]> granicus.if.org Git - postgresql/log
postgresql
20 years agoAdd ALTER INDEX, particularly for moving tablespaces.
Bruce Momjian [Fri, 20 Aug 2004 04:29:33 +0000 (04:29 +0000)]
Add ALTER INDEX, particularly for moving tablespaces.

Gavin Sherry

20 years ago> Please find attached a submission to add a "exit on error" option to
Bruce Momjian [Fri, 20 Aug 2004 04:20:23 +0000 (04:20 +0000)]
> Please find attached a submission to add a "exit on error" option to
> pg_restore, as it seems that some people have scripts that rely on the
> previous "abort on error" default behavior when restoring data with a
> direct connection.
>
> Fabien Coelho

20 years ago> After all that about numbering centuries and millenia correctly,
Bruce Momjian [Fri, 20 Aug 2004 03:45:14 +0000 (03:45 +0000)]
> After all that about numbering centuries and millenia correctly,
> why does CVS tip still give me
>
> regression=# select extract(century from now());
>  date_part
> -----------
>         20
> (1 row)
> [ ... looks in code ... ]
>
> Apparently it's because you fixed only timestamp_part, and not
> timestamptz_part.  I'm not too sure about what timestamp_trunc or
> timestamptz_trunc should do, but they may be wrong as well.

Sigh... as usual, what is not tested does not work:-(

> Could we have a more complete patch?

Please find a submission attached. I hope it really fixes all decade,
century and millenium issues for extract and *_trunc functions on
interval
and other timestamp types. If someone could check that the results
are reasonnable, it would be great.

I indeed overlooked the fact that there were two functions. The patch
fixes the code so that both variants agree.

I added comments to interval extractions, because it relies on the C
division to have a negative remainder: -7/10 = 0 and remains -7.

As for *_trunc functions, I have chosen to put the first year of the
century or millennium: -100, 1, 101... 1001 2001 etc. Indeed, I don't
think it would make sense to put 2000 (last year of the 2nd millennium)
for rounding all years of the third millenium.

I also fixed the code so that all decades last 10 years and decade 199
means the 1990's.

I have added some tests that are relevant to deal with tricky cases. The
formula may be simplified, but all these cases must pass. Please keep
them.

Fabien Coelho

20 years agoAdd:
Bruce Momjian [Fri, 20 Aug 2004 03:06:00 +0000 (03:06 +0000)]
Add:

> * Add a function that returns the 'uptime' of the postmaster

20 years agoUpdate wording on commented postgresql.conf values:
Bruce Momjian [Thu, 19 Aug 2004 22:25:09 +0000 (22:25 +0000)]
Update wording on commented postgresql.conf values:

>   Another option is to allow commented values to return to their
>   default values.

20 years agoRepair some issues with column aliases and RowExpr construction in the
Tom Lane [Thu, 19 Aug 2004 20:57:41 +0000 (20:57 +0000)]
Repair some issues with column aliases and RowExpr construction in the
presence of dropped columns.  Document the already-presumed fact that
eref aliases in relation RTEs are supposed to have entries for dropped
columns; cause the user alias structs to have such entries too, so that
there's always a one-to-one mapping to the underlying physical attnums.
Adjust expandRTE() and related code to handle the case where a column
that is part of a JOIN has been dropped.  Generalize expandRTE()'s API
so that it can be used in a couple of places that formerly rolled their
own implementation of the same logic.  Fix ruleutils.c to suppress
display of aliases for columns that were dropped since the rule was made.

20 years agoUse backslash in %USERPROFILE%\.pgpass.
Bruce Momjian [Thu, 19 Aug 2004 16:39:13 +0000 (16:39 +0000)]
Use backslash in %USERPROFILE%\.pgpass.

20 years agoAdd:
Bruce Momjian [Thu, 19 Aug 2004 02:36:24 +0000 (02:36 +0000)]
Add:

>  o Allow COPY to optionally include column headings as the first line

20 years agoDon't use get_home_path so libpq doesn't pull path.c into the library.
Bruce Momjian [Thu, 19 Aug 2004 00:00:34 +0000 (00:00 +0000)]
Don't use get_home_path so libpq doesn't pull path.c into the library.

