]> granicus.if.org Git - postgresql/log
postgresql
26 years agoConverted to page.sgml. Not needed in a man page.
Thomas G. Lockhart [Sat, 15 Aug 1998 13:01:26 +0000 (13:01 +0000)]
Converted to page.sgml. Not needed in a man page.

26 years agoInformation moved to sgml source files.
Thomas G. Lockhart [Sat, 15 Aug 1998 07:03:59 +0000 (07:03 +0000)]
Information moved to sgml source files.
The "Oracle compatibility" page should have always been in with functions
 anyway. The BKI information is not really appropriate for a man page.

26 years agoMake minor formatting improvements to statement synopses.
Thomas G. Lockhart [Sat, 15 Aug 1998 07:00:37 +0000 (07:00 +0000)]
Make minor formatting improvements to statement synopses.
Make sure all files end with a newline.

26 years agoUpdate for new information on PQsetNoticeProcessor().
Thomas G. Lockhart [Sat, 15 Aug 1998 06:56:36 +0000 (06:56 +0000)]
Update for new information on PQsetNoticeProcessor().
Now back in sync with libpq.3 man page, which should be obsolete soon.

26 years agoChange Id field in chapter tag to change html output file name.
Thomas G. Lockhart [Sat, 15 Aug 1998 06:55:05 +0000 (06:55 +0000)]
Change Id field in chapter tag to change html output file name.

26 years agoInclude working list of all documentation sources, with current status
Thomas G. Lockhart [Sat, 15 Aug 1998 06:53:52 +0000 (06:53 +0000)]
Include working list of all documentation sources, with current status
 and plans for some of them.

26 years agoInclude new chapters.
Thomas G. Lockhart [Sat, 15 Aug 1998 06:52:54 +0000 (06:52 +0000)]
Include new chapters.

26 years agoNice exposition on indices and keys from Herouth Maoz which appeared
Thomas G. Lockhart [Sat, 15 Aug 1998 06:52:03 +0000 (06:52 +0000)]
Nice exposition on indices and keys from Herouth Maoz which appeared
 on the mailing lists a while ago. Maybe slightly changed to fit docs.
Will go into the User's Guide.

26 years agoConvert information from man pages. Not really suitable for man pages
Thomas G. Lockhart [Sat, 15 Aug 1998 06:49:33 +0000 (06:49 +0000)]
Convert information from man pages. Not really suitable for man pages
 anyway, so this will replace them as of now.

26 years agoRemove single-argument trim() function from table.
Thomas G. Lockhart [Sat, 15 Aug 1998 06:47:39 +0000 (06:47 +0000)]
Remove single-argument trim() function from table.
Never seen because the parser frontend converts all trim() calls to
 btrim(), ltrim(), and rtime() calls before execution.

26 years agoSome old cleanup fixes for close_ps() from G. Thaker.
Thomas G. Lockhart [Sat, 15 Aug 1998 06:45:10 +0000 (06:45 +0000)]
Some old cleanup fixes for close_ps() from G. Thaker.

26 years agoAllow binary-compatible indices to be considered when checking for valid
Thomas G. Lockhart [Fri, 14 Aug 1998 16:13:07 +0000 (16:13 +0000)]
Allow binary-compatible indices to be considered when checking for valid
 indices for restriction clauses containing a constant.
Note that if an index does not match directly (usually because the types
 on both side of the clause don't match), and if a binary-compatible index
 is identified, then the operator function will be replaced by a new
 one. Should not be a problem, but be sure that if types are listed as
 being binary compatible (in parse_coerce.h) then the comparison functions
 are also binary-compatible, giving equivalent results.

26 years agoInclude OID as a built-in type.
Thomas G. Lockhart [Fri, 14 Aug 1998 16:07:00 +0000 (16:07 +0000)]
Include OID as a built-in type.

26 years agoCheck for bad result from pg_id. A bad result can come from shared library
Thomas G. Lockhart [Fri, 14 Aug 1998 16:05:51 +0000 (16:05 +0000)]
Check for bad result from pg_id. A bad result can come from shared library
 trouble, and the name of the shared library has been changed recently.
 Had to rerun ldconfig on my machine to get it working again.
Give an error message with a helpful hint if so...

