OSDN Git Service

pg-rex/syncrep.git
19 years agoDisallow the combination VACUUM FULL FREEZE for safety's sake, for the
Tom Lane [Thu, 2 Dec 2004 19:28:49 +0000 (19:28 +0000)]
Disallow the combination VACUUM FULL FREEZE for safety's sake, for the
reasons I outlined in pghackers a few days ago.

Also, undo someone's overly optimistic decision to reduce tuple state
checks from if (...) elog() to Asserts.  If I trusted this code more,
I might think it was a good idea to disable these checks in production
installations.  But I don't.

19 years agoAdd:
Bruce Momjian [Thu, 2 Dec 2004 18:13:47 +0000 (18:13 +0000)]
Add:

> * Fix psql's display of schema information (Neil)

19 years agoDon't bother adding to cflags/cppflags, just set them because configure
Bruce Momjian [Thu, 2 Dec 2004 18:11:40 +0000 (18:11 +0000)]
Don't bother adding to cflags/cppflags, just set them because configure
handles that, and make solaris debug use no optimization.

19 years agoWording improvement.
Bruce Momjian [Thu, 2 Dec 2004 17:39:54 +0000 (17:39 +0000)]
Wording improvement.

19 years agoDocument that btrim()'s second parameter is optional, and defaults to spaces.
Bruce Momjian [Thu, 2 Dec 2004 17:14:54 +0000 (17:14 +0000)]
Document that btrim()'s second parameter is optional, and defaults to spaces.

19 years agoUpdate Brazilian FAQ.
Bruce Momjian [Thu, 2 Dec 2004 15:49:25 +0000 (15:49 +0000)]
Update Brazilian FAQ.

Euler Taveira de Oliveira

19 years agoRework libpq threaded SIGPIPE handling to avoid interference with
Bruce Momjian [Thu, 2 Dec 2004 15:32:54 +0000 (15:32 +0000)]
Rework libpq threaded SIGPIPE handling to avoid interference with
calling applications.  This is done by blocking sigpipe in the libpq
thread and using sigpending/sigwait to possibily discard any sigpipe we
generated.

19 years agoFix a potential free() of string literal in oid2name, per report from
Neil Conway [Thu, 2 Dec 2004 06:14:50 +0000 (06:14 +0000)]
Fix a potential free() of string literal in oid2name, per report from
Michael Fuhr. Along the way, fix a bunch of accesses to uninitialized
memory, add a mystrdup() routine and use it to cleanup some code.

19 years agoAdd:
Bruce Momjian [Thu, 2 Dec 2004 04:40:37 +0000 (04:40 +0000)]
Add:

>
> * Allow FOR UPDATE queries to do NOWAIT locks
>

19 years agoThis is a little patch to correct the documentation on CVS. The URL for
Neil Conway [Thu, 2 Dec 2004 04:12:12 +0000 (04:12 +0000)]
This is a little patch to correct the documentation on CVS. The URL for
downloading CVS at cyclic.com site is long defunct, and I changed the text
to not overtly recommend CVS 1.10, a now fairly old version.

Jon Jensen

19 years agoUpdate Japanese FAQ.
Bruce Momjian [Thu, 2 Dec 2004 03:46:47 +0000 (03:46 +0000)]
Update Japanese FAQ.

Jun Kuwamura

19 years agoTeach regex_fixed_prefix() the correct handling of advanced regex
Tom Lane [Thu, 2 Dec 2004 02:45:07 +0000 (02:45 +0000)]
Teach regex_fixed_prefix() the correct handling of advanced regex
escapes --- they aren't simply quoted characters.  Problem noted by
Antti Salmela.  Also fix problem with incorrect handling of multibyte
characters when followed by a quantifier.

19 years agoMake some adjustments to reduce platform dependencies in plan selection.
Tom Lane [Thu, 2 Dec 2004 01:34:18 +0000 (01:34 +0000)]
Make some adjustments to reduce platform dependencies in plan selection.
In particular, there was a mathematical tie between the two possible
nestloop-with-materialized-inner-scan plans for a join (ie, we computed
the same cost with either input on the inside), resulting in a roundoff
error driven choice, if the relations were both small enough to fit in
sort_mem.  Add a small cost factor to ensure we prefer materializing the
smaller input.  This changes several regression test plans, but with any
luck we will now have more stability across platforms.

