]> granicus.if.org Git - postgresql/log
postgresql
20 years agoGUC assign hooks that look at external state in deciding whether a
Tom Lane [Fri, 24 Sep 2004 19:43:03 +0000 (19:43 +0000)]
GUC assign hooks that look at external state in deciding whether a
setting is valid must ignore that state and permit the assignment anyway
when source is PGC_S_OVERRIDE.  Otherwise they may disallow a rollback
at transaction abort, which is The Wrong Thing.  Per example from
Michael Fuhr 12-Sep-04.

20 years agoFix slightly misleading description of PERFORM syntax.
Tom Lane [Fri, 24 Sep 2004 18:54:14 +0000 (18:54 +0000)]
Fix slightly misleading description of PERFORM syntax.

20 years agoCleanup some ancient Ultrix / Alpha code in main() that is intended to
Neil Conway [Fri, 24 Sep 2004 06:29:07 +0000 (06:29 +0000)]
Cleanup some ancient Ultrix / Alpha code in main() that is intended to
modify how unaligned memory accesses are dealt with. Document that this
is really what is going on, and merge the NOFIXADE and NOPRINTADE code
paths.

20 years agoshmdt takes a void* parameter, per Kris Jurka.
Tom Lane [Fri, 24 Sep 2004 05:27:35 +0000 (05:27 +0000)]
shmdt takes a void* parameter, per Kris Jurka.

20 years agoFix gratuitous weirdness in function prototype, per Kris Jurka.
Tom Lane [Fri, 24 Sep 2004 05:16:35 +0000 (05:16 +0000)]
Fix gratuitous weirdness in function prototype, per Kris Jurka.

20 years agoOkay, I'm tired of reading gripes from OS X users ... add a regression
Tom Lane [Fri, 24 Sep 2004 04:58:46 +0000 (04:58 +0000)]
Okay, I'm tired of reading gripes from OS X users ... add a regression
variant file to cover OS X 10.3's bizarre minus-zero behavior.

20 years agoGet rid of /*-inside-comment warning. My fault.
Tom Lane [Fri, 24 Sep 2004 01:48:43 +0000 (01:48 +0000)]
Get rid of /*-inside-comment warning.  My fault.

20 years agoExecProcAppend() wasn't called ExecAppend() because the latter name was
Neil Conway [Fri, 24 Sep 2004 01:36:37 +0000 (01:36 +0000)]
ExecProcAppend() wasn't called ExecAppend() because the latter name was
formerly used in execMain. Since that is no longer the case, this patch
renames ExecProcAppend() to ExecAppend() for the sake of consistency.

20 years agoMinor improvements to the release notes.
Neil Conway [Fri, 24 Sep 2004 01:31:32 +0000 (01:31 +0000)]
Minor improvements to the release notes.

20 years agoFix TAS assembly stuff for Solaris/386. (I'm not in a position to
Tom Lane [Fri, 24 Sep 2004 00:21:32 +0000 (00:21 +0000)]
Fix TAS assembly stuff for Solaris/386.  (I'm not in a position to
actually test this, but it couldn't be broken any worse than it was...)

20 years agoImprove description of Execute message behavior, per Oliver Jowett.
Tom Lane [Thu, 23 Sep 2004 23:35:07 +0000 (23:35 +0000)]
Improve description of Execute message behavior, per Oliver Jowett.

20 years agoFix ALTER TABLE OWNER to adjust the ownership of dependent sequences,
Tom Lane [Thu, 23 Sep 2004 23:20:24 +0000 (23:20 +0000)]
Fix ALTER TABLE OWNER to adjust the ownership of dependent sequences,
not only indexes.  Alvaro Herrera, with some kibitzing by Tom Lane.

20 years agoIf we're going to print unrecognized result codes from SSL_get_error
Tom Lane [Thu, 23 Sep 2004 20:27:50 +0000 (20:27 +0000)]
If we're going to print unrecognized result codes from SSL_get_error
in open_client_SSL, surely we should do it everywhere.  Also make
message formatting conform to style guide.

20 years agoThis patch attempts to outline the supported level of SSL within libpq.
Bruce Momjian [Thu, 23 Sep 2004 13:31:09 +0000 (13:31 +0000)]
This patch attempts to outline the supported level of SSL within libpq.
I haven't mentioned any of
~/.postgresql/{root.crt,postgresql.crt,postresql.key} even though they
are checked for in the code, since they do not appear to be supported. I
base this on discussions in pgsql-hackers.

