]> granicus.if.org Git - postgresql/log
postgresql
18 years agoFixed a buffer overrun that was masked on Linux systems.
Michael Meskes [Fri, 18 Aug 2006 16:30:53 +0000 (16:30 +0000)]
Fixed a buffer overrun that was masked on Linux systems.

18 years agoAdd space before SP4.
Bruce Momjian [Fri, 18 Aug 2006 16:27:40 +0000 (16:27 +0000)]
Add space before SP4.

18 years agoUpdate Win2000SP4 mention.
Bruce Momjian [Fri, 18 Aug 2006 16:26:59 +0000 (16:26 +0000)]
Update Win2000SP4 mention.

18 years agoNow that we've rearranged relation open to get a lock before touching
Tom Lane [Fri, 18 Aug 2006 16:09:13 +0000 (16:09 +0000)]
Now that we've rearranged relation open to get a lock before touching
the rel, it's easy to get rid of the narrow race-condition window that
used to exist in VACUUM and CLUSTER.  Did some minor code-beautification
work in the same area, too.

18 years agoChanged lexer to no longer use the default rule.
Michael Meskes [Fri, 18 Aug 2006 15:59:35 +0000 (15:59 +0000)]
Changed lexer to no longer use the default rule.
Synced parser and keyword list.
Fixed parsing of CONNECT statement so it accepts a C string again.

18 years agoMention only Windows 2000SP4 works.
Bruce Momjian [Fri, 18 Aug 2006 15:51:28 +0000 (15:51 +0000)]
Mention only Windows 2000SP4 works.

18 years agoUpdate Win32 CMD quoting rules information location, CMD /?, from
Bruce Momjian [Fri, 18 Aug 2006 15:47:08 +0000 (15:47 +0000)]
Update Win32 CMD quoting rules information location, CMD /?, from
Magnus.

18 years agoImplement archive_timeout feature to force xlog file switches to occur no more
Tom Lane [Thu, 17 Aug 2006 23:04:10 +0000 (23:04 +0000)]
Implement archive_timeout feature to force xlog file switches to occur no more
than N seconds apart.  This allows a simple, if not very high performance,
means of guaranteeing that a PITR archive is no more than N seconds behind
real time.  Also make pg_current_xlog_location return the WAL Write pointer,
add pg_current_xlog_insert_location to return the Insert pointer, and fix
pg_xlogfile_name_offset to return its results as a two-element record instead
of a smashed-together string, as per recent discussion.

Simon Riggs

18 years agoProduce a clean failure in configure when dtrace is selected but cannot be
Peter Eisentraut [Thu, 17 Aug 2006 17:25:43 +0000 (17:25 +0000)]
Produce a clean failure in configure when dtrace is selected but cannot be
found.  Besides stopping those early who have no dtrace installed
whatsoever, this will also alert those who have dtrace in /usr/sbin, which
might not be in the path, which would produce confusing failures much later
in the build process.

Add documentation about pointing configure to find dtrace.

18 years agoFix an oversight in mergejoin planning: the planner would reject a
Tom Lane [Thu, 17 Aug 2006 17:06:37 +0000 (17:06 +0000)]
Fix an oversight in mergejoin planning: the planner would reject a
mergejoin possibility where the inner rel was less well sorted than
the outer (ie, it matches some but not all of the merge clauses that
can work with the outer), if the inner path in question is also the
overall cheapest path for its rel.  This is an old bug, but I'm not
sure it's worth back-patching, because it's such a corner case.
Noted while investigating a test case from Peter Hardman.

18 years agoTeach convert_subquery_pathkeys() to handle the case where the
Tom Lane [Thu, 17 Aug 2006 17:02:49 +0000 (17:02 +0000)]
Teach convert_subquery_pathkeys() to handle the case where the
subquery's pathkey is a RelabelType applied to something that appears
in the subquery's output; for example where the subquery returns a
varchar Var and the sort order is shown as that Var coerced to text.
This comes up because varchar doesn't have its own sort operator.
Per example from Peter Hardman.

