OSDN Git Service

pg-rex/syncrep.git
27 years agoFrom: PortSite <info@portsite.nl>
Marc G. Fournier [Fri, 6 Jun 1997 01:34:59 +0000 (01:34 +0000)]
From: PortSite <info@portsite.nl>

Install file says: Type flex -- version
There is a space between -- and version that shouldn't
be there :-(

27 years agoFrom: "D'Arcy J.M. Cain" <darcy@druid.net>
Marc G. Fournier [Fri, 6 Jun 1997 00:38:17 +0000 (00:38 +0000)]
From: "D'Arcy J.M. Cain" <darcy@druid.net>
Subject: [HACKERS] src.original/./backend/lib/fstack.c

Another change I suggested.  I bracket an unused function and add a
return to quiet the compiler.  In addition I added an internal
consistency check.

27 years agoFrom: "D'Arcy J.M. Cain" <darcy@druid.net>
Marc G. Fournier [Fri, 6 Jun 1997 00:37:23 +0000 (00:37 +0000)]
From: "D'Arcy J.M. Cain" <darcy@druid.net>
Subject: [HACKERS] backend/optimizer/geqo/geqo_erx.c

I sent these changes in with a bunch of others.  Some were folded in but
others, like these, were not.  I am not sure why so I am resending this
to the developers list by itself for discussion.

The readon why I suggest these changes is that the compiler can't tell
that minimum_count is initialized before it is used.  The tests that I
add in here will cause an immediate error if it doesn't.  As the comments
below suggest, if it is 100% guaranteed that the variable will always
be initialized then how this is so should be commented here.  I don't
know how much strain the actual test puts on the performance but if it
isn't too much then maybe leave it in for absolute safety anyway.

There are also a few returns just to stop warnings.

27 years agoFix for large objects and case sensitivity, from Raymond Toy.
Bruce Momjian [Thu, 5 Jun 1997 22:59:45 +0000 (22:59 +0000)]
Fix for large objects and case sensitivity, from Raymond Toy.

27 years agobsdi configure detection fix.
Bruce Momjian [Thu, 5 Jun 1997 17:36:43 +0000 (17:36 +0000)]
bsdi configure detection fix.

27 years agoallpaths.c:find_join_paths(): compute_rel_size() shouldn't be called
Vadim B. Mikheev [Thu, 5 Jun 1997 09:33:52 +0000 (09:33 +0000)]
allpaths.c:find_join_paths(): compute_rel_size() shouldn't be called
for join-relations. Sizes already computed by
prune_rel_paths():compute_joinrel_size().

joinrels.c:
<                   if ( _use_right_sided_plans_ )
---
>                   if ( _use_right_sided_plans_ &&
>                               length (outer_rel->relids) > 1 )

 - r_plans are useful when outer_rel is join-relation... It
decreases the size of search space...

27 years agoDefinition for RelationForgetRelation().
Vadim B. Mikheev [Wed, 4 Jun 1997 09:01:49 +0000 (09:01 +0000)]
Definition for RelationForgetRelation().

27 years agoheap_destroy() now calls RelationForgetRelation() to really flush
Vadim B. Mikheev [Wed, 4 Jun 1997 08:59:22 +0000 (08:59 +0000)]
heap_destroy() now calls RelationForgetRelation() to really flush
the relation from the relcache.

27 years agoNew func RelationForgetRelation();
Vadim B. Mikheev [Wed, 4 Jun 1997 08:56:51 +0000 (08:56 +0000)]
New func RelationForgetRelation();
 *         RelationFlushRelation + if the relation is local then get rid of
 *         the relation descriptor from the newly created relation list.

27 years agoUPdate TODO/HISTORY again.
Bruce Momjian [Tue, 3 Jun 1997 20:36:48 +0000 (20:36 +0000)]
UPdate TODO/HISTORY again.

27 years agoUpdate TODO for release
Bruce Momjian [Tue, 3 Jun 1997 20:29:57 +0000 (20:29 +0000)]
Update TODO for release

27 years agoAdd bsdi to template detection.
Bruce Momjian [Tue, 3 Jun 1997 20:08:06 +0000 (20:08 +0000)]
Add bsdi to template detection.

