on relations with indexes.

24 years agoBack-patch fix to grab read lock on a buffer while it is written out.
Tom Lane [Mon, 25 Sep 2000 04:34:10 +0000 (04:34 +0000)]
Back-patch fix to grab read lock on a buffer while it is written out.

24 years agoBack-patch code to deduce implied equalities from transitivity of
Tom Lane [Sat, 23 Sep 2000 23:50:47 +0000 (23:50 +0000)]
Back-patch code to deduce implied equalities from transitivity of
mergejoin clauses, and add these equalities to the given WHERE clauses.
This is necessary to ensure that sort keys we think are equivalent
really are equivalent as soon as their rels have been joined.  Without
this, 7.0 may create an incorrect mergejoin plan.

24 years agoBack-patch fix for subselect in targetlist of Append node.
Tom Lane [Sat, 23 Sep 2000 23:41:05 +0000 (23:41 +0000)]
Back-patch fix for subselect in targetlist of Append node.

24 years agoBack-patch fix to make pg_dump dump 'iscachable' flag for functions.
Tom Lane [Sat, 23 Sep 2000 23:36:17 +0000 (23:36 +0000)]
Back-patch fix to make pg_dump dump 'iscachable' flag for functions.

24 years agoBack-patch fix that allows AllocateFile() to return errno=ENFILE/EMFILE
Tom Lane [Sat, 23 Sep 2000 23:31:24 +0000 (23:31 +0000)]
Back-patch fix that allows AllocateFile() to return errno=ENFILE/EMFILE
after we are no longer able to close any more VFDs.  This is needed to
avoid postmaster crash under out-of-file-descriptors conditions.

24 years agoBack-patch fix for '.' not in PATH at build time, per SL Baur.
Tom Lane [Sat, 23 Sep 2000 22:51:19 +0000 (22:51 +0000)]
Back-patch fix for '.' not in PATH at build time, per SL Baur.

24 years agoBack-patch fix to include kernel errno message in all smgr elog messages.
Tom Lane [Sat, 23 Sep 2000 22:17:12 +0000 (22:17 +0000)]
Back-patch fix to include kernel errno message in all smgr elog messages.

24 years agoBack-patch fix for 'Sorcerer's Apprentice' syndrome wherein md.c would
Tom Lane [Sat, 23 Sep 2000 22:11:41 +0000 (22:11 +0000)]
Back-patch fix for 'Sorcerer's Apprentice' syndrome wherein md.c would
create a vast quantity of zero-length files if asked to access a block
number far beyond the actual end of a relation.

24 years agoBack-patch fix for erroneous use of strcmp().
Tom Lane [Sat, 23 Sep 2000 21:32:55 +0000 (21:32 +0000)]
Back-patch fix for erroneous use of strcmp().

24 years agoBack-patch fix for erroneous selectivity of not-equals.
Tom Lane [Sat, 23 Sep 2000 21:27:05 +0000 (21:27 +0000)]
Back-patch fix for erroneous selectivity of not-equals.

24 years agoApply Jeroen van Vianen's patch for failure to check heap_openr failure
Tom Lane [Sat, 23 Sep 2000 21:15:26 +0000 (21:15 +0000)]
Apply Jeroen van Vianen's patch for failure to check heap_openr failure
in RI triggers.  This is fixed in another way in current sources, but
might as well apply this patch to REL7_0 branch so that 7.0.3 need not
suffer this crash.

24 years agoBack-patch fix to copy sub-Query nodes before planning them. This
Tom Lane [Sat, 23 Sep 2000 21:00:05 +0000 (21:00 +0000)]
Back-patch fix to copy sub-Query nodes before planning them.  This
fixes problems with subselects appearing in contexts like COALESCE or
BETWEEN, where parser will make multiple links to same subexpression.

24 years agoBack-patch fix to remove bogus use of int4out().
Tom Lane [Sat, 23 Sep 2000 19:30:04 +0000 (19:30 +0000)]
Back-patch fix to remove bogus use of int4out().

24 years agoBack-patch fix for erroneous free() of getpwuid() result.
Tom Lane [Sat, 23 Sep 2000 19:17:06 +0000 (19:17 +0000)]
Back-patch fix for erroneous free() of getpwuid() result.

24 years agoCancel request while waiting for a lock should try to wake
Hiroshi Inoue [Wed, 20 Sep 2000 07:20:27 +0000 (07:20 +0000)]
Cancel request while waiting for a lock should try to wake
up sleeping processes.