18 years agoAdd:
Bruce Momjian [Thu, 17 Aug 2006 06:48:38 +0000 (06:48 +0000)]
Add:

< * Add support for arrays of domains
>  o Add support for arrays of domains
>  o Add support for arrays of complex types

18 years agoAdd:
Bruce Momjian [Thu, 17 Aug 2006 06:45:12 +0000 (06:45 +0000)]
Add:

> * Support a data type with specific enumerated values (ENUM)
>
>   http://archives.postgresql.org/pgsql-hackers/2006-08/msg00979.php

18 years agoAdd:
Bruce Momjian [Wed, 16 Aug 2006 14:59:26 +0000 (14:59 +0000)]
Add:

> * Add support for arrays of domains

18 years agoPut probes.o file in right directory, fixes dtrace compilation
Peter Eisentraut [Wed, 16 Aug 2006 14:57:37 +0000 (14:57 +0000)]
Put probes.o file in right directory, fixes dtrace compilation

18 years agoSeems some C compilers think 'restrict' is a fully reserved word.
Tom Lane [Wed, 16 Aug 2006 04:32:49 +0000 (04:32 +0000)]
Seems some C compilers think 'restrict' is a fully reserved word.
Per buildfarm results from warthog.

18 years agoAdd a hack so that get_type_io_data() can work from bootstrap.c's
Tom Lane [Tue, 15 Aug 2006 22:36:17 +0000 (22:36 +0000)]
Add a hack so that get_type_io_data() can work from bootstrap.c's
internal TypInfo table in bootstrap mode.  This allows array_in and
array_out to be used during early bootstrap, which eliminates the
former obstacle to giving OUT parameters to built-in functions.

18 years agoAdd hooks to allow debugging and performance measurement plugins
Tom Lane [Tue, 15 Aug 2006 19:01:17 +0000 (19:01 +0000)]
Add hooks to allow debugging and performance measurement plugins
to instrument PL/pgSQL.  Korry Douglas

18 years agoAdd server support for "plugin" libraries that can be used for add-on tasks
Tom Lane [Tue, 15 Aug 2006 18:26:59 +0000 (18:26 +0000)]
Add server support for "plugin" libraries that can be used for add-on tasks
such as debugging and performance measurement.  This consists of two features:
a table of "rendezvous variables" that allows separately-loaded shared
libraries to communicate, and a new GUC setting "local_preload_libraries"
that allows libraries to be loaded into specific sessions without explicit
cooperation from the client application.  To make local_preload_libraries
as flexible as possible, we do not restrict its use to superusers; instead,
it is restricted to load only libraries stored in $libdir/plugins/.  The
existing LOAD command has also been modified to allow non-superusers to
LOAD libraries stored in this directory.

This patch also renames the existing GUC variable preload_libraries to
shared_preload_libraries (after a suggestion by Simon Riggs) and does some
code refactoring in dfmgr.c to improve clarity.

Korry Douglas, with a little help from Tom Lane.

18 years agoFix pgbench to handle empty lines in script files as documented.
Tatsuo Ishii [Tue, 15 Aug 2006 13:05:30 +0000 (13:05 +0000)]
Fix pgbench to handle empty lines in script files as documented.
patches contributed by Itagaki Takahiro.

18 years agoFixed a typo and made two mallocs Coverity friendly.
Michael Meskes [Tue, 15 Aug 2006 12:46:25 +0000 (12:46 +0000)]
Fixed a typo and made two mallocs Coverity friendly.

18 years agoSorry, forgot the new files.
Michael Meskes [Tue, 15 Aug 2006 06:48:24 +0000 (06:48 +0000)]
Sorry, forgot the new files.

18 years agoAdded lots of SoC stuff made by Joachim.
Michael Meskes [Tue, 15 Aug 2006 06:40:20 +0000 (06:40 +0000)]
Added lots of SoC stuff made by Joachim.
Fixed broken newline on Windows.
Fixed a nasty buffer underrun that only occured when using Informix
no_indicator NULL setting on timestamps and intervals.

18 years agoCause '*' and 'foo.*' notations to mark the referenced RTE(s) as
Tom Lane [Mon, 14 Aug 2006 23:39:32 +0000 (23:39 +0000)]
Cause '*' and 'foo.*' notations to mark the referenced RTE(s) as
requiring read permissions.  Up till now there was no possible case
in which the RTEs wouldn't already have ACL_SELECT set ... but now that
you can say something like 'INSERT INTO foo ... RETURNING *' this is
an essential step.  With this commit, a RETURNING clause adds the
requirement for SELECT permissions on the target table if and only if
the clause actually reads the value of at least one target-table column.

