]> granicus.if.org Git - postgresql/log
postgresql
20 years agoRemove mmap TODO.detail, now in TODO.
Bruce Momjian [Tue, 9 Nov 2004 03:57:16 +0000 (03:57 +0000)]
Remove mmap TODO.detail, now in TODO.

20 years agoRemove inheritance, already in TODO.
Bruce Momjian [Tue, 9 Nov 2004 03:01:48 +0000 (03:01 +0000)]
Remove inheritance, already in TODO.

20 years agoUse a hopefully-more-reliable method of detecting default selectivity
Tom Lane [Tue, 9 Nov 2004 00:34:46 +0000 (00:34 +0000)]
Use a hopefully-more-reliable method of detecting default selectivity
estimates when combining the estimates for a range query.  As pointed out
by Miquel van Smoorenburg, the existing check for an impossible combined
result would quite possibly fail to detect one default and one non-default
input.  It seems better to use the default range query estimate in such
cases.  To do so, add a check for an estimate of exactly DEFAULT_INEQ_SEL.
This is a bit ugly because it introduces additional coupling between
clauselist_selectivity and scalarltsel/scalargtsel, but it's not like
there wasn't plenty already...

20 years agoRemove fadvise TODO.detail. Merged into TODO.
Bruce Momjian [Mon, 8 Nov 2004 20:16:43 +0000 (20:16 +0000)]
Remove fadvise TODO.detail.  Merged into TODO.

20 years agoUpdate:
Bruce Momjian [Mon, 8 Nov 2004 20:16:20 +0000 (20:16 +0000)]
Update:

<   posix_fadvise() [fadvise]
>   posix_fadvise()
>
>   Posix_fadvise() can control both sequential/random file caching and
>   free-behind behavior, but it is unclear how the setting affects other
>   backends that also have the file open, and the feature is not supported
>   on all operating systems.
>

20 years agoRecommend that archive_command be coded to not overwrite existing files.
Tom Lane [Mon, 8 Nov 2004 18:01:28 +0000 (18:01 +0000)]
Recommend that archive_command be coded to not overwrite existing files.
Add explicit documentation of the recovery configuration settings.  Other
minor improvements in the PITR docs.  Simon Riggs, some editorialization
by Tom Lane.

20 years agoUpdate workding on why we use is_absolute_path() as a macro.
Bruce Momjian [Mon, 8 Nov 2004 16:34:23 +0000 (16:34 +0000)]
Update workding on why we use is_absolute_path() as a macro.

20 years agoUse correct function name in example
Peter Eisentraut [Mon, 8 Nov 2004 15:57:05 +0000 (15:57 +0000)]
Use correct function name in example

20 years agoRemove count TODO.detail now that info is in TODO.
Bruce Momjian [Mon, 8 Nov 2004 15:35:52 +0000 (15:35 +0000)]
Remove count TODO.detail now that info is in TODO.

20 years agoRemove count TODO.detail.
Bruce Momjian [Mon, 8 Nov 2004 15:32:40 +0000 (15:32 +0000)]
Remove count TODO.detail.

20 years agoRemove atttypmod TODO.detail and merge into TODO list.
Bruce Momjian [Mon, 8 Nov 2004 15:19:31 +0000 (15:19 +0000)]
Remove atttypmod TODO.detail and merge into TODO list.

20 years agoUpdate item description:
Bruce Momjian [Mon, 8 Nov 2004 15:18:16 +0000 (15:18 +0000)]
Update item description:

< * CREATE TABLE AS can not determine column lengths from expressions [atttypmod]
> * Allow CREATE TABLE AS to determine column lengths for complex
>   expressions like SELECT col1 || col2

20 years agoRemove view TODO.detail.
Bruce Momjian [Mon, 8 Nov 2004 14:55:05 +0000 (14:55 +0000)]
Remove view TODO.detail.