Dominic Mitchell

20 years agoThis patch logs the error code in the default case, so that the user
Bruce Momjian [Thu, 23 Sep 2004 13:20:45 +0000 (13:20 +0000)]
This patch logs the error code in the default case, so that the user
stands a chance of looking it up.  "Unrecognised error" is always
disheartening.  :-)

Dominic Mitchell

20 years agoThis patch attempts to note the use of the root.crt file in the server.
Bruce Momjian [Thu, 23 Sep 2004 13:16:02 +0000 (13:16 +0000)]
This patch attempts to note the use of the root.crt file in the server.
Given that PostgreSQL will output a message complaining about it's
absence if you're using SSL mode, I feel it's important that it gets a
mention in the documentation at some point.

Dominic Mitchell

20 years agoUpdate CREATE TABLE AS docs to reflect the fact that CREATE TABLE AS was
Neil Conway [Thu, 23 Sep 2004 03:43:57 +0000 (03:43 +0000)]
Update CREATE TABLE AS docs to reflect the fact that CREATE TABLE AS was
added to the SQL:2003 standard.

20 years agoMinor tweak to pg_controldata usage information, per suggestion from
Neil Conway [Thu, 23 Sep 2004 00:47:44 +0000 (00:47 +0000)]
Minor tweak to pg_controldata usage information, per suggestion from
Dennis Björklund. Also, remove some redundant #include directives.

20 years agoUpdate installation docs for PL/Python to note that distutils is now
Neil Conway [Thu, 23 Sep 2004 00:30:13 +0000 (00:30 +0000)]
Update installation docs for PL/Python to note that distutils is now
required. Original patch from James William Pye, editorializing by
Neil Conway.

20 years agoArrange for hash join to skip scanning the outer relation if it detects
Tom Lane [Wed, 22 Sep 2004 19:13:52 +0000 (19:13 +0000)]
Arrange for hash join to skip scanning the outer relation if it detects
that the inner one is completely empty.  Per recent discussion.  Also some
cosmetic cleanups in nearby code.

20 years agoIssue a CHECKPOINT just after creating the regression database. Without
Tom Lane [Wed, 22 Sep 2004 19:11:19 +0000 (19:11 +0000)]
Issue a CHECKPOINT just after creating the regression database.  Without
this, it's hard to debug core-dump test failures, because WAL replay will
enthusiastically remove the core file (along with the rest of the
regression database directory).  Per recent discussion, not to mention
bitter experience.

20 years agoAdjust ExecMakeTableFunctionResult to produce a single all-nulls row
Tom Lane [Wed, 22 Sep 2004 17:41:51 +0000 (17:41 +0000)]
Adjust ExecMakeTableFunctionResult to produce a single all-nulls row
when a function that returns a single tuple (not a setof tuple) returns
NULL.  This seems to be the most consistent behavior.  It would have
taken a bit less code to make it return an empty table (zero rows) but
ISTM a non-SETOF function ought always return exactly one row.  Per
bug report from Ivan-Sun1.

20 years agoMinor cleanup of libpq/LO examples: fix some memory leaks, update a comment
Neil Conway [Wed, 22 Sep 2004 05:12:45 +0000 (05:12 +0000)]
Minor cleanup of libpq/LO examples: fix some memory leaks, update a comment
or two.

20 years agoSome improvements for the tab-completion of psql. This should
Neil Conway [Wed, 22 Sep 2004 04:25:16 +0000 (04:25 +0000)]
Some improvements for the tab-completion of psql. This should
address all of the items in the todo list and adds some new
things as well. Specifically:

* Add support for ALTER SEQUENCE ...
* Add "RENAME TO" for ALTER TRIGGER xx ON yy
* Pick proper table for ALTER TRIGGER xx ON ...
* Support for ALTER USER xxx ...
* Fix ALTER GROUP xxx DROP ...
* Fix ALTER DOMAIN xxx DROP ...
* Remove "OWNER TO" from ALTER DOMAIN xx DROP ...
* Fix ALTER DOMAIN xx SET DEFAULT ..
* Prevent ALTER INDEX xxx SET TABLESPACE from using "TO"
* Support for ALTER LANGUAGE xxx (RENAME TO)
* More support for ALTER TABLE xxx ALTER COLUMN xxx ...
* More support for COPY

Greg Sabino Mullane

20 years agoThis patch from Alvaro Herrera adds transaction ID to the list of
Neil Conway [Wed, 22 Sep 2004 03:55:27 +0000 (03:55 +0000)]
This patch from Alvaro Herrera adds transaction ID to the list of
log_line_prefix escapes. The escape sequence used for this is %x.
%x previously meant "postmaster et al. stop here" -- this has been
renamed to %q.

