OSDN Git Service

pg-rex/syncrep.git
21 years agoAdd:
Bruce Momjian [Sun, 20 Oct 2002 02:30:04 +0000 (02:30 +0000)]
Add:

>  o Add GUC parameter to control the maximum number of rewrite cycles

21 years agoAdd warning that autocommit=off is not well-supported yet.
Tom Lane [Sun, 20 Oct 2002 01:43:32 +0000 (01:43 +0000)]
Add warning that autocommit=off is not well-supported yet.

21 years agoMove BeOS and QNX4 to unsupported-platform list, until someone cares
Tom Lane [Sun, 20 Oct 2002 01:38:18 +0000 (01:38 +0000)]
Move BeOS and QNX4 to unsupported-platform list, until someone cares
to expend the effort to update them for new semaphore code.

21 years agoRule rewriter was doing the wrong thing with conditional INSTEAD rules
Tom Lane [Sun, 20 Oct 2002 00:58:55 +0000 (00:58 +0000)]
Rule rewriter was doing the wrong thing with conditional INSTEAD rules
whose conditions might yield NULL.  The negated qual to attach to the
original query is properly 'x IS NOT TRUE', not 'NOT x'.  This fix
produces correct behavior, but we may be taking a performance hit because
the planner is much stupider about IS NOT TRUE than it is about NOT
clauses.  Future TODO: teach prepqual, other parts of planner how to
cope with BooleanTest clauses more effectively.

21 years agoDisallow aggregate functions in rule WHERE clauses. Per gripe from
Tom Lane [Sun, 20 Oct 2002 00:31:53 +0000 (00:31 +0000)]
Disallow aggregate functions in rule WHERE clauses.  Per gripe from
Fritz Lehmann-Grube back in January.

21 years agoApplied patch submitted by Mike Beachy to give a better error message if
Barry Lind [Sun, 20 Oct 2002 00:10:55 +0000 (00:10 +0000)]
Applied patch submitted by Mike Beachy to give a better error message if
configure hasn't been run before trying to build.
Also cleaned up the README file and removed some obsolete files.

 Modified Files:
  jdbc/README jdbc/build.xml
 Removed Files:
  jdbc/CHANGELOG jdbc/Implementation jdbc/jdbc.jpx

21 years agoFix inconsistent formatting.
Tom Lane [Sat, 19 Oct 2002 23:09:20 +0000 (23:09 +0000)]
Fix inconsistent formatting.

21 years agoFix incomplete definition of ALTER TABLE ADD/DROP CONSTRAINT syntax.
Tom Lane [Sat, 19 Oct 2002 22:51:45 +0000 (22:51 +0000)]
Fix incomplete definition of ALTER TABLE ADD/DROP CONSTRAINT syntax.
Add some verbiage about recent tweaks to behavior of ADD and DROP
COLUMN when there are descendant tables.

21 years agoFix within-function memory leaks in the various PLs' interfaces to
Tom Lane [Sat, 19 Oct 2002 22:10:58 +0000 (22:10 +0000)]
Fix within-function memory leaks in the various PLs' interfaces to
SPI_prepare: they all save the prepared plan into topCxt, and so the
procCxt copy that's actually returned by SPI_prepare ought to be freed.
Diagnosis and plpython fix by Nigel Andrews, followup for other PLs
by Tom Lane.

21 years agoApplied patch submitted by Kris Jurka to result in a better error message
Barry Lind [Sat, 19 Oct 2002 22:10:36 +0000 (22:10 +0000)]
Applied patch submitted by Kris Jurka to result in a better error message
under some circumstances and handle negative money values better.

 Modified Files:
  jdbc/org/postgresql/jdbc1/AbstractJdbc1ResultSet.java

21 years agoFix pltcl self-test for recent changes to elog.
Tom Lane [Sat, 19 Oct 2002 22:01:45 +0000 (22:01 +0000)]
Fix pltcl self-test for recent changes to elog.

21 years agoBacked out part of the change from 1.6. The attempt to support int8 binds
Barry Lind [Sat, 19 Oct 2002 21:53:42 +0000 (21:53 +0000)]
Backed out part of the change from 1.6.  The attempt to support int8 binds
in such a way that indexes on int8 columns would be used (by quoting the value)
caused other problems.  Will need to wait for the backend to properly fix
the root problem.

