OSDN Git Service

pg-rex/syncrep.git
19 years agoFlushRelationBuffers was also being a bit cavalier about whether the
Tom Lane [Tue, 31 Aug 2004 16:13:06 +0000 (16:13 +0000)]
FlushRelationBuffers was also being a bit cavalier about whether the
relation is already opened by smgr.

19 years agocopy_relation_data was mistakenly assuming that the source relation
Tom Lane [Tue, 31 Aug 2004 15:56:39 +0000 (15:56 +0000)]
copy_relation_data was mistakenly assuming that the source relation
would always be already open at the smgr level.  Per bug report from
Fabien Coelho.

19 years agoDefine lstat with parameters, rather than just redefining the symbol.
Bruce Momjian [Tue, 31 Aug 2004 11:29:56 +0000 (11:29 +0000)]
Define lstat with parameters, rather than just redefining the symbol.

19 years agoFix typo in lstat() macro exposed by new tablespace code.
Bruce Momjian [Tue, 31 Aug 2004 11:25:33 +0000 (11:25 +0000)]
Fix typo in lstat() macro exposed by new tablespace code.

19 years agoReplace log_filename_prefix with more general log_filename parameter,
Tom Lane [Tue, 31 Aug 2004 04:53:44 +0000 (04:53 +0000)]
Replace log_filename_prefix with more general log_filename parameter,
to allow DBA to choose the form in which log filenames reflect the
current time.  Also allow for truncating instead of appending to
pre-existing files --- this is convenient when the log filename pattern
rewrites the same names cyclically.  Per Ed L.

19 years agotag configure beta2
PostgreSQL Daemon [Tue, 31 Aug 2004 04:08:33 +0000 (04:08 +0000)]
tag configure beta2

19 years agoImprove spinlock selftest to make it able to detect misdeclaration of
Tom Lane [Mon, 30 Aug 2004 23:47:20 +0000 (23:47 +0000)]
Improve spinlock selftest to make it able to detect misdeclaration of
the slock_t datatype (ie, declared type smaller than what the hardware
TAS instruction needs).

19 years agoslock_t must be int not char for MIPS. 7.4 got this right, but the
Tom Lane [Mon, 30 Aug 2004 22:49:07 +0000 (22:49 +0000)]
slock_t must be int not char for MIPS.  7.4 got this right, but the
info was apparently mistranscribed in s_lock code rearrangement.

19 years agoAvoid ambiguity, as per suggestion from Peter Stricker.
Tom Lane [Mon, 30 Aug 2004 21:29:12 +0000 (21:29 +0000)]
Avoid ambiguity, as per suggestion from Peter Stricker.

19 years agoRather than referring to src/tutorial/README, put the build instructions
Tom Lane [Mon, 30 Aug 2004 21:25:27 +0000 (21:25 +0000)]
Rather than referring to src/tutorial/README, put the build instructions
for the tutorial files right here.

19 years agoFix obviously-utterly-untested noTocComments code.
Tom Lane [Mon, 30 Aug 2004 19:44:14 +0000 (19:44 +0000)]
Fix obviously-utterly-untested noTocComments code.

19 years agoEnsure that the remainder of the current pg_clog page is zeroed during
Tom Lane [Mon, 30 Aug 2004 19:00:42 +0000 (19:00 +0000)]
Ensure that the remainder of the current pg_clog page is zeroed during
startup, just to be sure that there's no leftover junk there.

19 years agoFix failure to advance nextXID beyond subtransactions whose XIDs appear
Tom Lane [Mon, 30 Aug 2004 19:00:03 +0000 (19:00 +0000)]
Fix failure to advance nextXID beyond subtransactions whose XIDs appear
only within COMMIT or ABORT records.

19 years agoRemove item because it is on the open item list:
Bruce Momjian [Mon, 30 Aug 2004 16:08:20 +0000 (16:08 +0000)]
Remove item because it is on the open item list:

< * Fix oid2name and dbsize for tablespaces

19 years agoTweak md.c logic to cope with the situation where WAL replay tries to
Tom Lane [Mon, 30 Aug 2004 03:52:43 +0000 (03:52 +0000)]
Tweak md.c logic to cope with the situation where WAL replay tries to
write into a high-numbered segment of a relation that was later deleted.
We need to temporarily recreate missing segment files, instead of
failing.

19 years agoDept. of second thoughts: it'd be a good idea to flush buffers
Tom Lane [Mon, 30 Aug 2004 03:50:24 +0000 (03:50 +0000)]
Dept. of second thoughts: it'd be a good idea to flush buffers
during replay of CREATE DATABASE as well as the first time around.
Else it's possible that the copy operation will copy obsolete blocks.
We are still a long way from guaranteeing anything about using a
recently-written database as a CREATE template, but this seems needed
to ensure the existing behavior holds up during replay.