27 years agoUpdate comments for regression testing.
Thomas G. Lockhart [Tue, 3 Jun 1997 15:25:45 +0000 (15:25 +0000)]
Update comments for regression testing.
Include paragraph on upgrading old databases containing path or polygon types.

27 years agoUse error message syntax consistant with other messages from the same
Thomas G. Lockhart [Tue, 3 Jun 1997 14:34:51 +0000 (14:34 +0000)]
Use error message syntax consistant with other messages from the same
reference platform (Linux/gcc-lib/i686).

27 years agoUpdate to reflect new "drop aggregate" syntax and messages.
Thomas G. Lockhart [Tue, 3 Jun 1997 14:31:46 +0000 (14:31 +0000)]
Update to reflect new "drop aggregate" syntax and messages.

27 years agoUpdate to show new path and polygon syntax.
Thomas G. Lockhart [Tue, 3 Jun 1997 14:29:51 +0000 (14:29 +0000)]
Update to show new path and polygon syntax.

27 years agoUpdate path and polygon syntax to new conventions.
Thomas G. Lockhart [Tue, 3 Jun 1997 14:24:12 +0000 (14:24 +0000)]
Update path and polygon syntax to new conventions.

27 years agoUpdate to use new path geometric type syntax.
Thomas G. Lockhart [Tue, 3 Jun 1997 14:21:36 +0000 (14:21 +0000)]
Update to use new path geometric type syntax.
Order in streets.data changed and is now alphabetical (sorry).

27 years agoModify banner to suggest using PST8PDT time zone rather than full specifier
Thomas G. Lockhart [Tue, 3 Jun 1997 14:19:28 +0000 (14:19 +0000)]
Modify banner to suggest using PST8PDT time zone rather than full specifier
string.

27 years agoUpdate to suggest using PST8PDT time zone rather than full time zone string.
Thomas G. Lockhart [Tue, 3 Jun 1997 14:18:38 +0000 (14:18 +0000)]
Update to suggest using PST8PDT time zone rather than full time zone string.
Include section on error message differences.

27 years agoAdd upgradepath(), isoldpath(), upgradepoly() and revertpoly() to allow
Thomas G. Lockhart [Tue, 3 Jun 1997 14:11:28 +0000 (14:11 +0000)]
Add upgradepath(), isoldpath(), upgradepoly() and revertpoly() to allow
upgrading from existing pre-v6.1 path and polygon geometric data types.

27 years agoPut parens around macro arguments for safety per D'Arcy's suggestion.
Thomas G. Lockhart [Tue, 3 Jun 1997 14:07:24 +0000 (14:07 +0000)]
Put parens around macro arguments for safety per D'Arcy's suggestion.

27 years agoAdd upgradepath(), isoldpath(), upgradepoly() and revertpoly() to allow
Thomas G. Lockhart [Tue, 3 Jun 1997 14:01:22 +0000 (14:01 +0000)]
Add upgradepath(), isoldpath(), upgradepoly() and revertpoly() to allow
migration from pre-v6.1 geometric data types.
Only allow new input syntax for paths and polygons.

27 years agoUse finite() macro if available to check returns from pow() and exp().
Thomas G. Lockhart [Tue, 3 Jun 1997 13:58:06 +0000 (13:58 +0000)]
Use finite() macro if available to check returns from pow() and exp().

27 years agoFix a few DATEDEBUG print statements.
Thomas G. Lockhart [Tue, 3 Jun 1997 13:56:32 +0000 (13:56 +0000)]
Fix a few DATEDEBUG print statements.

27 years agocc1: warnings being treated as errors
Vadim B. Mikheev [Tue, 3 Jun 1997 06:32:56 +0000 (06:32 +0000)]
cc1: warnings being treated as errors
dt.c: In function `timespan2tm':
dt.c:1722: warning: unused variable `funit'
dt.c:1722: warning: unused variable `iunit'

-> got rid of them.