 Modified Files:
  jdbc/org/postgresql/jdbc1/AbstractJdbc1Statement.java

21 years agoFix case where a function in FROM returns a scalar type, but is
Tom Lane [Sat, 19 Oct 2002 21:23:20 +0000 (21:23 +0000)]
Fix case where a function in FROM returns a scalar type, but is
referred to with whole-tuple syntax.

21 years agoMake psql's \df display functions that return sets as having return type
Tom Lane [Sat, 19 Oct 2002 20:50:44 +0000 (20:50 +0000)]
Make psql's \df display functions that return sets as having return type
'setof something'; formerly you could not tell at all that the function
returns set.

21 years agoInvert logic in pg_exec_query_string() so that we set a snapshot for
Tom Lane [Sat, 19 Oct 2002 20:15:09 +0000 (20:15 +0000)]
Invert logic in pg_exec_query_string() so that we set a snapshot for
all utility statement types *except* a short list, per discussion a few
days ago.  Add missing SetQuerySnapshot calls in VACUUM and REINDEX,
and guard against calling REINDEX DATABASE from a function (has same
problem as VACUUM).

21 years agoFix rewrite code so that rules are in fact executed in order by name,
Tom Lane [Sat, 19 Oct 2002 19:00:47 +0000 (19:00 +0000)]
Fix rewrite code so that rules are in fact executed in order by name,
rather than being reordered according to INSTEAD attribute for
implementation convenience.
Also, increase compiled-in recursion depth limit from 10 to 100 rewrite
cycles.  10 seems pretty marginal for situations where multiple rules
exist for the same query.  There was a complaint about this recently,
so I'm going to bump it up.  (Perhaps we should make the limit a GUC
parameter, but that's too close to being a new feature to do in beta.)

21 years agoBack out Alvaro's patch until regression tests pass.
Bruce Momjian [Sat, 19 Oct 2002 03:01:09 +0000 (03:01 +0000)]
Back out Alvaro's patch until regression tests pass.

21 years agoFix range-query estimation to not double-exclude NULLs, per gripe from
Tom Lane [Sat, 19 Oct 2002 02:56:16 +0000 (02:56 +0000)]
Fix range-query estimation to not double-exclude NULLs, per gripe from
Ray Ontko 28-June-02.  Also, fix prefix_selectivity for NAME lefthand
variables (it was bogusly assuming binary compatibility), and adjust
make_greater_string() to not call pg_mbcliplen() with invalid multibyte
data (this last per bug report that I can't find at the moment, but it
was in July '02).

21 years agoAdd:
Bruce Momjian [Sat, 19 Oct 2002 02:31:10 +0000 (02:31 +0000)]
Add:

> * Move /contrib/retep to gborg.postgresql.org

21 years agoFix compile failure caused by new patch.
Bruce Momjian [Sat, 19 Oct 2002 02:25:51 +0000 (02:25 +0000)]
Fix compile failure caused by new patch.

21 years agoRemove mention of utils/getopt.c. Not used anymore.
Bruce Momjian [Sat, 19 Oct 2002 02:23:26 +0000 (02:23 +0000)]
Remove mention of utils/getopt.c.  Not used anymore.

21 years agoIt includes
Bruce Momjian [Sat, 19 Oct 2002 02:16:40 +0000 (02:16 +0000)]
It includes
-Support for mirroring tables in different Schema's
-Improved documentation for compiling with 7.1.x and 7.2.x
-Fixes a buffer overrun bug.

Steven Singer

21 years ago> Huh, I don't know where I got the idea you were (or someone else was?)
Bruce Momjian [Sat, 19 Oct 2002 02:09:45 +0000 (02:09 +0000)]
> Huh, I don't know where I got the idea you were (or someone else was?)
> in the position that attislocal should be reset.  I'll clean everything
> up and submit the patch I had originally made.

All right, this is it.  This patch merely checks if child tables have
the column.  If atttypid and atttypmod are the same, the attributes'
attinhcount is incremented; else the operation is aborted.  If child
tables don't have the column, recursively add it.

attislocal is not touched in any case.