19 years agoAdd Win32 service capability to pg_autovacuum.
Bruce Momjian [Mon, 30 Aug 2004 03:31:29 +0000 (03:31 +0000)]
Add Win32 service capability to pg_autovacuum.

Dave Page

19 years agoAnother pgindent run with lib typedefs added.
Bruce Momjian [Mon, 30 Aug 2004 02:54:42 +0000 (02:54 +0000)]
Another pgindent run with lib typedefs added.

19 years agoAdd note that contrib/rserv is gone.
Tom Lane [Mon, 30 Aug 2004 00:47:31 +0000 (00:47 +0000)]
Add note that contrib/rserv is gone.

19 years agoAdd explicit note that empty-string array element values now have to
Tom Lane [Mon, 30 Aug 2004 00:33:52 +0000 (00:33 +0000)]
Add explicit note that empty-string array element values now have to
be written with quotes.  Minor copy-editing too.

19 years agoAdd WAL logging for CREATE/DROP DATABASE and CREATE/DROP TABLESPACE.
Tom Lane [Sun, 29 Aug 2004 21:08:48 +0000 (21:08 +0000)]
Add WAL logging for CREATE/DROP DATABASE and CREATE/DROP TABLESPACE.
Fix TablespaceCreateDbspace() to be able to create a dummy directory
in place of a dropped tablespace's symlink.  This eliminates the open
problem of a PANIC during WAL replay when a replayed action attempts
to touch a file in a since-deleted tablespace.  It also makes for a
significant improvement in the usability of PITR replay.

19 years agoUpdate typedefs with /lib info.
Bruce Momjian [Sun, 29 Aug 2004 17:31:42 +0000 (17:31 +0000)]
Update typedefs with /lib info.

19 years agoReplace bcopy by memmove for more portability.
Tom Lane [Sun, 29 Aug 2004 16:43:05 +0000 (16:43 +0000)]
Replace bcopy by memmove for more portability.

19 years agoWiden xl_len field of XLogRecord header to 32 bits, so that we'll have
Tom Lane [Sun, 29 Aug 2004 16:34:48 +0000 (16:34 +0000)]
Widen xl_len field of XLogRecord header to 32 bits, so that we'll have
a more tolerable limit on the number of subtransactions or deleted files
in COMMIT and ABORT records.  Buy back the extra space by eliminating the
xl_xact_prev field, which isn't being used for anything and is rather
unlikely ever to be used for anything.
This does not force initdb, but you do need to do pg_resetxlog if you
want to upgrade an existing 8.0 installation without initdb.

19 years agoPgindent run for 8.0.
Bruce Momjian [Sun, 29 Aug 2004 05:07:03 +0000 (05:07 +0000)]
Pgindent run for 8.0.

19 years agoUpdate with new typedefs. Remove java and c++ parts of readme.
Bruce Momjian [Sun, 29 Aug 2004 04:49:45 +0000 (04:49 +0000)]
Update with new typedefs.  Remove java and c++ parts of readme.

19 years agoUpdate copyright to 2004.
Bruce Momjian [Sun, 29 Aug 2004 04:13:13 +0000 (04:13 +0000)]
Update copyright to 2004.

19 years agoFix for postmaster.c function win32_waitpid(int *exitstatus) call to
Bruce Momjian [Sun, 29 Aug 2004 03:16:30 +0000 (03:16 +0000)]
Fix for postc function win32_waitpid(int *exitstatus) call to
Win32 WaitForMultipleObjects:

ret = WaitForMultipleObjects(win32_numChildren, win32_childHNDArray,
FALSE, 0);

Problem is 'win32_numChildren' could be more then 64 ( function supports
), problem basically arise ( kills postgres ) when you create more then
64 connections and terminate some of them sill leaving more then 64.

Claudio Natoli

19 years agoAdjust regression expected file for new grant language hint.
Bruce Momjian [Sun, 29 Aug 2004 03:14:07 +0000 (03:14 +0000)]
Adjust regression expected file for new grant language hint.

19 years agoAdd hint about using GRANT with non-trusted languages.
Bruce Momjian [Sun, 29 Aug 2004 03:04:15 +0000 (03:04 +0000)]
Add hint about using GRANT with non-trusted languages.

James William Pye

19 years agoSeveral Cygwin fixes pointed out by Reini Urban.
Bruce Momjian [Sun, 29 Aug 2004 02:58:50 +0000 (02:58 +0000)]
Several Cygwin fixes pointed out by Reini Urban.

