]> granicus.if.org Git - postgresql/log
postgresql
24 years agoUpdate FAQ.
Bruce Momjian [Thu, 12 Oct 2000 03:12:41 +0000 (03:12 +0000)]
Update FAQ.

24 years agoUpdate TODO list.
Bruce Momjian [Thu, 12 Oct 2000 03:04:44 +0000 (03:04 +0000)]
Update TODO list.

24 years agoUpdate TODO for LIKE/ESCAPE.
Bruce Momjian [Thu, 12 Oct 2000 01:23:37 +0000 (01:23 +0000)]
Update TODO for LIKE/ESCAPE.

24 years agoUpdate TODO list.
Bruce Momjian [Thu, 12 Oct 2000 01:22:39 +0000 (01:22 +0000)]
Update TODO list.

24 years agoFix temp relation handling for indexes, cleanup
Bruce Momjian [Wed, 11 Oct 2000 21:28:19 +0000 (21:28 +0000)]
Fix temp relation handling for indexes, cleanup

24 years agoautoconf
Bruce Momjian [Wed, 11 Oct 2000 19:19:51 +0000 (19:19 +0000)]
autoconf

24 years agoCorrect error in description of AccessExclusiveLock mode
Tom Lane [Wed, 11 Oct 2000 18:29:52 +0000 (18:29 +0000)]
Correct error in description of AccessExclusiveLock mode
(failed to say that it conflicts with AccessShareLock).

24 years agoAdd pooling discussion.
Bruce Momjian [Wed, 11 Oct 2000 18:09:38 +0000 (18:09 +0000)]
Add pooling discussion.

24 years agoOops, back out my unintended changes.
Bruce Momjian [Wed, 11 Oct 2000 17:58:01 +0000 (17:58 +0000)]
Oops, back out my unintended changes.

24 years agoUpdate make_mkid
Bruce Momjian [Wed, 11 Oct 2000 17:55:34 +0000 (17:55 +0000)]
Update make_mkid

24 years agoImprove MVCC discussion.
Tom Lane [Wed, 11 Oct 2000 17:38:36 +0000 (17:38 +0000)]
Improve MVCC discussion.

24 years agoAdded new SQL function setval(seq,val,bool) to restore is_called as well as value
Philip Warner [Wed, 11 Oct 2000 15:31:34 +0000 (15:31 +0000)]
Added new SQL function setval(seq,val,bool) to restore is_called as well as value
(will be used in a future pg_dump).

24 years agoSynced preproc.y with gram.y
Michael Meskes [Wed, 11 Oct 2000 14:56:37 +0000 (14:56 +0000)]
Synced preproc.y with gram.y

24 years agopg_hba.conf mention fix.
Bruce Momjian [Wed, 11 Oct 2000 01:24:01 +0000 (01:24 +0000)]
pg_hba.conf mention fix.

24 years agoMore pg_config updates.
Bruce Momjian [Tue, 10 Oct 2000 22:03:31 +0000 (22:03 +0000)]
More pg_config updates.

24 years agoRename pg-config to pg_config to be consistent with other pg commands.
Bruce Momjian [Tue, 10 Oct 2000 22:01:55 +0000 (22:01 +0000)]
Rename pg-config to pg_config to be consistent with other pg commands.

24 years agoUnify solaris_i386 and solaris_sparc templates. They were almost identical
Peter Eisentraut [Tue, 10 Oct 2000 21:22:29 +0000 (21:22 +0000)]
Unify solaris_i386 and solaris_sparc templates.  They were almost identical
anyway, the rest being due to them not being kept in sync.  Add configure
test for lorder and use it (on Solaris) when found.

24 years agoSome dynloader files missed the template rename a while back.
Peter Eisentraut [Tue, 10 Oct 2000 21:08:04 +0000 (21:08 +0000)]
Some dynloader files missed the template rename a while back.

24 years agoThis is a stupid garden variety bug and I'm not sure why I didn't catch
Bruce Momjian [Tue, 10 Oct 2000 17:13:30 +0000 (17:13 +0000)]
This is a stupid garden variety bug and I'm not sure why I didn't catch
it previously.  The patch included is against fairly current sources, but
it may apply cleanly against 7.0.2 as well.