20 years agoUpdate view info:
Bruce Momjian [Mon, 8 Nov 2004 14:54:00 +0000 (14:54 +0000)]
Update view info:

< * Automatically create rules on views so they are updateable, per SQL99 [view]
> * Automatically create rules on views so they are updateable, per SQL99
>
>   We can only auto-create rules for simple views.  For more complex
>   cases users will still have to write rules.
>

20 years agoRemove win32 TODO.detail.
Bruce Momjian [Mon, 8 Nov 2004 14:46:52 +0000 (14:46 +0000)]
Remove win32 TODO.detail.

20 years agoAllow win32/cygwin link against the first matching library symbol rather
Bruce Momjian [Mon, 8 Nov 2004 05:23:26 +0000 (05:23 +0000)]
Allow win32/cygwin link against the first matching library symbol rather
than erroring out.  This is the Unix behavior.

20 years agoKris Jurka pointed out that the qualified_name production wasn't
Tom Lane [Mon, 8 Nov 2004 04:02:20 +0000 (04:02 +0000)]
Kris Jurka pointed out that the qualified_name production wasn't
working as intended --- for some reason, FROM a.b.c was getting
parsed as if it were a function name and not a qualified name.
I think there must be a bug in bison, because it should have
complained that the grammar was ambiguous.  Anyway, fix it along
the same lines previously used for func_name vs columnref, and get
rid of the right-recursion in attrs that seems to have confused
bison.

20 years agoFix unportable code in SockAddr_cidr_mask: you can't assume that
Tom Lane [Mon, 8 Nov 2004 01:54:40 +0000 (01:54 +0000)]
Fix unportable code in SockAddr_cidr_mask: you can't assume that
shifting left by full word width gives zero.  Per bug report from
Tyson Thomson.

20 years agoNew translation
Peter Eisentraut [Sun, 7 Nov 2004 23:38:19 +0000 (23:38 +0000)]
New translation

20 years agoTranslation update
Peter Eisentraut [Sun, 7 Nov 2004 23:29:35 +0000 (23:29 +0000)]
Translation update

20 years agoTranslation updates
Peter Eisentraut [Sun, 7 Nov 2004 23:28:20 +0000 (23:28 +0000)]
Translation updates

20 years agoRemoves duplicate slashes from the path in canonicalize_path(). It
Bruce Momjian [Sun, 7 Nov 2004 02:12:17 +0000 (02:12 +0000)]
Removes duplicate slashes from the path in canonicalize_path().  It
preserve double leading slashes on Win32.

e.g.    ////a////b => /a/b

20 years agoAdd code to find_my_exec() to resolve a symbolic link down to the
Tom Lane [Sat, 6 Nov 2004 23:06:29 +0000 (23:06 +0000)]
Add code to find_my_exec() to resolve a symbolic link down to the
actual executable location.  This allows people to continue to use
setups where, eg, postmaster is symlinked from a convenient place.
Per gripe from Josh Berkus.

20 years agoSimplify relative-path logic, on the assumption that compiled-in paths
Tom Lane [Sat, 6 Nov 2004 21:39:45 +0000 (21:39 +0000)]
Simplify relative-path logic, on the assumption that compiled-in paths
won't have trivial differences in their common prefix.

20 years agoUpdate pg_dump to use SET DEFAULT_TABLESPACE instead of explicit
Tom Lane [Sat, 6 Nov 2004 19:36:02 +0000 (19:36 +0000)]
Update pg_dump to use SET DEFAULT_TABLESPACE instead of explicit
tablespace clauses; this should improve compatibility of dump files.
Philip Warner, some rework by Tom Lane.

20 years agoRemove directory from cmd.exe default shell.
Bruce Momjian [Sat, 6 Nov 2004 17:56:40 +0000 (17:56 +0000)]
Remove directory from cmd.exe default shell.

20 years agoWhen implementing a coercion to a domain type with a combined
Tom Lane [Sat, 6 Nov 2004 17:46:38 +0000 (17:46 +0000)]
When implementing a coercion to a domain type with a combined
type-and-length coercion function, make sure that the coercion function
is told the correct typmod.  Fixes Kris Jurka's example of a domain
over bit(N).