20 years agoClean up some random departures from project's standard declaration style.
Tom Lane [Wed, 18 Aug 2004 19:27:13 +0000 (19:27 +0000)]
Clean up some random departures from project's standard declaration style.

20 years agoRemove NT4 mention in release notes.
Bruce Momjian [Wed, 18 Aug 2004 15:48:36 +0000 (15:48 +0000)]
Remove NT4 mention in release notes.

20 years agoRemove as duplicate:
Bruce Momjian [Wed, 18 Aug 2004 04:00:32 +0000 (04:00 +0000)]
Remove as duplicate:

<  o Allow pg_dumpall to use non-text output formats

20 years agoAdd:
Bruce Momjian [Wed, 18 Aug 2004 03:51:35 +0000 (03:51 +0000)]
Add:

> * Move some /contrib modules out to their own project sites

20 years agoThe enclose patch clarifies and makes a more useful example for the
Bruce Momjian [Wed, 18 Aug 2004 03:37:56 +0000 (03:37 +0000)]
The enclose patch clarifies and makes a more useful example for the
Global Values in PL/Perl section of the documents.

David Fetter

20 years agoAdd pg_dump section:
Bruce Momjian [Wed, 18 Aug 2004 03:19:42 +0000 (03:19 +0000)]
Add pg_dump section:

< * -Allow pg_dump to dump CREATE CONVERSION (Christopher)
< * -Make pg_restore continue after errors, so it acts more like pg_dump scripts
485,486d482
< * Allow pg_dumpall to use non-text output formats
< * Have pg_dump use multi-statement transactions for INSERT dumps
493,496d488
< * Allow pg_dump to use multiple -t and -n switches
<
<   This should be done by allowing a '-t schema.table' syntax.
<
498a491,512
>
> * pg_dump
>  o Allow pg_dumpall to use non-text output formats
>  o Have pg_dump use multi-statement transactions for INSERT dumps
>  o -Allow pg_dump to dump CREATE CONVERSION (Christopher)
>  o -Make pg_restore continue after errors, so it acts more like pg_dump
>    scripts
>  o Allow pg_dump to use multiple -t and -n switches
>
>    This should be done by allowing a '-t schema.table' syntax.
>
>  o Add dumping of comments on composite type columns
>  o Add dumping of comments on index columns
>  o Replace crude DELETE FROM method of pg_dumpall for cleaning of
>    users and groups with separate DROP commands
>  o Add dumping and restoring of LOB comments
>  o Stop dumping CASCADE on DROP TYPE commands in clean mode
>  o Add full object name to the tag field.  eg. for operators we need
>    '=(integer, integer)', instead of just '='.
>  o Add pg_dumpall custom format dumps. This is probably best done by
>    combining pg_dump and pg_dumpall into a single binary
>  o Add CSV output format

20 years agoReplace incorrect example for quote_ident
Bruce Momjian [Wed, 18 Aug 2004 03:06:56 +0000 (03:06 +0000)]
Replace incorrect example for quote_ident

Greg Sabino Mullan

20 years agoUpdate release history for 7.4.5.
Tom Lane [Wed, 18 Aug 2004 03:03:32 +0000 (03:03 +0000)]
Update release history for 7.4.5.

20 years agoAdd get_home_path() to use USERPROFILE on Win32 and HOME on Unix.
Bruce Momjian [Wed, 18 Aug 2004 02:59:12 +0000 (02:59 +0000)]
Add get_home_path() to use USERPROFILE on Win32 and HOME on Unix.