On Fri, 6 Oct 2000, Vilson farias wrote:

> I found a irregular behavior with constraints.
>
> I can only set a referencial integrity between these tables when there are
> no data, even if there are no change to referential integrity violation.

24 years ago- Be more paranoid when getting views
Philip Warner [Tue, 10 Oct 2000 13:55:28 +0000 (13:55 +0000)]
- Be more paranoid when getting views
- Support for 'isstrict' procedure attribute.
- Disable --blobs and --table
(replaced prior to attempting to fix sequence dump problems)

24 years agoRemove override of CFLAGS.
Bruce Momjian [Tue, 10 Oct 2000 13:04:46 +0000 (13:04 +0000)]
Remove override of CFLAGS.

24 years agoAdd spaces to manual page.
Bruce Momjian [Tue, 10 Oct 2000 04:42:43 +0000 (04:42 +0000)]
Add spaces to manual page.

24 years agoUpdate FAQ.
Bruce Momjian [Mon, 9 Oct 2000 20:21:01 +0000 (20:21 +0000)]
Update FAQ.

24 years agoUpdate TODO list.
Bruce Momjian [Mon, 9 Oct 2000 19:41:27 +0000 (19:41 +0000)]
Update TODO list.

24 years agoBack out Gunnar R|nning jdbc changes.
Bruce Momjian [Mon, 9 Oct 2000 16:48:19 +0000 (16:48 +0000)]
Back out Gunnar R|nning jdbc changes.

24 years agoCosmetic cleanups of Beos port
Peter Eisentraut [Mon, 9 Oct 2000 16:42:55 +0000 (16:42 +0000)]
Cosmetic cleanups of Beos port

24 years agoUpdate FAQ.
Bruce Momjian [Mon, 9 Oct 2000 07:38:53 +0000 (07:38 +0000)]
Update FAQ.

24 years agoUpdate TODO list.
Bruce Momjian [Mon, 9 Oct 2000 07:38:35 +0000 (07:38 +0000)]
Update TODO list.

24 years agoI have compiled and tested PostgreSQL 7.1devel on UnixWare 7.1. During the
Bruce Momjian [Mon, 9 Oct 2000 03:49:42 +0000 (03:49 +0000)]
I have compiled and tested PostgreSQL 7.1devel on UnixWare 7.1.  During the
process, the need for changes to the FAQ_SCO document was uncovered.  The
attach patch file implements thost changes.

Billy G. Allie

24 years agoAppend "/postgresql" to (certain) installation subdirectories when
Peter Eisentraut [Sun, 8 Oct 2000 21:13:27 +0000 (21:13 +0000)]
Append "/postgresql" to (certain) installation subdirectories when
installing into a shared location.  Also Makefile.global organizational
cleanup.

24 years agoinitlocation must set env before postmaster start.
Bruce Momjian [Sun, 8 Oct 2000 19:44:01 +0000 (19:44 +0000)]
initlocation must set env before postmaster start.

24 years agoOkay, I have some new code in place that hopefully should work better. I
Bruce Momjian [Sun, 8 Oct 2000 19:37:56 +0000 (19:37 +0000)]
Okay, I have some new code in place that hopefully should work better. I
couldn't produce a full patch using cvs diff -c this time since I have
created new files and anonymous cvs usage doesn't allow you to
adds. I'm supplying the modified src/interfaces/jdbc as a tarball at :
http://www.candleweb.no/~gunnar/projects/pgsql/postgres-jdbc-2000-10-05.tgz

The new files that should be added are :

? org/postgresql/PGStatement.java
? org/postgresql/ObjectPool.java
? org/postgresql/ObjectPoolFactory.java

There is now a global static pool of free byte arrays and used byte arrays
connected to a statement object. This is the role of the new PGStatement
class. Access to the global free array is synchronized, while we rely on
the PG_Stream synchronization for the used array.

My measurements show that the perfomance boost on this code is not quite as
big as my last shot, but it is still an improvement. Maybe some of the
difference is due to the new synchronization on the global array. I think I
will look into choosing between on a connection level and global level.