20 years agoReduce formatting entropy
Peter Eisentraut [Sat, 6 Nov 2004 14:32:10 +0000 (14:32 +0000)]
Reduce formatting entropy

20 years agoRemove:
Bruce Momjian [Sat, 6 Nov 2004 05:38:19 +0000 (05:38 +0000)]
Remove:

* Allow database recovery where tablespaces can't be created

  When a pg_dump is restored, all tablespaces will attempt to be created
  in their original locations. If this fails, the user must be able to
  adjust the restore process.

20 years agoNow that we advertize only CIDR address in pg_hba.conf, remove
Bruce Momjian [Sat, 6 Nov 2004 05:32:05 +0000 (05:32 +0000)]
Now that we advertize only CIDR address in pg_hba.conf, remove
duplicates sample entries from comments.

20 years agoVariable initializer no longer necessary.
Bruce Momjian [Sat, 6 Nov 2004 05:20:41 +0000 (05:20 +0000)]
Variable initializer no longer necessary.

20 years agoTests for COMSPEC in the right place and supply a reasonable shell
Bruce Momjian [Sat, 6 Nov 2004 04:29:40 +0000 (04:29 +0000)]
Tests for COMSPEC in the right place and supply a reasonable shell
default on Win32.

20 years agoFix sizeof(EXE) arithmetic, per Andrew Dunstan.
Tom Lane [Sat, 6 Nov 2004 04:24:14 +0000 (04:24 +0000)]
Fix sizeof(EXE) arithmetic, per Andrew Dunstan.

20 years ago*** empty log message ***
Bruce Momjian [Sat, 6 Nov 2004 03:25:20 +0000 (03:25 +0000)]
*** empty log message ***

20 years agoUpdate "dash" to "hyphen".
Bruce Momjian [Sat, 6 Nov 2004 03:24:21 +0000 (03:24 +0000)]
Update "dash" to "hyphen".

20 years agoCode cleanup in path.c and exec.c. Handle Windows drive and network specs
Tom Lane [Sat, 6 Nov 2004 01:16:22 +0000 (01:16 +0000)]
Code cleanup in path.c and exec.c.  Handle Windows drive and network specs
everywhere not just some places, get rid of . and .. when joining path
sections together.  This should eliminate most of the ugly paths like
/foo/bar/./baz that we've been generating.

20 years agopred_test() logic was being too narrow-minded about where it might find
Tom Lane [Fri, 5 Nov 2004 20:45:10 +0000 (20:45 +0000)]
pred_test() logic was being too narrow-minded about where it might find
RestrictInfo nodes in the query expression.  Per example from James Robinson.

20 years agoCreate 'default_tablespace' GUC variable that supplies a TABLESPACE
Tom Lane [Fri, 5 Nov 2004 19:17:13 +0000 (19:17 +0000)]
Create 'default_tablespace' GUC variable that supplies a TABLESPACE
clause implicitly whenever one is not given explicitly.  Remove concept
of a schema having an associated tablespace, and simplify the rules for
selecting a default tablespace for a table or index.  It's now just
(a) explicit TABLESPACE clause; (b) default_tablespace if that's not an
empty string; (c) database's default.  This will allow pg_dump to use
SET commands instead of tablespace clauses to determine object locations
(but I didn't actually make it do so).  All per recent discussions.

20 years agoSmall message clarifications
Peter Eisentraut [Fri, 5 Nov 2004 17:11:34 +0000 (17:11 +0000)]
Small message clarifications

20 years agoTranslation update
Peter Eisentraut [Fri, 5 Nov 2004 17:08:11 +0000 (17:08 +0000)]
Translation update

20 years agoFix typo:
Bruce Momjian [Fri, 5 Nov 2004 03:11:54 +0000 (03:11 +0000)]
Fix typo:

<   one column or expression indexes, perhaps using per=index statistics
>   one column or expression indexes, perhaps using per-index statistics

20 years agoAdd items:
Bruce Momjian [Fri, 5 Nov 2004 03:04:36 +0000 (03:04 +0000)]
Add items:

< * Add transaction_timestamp(), statement_timestamp(), clock_timestamp() functionality
> * Add transaction_timestamp(), statement_timestamp(), clock_timestamp()
>   functionality
278c279
<   one column
>   one column or expression indexes, perhaps using per=index statistics
846a848
> * Remove kerberos4 from source tree?