20 years agoFix bug introduced into _bt_getstackbuf() on 2003-Feb-21: the initial
Tom Lane [Tue, 17 Aug 2004 23:15:33 +0000 (23:15 +0000)]
Fix bug introduced into _bt_getstackbuf() on 2003-Feb-21: the initial
value of 'start' could be past the end of the page, if the page was
split by some concurrent inserting process since we visited it.  In
this situation the code could look at bogus entries and possibly find
a match (since after all those entries still contain what they had
before the split).  This would lead to 'specified item offset is too large'
followed by 'PANIC: failed to add item to the page', as reported by Joe
Conway for scenarios involving heavy concurrent insertion activity.

20 years agoStandardize on the assumption that the arguments of a RowExpr correspond
Tom Lane [Tue, 17 Aug 2004 18:47:09 +0000 (18:47 +0000)]
Standardize on the assumption that the arguments of a RowExpr correspond
to the physical layout of the rowtype, ie, there are dummy arguments
corresponding to any dropped columns in the rowtype.  We formerly had a
couple of places that did it this way and several others that did not.
Fixes Gaetano Mendola's "cache lookup failed for type 0" bug of 5-Aug.

20 years agoAdd:
Bruce Momjian [Tue, 17 Aug 2004 17:14:56 +0000 (17:14 +0000)]
Add:

< * -Allow savepoints / nested transactions [transactions] (Alvaro)
> * -Allow savepoints / nested transactions (Alvaro)
348a349,353
> * Add an option to automatically use savepoints for each statement in a
>   multi-statement transaction.
>
>   When enabled, this would allow errors in multi-statement transactions
>   to be automatically ignored.

20 years agoRemove transactions TODO.detail.
Bruce Momjian [Tue, 17 Aug 2004 17:11:44 +0000 (17:11 +0000)]
Remove transactions TODO.detail.

20 years agoChain on to SIGPIPE handler rather than just do action on default.
Bruce Momjian [Tue, 17 Aug 2004 16:54:47 +0000 (16:54 +0000)]
Chain on to SIGPIPE handler rather than just do action on default.

Always create thread-specific variable.

20 years agoFix agressive collection of thread flags.
Bruce Momjian [Tue, 17 Aug 2004 15:19:09 +0000 (15:19 +0000)]
Fix agressive collection of thread flags.

20 years agoAdd comment about portability function name inconsistency.
Bruce Momjian [Tue, 17 Aug 2004 14:38:38 +0000 (14:38 +0000)]
Add comment about portability function name inconsistency.

20 years agoFix linking problem when enabling thread safety on Darwin: uninitialized
Tom Lane [Tue, 17 Aug 2004 04:24:23 +0000 (04:24 +0000)]
Fix linking problem when enabling thread safety on Darwin: uninitialized
global variables are problematic on this platform.  Simplest solution
seems to be to initialize pthread key variable to 0.  Also, rename this
variable and check_sigpipe_handler to something involving "pq" to
avoid gratuitous pollution of application namespace.

20 years agoAdd DLLIMPORT for PostGIS.
Bruce Momjian [Tue, 17 Aug 2004 02:51:32 +0000 (02:51 +0000)]
Add DLLIMPORT for PostGIS.

20 years agoMove io.h include higher in the file so it doesn't conflict with
Bruce Momjian [Tue, 17 Aug 2004 02:44:13 +0000 (02:44 +0000)]
Move io.h include higher in the file so it doesn't conflict with
rename/unlink defined later.  Problem exists on MS VC.

Andrew Francis

20 years agoFix syntax error just introduced.
Bruce Momjian [Mon, 16 Aug 2004 23:49:58 +0000 (23:49 +0000)]
Fix syntax error just introduced.

20 years agoautoconf for recent thread check changes.
Bruce Momjian [Mon, 16 Aug 2004 23:41:44 +0000 (23:41 +0000)]
autoconf for recent thread check changes.

20 years agoAdd trivial NULL statement to plpgsql, for Oracle compatibility.
Tom Lane [Mon, 16 Aug 2004 17:52:06 +0000 (17:52 +0000)]
Add trivial NULL statement to plpgsql, for Oracle compatibility.