26 years agofix typo.
Bruce Momjian [Tue, 11 Aug 1998 22:39:33 +0000 (22:39 +0000)]
fix typo.

26 years agoStrategy cleanups.
Bruce Momjian [Tue, 11 Aug 1998 20:55:52 +0000 (20:55 +0000)]
Strategy cleanups.

26 years agoindex strategy cleanup
Bruce Momjian [Tue, 11 Aug 1998 19:32:39 +0000 (19:32 +0000)]
index strategy cleanup

26 years ago the following patch fixes a bug in the oracle compatibility
Bruce Momjian [Tue, 11 Aug 1998 18:38:07 +0000 (18:38 +0000)]
the  following  patch fixes a bug in the oracle compatibility
    functions btrim() ltrim() and rtrim().

    The error was that the character after the set  was  included
    in the tests (ptr2 pointed to the character after the vardata
    part of set if no match found,  so  comparing  *ptr  or  *end
    against *ptr2 MAY match -> strip).

Jan

--

#======================================================================#
# It's easier to get forgiveness for being wrong than for being
right. # # Let's break this rule - forgive me.
# #======================================== jwieck@debis.com (Jan
Wieck) #

26 years agoThis one cleans the cursor problems ecpg had so far. It is now able
Bruce Momjian [Tue, 11 Aug 1998 18:33:37 +0000 (18:33 +0000)]
This one cleans the cursor problems ecpg had so far. It is now able
to understand cursors with variables.

Michael

26 years agoChange owner from oid to int4 type.
Bruce Momjian [Tue, 11 Aug 1998 18:28:49 +0000 (18:28 +0000)]
Change owner from oid to int4 type.

26 years agoMore op_class cleanup.
Bruce Momjian [Tue, 11 Aug 1998 14:32:03 +0000 (14:32 +0000)]
More op_class cleanup.

26 years agoRemove NOBTREE defines, and make findoidlinks handle regproc.
Bruce Momjian [Tue, 11 Aug 1998 05:32:46 +0000 (05:32 +0000)]
Remove NOBTREE defines, and make findoidlinks handle regproc.

26 years agocleanup.
Bruce Momjian [Tue, 11 Aug 1998 05:09:30 +0000 (05:09 +0000)]
cleanup.

26 years agoNew findoidjoins examines oid columns to find join relationships.
Bruce Momjian [Tue, 11 Aug 1998 02:32:00 +0000 (02:32 +0000)]
New findoidjoins examines oid columns to find join relationships.

26 years agoAdd new \w write command to psql.
Bruce Momjian [Mon, 10 Aug 1998 20:31:42 +0000 (20:31 +0000)]
Add new \w write command to psql.

26 years agoUpdate flowchart.
Bruce Momjian [Mon, 10 Aug 1998 14:39:45 +0000 (14:39 +0000)]
Update flowchart.

26 years agoUpdate flowchart.
Bruce Momjian [Mon, 10 Aug 1998 14:32:46 +0000 (14:32 +0000)]
Update flowchart.

26 years agoCleanups of optimizer.
Bruce Momjian [Mon, 10 Aug 1998 04:49:39 +0000 (04:49 +0000)]
Cleanups of optimizer.

26 years agoCleanup optimizer function names and clarify code.
Bruce Momjian [Mon, 10 Aug 1998 02:26:40 +0000 (02:26 +0000)]
Cleanup optimizer function names and clarify code.

26 years agoUpdate pgindent.
Bruce Momjian [Sun, 9 Aug 1998 17:57:31 +0000 (17:57 +0000)]
Update pgindent.

26 years agoNew pgindent.
Bruce Momjian [Sun, 9 Aug 1998 04:59:10 +0000 (04:59 +0000)]
New pgindent.

26 years agocleanups.
Bruce Momjian [Sun, 9 Aug 1998 04:17:38 +0000 (04:17 +0000)]
cleanups.