19 years agoAllow libpq to build on MS Visual Studio .NET 2003 on Windows XP.
Bruce Momjian [Wed, 1 Dec 2004 23:42:26 +0000 (23:42 +0000)]
Allow libpq to build on MS Visual Studio .NET 2003 on Windows XP.

19 years agoUpdate French FAQ.
Bruce Momjian [Wed, 1 Dec 2004 22:17:16 +0000 (22:17 +0000)]
Update French FAQ.

Guillaume LELARGE

19 years agoFix timestamptz_age() to do calculation in local timezone not GMT, per bug 1332.
Tom Lane [Wed, 1 Dec 2004 19:57:49 +0000 (19:57 +0000)]
Fix timestamptz_age() to do calculation in local timezone not GMT, per bug 1332.

19 years agoClarify documentation of age() functions.
Tom Lane [Wed, 1 Dec 2004 19:32:12 +0000 (19:32 +0000)]
Clarify documentation of age() functions.

19 years agoChange planner to use the current true disk file size as its estimate of
Tom Lane [Wed, 1 Dec 2004 19:00:56 +0000 (19:00 +0000)]
Change planner to use the current true disk file size as its estimate of
a relation's number of blocks, rather than the possibly-obsolete value
in pg_class.relpages.  Scale the value in pg_class.reltuples correspondingly
to arrive at a hopefully more accurate number of rows.  When pg_class
contains 0/0, estimate a tuple width from the column datatypes and divide
that into current file size to estimate number of rows.  This improved
methodology allows us to jettison the ancient hacks that put bogus default
values into pg_class when a table is first created.  Also, per a suggestion
from Simon, make VACUUM (but not VACUUM FULL or ANALYZE) adjust the value
it puts into pg_class.reltuples to try to represent the mean tuple density
instead of the minimal density that actually prevails just after VACUUM.
These changes alter the plans selected for certain regression tests, so
update the expected files accordingly.  (I removed join_1.out because
it's not clear if it still applies; we can add back any variant versions
as they are shown to be needed.)

19 years agoAdd:
Bruce Momjian [Wed, 1 Dec 2004 17:39:18 +0000 (17:39 +0000)]
Add:

> * Allow ecpg to work with MSVC and BCC

19 years agoUpdate with Gavin's additions.
Bruce Momjian [Wed, 1 Dec 2004 17:21:58 +0000 (17:21 +0000)]
Update with Gavin's additions.

19 years agoRemove question mark on COPY CSV fix:
Bruce Momjian [Tue, 30 Nov 2004 22:15:21 +0000 (22:15 +0000)]
Remove question mark on COPY CSV fix:

<    returns in data?
>    returns in data

19 years agoAdd \n to end of output for psql \s with no history file name.
Bruce Momjian [Tue, 30 Nov 2004 20:00:34 +0000 (20:00 +0000)]
Add \n to end of output for psql \s with no history file name.

This prevents the next psql prompt from showing up on the end of the
last history line.

19 years agoUpdate Turkish FAQ.
Bruce Momjian [Tue, 30 Nov 2004 19:42:38 +0000 (19:42 +0000)]
Update Turkish FAQ.

Devrim GUNDUZ

19 years agoRemoved:
Bruce Momjian [Tue, 30 Nov 2004 19:32:24 +0000 (19:32 +0000)]
Removed:

< * Make log_min_duration_statement output when the duration is reached rather
<   than when the statement completes
<
<   This prints long queries while they are running, making trouble shooting
<   easier.  Also, it eliminates the need for log_statement because it
<   would now be the same as a log_min_duration_statement of zero.

19 years agoQuote SHELL arguments only on Win32.
Bruce Momjian [Tue, 30 Nov 2004 19:01:28 +0000 (19:01 +0000)]
Quote SHELL arguments only on Win32.