20 years agoFix psql \e and \! for Win32.
Bruce Momjian [Thu, 4 Nov 2004 22:25:14 +0000 (22:25 +0000)]
Fix psql \e and \! for Win32.

20 years agoMinor documentation updates from Simon Riggs.
Tom Lane [Thu, 4 Nov 2004 19:08:42 +0000 (19:08 +0000)]
Minor documentation updates from Simon Riggs.

20 years agoAdd:
Bruce Momjian [Thu, 4 Nov 2004 17:09:54 +0000 (17:09 +0000)]
Add:

>
> * Add fillfactor to control reserved free space during index creation

20 years agoAdd:
Bruce Momjian [Thu, 4 Nov 2004 16:53:58 +0000 (16:53 +0000)]
Add:

> * Allow accurate statistics to be collected on indexes with more than
>   one column

20 years agoUpdates from Simon Riggs:
Bruce Momjian [Thu, 4 Nov 2004 16:49:53 +0000 (16:49 +0000)]
Updates from Simon Riggs:

<   that can spam more than one table.
>   that can span more than one table.
239c239
<   rather than just col1
>   rather than just col1; also called skip-scanning.
641c641,642
< * Add free-behind capability for large sequential scans [fadvise]
> * Allow free-behind capability for large sequential scans, perhaps using
>   posix_fadvise() [fadvise]

20 years agoUpdate bitmap index entries:
Bruce Momjian [Thu, 4 Nov 2004 16:19:46 +0000 (16:19 +0000)]
Update bitmap index entries:

< * Allow the creation of bitmap indexes which can be quickly combined
<   with other bitmap indexes
> * Allow non-bitmap indexes to be combined by creating bitmaps in memory
259,261c258,259
<   combined.  Such indexes could be more compact if there are few unique
<   value.  Also, perhaps they can be lossy requiring a scan of the heap page
<   to find matching rows.
>   combined.  They can index by tid or can be lossy requiring a scan of the
>   heap page to find matching rows.
263c261,262
< * Allow non-bitmap indexes to be combined
> * Allow the creation of on-disk bitmap indexes which can be quickly
>   combined with other bitmap indexes
265,266c264
<   Do lookups on non-bitmap indexes and create bitmaps in memory that can be
<   combined with other indexes.
>   Such indexes could be more compact if there are few unique value.

20 years agoFix build break, per report from Kris Jurka. Rename README to README.mysql,
Neil Conway [Thu, 4 Nov 2004 09:12:35 +0000 (09:12 +0000)]
Fix build break, per report from Kris Jurka. Rename README to README.mysql,
for consistency.

20 years agoContrib build fixes:
Neil Conway [Thu, 4 Nov 2004 06:09:26 +0000 (06:09 +0000)]
Contrib build fixes:

- add some additional files to the dbmirror install (approved by
ssinger)

- add a makefile for contrib/mysql, and add mysql to the list of
contribs build by default

- use xml2-config to pickup -I flags for libxml2 in contrib/xml and
contrib/xml2

Original work from Martin Pitt of Debian, minor cleanups by Neil
Conway.

20 years agoUpdate:
Bruce Momjian [Thu, 4 Nov 2004 03:57:58 +0000 (03:57 +0000)]
Update:

< * Use bitmaps to combine existing indexes [performance]
> * Allow the creation of bitmap indexes which can be quickly combined
>   with other bitmap indexes
255,257c256,266
<   Bitmap indexes allow single indexed columns to be combined to
<   dynamically create a composite index to match a specific query. Each
<   index is a bitmap, and the bitmaps are AND'ed or OR'ed to be combined.
>   Bitmap indexes index single columns that can be combined with other bitmap
>   indexes to dynamically create a composite index to match a specific query.
>   Each index is a bitmap, and the bitmaps are bitwise AND'ed or OR'ed to be
>   combined.  Such indexes could be more compact if there are few unique
>   value.  Also, perhaps they can be lossy requiring a scan of the heap page
>   to find matching rows.
>
> * Allow non-bitmap indexes to be combined
>
>   Do lookups on non-bitmap indexes and create bitmaps in memory that can be
>   combined with other indexes.