26 years agoThe attached patch implements some changes that were discussed a
Bruce Momjian [Sun, 9 Aug 1998 02:59:33 +0000 (02:59 +0000)]
The attached patch implements some changes that were discussed a
couple weeks ago on the hackers and interfaces lists:

1. When the backend sends a NOTICE message and closes the connection
   (typically, because it was told to by the postmaster after
   another backend coredumped), libpq will now print the notice
   and close the connection cleanly.  Formerly, the frontend app
   would usually terminate ungracefully due to a SIGPIPE.  (I am
   not sure if 6.3.2 behaved that way, but the current cvs sources
   do...)

2. libpq's various printouts to stderr are now fed through a single
   "notice processor" routine, which can be overridden by the
   application to direct notices someplace else.  This should ease
   porting libpq to Windows.

I also noticed and fixed a problem in PQprint: when sending output
to a pager subprocess, it would disable SIGPIPE in case the pager
terminates early (this is good) --- but afterwards it reset SIGPIPE
to SIG_DFL, rather than restoring the application's prior setting
(bad).

regards, tom lane

26 years agoOPTIMIZER_DEBUG additions.
Bruce Momjian [Fri, 7 Aug 1998 05:02:32 +0000 (05:02 +0000)]
OPTIMIZER_DEBUG additions.

26 years agoMake large objects their own relkind type. Fix dups in pg_class_mb
Bruce Momjian [Thu, 6 Aug 1998 05:13:14 +0000 (05:13 +0000)]
Make large objects their own relkind type.  Fix dups in pg_class_mb
files.  Fix sequence creation hack for relkind type.

26 years agoflowchart update
Bruce Momjian [Wed, 5 Aug 1998 19:14:23 +0000 (19:14 +0000)]
flowchart update

26 years agoflowchart update
Bruce Momjian [Wed, 5 Aug 1998 19:01:58 +0000 (19:01 +0000)]
flowchart update

26 years agoFix for \d index display.
Bruce Momjian [Wed, 5 Aug 1998 16:23:40 +0000 (16:23 +0000)]
Fix for \d index display.

26 years agoFrom: David Hartwig <daybee@bellatlantic.net>
Marc G. Fournier [Wed, 5 Aug 1998 04:49:19 +0000 (04:49 +0000)]
From: David Hartwig <daybee@bellatlantic.net>

I have attached a patch to allow GROUP BY and/or ORDER BY function or
expressions.  Note worthy items:

1. The expression or function need not be in the target list.
Example:
            SELECT  name FROM foo GROUP BY lower(name);

2.   Simplified the grammar to use expressions only.

3.  Cleaned up earlier patch in this area to make use of existing
utility functions.

3.  Reduced some of the members in the SortGroupBy parse node.   The
original data members were redundant with the new expression node.
(MUST do a "make clean" now)

4.  Added a new parse node "JoinUsing".   The JOIN USING clause was
overloading this SortGroupBy structure.   With the afore mentioned
reduction of members, the two clauses lost all their commonality.

5.  A bug still exist where, if a function or expression is GROUPed BY,
and an aggregate function does not include a attribute from the
expression or function, the backend crashes.   (or something like
that)   The bug pre-dates this patch.    Example:

    SELECT lower(a) AS lowcase, count(b) FROM foo GROUP BY lowcase;
                 *** BOOM  ***

    --Also when not in target list
    SELECT  count(b) FROM foo GROUP BY lower(a);
                *** BOOM  AGAIN ***

26 years agoFrom: Dr. Michael Meskes <meskes@online-club.de>
Marc G. Fournier [Wed, 5 Aug 1998 04:47:54 +0000 (04:47 +0000)]
From: Dr. Michael Meskes <meskes@online-club.de>

So this should finally get cursors working. There was an ugly bug in it.

26 years agoOR processing cleanup.
Bruce Momjian [Tue, 4 Aug 1998 18:42:39 +0000 (18:42 +0000)]
OR processing cleanup.