19 years agoAdd:
Bruce Momjian [Tue, 30 Nov 2004 18:20:35 +0000 (18:20 +0000)]
Add:

> * Make log_min_duration_statement output when the duration is reached rather
>   than when the statement completes
>
>   This prints long queries while they are running, making trouble shooting
>   easier.  Also, it eliminates the need for log_statement because it
>   would now be the same as a log_min_duration_statement of zero.

19 years agoFix readline/libedit selection code to prefer readline over libedit
Tom Lane [Tue, 30 Nov 2004 06:13:04 +0000 (06:13 +0000)]
Fix readline/libedit selection code to prefer readline over libedit
reliably (ie, regardless of which libraries they depend on).  Also
make sure that we don't select headers that obviously belong to the
wrong one of the two libraries.  This was discussed back around 4-Sep
but seems to have slipped through the cracks.  The header selection
could be checked more closely, perhaps, but let's see if this is good
enough.

19 years agoUpdate wording:
Bruce Momjian [Tue, 30 Nov 2004 05:06:33 +0000 (05:06 +0000)]
Update wording:

< you would like to work on an item, please read the developer's FAQ
> you would like to work on an item, please read the Developer's FAQ

19 years agoUpdated TODO list description:
Bruce Momjian [Tue, 30 Nov 2004 05:03:47 +0000 (05:03 +0000)]
Updated TODO list description:

9c5
< Last updated: Mon Nov 29 23:33:09 EST 2004
> Last updated: Tue Nov 30 00:03:37 EST 2004
13a10,16
> #A hyphen, "-", marks changes that will appear in the upcoming 8.1 release.#
>
> Bracketed items, "[]", have more detail.
>
> This list contains all known PostgreSQL bugs and feature requests. If
> you would like to work on an item, please read the developer's FAQ
> first.

19 years agoAdd:
Bruce Momjian [Tue, 30 Nov 2004 04:33:14 +0000 (04:33 +0000)]
Add:

> * Prevent INET cast to CIDR from droping netmask, SELECT '1.1.1.1'::inet::cidr

19 years agoAdd:
Bruce Momjian [Tue, 30 Nov 2004 03:53:00 +0000 (03:53 +0000)]
Add:

>  o Allow COPY FROM ... CVS to interpret newlines and carriage
>    returns in data?
>
>    This would require major refactoring of the copy source code.

19 years agoFix a few typos in comments.
Neil Conway [Tue, 30 Nov 2004 03:50:29 +0000 (03:50 +0000)]
Fix a few typos in comments.

19 years agoAdd:
Bruce Momjian [Tue, 30 Nov 2004 03:27:10 +0000 (03:27 +0000)]
Add:

> * Improve NLS maintenace of libpgport messages linked onto applications
>

19 years agoUpdate plperl to use ereport() not elog() for user-facing messages,
Tom Lane [Mon, 29 Nov 2004 20:11:06 +0000 (20:11 +0000)]
Update plperl to use ereport() not elog() for user-facing messages,
so that they will be translatable.  Give messages some semblance of
conformance to the style guide.

19 years agoAdd:
Bruce Momjian [Mon, 29 Nov 2004 03:17:30 +0000 (03:17 +0000)]
Add:
<
> * Allow cross-compiling by generating the zic database on the target system

19 years agoClean up initdb's error handling so that it prints something more
Tom Lane [Mon, 29 Nov 2004 03:05:03 +0000 (03:05 +0000)]
Clean up initdb's error handling so that it prints something more
useful than just \'failed\' when there's a problem.  Per gripe from
Chris Albertson.

In an unrelated change, use VACUUM FULL; VACUUM FREEZE; rather than
a single VACUUM FULL FREEZE command, to respond to my worries of a
couple days ago about the reliability of doing this in one go.

19 years agoUpdate Turkish FAQ.
Bruce Momjian [Mon, 29 Nov 2004 03:04:31 +0000 (03:04 +0000)]
Update Turkish FAQ.