19 years agoAdd comment on palloc use and DLLIMPORT>
Bruce Momjian [Sun, 29 Aug 2004 01:44:02 +0000 (01:44 +0000)]
Add comment on palloc use and DLLIMPORT>

19 years ago>>> I understand your disliking of non-posix stuff. OTOH,
Bruce Momjian [Sun, 29 Aug 2004 00:38:03 +0000 (00:38 +0000)]
>>> I understand your disliking of non-posix stuff. OTOH,
>>GetLastError will
>>> give much more details than errno.
>>
>>How much more, really?  That mapping table gave me the impression that
>>the win32 error codes aren't all that much more detailed than errno...
>
>The mapping table is not complete. My winerror.h from the SDK
>lists 2209
>error codes, whereas errno.h lists 42...
>
>I still don't think we'll get that much more stuff. Right now,
>the Win32
>code paths that actually use the more advanced functions already write
>out the error number in case something happens. We can keep doing that
>for the other paths (ereport the error *number* when the mapping does
>not have a match). The map to errno will catch almost all cases, I
>think. And in the corner cases we can do with just the number, and use
>"net helpmsg" to get the actual message when checking...

Here's an attempt on this. new file goes in backend/port/win32.

Magnus Hagander

19 years agoThe attached patch improves pg_ctl's win32 service code to eliminate
Bruce Momjian [Sat, 28 Aug 2004 23:26:37 +0000 (23:26 +0000)]
The attached patch improves pg_ctl's win32 service code to eliminate
some possible causes of the stale postmaster.pid problem that some users
have reported.

- The service did not properly report that it accepts
SERVICE_CONTROL_SHUTDOWN events, thus it's possible the SCM simply
killed the postmaster on shutdown.

- 'WaitHints' are now given to the SCM to prevent it timing out if
pg_ctl doesn't respond to a control event quickly enough.

- During shutdown, the service checkpoint counter is incremented every
five seconds for up to a minute to prevent the SCM timing out and
assuming the service is not responding.

Dave Page

19 years ago> Am Dienstag, 17. August 2004 14:26 schrieb Fabien COELHO:
Bruce Momjian [Sat, 28 Aug 2004 22:55:06 +0000 (22:55 +0000)]
> Am Dienstag, 17. August 2004 14:26 schrieb Fabien COELHO:
> > The patch adds missing the "libpgport.a" file to the installation under
> > "install-all-headers". It is needed by some contribs. I install the
> > library in "pkglibdir", but I was wondering whether it should be "libdir"?

Please find attached a small patch against current CVS head that fixes
pgport library installation so that it goes to libdir instead of
pkglibdir. It works for me.

Fabien Coelho

19 years agoFix high-bit comparison compiler warning in pg_dump.
Bruce Momjian [Sat, 28 Aug 2004 22:52:50 +0000 (22:52 +0000)]
Fix high-bit comparison compiler warning in pg_dump.

Philip Warner

19 years agoAdd regression tests for ALTER INDEX.
Bruce Momjian [Sat, 28 Aug 2004 22:06:04 +0000 (22:06 +0000)]
Add regression tests for ALTER INDEX.

Gavin Sherry

19 years agoNow that TransactionIdDidAbort doesn't think it should try to modify
Tom Lane [Sat, 28 Aug 2004 22:04:12 +0000 (22:04 +0000)]
Now that TransactionIdDidAbort doesn't think it should try to modify
pg_clog, there's no reason to do abort marking of subtransactions in a
nonintuitive order.

19 years agoPropogate pg_ctl -D to the postmaster for command-line identification.
Bruce Momjian [Sat, 28 Aug 2004 22:04:01 +0000 (22:04 +0000)]
Propogate pg_ctl -D to the postmaster for command-line identification.

19 years agoAdd missing Assert to make TransactionIdDidAbort more consistent with
Tom Lane [Sat, 28 Aug 2004 21:58:59 +0000 (21:58 +0000)]
Add missing Assert to make TransactionIdDidAbort more consistent with
TransactionIdDidCommit.

19 years agoRemove contrib/rserv, since there are better replication solutions
Tom Lane [Sat, 28 Aug 2004 21:36:25 +0000 (21:36 +0000)]
Remove contrib/rserv, since there are better replication solutions
available elsewhere (including a descendant project on gborg).