20 years agoAdd:
Bruce Momjian [Mon, 16 Aug 2004 16:14:31 +0000 (16:14 +0000)]
Add:

>  o Add ALTER INDEX syntax to work like ALTER TABLE indexname

20 years agoThrow error if initdb -L is not an absolute path.
Bruce Momjian [Mon, 16 Aug 2004 15:44:03 +0000 (15:44 +0000)]
Throw error if initdb -L is not an absolute path.

20 years agoAdd:
Bruce Momjian [Mon, 16 Aug 2004 04:38:46 +0000 (04:38 +0000)]
Add:

> * Set proper permissions on non-system schemas during db creation
>
>   Currently all schemas are owned by the super-user because they are
>   copied from the template1 database.
>

20 years agoFix for adding \n for zero-length win32 read_pipe return
Bruce Momjian [Mon, 16 Aug 2004 02:46:36 +0000 (02:46 +0000)]
Fix for adding \n for zero-length win32 read_pipe return

20 years agoMark server_encoding and integer_datetimes as GUC_REPORT, per previous
Tom Lane [Mon, 16 Aug 2004 02:12:29 +0000 (02:12 +0000)]
Mark server_encoding and integer_datetimes as GUC_REPORT, per previous
proposals by Oliver Jowett.  Update documentation.

20 years agoClarify need for \r\n -> \n translation in version checking code.
Bruce Momjian [Mon, 16 Aug 2004 01:26:31 +0000 (01:26 +0000)]
Clarify need for \r\n -> \n translation in version checking code.

20 years agoDept. of further reflection: I looked around to see if any other callers
Tom Lane [Sun, 15 Aug 2004 23:44:46 +0000 (23:44 +0000)]
Dept. of further reflection: I looked around to see if any other callers
of XLogInsert had the same sort of checkpoint interlock problem as
RecordTransactionCommit, and indeed I found some.  Btree index build
and ALTER TABLE SET TABLESPACE write data outside the friendly confines
of the buffer manager, and therefore they have to take their own
responsibility for checkpoint interlock.  The easiest solution seems to
be to force smgrimmedsync at the end of the index build or table copy,
even when the operation is being WAL-logged.  This is sufficient since
the new index or table will be of interest to no one if we don't get
as far as committing the current transaction.

20 years agoXmin calculations should consider only top transaction IDs, and
Tom Lane [Sun, 15 Aug 2004 17:03:36 +0000 (17:03 +0000)]
Xmin calculations should consider only top transaction IDs, and
therefore starting with GetCurrentTransactionId is wrong.  Fixes
miscomputation of RecentGlobalXmin leading to bizarre behavior
reported by Gavin Sherry.

20 years agoSpecify SA_NOCLDSTOP when enabling SIGCHLD, per suggestion from
Tom Lane [Sun, 15 Aug 2004 05:25:10 +0000 (05:25 +0000)]
Specify SA_NOCLDSTOP when enabling SIGCHLD, per suggestion from
Oliver Jowett.

20 years agoQuote PERL expansion for Win32 path that might have spaces.
Bruce Momjian [Sun, 15 Aug 2004 00:41:51 +0000 (00:41 +0000)]
Quote PERL expansion for Win32 path that might have spaces.

20 years agoFix trivial typo.
Tom Lane [Sun, 15 Aug 2004 00:27:14 +0000 (00:27 +0000)]
Fix trivial typo.

20 years agoMinor copy-editing for 7.4.4, 7.3.7, 7.2.5 release notes.
Tom Lane [Sun, 15 Aug 2004 00:09:24 +0000 (00:09 +0000)]
Minor copy-editing for 7.4.4, 7.3.7, 7.2.5 release notes.

20 years agoFix typo, per Andreas Seltenreich.
Tom Lane [Sat, 14 Aug 2004 23:49:07 +0000 (23:49 +0000)]
Fix typo, per Andreas Seltenreich.