27 years agoSET geqo TO 'on' restores _use_geqo_rels_ to GEQO_RELS, not
Vadim B. Mikheev [Tue, 3 Jun 1997 06:29:31 +0000 (06:29 +0000)]
SET geqo TO 'on' restores _use_geqo_rels_ to GEQO_RELS, not
to last specified with 'on' #.
What is better ?

27 years agocc1: warnings being treated as errors
Vadim B. Mikheev [Tue, 3 Jun 1997 06:23:24 +0000 (06:23 +0000)]
cc1: warnings being treated as errors
remove.c: In function `RemoveAggregate':
remove.c:413: warning: unused variable `typename'

-> got rid of it.

27 years agoSET var TO '...'
Vadim B. Mikheev [Tue, 3 Jun 1997 06:19:11 +0000 (06:19 +0000)]
SET var TO '...'
        ^^ - added

27 years agocc1: warnings being treated as errors
Vadim B. Mikheev [Tue, 3 Jun 1997 06:17:34 +0000 (06:17 +0000)]
cc1: warnings being treated as errors
psql.c: In function `HandleSlashCmds':
psql.c:1141: warning: `optarg3' might be used uninitialized in this function
psql.c:1157: warning: `optarg3' might be used uninitialized in this function

-> char           *optarg3 = NULL;

27 years agoCleanups.
Vadim B. Mikheev [Tue, 3 Jun 1997 04:14:31 +0000 (04:14 +0000)]
Cleanups.

27 years agoAdd \dt and \di to psql help.
Bruce Momjian [Tue, 3 Jun 1997 03:35:46 +0000 (03:35 +0000)]
Add \dt and \di to psql help.

27 years agoNew description of currval.
Vadim B. Mikheev [Tue, 3 Jun 1997 03:19:08 +0000 (03:19 +0000)]
New description of currval.

27 years ago#define GEQO_RELS 6 (not 7)
Vadim B. Mikheev [Tue, 3 Jun 1997 03:15:50 +0000 (03:15 +0000)]
#define GEQO_RELS 6 (not 7)

27 years agoFix deleted tuples re-incarnation possible when vacuum transaction
Vadim B. Mikheev [Tue, 3 Jun 1997 01:29:26 +0000 (01:29 +0000)]
Fix deleted tuples re-incarnation possible when vacuum transaction
start time equal to tuple->t_tmax.

Privent shrinking if there are tuples modifyed by running transactions
(it concerns system relations only, currently).

27 years ago*** empty log message ***
Edmund Mergl [Mon, 2 Jun 1997 19:42:14 +0000 (19:42 +0000)]
*** empty log message ***

27 years agoUpdate man and psqlHelp for new SET GEQO=# option.
Bruce Momjian [Mon, 2 Jun 1997 14:12:40 +0000 (14:12 +0000)]
Update man and psqlHelp for new SET GEQO=# option.

27 years agoAllow for EOF or \0 as input() return from lex.
Bruce Momjian [Mon, 2 Jun 1997 13:55:19 +0000 (13:55 +0000)]
Allow for EOF or \0 as input() return from lex.

27 years agoOracle like currval behaviour.
Vadim B. Mikheev [Mon, 2 Jun 1997 11:22:52 +0000 (11:22 +0000)]
Oracle like currval behaviour.

27 years ago#define GEQO_RELS 7
Vadim B. Mikheev [Mon, 2 Jun 1997 11:19:33 +0000 (11:19 +0000)]
#define GEQO_RELS 7
moved from geqo.h to internal.h.

27 years agoUse GEQO if _use_geqo_ is TRUE and # of relations is >= _use_geqo_rels_
Vadim B. Mikheev [Mon, 2 Jun 1997 11:14:40 +0000 (11:14 +0000)]
Use GEQO if _use_geqo_ is TRUE and # of relations is >= _use_geqo_rels_
(both are settable via SET geqo TO ...).

27 years agoSET var TO 'a=b'
Vadim B. Mikheev [Mon, 2 Jun 1997 11:00:57 +0000 (11:00 +0000)]
SET var TO 'a=b'
             ^^ is supported by get_token now.
(SET geqo TO 'on=XXX' works now).