Alvaro Herrera

21 years agoThis patch adds some missing functions for float8 math operations,
Bruce Momjian [Sat, 19 Oct 2002 02:08:19 +0000 (02:08 +0000)]
This patch adds some missing functions for float8 math operations,
specifically ceil(), floor(), and sign(). There may be other functions
that need to be added, but this is a start. I've included some simple
regression tests.

Neil Conway

21 years agoMake regression tests safe for autocommit = 'off'.
Bruce Momjian [Sat, 19 Oct 2002 01:35:43 +0000 (01:35 +0000)]
Make regression tests safe for autocommit = 'off'.

21 years agoAdd missing #include <errno.h>.
Tom Lane [Sat, 19 Oct 2002 00:25:36 +0000 (00:25 +0000)]
Add missing #include <errno.h>.

21 years agoFix psql's \copy to accept table names containing schemas, as well as
Tom Lane [Sat, 19 Oct 2002 00:22:14 +0000 (00:22 +0000)]
Fix psql's \copy to accept table names containing schemas, as well as
a column list.  Bring its parsing of quoted names and quoted strings
somewhat up to speed --- I believe it now handles all non-error cases
the same way the backend would, but weird boundary conditions are not
necessarily done the same way.

21 years agoImprove formatting of --help output.
Peter Eisentraut [Fri, 18 Oct 2002 22:05:36 +0000 (22:05 +0000)]
Improve formatting of --help output.

21 years agoFix breakage that had crept into setlocale() usage: once again we've
Tom Lane [Fri, 18 Oct 2002 20:44:02 +0000 (20:44 +0000)]
Fix breakage that had crept into setlocale() usage: once again we've
been bit by the fact that the locale functions return pointers to
modifiable variables.  I added some comments that might help us avoid
the mistake in future.

21 years agoMake 'dummy' declarations in header files be 'extern int no_such_variable'
Tom Lane [Fri, 18 Oct 2002 20:33:57 +0000 (20:33 +0000)]
Make 'dummy' declarations in header files be 'extern int no_such_variable'
instead of 'extern int errno'; the latter is unsafe according to the
ANSI C standard, as well as in practice on some platforms.

21 years agoUn-break createlang build.
Tom Lane [Fri, 18 Oct 2002 19:35:28 +0000 (19:35 +0000)]
Un-break createlang build.

21 years agoClean up for dblink autocommit OFF handling.
Bruce Momjian [Fri, 18 Oct 2002 19:11:27 +0000 (19:11 +0000)]
Clean up for dblink autocommit OFF handling.

21 years agoUpdate /contrib for "autocommit TO 'on'".
Bruce Momjian [Fri, 18 Oct 2002 18:41:22 +0000 (18:41 +0000)]
Update /contrib for "autocommit TO 'on'".

Create objects in public schema.

Make spacing/capitalization consistent.

Remove transaction block use for object creation.

Remove unneeded function GRANTs.

21 years agoImprove discussion of FOR UPDATE.
Tom Lane [Fri, 18 Oct 2002 18:26:22 +0000 (18:26 +0000)]
Improve discussion of FOR UPDATE.

21 years agoUpdate for tables returning function, item 5.3, from Joe Conway.
Bruce Momjian [Thu, 17 Oct 2002 23:34:10 +0000 (23:34 +0000)]
Update for tables returning function, item 5.3, from Joe Conway.

21 years agoFix in updateable result sets to handle binding null values correctly
Barry Lind [Thu, 17 Oct 2002 19:17:08 +0000 (19:17 +0000)]
Fix in updateable result sets to handle binding null values correctly
Patch submitted by Kris Jurka (applied with some modifications)

 Modified Files:
  jdbc/org/postgresql/jdbc2/AbstractJdbc2ResultSet.java

21 years agoFix free-slot search in PgSetResultId so it actually works.
Tom Lane [Thu, 17 Oct 2002 14:53:32 +0000 (14:53 +0000)]
Fix free-slot search in PgSetResultId so it actually works.

21 years agoImprove formatting of error message.
Tom Lane [Thu, 17 Oct 2002 14:51:50 +0000 (14:51 +0000)]
Improve formatting of error message.