20 years agoAdd discussion and example about predicate locking and why "serializable"
Tom Lane [Sat, 14 Aug 2004 22:18:23 +0000 (22:18 +0000)]
Add discussion and example about predicate locking and why "serializable"
mode isn't really serializable.  I had thought this was covered already
in our docs, but I sure can't find it.

20 years agoMinor editorializing.
Tom Lane [Sat, 14 Aug 2004 22:17:08 +0000 (22:17 +0000)]
Minor editorializing.

20 years agoTranslation updates
Peter Eisentraut [Sat, 14 Aug 2004 19:16:07 +0000 (19:16 +0000)]
Translation updates

20 years agoAdd:
Bruce Momjian [Sat, 14 Aug 2004 01:30:49 +0000 (01:30 +0000)]
Add:

>
> * Allow buffered WAL writes and fsync
>
>   Instead of guaranteeing recovery of all committed transactions, this
>   would provide improved performance by delaying WAL writes and fsync
>   so an abrupt operating system restart might lose a few seconds of
>   committed transactions but still be consistent.  We could perhaps
>   remove the 'fsync' parameter (which results in an an inconsistent
>   database) in favor of this capability.

20 years agoFix psql's COPY support to deal with \r\n line endings.
Tom Lane [Fri, 13 Aug 2004 22:59:29 +0000 (22:59 +0000)]
Fix psql's COPY support to deal with \r\n line endings.
Andrew Dunstan, some further hacking by Tom Lane.

20 years agoFix breakage with PUBLIC schema. Try to untwist the remarkably contorted
Tom Lane [Fri, 13 Aug 2004 21:37:28 +0000 (21:37 +0000)]
Fix breakage with PUBLIC schema.  Try to untwist the remarkably contorted
logic a little bit.

20 years agoGive a more specific error message for "you can't do that" error cases
Tom Lane [Fri, 13 Aug 2004 18:47:56 +0000 (18:47 +0000)]
Give a more specific error message for "you can't do that" error cases
in plpgsql, particularly trying to begin/end/rollback a transaction.

20 years agoPreliminary release notes for 7.4.4, 7.3.7, 7.2.5. Will add to the
Tom Lane [Fri, 13 Aug 2004 18:22:12 +0000 (18:22 +0000)]
Preliminary release notes for 7.4.4, 7.3.7, 7.2.5.  Will add to the
back branches later.

20 years agoNew translations
Peter Eisentraut [Fri, 13 Aug 2004 16:47:30 +0000 (16:47 +0000)]
New translations

20 years agoTranslation update
Peter Eisentraut [Fri, 13 Aug 2004 16:43:06 +0000 (16:43 +0000)]
Translation update

20 years agoTranslation updates
Peter Eisentraut [Fri, 13 Aug 2004 16:34:31 +0000 (16:34 +0000)]
Translation updates

20 years agoMention log_statement_stats can not be enabled with the other options.
Bruce Momjian [Fri, 13 Aug 2004 16:29:55 +0000 (16:29 +0000)]
Mention log_statement_stats can not be enabled with the other options.

20 years agoClarify trigger function return convention, per Thomas Hallgren.
Tom Lane [Fri, 13 Aug 2004 16:17:19 +0000 (16:17 +0000)]
Clarify trigger function return convention, per Thomas Hallgren.

20 years agoFix core dumps, inability to count, etc associated with canonicalize_path
Tom Lane [Fri, 13 Aug 2004 14:47:23 +0000 (14:47 +0000)]
Fix core dumps, inability to count, etc associated with canonicalize_path
patches.

20 years agoChange order of operations in ALTER TABLE SET TABLESPACE so that we
Tom Lane [Fri, 13 Aug 2004 04:50:28 +0000 (04:50 +0000)]
Change order of operations in ALTER TABLE SET TABLESPACE so that we
don't hold an open file reference to the original table at the end.
This is a good thing in any case, particularly so on Windows which
cannot drop the table file otherwise.

20 years agoMention the installer doesn't run on NT4.
Bruce Momjian [Fri, 13 Aug 2004 03:00:07 +0000 (03:00 +0000)]
Mention the installer doesn't run on NT4.