I have also started experimented with improving the performance of the
various conversions. The problem here is ofcourse related handle the
various encodings. One thing I found to speed up ResultSet.getInt() a lot
was to do custom conversion on the byte array into int instead of going
through the getString() to do the conversion. But I'm unsure if this is
portable, can we assume that a digit never can be represented by more than
one byte ? It works fine in my iso-latin-8859-1 environment, but what about
other environments ? Maybe we could provide different ResultSet
implementations depending on the encoding used or delegate some methods of
the result set to an "converter class".

Check the org/postgresql/jdbc2/FastResultSet.java in the tarball above to
see the modified getInt() method.

Regards,

        Gunnar

24 years agoautoconf
Bruce Momjian [Sun, 8 Oct 2000 13:52:22 +0000 (13:52 +0000)]
autoconf

24 years agoThis removes the LDFLAGS from the template and adds an autoconf check
Bruce Momjian [Sun, 8 Oct 2000 13:49:47 +0000 (13:49 +0000)]
This removes the LDFLAGS from the template and adds an autoconf check
for the library.  not sure if this will cause problems on other
platforms, but if it does it can be easily fixed.  Also remove the
references to the GeekGadgets includes as the majority of users don't
have them installed and they foul the build process.  We can document
that adding them if you have them installed is a good idea.

David Reid

24 years agoThese aren't used anymore.
Peter Eisentraut [Sun, 8 Oct 2000 13:25:11 +0000 (13:25 +0000)]
These aren't used anymore.

24 years agoSplit out Developer's from Programmer's guide.
Peter Eisentraut [Sun, 8 Oct 2000 13:24:09 +0000 (13:24 +0000)]
Split out Developer's from Programmer's guide.

24 years agomarkup repair
Peter Eisentraut [Sun, 8 Oct 2000 13:22:24 +0000 (13:22 +0000)]
markup repair

24 years agoAdd runtime configuration option "silent_mode".
Tatsuo Ishii [Sun, 8 Oct 2000 09:25:38 +0000 (09:25 +0000)]
Add runtime configuration option "silent_mode".
This is equivalent to postmaster's -S option.

24 years agoTom Lane wrote:
Bruce Momjian [Sun, 8 Oct 2000 04:38:21 +0000 (04:38 +0000)]
Tom Lane wrote:

> > For a while I though it might be because we are using an alpha TAS in
> > the spinlock rather than the old semaphore. I replaced our spinlock
> > with the standard one and it made no difference. We have been running
> > with our spinlock implementation for nearly 2 months on a production
> > database now without a hitch, so I think it is ok. Did I ever submit
> > any patches for the Alpha spinlock?
>
> Not that I recall.  We did get some advice from some Alpha gurus at DEC
> who seemed to think the existing TAS code is OK.  What was it that you
> felt needed to be improved?

The current code uses semaphores, which has the advantage that it works
well even on multi-processor machines, but the disadvantage that it is not
the fastest way possible. Writing a spinlock on Alpha for SMP machines is
very difficult, as you need to deal with memory barriers. A real mess. But
then one of the people at Compaq pointed out to us that there is a
ready-made routine on Alpha. We implemented it with the two patches below.
I ran tests with lots of parallel back-ends and got around a 10% speed
increase. I include the two patches. Perhaps some of the other people
running Tru64 can have a look at these as well.

Cheers,

Adriaan Joubert

24 years agoBack out:
Bruce Momjian [Sun, 8 Oct 2000 03:53:15 +0000 (03:53 +0000)]
Back out:

> this is patch v 0.4 to support transactions with BLOBs.
> All BLOBs are in one table. You need to make initdb.
>
> --
> Sincerely Yours,
> Denis Perchine

24 years agoHello,
Bruce Momjian [Sun, 8 Oct 2000 03:18:57 +0000 (03:18 +0000)]
Hello,

this is patch v 0.4 to support transactions with BLOBs.
All BLOBs are in one table. You need to make initdb.

--
Sincerely Yours,
Denis Perchine

24 years agoRevert rules.out to non-broken state. I will with difficulty hold my
Tom Lane [Sun, 8 Oct 2000 02:51:50 +0000 (02:51 +0000)]
Revert rules.out to non-broken state.  I will with difficulty hold my
tongue about the sloppiness of (a) submitting and (b) committing this
file.