21 years agoFixed support in jdbc for 7.3 server autocommit. With these changes the
Barry Lind [Thu, 17 Oct 2002 05:33:52 +0000 (05:33 +0000)]
Fixed support in jdbc for 7.3 server autocommit.  With these changes the
jdbc regression tests pass for both autocommit on and autocommit off

 Modified Files:
  jdbc/org/postgresql/jdbc1/AbstractJdbc1Connection.java
  jdbc/org/postgresql/test/jdbc2/ConnectionTest.java

21 years agoAdd:
Bruce Momjian [Thu, 17 Oct 2002 03:58:22 +0000 (03:58 +0000)]
Add:

>  o Allow copy to understand \x as hex

21 years agoPut the disk information all in one chapter.
Peter Eisentraut [Wed, 16 Oct 2002 22:06:33 +0000 (22:06 +0000)]
Put the disk information all in one chapter.

21 years agoMake pg_dump save for autocommit = off.
Bruce Momjian [Wed, 16 Oct 2002 05:46:54 +0000 (05:46 +0000)]
Make pg_dump save for autocommit = off.

21 years agoFix from Joe on timeout code.
Bruce Momjian [Wed, 16 Oct 2002 04:38:00 +0000 (04:38 +0000)]
Fix from Joe on timeout code.

21 years agoImprove appearance of SET command.
Bruce Momjian [Wed, 16 Oct 2002 03:44:28 +0000 (03:44 +0000)]
Improve appearance of SET command.

21 years agoFix script to handle autocommit = 'off' by prepending autocommit 'on' to
Bruce Momjian [Wed, 16 Oct 2002 03:24:09 +0000 (03:24 +0000)]
Fix script to handle autocommit = 'off' by prepending autocommit 'on' to
the start of the psql commandline.  This is better than adding BEGIN/END
because it handles multiple queries well, and allows the return code for
psql to return the proper value.

21 years agoFix connection_timeout to use time() and handle timeout == 1.
Bruce Momjian [Wed, 16 Oct 2002 02:55:30 +0000 (02:55 +0000)]
Fix connection_timeout to use time() and handle timeout == 1.

Code cleanup.

21 years agoRemove:
Bruce Momjian [Tue, 15 Oct 2002 21:34:10 +0000 (21:34 +0000)]
Remove:

< * Rename oid2name to relfilenode2name and install by default
< * Move /contrib/oid2name to /bin, rename to relfilenode2name

21 years agoAdd:
Bruce Momjian [Tue, 15 Oct 2002 18:54:56 +0000 (18:54 +0000)]
Add:

> * Move /contrib/oid2name to /bin, rename to relfilenode2name

21 years agoFix unintentional breakage of COPY TO/FROM stdin. Mea culpa.
Tom Lane [Tue, 15 Oct 2002 16:44:21 +0000 (16:44 +0000)]
Fix unintentional breakage of COPY TO/FROM stdin.  Mea culpa.

21 years agoFix Linux dynloader code for pre-HAVE_DLOPEN systems, which evidently
Tom Lane [Tue, 15 Oct 2002 16:04:17 +0000 (16:04 +0000)]
Fix Linux dynloader code for pre-HAVE_DLOPEN systems, which evidently
are still in use out there.  Per report from Brendan LeFebvre.

21 years agoFix psql to cope with autocommit off, at least during startup.
Tom Lane [Tue, 15 Oct 2002 02:24:16 +0000 (02:24 +0000)]
Fix psql to cope with autocommit off, at least during startup.
Behavior of backslash commands (especially for large objects)
may still require some thought.

21 years agoFix libpq startup code to work correctly in autocommit off mode.
Tom Lane [Tue, 15 Oct 2002 01:48:25 +0000 (01:48 +0000)]
Fix libpq startup code to work correctly in autocommit off mode.
In passing, fix breakage for case where PGCLIENTENCODING is set in
environment.