20 years agoClarify TODO item: we ought to implement updatable views per SQL99, not
Neil Conway [Tue, 21 Sep 2004 05:38:54 +0000 (05:38 +0000)]
Clarify TODO item: we ought to implement updatable views per SQL99, not
SQL92 (the former defines a much more ambitious subset of views that
need to be updatable).

20 years agoRotate on time boundaries that are sensible per local time rather than GMT.
Tom Lane [Tue, 21 Sep 2004 00:21:25 +0000 (00:21 +0000)]
Rotate on time boundaries that are sensible per local time rather than GMT.
Also, avoid truncating the file we just wrote into, which might otherwise
easily happen at DST boundaries.  Ed L. and Tom Lane.

20 years agoChange some of the existing plpgsql regression test cases so that they
Tom Lane [Mon, 20 Sep 2004 23:00:12 +0000 (23:00 +0000)]
Change some of the existing plpgsql regression test cases so that they
exercise dollar quoting and named function parameters.  AFAICS we had
no tests of either feature before.

20 years agoSome editorializing on the docs for the dollar-quoting feature: fix
Tom Lane [Mon, 20 Sep 2004 22:48:29 +0000 (22:48 +0000)]
Some editorializing on the docs for the dollar-quoting feature: fix
grammar, don't drop discussions into the middle of unrelated discussions,
etc.

20 years agoFix some shortcomings in psql's autocommit-off mode concerning detection
Tom Lane [Mon, 20 Sep 2004 18:51:19 +0000 (18:51 +0000)]
Fix some shortcomings in psql's autocommit-off mode concerning detection
of commands for which a transaction block should not be forced.  Recognize
VACUUM and other PreventTransactionChain commands; handle nested /* .. */
comments correctly; handle multibyte encodings correctly.
Michael Paesold with some kibitzing from Tom Lane.

20 years agoDocumentation improvements, per Josh Berkus.
Tom Lane [Mon, 20 Sep 2004 17:53:59 +0000 (17:53 +0000)]
Documentation improvements, per Josh Berkus.

20 years agoTranslation update
Peter Eisentraut [Mon, 20 Sep 2004 08:15:29 +0000 (08:15 +0000)]
Translation update

20 years agoTranslation updates
Peter Eisentraut [Mon, 20 Sep 2004 07:57:14 +0000 (07:57 +0000)]
Translation updates

20 years agoTranslation updates
Peter Eisentraut [Mon, 20 Sep 2004 07:49:54 +0000 (07:49 +0000)]
Translation updates

20 years agoTranslation updates
Peter Eisentraut [Mon, 20 Sep 2004 07:44:42 +0000 (07:44 +0000)]
Translation updates

20 years agoNow that I look, table 8-1 has some other marginal misstatements in it.
Tom Lane [Mon, 20 Sep 2004 04:19:50 +0000 (04:19 +0000)]
Now that I look, table 8-1 has some other marginal misstatements in it.
If we're going to put brackets around optional precision specs in some
entries, we'd better do it in all.

20 years agoMinor documentation cleanup and improvement.
Neil Conway [Mon, 20 Sep 2004 00:04:19 +0000 (00:04 +0000)]
Minor documentation cleanup and improvement.

20 years agoAdd defenses against plpython functions being declared to take or return
Tom Lane [Sun, 19 Sep 2004 23:38:21 +0000 (23:38 +0000)]
Add defenses against plpython functions being declared to take or return
pseudotypes.  Not sure why I neglected to add these checks at the same
time I added them to the other PLs, but it seems I did.

20 years agoFix estimate_num_groups to be able to use expression-index statistics
Tom Lane [Sat, 18 Sep 2004 19:39:50 +0000 (19:39 +0000)]
Fix estimate_num_groups to be able to use expression-index statistics
when there is an expressional index matching a GROUP BY item.

20 years agoPut the brackets in the right places in timestamp entries in table 8-1.
Tom Lane [Sat, 18 Sep 2004 15:28:03 +0000 (15:28 +0000)]
Put the brackets in the right places in timestamp entries in table 8-1.
Spotted by Josh Purinton.