24 years agoBack-patch fix to ensure that VACUUM always calls FlushRelationBuffers.
Tom Lane [Tue, 19 Sep 2000 21:01:04 +0000 (21:01 +0000)]
Back-patch fix to ensure that VACUUM always calls FlushRelationBuffers.

24 years agoAllow PL/pgSQL accept non ascii identifiers
Tatsuo Ishii [Fri, 15 Sep 2000 12:08:56 +0000 (12:08 +0000)]
Allow PL/pgSQL accept non ascii identifiers

24 years agoBack-patch fix for bogus plans involving non-mark/restorable plan
Tom Lane [Fri, 8 Sep 2000 02:11:33 +0000 (02:11 +0000)]
Back-patch fix for bogus plans involving non-mark/restorable plan
as inner plan of a mergejoin.

24 years agoFix psql crash. If MULTIBYTE is enabled, \l+ dumps core due to
Tatsuo Ishii [Thu, 7 Sep 2000 04:53:32 +0000 (04:53 +0000)]
Fix psql crash. If MULTIBYTE is enabled, \l+ dumps core due to
SQL buffer in listAllDbs is just too small.

24 years agoBack-patch fix to ensure we abort any open transaction at backend exit.
Tom Lane [Wed, 30 Aug 2000 21:19:32 +0000 (21:19 +0000)]
Back-patch fix to ensure we abort any open transaction at backend exit.

24 years agooops, in v7.x its USE_SYSLOG, not ENABLE_SYSLOG
Marc G. Fournier [Mon, 28 Aug 2000 20:27:32 +0000 (20:27 +0000)]
oops, in v7.x its USE_SYSLOG, not ENABLE_SYSLOG
modify config.h.in so that it gets set by configure properly

24 years agobackpatch the --enable-syslog functionality to REL7_0 branch
Marc G. Fournier [Mon, 28 Aug 2000 20:04:39 +0000 (20:04 +0000)]
backpatch the --enable-syslog functionality to REL7_0 branch

24 years agoBack-patch primary fix for planner recursion bug.
Tom Lane [Thu, 27 Jul 2000 23:53:29 +0000 (23:53 +0000)]
Back-patch primary fix for planner recursion bug.

24 years agoSELECT ... FOR UPDATE neglects duplicate key checking.
Tatsuo Ishii [Wed, 26 Jul 2000 13:09:30 +0000 (13:09 +0000)]
SELECT ... FOR UPDATE neglects duplicate key checking.
patches submitted by Hiroshi Inoue.

24 years agoBackpatch backwards-index-scan fix.
Tom Lane [Thu, 13 Jul 2000 05:52:18 +0000 (05:52 +0000)]
Backpatch backwards-index-scan fix.

24 years agoBack-patch StrNCpy fix.
Tom Lane [Fri, 7 Jul 2000 21:29:57 +0000 (21:29 +0000)]
Back-patch StrNCpy fix.

25 years agoBack-patch COPY WITH OIDS leak fix.
Tom Lane [Wed, 28 Jun 2000 06:13:01 +0000 (06:13 +0000)]
Back-patch COPY WITH OIDS leak fix.

25 years agoUpdate FAQ.
Bruce Momjian [Sat, 24 Jun 2000 01:51:33 +0000 (01:51 +0000)]
Update FAQ.

25 years agoBack-patch large-object fix.
Tom Lane [Thu, 15 Jun 2000 06:10:27 +0000 (06:10 +0000)]
Back-patch large-object fix.

25 years agoBackpatch jdbc fixes into 7.0.X.
Bruce Momjian [Thu, 15 Jun 2000 04:12:41 +0000 (04:12 +0000)]
Backpatch jdbc fixes into 7.0.X.

25 years agoFix markup to allow doc building.
Thomas G. Lockhart [Wed, 14 Jun 2000 13:18:59 +0000 (13:18 +0000)]
Fix markup to allow doc building.

25 years agoBrand 7.1 release. Also update jdbc version in release branch.
Bruce Momjian [Mon, 12 Jun 2000 22:37:07 +0000 (22:37 +0000)]
Brand 7.1 release.  Also update jdbc version in release branch.