19 years agoRearrange order of operations in heap_drop_with_catalog and index_drop
Tom Lane [Sat, 28 Aug 2004 21:05:26 +0000 (21:05 +0000)]
Rearrange order of operations in heap_drop_with_catalog and index_drop
so that we close and flush the doomed relation's relcache entry before
we start to delete the underlying catalog rows, rather than afterwards.
For awhile yesterday I thought that an unexpected relcache entry rebuild
partway through this sequence might explain the infrequent parallel
regression failures we were chasing.  It doesn't, mainly because there's
no CommandCounterIncrement in the sequence and so the deletions aren't
"really" done yet.  But it sure seems like trouble waiting to happen.

19 years ago> > Another issue is that when installing the Postgres service with
Bruce Momjian [Sat, 28 Aug 2004 21:01:38 +0000 (21:01 +0000)]
> > Another issue is that when installing the Postgres service with
> > "pg_ctl register -w ...." the "-w" parameter was not put in
> the registry "ImagePath"
> > value for the Postgres service. (I added it manually to test.) So I
> > suspect that "pg_ctl register" will need to be enhanced to add the
> > "-w" parameter to the registry settings.

Dave Page

19 years agoUse dynamic buffer for token buffer in win32 admin check
Bruce Momjian [Sat, 28 Aug 2004 21:00:35 +0000 (21:00 +0000)]
Use dynamic buffer for token buffer in win32 admin check

Magnus Hagander

19 years agoFix relcache to account properly for subtransaction status of 'new'
Tom Lane [Sat, 28 Aug 2004 20:31:44 +0000 (20:31 +0000)]
Fix relcache to account properly for subtransaction status of 'new'
relcache entries.  Also, change TransactionIdIsCurrentTransactionId()
so that if consulted during transaction abort, it will not say that
the aborted xact is still current.  (It would be better to ensure that
it's never called at all during abort, but I'm not sure we can easily
guarantee that.)  In combination, these fix a crash we have seen
occasionally during parallel regression tests of 8.0.

19 years agoFurther tightening of the array literal parser. Prevent junk
Joe Conway [Sat, 28 Aug 2004 19:31:29 +0000 (19:31 +0000)]
Further tightening of the array literal parser. Prevent junk
from being accepted after the outer right brace. Per report from
Markus Bertheau.

Also add regression test cases for this change, and for previous
recent array literal parser changes.

19 years agoCan't truncate pg_subtrans during a recovery checkpoint --- subtrans
Tom Lane [Sat, 28 Aug 2004 18:18:03 +0000 (18:18 +0000)]
Can't truncate pg_subtrans during a recovery checkpoint --- subtrans
module isn't fully initialized yet.

19 years agoAdd missing semicolon; some bison versions warn of this.
Tom Lane [Sat, 28 Aug 2004 18:04:51 +0000 (18:04 +0000)]
Add missing semicolon; some bison versions warn of this.

19 years agoFix Windows emulation of kill(pid, 0). This will now succeed, but only
Tom Lane [Fri, 27 Aug 2004 18:31:48 +0000 (18:31 +0000)]
Fix Windows emulation of kill(pid, 0).  This will now succeed, but only
if the target PID is a PG postmaster or backend --- for our purposes that
is actually better than the Unix behavior.  Per Dave Page and Andrew Dunstan.

19 years agoIntroduce local hash table for lock state, as per recent proposal.
Tom Lane [Fri, 27 Aug 2004 17:07:42 +0000 (17:07 +0000)]
Introduce local hash table for lock state, as per recent proposal.
PROCLOCK structs in shared memory now have only a bitmask for held
locks, rather than counts (making them 40 bytes smaller, which is a
good thing).  Multiple locks within a transaction are counted in the
local hash table instead, and we have provision for tracking which
ResourceOwner each count belongs to.  Solves recently reported problem
with memory leakage within long transactions.

19 years agoUpdate wording:
Bruce Momjian [Fri, 27 Aug 2004 04:34:28 +0000 (04:34 +0000)]
Update wording:

<  This would require some background daemon to maintain clustering
>  This might require some background daemon to maintain clustering
397,398c397,398
<  paritally filled for easier reorganization.  It also might require
<  creating a merged heap/index data file so an index lookup would
>  paritally filled for easier reorganization.  Another idea would
>         be to create a merged heap/index data file so an index lookup would

19 years agoUpdate win32 comments.
Bruce Momjian [Fri, 27 Aug 2004 02:09:18 +0000 (02:09 +0000)]
Update win32 comments.

19 years agoUpdate:
Bruce Momjian [Fri, 27 Aug 2004 01:44:49 +0000 (01:44 +0000)]
Update:

<  This would require some background daemon to restore clustering
>  This would require some background daemon to maintain clustering
397c397,399
<  paritally filled for easier reorganization.
>  paritally filled for easier reorganization.  It also might require
>  creating a merged heap/index data file so an index lookup would
>  automatically access the heap data too.