20 years agoHonor TMPDIR.
Bruce Momjian [Thu, 4 Nov 2004 02:04:10 +0000 (02:04 +0000)]
Honor TMPDIR.

Add the script name to the tmp directory name.

Move trap up now that the dir is more unique.

20 years agoMove trap to after the directory is created.
Bruce Momjian [Wed, 3 Nov 2004 23:06:52 +0000 (23:06 +0000)]
Move trap to after the directory is created.

20 years agoCreate temporary files securely.
Bruce Momjian [Wed, 3 Nov 2004 22:46:15 +0000 (22:46 +0000)]
Create temporary files securely.

20 years agoUpdate tablespace temp usage item:
Bruce Momjian [Wed, 3 Nov 2004 22:20:52 +0000 (22:20 +0000)]
Update tablespace temp usage item:

<   This perhaps should use a round-robin allocation system where several
<   tablespaces are used in a cycle.  The cycle pointer should be global.
>   It could start with a random tablespace from a supplied list and cycle
>   through the list.

20 years agoAdd idea on auto-cluster:
Bruce Momjian [Wed, 3 Nov 2004 19:38:20 +0000 (19:38 +0000)]
Add idea on auto-cluster:

<    store heap rows in hashed groups.
>    store heap rows in hashed groups, perhaps using a user-supplied
>    hash function.

20 years agoAdd idea on auto-cluster:
Bruce Momjian [Wed, 3 Nov 2004 19:28:11 +0000 (19:28 +0000)]
Add idea on auto-cluster:

<    automatically access the heap data too.
>    automatically access the heap data too.  A third idea would be to
>    store heap rows in hashed groups.

20 years agoDon't create $DESTDIR/include/internal/lib, as it is no longer used.
Neil Conway [Wed, 3 Nov 2004 07:38:08 +0000 (07:38 +0000)]
Don't create $DESTDIR/include/internal/lib, as it is no longer used.

20 years agoTab completion for CREATE TABLESPACE arguments. Patch by Joachim Wieland.
Peter Eisentraut [Tue, 2 Nov 2004 16:10:05 +0000 (16:10 +0000)]
Tab completion for CREATE TABLESPACE arguments.  Patch by Joachim Wieland.

20 years agoUpdate:
Bruce Momjian [Tue, 2 Nov 2004 14:49:27 +0000 (14:49 +0000)]
Update:

< * Add a GUC variable to control the tablespace for temporary objects
> * Add a GUC variable to control the tablespace for temporary objects and
>   sort files
>
>   This perhaps should use a round-robin allocation system where several
>   tablespaces are used in a cycle.  The cycle pointer should be global.
>

20 years agoTranslation updates
Peter Eisentraut [Tue, 2 Nov 2004 09:28:41 +0000 (09:28 +0000)]
Translation updates

20 years agoTranslation updates
Peter Eisentraut [Tue, 2 Nov 2004 09:20:22 +0000 (09:20 +0000)]
Translation updates

20 years agoMove pthread.h test up to use PTHREAD_CFLAGS.
Bruce Momjian [Tue, 2 Nov 2004 05:44:45 +0000 (05:44 +0000)]
Move pthread.h test up to use PTHREAD_CFLAGS.

Kris Jurka

20 years agoAdd comment to postmaster.c that get_progname() will call exit if it
Bruce Momjian [Tue, 2 Nov 2004 03:34:50 +0000 (03:34 +0000)]
Add comment to postmaster.c that get_progname() will call exit if it
can't strdup().