24 years agoRemove S-Tree_indexes.html.gz
Bruce Momjian [Sat, 7 Oct 2000 22:14:09 +0000 (22:14 +0000)]
Remove S-Tree_indexes.html.gz

24 years agoTo create a temporary installation, the regression test driver should use
Peter Eisentraut [Sat, 7 Oct 2000 20:23:03 +0000 (20:23 +0000)]
To create a temporary installation, the regression test driver should use
the make program that was used to build itself.  This seems more correct
than guessing around for some gmake.

24 years ago Thanks for the patch
Bruce Momjian [Sat, 7 Oct 2000 19:48:27 +0000 (19:48 +0000)]
    Thanks for the patch

        It builds and run nicely now

        it seem that the resultmap file have not been updated.

Cyril VELTER

24 years agoRemove stuff that shouldn't be here.
Peter Eisentraut [Sat, 7 Oct 2000 18:59:12 +0000 (18:59 +0000)]
Remove stuff that shouldn't be here.

24 years agoSeparate -Wl,-E or equivalent into separate make variable, since it's only
Peter Eisentraut [Sat, 7 Oct 2000 18:43:23 +0000 (18:43 +0000)]
Separate -Wl,-E or equivalent into separate make variable, since it's only
necessary for the postgres/postmaster link, not every link.

24 years agoautoconf
Bruce Momjian [Sat, 7 Oct 2000 14:57:17 +0000 (14:57 +0000)]
autoconf

24 years agoBeos regression patch.
Bruce Momjian [Sat, 7 Oct 2000 14:55:16 +0000 (14:55 +0000)]
Beos regression patch.

24 years agoautoconf
Bruce Momjian [Sat, 7 Oct 2000 14:40:10 +0000 (14:40 +0000)]
autoconf

24 years ago The beos port in the source tree doesn't even compile. and even
Bruce Momjian [Sat, 7 Oct 2000 14:39:21 +0000 (14:39 +0000)]
    The beos port in the source tree doesn't even compile. and even
after that dynamic loading isn't working and shared memory handling is
broken.

        Attached with this message, there is a Zip file which contain :

        * beos.diff = patch file generated with difforig
        * beos = folder with beos support files which need to be moved in /
src/backend/port
        * expected = foler with three file for message and precision
difference in regression test
        * regression.diff = rule problem (need to kill the backend manualy)
        * dynloader = dynloader files (they are also in the pacth files,
but there is so much modification that I have join full files)

        Everything works except a problem in 'rules' Is there some problems
with rules in the current tree ? It used to works with last week tree.

Cyril VELTER

24 years agoNew beos files. FAQ cleanup.
Bruce Momjian [Sat, 7 Oct 2000 14:16:02 +0000 (14:16 +0000)]
New beos files.  FAQ cleanup.

24 years agoSecond thoughts about where to put DeferredTriggerEndQuery() call.
Tom Lane [Sat, 7 Oct 2000 04:00:41 +0000 (04:00 +0000)]
Second thoughts about where to put DeferredTriggerEndQuery() call.

24 years agoArrange that no database accesses are attempted during parser() --- this
Tom Lane [Sat, 7 Oct 2000 00:58:23 +0000 (00:58 +0000)]
Arrange that no database accesses are attempted during parser() --- this
took some rejiggering of typename and ACL parsing, as well as moving
parse_analyze call out of parser().  Restructure postgres.c processing
so that parse analysis and rewrite are skipped when in abort-transaction
state.  Only COMMIT and ABORT statements will be processed beyond the raw
parser() phase.  This addresses problem of parser failing with database access
errors while in aborted state (see pghackers discussions around 7/28/00).
Also fix some bugs with COMMIT/ABORT statements appearing in the middle of
a single query input string.
Function, operator, and aggregate arguments/results can now use full
TypeName production, in particular foo[] for array types.
DROP OPERATOR and COMMENT ON OPERATOR were broken for unary operators.
Allow CREATE AGGREGATE to accept unquoted numeric constants for initcond.