Devrim GUNDUZ

19 years agoAdd comment explaining possible compiler warning:
Bruce Momjian [Mon, 29 Nov 2004 03:01:54 +0000 (03:01 +0000)]
Add comment explaining possible compiler warning:

    /*
     *  Some compilers with throw a warning knowing this test can never be
     *  true because off_t can't exceed the compared maximum.
     */
    if (th->fileLen > MAX_TAR_MEMBER_FILELEN)
        die_horribly(AH, modulename, "archive member too large for tar format\n");

19 years agoAdd:
Bruce Momjian [Mon, 29 Nov 2004 02:29:28 +0000 (02:29 +0000)]
Add:

> * Auto-vacuum
>  o Move into the backend code
>  o Scan the buffer cache to find free space or use background writer
>  o Use free-space map information to guide refilling

19 years agoTypo fixes.
Bruce Momjian [Mon, 29 Nov 2004 01:58:10 +0000 (01:58 +0000)]
Typo fixes.

Euler Taveira de Oliveira

19 years agoHave initdb display relative paths to start postmaster if used to invoke
Bruce Momjian [Mon, 29 Nov 2004 01:14:45 +0000 (01:14 +0000)]
Have initdb display relative paths to start postmaster if used to invoke
initdb, and display in a path-native way.

19 years agoMake pg_dumplo schema-aware. Karel Zak
Tom Lane [Sun, 28 Nov 2004 23:49:49 +0000 (23:49 +0000)]
Make pg_dumplo schema-aware.  Karel Zak

19 years agoAvoid scribbling on original parsetree during DECLARE CURSOR. This
Tom Lane [Sun, 28 Nov 2004 22:16:31 +0000 (22:16 +0000)]
Avoid scribbling on original parsetree during DECLARE CURSOR.  This
prevents problems when the DECLARE is in a portal and is executed
repeatedly, as is possible in v3 protocol.  Per analysis by Oliver
Jowett, though I didn't use his patch exactly.

19 years agoAdd description:
Bruce Momjian [Sun, 28 Nov 2004 05:12:19 +0000 (05:12 +0000)]
Add description:

<   information, either by name or offset from UTC
>   information, either zone name or offset from UTC
>
>   If the TIMESTAMP value is stored with a time zone name, interval
>   computations should adjust based on the time zone rules, e.g. adding
>   24 hours to a timestamp would yield a different result from adding one
>   day.
>

19 years agoMega-clarifications from Joachim Wieland.
Bruce Momjian [Sun, 28 Nov 2004 04:56:04 +0000 (04:56 +0000)]
Mega-clarifications from Joachim Wieland.

19 years agoWork around lack of NLS support in libpgport by making those components
Peter Eisentraut [Sat, 27 Nov 2004 22:44:15 +0000 (22:44 +0000)]
Work around lack of NLS support in libpgport by making those components
who use it scan the relevant source files for their own catalog.  It
creates a bit of duplicate work for translators, but it gets the job done
for now.

19 years agoCorrectly name the Windows operating systems
Peter Eisentraut [Sat, 27 Nov 2004 21:56:05 +0000 (21:56 +0000)]
Correctly name the Windows operating systems

19 years agoUpdate of conformance information to SQL:2003
Peter Eisentraut [Sat, 27 Nov 2004 21:27:08 +0000 (21:27 +0000)]
Update of conformance information to SQL:2003
by Troels Arvin, Simon Riggs, Elein Mustain

Make spelling of SQL standard names uniform.

19 years agoFix HAVE_OPTRESET to be HAVE_INT_OPTRESET. Typos spotted by Lorne Sunley.
Tom Lane [Sat, 27 Nov 2004 18:51:08 +0000 (18:51 +0000)]
Fix HAVE_OPTRESET to be HAVE_INT_OPTRESET.  Typos spotted by Lorne Sunley.

19 years agoAdd web development faq item.
Bruce Momjian [Sat, 27 Nov 2004 06:02:53 +0000 (06:02 +0000)]
Add web development faq item.

19 years agoAdd mention of Cygwin. Update version number.
Bruce Momjian [Sat, 27 Nov 2004 05:15:21 +0000 (05:15 +0000)]
Add mention of Cygwin. Update version number.