20 years agoAdd comment about memory failure of get_progname exiting the postmaster:
Bruce Momjian [Tue, 2 Nov 2004 03:09:06 +0000 (03:09 +0000)]
Add comment about memory failure of get_progname exiting the postmaster:

            exit(1);    /* This could exit the postmaster */

20 years agotimestamptz_trunc() should only recalculate the timezone when truncating
Tom Lane [Mon, 1 Nov 2004 22:00:30 +0000 (22:00 +0000)]
timestamptz_trunc() should only recalculate the timezone when truncating
to DAY precision or coarser; leave the timezone alone when precision is
HOUR or less.  This avoids surprises for inputs near a DST transition
time, as per example from Matthew Gabeler-Lee.  (The only reason we
recalculate at all is so that outputs that are supposed to represent
days will come out as local midnight, and that's not relevant for sub-day
precision.)

20 years agoInvent pg_next_dst_boundary() and rewrite DetermineLocalTimeZone() to
Tom Lane [Mon, 1 Nov 2004 21:34:44 +0000 (21:34 +0000)]
Invent pg_next_dst_boundary() and rewrite DetermineLocalTimeZone() to
use it, as per my proposal of yesterday.  This gives us a means of
determining the zone offset to impute to an unlabeled timestamp that
is both efficient and reliable, unlike all our previous tries involving
mktime() and localtime().  The behavior for invalid or ambiguous times
at a DST transition is fixed to be really and truly "assume standard
time", fixing a bug that has come and gone repeatedly but was back
again in 7.4.  (There is some ongoing discussion about whether we should
raise an error instead, but for the moment I'll make it do what it was
previously intended to do.)

20 years agoThat's just the lovely way windows handles a "segfault". Gotta be really
Bruce Momjian [Mon, 1 Nov 2004 19:21:50 +0000 (19:21 +0000)]
That's just the lovely way windows handles a "segfault". Gotta be really
interesting for MS to catch all those dumps...

Anyway. Oops. Seems I ran my regression tests with the old psql, and
just managed to update the backend, when I tested that patch. Turns out
there are codepaths where we'd access the Critical Section before it was
initialized. Attached patch breaks the initializeation off to a separate
part and adds that one to a much earlier position in the program.

Magnus Hagander

20 years agoUpdate comment to point to proper file.
Bruce Momjian [Mon, 1 Nov 2004 14:33:10 +0000 (14:33 +0000)]
Update comment to point to proper file.

20 years agofixed bug in segfault REVOKE statement
Dave Cramer [Mon, 1 Nov 2004 13:17:12 +0000 (13:17 +0000)]
fixed bug in segfault REVOKE statement

20 years agoTrivial fixes for English grammar in contrib/btree_gist and contrib/rtree_gist
Neil Conway [Mon, 1 Nov 2004 11:11:13 +0000 (11:11 +0000)]
Trivial fixes for English grammar in contrib/btree_gist and contrib/rtree_gist
documentation.

20 years agoAdjust use of sizeof(EXE).
Bruce Momjian [Mon, 1 Nov 2004 04:46:56 +0000 (04:46 +0000)]
Adjust use of sizeof(EXE).

20 years agoClarify coding of .exe patch
Bruce Momjian [Mon, 1 Nov 2004 04:25:18 +0000 (04:25 +0000)]
Clarify coding of .exe patch

20 years agoUpdate:
Bruce Momjian [Mon, 1 Nov 2004 02:22:05 +0000 (02:22 +0000)]
Update:

< * Add a GUC variable to control temporary and TOAST tablespace usage
> * Add a GUC variable to control the tablespace for temporary objects

20 years agoAdd:
Bruce Momjian [Mon, 1 Nov 2004 00:28:40 +0000 (00:28 +0000)]
Add:

> * Add a GUC variable to control temporary and TOAST tablespace usage

20 years agoUse psql -q option instead of 2>/dev/null during first invocation
Tom Lane [Sun, 31 Oct 2004 19:14:16 +0000 (19:14 +0000)]
Use psql -q option instead of 2>/dev/null during first invocation
of psql; this should make it easier to diagnose client-side problems,
such as library version mismatch.  Also, consistently use -X option
to avoid problems from weird .psqlrc settings.