27 years agoFrom: Igor <igor@sba.miami.edu>
Marc G. Fournier [Mon, 2 Jun 1997 03:04:50 +0000 (03:04 +0000)]
From: Igor <igor@sba.miami.edu>
Subject: [PATCHES] patch for a memory leak

Well...I screwed up and posted the wrong patch for psql originally..
The patch for that patch wposted below will fix it..

27 years agoFrom: David Friend <dfriend@atlsci.atlsci.com>
Marc G. Fournier [Mon, 2 Jun 1997 03:02:43 +0000 (03:02 +0000)]
From: David Friend <dfriend@atlsci.atlsci.com>
Subject: [PATCHES] INSTALL patch

This patch makes the following changes to the INSTALL instructions:
  - Before step 1, describe disk space requirements.
  - Step 1 now defines a "tested" platform.
  - Add step 3a on checking for disk space.
  - Added new step 27 asking for feedback.

27 years agoFrom: Igor <igor@sba.miami.edu>
Marc G. Fournier [Mon, 2 Jun 1997 02:53:00 +0000 (02:53 +0000)]
From: Igor <igor@sba.miami.edu>
Subject: [PATCHES] Another destroydb patch

This is a patch to my previous destroydb patch cause some people wanted
slightly different behavior. After this patch is applied, destroydb
will destroy a database as usual, but if added -i flag (which could be
aliased like rm -i) would ask for confirmation.

27 years agoFrom: Igor <igor@sba.miami.edu>
Marc G. Fournier [Mon, 2 Jun 1997 02:52:06 +0000 (02:52 +0000)]
From: Igor <igor@sba.miami.edu>
Subject: [PATCHES] pg_dump memory leak patch

This patch fixes a HUGE memory leak problem in pg_dump.
Pretty much anything that was allocated was never freed and Purify
reported about 40% possible memory leak and 6% actual leak. I added
functions to clear out all the allocated structures. After the patch
Purify returns 0 for number of bytes leaked...

27 years agoOrdering of libraries changed so that -lbsd is *after* -lm
Marc G. Fournier [Sun, 1 Jun 1997 16:25:34 +0000 (16:25 +0000)]
Ordering of libraries changed so that -lbsd is *after* -lm

Fixes a problem Darren reports with AIX, and doesn't seem to break either
FreeBSD or Linux(David?)...

27 years agoFrom: Darren King <aixssd!darrenk@abs.net>
Marc G. Fournier [Sun, 1 Jun 1997 15:53:24 +0000 (15:53 +0000)]
From: Darren King <aixssd!darrenk@abs.net>
Subject: [PATCHES] psql - \dt,\di commands.

I sent this a couple of months ago in re a request by Maxim
Kozin, but I had the patch reversed, creating some confusion
over applying it.

Here's a more complete version.

Adds \dt to list only tables/views and \di to list only
indicies.  \d will still work as before.

27 years agoFrom: Phil Nelson <phil@cs.wwu.edu>
Marc G. Fournier [Sun, 1 Jun 1997 15:51:10 +0000 (15:51 +0000)]
From: Phil Nelson <phil@cs.wwu.edu>
Subject: [PATCHES] Support for NetBSD/sparc

The following is a patch to allow BSD44_derived to support NetBSD/sparc.

27 years agoFrom: Darren King <aixssd!darrenk@abs.net>
Marc G. Fournier [Sun, 1 Jun 1997 15:50:08 +0000 (15:50 +0000)]
From: Darren King <aixssd!darrenk@abs.net>
Subject: [PATCHES] aix.h needs POSIX...

27 years agoFrom: Robert Bruccoleri <bruc@bms.com>
Marc G. Fournier [Sun, 1 Jun 1997 15:41:40 +0000 (15:41 +0000)]
From: Robert Bruccoleri <bruc@bms.com>

One more Irix related patch for Dynamic Linking

27 years agoFrom: Igor <igor@sba.miami.edu>
Marc G. Fournier [Sun, 1 Jun 1997 15:40:08 +0000 (15:40 +0000)]
From: Igor <igor@sba.miami.edu>
Subject: [PATCHES] destroydb patch

I am including a patch for destroydb to ask for confirmation before
deleting databases (after I accidentally deleted mine)...destroydb -y
would force delete without any confirmation.