18 years agoFix oversight in initial implementation of PORTAL_ONE_RETURNING mode: we
Tom Lane [Mon, 14 Aug 2006 22:57:15 +0000 (22:57 +0000)]
Fix oversight in initial implementation of PORTAL_ONE_RETURNING mode: we
cannot assume that there's exactly one Query in the Portal, as we can for
ONE_SELECT mode, because non-SELECT queries might have extra queries added
during rule rewrites.  Fix things up so that we'll use ONE_RETURNING mode
when a Portal contains one primary (canSetTag) query and that query has
a RETURNING list.  This appears to be a second showstopper reason for running
the Portal to completion before we start to hand anything back --- we want
to be sure that the rule-added queries get run too.

18 years agoSupport INSERT/UPDATE/DELETE RETURNING in plpgsql, with rowcount checking
Tom Lane [Mon, 14 Aug 2006 21:14:42 +0000 (21:14 +0000)]
Support INSERT/UPDATE/DELETE RETURNING in plpgsql, with rowcount checking
as per yesterday's proposal.  Also make things a tad more orthogonal by
adding the recent STRICT addition to EXECUTE INTO.
Jonah Harris and Tom Lane

18 years agoWhen executing a list of queries derived from rule expansion,
Tom Lane [Mon, 14 Aug 2006 13:40:18 +0000 (13:40 +0000)]
When executing a list of queries derived from rule expansion,
_SPI_execute_plan's return code should reflect the type of the query
that is marked canSetTag, not necessarily the last one in the list.

This is arguably a bug fix, but I'm hesitant to back-patch it because
it's the sort of subtle change that might break someone's code, and it's
best not to do that kind of thing in point releases.

18 years agoRemove hash_destroy calls in hash_create's failure paths. As noted by
Tom Lane [Mon, 14 Aug 2006 12:39:55 +0000 (12:39 +0000)]
Remove hash_destroy calls in hash_create's failure paths.  As noted by
a Coverity warning, these are risky since the hashtable isn't necessarily
fully set up yet.  They're unnecessary anyway: a deletable hashtable
should be in a memory context that will be cleared following elog(ERROR).
Per report from Martijn van Oosterhout.

18 years agoRevert (again) GUC patch to return commented fields to their default
Bruce Momjian [Mon, 14 Aug 2006 02:27:27 +0000 (02:27 +0000)]
Revert (again) GUC patch to return commented fields to their default
values, due to concern about the patch.

18 years agoGet rid of "lookahead" functionality in plpgsql's yylex() function,
Tom Lane [Mon, 14 Aug 2006 00:46:53 +0000 (00:46 +0000)]
Get rid of "lookahead" functionality in plpgsql's yylex() function,
and instead make the grammar production for the RETURN statement do the
heavy lifting.  The lookahead idea was copied from the main parser, but
it does not work in plpgsql's parser because here gram.y looks explicitly
at the scanner's yytext variable, which will be out of sync after a
failed lookahead step.  A minimal example is

create or replace function foo() returns void language plpgsql as '
begin
  perform return foo bar;
end';

which can be seen by testing to deliver "foo foo bar" to the main parser
instead of the expected "return foo bar".  This isn't a huge bug since
RETURN is not found in the main grammar, but it could bite someone who
tried to use "return" as an identifier.

Back-patch to 8.1.  Bug exists further back, but HEAD patch doesn't apply
cleanly, and given the lack of field complaints it doesn't seem worth
the effort to develop adjusted patches.

18 years agoFix core dump in duration logging for a V3-protocol Execute message
Tom Lane [Sun, 13 Aug 2006 22:18:08 +0000 (22:18 +0000)]
Fix core dump in duration logging for a V3-protocol Execute message
when what's being executed is a COMMIT or ROLLBACK.  Per report from
Sergey Koposov.  Backpatch to 8.1; 8.0 and before don't have the bug
due to lack of any logging at all here.

18 years agoCause psql to report both the returned data and the command status tag
Tom Lane [Sun, 13 Aug 2006 21:10:04 +0000 (21:10 +0000)]
Cause psql to report both the returned data and the command status tag
for INSERT/UPDATE/DELETE RETURNING.  Per discussion.

18 years agoIf test postmaster fails to start within 60 seconds, try to kill -9 it
Tom Lane [Sun, 13 Aug 2006 20:39:07 +0000 (20:39 +0000)]
If test postmaster fails to start within 60 seconds, try to kill -9 it
so that it won't interfere with later trials.  Per recent buildfarm
experience.  Anyone know how to do this on Windows?