19 years agoAdd to CPPFLAGS/CFLAGS in template files, don't over-ride them.
Bruce Momjian [Sat, 27 Nov 2004 05:03:26 +0000 (05:03 +0000)]
Add to CPPFLAGS/CFLAGS in template files, don't over-ride them.

19 years agoThis adds mention of my latest tweak to the tsearch2/pg_trgm
Tom Lane [Sat, 27 Nov 2004 00:01:02 +0000 (00:01 +0000)]
This adds mention of my latest tweak to the tsearch2/pg_trgm
integration.  It is much better to create a word list of unstemmed words
than stemmed ones.

Chris K-L

19 years agoRelOptInfo.pages should really be declared as BlockNumber, not long.
Tom Lane [Fri, 26 Nov 2004 21:08:35 +0000 (21:08 +0000)]
RelOptInfo.pages should really be declared as BlockNumber, not long.

19 years agoInstall Tcl regex fixes to sync our regex engine with Tcl 8.4.8 (up from
Tom Lane [Wed, 24 Nov 2004 22:56:54 +0000 (22:56 +0000)]
Install Tcl regex fixes to sync our regex engine with Tcl 8.4.8 (up from
8.4.1).  This corrects some curious regex bugs, though not the greediness
issue I was hoping to find a solution for :-(

19 years agoOur interface code for Spencer's regexp package was checking for regexp
Tom Lane [Wed, 24 Nov 2004 22:44:07 +0000 (22:44 +0000)]
Our interface code for Spencer's regexp package was checking for regexp
error conditions during regexp compile, but not during regexp execution;
any sort of "can't happen" errors would be treated as no-match instead
of being reported as they should be.  Noticed while trying to duplicate
a reported Tcl bug.

19 years agoA client_encoding specification coming from the connection request has
Tom Lane [Wed, 24 Nov 2004 19:51:05 +0000 (19:51 +0000)]
A client_encoding specification coming from the connection request has
to be processed by GUC before InitPostgres, because any required lookup
of the encoding conversion function has to be done during InitializeClientEncoding.
So, I broke this last week by moving GUC processing to after InitPostgres :-(.
What we can do as a compromise is process non-SUSET variables during
command line scanning (the same as before), and postpone the processing
of only SUSET variables.  None of the SUSET variables need to be set
before InitPostgres.

19 years agoAvoid getting bit by roundoff error while checking $Safe::VERSION.
Tom Lane [Wed, 24 Nov 2004 18:47:38 +0000 (18:47 +0000)]
Avoid getting bit by roundoff error while checking $Safe::VERSION.
Per report from Mark Kirkwood.

19 years agoTrivial fix for a bug introduced in r1.94 of help.c: we want to only
Neil Conway [Wed, 24 Nov 2004 04:58:42 +0000 (04:58 +0000)]
Trivial fix for a bug introduced in r1.94 of help.c: we want to only
include "\s" in \? output when readline is enabled, but that commit
supressed "\s" whether readline was enabled or not.

19 years agoAssert that BufferIsPinned() in IncrBufferRefCount(), rather than using
Neil Conway [Wed, 24 Nov 2004 02:56:17 +0000 (02:56 +0000)]
Assert that BufferIsPinned() in IncrBufferRefCount(), rather than using
a home-brewed combination of assertions that boiled down to the same
thing.

19 years agoPrevent pgcrypto from successfully compiling if no valid random source
Neil Conway [Tue, 23 Nov 2004 23:44:08 +0000 (23:44 +0000)]
Prevent pgcrypto from successfully compiling if no valid random source
has been defined. Previously, pgcrypto would compile but would be
unusable.

19 years agoNew translations
Peter Eisentraut [Tue, 23 Nov 2004 23:13:26 +0000 (23:13 +0000)]
New translations

19 years agoTranslation updates
Peter Eisentraut [Tue, 23 Nov 2004 22:16:04 +0000 (22:16 +0000)]
Translation updates

19 years agoClean files generated by htmlhelp target.
Peter Eisentraut [Tue, 23 Nov 2004 21:41:24 +0000 (21:41 +0000)]
Clean files generated by htmlhelp target.