20 years agoMove comment at end of line to a separate line so trailing tabs don't
Peter Eisentraut [Sat, 18 Sep 2004 13:28:54 +0000 (13:28 +0000)]
Move comment at end of line to a separate line so trailing tabs don't
become part of the make variable value.

20 years agoFix memory leak in tokenize_file, per report from Vadim Passynkov.
Tom Lane [Sat, 18 Sep 2004 01:22:58 +0000 (01:22 +0000)]
Fix memory leak in tokenize_file, per report from Vadim Passynkov.

20 years agoImprove documentation of log rotation options, per Ed L.
Tom Lane [Fri, 17 Sep 2004 22:40:46 +0000 (22:40 +0000)]
Improve documentation of log rotation options, per Ed L.

20 years agoAdd support for GNU/Hurd and GNU/*BSD. For our purposes, they behave just
Peter Eisentraut [Fri, 17 Sep 2004 22:31:59 +0000 (22:31 +0000)]
Add support for GNU/Hurd and GNU/*BSD.  For our purposes, they behave just
like Linux; we just need to recognize the system types.

20 years agoNew config.guess and config.sub
Peter Eisentraut [Fri, 17 Sep 2004 22:09:21 +0000 (22:09 +0000)]
New config.guess and config.sub

20 years agoAllow WIN1250 as server encoding.
Peter Eisentraut [Fri, 17 Sep 2004 21:59:57 +0000 (21:59 +0000)]
Allow WIN1250 as server encoding.

20 years agoHere is a patch bringing oid2name into the 21st century.
Tom Lane [Fri, 17 Sep 2004 21:14:19 +0000 (21:14 +0000)]
Here is a patch bringing oid2name into the 21st century.

Alvaro Herrera

20 years agoHashed LEFT JOIN would miss outer tuples with no inner match if the join
Tom Lane [Fri, 17 Sep 2004 18:28:53 +0000 (18:28 +0000)]
Hashed LEFT JOIN would miss outer tuples with no inner match if the join
was large enough to be batched and the tuples fell into a batch where
there were no inner tuples at all.  Thanks to Xiaoyu Wang for finding a
test case that exposed this long-standing bug.

20 years agoNow that xmax and cmin are distinct fields again, we should zero xmax when
Tom Lane [Fri, 17 Sep 2004 18:09:55 +0000 (18:09 +0000)]
Now that xmax and cmin are distinct fields again, we should zero xmax when
creating a new tuple.  This is just for debugging sanity, though, since
nothing should be paying any attention to xmax when the HEAP_XMAX_INVALID
bit is set.

20 years agoHyperlink a reference to DROP CAST in the CREATE CAST reference page.
Neil Conway [Fri, 17 Sep 2004 02:06:33 +0000 (02:06 +0000)]
Hyperlink a reference to DROP CAST in the CREATE CAST reference page.

20 years agoMake discovery of python_configdir architecture independent. Solution
Joe Conway [Thu, 16 Sep 2004 23:30:30 +0000 (23:30 +0000)]
Make discovery of python_configdir architecture independent. Solution
from James William Pye.

20 years agoFix oversight: there's no reason for PG_TRY to use sigsetjmp(buf,1)
Tom Lane [Thu, 16 Sep 2004 21:59:17 +0000 (21:59 +0000)]
Fix oversight: there's no reason for PG_TRY to use sigsetjmp(buf,1)
since we don't change the signal mask during normal backend operations.
Use sigsetjmp(buf,0) to avoid many unnecessary kernel calls.

20 years agoRemove contrib/pg_logger, per recent discussion.
Tom Lane [Thu, 16 Sep 2004 21:20:20 +0000 (21:20 +0000)]
Remove contrib/pg_logger, per recent discussion.

20 years agoRemove erroneous Assert, per example from Kris Jurka.
Tom Lane [Thu, 16 Sep 2004 21:08:12 +0000 (21:08 +0000)]
Remove erroneous Assert, per example from Kris Jurka.

20 years agoAdd some marginal tweaks to eliminate memory leakages associated with
Tom Lane [Thu, 16 Sep 2004 20:17:49 +0000 (20:17 +0000)]
Add some marginal tweaks to eliminate memory leakages associated with
subtransactions.  Trivial subxacts (such as a plpgsql exception block
containing no database access) now demonstrably leak zero bytes.