18 years agoExtend 'guc' regression test to check manipulations of datestyle as
Tom Lane [Sun, 13 Aug 2006 19:31:06 +0000 (19:31 +0000)]
Extend 'guc' regression test to check manipulations of datestyle as
well as vacuum_cost_delay.  Since datestyle is a string variable,
this exercises memory allocation issues that might not appear when
modifying an integer GUC variable.  Also, we can observe the side
effects of changing datestyle to check that assign hooks are called
at the right times.

18 years agoOriginal coding of 'returning' regression test inadvertently chose a
Tom Lane [Sun, 13 Aug 2006 17:58:49 +0000 (17:58 +0000)]
Original coding of 'returning' regression test inadvertently chose a
nonunique join value, leading to plan-choice-dependent results ... and
it seems some platforms will choose a different plan.  Tweak the test
so that it has well-defined results.  Per report from Olivier Prenant.

18 years agoBack out plperl OUT hash/array parameter patch, again.
Bruce Momjian [Sun, 13 Aug 2006 17:31:10 +0000 (17:31 +0000)]
Back out plperl OUT hash/array parameter patch, again.

18 years agoFix recent guc comment-to-default patch for custom variables.
Bruce Momjian [Sun, 13 Aug 2006 15:37:02 +0000 (15:37 +0000)]
Fix recent guc comment-to-default patch for custom variables.

18 years agoApplied patch for VPATH builds by Alvaro Herrera <alvherre@commandprompt.com>
Michael Meskes [Sun, 13 Aug 2006 10:18:31 +0000 (10:18 +0000)]
Applied patch for VPATH builds by Alvaro Herrera <alvherre@commandprompt.com>
Merged dyntest.pgc and dyntest2.pgc.
Hopefully fixed the last Coverity reports (finally)

18 years agoRe-apply plperl patch that allows OUT parameters to be placed into Perl
Bruce Momjian [Sun, 13 Aug 2006 02:37:11 +0000 (02:37 +0000)]
Re-apply plperl patch that allows OUT parameters to be placed into Perl
hash and array variables.  (regression output updated)

18 years agoRecommit patch to allow commented GUC variables to return to their
Bruce Momjian [Sun, 13 Aug 2006 02:22:24 +0000 (02:22 +0000)]
Recommit patch to allow commented GUC variables to return to their
default values.

18 years agoRe-apply guc cleanup patch, with memory allocation bugs fixed.
Bruce Momjian [Sun, 13 Aug 2006 01:30:17 +0000 (01:30 +0000)]
Re-apply guc cleanup patch, with memory allocation bugs fixed.

18 years agoTweak SPI_cursor_open to allow INSERT/UPDATE/DELETE RETURNING; this was
Tom Lane [Sat, 12 Aug 2006 20:05:56 +0000 (20:05 +0000)]
Tweak SPI_cursor_open to allow INSERT/UPDATE/DELETE RETURNING; this was
merely a matter of fixing the error check, since the underlying Portal
infrastructure already handles it.  This in turn allows these statements
to be used in some existing plpgsql and plperl contexts, such as a
plpgsql FOR loop.  Also, do some marginal code cleanup in places that
were being sloppy about distinguishing SELECT from SELECT INTO.

18 years agoRemove ancient, obsolete comment.
Tom Lane [Sat, 12 Aug 2006 18:58:54 +0000 (18:58 +0000)]
Remove ancient, obsolete comment.

18 years agoBack out patch for plperl to handle OUT paramaters into arrays and
Bruce Momjian [Sat, 12 Aug 2006 04:16:45 +0000 (04:16 +0000)]
Back out patch for plperl to handle OUT paramaters into arrays and
hashes.  Was causing regression failures.

18 years agoBack out patch to reorganize guc processing. Was causing regression
Bruce Momjian [Sat, 12 Aug 2006 04:12:41 +0000 (04:12 +0000)]
Back out patch to reorganize guc processing.  Was causing regression
failures.

18 years agoBack out patch that allowed commented guc variables to return to their
Bruce Momjian [Sat, 12 Aug 2006 04:11:50 +0000 (04:11 +0000)]
Back out patch that allowed commented guc variables to return to their
default values.  Was causing regression failures.