20 years agoInvent a new, more thread-safe version of PQrequestCancel, called PQcancel.
Tom Lane [Sat, 30 Oct 2004 23:11:27 +0000 (23:11 +0000)]
Invent a new, more thread-safe version of PQrequestCancel, called PQcancel.
Use this new function in psql.  Implement query cancellation in psql for
Windows.  Code by Magnus Hagander, documentation and minor editorialization
by Tom Lane.

20 years agoI found a corner case in which it is possible for RI_FKey_check's call
Tom Lane [Sat, 30 Oct 2004 20:53:06 +0000 (20:53 +0000)]
I found a corner case in which it is possible for RI_FKey_check's call
of HeapTupleSatisfiesItself() to trigger a hint-bit update on the tuple:
if the row was updated or deleted by a subtransaction of my own transaction
that was later rolled back.  This cannot occur in pre-8.0 of course, so
the hint-bit patch applied a couple weeks ago is OK for existing releases.
But for 8.0 it seems we had better fix things so that RI_FKey_check can
pass the correct buffer number to HeapTupleSatisfiesItself.  Accordingly,
add fields to the TriggerData struct to carry the buffer ID(s) for the
old and new tuple(s).  There are other possible solutions but this one
seems cleanest; it will allow other AFTER-trigger functions to safely
do tqual.c calls if they want to.  Put new fields at end of struct so
that there is no API breakage.

20 years agoChange COMMIT back to the old behavior of emitting command tag COMMIT,
Tom Lane [Sat, 30 Oct 2004 20:44:43 +0000 (20:44 +0000)]
Change COMMIT back to the old behavior of emitting command tag COMMIT,
not ROLLBACK, for the case of COMMIT outside a transaction block.
Alvaro Herrera

20 years agoTranslation update
Peter Eisentraut [Sat, 30 Oct 2004 08:23:32 +0000 (08:23 +0000)]
Translation update

20 years agoRearrange order of pre-commit operations: must close cursors before doing
Tom Lane [Fri, 29 Oct 2004 22:19:53 +0000 (22:19 +0000)]
Rearrange order of pre-commit operations: must close cursors before doing
ON COMMIT actions.  Per bug report from Michael Guerin.

20 years agoFix broken example for PREPARE.
Tom Lane [Fri, 29 Oct 2004 19:40:33 +0000 (19:40 +0000)]
Fix broken example for PREPARE.

20 years agoNo need to set errorMessage length --- already set in goto target.
Bruce Momjian [Fri, 29 Oct 2004 19:30:02 +0000 (19:30 +0000)]
No need to set errorMessage length --- already set in goto target.

20 years agoFix failure to think clearly about encoding conversion errors in COPY.
Tom Lane [Fri, 29 Oct 2004 19:18:22 +0000 (19:18 +0000)]
Fix failure to think clearly about encoding conversion errors in COPY.
We can't regurgitate the unconverted string as I first thought, because
the elog.c mechanisms will assume the error message data is in the server
encoding and attempt a reverse conversion.  Eventually it might be worth
providing a short-circuit path to support this, but for now the simplest
solution is to abandon trying to report back the line contents after a
conversion failure.  Per bug report from Sil Lee, 27-Oct-2004.

20 years agoSet errorMessage length in PQrequestCancel() in all places.
Bruce Momjian [Fri, 29 Oct 2004 17:53:13 +0000 (17:53 +0000)]
Set errorMessage length in PQrequestCancel() in all places.

20 years agoImprovements to the CREATE DATABASE reference page's description of the
Neil Conway [Fri, 29 Oct 2004 03:17:22 +0000 (03:17 +0000)]
Improvements to the CREATE DATABASE reference page's description of the
TABLESPACE clause.