25 years agoupdate developers faq
Bruce Momjian [Sat, 10 Jun 2000 02:05:55 +0000 (02:05 +0000)]
update developers faq

25 years agoUpdate bsdi faq.
Bruce Momjian [Fri, 9 Jun 2000 16:35:54 +0000 (16:35 +0000)]
Update bsdi faq.

25 years agoUpdate FAQ.
Bruce Momjian [Fri, 9 Jun 2000 12:20:59 +0000 (12:20 +0000)]
Update FAQ.

25 years agoBackpatch SearchSysCacheTupleCopy fix for init_fcache.
Tom Lane [Tue, 6 Jun 2000 17:45:05 +0000 (17:45 +0000)]
Backpatch SearchSysCacheTupleCopy fix for init_fcache.

25 years agoBack-patch typeTypeName() fix into REL7_0.
Tom Lane [Tue, 6 Jun 2000 16:51:24 +0000 (16:51 +0000)]
Back-patch typeTypeName() fix into REL7_0.

25 years agoUpdate install for upgraders. REL7_0_2
Bruce Momjian [Mon, 5 Jun 2000 17:02:27 +0000 (17:02 +0000)]
Update install for upgraders.

25 years agoFix error message wording
Bruce Momjian [Mon, 5 Jun 2000 11:13:55 +0000 (11:13 +0000)]
Fix error message wording

25 years agoUpdate for 7.0.2.
Bruce Momjian [Mon, 5 Jun 2000 10:59:19 +0000 (10:59 +0000)]
Update for 7.0.2.

25 years agoupdate
Bruce Momjian [Fri, 2 Jun 2000 16:34:42 +0000 (16:34 +0000)]
update

25 years agoUpdate faq's.
Bruce Momjian [Fri, 2 Jun 2000 02:28:13 +0000 (02:28 +0000)]
Update faq's.

25 years agoAdd fix to index length
Bruce Momjian [Fri, 2 Jun 2000 01:26:29 +0000 (01:26 +0000)]
Add fix to index length

25 years agoLexer defended us against overlength plain identifiers, but not against
Tom Lane [Thu, 1 Jun 2000 22:23:10 +0000 (22:23 +0000)]
Lexer defended us against overlength plain identifiers, but not against
overlength quoted identifiers.  Death and destruction ensue...

25 years agoUpdate 7.0.1 for Tatsuo
Bruce Momjian [Thu, 1 Jun 2000 06:03:49 +0000 (06:03 +0000)]
Update 7.0.1 for Tatsuo

25 years agoMention pg_upgrade
Bruce Momjian [Thu, 1 Jun 2000 05:58:56 +0000 (05:58 +0000)]
Mention pg_upgrade

25 years agoFixups for 7.0.1
Bruce Momjian [Thu, 1 Jun 2000 05:14:35 +0000 (05:14 +0000)]
Fixups for 7.0.1

25 years agoBack-patch fix to recognize clause pairs involving Params
Tom Lane [Wed, 31 May 2000 15:43:31 +0000 (15:43 +0000)]
Back-patch fix to recognize clause pairs involving Params
as being range queries.

25 years agoUpdate pg_hba.conf with more examples
Bruce Momjian [Tue, 30 May 2000 17:40:14 +0000 (17:40 +0000)]
Update pg_hba.conf with more examples

25 years agoUpdate pg_hba.conf.sample with better examples and descriptions
Bruce Momjian [Tue, 30 May 2000 17:18:43 +0000 (17:18 +0000)]
Update pg_hba.conf.sample with better examples and descriptions

25 years agoCopy pg_ident.conf.sample into /lib directory. Rename USERAUTH to
Bruce Momjian [Tue, 30 May 2000 16:36:34 +0000 (16:36 +0000)]
Copy pg_ident.conf.sample into /lib directory.  Rename USERAUTH to
AUTHTYPE in config file.  Patch both branches.

25 years agoBack-patch tz name length fixes into REL7_0.
Tom Lane [Mon, 29 May 2000 19:19:07 +0000 (19:19 +0000)]
Back-patch tz name length fixes into REL7_0.

25 years agoarray_map() needs to guard against a NULL input array. This problem
Tom Lane [Sun, 28 May 2000 17:43:34 +0000 (17:43 +0000)]
array_map() needs to guard against a NULL input array.  This problem
should go away in 7.1 with the new fmgr, but for 7.0.1 ...