21 years agoMake SPI's execution of querystrings follow the rules agreed to for
Tom Lane [Mon, 14 Oct 2002 23:49:20 +0000 (23:49 +0000)]
Make SPI's execution of querystrings follow the rules agreed to for
command status at the interactive level.  SPI_processed, etc are set
in the same way as the returned command status would have been set if
the same querystring were issued interactively.  Per gripe from
Michael Paesold 25-Sep-02.

21 years agoNone.
Bruce Momjian [Mon, 14 Oct 2002 22:43:45 +0000 (22:43 +0000)]
None.

21 years agoAdjust handling of command status strings in the presence of rules,
Tom Lane [Mon, 14 Oct 2002 22:14:35 +0000 (22:14 +0000)]
Adjust handling of command status strings in the presence of rules,
as per recent pghackers discussions.  initdb forced due to change in
fields of stored Query nodes.

21 years agoAdd missing entries for bit<->int4 and bit<->int8 conversions.
Tom Lane [Mon, 14 Oct 2002 22:12:49 +0000 (22:12 +0000)]
Add missing entries for bit<->int4 and bit<->int8 conversions.

21 years agoTranslation updates
Peter Eisentraut [Mon, 14 Oct 2002 19:04:28 +0000 (19:04 +0000)]
Translation updates

21 years agoMention that resetting the timeout may be wrong on select retry.
Bruce Momjian [Mon, 14 Oct 2002 18:11:17 +0000 (18:11 +0000)]
Mention that resetting the timeout may be wrong on select retry.

21 years agoRestore ptmp_timeout for cases where no timeout is passed.
Bruce Momjian [Mon, 14 Oct 2002 17:33:08 +0000 (17:33 +0000)]
Restore ptmp_timeout for cases where no timeout is passed.

21 years agolibpq connection_timeout doesn't do subsecond timing, so make the code
Bruce Momjian [Mon, 14 Oct 2002 17:15:11 +0000 (17:15 +0000)]
libpq connection_timeout doesn't do subsecond timing, so make the code
clear on that point.

21 years agoArrange to copy relcache's trigdesc structure at the start of any
Tom Lane [Mon, 14 Oct 2002 16:51:30 +0000 (16:51 +0000)]
Arrange to copy relcache's trigdesc structure at the start of any
query that uses it.  This ensures that triggers will be applied consistently
throughout a query even if someone commits changes to the relation's
pg_class.reltriggers field meanwhile.  Per crash report from Laurette Cisneros.
While at it, simplify memory management in relcache.c, which no longer
needs the old hack to try to keep trigger info in the same place over
a relcache entry rebuild.  (Should try to fix rd_att and rewrite-rule
access similarly, someday.)  And make RelationBuildTriggers simpler and
more robust by making it build the trigdesc in working memory and then
CopyTriggerDesc() into cache memory.

21 years ago- Link the entries in the table to the catalog heading
Bruce Momjian [Mon, 14 Oct 2002 04:29:23 +0000 (04:29 +0000)]
- Link the entries in the table to the catalog heading
- Wrap them in the <database class="table"> tags, since thats what they
are (no markup rules for this, so it inherits from parent -- no style
change)
- Mention that pg_database, pg_shadow, and pg_group are global, and the
rest are local to the specific DB. (I believe this is correct).

> Works for me, though I suppose we could explain what the exceptions are
> like in general terms.  Perhaps something like
>
> 'Most system catalogs are copied from the template database during
> database creation, and are thereafter database-specific.  A few
> catalogs are physically shared across all databases in an installation;
> these are marked in the descriptions of the individual catalogs.'

Ok, new patch.

Rod Taylor

21 years agoThis trivial patch fixes a bunch of spelling mistakes in the
Bruce Momjian [Mon, 14 Oct 2002 04:27:25 +0000 (04:27 +0000)]
This trivial patch fixes a bunch of spelling mistakes in the
contrib/dbmirror/README.dbmirror doc file.

Neil Conway

21 years agoAs Niel so nicely pointed out this morning, the output of EXPLAIN
Bruce Momjian [Mon, 14 Oct 2002 04:26:54 +0000 (04:26 +0000)]
As Niel so nicely pointed out this morning, the output of EXPLAIN
ANALYZE is not quite clear when branches of the query are never
executed. So this tiny patch fixes that.