18 years agoDone:
Bruce Momjian [Sat, 12 Aug 2006 03:50:27 +0000 (03:50 +0000)]
Done:

<  o Allow INSERT/UPDATE ... RETURNING new.col or old.col
<
<    This is useful for returning the auto-generated key for an INSERT.
<    One complication is how to handle rules that run as part of
<    the insert.
<    http://archives.postgresql.org/pgsql-patches/2005-07/msg00568.php
>  o -Allow INSERT/UPDATE ... RETURNING new.col or old.col

18 years agoFix wording, per Neil.
Bruce Momjian [Sat, 12 Aug 2006 03:48:32 +0000 (03:48 +0000)]
Fix wording, per Neil.

18 years agoAdd INSERT/UPDATE/DELETE RETURNING, with basic docs and regression tests.
Tom Lane [Sat, 12 Aug 2006 02:52:06 +0000 (02:52 +0000)]
Add INSERT/UPDATE/DELETE RETURNING, with basic docs and regression tests.
plpgsql support to come later.  Along the way, convert execMain's
SELECT INTO support into a DestReceiver, in order to eliminate some ugly
special cases.

Jonah Harris and Tom Lane

18 years agoSet Win32 server-size socket buffer to 32k, for performance reasons.
Bruce Momjian [Fri, 11 Aug 2006 20:44:20 +0000 (20:44 +0000)]
Set Win32 server-size socket buffer to 32k, for performance reasons.

Yoshiyuki Asaba

18 years agoDone:
Bruce Momjian [Fri, 11 Aug 2006 20:17:26 +0000 (20:17 +0000)]
Done:

<  o %Allow commenting of variables in postgresql.conf to restore them
>  o -Allow commenting of variables in postgresql.conf to restore them

18 years agoRevert undesired FAQ_DEV change.
Bruce Momjian [Fri, 11 Aug 2006 20:16:50 +0000 (20:16 +0000)]
Revert undesired  FAQ_DEV change.

18 years agoAllow commenting of variables in postgresql.conf to restore them to
Bruce Momjian [Fri, 11 Aug 2006 20:15:16 +0000 (20:15 +0000)]
Allow commenting of variables in postgresql.conf to restore them to
defaults.

Zdenek Kotala

18 years agoRefactor GUC set_config_option function:
Bruce Momjian [Fri, 11 Aug 2006 20:08:28 +0000 (20:08 +0000)]
Refactor GUC set_config_option function:

The main reason for refactoring was that set_config_option() was too
overloaded function and its behavior did not consistent. Old version of
set_config_function hides some messages. For example if you type:

tcp_port = 5432.1

then old implementation ignore this error without any message to log
file in the signal context (configuration reload). Main problem was that
semantic analysis of postgresql.conf is not perform in the
ProcessConfigFile function, but in the set_config_options *after*
context check. This skipped check for variables with PG_POSTMASTER
context. There was request from Joachim Wieland to add more messages
about ignored changes in the config file as well.

Zdenek Kotala

18 years agoplperl:
Bruce Momjian [Fri, 11 Aug 2006 19:42:35 +0000 (19:42 +0000)]
plperl:

Allow conversion from perl to postgresql array in OUT parameters. Second,
allow hash form output from procedures with one OUT argument.

Pavel Stehule

18 years agoMake the psql line counter 64-bit so it can handle files > 4gig lines.
Bruce Momjian [Fri, 11 Aug 2006 19:20:59 +0000 (19:20 +0000)]
Make the psql line counter 64-bit so it can handle files > 4gig lines.

David Fetter

18 years agoAdd Neil's presentation to FAQ_DEV.
Bruce Momjian [Fri, 11 Aug 2006 19:18:59 +0000 (19:18 +0000)]
Add Neil's presentation to FAQ_DEV.

18 years agoFix typo.
Alvaro Herrera [Fri, 11 Aug 2006 18:51:50 +0000 (18:51 +0000)]
Fix typo.

18 years agoUpdate for company URL.
Bruce Momjian [Fri, 11 Aug 2006 18:47:57 +0000 (18:47 +0000)]
Update for company URL.

Ian Barwick

18 years agoAdd URLs
Bruce Momjian [Fri, 11 Aug 2006 15:53:39 +0000 (15:53 +0000)]
Add URLs

< * Allow encoding on a per-column basis
> * Allow encoding on a per-column basis optionally using the ICU library:
>   http://archives.postgresql.org/pgsql-patches/2005-08/msg00309.php
>   http://archives.postgresql.org/pgsql-patches/2006-03/msg00233.php

18 years agoFix UNION/INTERSECT/EXCEPT so that when two inputs being merged have
Tom Lane [Thu, 10 Aug 2006 02:36:29 +0000 (02:36 +0000)]
Fix UNION/INTERSECT/EXCEPT so that when two inputs being merged have
same data type and same typmod, we show that typmod as the output
typmod, rather than generic -1.  This responds to several complaints
over the past few years about UNIONs unexpectedly dropping length or
precision info.