20 years agoAdd:
Bruce Momjian [Fri, 13 Aug 2004 02:47:53 +0000 (02:47 +0000)]
Add:

>  o Add ALTER INDEX that works just like ALTER TABLE already does
>    on an index

20 years agoAllow commas in BEGIN, START TRANSACTION, and SET TRANSACTION, as required
Tom Lane [Thu, 12 Aug 2004 21:00:34 +0000 (21:00 +0000)]
Allow commas in BEGIN, START TRANSACTION, and SET TRANSACTION, as required
by the SQL standard.  For backwards compatibility, however, continue to
accept the syntax without.  Minor editorialization in the reference pages
for these commands, too.

20 years agoAdd:
Bruce Momjian [Thu, 12 Aug 2004 19:45:24 +0000 (19:45 +0000)]
Add:

> * Allow finer control over the caching of prepared query plans
>
>   Currently, queries prepared via the libpq API are planned on first
>   execute using the supplied parameters --- allow SQL PREPARE to do the
>   same.  Also, allow control over replanning prepared queries either
>   manually or automatically when statistics for execute parameters
>   differ dramatically from those used during planning.
>

20 years agodbsize modification to support tablespaces
Bruce Momjian [Thu, 12 Aug 2004 19:18:05 +0000 (19:18 +0000)]
dbsize modification to support tablespaces

Gavin Sherry

20 years agoAllow optional SAVEPOINT keyword in RELEASE and ROLLBACK TO, for greater
Tom Lane [Thu, 12 Aug 2004 19:12:21 +0000 (19:12 +0000)]
Allow optional SAVEPOINT keyword in RELEASE and ROLLBACK TO, for greater
compliance with SQL2003 spec syntax.

Oliver Jowett

20 years agoCleanup Win32 COPY handling, and move archive examples to SGML.
Bruce Momjian [Thu, 12 Aug 2004 19:03:44 +0000 (19:03 +0000)]
Cleanup Win32 COPY handling, and move archive examples to SGML.

20 years agoAdd mention of "WIN32" COPY.
Bruce Momjian [Thu, 12 Aug 2004 18:34:45 +0000 (18:34 +0000)]
Add mention of "WIN32" COPY.

20 years agoAdd make_native_path() because Win32 COPY is an internal CMD.EXE command
Bruce Momjian [Thu, 12 Aug 2004 18:32:52 +0000 (18:32 +0000)]
Add make_native_path() because Win32 COPY is an internal CMD.EXE command
and doesn't process forward slashes in the same way as external
commands.  Quoting the first argument to COPY does not convert forward
to backward slashes, but COPY does properly process quoted forward
slashes in the second argument.

Win32 COPY works with quoted forward slashes in the first argument only if the
current directory is the same as the directory of the first argument.

20 years agoBe more aggressive about adding flags to thread compiles. The configure
Bruce Momjian [Thu, 12 Aug 2004 16:39:50 +0000 (16:39 +0000)]
Be more aggressive about adding flags to thread compiles.  The configure
test only tests for building a binary, not building a shared library.

On Linux, you can build a binary with -pthread, but you can't build a
binary that uses a threaded shared library unless you also use -pthread
when building the binary, or adding -lpthread to the shared library
build.  This patch has the effect of doing the later by adding both
-pthread and -lpthread when building libpq.

20 years agoMinor style cleanup of thread test script.
Bruce Momjian [Thu, 12 Aug 2004 14:58:37 +0000 (14:58 +0000)]
Minor style cleanup of thread test script.

20 years agoCause initdb to actually accept -s as intended, and fix some typos in
Tom Lane [Wed, 11 Aug 2004 23:28:54 +0000 (23:28 +0000)]
Cause initdb to actually accept -s as intended, and fix some typos in
a comment.

Jon Jensen

20 years agoFix broken parsing of CREATE TABLE AS (looks like a mis-sync with main
Tom Lane [Wed, 11 Aug 2004 22:50:35 +0000 (22:50 +0000)]
Fix broken parsing of CREATE TABLE AS (looks like a mis-sync with main
parser).