20 years agoRecentXmin is too recent to use as the cutoff point for accessing
Tom Lane [Thu, 16 Sep 2004 18:35:23 +0000 (18:35 +0000)]
RecentXmin is too recent to use as the cutoff point for accessing
pg_subtrans --- what we need is the oldest xmin of any snapshot in use
in the current top transaction.  Introduce a new variable TransactionXmin
to play this role.  Fixes intermittent regression failure reported by
Neil Conway.

20 years agoRestructure subtransaction handling to reduce resource consumption,
Tom Lane [Thu, 16 Sep 2004 16:58:44 +0000 (16:58 +0000)]
Restructure subtransaction handling to reduce resource consumption,
as per recent discussions.  Invent SubTransactionIds that are managed like
CommandIds (ie, counter is reset at start of each top transaction), and
use these instead of TransactionIds to keep track of subtransaction status
in those modules that need it.  This means that a subtransaction does not
need an XID unless it actually inserts/modifies rows in the database.
Accordingly, don't assign it an XID nor take a lock on the XID until it
tries to do that.  This saves a lot of overhead for subtransactions that
are only used for error recovery (eg plpgsql exceptions).  Also, arrange
to release a subtransaction's XID lock as soon as the subtransaction
exits, in both the commit and abort cases.  This avoids holding many
unique locks after a long series of subtransactions.  The price is some
additional overhead in XactLockTableWait, but that seems acceptable.
Finally, restructure the state machine in xact.c to have a more orthogonal
set of states for subtransactions.

20 years agoMinor doc tweak: mention that function names can be optionally schema-
Neil Conway [Thu, 16 Sep 2004 04:16:08 +0000 (04:16 +0000)]
Minor doc tweak: mention that function names can be optionally schema-
qualified.

20 years agoFix a read of uninitialized memory in array_out(). Perform some minor
Neil Conway [Thu, 16 Sep 2004 03:15:54 +0000 (03:15 +0000)]
Fix a read of uninitialized memory in array_out(). Perform some minor
cosmetic code cleanup at the same time.

20 years agoCosmetic PL/PgSQL fix: declare the second parameter plpgsql_dstring_append
Neil Conway [Tue, 14 Sep 2004 23:46:46 +0000 (23:46 +0000)]
Cosmetic PL/PgSQL fix: declare the second parameter plpgsql_dstring_append
as a const char *, so that we don't need to cast away a const in gram.y

20 years agoMark as done:
Bruce Momjian [Tue, 14 Sep 2004 14:34:45 +0000 (14:34 +0000)]
Mark as done:

< * Point-in-time data recovery using backup and write-ahead log,
< * Create native Win32 port, http://momjian.postgresql.org/main/writings/pgsql/project/win32.html
> * -Point-in-time data recovery using backup and write-ahead log
> * -Create native Win32 port
470c470
<     o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW
>  o Fix PL/pgSQL RENAME to work on variables other than OLD/NEW

20 years agoNew translation
Peter Eisentraut [Tue, 14 Sep 2004 05:50:33 +0000 (05:50 +0000)]
New translation

20 years agoTranslation update
Peter Eisentraut [Tue, 14 Sep 2004 05:44:40 +0000 (05:44 +0000)]
Translation update

20 years agoFix contrib/cube and contrib/seg to compile on Windows.
Tom Lane [Tue, 14 Sep 2004 04:21:38 +0000 (04:21 +0000)]
Fix contrib/cube and contrib/seg to compile on Windows.
Andreas Pflug

20 years agoFix some minor issues with the new Win32 service code for autovacuum,
Tom Lane [Tue, 14 Sep 2004 04:06:20 +0000 (04:06 +0000)]
Fix some minor issues with the new Win32 service code for autovacuum,
and add documentation.  Dave Page

20 years agoAdjust tsearch2.sql to avoid use of COPY FROM STDIN, so as to
Tom Lane [Tue, 14 Sep 2004 03:58:54 +0000 (03:58 +0000)]
Adjust tsearch2.sql to avoid use of COPY FROM STDIN, so as to
simplify life for the Win32 installer.  Per Dave Page.