19 years agoAdd htmlhelp stylesheet and build target.
Peter Eisentraut [Tue, 23 Nov 2004 16:27:59 +0000 (16:27 +0000)]
Add htmlhelp stylesheet and build target.

19 years agoFurther plperl cleanup: be more paranoid about checking the type of
Tom Lane [Tue, 23 Nov 2004 00:21:24 +0000 (00:21 +0000)]
Further plperl cleanup: be more paranoid about checking the type of
data returned from Perl.  Consolidate multiple bits of code to convert
a Perl hash to a tuple, and drive the conversion off the keys present
in the hash rather than the tuple column names, so we detect error if
the hash contains keys it shouldn't.  (This means keys not in the hash
will silently default to NULL, which seems ok to me.)  Fix a bunch of
reference-count leaks too.

19 years agoThe beginnings of a regression test for plperl. Right now it only
Tom Lane [Mon, 22 Nov 2004 20:32:59 +0000 (20:32 +0000)]
The beginnings of a regression test for plperl.  Right now it only
covers return value processing, but that was the most broken stuff...

19 years agoTry to instill some sanity in plperl's function result processing.
Tom Lane [Mon, 22 Nov 2004 20:31:53 +0000 (20:31 +0000)]
Try to instill some sanity in plperl's function result processing.
Get rid of static variables for SETOF result, don't crash when called
from non-FROM context, eliminate dead code, etc.

19 years agoFix typo in release notes. Michael Fuhr.
Neil Conway [Mon, 22 Nov 2004 07:30:22 +0000 (07:30 +0000)]
Fix typo in release notes. Michael Fuhr.

19 years agoupdate us to beta5
PostgreSQL Daemon [Mon, 22 Nov 2004 03:06:35 +0000 (03:06 +0000)]
update us to beta5

19 years agoUpdate release notes for recent changes; also minor editorialization.
Tom Lane [Mon, 22 Nov 2004 00:37:59 +0000 (00:37 +0000)]
Update release notes for recent changes; also minor editorialization.

19 years agoFix rounding problem in dynahash.c's decision about when the target
Tom Lane [Sun, 21 Nov 2004 22:57:00 +0000 (22:57 +0000)]
Fix rounding problem in dynahash.c's decision about when the target
fill factor has been exceeded.  We usually run with ffactor == 1, but
the way the test was coded, it wouldn't split a bucket until the actual
fill factor reached 2.0, because of use of integer division.  Change
from > to >= so that it will split more aggressively when the table
starts to get full.

19 years agoReduce the default size of the PortalHashTable in order to save a
Tom Lane [Sun, 21 Nov 2004 22:48:01 +0000 (22:48 +0000)]
Reduce the default size of the PortalHashTable in order to save a
few cycles during transaction exit.  A typical session probably
wouldn't have as many as half a dozen portals open at once, so the
original value of 64 seems far larger than needed.

19 years agoWhile fixing plperl and pltcl, I realized plpgsql wasn't doing
Tom Lane [Sun, 21 Nov 2004 22:27:34 +0000 (22:27 +0000)]
While fixing plperl and pltcl, I realized plpgsql wasn't doing
subtransactions quite right either: the ReleaseCurrentSubTransaction
call should occur inside the PG_TRY, so that the proper path is taken
if an error occurs during subtransaction commit.  This assumes that
AbortSubTransaction can cope with the state left behind if
CommitSubTransaction fails partway through, but we were already
requiring that.

19 years agoSuppress remaining compile warnings, and add a comment about why
Tom Lane [Sun, 21 Nov 2004 22:13:37 +0000 (22:13 +0000)]
Suppress remaining compile warnings, and add a comment about why
it's not really broken.  Andrew Dunstan

19 years agoFix plperl and pltcl error handling per my previous proposal. SPI
Tom Lane [Sun, 21 Nov 2004 21:17:07 +0000 (21:17 +0000)]
Fix plperl and pltcl error handling per my previous proposal.  SPI
operations are now run as subtransactions, so that errors in them
can be reported as ordinary Perl or Tcl errors and caught by the
normal error handling convention of those languages.  Also do some
minor code cleanup in pltcl.c: extract a large chunk of duplicated
code in pltcl_SPI_execute and pltcl_SPI_execute_plan into a shared
subroutine.