20 years agoAvoid crashing when restoring a saved GUC session_authorization value
Tom Lane [Wed, 11 Aug 2004 21:10:37 +0000 (21:10 +0000)]
Avoid crashing when restoring a saved GUC session_authorization value
that refers to a now-deleted userid.  Per gripe from Chris Ochs.

20 years agoAdd PQserverVersion() to libpq to provide more-convenient access to
Tom Lane [Wed, 11 Aug 2004 18:06:01 +0000 (18:06 +0000)]
Add PQserverVersion() to libpq to provide more-convenient access to
the server version number.  This commit also removes bogus DOS line
endings from libpqddll.def.

Greg Sabino Mullane

20 years agoWork around broken strtod() that's present in many Solaris releases.
Tom Lane [Wed, 11 Aug 2004 17:20:50 +0000 (17:20 +0000)]
Work around broken strtod() that's present in many Solaris releases.
Thanks to Michael Fuhr for identifying the problem.

20 years agoFix function definition that somehow missed being ANSI-fied, and align
Tom Lane [Wed, 11 Aug 2004 16:53:28 +0000 (16:53 +0000)]
Fix function definition that somehow missed being ANSI-fied, and align
it with previous prototype to suppress complaints from picky compilers,
per report from Scott Bailey.  Also, remove substitute strerror
definition --- not needed, since we link this with libpgport.

20 years agoAllow compilation when CODESET is not defined (OpenBSD).
Peter Eisentraut [Wed, 11 Aug 2004 11:06:23 +0000 (11:06 +0000)]
Allow compilation when CODESET is not defined (OpenBSD).

20 years agoTranslation update
Peter Eisentraut [Wed, 11 Aug 2004 09:27:20 +0000 (09:27 +0000)]
Translation update

20 years agoAdd translation from 7.4 branch to head.
Peter Eisentraut [Wed, 11 Aug 2004 09:00:27 +0000 (09:00 +0000)]
Add translation from 7.4 branch to head.

20 years agoFix failure to guarantee that a checkpoint will write out pg_clog updates
Tom Lane [Wed, 11 Aug 2004 04:07:16 +0000 (04:07 +0000)]
Fix failure to guarantee that a checkpoint will write out pg_clog updates
for transaction commits that occurred just before the checkpoint.  This is
an EXTREMELY serious bug --- kudos to Satoshi Okada for creating a
reproducible test case to prove its existence.

20 years agoHashed crosstab was dying with an SPI_finish error when the source SQL
Joe Conway [Wed, 11 Aug 2004 00:49:35 +0000 (00:49 +0000)]
Hashed crosstab was dying with an SPI_finish error when the source SQL
produced no rows. Now it returns 0 rows instead. Adjusted regression
test for this case.

20 years agoSet DYLD_LIBRARY_PATH so that 'make check' works without prior
Tom Lane [Tue, 10 Aug 2004 22:24:06 +0000 (22:24 +0000)]
Set DYLD_LIBRARY_PATH so that 'make check' works without prior
'make install' on OS X.  Per suggestion from Adam Witney.

20 years agoNew translation
Peter Eisentraut [Tue, 10 Aug 2004 19:10:21 +0000 (19:10 +0000)]
New translation

20 years agoTranslation updates
Peter Eisentraut [Tue, 10 Aug 2004 19:06:21 +0000 (19:06 +0000)]
Translation updates

20 years agoTranslation update
Peter Eisentraut [Tue, 10 Aug 2004 18:59:57 +0000 (18:59 +0000)]
Translation update

20 years agoTranslation updates
Peter Eisentraut [Tue, 10 Aug 2004 18:42:22 +0000 (18:42 +0000)]
Translation updates

20 years agoTranslation update
Peter Eisentraut [Tue, 10 Aug 2004 18:38:19 +0000 (18:38 +0000)]
Translation update