20 years agoWin32 compile fix for misc_utils.
Tom Lane [Tue, 14 Sep 2004 03:50:17 +0000 (03:50 +0000)]
Win32 compile fix for misc_utils.
Claudio Natoli

20 years agoWin32 compile fixes for pgbench, pgcrypto, and tsearch.
Tom Lane [Tue, 14 Sep 2004 03:39:55 +0000 (03:39 +0000)]
Win32 compile fixes for pgbench, pgcrypto, and tsearch.
Claudio Natoli

20 years agoFix contrib/dbase to compile under Win32. Laurent Ballester
Tom Lane [Tue, 14 Sep 2004 03:28:28 +0000 (03:28 +0000)]
Fix contrib/dbase to compile under Win32.  Laurent Ballester

20 years agoMake pltcl work on Win32. Magnus Hagander
Tom Lane [Tue, 14 Sep 2004 03:21:27 +0000 (03:21 +0000)]
Make pltcl work on Win32.  Magnus Hagander

20 years agoFix small memory leak in psql.
Neil Conway [Mon, 13 Sep 2004 23:07:12 +0000 (23:07 +0000)]
Fix small memory leak in psql.

20 years agoTranslation update
Peter Eisentraut [Mon, 13 Sep 2004 20:53:10 +0000 (20:53 +0000)]
Translation update

20 years agoRedesign query-snapshot timing so that volatile functions in READ COMMITTED
Tom Lane [Mon, 13 Sep 2004 20:10:13 +0000 (20:10 +0000)]
Redesign query-snapshot timing so that volatile functions in READ COMMITTED
mode see a fresh snapshot for each command in the function, rather than
using the latest interactive command's snapshot.  Also, suppress fresh
snapshots as well as CommandCounterIncrement inside STABLE and IMMUTABLE
functions, instead using the snapshot taken for the most closely nested
regular query.  (This behavior is only sane for read-only functions, so
the patch also enforces that such functions contain only SELECT commands.)
As per my proposal of 6-Sep-2004; I note that I floated essentially the
same proposal on 19-Jun-2002, but that discussion tailed off without any
action.  Since 8.0 seems like the right place to be taking possibly
nontrivial backwards compatibility hits, let's get it done now.

20 years agoTranslation updates
Peter Eisentraut [Mon, 13 Sep 2004 12:43:27 +0000 (12:43 +0000)]
Translation updates

20 years agoTranslation updates
Peter Eisentraut [Mon, 13 Sep 2004 12:35:18 +0000 (12:35 +0000)]
Translation updates

20 years agoNew translations
Peter Eisentraut [Mon, 13 Sep 2004 12:20:24 +0000 (12:20 +0000)]
New translations

20 years agoNew translation
Peter Eisentraut [Mon, 13 Sep 2004 12:14:50 +0000 (12:14 +0000)]
New translation

20 years agoTranslation updates
Peter Eisentraut [Mon, 13 Sep 2004 12:11:22 +0000 (12:11 +0000)]
Translation updates

20 years agoFix two typos in comments.
Neil Conway [Mon, 13 Sep 2004 01:45:32 +0000 (01:45 +0000)]
Fix two typos in comments.

20 years agoRe-add brace removal code but comment it out so we know why we removed
Bruce Momjian [Sun, 12 Sep 2004 22:21:30 +0000 (22:21 +0000)]
Re-add brace removal code but comment it out so we know why we removed
it and have it in case we need it for some special case.

20 years agoRemove code that delete braces around single statements.
Bruce Momjian [Sun, 12 Sep 2004 22:11:27 +0000 (22:11 +0000)]
Remove code that delete braces around single statements.

20 years agoWhen LockAcquire fails at the stage of creating a proclock object, be
Tom Lane [Sun, 12 Sep 2004 18:30:50 +0000 (18:30 +0000)]
When LockAcquire fails at the stage of creating a proclock object, be
sure to clean up the already-created lock object, if it has no other
references.  Avoids possibly-permanent leak of shared memory.

20 years agoRenumber SnapshotNow and the other special snapshot codes so that
Tom Lane [Sat, 11 Sep 2004 18:28:34 +0000 (18:28 +0000)]
Renumber SnapshotNow and the other special snapshot codes so that
((Snapshot) NULL) can no longer be confused with a valid snapshot,
as per my recent suggestion.  Define a macro InvalidSnapshot for 0.
Use InvalidSnapshot instead of SnapshotAny as the do-nothing special
case for heap_update and heap_delete crosschecks; this seems a little
cleaner even though the behavior is really the same.