19 years agoFix user locks. Broken some time ago for all platforms by Windows-related
Tom Lane [Thu, 26 Aug 2004 17:23:30 +0000 (17:23 +0000)]
Fix user locks.  Broken some time ago for all platforms by Windows-related
changes.

19 years agoImprove some comments.
Tom Lane [Thu, 26 Aug 2004 17:22:28 +0000 (17:22 +0000)]
Improve some comments.

19 years agoNew Win32 wording that mentions "code".
Bruce Momjian [Thu, 26 Aug 2004 16:54:36 +0000 (16:54 +0000)]
New Win32 wording that mentions "code".

19 years agoUpdate documentation to prefer CIDR format for pg_hba.conf and use new
Bruce Momjian [Thu, 26 Aug 2004 16:50:05 +0000 (16:50 +0000)]
Update documentation to prefer CIDR format for pg_hba.conf and use new
CIDR column in examples first.

19 years agoUpdate Win32 wording.
Bruce Momjian [Thu, 26 Aug 2004 13:44:38 +0000 (13:44 +0000)]
Update Win32 wording.

19 years agoAdd warning about Win32 bugs in this release.
Bruce Momjian [Thu, 26 Aug 2004 13:08:59 +0000 (13:08 +0000)]
Add warning about Win32 bugs in this release.

19 years agoUpdate Japanese FAQ.
Bruce Momjian [Thu, 26 Aug 2004 04:09:59 +0000 (04:09 +0000)]
Update Japanese FAQ.

Jun Kuwamura

19 years agoUpdate tcp connection faq.
Bruce Momjian [Thu, 26 Aug 2004 03:18:25 +0000 (03:18 +0000)]
Update tcp connection faq.

19 years agoAdd:
Bruce Momjian [Thu, 26 Aug 2004 03:09:13 +0000 (03:09 +0000)]
Add:

> * Merge hardwired timezone names with the TZ database; allow either kind
>   everywhere a TZ name is currently taken
> * Allow customization of the known set of TZ names (generalize the
>   present australian_timezones hack)

19 years agoFix typo in comment, per Andrew Dunstan.
Tom Lane [Wed, 25 Aug 2004 20:07:57 +0000 (20:07 +0000)]
Fix typo in comment, per Andrew Dunstan.

19 years agoRevise ResourceOwner code to avoid accumulating ResourceOwner objects
Tom Lane [Wed, 25 Aug 2004 18:43:43 +0000 (18:43 +0000)]
Revise ResourceOwner code to avoid accumulating ResourceOwner objects
for every command executed within a transaction.  For long transactions
this was a significant memory leak.  Instead, we can delete a portal's
or subtransaction's ResourceOwner immediately, if we physically transfer
the information about its locks up to the parent owner.  This does not
fully solve the leak problem; we need to do something about counting
multiple acquisitions of the same lock in order to fix it.  But it's a
necessary step along the way.

19 years agoTranslation update
Peter Eisentraut [Tue, 24 Aug 2004 21:45:40 +0000 (21:45 +0000)]
Translation update

19 years agoTranslation update
Peter Eisentraut [Tue, 24 Aug 2004 21:35:19 +0000 (21:35 +0000)]
Translation update

19 years agoAllow second and subsequent names in a qualified (dotted) name to be
Tom Lane [Tue, 24 Aug 2004 20:41:40 +0000 (20:41 +0000)]
Allow second and subsequent names in a qualified (dotted) name to be
ColLabel instead of just ColId --- that is, any keyword can appear after
a dot and it will be taken as an identifier.  Fixes problems with names
that are okay as standalone function names but fail when qualified.

19 years agoUpdate item:
Bruce Momjian [Tue, 24 Aug 2004 11:09:44 +0000 (11:09 +0000)]
Update item:

< * Implement dirty reads or shared row locks and use them in RI triggers (?)
> * Implement dirty reads or shared row locks and use them in RI triggers
>
>   Adding shared locks requires recording the table/rows numbers in a
>   shared area, and this could potentially be a large amount of data.
>   One idea is to store the table/row numbers in a separate table and set
>   a bit on the row indicating looking in this new table is required to
>   find any shared row locks.
>

19 years agoVarious minor improvements to the SGML documentation: fix some incorrect
Neil Conway [Tue, 24 Aug 2004 00:06:51 +0000 (00:06 +0000)]
Various minor improvements to the SGML documentation: fix some incorrect
SGML markup, add a "deprecated features" section to the 8.0 release
notes, untabify release.sgml and runtime.sgml, and make some other
minor improvements.