27 years agoFrom: Igor <igor@sba.miami.edu>
Marc G. Fournier [Sun, 1 Jun 1997 15:39:08 +0000 (15:39 +0000)]
From: Igor <igor@sba.miami.edu>
Subject: [PATCHES] memory leak patches in libpq and psql

A couple of small memory leak patches (detected with Purify) primarily
in libpq.

* Fixed (NULL) border problem in psql (run psql, do \m, then select
  something from a table...row separators will be nulls)
* Fixed memory leak with the abovementioned border not being freed
  properly.
* Fixed memory leak in freePGconn() not freeing conn->port
* Fixed up PQclear() to free parts of PGresult only if these
  parts are not null.
* Fixed a decent memory leak that occured after executing every command
  in psql. PGresult *results was not freed most of the time.

There is still a leak being detected (2 bytes) in readline functions, but
I think this is old readline library. I will install new one and test it.

27 years agoAdded \n to error message.
Bruce Momjian [Sun, 1 Jun 1997 04:59:25 +0000 (04:59 +0000)]
Added \n to error message.

27 years agoAdd R_PLANS to psql help, and SET cleanup.
Bruce Momjian [Sun, 1 Jun 1997 04:53:44 +0000 (04:53 +0000)]
Add R_PLANS to psql help,  and SET cleanup.

27 years agoDROP agg psql help change from Darren.
Bruce Momjian [Sun, 1 Jun 1997 04:32:41 +0000 (04:32 +0000)]
DROP agg psql help change from Darren.

27 years agoSET R_PLANS added to manual page.
Bruce Momjian [Sun, 1 Jun 1997 04:23:12 +0000 (04:23 +0000)]
SET R_PLANS added to manual page.

27 years agoPut back old lines.
Bruce Momjian [Sun, 1 Jun 1997 04:16:16 +0000 (04:16 +0000)]
Put back old lines.

27 years agoAdd bsdi 3.0 to configure.
Bruce Momjian [Sun, 1 Jun 1997 04:05:26 +0000 (04:05 +0000)]
Add bsdi 3.0 to configure.

27 years agoCleanup for AIX from Darren.
Bruce Momjian [Sun, 1 Jun 1997 03:39:28 +0000 (03:39 +0000)]
Cleanup for AIX from Darren.

27 years agoEscape newline in select output, from A. Duursma.
Bruce Momjian [Sun, 1 Jun 1997 03:18:32 +0000 (03:18 +0000)]
Escape newline in select output, from A. Duursma.

27 years agoEnable GEQO for more than six tables, from Martin.
Bruce Momjian [Sun, 1 Jun 1997 02:56:00 +0000 (02:56 +0000)]
Enable GEQO for more than six tables, from Martin.

27 years agoChanged 1943 to 1947 to fix, other cleanups.
Bruce Momjian [Sun, 1 Jun 1997 02:30:25 +0000 (02:30 +0000)]
Changed 1943 to 1947 to fix, other cleanups.

27 years agoFix for SELECT * INTO TABLE for char(), varchar() fields.
Bruce Momjian [Sat, 31 May 1997 16:52:19 +0000 (16:52 +0000)]
Fix for SELECT * INTO TABLE for char(), varchar() fields.

27 years agoIf there is no table in RangeTable for colname then
Vadim B. Mikheev [Sat, 31 May 1997 07:10:25 +0000 (07:10 +0000)]
If there is no table in RangeTable for colname then
elog (WARN, "attribute %s not found", colname);

27 years agoIf we have to split leaf page in the chain of duplicates
Vadim B. Mikheev [Sat, 31 May 1997 06:35:56 +0000 (06:35 +0000)]
If we have to split leaf page in the chain of duplicates
then we try to look at our right sibling first, but not farther,
as it was in yesterday fix.

27 years agoAdded: new BTP_CHAIN flag (if hikey == firstkey then it's not
Vadim B. Mikheev [Fri, 30 May 1997 18:40:02 +0000 (18:40 +0000)]
Added: new BTP_CHAIN flag (if hikey == firstkey then it's not
last page in chain of duplicates).