24 years agoFix silly typo that led to 'tleIsArrayAssign: I'm confused' message.
Tom Lane [Thu, 5 Oct 2000 21:52:08 +0000 (21:52 +0000)]
Fix silly typo that led to 'tleIsArrayAssign: I'm confused' message.
Indeed it was.

24 years agoClean up gcc warning, fix inappropriate choice of configuration symbol.
Tom Lane [Thu, 5 Oct 2000 20:18:33 +0000 (20:18 +0000)]
Clean up gcc warning, fix inappropriate choice of configuration symbol.

24 years agoSuppress gcc warnings.
Tom Lane [Thu, 5 Oct 2000 20:10:20 +0000 (20:10 +0000)]
Suppress gcc warnings.

24 years agoFinal proofreader changes to docs.
Bruce Momjian [Thu, 5 Oct 2000 19:57:23 +0000 (19:57 +0000)]
Final proofreader changes to docs.

24 years agoAdd proofreader's changes to docs.
Bruce Momjian [Thu, 5 Oct 2000 19:48:34 +0000 (19:48 +0000)]
Add proofreader's changes to docs.
Fix misspelling of disbursion to dispersion.

24 years agoReimplementation of UNION/INTERSECT/EXCEPT. INTERSECT/EXCEPT now meet the
Tom Lane [Thu, 5 Oct 2000 19:11:39 +0000 (19:11 +0000)]
Reimplementation of UNION/INTERSECT/EXCEPT.  INTERSECT/EXCEPT now meet the
SQL92 semantics, including support for ALL option.  All three can be used
in subqueries and views.  DISTINCT and ORDER BY work now in views, too.
This rewrite fixes many problems with cross-datatype UNIONs and INSERT/SELECT
where the SELECT yields different datatypes than the INSERT needs.  I did
that by making UNION subqueries and SELECT in INSERT be treated like
subselects-in-FROM, thereby allowing an extra level of targetlist where the
datatype conversions can be inserted safely.
INITDB NEEDED!

24 years agoAllow overriding HSTYLE and PSTYLE from environment
Peter Eisentraut [Thu, 5 Oct 2000 16:57:23 +0000 (16:57 +0000)]
Allow overriding HSTYLE and PSTYLE from environment

24 years agoUpdate FAQ.
Bruce Momjian [Thu, 5 Oct 2000 05:14:58 +0000 (05:14 +0000)]
Update FAQ.

24 years agoUpdate FAQ.
Bruce Momjian [Thu, 5 Oct 2000 04:58:24 +0000 (04:58 +0000)]
Update FAQ.

24 years agoUpdate FAQ.
Bruce Momjian [Thu, 5 Oct 2000 04:57:27 +0000 (04:57 +0000)]
Update FAQ.

24 years agoUpdate FAQ.
Bruce Momjian [Thu, 5 Oct 2000 04:52:18 +0000 (04:52 +0000)]
Update FAQ.

24 years agoUpdate FAQ.
Bruce Momjian [Thu, 5 Oct 2000 04:49:12 +0000 (04:49 +0000)]
Update FAQ.

24 years agoUpdate soundex to new fmgr interface and fix algorithm
Peter Eisentraut [Wed, 4 Oct 2000 19:25:34 +0000 (19:25 +0000)]
Update soundex to new fmgr interface and fix algorithm

24 years agoConvert macaddr documentation to DocBook, update outdated information.
Peter Eisentraut [Wed, 4 Oct 2000 15:47:45 +0000 (15:47 +0000)]
Convert macaddr documentation to DocBook, update outdated information.

24 years agoAdded patch by Cristof for catching yet another NOTICE.
Michael Meskes [Wed, 4 Oct 2000 12:39:45 +0000 (12:39 +0000)]
Added patch by Cristof for catching yet another NOTICE.

24 years agoremoved binary file
Michael Meskes [Wed, 4 Oct 2000 10:11:33 +0000 (10:11 +0000)]
removed binary file

CV: ----------------------------------------------------------------------

24 years agoBtree WAL logging.
Vadim B. Mikheev [Wed, 4 Oct 2000 00:04:43 +0000 (00:04 +0000)]
Btree WAL logging.