18 years agoReturn inline to win32.h because code was OK, but keep additional
Bruce Momjian [Thu, 10 Aug 2006 01:41:29 +0000 (01:41 +0000)]
Return inline to win32.h because code was OK, but keep additional
comment.  8.1.X is not affected by this commit.

18 years agoMove "#define inline __inline" from port/win32.h to c.h because Win32
Bruce Momjian [Thu, 10 Aug 2006 01:35:21 +0000 (01:35 +0000)]
Move "#define inline __inline" from port/win32.h to c.h because Win32
interface builds like libpq need it.

Backpatch addition to 8.1.X.

18 years agoFix display of log duration so it is milliseconds.microseconds "ms".
Bruce Momjian [Thu, 10 Aug 2006 00:44:01 +0000 (00:44 +0000)]
Fix display of log duration so it is milliseconds.microseconds "ms".

Greg Sabino Mullane

18 years agoFix inadequate makefile-fu in ecpg regression tests. I think this
Tom Lane [Wed, 9 Aug 2006 22:48:17 +0000 (22:48 +0000)]
Fix inadequate makefile-fu in ecpg regression tests.  I think this
will fix current platypus build failure.

18 years agoRound microseconds on setitimer upwards.
Bruce Momjian [Wed, 9 Aug 2006 21:18:13 +0000 (21:18 +0000)]
Round microseconds on setitimer upwards.

18 years agoOn Win32, make minimum setitimer() sleep be 1ms, so sleeps < 1ms aren't
Bruce Momjian [Wed, 9 Aug 2006 20:40:56 +0000 (20:40 +0000)]
On Win32, make minimum setitimer() sleep be 1ms, so sleeps < 1ms aren't
rounded down to zero.

Backpatch to 8.1.X.

18 years agoAdd URL for typmod:
Bruce Momjian [Wed, 9 Aug 2006 18:10:34 +0000 (18:10 +0000)]
Add URL for typmod:

>   http://archives.postgresql.org/pgsql-hackers/2006-08/msg00149.php

18 years agoFix statement_timeout on Win32 so that it properly treats micro-seconds
Bruce Momjian [Wed, 9 Aug 2006 17:47:03 +0000 (17:47 +0000)]
Fix statement_timeout on Win32 so that it properly treats micro-seconds
as micro-seconds, rather than as 100 microseconds, as it does now.  This
actually fixes all setitimer calls on Win32, but statement_timeout is
the most visible fix.

Backpatch to 8.1.X.  8.0 works as documented.

18 years agoFormatting improvement.
Bruce Momjian [Wed, 9 Aug 2006 17:33:52 +0000 (17:33 +0000)]
Formatting improvement.

18 years agoAdd URL:
Bruce Momjian [Wed, 9 Aug 2006 16:42:19 +0000 (16:42 +0000)]
Add URL:

>    http://archives.postgresql.org/pgsql-performance/2004-08/msg00349.php

18 years agoAdd TODO items we don't want (Idea from Josh Berkus)
Bruce Momjian [Wed, 9 Aug 2006 16:35:51 +0000 (16:35 +0000)]
Add TODO items we don't want (Idea from Josh Berkus)

>
>
> Features We Do _Not_ Want
> =========================
>
> * All backends running as threads in a single process (not want)
>
>   This eliminates the process protection we get from the current setup.
>   Thread creation is usually the same overhead as process creation on
>   modern systems, so it seems unwise to use a pure threaded model.
>
> * Optimizer hints (not want)
>
>   Optimizer hints are used to work around problems in the optimizer.  We
>   would rather have the problems reported and fixed.
>   http://archives.postgresql.org/pgsql-hackers/2006-08/msg00506.php

18 years agoUpdate packages:
Bruce Momjian [Wed, 9 Aug 2006 14:55:39 +0000 (14:55 +0000)]
Update packages:

<    A package would be a schema with public/private variables,
>    A package would be a schema with session-local variables,
>    http://archives.postgresql.org/pgsql-hackers/2006-08/msg00384.php

18 years agoFixed some memory bugs that somehow reappeared.
Michael Meskes [Wed, 9 Aug 2006 09:08:32 +0000 (09:08 +0000)]
Fixed some memory bugs that somehow reappeared.
Also fixed a new Coverity report.