27 years agoFix duplicates handling.
Vadim B. Mikheev [Fri, 30 May 1997 18:35:40 +0000 (18:35 +0000)]
Fix duplicates handling.

27 years agoClean up support for USE_POSIX_TIME, ! HAVE_INT_TIMEZONE machines.
Thomas G. Lockhart [Fri, 30 May 1997 15:02:51 +0000 (15:02 +0000)]
Clean up support for USE_POSIX_TIME, ! HAVE_INT_TIMEZONE machines.
Remove references to modf() which is buggy on some platforms (Sparc/Linux).

27 years agoEnsure double alignment for time and arrays of circles.
Thomas G. Lockhart [Mon, 26 May 1997 16:55:23 +0000 (16:55 +0000)]
Ensure double alignment for time and arrays of circles.

27 years agoFrom: David Friend <dfriend@atlsci.atlsci.com>
Marc G. Fournier [Mon, 26 May 1997 00:43:41 +0000 (00:43 +0000)]
From: David Friend <dfriend@atlsci.atlsci.com>
Subject: [PATCHES] INSTALL changes

This patch modifies the INSTALL file.  The changes are:
  - SPARC/Linux-ELF was added to the list of supported platforms.
    The special notes for it at the bottom of the file were removed.
  - Changed "database server" to "RDBMS database server".
  - Modified step 6 so that when you restore your database the
    tables will be owned by the original owners instead of the
    postgresql superuser.
  - Modified step 19 on diagnosing the regression tests for the
    SPARC Linux platform with a beta release.
  - Other minor changes.

27 years agoFrom: David Friend <dfriend@atlsci.atlsci.com>
Marc G. Fournier [Mon, 26 May 1997 00:26:56 +0000 (00:26 +0000)]
From: David Friend <dfriend@atlsci.atlsci.com>
Subject: [PATCHES] pqcomprim.c patch

This is the patch by Robert Bruccoleri to fix the endian problem.
(Actually, it's the reverse of his patch.  He must have gotten the
order wrong.)

27 years ago-m <arg> -M weren't setup properly in getopts()...
Marc G. Fournier [Sun, 25 May 1997 15:37:15 +0000 (15:37 +0000)]
-m <arg> -M weren't setup properly in getopts()...

These are undocumented in the man pages though...should they be removed?

Reported by: "A. Duursma" <it@portsite.nl>

27 years agoFix core dump on \c.
Bruce Momjian [Sat, 24 May 1997 14:38:05 +0000 (14:38 +0000)]
Fix core dump on \c.

27 years agoFix \c option after breaking it.
Bruce Momjian [Sat, 24 May 1997 01:45:17 +0000 (01:45 +0000)]
Fix \c option after breaking it.

27 years agoClean up redundant tests for valid pointers in geometric types.
Thomas G. Lockhart [Fri, 23 May 1997 05:24:53 +0000 (05:24 +0000)]
Clean up redundant tests for valid pointers in geometric types.
Fix up decoder field masks for timespan and reltime.

27 years agoEnable new date and time definitions to allow Solaris and Irix -O2 compilation.
Thomas G. Lockhart [Fri, 23 May 1997 05:10:03 +0000 (05:10 +0000)]
Enable new date and time definitions to allow Solaris and Irix -O2 compilation.
Move date and time field type labels to less than 32 to allow use as masks.

27 years agoAllocateFile():
Vadim B. Mikheev [Fri, 23 May 1997 02:56:48 +0000 (02:56 +0000)]
AllocateFile():
   fdleft = pg_nofile() - allocatedFiles - nfile;
looks more realistic, but too noisy -
   fdleft = pg_nofile() - allocatedFiles;
restored.

27 years agoOnly free memory if environment value is replaced.
Bruce Momjian [Fri, 23 May 1997 01:53:42 +0000 (01:53 +0000)]
Only free memory if environment value is replaced.

27 years agoComment on fix added.
Bruce Momjian [Fri, 23 May 1997 00:20:20 +0000 (00:20 +0000)]
Comment on fix added.