19 years agoFix typo.
Bruce Momjian [Sat, 20 Nov 2004 22:28:39 +0000 (22:28 +0000)]
Fix typo.

19 years agoPatch of 2004-03-30 corrected date_part(timestamp) for extracting
Tom Lane [Sat, 20 Nov 2004 22:12:44 +0000 (22:12 +0000)]
Patch of 2004-03-30 corrected date_part(timestamp) for extracting
the year from a BC date, but failed to make the same fix in
date_part(timestamptz).

19 years agoFix typo, per Bruno Wolff.
Tom Lane [Sat, 20 Nov 2004 21:44:24 +0000 (21:44 +0000)]
Fix typo, per Bruno Wolff.

19 years agoFix up makefile dependencies for pg_config_paths.h.
Tom Lane [Sat, 20 Nov 2004 21:13:06 +0000 (21:13 +0000)]
Fix up makefile dependencies for pg_config_paths.h.

19 years agoAvoid scanning the relcache during AtEOSubXact_RelationCache when there
Tom Lane [Sat, 20 Nov 2004 20:19:52 +0000 (20:19 +0000)]
Avoid scanning the relcache during AtEOSubXact_RelationCache when there
is nothing to do, which is most of the time.  This is another simple
improvement to cut subtransaction entry/exit overhead.

19 years agoReduce the default size of the local lock hash table. There's usually
Tom Lane [Sat, 20 Nov 2004 20:16:54 +0000 (20:16 +0000)]
Reduce the default size of the local lock hash table.  There's usually
no need for it to be nearly as big as the global hash table, and since
it's not in shared memory it can grow if it does need to be bigger.
By reducing the size, we speed up hash_seq_search(), which saves a
significant fraction of subtransaction entry/exit overhead.

19 years agoFix plperl's elog() function to convert elog(ERROR) into Perl croak(),
Tom Lane [Sat, 20 Nov 2004 19:07:40 +0000 (19:07 +0000)]
Fix plperl's elog() function to convert elog(ERROR) into Perl croak(),
rather than longjmp'ing clear out of Perl and thereby leaving Perl in
a broken state.  Also some minor prettification of error messages.
Still need to do something with spi_exec_query() error handling.

19 years agoFix one more place where we were expecting lcons() to be nondestructive
Tom Lane [Sat, 20 Nov 2004 17:59:31 +0000 (17:59 +0000)]
Fix one more place where we were expecting lcons() to be nondestructive
to the original List; per report from Sebastian BÎck.  I think this is
the last such bug --- I examined every lcons() call in the backend and
the rest seem OK --- but it's nervous-making that we're still finding
'em so many months after the List rewrite went in.

19 years agoOn some HPUX versions/hardware, it seems we need to include libgcc.a in
Tom Lane [Sat, 20 Nov 2004 03:52:00 +0000 (03:52 +0000)]
On some HPUX versions/hardware, it seems we need to include libgcc.a in
the regression test shared libraries.

19 years agoFix unportable isdigit() call --- must cast arg to unsigned char.
Tom Lane [Sat, 20 Nov 2004 02:09:47 +0000 (02:09 +0000)]
Fix unportable isdigit() call --- must cast arg to unsigned char.

19 years agoMove pgstat_report_tabstat() call so that stats are not reported to the
Tom Lane [Sat, 20 Nov 2004 00:48:58 +0000 (00:48 +0000)]
Move pgstat_report_tabstat() call so that stats are not reported to the
collector until the transaction commits.  Per recent discussion, this
should avoid confusing autovacuum when an updating transaction runs for
a long time.

19 years agoMake sure we produce a useful error message for the case where the
Tom Lane [Sat, 20 Nov 2004 00:35:13 +0000 (00:35 +0000)]
Make sure we produce a useful error message for the case where the
postgresql.crt file simply isn't there, too.