25 years agofastpath neglected to pass a valid isNull flag to the function being
Tom Lane [Sun, 28 May 2000 15:40:25 +0000 (15:40 +0000)]
fastpath neglected to pass a valid isNull flag to the function being
called.  We hadn't noticed this because C-coded functions mostly don't
pay attention to isNull anyway ... but plpgsql functions sure do ...
This is fixed in another way for 7.1, but 7.0.1 needs this patch.

25 years agoMake EUC_CN support more robust.
Tatsuo Ishii [Sun, 28 May 2000 04:57:06 +0000 (04:57 +0000)]
Make EUC_CN support more robust.

25 years agoChange date of documents.
Tatsuo Ishii [Sun, 28 May 2000 04:28:55 +0000 (04:28 +0000)]
Change date of documents.

25 years agofix typo
Bruce Momjian [Sun, 28 May 2000 02:02:07 +0000 (02:02 +0000)]
fix typo

25 years agoAdd README.kerbros
Bruce Momjian [Sat, 27 May 2000 03:42:49 +0000 (03:42 +0000)]
Add README.kerbros

25 years agoImprove pg_hba.conf examples
Bruce Momjian [Sat, 27 May 2000 01:32:01 +0000 (01:32 +0000)]
Improve pg_hba.conf examples

25 years agoReduce eqsel()'s fudge-factor for estimating the frequency of values
Tom Lane [Fri, 26 May 2000 17:19:15 +0000 (17:19 +0000)]
Reduce eqsel()'s fudge-factor for estimating the frequency of values
other than the most common value in a column.  We had had 0.5, make it
0.1 to make it more likely that an indexscan will be chosen.  Really
need better statistics instead, but this should stem the bleeding
meanwhile ...

25 years agoFix \h SELECT to not show SELECT INTO.
Bruce Momjian [Fri, 26 May 2000 15:47:18 +0000 (15:47 +0000)]
Fix \h SELECT to not show SELECT INTO.

25 years agoFix too long syslog message problem
Tatsuo Ishii [Fri, 26 May 2000 12:17:48 +0000 (12:17 +0000)]
Fix too long syslog message problem

25 years agoGenerate a reasonable error message when an aggregate function is applied
Tom Lane [Fri, 26 May 2000 03:56:40 +0000 (03:56 +0000)]
Generate a reasonable error message when an aggregate function is applied
to an undecorated relation name (cf. example from Ed Loehr, 5/25/00).

25 years agoSpell fix
Bruce Momjian [Fri, 26 May 2000 03:15:18 +0000 (03:15 +0000)]
Spell fix

25 years agoFix initlocation syntax example for createdb
Bruce Momjian [Fri, 26 May 2000 02:07:15 +0000 (02:07 +0000)]
Fix initlocation syntax example for createdb

25 years agoMore paranoia about global variables containing references to long-
Tom Lane [Fri, 26 May 2000 01:38:08 +0000 (01:38 +0000)]
More paranoia about global variables containing references to long-
since-closed file descriptors...

25 years agoAfter closing frontend socket, set MyProcPort->sock = -1 to ensure that
Tom Lane [Fri, 26 May 2000 01:26:19 +0000 (01:26 +0000)]
After closing frontend socket, set MyProcPort->sock = -1 to ensure that
subsequent I/O attempts fail cleanly.  I'm speculating about failure
scenarios in which we do pq_close, then something in a proc_exit routine
opens a file (re-using that kernel FD number), then something else
fails and tries to write an elog message to the frontend ... message
ends up in opened file, oops.  No known examples of this but it seems
like a potential hole.

25 years agoAdd some elog(DEBUG)'s to help diagnose mdblindwrt failures.
Tom Lane [Thu, 25 May 2000 23:30:20 +0000 (23:30 +0000)]
Add some elog(DEBUG)'s to help diagnose mdblindwrt failures.

25 years agoClean up sloppy coding of _outAExpr().
Tom Lane [Thu, 25 May 2000 22:43:12 +0000 (22:43 +0000)]
Clean up sloppy coding of _outAExpr().

25 years agoModify raw parsetree representation returned by gram.y for SubLinks:
Tom Lane [Thu, 25 May 2000 22:42:19 +0000 (22:42 +0000)]
Modify raw parsetree representation returned by gram.y for SubLinks:
the oper field should be a valid Node structure so it can be dumped by
outfuncs.c without risk of coredump.  (We had been using a raw pointer
to character string, which surely is NOT a valid Node.)  This doesn't
cause any backwards compatibility problems for stored rules, since
raw unanalyzed parsetrees are never stored.

