]> granicus.if.org Git - postgresql/log
postgresql
21 years agoApplied patch from kho@redhat.com to fix a problem with trying to use a fetch
Barry Lind [Wed, 6 Aug 2003 05:53:13 +0000 (05:53 +0000)]
Applied patch from kho@redhat.com to fix a problem with trying to use a fetch
when a cursor wasn't being used.

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

21 years agoAdd ecpg thread testing file.
Bruce Momjian [Wed, 6 Aug 2003 02:19:51 +0000 (02:19 +0000)]
Add ecpg thread testing file.

21 years agoImprove documentation of ParseDateTime(). Reorder tests to prevent
Tom Lane [Tue, 5 Aug 2003 18:30:21 +0000 (18:30 +0000)]
Improve documentation of ParseDateTime().  Reorder tests to prevent
writing one more value into return arrays than will fit.  This is
potentially a stack smash, though I do not think it is a problem in
current uses of the routine, since a failure return causes elog anyway.

21 years agoFix several places where fractional-second inputs were misprocessed
Tom Lane [Tue, 5 Aug 2003 17:39:19 +0000 (17:39 +0000)]
Fix several places where fractional-second inputs were misprocessed
in HAVE_INT64_TIMESTAMP cases, including two potential stack smashes
when more than six fractional digits were supplied.  Per bug report
from Philipp Reisner.

21 years agoModify:
Bruce Momjian [Tue, 5 Aug 2003 15:31:33 +0000 (15:31 +0000)]
Modify:

Prevent interval from supressing ':00' seconds display

21 years agoRemove Tom mention on array items.
Bruce Momjian [Tue, 5 Aug 2003 15:04:16 +0000 (15:04 +0000)]
Remove Tom mention on array items.

21 years agoAdd mention of :00 seconds.
Bruce Momjian [Tue, 5 Aug 2003 14:37:49 +0000 (14:37 +0000)]
Add mention of :00 seconds.

21 years agoUpdate thread wording.
Bruce Momjian [Tue, 5 Aug 2003 14:35:12 +0000 (14:35 +0000)]
Update thread wording.

21 years agoModify tsearch2 to be Oleg,Teodor
Bruce Momjian [Tue, 5 Aug 2003 14:33:30 +0000 (14:33 +0000)]
Modify tsearch2 to be Oleg,Teodor

21 years agoAdd tsearch2 mention.
Bruce Momjian [Tue, 5 Aug 2003 05:07:50 +0000 (05:07 +0000)]
Add tsearch2 mention.

21 years agoAdd Philip Yarra to threads item, remove my name.
Bruce Momjian [Tue, 5 Aug 2003 05:04:50 +0000 (05:04 +0000)]
Add Philip Yarra to threads item, remove my name.

21 years agoUpdate release items from Joe Conway.
Bruce Momjian [Tue, 5 Aug 2003 05:01:50 +0000 (05:01 +0000)]
Update release items from Joe Conway.

21 years agocan't mix and match .gz and .bz2 in here ... won't build REL7_4_BETA1
PostgreSQL Daemon [Tue, 5 Aug 2003 04:55:58 +0000 (04:55 +0000)]
can't mix and match .gz and .bz2 in here ... won't build

21 years agoseeing if building bz2 distributions actually works ...
PostgreSQL Daemon [Tue, 5 Aug 2003 04:40:29 +0000 (04:40 +0000)]
seeing if building bz2 distributions actually works ...

21 years agoremove src/data from target list
PostgreSQL Daemon [Tue, 5 Aug 2003 04:18:51 +0000 (04:18 +0000)]
remove src/data from target list

21 years agoRemove MOVE ALL item.
Bruce Momjian [Tue, 5 Aug 2003 03:30:35 +0000 (03:30 +0000)]
Remove MOVE ALL item.

21 years agoFix some copyright notices that weren't updated. Improve copyright tool
Tom Lane [Mon, 4 Aug 2003 23:59:41 +0000 (23:59 +0000)]
Fix some copyright notices that weren't updated.  Improve copyright tool
so it won't miss 'em again.

21 years agochange tag to 7.4beta1 and update the Copyright to 2003
PostgreSQL Daemon [Mon, 4 Aug 2003 22:30:30 +0000 (22:30 +0000)]
change tag to 7.4beta1 and update the Copyright to 2003

Guess what folks?  We are now in Beta!!

21 years agoUpdate vacuumlo to be properly schema-aware. Improve documentation.
Tom Lane [Mon, 4 Aug 2003 22:03:39 +0000 (22:03 +0000)]
Update vacuumlo to be properly schema-aware.  Improve documentation.

21 years agoRe-add USE_THREADS, used by ecpg.
Bruce Momjian [Mon, 4 Aug 2003 21:26:26 +0000 (21:26 +0000)]
Re-add USE_THREADS, used by ecpg.

21 years agoFix erroneous direct reference to postgres.tar.gz, per bug report
Tom Lane [Mon, 4 Aug 2003 20:52:48 +0000 (20:52 +0000)]
Fix erroneous direct reference to postgres.tar.gz, per bug report
from Garrick Staples.