26 years agoAdd Indices display to \d command.
Bruce Momjian [Tue, 4 Aug 1998 18:29:41 +0000 (18:29 +0000)]
Add Indices display to \d command.

26 years agoFix encoding grammer problem.
Bruce Momjian [Tue, 4 Aug 1998 17:37:48 +0000 (17:37 +0000)]
Fix encoding grammer problem.

26 years agoMergeSort was sometimes called mergejoin and was confusing. Now
Bruce Momjian [Tue, 4 Aug 1998 16:44:31 +0000 (16:44 +0000)]
MergeSort was sometimes called mergejoin and was confusing.  Now
it is now only mergejoin.

26 years agoEXPLAIN VERBOSE prints the plan, and now pretty-prints the plan to
Bruce Momjian [Tue, 4 Aug 1998 15:00:28 +0000 (15:00 +0000)]
EXPLAIN VERBOSE prints the plan, and now pretty-prints the plan to
the postmaster log file.

26 years agoMake EXPLAIN show output more clearly.
Bruce Momjian [Tue, 4 Aug 1998 04:50:15 +0000 (04:50 +0000)]
Make EXPLAIN show output more clearly.

26 years agoUpdate flow chart.
Bruce Momjian [Tue, 4 Aug 1998 00:42:14 +0000 (00:42 +0000)]
Update flow chart.

26 years agoUpdate mark/reset index code for multiple indexes, (OR code).
Bruce Momjian [Mon, 3 Aug 1998 19:41:35 +0000 (19:41 +0000)]
Update mark/reset index code for multiple indexes, (OR code).
Thanks for Vadim for fixes.

26 years agoFix typo in man pages.
Bruce Momjian [Mon, 3 Aug 1998 05:54:30 +0000 (05:54 +0000)]
Fix typo in man pages.

26 years agoCleanup of OR processing.
Bruce Momjian [Mon, 3 Aug 1998 05:49:24 +0000 (05:49 +0000)]
Cleanup of OR processing.

26 years agoAllows the following query to succeed: "SELECT NULL ORDER BY 1;"
Thomas G. Lockhart [Sun, 2 Aug 1998 13:34:26 +0000 (13:34 +0000)]
Allows the following query to succeed: "SELECT NULL ORDER BY 1;"
There are three or four cases in transformSortClause() and I had fixed
only one case for UNION. A second case is now fixed, in the same way; I
assigned INT4OID to the column type for the "won't actually happen"
sort. Didn't want to skip the code entirely, since the backend needs to
_try_ a sort to get the NULLs right. I'm not certain under what
circumstances the other cases are invoked and these are not yet
fixed up, though perhaps they don't need to be...

26 years agoFix for OR handling with multiple indexes.
Bruce Momjian [Sun, 2 Aug 1998 07:10:38 +0000 (07:10 +0000)]
Fix for OR handling with multiple indexes.

26 years agorun autoconf
Bruce Momjian [Sun, 2 Aug 1998 00:34:05 +0000 (00:34 +0000)]
run autoconf

26 years agoFix MULTIBYTE typo.
Bruce Momjian [Sun, 2 Aug 1998 00:21:39 +0000 (00:21 +0000)]
Fix MULTIBYTE typo.

26 years agoFix duplicate inside initdb.sh
Bruce Momjian [Sat, 1 Aug 1998 22:57:41 +0000 (22:57 +0000)]
Fix duplicate inside initdb.sh

26 years agoCleanup of OR fix.
Bruce Momjian [Sat, 1 Aug 1998 22:44:55 +0000 (22:44 +0000)]
Cleanup of OR fix.

26 years agoAllow index use with OR clauses.
Bruce Momjian [Sat, 1 Aug 1998 22:12:13 +0000 (22:12 +0000)]
Allow index use with OR clauses.

26 years agoAdrian Hall reported a problem to me that snprintf() doesn't exist in, at
Marc G. Fournier [Sat, 1 Aug 1998 19:30:29 +0000 (19:30 +0000)]
Adrian Hall reported a problem to me that snprintf() doesn't exist in, at
least, Solaris 2.5.1.  We use it in backend/utils/adt/int8.c.