25 years agoDo table renaming in a sane order: physical file rename must happen
Tom Lane [Thu, 25 May 2000 21:30:20 +0000 (21:30 +0000)]
Do table renaming in a sane order: physical file rename must happen
*last*, after all updating of system catalogs.  In old code, an error
detected during TypeRename left the relation hosed.  Also, add a call
to flush the relation's relcache entry, rather than trusting to shared
cache invalidation to flush it for us.

25 years agoheap_drop() should flush the relcache entry for the
Tom Lane [Thu, 25 May 2000 21:25:32 +0000 (21:25 +0000)]
heap_drop() should flush the relcache entry for the
relation being dropped.

25 years agoQuote database name so that not-all-lowercase names are handled safely.
Tom Lane [Thu, 25 May 2000 20:18:15 +0000 (20:18 +0000)]
Quote database name so that not-all-lowercase names are handled safely.

25 years ago! * Users doing non-blocking connections need to handle the reset
Bruce Momjian [Thu, 25 May 2000 19:09:55 +0000 (19:09 +0000)]
!        * Users doing non-blocking connections need to handle the reset
!        * themselves, they'll need to check the connection status if we
!        * return an error.

Alfred Perlstein

25 years agoUpdate TODO list.
Bruce Momjian [Thu, 25 May 2000 16:10:01 +0000 (16:10 +0000)]
Update TODO list.

25 years agoSpell fix
Bruce Momjian [Thu, 25 May 2000 15:32:03 +0000 (15:32 +0000)]
Spell fix

25 years agoOn solaris, createdb/dropdb fails because of strange behavior of system().
Tatsuo Ishii [Thu, 25 May 2000 06:53:43 +0000 (06:53 +0000)]
On solaris, createdb/dropdb fails because of strange behavior of system().
(it returns error with errno ECHILD upon successful completion of commands).
This fix ignores an error from system() if errno == ECHILD.

25 years agoMake the world safe for QNX's busted shell ...
Tom Lane [Wed, 24 May 2000 22:43:59 +0000 (22:43 +0000)]
Make the world safe for QNX's busted shell ...

25 years agoMake the world safe for QNX's busted shell ...
Tom Lane [Wed, 24 May 2000 22:32:59 +0000 (22:32 +0000)]
Make the world safe for QNX's busted shell ...

25 years agoadd a --with-setproctitle switch to configure that leaves the use of setproctitle()
Marc G. Fournier [Wed, 24 May 2000 14:58:21 +0000 (14:58 +0000)]
add a --with-setproctitle switch to configure that leaves the use of setproctitle()
disabled by default ...

25 years agoUpdate so init displays status with setproctitle too, like the rest of them.
Bruce Momjian [Wed, 24 May 2000 00:24:17 +0000 (00:24 +0000)]
Update so init displays status with setproctitle too, like the rest of them.

25 years agoMake setproctitle update for every query.
Bruce Momjian [Wed, 24 May 2000 00:14:26 +0000 (00:14 +0000)]
Make setproctitle update for every query.

25 years agoComment out no-op ps updates for setproctitle(), until we figure out
Bruce Momjian [Tue, 23 May 2000 22:50:55 +0000 (22:50 +0000)]
Comment out no-op ps updates for setproctitle(), until we figure out
what to do.

25 years agocomment cleanup
Bruce Momjian [Tue, 23 May 2000 22:16:27 +0000 (22:16 +0000)]
comment cleanup

25 years agopg_ctl does not use -S by default anymore so update documentation
Bruce Momjian [Tue, 23 May 2000 21:46:57 +0000 (21:46 +0000)]
pg_ctl does not use -S by default anymore so update documentation

25 years agoFix problem in which sloppily-coded test in ExecInitIndexScan would
Tom Lane [Tue, 23 May 2000 16:56:37 +0000 (16:56 +0000)]
Fix problem in which sloppily-coded test in ExecInitIndexScan would
think that both sides of indexqual look like index keys.  An example is
create table inside (f1 float8 primary key);
create table outside (g1 float8, g2 float8);
select * from inside,outside where f1 = atan2(g1+1, g2);
ERROR:  ExecInitIndexScan: both left and right ops are rel-vars
(note that failure is potentially platform-dependent).  Solution is a
cleanup I had had in mind to make anyway: functional index keys should
be represented as Var nodes in the fixed indexqual, just like regular
index keys.