20 years agoEnsure that pg_largeobject references opened by lo_import() or lo_export()
Tom Lane [Sat, 11 Sep 2004 15:56:46 +0000 (15:56 +0000)]
Ensure that pg_largeobject references opened by lo_import() or lo_export()
will be cleaned up at end of transaction, even when there is no other LO
operation in the transaction.  Per bug report from Daniel Schuchardt.

20 years agoSend thread test output to file descriptor 5 like configure does rather
Bruce Momjian [Sat, 11 Sep 2004 02:12:17 +0000 (02:12 +0000)]
Send thread test output to file descriptor 5 like configure does rather
than /dev/null, which Win32 doesn't have.

20 years agoCheck for ignored thread compiler options to reduce compiler noise.
Bruce Momjian [Sat, 11 Sep 2004 00:03:06 +0000 (00:03 +0000)]
Check for ignored thread compiler options to reduce compiler noise.

20 years agoDone:
Bruce Momjian [Fri, 10 Sep 2004 22:38:52 +0000 (22:38 +0000)]
Done:

< * Have AFTER triggers execute after the appropriate SQL statement in a
> * -Have AFTER triggers execute after the appropriate SQL statement in a

20 years agoFix some problems with restoring databases owned by non-superusers,
Tom Lane [Fri, 10 Sep 2004 20:05:18 +0000 (20:05 +0000)]
Fix some problems with restoring databases owned by non-superusers,
as per bug #1249; and remove the last vestiges of using \connect to
change authorization.

20 years agoFire non-deferred AFTER triggers immediately upon query completion,
Tom Lane [Fri, 10 Sep 2004 18:40:09 +0000 (18:40 +0000)]
Fire non-deferred AFTER triggers immediately upon query completion,
rather than when returning to the idle loop.  This makes no particular
difference for interactively-issued queries, but it makes a big difference
for queries issued within functions: trigger execution now occurs before
the calling function is allowed to proceed.  This responds to numerous
complaints about nonintuitive behavior of foreign key checking, such as
http://archives.postgresql.org/pgsql-bugs/2004-09/msg00020.php, and
appears to be required by the SQL99 spec.
Also take the opportunity to simplify the data structures used for the
pending-trigger list, rename them for more clarity, and squeeze out a
bit of space.

20 years agoAdd:
Bruce Momjian [Fri, 10 Sep 2004 16:35:56 +0000 (16:35 +0000)]
Add:

> * Make row-wise comparisons work per SQL spec
>

20 years agoFix #if define
Bruce Momjian [Fri, 10 Sep 2004 15:51:47 +0000 (15:51 +0000)]
Fix #if define

20 years agoFix palloc call from /port for Cygwin.
Bruce Momjian [Fri, 10 Sep 2004 15:23:51 +0000 (15:23 +0000)]
Fix palloc call from /port for Cygwin.

20 years agoProperly include port file for Cygwin.
Bruce Momjian [Fri, 10 Sep 2004 15:20:19 +0000 (15:20 +0000)]
Properly include port file for Cygwin.

20 years agoMinor cleanup.
Bruce Momjian [Fri, 10 Sep 2004 14:27:37 +0000 (14:27 +0000)]
Minor cleanup.

20 years agoMore cleanup.
Bruce Momjian [Fri, 10 Sep 2004 14:24:25 +0000 (14:24 +0000)]
More cleanup.

20 years agoUse $PORTNAME consistently instead of $host_os, where appropriate.
Bruce Momjian [Fri, 10 Sep 2004 13:53:40 +0000 (13:53 +0000)]
Use $PORTNAME consistently instead of $host_os, where appropriate.

20 years agoAdd:
Bruce Momjian [Fri, 10 Sep 2004 10:03:41 +0000 (10:03 +0000)]
Add:

> * Remove Win32 rename/unlink looping if unnecessary

20 years agoAdd mention loops over unlink/rename might not be needed.
Bruce Momjian [Fri, 10 Sep 2004 09:58:21 +0000 (09:58 +0000)]
Add mention loops over unlink/rename might not be needed.

20 years agoMove undef of rename/unlink so system declarations are unmodified.
Bruce Momjian [Fri, 10 Sep 2004 09:53:08 +0000 (09:53 +0000)]
Move undef of rename/unlink so system declarations are unmodified.

20 years agoAdjust quotes for win32 build of psql.
Bruce Momjian [Fri, 10 Sep 2004 09:45:21 +0000 (09:45 +0000)]
Adjust quotes for win32 build of psql.