Add a check to configure so that we see if it exists or not, and, if not,
compile in snprintf.c from backend/port, which was taken from, and falls under
the same Berkeley license as us, the FreeBSD libc/stdio ...

26 years agoLmgr cleanup, new locking modes for LLL.
Vadim B. Mikheev [Sat, 1 Aug 1998 15:26:38 +0000 (15:26 +0000)]
Lmgr cleanup, new locking modes for LLL.

26 years agoMatching clauses to functional indices was broken, now fixed.
Vadim B. Mikheev [Fri, 31 Jul 1998 15:10:40 +0000 (15:10 +0000)]
Matching clauses to functional indices was broken, now fixed.

26 years agoFix scan adjustment.
Vadim B. Mikheev [Thu, 30 Jul 1998 05:05:05 +0000 (05:05 +0000)]
Fix scan adjustment.

26 years agoRemoved from distribution. Completely converted to SGML.
Thomas G. Lockhart [Wed, 29 Jul 1998 15:23:57 +0000 (15:23 +0000)]
Removed from distribution. Completely converted to SGML.
This was the Postgres95 documentation written by Jolly and Chu...

26 years agoUse libpq only once to avoid file naming conflicts in the html output.
Thomas G. Lockhart [Wed, 29 Jul 1998 07:08:00 +0000 (07:08 +0000)]
Use libpq only once to avoid file naming conflicts in the html output.

26 years agoSomehow had substituted this file for the contents of allfiles.sgml.
Thomas G. Lockhart [Wed, 29 Jul 1998 07:01:58 +0000 (07:01 +0000)]
Somehow had substituted this file for the contents of allfiles.sgml.
This should be the right thing now...

26 years agoAdd id field to chapter and book tags to allow output file names
Thomas G. Lockhart [Wed, 29 Jul 1998 06:50:04 +0000 (06:50 +0000)]
Add id field to chapter and book tags to allow output file names
 to be meaningful.

26 years agoAdd port information for NS32532.
Thomas G. Lockhart [Wed, 29 Jul 1998 06:35:19 +0000 (06:35 +0000)]
Add port information for NS32532.
Fix Id tag in chapter headings to allow meaningful file names in html.

26 years agoInclude information on documentation products and sources.
Thomas G. Lockhart [Wed, 29 Jul 1998 06:33:37 +0000 (06:33 +0000)]
Include information on documentation products and sources.
Include tables of possible source files throughout tree.

26 years agoAdd missing information.
Thomas G. Lockhart [Wed, 29 Jul 1998 06:32:47 +0000 (06:32 +0000)]
Add missing information.
Fix examples for regex operators.

26 years agoUse option to jade to produce html file names from id field in tag.
Thomas G. Lockhart [Wed, 29 Jul 1998 06:29:04 +0000 (06:29 +0000)]
Use option to jade to produce html file names from id field in tag.

26 years agoAdd new files from Oliver Elphick.
Thomas G. Lockhart [Wed, 29 Jul 1998 06:23:26 +0000 (06:23 +0000)]
Add new files from Oliver Elphick.

26 years agoFix minor end-of-file formatting.
Thomas G. Lockhart [Wed, 29 Jul 1998 06:21:57 +0000 (06:21 +0000)]
Fix minor end-of-file formatting.

26 years agoInformation moved to doc/src/sgml/release.sgml.
Thomas G. Lockhart [Wed, 29 Jul 1998 06:13:48 +0000 (06:13 +0000)]
Information moved to doc/src/sgml/release.sgml.

26 years agoInformation moved to doc/src/sgml/geqo.sgml.
Thomas G. Lockhart [Wed, 29 Jul 1998 06:07:54 +0000 (06:07 +0000)]
Information moved to doc/src/sgml/geqo.sgml.

26 years agoUse Snapshot in heap access methods.
Vadim B. Mikheev [Mon, 27 Jul 1998 19:38:40 +0000 (19:38 +0000)]
Use Snapshot in heap access methods.