20 years agoMinor improvements to the tablespace documentation.
Neil Conway [Fri, 29 Oct 2004 02:11:18 +0000 (02:11 +0000)]
Minor improvements to the tablespace documentation.

20 years agoAdd DEBUG1-level logging of checkpoint start and end. Also, reduce the
Tom Lane [Fri, 29 Oct 2004 00:16:08 +0000 (00:16 +0000)]
Add DEBUG1-level logging of checkpoint start and end.  Also, reduce the
'recycled log files' and 'removed log files' messages from DEBUG1 to
DEBUG2, replacing them with a count of files added/removed/recycled in
the checkpoint end message, as per suggestion from Simon Riggs.

20 years agoThis makes dblink pass its installcheck test on platforms where
Tom Lane [Thu, 28 Oct 2004 22:31:11 +0000 (22:31 +0000)]
This makes dblink pass its installcheck test on platforms where
snprintf(data, len, %s, NULL) crash.

The code was trying to find a connection by name when it already had an
unnamed connection and did not have a name to search with.

Kris Jurka

20 years agoCode cleanup in dirmod.c. Andrew Dunstan, some further mods by moi.
Tom Lane [Thu, 28 Oct 2004 22:09:31 +0000 (22:09 +0000)]
Code cleanup in dirmod.c.  Andrew Dunstan, some further mods by moi.

20 years agoFix to_number for the case of a trailing S.
Tom Lane [Thu, 28 Oct 2004 18:55:08 +0000 (18:55 +0000)]
Fix to_number for the case of a trailing S.
Karel Zak

20 years agoTranslation update
Peter Eisentraut [Thu, 28 Oct 2004 09:38:27 +0000 (09:38 +0000)]
Translation update

20 years agoTranslation updates
Peter Eisentraut [Thu, 28 Oct 2004 09:01:06 +0000 (09:01 +0000)]
Translation updates

20 years agoTranslation updates
Peter Eisentraut [Thu, 28 Oct 2004 08:54:09 +0000 (08:54 +0000)]
Translation updates

20 years agoUse AllocateFile(), FreeFile() and palloc() rather than fopen(), fclose()
Neil Conway [Thu, 28 Oct 2004 01:38:41 +0000 (01:38 +0000)]
Use AllocateFile(), FreeFile() and palloc() rather than fopen(), fclose()
and malloc() in pgstat.c, respectively. This simplifies error recovery,
as well as being more consistent with the rest of the backend.

20 years agoOn Windows, force a checkpoint just before dropping a database's physical
Tom Lane [Thu, 28 Oct 2004 00:39:59 +0000 (00:39 +0000)]
On Windows, force a checkpoint just before dropping a database's physical
files and directories.  This ensures that the bgwriter will close any open
file references it is holding for files therein, which is needed for the
rmdir() to succeed.  Andrew Dunstan and Tom Lane.

20 years agoHave pg_ctl status always output to stdout, even stopped status, not stderr.
Bruce Momjian [Wed, 27 Oct 2004 19:44:14 +0000 (19:44 +0000)]
Have pg_ctl status always output to stdout, even stopped status, not stderr.

20 years agoFix Cygwin compile for timezone.
Bruce Momjian [Wed, 27 Oct 2004 19:14:07 +0000 (19:14 +0000)]
Fix Cygwin compile for timezone.

20 years agoWhen displaying a Var that is a reference to a column of an unnamed join,
Tom Lane [Wed, 27 Oct 2004 18:09:41 +0000 (18:09 +0000)]
When displaying a Var that is a reference to a column of an unnamed join,
try to display it as a reference to the underlying column instead.  This
is a legitimate substitution (it wouldn't be for a named join) and it
fixes some cases where the display would otherwise be ambiguous.  Per
example from Sim Zacks.

20 years agoCanonicalize Win32 path coming in from pg_ctl -D, idea from Magnus.
Bruce Momjian [Wed, 27 Oct 2004 17:17:09 +0000 (17:17 +0000)]
Canonicalize Win32 path coming in from pg_ctl -D, idea from Magnus.