21 years agoSub-Makefiles need to explicitly add CFLAGS_SL to CFLAGS, else their
Tom Lane [Mon, 4 Aug 2003 20:34:26 +0000 (20:34 +0000)]
Sub-Makefiles need to explicitly add CFLAGS_SL to CFLAGS, else their
object files do not get built with -fpic.

21 years agoFix to build correctly outside source tree.
Tom Lane [Mon, 4 Aug 2003 19:52:37 +0000 (19:52 +0000)]
Fix to build correctly outside source tree.

21 years agoMust print server's failure message before trying reconnect, not after.
Tom Lane [Mon, 4 Aug 2003 19:10:40 +0000 (19:10 +0000)]
Must print server's failure message before trying reconnect, not after.

21 years agoFix pltcl and plpython to support STATEMENT triggers.
Tom Lane [Mon, 4 Aug 2003 18:40:50 +0000 (18:40 +0000)]
Fix pltcl and plpython to support STATEMENT triggers.

Joe Conway

21 years agoSSL_read/SSL_write do not approximate the return conventions of recv()
Tom Lane [Mon, 4 Aug 2003 17:58:14 +0000 (17:58 +0000)]
SSL_read/SSL_write do not approximate the return conventions of recv()
and send() very well at all; and in any case we can't use retval==0
for EOF due to race conditions.  Make the same fixes in the backend as
are required in libpq.

21 years agoFix some more problems with testing error returns from SSL.
Tom Lane [Mon, 4 Aug 2003 17:25:14 +0000 (17:25 +0000)]
Fix some more problems with testing error returns from SSL.

21 years agoFix thread handling in configure.
Bruce Momjian [Mon, 4 Aug 2003 16:48:03 +0000 (16:48 +0000)]
Fix thread handling in configure.

21 years agoMinor cleanups in S_LOCK_TEST code.
Tom Lane [Mon, 4 Aug 2003 15:28:33 +0000 (15:28 +0000)]
Minor cleanups in S_LOCK_TEST code.

21 years agoHPUX's horology behavior is OS-dependent, not hardware-dependent.
Tom Lane [Mon, 4 Aug 2003 15:06:45 +0000 (15:06 +0000)]
HPUX's horology behavior is OS-dependent, not hardware-dependent.

21 years agomake sub-Makefiles in the sub-directories
Teodor Sigaev [Mon, 4 Aug 2003 14:54:47 +0000 (14:54 +0000)]
make sub-Makefiles in the sub-directories

21 years agoDocs fixes
Teodor Sigaev [Mon, 4 Aug 2003 14:11:08 +0000 (14:11 +0000)]
Docs fixes

21 years agoTranslation updates
Peter Eisentraut [Mon, 4 Aug 2003 14:01:37 +0000 (14:01 +0000)]
Translation updates

21 years agoAdd a note that AND and OR are commutative. Apparently, they are not in
Peter Eisentraut [Mon, 4 Aug 2003 14:00:14 +0000 (14:00 +0000)]
Add a note that AND and OR are commutative.  Apparently, they are not in
certain other products.

21 years agoRemove --enable-recode feature, since it's been broken by IPv6 changes,
Tom Lane [Mon, 4 Aug 2003 04:03:10 +0000 (04:03 +0000)]
Remove --enable-recode feature, since it's been broken by IPv6 changes,
and seems to have too few users to justify maintaining.

21 years agoUpdate copyrights to 2003.
Bruce Momjian [Mon, 4 Aug 2003 02:40:20 +0000 (02:40 +0000)]
Update copyrights to 2003.

21 years agoFix for 2003 again.
Bruce Momjian [Mon, 4 Aug 2003 02:27:25 +0000 (02:27 +0000)]
Fix for 2003 again.

21 years agoUpdate copyright script for 2003.
Bruce Momjian [Mon, 4 Aug 2003 02:22:37 +0000 (02:22 +0000)]
Update copyright script for 2003.

21 years agoPut back braces removed by pgindent (not really pgindent's fault).
Tom Lane [Mon, 4 Aug 2003 01:57:58 +0000 (01:57 +0000)]
Put back braces removed by pgindent (not really pgindent's fault).

21 years agopgindent run.
Bruce Momjian [Mon, 4 Aug 2003 00:43:34 +0000 (00:43 +0000)]
pgindent run.

21 years agoRemove extra paren in NOT_USED code, found by pgindent.
Bruce Momjian [Mon, 4 Aug 2003 00:26:49 +0000 (00:26 +0000)]
Remove extra paren in NOT_USED code, found by pgindent.

21 years agoTighten inline_function's test for overly complex parameters. This
Tom Lane [Sun, 3 Aug 2003 23:46:37 +0000 (23:46 +0000)]
Tighten inline_function's test for overly complex parameters.  This
should catch most situations where repeated inlining blows up the
expression complexity unreasonably, as in Joe Conway's recent example.