27 years agoPreserve putenv() memory in psql \connect user fix.
Bruce Momjian [Thu, 22 May 1997 23:57:27 +0000 (23:57 +0000)]
Preserve putenv() memory in psql \connect user fix.

27 years agoFix backslash commands broken when \connect user added.
Bruce Momjian [Thu, 22 May 1997 18:55:40 +0000 (18:55 +0000)]
Fix backslash commands broken when \connect user added.

27 years agoRelationPurgeLocalRelation():
Vadim B. Mikheev [Thu, 22 May 1997 17:24:20 +0000 (17:24 +0000)]
RelationPurgeLocalRelation():
            /*
             * RelationFlushRelation () below will flush relation information
             * from the cache. We must call smgrclose to flush relation
             * information from SMGR & FMGR, too. We assume that for temp
             * relations smgrunlink is already called by heap_destroyr
             * and we skip smgrclose for them.          - vadim 05/22/97
             */
            smgrclose(reln->rd_rel->relsmgr, reln);

 - it avoids memory leaks in SMGR & VFD.

RelationFlushRelation():
   there is no more call FileInvalidate(RelationGetSystemPort(relation));
   - invalid (FileInvalidate() expects File, not SMGR' fd)
   - unuseful anyway.

27 years ago1. Fix md memory leak:
Vadim B. Mikheev [Thu, 22 May 1997 17:08:35 +0000 (17:08 +0000)]
1. Fix md memory leak:
   mdunlink() and mdclose() (too !!!) now free MdfdVec for relation
   and add it to free list, so it may be re-used for another relation
   later.
2. Fix VFD-manager memory leak (found by Massimo ... and me):
   mdunlink() has to call FileUnlink() to free allocation for fileName
   and add the Vfd slot to the free list.

27 years ago1. Cleanup (no more FreeFd - unuseful; others).
Vadim B. Mikheev [Thu, 22 May 1997 16:51:19 +0000 (16:51 +0000)]
1. Cleanup (no more FreeFd - unuseful; others).
2. FreeFile() has to do nothing with nfile (# of files opened
   by VFD manager).

27 years agogist.h:
Vadim B. Mikheev [Thu, 22 May 1997 16:09:57 +0000 (16:09 +0000)]
gist.h:
/*
** You can have as many strategies as you please in GiSTs, as
** long as your consistent method can handle them
*/
#define GISTNStrategies                 100
                                        ^^^
- too big number:

strat.h->StrategyEvaluationData->StrategyExpression  expression[12]
                                                                ^^
 - so 12 is real max # of strategies, or StrategyEvaluationIsValid
crashes backend (called if CASSER defined).

27 years ago1. Pass GISTENTRYs to giststate->penaltyFn by pointers, not by vals.
Vadim B. Mikheev [Thu, 22 May 1997 16:01:33 +0000 (16:01 +0000)]
1. Pass GISTENTRYs to giststate->penaltyFn by pointers, not by vals.
2. Re-initialize keys in gistrescan (if gist used in inner scan).

27 years agoFrom: Darren King <aixssd!darrenk@abs.net>
Marc G. Fournier [Thu, 22 May 1997 00:24:07 +0000 (00:24 +0000)]
From: Darren King <aixssd!darrenk@abs.net>
To: pgsql-patches@postgreSQL.org
Subject: [PATCHES] DROP AGGREGATE gram.y typo...

Somehow I dropped a comma in the gram.y part (line 129) of my
patch for drop aggregate.  Here's a correct patch for gram.y.

PS. I hope I got the right comma, manually applied :) (scrappy)

27 years agoFrom: Darren King <aixssd!darrenk@abs.net>
Marc G. Fournier [Thu, 22 May 1997 00:17:24 +0000 (00:17 +0000)]
From: Darren King <aixssd!darrenk@abs.net>
Subject: [PATCHES] DROP AGGREGATE patch/fix.

Here's a patch that fixes the DROP AGGREGATE command to delete
the desired aggregate for a specific type.