25 years agoTwiddle C++ configuration checks so that we #include <string> (if
Tom Lane [Tue, 23 May 2000 04:55:44 +0000 (04:55 +0000)]
Twiddle C++ configuration checks so that we #include <string> (if
it exists) before testing 'using namespace std'.  This is necessary
on some C++ setups where the compiler won't take a 'using' until
you've included a header that mentions namespace std.  (Pretty braindead
if you ask me, but...)

25 years agoTransmit -I switches selected by configure (particularly those from
Tom Lane [Tue, 23 May 2000 02:12:46 +0000 (02:12 +0000)]
Transmit -I switches selected by configure (particularly those from
--with-includes) to makefiles for pltcl and plperl, so that these
switches will be used even though we do not want other top-level
CFLAGS.  Ain't it fun trying to support multiple-compiler platforms?

25 years agoCast parameters to ckfree to (char *) to suppress compiler warnings.
Tom Lane [Tue, 23 May 2000 01:59:05 +0000 (01:59 +0000)]
Cast parameters to ckfree to (char *) to suppress compiler warnings.

25 years agoIrix fixes from Murad Nayal
Bruce Momjian [Mon, 22 May 2000 23:56:44 +0000 (23:56 +0000)]
Irix fixes from Murad Nayal

25 years agoReformatted the install file as it used to be
Peter Eisentraut [Mon, 22 May 2000 22:04:47 +0000 (22:04 +0000)]
Reformatted the install file as it used to be

25 years agoCleanup
Bruce Momjian [Mon, 22 May 2000 02:41:39 +0000 (02:41 +0000)]
Cleanup

25 years agoI am attempting to integrate postgres (v 7.0) with an open source
Bruce Momjian [Mon, 22 May 2000 02:34:23 +0000 (02:34 +0000)]
I am attempting to integrate postgres (v 7.0) with an open source
project I am working on (Recall - a distributed, fault-tolerant,
replicated, storage framework @ http://www.fault-tolerant.org).
Recall is written in C++.  I need to include the postgres headers and
there are some problems when including the headers w/C++.

Attached is a patch generated from postgres/src that fixes my problems.
I was hoping to get this into the main source.  It's very small (2k) and
3 files are changed: backend/utils/fmgr/fmgr.c,
backend/utils/Gen_fmgrtab.sh.in, and include/access/tupdesc.h.

In C++, you get a multiply defined symbol because the variable
(FmgrInfo *fmgr_pl_finfo) is defined in the header (the patch moves it
to the .c file).  The other problem in tupdesc.h is the use of typeid
is a problem in c++ (I renamed it to oidtypeid).

Thanks,
Neal Norwitz

25 years agoRemove calls to getprotobyname(), which we now know leaks memory on
Tom Lane [Sun, 21 May 2000 21:19:53 +0000 (21:19 +0000)]
Remove calls to getprotobyname(), which we now know leaks memory on
some platforms --- and I also see that it is documented as not thread-
safe on HPUX and possibly other platforms.  No good reason not to just
use IPPROTO_TCP constant from <netinet/in.h> instead.

25 years agoRepair memory leaks that caused CacheCxt to grow without bound. We
Tom Lane [Sun, 21 May 2000 02:28:55 +0000 (02:28 +0000)]
Repair memory leaks that caused CacheCxt to grow without bound.  We
really ought to fix relcache entry construction so that it does not
do so much with CurrentMemoryContext = CacheCxt.  As is, relatively
harmless leaks in either sequential or index scanning translate to
permanent leaks if they occur when called from relcache build.
For the moment, however, the path of least resistance is to repair
all such leaks...

25 years agoAdd debug code to aid in memory-leak tracking: if SHOW_MEMORY_STATS is
Tom Lane [Sun, 21 May 2000 02:23:30 +0000 (02:23 +0000)]
Add debug code to aid in memory-leak tracking: if SHOW_MEMORY_STATS is
defined then statistics about memory usage of all the global memory
contexts are printed after each commit.

25 years agoClean up grotty references to CacheCxt (externs inside functions,
Tom Lane [Sat, 20 May 2000 23:11:29 +0000 (23:11 +0000)]
Clean up grotty references to CacheCxt (externs inside functions,
duplicate global declarations, no points for style at all!)