24 years agoBetter coding of readline tests
Peter Eisentraut [Tue, 3 Oct 2000 19:50:21 +0000 (19:50 +0000)]
Better coding of readline tests

24 years agoFix to make output format identical on echo -n and echo \c platforms. This
Peter Eisentraut [Tue, 3 Oct 2000 19:37:39 +0000 (19:37 +0000)]
Fix to make output format identical on echo -n and echo \c platforms.  This
had caused the test counts to be wrong on the latter group.

24 years agoTreat empty connection parameters as is, in particular to enable overriding
Peter Eisentraut [Tue, 3 Oct 2000 19:16:17 +0000 (19:16 +0000)]
Treat empty connection parameters as is, in particular to enable overriding
environment variables with "nothing".  Empty host parameter indicates
Unix socket.

24 years agoOne check for sys/un.h ought to be enough. Zero checks for sys/shm.h are not.
Peter Eisentraut [Tue, 3 Oct 2000 13:39:24 +0000 (13:39 +0000)]
One check for sys/un.h ought to be enough.  Zero checks for sys/shm.h are not.

24 years agoCleanup of fe-connect ifdef platform problem added with beos.
Bruce Momjian [Tue, 3 Oct 2000 03:39:46 +0000 (03:39 +0000)]
Cleanup of fe-connect ifdef platform problem added with beos.

24 years agoautoconf
Bruce Momjian [Tue, 3 Oct 2000 03:20:13 +0000 (03:20 +0000)]
autoconf

24 years agoUpdate configure.in
Bruce Momjian [Tue, 3 Oct 2000 03:20:00 +0000 (03:20 +0000)]
Update configure.in

24 years agoautoconf
Bruce Momjian [Tue, 3 Oct 2000 03:15:44 +0000 (03:15 +0000)]
autoconf

24 years agoAdd configure.in change.
Bruce Momjian [Tue, 3 Oct 2000 03:14:53 +0000 (03:14 +0000)]
Add configure.in change.

24 years agoNew diff that now covers the entire tree. Applying this gets postgresql
Bruce Momjian [Tue, 3 Oct 2000 03:11:26 +0000 (03:11 +0000)]
New diff that now covers the entire tree.  Applying this gets postgresql

working on the VERY latest version of BeOS.  I'm sure there will be
alot of comments, but then if there weren't I'd be disappointed!

Thanks for your continuing efforts to get this into your tree.

Haven't bothered with the new files as they haven't changed.

BTW Peter, the compiler is "broken" about the bool define and so on.
I'm filing a bug report to try and get it addressed.  Hopefully then we
can tidy up the code a bit.

I await the replies with interest :)

David Reid

24 years ago"installcheck" doesn't need to depend on "all" since we depend on the user
Peter Eisentraut [Mon, 2 Oct 2000 22:21:21 +0000 (22:21 +0000)]
"installcheck" doesn't need to depend on "all" since we depend on the user
to start up a postmaster anyway.

24 years agoMop up from caddr_t -> Datum conversion to make things extra type safe
Peter Eisentraut [Mon, 2 Oct 2000 21:45:33 +0000 (21:45 +0000)]
Mop up from caddr_t -> Datum conversion to make things extra type safe

24 years agoBanish caddr_t (mostly), use Datum where appropriate.
Peter Eisentraut [Mon, 2 Oct 2000 19:42:56 +0000 (19:42 +0000)]
Banish caddr_t (mostly), use Datum where appropriate.

24 years agoUpdate TODO list.
Bruce Momjian [Mon, 2 Oct 2000 17:36:56 +0000 (17:36 +0000)]
Update TODO list.

24 years agoautoconf
Bruce Momjian [Mon, 2 Oct 2000 17:18:50 +0000 (17:18 +0000)]
autoconf

24 years agoHere is the first batch of files and diffs for the BeOS port. I've run into
Bruce Momjian [Mon, 2 Oct 2000 17:16:01 +0000 (17:16 +0000)]
Here is the first batch of files and diffs for the BeOS port.  I've run into
problems with some bits of it, but when all the patches are in it'll build
and we can fix it from there :)  I've got a version that builds and runs and
that is the basis for these patches.