20 years agoUpdate DELETE FROM:
Bruce Momjian [Tue, 10 Aug 2004 17:30:47 +0000 (17:30 +0000)]
Update DELETE FROM:

< * Allow DELETE to handle table aliases for self-joins
> * Allow an alias to be provided for the target table in UPDATE/DELETE
276,279c276,282
<   There is no way to create a table alias for the deleted table for use
<   in the DELETE WHERE clause.  The agreed approach is to allow a USING
<   clause to specify additional tables.  UPDATE already has an optional
<   FROM clause for this purpose.
>   This is not SQL-spec but many DBMSs allow it.
>
> * Allow additional tables to be specified in DELETE for joins
>
>   UPDATE already allows this (UPDATE...FROM) but we need similar
>   functionality in DELETE.  It's been agreed that the keyword should
>   be USING, to avoid anything as confusing as DELETE FROM a FROM b.

20 years agoUpdate DELETE FROM TODO info.
Bruce Momjian [Tue, 10 Aug 2004 16:29:43 +0000 (16:29 +0000)]
Update DELETE FROM TODO info.

20 years agoAdd:
Bruce Momjian [Tue, 10 Aug 2004 01:38:59 +0000 (01:38 +0000)]
Add:

* Add COMMENT ON for all cluster global objects (users, groups,
  databases and tablespaces)

20 years agoUpdate documentation to reflect the fact that we now know exactly what
Tom Lane [Tue, 10 Aug 2004 00:55:08 +0000 (00:55 +0000)]
Update documentation to reflect the fact that we now know exactly what
time zone names we support.

20 years agoRemoved, shared table:
Bruce Momjian [Mon, 9 Aug 2004 22:48:01 +0000 (22:48 +0000)]
Removed, shared table:

< * Add COMMENT for tablespaces

20 years agoAdd:
Bruce Momjian [Mon, 9 Aug 2004 22:39:18 +0000 (22:39 +0000)]
Add:

> * Add COMMENT for tablespaces

20 years agosyslogger.c needs <sys/time.h> on some platforms, per Greg Mullane.
Tom Lane [Mon, 9 Aug 2004 20:28:48 +0000 (20:28 +0000)]
syslogger.c needs <sys/time.h> on some platforms, per Greg Mullane.

20 years agoPath-mangling logic was failing to account for paths containing mentions
Tom Lane [Mon, 9 Aug 2004 20:20:47 +0000 (20:20 +0000)]
Path-mangling logic was failing to account for paths containing mentions
of '.' or '..'.  Extend canonicalize_path() to trim off trailing occurrences
of these things, and use it to fix up paths where needed (which I think is
only after places where we trim the last path component, but maybe some
others will turn up).  Fixes Josh's complaint that './initdb' does not
work.

20 years agoWhen expanding %p in archive_command or restore_command, translate
Tom Lane [Mon, 9 Aug 2004 16:26:06 +0000 (16:26 +0000)]
When expanding %p in archive_command or restore_command, translate
slashes to backslashes #ifdef WIN32.  This is to cope with the fact
that Windows seems exceedingly unfriendly to slashes in shell commands,
as per recent discussion.

20 years agoRemove no-longer-relevant information about trying to force your OS
Tom Lane [Mon, 9 Aug 2004 05:34:39 +0000 (05:34 +0000)]
Remove no-longer-relevant information about trying to force your OS
to support PST8PDT time zone for the regression tests.

20 years agoFix Win32 pg_dumpall check. REL8_0_0BETA1
Bruce Momjian [Mon, 9 Aug 2004 03:12:38 +0000 (03:12 +0000)]
Fix Win32 pg_dumpall check.

20 years agoImprove comment.
Bruce Momjian [Mon, 9 Aug 2004 02:12:51 +0000 (02:12 +0000)]
Improve comment.

20 years agoStamp libpq.rc with 8.0 version.
Bruce Momjian [Mon, 9 Aug 2004 01:55:42 +0000 (01:55 +0000)]
Stamp libpq.rc with 8.0 version.