The patch is attached and can also be found at:
http://svana.org/kleptog/pgsql/pgsql-explain.patch

Martijn van Oosterhout

21 years agoI have attached two patches as per:
Bruce Momjian [Mon, 14 Oct 2002 04:20:52 +0000 (04:20 +0000)]
I have attached two patches as per:

1) pltcl:
Add SPI_freetuptable() calls to avoid memory leaks (Me + Neil Conway)
Change sprintf()s to snprintf()s (Neil Conway)
Remove header files included elsewhere (Neil Conway)

2)plpython:
Add SPI_freetuptable() calls to avoid memory leaks
Cosemtic change to remove a compiler warning

Notes:

I have tested pltcl.c for
 a) the original leak problem reported for the repeated call of spi_exec
in a TCL fragment
and
 b) the subsequent report resulting from the use of spi_exec -array
in a TCL
fragment.

The plpython.c patch is exactly the same as that applied to make
revision 1.23,
the plpython_schema.sql and feature.expected sections of the patch are
also the
same as last submited, applied and subsequently reversed out. It remains
untested by me (other than via make check). However, this should be safe
provided PyString_FromString() _copies_ the given string to make a
PyObject.

Nigel J. Andrews

21 years agoImprovements to Ian Barwick patch.
Bruce Momjian [Mon, 14 Oct 2002 03:15:33 +0000 (03:15 +0000)]
Improvements to Ian Barwick patch.

21 years agoFAQ updates from Ian Barwick.
Bruce Momjian [Mon, 14 Oct 2002 02:50:28 +0000 (02:50 +0000)]
FAQ updates from Ian Barwick.

21 years agoMake SET really not start a transaction.
Tom Lane [Sun, 13 Oct 2002 16:55:05 +0000 (16:55 +0000)]
Make SET really not start a transaction.

21 years agoMake macaddr_in reject trailing garbage (except whitespace).
Tom Lane [Sun, 13 Oct 2002 15:39:17 +0000 (15:39 +0000)]
Make macaddr_in reject trailing garbage (except whitespace).
Per gripe from Patrick Welche, 13-Oct-2002.

21 years agopsql thought that backslash is an escape character inside double quotes.
Tom Lane [Sat, 12 Oct 2002 23:09:34 +0000 (23:09 +0000)]
psql thought that backslash is an escape character inside double quotes.
It isn't.

21 years agoFix for bug #795: two clauses that seem redundant are not really, if
Tom Lane [Sat, 12 Oct 2002 22:24:49 +0000 (22:24 +0000)]
Fix for bug #795: two clauses that seem redundant are not really, if
one is pushed down into an outer join and the other is not.

21 years agoUpdate README for oid2name.
Bruce Momjian [Sat, 12 Oct 2002 19:15:09 +0000 (19:15 +0000)]
Update README for oid2name.

21 years agoAdd:
Bruce Momjian [Sat, 12 Oct 2002 19:03:41 +0000 (19:03 +0000)]
Add:

> * Rename oid2name to relfilenode2name and install by default

21 years agoAdded:
Bruce Momjian [Sat, 12 Oct 2002 17:01:19 +0000 (17:01 +0000)]
Added:

>  o Allow psql \copy to specify column names

21 years agoCompute version number for docs on the fly.
Peter Eisentraut [Sat, 12 Oct 2002 16:34:28 +0000 (16:34 +0000)]
Compute version number for docs on the fly.

21 years agoHandle indentation of verbatim environments in HTML output via CSS.
Peter Eisentraut [Sat, 12 Oct 2002 16:33:43 +0000 (16:33 +0000)]
Handle indentation of verbatim environments in HTML output via CSS.

21 years agoFix linking problem.
Peter Eisentraut [Sat, 12 Oct 2002 16:31:55 +0000 (16:31 +0000)]
Fix linking problem.

21 years agoReplace &version; by appropriate version.
Peter Eisentraut [Sat, 12 Oct 2002 16:29:51 +0000 (16:29 +0000)]
Replace &version; by appropriate version.

21 years agoAssorted reference page updates
Peter Eisentraut [Fri, 11 Oct 2002 23:03:48 +0000 (23:03 +0000)]
Assorted reference page updates

21 years agoUpdate FAQ.
Bruce Momjian [Fri, 11 Oct 2002 17:55:17 +0000 (17:55 +0000)]
Update FAQ.