26 years agoMissed a few files in the last round of commits from Tatsuo, as well
Marc G. Fournier [Mon, 27 Jul 1998 03:21:58 +0000 (03:21 +0000)]
Missed a few files in the last round of commits from Tatsuo, as well
as needed to run autoconf ...

26 years agoFrom: t-ishii@sra.co.jp
Marc G. Fournier [Sun, 26 Jul 1998 04:31:41 +0000 (04:31 +0000)]
From: t-ishii@sra.co.jp

As Bruce mentioned, this is due to the conflict among changes we made.
Included patches should fix the problem(I changed all MB to
MULTIBYTE). Please let me know if you have further problem.

P.S. I did not include pathces to configure and gram.c to save the
file size(configure.in and gram.y modified).

26 years agoI missed addding initdb.sh ...
Marc G. Fournier [Sun, 26 Jul 1998 04:22:42 +0000 (04:22 +0000)]
I missed addding initdb.sh ...

26 years agoUpdate flowchart. Add arrow.
Bruce Momjian [Sun, 26 Jul 1998 02:17:55 +0000 (02:17 +0000)]
Update flowchart.  Add arrow.

26 years agoFix compile error. Make transaction/work optional on all transaction
Bruce Momjian [Sun, 26 Jul 1998 01:18:09 +0000 (01:18 +0000)]
Fix compile error.  Make transaction/work optional on all transaction
statements.  More cleanups of psql help.  Fix for shift/reduce on
UNION in subselect.

26 years agoUpdate psql help syntax to remove <> and uppercaese keywords.
Bruce Momjian [Sat, 25 Jul 1998 00:17:30 +0000 (00:17 +0000)]
Update psql help syntax to remove <> and uppercaese keywords.

26 years agoAdd mention of actual SQL command to create*/destroy* commands.
Bruce Momjian [Fri, 24 Jul 1998 16:44:04 +0000 (16:44 +0000)]
Add mention of actual SQL command to create*/destroy* commands.

26 years agoAdd mention of SQL commands to create/destroy man pages.
Bruce Momjian [Fri, 24 Jul 1998 16:36:56 +0000 (16:36 +0000)]
Add mention of SQL commands to create/destroy man pages.

26 years agoAdd mention of psql and template1 to create*, destroy* command
Bruce Momjian [Fri, 24 Jul 1998 16:10:19 +0000 (16:10 +0000)]
Add mention of psql and template1 to create*, destroy* command
manual pages.

26 years agoFix HAVING patch missing cast.
Bruce Momjian [Fri, 24 Jul 1998 15:54:10 +0000 (15:54 +0000)]
Fix HAVING patch missing cast.

26 years agoI'm sorry, but I think I introduced a little bug with my last patch.
Bruce Momjian [Fri, 24 Jul 1998 04:03:10 +0000 (04:03 +0000)]
I'm sorry, but I think I introduced a little bug with my last patch.
Everyone using an [NOT] EXISTS subquery will have noticed that
already.

The bug is in "subselect.c" in the function "SS_process_sublinks()".

Here the whole function as it *SHOULD BE*:

Stephan

26 years agoI really hope that I haven't missed anything in this one...
Marc G. Fournier [Fri, 24 Jul 1998 03:32:46 +0000 (03:32 +0000)]
I really hope that I haven't missed anything in this one...

From: t-ishii@sra.co.jp

Attached are patches to enhance the multi-byte support.  (patches are
against 7/18 snapshot)

* determine encoding at initdb/createdb rather than compile time

Now initdb/createdb has an option to specify the encoding. Also, I
modified the syntax of CREATE DATABASE to accept encoding option. See
README.mb for more details.

For this purpose I have added new column "encoding" to pg_database.
Also pg_attribute and pg_class are changed to catch up the
modification to pg_database.  Actually I haved added pg_database_mb.h,
pg_attribute_mb.h and pg_class_mb.h. These are used only when MB is
enabled. The reason having separate files is I couldn't find a way to
use ifdef or whatever in those files. I have to admit it looks
ugly. No way.