19 years agoImprove error reporting for SSL connection failures. Remove redundant
Tom Lane [Sat, 20 Nov 2004 00:18:18 +0000 (00:18 +0000)]
Improve error reporting for SSL connection failures.  Remove redundant
free operations in client_cert_cb --- openssl will also attempt to free
these structures, resulting in core dumps.

19 years agoFix incorrect documentation of plperl's method for accessing the number
Tom Lane [Fri, 19 Nov 2004 23:22:54 +0000 (23:22 +0000)]
Fix incorrect documentation of plperl's method for accessing the number
of rows processed by a SPI query (David Fetter); also some other minor
editorial cleanup (Tom Lane).

19 years agoClean up rpath handling for HPUX --- we can't use the cc-style rpath
Tom Lane [Fri, 19 Nov 2004 21:27:42 +0000 (21:27 +0000)]
Clean up rpath handling for HPUX --- we can't use the cc-style rpath
switch syntax when calling ld directly.

19 years agoSet rpath for plperl and plpython to ensure we can find the .so files
Tom Lane [Fri, 19 Nov 2004 19:23:01 +0000 (19:23 +0000)]
Set rpath for plperl and plpython to ensure we can find the .so files
for the languages even when not installed in a standard directory.
pltcl may need this treatment as well, but we don't have the right path
conveniently available, so I'll leave it alone as long as there aren't
actual reports of trouble.

19 years agoModify the platform-specific makefiles so that macro 'rpath' is defined
Tom Lane [Fri, 19 Nov 2004 00:41:39 +0000 (00:41 +0000)]
Modify the platform-specific makefiles so that macro 'rpath' is defined
in terms of macro 'rpathdir', as I proposed a few weeks ago.  In itself
this commit shouldn't change the behavior at all, but it opens the door
to using special rpaths for the PL shared libraries, as seems to be
needed for plperl in particular.

19 years agoFix memory clobber problem reported by John Hansen: plperl_safe_init()
Tom Lane [Thu, 18 Nov 2004 21:35:42 +0000 (21:35 +0000)]
Fix memory clobber problem reported by John Hansen: plperl_safe_init()
may expand the Perl stack, therefore we must SPAGAIN to reload the local
stack pointer after calling it.  Also a couple other marginal readability
improvements.

19 years agoWhoops, missed converting the other sleep() call to pg_usleep().
Tom Lane [Thu, 18 Nov 2004 17:13:38 +0000 (17:13 +0000)]
Whoops, missed converting the other sleep() call to pg_usleep().

19 years agoTranslation update
Peter Eisentraut [Thu, 18 Nov 2004 10:09:59 +0000 (10:09 +0000)]
Translation update

19 years agoForce pg_database updates out to disk immediately after ALTER DATABASE;
Tom Lane [Thu, 18 Nov 2004 01:14:26 +0000 (01:14 +0000)]
Force pg_database updates out to disk immediately after ALTER DATABASE;
this is to avoid scenarios where incoming backends find no live copies
of a database's row because the only live copy is in an as-yet-unwritten
shared buffer, which they can't see.  Also, use FlushRelationBuffers()
for forcing out pg_database, instead of the much more expensive BufferSync().
There's no need to write out pages belonging to other relations.

19 years agoInclude <getopt.h> if available, to avoid warnings on Windows.
Tom Lane [Wed, 17 Nov 2004 21:37:47 +0000 (21:37 +0000)]
Include <getopt.h> if available, to avoid warnings on Windows.

19 years agoRename vacuum_cost_limit switch from -N to -l to avoid conflict with
Tom Lane [Wed, 17 Nov 2004 21:30:36 +0000 (21:30 +0000)]
Rename vacuum_cost_limit switch from -N to -l to avoid conflict with
existing Windows-only switch.

19 years agoRemove ill-considered suppression of gcc warnings in plperl, and fix
Tom Lane [Wed, 17 Nov 2004 21:23:36 +0000 (21:23 +0000)]
Remove ill-considered suppression of gcc warnings in plperl, and fix
some of the bugs exposed thereby.  The remaining 'might be used uninitialized'
warnings look like live bugs, but I am not familiar enough with Perl/C hacking
to tell how to fix them.