21 years agoUpdate Russian FAQ.
Bruce Momjian [Fri, 11 Oct 2002 17:45:40 +0000 (17:45 +0000)]
Update Russian FAQ.

21 years agoAdd replication, encryption, and cross database FAQ items.
Bruce Momjian [Fri, 11 Oct 2002 05:02:24 +0000 (05:02 +0000)]
Add replication, encryption, and cross database FAQ items.

21 years agoAdd tv_sec change for connection timeout suggested by author.
Bruce Momjian [Fri, 11 Oct 2002 04:41:59 +0000 (04:41 +0000)]
Add tv_sec change for connection timeout suggested by author.

21 years agoOops, back out newNode changes. We are not ready for that yet.
Bruce Momjian [Fri, 11 Oct 2002 04:16:44 +0000 (04:16 +0000)]
Oops, back out newNode changes.  We are not ready for that yet.

21 years agoPrevent tv_sec from becoming negative in connection timeout code.
Bruce Momjian [Fri, 11 Oct 2002 04:12:14 +0000 (04:12 +0000)]
Prevent tv_sec from becoming negative in connection timeout code.

21 years agoUpdate:
Bruce Momjian [Fri, 11 Oct 2002 03:32:45 +0000 (03:32 +0000)]
Update:

> * -Add pg_backend_pid() function to backend

21 years agoUpdate IN/EXISTS item.
Bruce Momjian [Thu, 10 Oct 2002 03:15:19 +0000 (03:15 +0000)]
Update IN/EXISTS item.

21 years agoset.patch updates an example in ref/set.sgml to have microsecond
Bruce Momjian [Wed, 9 Oct 2002 16:27:48 +0000 (16:27 +0000)]
set.patch updates an example in ref/set.sgml to have microsecond
precision.

vacuum.patch updates ref/vacuum.sgml to explicitly state that an
exclusive lock is not obtained during normal (non-FULL) vacuum.

  Rod Taylor

21 years agoLock on the rule relation wasn't removed after adding the comment.
Bruce Momjian [Wed, 9 Oct 2002 16:26:46 +0000 (16:26 +0000)]
Lock on the rule relation wasn't removed after adding the comment.

Added Tom's patch fix for heap_close.

Rod Taylor

21 years ago Well, this patch makes Makefile for contrib/rserv use the
Bruce Momjian [Wed, 9 Oct 2002 16:23:55 +0000 (16:23 +0000)]
   Well, this patch makes Makefile for contrib/rserv use the
   contrib/contrib-global.mk library and _generally_ behave like
   Makefiles for other contrib modules.
   Besides it fixes Perl's interpolation of $libdir variable, which
   should be passed to backend instead. This patch is done against
   PostgreSQL 7.3b2

   Besides, I want to thank Peter Eisentraut for his very friendly and
   helpful attitude and politely ask him to check whether contrib
   modules actually continue to work after he implements another
   major change to their build process.

Alexey Borzov

21 years ago> > > > and mb conversions (pg_ascii2mic and pg_mic2ascii not
Bruce Momjian [Wed, 9 Oct 2002 16:21:54 +0000 (16:21 +0000)]
> > > > and mb conversions (pg_ascii2mic and pg_mic2ascii not
> > > > found in the postmaster and not included from elsewhere)
> >
> > shared libs on AIX need to be able to resolve all symbols at linkage time.
> > Those two symbols are in backend/utils/SUBSYS.o but not in the postgres
> > executable.
>
> They are defined in backend/utils/mb/conv.c and declared in
> include/mb/pg_wchar.h.  They're also linked into the
> postmaster.  I don't see anything unusual.

Attached is a patch to fix the mb linking problems on AIX. As a nice side effect

it reduces the duplicate symbol warnings to linking libpq.so and libecpg.so
(all shlibs that are not postmaster loadable modules).

Please apply to current (only affects AIX).

The _LARGE_FILES problem is unfortunately still open, unless Peter
has fixed it per his recent idea.