* support for PGCLIENTENCODING when issuing COPY command

commands/copy.c modified.

* support for SQL92 syntax "SET NAMES"

See gram.y.

* support for LATIN2-5
* add UNICODE regression test case
* new test suite for MB

New directory test/mb added.

* clean up source files

Basic idea is to have MB's own subdirectory for easier maintenance.
These are include/mb and backend/utils/mb.

26 years agoVersion 06-30-0248
Byron Nikolaidis [Wed, 22 Jul 1998 23:47:48 +0000 (23:47 +0000)]
Version 06-30-0248

26 years agoFix _lo_commit: core if LO memmory context was not created (LO were not used?)
Vadim B. Mikheev [Wed, 22 Jul 1998 05:48:59 +0000 (05:48 +0000)]
Fix _lo_commit: core if LO memmory context was not created (LO were not used?)

26 years agoVariableCache (next XID generator) is placed in shmem.
Vadim B. Mikheev [Tue, 21 Jul 1998 06:17:39 +0000 (06:17 +0000)]
VariableCache (next XID generator) is placed in shmem.

26 years agoUpdate.
Bruce Momjian [Tue, 21 Jul 1998 04:48:31 +0000 (04:48 +0000)]
Update.

26 years agoUpdate.
Bruce Momjian [Tue, 21 Jul 1998 04:19:44 +0000 (04:19 +0000)]
Update.

26 years agoTheses buffer leaks are caused by indexes that are kept open between
Bruce Momjian [Tue, 21 Jul 1998 04:17:30 +0000 (04:17 +0000)]
Theses buffer leaks are caused by indexes that are kept open between
calls. Outside a transaction, the backend detects them as buffer
leaks; it sends a NOTICE, and frees them. This sometimes cause a
segmentation fault (at least on Linux). These indexes are initialized
on the first lo_read/lo_write/lo_tell call, and (normally) closed
on a lo_close call.  Thus the buffer leaks appear when lo direct
access functions are used, and not with lo_import/lo_export functions
(libpq version calls lo_close before ending the command, and the
backend version uses another path).

The included patches (against recent snapshot, and against 6.3.2)
cause indexes to be closed on transaction end (that is on explicit
'END' statment, or on command termination outside trasaction blocks),
thus preventing the buffer leaks while increasing performance inside
transactions. Some (all?) 'classic' memory leaks are also removed.

I hope it will be ok.

--- Pascal ANDRE, graduated from Ecole Centrale Paris andre@via.ecp.fr

26 years agotarget list fixes.
Bruce Momjian [Mon, 20 Jul 1998 21:18:35 +0000 (21:18 +0000)]
target list fixes.

26 years agomakeTargetEntry cleanup.
Bruce Momjian [Mon, 20 Jul 1998 20:48:54 +0000 (20:48 +0000)]
makeTargetEntry cleanup.

26 years agoMakefile removal of internal.c.
Bruce Momjian [Mon, 20 Jul 1998 20:01:48 +0000 (20:01 +0000)]
Makefile removal of internal.c.

26 years agoCleanup makeTargetEntry and remove internal.c.
Bruce Momjian [Mon, 20 Jul 1998 19:53:53 +0000 (19:53 +0000)]
Cleanup makeTargetEntry and remove internal.c.

26 years agoUse defines rather than constants for types.
Bruce Momjian [Mon, 20 Jul 1998 19:21:45 +0000 (19:21 +0000)]
Use defines rather than constants for types.

26 years agoFix problem brought in with 32K machine.
Bruce Momjian [Mon, 20 Jul 1998 17:45:49 +0000 (17:45 +0000)]
Fix problem brought in with 32K machine.

26 years agoCleanup use of 16 that should be NAMEDATALEN.
Bruce Momjian [Mon, 20 Jul 1998 16:57:18 +0000 (16:57 +0000)]
Cleanup use of 16 that should be NAMEDATALEN.

26 years agoCleanup Name usage.
Bruce Momjian [Mon, 20 Jul 1998 16:14:18 +0000 (16:14 +0000)]
Cleanup Name usage.