18 years agoFixed error handling in numeric conversion (Joachim).
Michael Meskes [Wed, 9 Aug 2006 07:30:56 +0000 (07:30 +0000)]
Fixed error handling in numeric conversion (Joachim).
Further regression cleanup.

18 years agoSuppress MSVC warnings about weak functions by using
Bruce Momjian [Wed, 9 Aug 2006 03:09:56 +0000 (03:09 +0000)]
Suppress MSVC warnings about weak functions by using
_CRT_SECURE_NO_DEPRECATE.

Hiroshi Saito

18 years agoUpdate packages description:
Bruce Momjian [Wed, 9 Aug 2006 02:48:10 +0000 (02:48 +0000)]
Update packages description:

<    A package would be a schema with its own variables,
<    private functions, and initialization functions.  It
>    A package would be a schema with public/private variables,
>    public/private functions, and initialization functions.  It

18 years agoUpdate packages:
Bruce Momjian [Tue, 8 Aug 2006 22:51:37 +0000 (22:51 +0000)]
Update packages:

<    private functions, and initialization functions
>    private functions, and initialization functions.  It
>    is also possible to implement these capabilities
>    in all schemas and not use a separate "packages"
>    syntax at all.

18 years agoFix win32.mak for config cleanup.
Bruce Momjian [Tue, 8 Aug 2006 22:44:05 +0000 (22:44 +0000)]
Fix win32.mak for config cleanup.

Hiroshi Saito

18 years agoUpdate:
Bruce Momjian [Tue, 8 Aug 2006 22:39:09 +0000 (22:39 +0000)]
Update:

<  o COPY FROM STDIN not implemented because not useful?
>  o Implement COPY FROM STDIN

18 years agoFix 'make clean' for change of regression.diffs file name.
Tom Lane [Tue, 8 Aug 2006 19:43:11 +0000 (19:43 +0000)]
Fix 'make clean' for change of regression.diffs file name.

18 years agoAdd a feature for automatic initialization and finalization of dynamically
Tom Lane [Tue, 8 Aug 2006 19:15:09 +0000 (19:15 +0000)]
Add a feature for automatic initialization and finalization of dynamically
loaded libraries: call functions _PG_init() and _PG_fini() if the library
defines such symbols.  Hence we no longer need to specify an initialization
function in preload_libraries: we can assume that the library used the
_PG_init() convention, instead.  This removes one source of pilot error
in use of preloaded libraries.  Original patch by Ralf Engelschall,
preload_libraries changes by me.

18 years agoReverse order of performance guides so powerpostgresql is first.
Bruce Momjian [Tue, 8 Aug 2006 19:05:24 +0000 (19:05 +0000)]
Reverse order of performance guides so powerpostgresql is first.

18 years agoAdd detail on packages:
Bruce Momjian [Tue, 8 Aug 2006 19:03:59 +0000 (19:03 +0000)]
Add detail on packages:

>
>    A package would be a schema with its own variables,
>    private functions, and initialization functions
>

18 years agoAdd package item:
Bruce Momjian [Tue, 8 Aug 2006 19:02:13 +0000 (19:02 +0000)]
Add package item:

<  o Add Oracle-style packages
>  o Add Oracle-style packages  (Pavel)
> * Pavel is Pavel Stehule <pavel.stehule@hotmail.com>

18 years agoFix definition of "errcode" for MSVC.
Bruce Momjian [Tue, 8 Aug 2006 18:49:14 +0000 (18:49 +0000)]
Fix definition of "errcode" for MSVC.

Hiroshi Saito

18 years agoDone:
Bruce Momjian [Tue, 8 Aug 2006 17:40:02 +0000 (17:40 +0000)]
Done:

<    o Automatically force archiving of partially-filled WAL files when
>    o -Automatically force archiving of partially-filled WAL files when
<
<      Doing this will allow administrators to know more easily when
<      the archive contains all the files needed for point-in-time
<      recovery.
<      http://archives.postgresql.org/pgsql-patches/2005-04/msg00121.php
<
<    o Add reporting of the current WAL file and offset, perhaps as
>    o -Add reporting of the current WAL file and offset, perhaps as
<
<      The offset allows parts of a WAL file to be archived using
<      an external program.
<

18 years agoFix bug introduced by last patch, thanks again to Mario Weilguni <mweilguni@sime...
Teodor Sigaev [Tue, 8 Aug 2006 15:45:18 +0000 (15:45 +0000)]
Fix bug introduced by last patch, thanks again to Mario Weilguni <mweilguni@sime.com>

18 years agoCleaned up log output a little bit more.
Michael Meskes [Tue, 8 Aug 2006 15:30:39 +0000 (15:30 +0000)]
Cleaned up log output a little bit more.