Zeugswetter Andreas SB SD

21 years ago> Alvaro Herrera <alvherre@atentus.com> writes:
Bruce Momjian [Wed, 9 Oct 2002 16:20:25 +0000 (16:20 +0000)]
> Alvaro Herrera <alvherre@atentus.com> writes:
> > I'm looking at pg_dump/common.c:flagInhAttrs() and suspect that it can
> > be more or less rewritten completely, and probably should to get rigth
> > all the cases mentioned in the past attisinherited discussion.  Is this
> > desirable for 7.3?  It can probably be hacked around and the rewrite
> > kept for 7.4, but I think it will be much simpler after the rewrite.
>
> If it's a bug then it's fair game to fix in 7.3.  But keep in mind that
> pg_dump has to behave at least somewhat sanely when called against older
> servers ... will your rewrite behave reasonably if the server does not
> offer attinhcount values?

Nah.  I don't think it's worth it: I had forgotten that older versions
should be supported.  I just left the code as is and added a
version-specific test.

This patch allows pg_dump to dump correctly local definition of columns.
In particular,

CREATE TABLE p1 (f1 int, f2 int);
CREATE TABLE p2 (f1 int);
CREATE TABLE c () INHERITS (p1, p2);
ALTER TABLE ONLY p1 DROP COLUMN f1;
CREATE TABLE p3 (f1 int);
CREATE TABLE c2 (f1 int) INHERITS (p3);

Will be dumped as
CREATE TABLE p1 (f2 int);
CREATE TABLE p2 (f1 int);
CREATE TABLE c (f1 int) INHERITS (p1, p2);
CREATE TABLE c2 (f1 int) INHERITS (p3);

(Previous version will dump
CREATE TABLE c () INHERITS (p1, p2)
CREATE TABLE c2 () INHERITS (p3) )

Alvaro Herrera

21 years agoHave SET not start transaction when autocommit off, with doc updates.
Bruce Momjian [Wed, 9 Oct 2002 04:59:38 +0000 (04:59 +0000)]
Have SET not start transaction when autocommit off, with doc updates.

21 years agoAdd:
Bruce Momjian [Wed, 9 Oct 2002 02:04:03 +0000 (02:04 +0000)]
Add:

>  o Allow SHOW of non-modifiable variables, like pg_controldata

21 years agoAgain improve MemSet comments.
Bruce Momjian [Tue, 8 Oct 2002 23:12:22 +0000 (23:12 +0000)]
Again improve MemSet comments.

21 years agoUpdate MemSet comments.
Bruce Momjian [Tue, 8 Oct 2002 19:17:58 +0000 (19:17 +0000)]
Update MemSet comments.

21 years agoMove responsibility for setting QuerySnapshot for utility statements
Tom Lane [Tue, 8 Oct 2002 17:17:19 +0000 (17:17 +0000)]
Move responsibility for setting QuerySnapshot for utility statements
into postgres.c; make sure it happens for all cases that seem to need it.
Perhaps it would be better to explicitly exclude just a few utility
statement types from setting a snapshot?

21 years agofixed missing apostrophe
Dave Cramer [Tue, 8 Oct 2002 01:47:55 +0000 (01:47 +0000)]
fixed missing apostrophe

21 years agoChange order of operations during XLogFlush so that we try to include
Tom Lane [Mon, 7 Oct 2002 17:04:30 +0000 (17:04 +0000)]
Change order of operations during XLogFlush so that we try to include
in our write/flush operation any WAL entries that got queued while we
were waiting to get the WALWriteLock.  This improves throughput when
transactions are small enough that several can be committed per WAL
write (ie, per disk revolution).

21 years agoAvoid PQisBusy/PQconsumeInput busy loop in case of PQisBusy returning
Tatsuo Ishii [Mon, 7 Oct 2002 05:10:02 +0000 (05:10 +0000)]
Avoid PQisBusy/PQconsumeInput busy loop in case of PQisBusy returning
false. per Tom Lane's suggestion. See:

Subject: Suggested change to pgbench
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Tatsuo Ishii <t-ishii@sra.co.jp>
Cc: pgsql-patches@postgreSQL.org
Date: Sun, 06 Oct 2002 12:37:27 -0400

for more details.