19 years agoRearrange pg_subtrans handling as per recent discussion. pg_subtrans
Tom Lane [Mon, 23 Aug 2004 23:22:45 +0000 (23:22 +0000)]
Rearrange pg_subtrans handling as per recent discussion.  pg_subtrans
updates are no longer WAL-logged nor even fsync'd; we do not need to,
since after a crash no old pg_subtrans data is needed again.  We truncate
pg_subtrans to RecentGlobalXmin at each checkpoint.  slru.c's API is
refactored a little bit to separate out the necessary decisions.

19 years agoMention removal of page images in PITR files:
Bruce Momjian [Mon, 23 Aug 2004 21:56:50 +0000 (21:56 +0000)]
Mention removal of page images in PITR files:

<   partial page writes during recovery.
>   partial page writes during recovery.  These pages can also be
>   eliminated from point-in-time archive files.

19 years agoUpdate that 8.0 will support MS Win natively.
Bruce Momjian [Mon, 23 Aug 2004 21:51:53 +0000 (21:51 +0000)]
Update that 8.0 will support MS Win natively.

19 years agoUpdate newest version number.
Bruce Momjian [Mon, 23 Aug 2004 16:16:43 +0000 (16:16 +0000)]
Update newest version number.

19 years agoTweak code so that pg_subtrans is never consulted for XIDs older than
Tom Lane [Sun, 22 Aug 2004 02:41:58 +0000 (02:41 +0000)]
Tweak code so that pg_subtrans is never consulted for XIDs older than
RecentXmin (== MyProc->xmin).  This ensures that it will be safe to
truncate pg_subtrans at RecentGlobalXmin, which should largely eliminate
any fear of bloat.  Along the way, eliminate SubTransXidsHaveCommonAncestor,
which isn't really needed and could not give a trustworthy result anyway
under the lookback restriction.
In an unrelated but nearby change, #ifdef out GetUndoRecPtr, which has
been dead code since 2001 and seems unlikely to ever be resurrected.

19 years agoCode review for ALTER INDEX patch.
Tom Lane [Sun, 22 Aug 2004 00:08:28 +0000 (00:08 +0000)]
Code review for ALTER INDEX patch.

19 years agoFix tab completion for ALTER INDEX.
Bruce Momjian [Sat, 21 Aug 2004 18:45:59 +0000 (18:45 +0000)]
Fix tab completion for ALTER INDEX.

19 years agoSeems it would be nice if the ALTER INDEX ref page were actually
Tom Lane [Sat, 21 Aug 2004 16:16:04 +0000 (16:16 +0000)]
Seems it would be nice if the ALTER INDEX ref page were actually
linked into the docs ...

19 years agoFix typo (extraneous comma).
Tom Lane [Sat, 21 Aug 2004 15:45:46 +0000 (15:45 +0000)]
Fix typo (extraneous comma).

19 years agoMention that to_char(interval) will be removed in 8.1.
Bruce Momjian [Sat, 21 Aug 2004 03:25:34 +0000 (03:25 +0000)]
Mention that to_char(interval) will be removed in 8.1.

19 years agoRemove Solaris bug mention in pg_hba.conf.
Bruce Momjian [Sat, 21 Aug 2004 03:21:57 +0000 (03:21 +0000)]
Remove Solaris bug mention in pg_hba.conf.

19 years agoFix pg_ctl -w to properly wait on server startup.
Bruce Momjian [Sat, 21 Aug 2004 03:12:55 +0000 (03:12 +0000)]
Fix pg_ctl -w to properly wait on server startup.

19 years agoRecent tightening of allowed array-literal syntax broke one test case
Tom Lane [Sat, 21 Aug 2004 00:26:54 +0000 (00:26 +0000)]
Recent tightening of allowed array-literal syntax broke one test case
in contrib/ltree.  Fix.

19 years agoWhack btree_gist code around until it has some small hope of building
Tom Lane [Sat, 21 Aug 2004 00:09:10 +0000 (00:09 +0000)]
Whack btree_gist code around until it has some small hope of building
on non-gcc compilers.

19 years agoRecognize plpgsql EXCEPTION condition names at function compile time
Tom Lane [Fri, 20 Aug 2004 22:00:14 +0000 (22:00 +0000)]
Recognize plpgsql EXCEPTION condition names at function compile time
instead of runtime, for better detection of invalid condition names
(and maybe a little more speed, too).