18 years agoClarified COPY FROM STDIN in ecpg todo
Michael Meskes [Tue, 8 Aug 2006 11:52:51 +0000 (11:52 +0000)]
Clarified COPY FROM STDIN in ecpg todo

18 years agoMade parser check for valid copy to/from stdin/stdout combinations.
Michael Meskes [Tue, 8 Aug 2006 11:51:25 +0000 (11:51 +0000)]
Made parser check for valid copy to/from stdin/stdout combinations.
Lots of small changes in regression test suite

18 years agoDone:
Bruce Momjian [Tue, 8 Aug 2006 01:24:36 +0000 (01:24 +0000)]
Done:

< * Allow protocol-level BIND parameter values to be logged
<
<   http://archives.postgresql.org/pgsql-hackers/2006-02/msg00165.php
> * -Allow protocol-level BIND parameter values to be logged, if text mode

18 years agoFor protocol-level prepare/bind/execute:
Bruce Momjian [Tue, 8 Aug 2006 01:23:15 +0000 (01:23 +0000)]
For protocol-level prepare/bind/execute:

o  print user name for all
o  print portal name if defined for all
o  print query for all
o  reduce log_statement header to single keyword
o  print bind parameters as DETAIL if text mode

18 years agoPut back another improperly-removed #include.
Tom Lane [Mon, 7 Aug 2006 21:56:25 +0000 (21:56 +0000)]
Put back another improperly-removed #include.

18 years agoOn some platforms, pg_usleep isn't interruptible by signals; fix
Tom Lane [Mon, 7 Aug 2006 17:41:42 +0000 (17:41 +0000)]
On some platforms, pg_usleep isn't interruptible by signals; fix
archiver to behave per original coder's expectation on these machines.
We already know this everywhere else AFAICT.

18 years agofix bug about modifying value in shared buffer,
Teodor Sigaev [Mon, 7 Aug 2006 17:39:04 +0000 (17:39 +0000)]
fix bug about modifying value in shared buffer,
what was a reason to corrupt index.
Thank to Mario Weilguni <mweilguni@sime.com> to
discover a bug.

18 years agoMake recovery from WAL be restartable, by executing a checkpoint-like
Tom Lane [Mon, 7 Aug 2006 16:57:57 +0000 (16:57 +0000)]
Make recovery from WAL be restartable, by executing a checkpoint-like
operation every so often.  This improves the usefulness of PITR log
shipping for hot standby: formerly, if the standby server crashed, it
was necessary to restart it from the last base backup and replay all
the WAL since then.  Now it will only need to reread about the same
amount of WAL as the master server would.  The behavior might also
come in handy during a long PITR replay sequence.  Simon Riggs,
with some editorialization by Tom Lane.

18 years agoJoachim fixed some bugs in numeric handling in pgtypeslib.
Michael Meskes [Mon, 7 Aug 2006 13:17:02 +0000 (13:17 +0000)]
Joachim fixed some bugs in numeric handling in pgtypeslib.
Fixed and cleaned up some regression tests. Also added a new one.

18 years agoTweak sanity_check regression test to display more tables (viz, those
Tom Lane [Sun, 6 Aug 2006 04:35:21 +0000 (04:35 +0000)]
Tweak sanity_check regression test to display more tables (viz, those
without indexes) but not to display temp tables.  It's a bit hard to
credit that sanity_check could get through a database-wide VACUUM
while the preceding create_index test is still trying to clean up
its temp tables ... but I see no other explanation for the current
failure report from buildfarm member sponge.

18 years agoAdd support for forcing a switch to a new xlog file; cause such a switch
Tom Lane [Sun, 6 Aug 2006 03:53:44 +0000 (03:53 +0000)]
Add support for forcing a switch to a new xlog file; cause such a switch
to happen automatically during pg_stop_backup().  Add some functions for
interrogating the current xlog insertion point and for easily extracting
WAL filenames from the hex WAL locations displayed by pg_stop_backup
and friends.  Simon Riggs with some editorialization by Tom Lane.