27 years agoOops, missed the commit on this one by Darren King also...new Makefile.aix
Marc G. Fournier [Thu, 22 May 1997 00:12:03 +0000 (00:12 +0000)]
Oops, missed the commit on this one by Darren King also...new Makefile.aix

27 years agoFrom: Darren King <aixssd!darrenk@abs.net>
Marc G. Fournier [Thu, 22 May 1997 00:11:29 +0000 (00:11 +0000)]
From: Darren King <aixssd!darrenk@abs.net>
Subject: [PATCHES] AIX make patch resubmitted.

Misc patches for AIX from Darren:

1)  New src/makefiles/Makefile.aix  This patch should only be
    applied if the following patch (4) is applied to backend/Makefile!
    Still looking into having configure determine the last line to do
    the shared link.  The 325 code will work for 41, so I put that in
    as the default.  Included a commented out 41 line for completeness.

*and*

4)  Patch the backend Makefile.  I've reviewed this patch with respect to the
    other ports that use MAKE_EXPORTS (svr4 and univel) as closely as I could
    and I don't see where it will break them.  If it does, please let me know
    and I'll rework it somehow.

27 years agoFrom: Darren King <aixssd!darrenk@abs.net>
Marc G. Fournier [Thu, 22 May 1997 00:07:30 +0000 (00:07 +0000)]
From: Darren King <aixssd!darrenk@abs.net>
Subject: [PATCHES] Re: [PORTS] AIX 6.1 fixes...

Here are the patches for the two things that wouldn't make it thru the AIX
compiler.  The geo_ops.c change is harmless I believe.  The nbtcompare.c patch
fixes me, but I don't know about any other ports.  Maybe wait on that one
until Vadim decides what to do about the unsigned vs signed chars varlena
issue.

27 years agoUpdated pg_dumpall and psql to preserve database owners.
Bruce Momjian [Wed, 21 May 1997 03:12:23 +0000 (03:12 +0000)]
Updated pg_dumpall and psql to preserve database owners.

27 years agoAdd SRCDIR to lextest
Bruce Momjian [Wed, 21 May 1997 00:45:24 +0000 (00:45 +0000)]
Add SRCDIR to lextest

27 years agoFix for lextest return value.
Bruce Momjian [Tue, 20 May 1997 21:35:16 +0000 (21:35 +0000)]
Fix for lextest return value.

27 years agoLittle memmory leak in RelationFlushRelation() in freeing
Vadim B. Mikheev [Tue, 20 May 1997 11:41:38 +0000 (11:41 +0000)]
Little memmory leak in RelationFlushRelation() in freeing
relation->rd_att (relation' TupleDesc).

27 years agoDefine FREE(x) as pfree.
Vadim B. Mikheev [Tue, 20 May 1997 11:37:33 +0000 (11:37 +0000)]
Define FREE(x) as pfree.

27 years agoShouldn't we use palloc instead of malloc ?
Vadim B. Mikheev [Tue, 20 May 1997 11:35:50 +0000 (11:35 +0000)]
Shouldn't we use palloc instead of malloc ?
Because of
 *      resetpsort  - resets (frees) malloc'd memory for an aborted Xaction
 *
 *      Not implemented yet.

27 years ago1. LocalBufferSync(): set nextFreeLocalBuf to 0 - we're freeing
Vadim B. Mikheev [Tue, 20 May 1997 11:30:32 +0000 (11:30 +0000)]
1. LocalBufferSync(): set nextFreeLocalBuf to 0 - we're freeing
   all local buffers @ xact commit, so accordingly nextFreeLocalBuf
   is first local buffer now.
   It helps to avoid unnecessary local buffer allocations in LocalBufferAlloc()
   latter ("memmory leaks" in 'order by').
2. ResetLocalBufferPool() lost allocated local buffers:

   memset(LocalBufferDescriptors, 0, sizeof(BufferDesc) * NLocBuffer);

   (local buffers leak @ xact aborts).

27 years agoConsider right-sided plans if _use_right_sided_plans_ is true.
Vadim B. Mikheev [Tue, 20 May 1997 10:37:26 +0000 (10:37 +0000)]
Consider right-sided plans if _use_right_sided_plans_ is true.