19 years ago>>At this stage of the game I would just change pg_hba.conf.sample to use
Bruce Momjian [Fri, 20 Aug 2004 20:23:06 +0000 (20:23 +0000)]
>>At this stage of the game I would just change pg_hba.conf.sample to use
>>'127.0.0.1/32' instead of '127.0.0.1 255.255.255.255'.
>>
>>
>
>Yeah, that's probably the path of least resistance.  Note that the
>comments and possibly the SGML docs need to be adjusted to match,
>however, so it's not quite a one-liner.

Andrew Dunstan

19 years agoAllow psql to use 7.4.X database by not referencing tablespaces.
Bruce Momjian [Fri, 20 Aug 2004 20:18:23 +0000 (20:18 +0000)]
Allow psql to use 7.4.X database by not referencing tablespaces.

Greg Sabino Mullan

19 years ago> Please find enclose a submission to fix these problems.
Bruce Momjian [Fri, 20 Aug 2004 20:13:10 +0000 (20:13 +0000)]
> Please find enclose a submission to fix these problems.
>
> The patch adds missing the "libpgport.a" file to the installation under
> "install-all-headers". It is needed by some contribs. I install the
> library in "pkglibdir", but I was wondering whether it should be "libdir"?
> I was wondering also whether it would make sense to have a "libpgport.so"?
>
> It fixes various macros which are used by contrib makefiles, especially
> libpq_*dir and LDFLAGS when used under PGXS. It seems to me that they are
> needed to
>
> It adds the ability to test and use PGXS with contribs, with "make
> USE_PGXS=1". Without the macro, this is exactly as before, there should be
> no difference, esp. wrt the vpath feature that seemed broken by previous
> submission. So it should not harm anybody, and it is useful at least to me.
>
> It fixes some inconsistencies in various contrib makefiles
> (useless override, ":=" instead of "=").

Fabien COELHO

19 years agoAdd ALTER SCHEMA item detail:
Bruce Momjian [Fri, 20 Aug 2004 20:07:12 +0000 (20:07 +0000)]
Add ALTER SCHEMA item detail:

<  o Allow databases, schemas, and indexes to be moved to different
<    tablespaces
>  o Allow databases and schemas to be moved to different tablespaces
>
>  One complexity is whether moving a schema should move all existing
>  schema objects or just define the location for future object creation.
>
382c385
<  o Add ALTER INDEX that works just like ALTER TABLE already does
>  o -Add ALTER INDEX that works just like ALTER TABLE already does
384d386
<  o Add ALTER INDEX syntax to work like ALTER TABLE indexname

19 years agoThey are two different problems; the TOC entry is important for any
Bruce Momjian [Fri, 20 Aug 2004 20:00:34 +0000 (20:00 +0000)]
They are two different problems; the TOC entry is important for any
multiline command  or to rerun the command easily later.

Whereas displaying the failed SQL command is a matter of fixing the
error
messages.

The latter is complicated by failed COPY commands which, with
die-on-errors
off, results in the data being processed as a command, so dumping the
command will dump all of the data.

In the case of long commands, should the whole command be dumped? eg.
(eg.
several pages of function definition).

In the case of the COPY command, I'm not sure what to do. Obviously, it
would be best to avoid sending the data, but the data and command are
combined (from memory). Also, the 'data' may be in the form of INSERT
statements.

Attached patch produces the first 125 chars of the command:

pg_restore: [archiver (db)] Error while PROCESSING TOC:
pg_restore: [archiver (db)] Error from TOC Entry 26; 1255 16449270
FUNCTION
plpgsql_call_handler() pjw
pg_restore: [archiver (db)] could not execute query: ERROR:  function
"plpgsql_call_handler" already exists with same argument types
     Command was: CREATE FUNCTION plpgsql_call_handler() RETURNS
language_handler
     AS '/var/lib/pgsql-8.0b1/lib/plpgsql', 'plpgsql_call_han...
pg_restore: [archiver (db)] Error from TOC Entry 27; 1255 16449271
FUNCTION
plpgsql_validator(oid) pjw
pg_restore: [archiver (db)] could not execute query: ERROR:  function
"plpgsql_validator" already exists with same argument types
     Command was: CREATE FUNCTION plpgsql_validator(oid) RETURNS void
     AS '/var/lib/pgsql-8.0b1/lib/plpgsql', 'plpgsql_validator'
     LANGU...

Philip Warner

19 years agoI notice that contrib/fuzzystrmatch/dmetaphone.c doesn't compile cleanly
Bruce Momjian [Fri, 20 Aug 2004 19:48:14 +0000 (19:48 +0000)]
I notice that contrib/fuzzystrmatch/dmetaphone.c doesn't compile cleanly
as it stands - it mixes declarations in code, C++-style. The attached
patch shifts declarations to the tops of functions and enables this file
to compile cleanly as C.