The first file has the new additional files that are required,
    template/beos
    backend/port/dynloader/beos.c
    backend/port/dynloader/beos.h
    include/port/beos.h
    makefiles/Makefile.beos

The second is a tarball of diffs against a few files.  I've added sys/ipc.h
to configure and config.h via configure.in and config.h.in and then started
adding the check as this file isn't needed on BeOS and having loads of
#ifdef BEOS isn't as obvious as #ifdef HAVE_SYS_IPC_H and isn't as
autconf'ish :)
Files touched are
    include/c.h
    configure.in
    include/config.h.in
    include/storage/ipc.h
    include/utils/int8.h

Let me know how these go.  I'll await a response before submitting any more.

Any problems just get in touch.

David Reid

24 years agoIn the meaning of bug-fix, the patch is not needed. Because you
Bruce Momjian [Mon, 2 Oct 2000 16:15:53 +0000 (16:15 +0000)]
In the meaning of bug-fix, the patch is not needed. Because you
have already modified "next_insert()" in 7.0-ecpglib.  However
in the meaning of speed-up, the patch will be needed.
--
Regards,
SAKAIDA Masaaki  -- Osaka, Japan

24 years agoCommunicate PGPORT and PGHOST via environment, not command line
Peter Eisentraut [Mon, 2 Oct 2000 11:47:30 +0000 (11:47 +0000)]
Communicate PGPORT and PGHOST via environment, not command line

24 years agoHonor $(SHELL) when invoking pg_regress.
Tom Lane [Mon, 2 Oct 2000 06:07:25 +0000 (06:07 +0000)]
Honor $(SHELL) when invoking pg_regress.

24 years agoFix silly typo that prevented resultmap from being honored by
Tom Lane [Mon, 2 Oct 2000 06:03:25 +0000 (06:03 +0000)]
Fix silly typo that prevented resultmap from being honored by
pg_regress.

24 years agoMake default ACL be consistent --- ie, starting point for ChangeAcl
Tom Lane [Mon, 2 Oct 2000 04:49:28 +0000 (04:49 +0000)]
Make default ACL be consistent --- ie, starting point for ChangeAcl
is the same as the access permissions granted when a relation's relacl
field is NULL, ie, owner=all rights, world=no rights.

24 years agoautoconf
Bruce Momjian [Mon, 2 Oct 2000 03:59:36 +0000 (03:59 +0000)]
autoconf

24 years agoFix configure types.h
Bruce Momjian [Mon, 2 Oct 2000 03:58:31 +0000 (03:58 +0000)]
Fix configure types.h

24 years agoAdd sys/types include for sockaddr test to configure
Bruce Momjian [Mon, 2 Oct 2000 03:55:15 +0000 (03:55 +0000)]
Add sys/types include for sockaddr test to configure

24 years agoAdd missing python files.
Bruce Momjian [Mon, 2 Oct 2000 03:46:24 +0000 (03:46 +0000)]
Add missing python files.

24 years agoUpdate for PyGreSQL 3.0, from D'Arcy J.M. Cain
Bruce Momjian [Mon, 2 Oct 2000 03:27:33 +0000 (03:27 +0000)]
Update for PyGreSQL 3.0, from D'Arcy J.M. Cain

24 years agoFix quoting bug
Peter Eisentraut [Sun, 1 Oct 2000 22:23:14 +0000 (22:23 +0000)]
Fix quoting bug

24 years agoFix test order
Peter Eisentraut [Sun, 1 Oct 2000 21:19:09 +0000 (21:19 +0000)]
Fix test order

24 years agoFix nasty TRUNCATE bug reported by Darrin Ladd. RelationTruncateIndexes
Tom Lane [Sat, 30 Sep 2000 18:28:53 +0000 (18:28 +0000)]
Fix nasty TRUNCATE bug reported by Darrin Ladd.  RelationTruncateIndexes
would close and then re-open rel being truncated.  Depending on the
luck of the draw, the re-opened relcache entry might or might not be
at the same physical location as before.  Unfortunately, if it wasn't
then heap_truncate would crash and burn, because it still had a pointer
at the old location.  Fix is to open and then close rel in
RelationTruncateIndexes, so that rel's refcount never goes to zero
until heap_truncate is done.