Richard Poole

19 years agoAdd psql tab completion TODO:
Bruce Momjian [Fri, 20 Aug 2004 19:30:39 +0000 (19:30 +0000)]
Add psql tab completion TODO:

< * -Have psql \dn show only visible temp schemas using current_schemas()
< * -Have psql '\i ~/<tab><tab>' actually load files it displays from home dir
484a483,484
> * -Have psql \dn show only visible temp schemas using current_schemas()
> * -Have psql '\i ~/<tab><tab>' actually load files it displays from home dir
516a517,527
>
> * psql tab completion
>
>  o Provide a list of conversions after ALTER CONVERSION?
>  o Support for ALTER SEQUENCE clauses
>  o Add RENAME TO to ALTER TRIGGER
>  o Support for ALTER USER
>  o Fix ALTER (GROUP|DOMAIN|...) <sth> DROP
>  o Support for ALTER LANGUAGE <sth> RENAME TO
>  o Improve support for COPY
>  o Improve support for ALTER TABLE

19 years agoAttached is the third version of my patch that adds/fixes several things
Bruce Momjian [Fri, 20 Aug 2004 19:24:59 +0000 (19:24 +0000)]
Attached is the third version of my patch that adds/fixes several things
to/in the psql-tabcomplete code. This diff includes the still missing
tab-complete support for TABLESPACE I already sent earlier. New in this
version of the patch is a small adaption of the tab-complete code to
support the adjusted SAVEPOINT-Syntax commited by Tom, as well as
completion of the only half working (and I think only by accident)
tabcomplete-suppport for "BEGIN [ TRANSACTION | WORK ]".

below is a complete list of the things I have changed with this patch:

*) add tablespace support for CREATE/DROP/ALTER and \db
*) sync the list of possible commands following ALTER with the docs (by
adding
AGGREGATE,CONVERSATION,DOMAIN,FUNCTION,LANGUAGE,OPERATOR,SEQUENCE,TABLESPACE
and TYPE)
*) provide a list of valid users after "OWNER TO"
*) tab-complete support for ALTER (AGGREGATE|CONVERSION|FUNCTION)
*) basic tab-complete support for ALTER DOMAIN
*) provide a list of suitable indexes following ALTER TABLE <sth>
CLUSTER ON(?)
*) add "CLUSTER ON" and "SET" to the ALTER TABLE <sth> - tab-complete
list(fixes incorrect/wrong tab-complete with ALTER TABLE <sth> SET
+<TAB> too)
*) provide a list of possible indexes following ALTER TABLE <sth> CLUSTER ON
*) provide list of possible commands(WITHOUT CLUSTER,WITHOUT OIDS,
TABLESPACE) following ALTER TABLE <sth> SET
*) sync "COMMENT ON" with docs by adding "CAST","CONVERSION","FUNCTION"
*) add ABSOLUT to the list of possible commands after FETCH
*) "END" was missing from the sql-commands overview (though it had
completion support!) - i know it's depreciated but we have ABORT and
others still in ...
*) fixes small buglet with ALTER (TRIGGER|CLUSTER) ON autocomplete
(CLUSTER ON +<TAB> would produce CLUSTER ON ON - same for TRIGGER ON)
*) adapt to new SAVEPOINT syntax
*) fix incomplete Support for BEGIN [ TRANSACTION | WORK ]

Stefan Kaltenbrunn

19 years agoThis patch allows pg_restore to recognize $-quotes in SQL queries. It
Bruce Momjian [Fri, 20 Aug 2004 16:07:15 +0000 (16:07 +0000)]
This patch allows pg_restore to recognize $-quotes in SQL queries. It
will treat any unquoted string that starts with a $ and has no preceding
identifier chars as a potential $-quote tag, it then makes sure that the
tag chars are valid. If so, it processes the $-quote.

Philip Warner

19 years agoAdd mention of environment variable conflict with uncommenting config
Bruce Momjian [Fri, 20 Aug 2004 15:20:27 +0000 (15:20 +0000)]
Add mention of environment variable conflict with uncommenting config
file variables:

<   Another option is to allow commented values to return to their
<   default values.
>   This has to address environment variables that are then overridden
>   by config file values.  Another option is to allow commented values
>   to return to their default values.

19 years agoFix typo in new ALTER INDEX manual page.
Bruce Momjian [Fri, 20 Aug 2004 14:56:57 +0000 (14:56 +0000)]
Fix typo in new ALTER INDEX manual page.

19 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

19 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

19 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

19 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

19 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.