21 years agoReconsider context for calling callback functions --- original idea
Tom Lane [Sun, 3 Aug 2003 23:44:44 +0000 (23:44 +0000)]
Reconsider context for calling callback functions --- original idea
that they aren't part of error processing is clearly faulty.

21 years agoUpdate comments.
Bruce Momjian [Sun, 3 Aug 2003 23:27:24 +0000 (23:27 +0000)]
Update comments.

21 years agoUpdate HISTORY file for 7.4.
Bruce Momjian [Sun, 3 Aug 2003 23:26:05 +0000 (23:26 +0000)]
Update HISTORY file for 7.4.

21 years agoDon't pull in libbind except on BeOS, per pgsql-bugs discussion around
Tom Lane [Sun, 3 Aug 2003 20:23:14 +0000 (20:23 +0000)]
Don't pull in libbind except on BeOS, per pgsql-bugs discussion around
12 Jun 2003.

21 years agoRemove docs that say psql \encoding doesn't track SET client_encoding.
Bruce Momjian [Sat, 2 Aug 2003 02:44:38 +0000 (02:44 +0000)]
Remove docs that say psql \encoding doesn't track SET client_encoding.

21 years agoComment out entries for IPv6 localhost connections, since they fail
Tom Lane [Fri, 1 Aug 2003 23:40:10 +0000 (23:40 +0000)]
Comment out entries for IPv6 localhost connections, since they fail
on machines with no IPv6 support.

21 years agoPostmaster erroneously rejected SSL connections on IPv6.
Tom Lane [Fri, 1 Aug 2003 23:25:00 +0000 (23:25 +0000)]
Postmaster erroneously rejected SSL connections on IPv6.

21 years agoFix some unprotected references to AF_UNIX ... wouldn't compile on
Tom Lane [Fri, 1 Aug 2003 23:24:28 +0000 (23:24 +0000)]
Fix some unprotected references to AF_UNIX ... wouldn't compile on
platforms without AF_UNIX sockets.

21 years agoinet_recv() wasn't IPv6-ready.
Tom Lane [Fri, 1 Aug 2003 23:22:52 +0000 (23:22 +0000)]
inet_recv() wasn't IPv6-ready.

21 years agoUpdate, add Greg:
Bruce Momjian [Fri, 1 Aug 2003 22:18:15 +0000 (22:18 +0000)]
Update, add Greg:

< * Fernando Nasser <fnasser@redhat.com> of Red Hat
< * Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
> * Fernando is Fernando Nasser <fnasser@redhat.com> of Red Hat
> * Gavin is Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
> * Greg is Greg Sabino Mullane <greg@turnstep.com>

21 years agoAdd:
Bruce Momjian [Fri, 1 Aug 2003 21:44:05 +0000 (21:44 +0000)]
Add:

<
>  o Allow PL/pgSQL EXECUTE query_var INTO record_var;

21 years agoCode review for sslmode patch: eliminate memory leak, avoid giving a
Tom Lane [Fri, 1 Aug 2003 21:27:27 +0000 (21:27 +0000)]
Code review for sslmode patch: eliminate memory leak, avoid giving a
completely useless error message in 'allow' case, don't retry connection
at the sendauth stage (by then the server will either let us in or not,
no point in wasting cycles on another try in the other SSL state).

21 years agoSince HPUX now exists for Itanium, we should decouple the assumption
Tom Lane [Fri, 1 Aug 2003 19:12:52 +0000 (19:12 +0000)]
Since HPUX now exists for Itanium, we should decouple the assumption
that OS=hpux is the same as CPU=hppa.  First steps at doing this.
With these patches, we still work on hppa with either gcc or HP's cc.
We might work on hpux/itanium with gcc, but I can't test it.  Definitely
will not work on hpux/itanium with non-gcc compiler, for lack of spinlock
code.

21 years agoGet the include ordering right for TCL includes vs whatever is coming
Tom Lane [Fri, 1 Aug 2003 19:01:49 +0000 (19:01 +0000)]
Get the include ordering right for TCL includes vs whatever is coming
from --with-includes.

21 years agoMissed two places to replace union member.
Michael Meskes [Fri, 1 Aug 2003 18:19:03 +0000 (18:19 +0000)]
Missed two places to replace union member.

21 years agoFix a few of the more blatantly unportable constructs in this file.
Tom Lane [Fri, 1 Aug 2003 18:10:43 +0000 (18:10 +0000)]
Fix a few of the more blatantly unportable constructs in this file.

21 years agoFix inconsistent static-vs-not-static declarations.
Tom Lane [Fri, 1 Aug 2003 18:03:57 +0000 (18:03 +0000)]
Fix inconsistent static-vs-not-static declarations.

21 years agoFix compile warning.
Tom Lane [Fri, 1 Aug 2003 16:46:06 +0000 (16:46 +0000)]
Fix compile warning.

21 years agoRussian translation updates by Serguei Mokhov
Peter Eisentraut [Fri, 1 Aug 2003 16:19:14 +0000 (16:19 +0000)]
Russian translation updates by Serguei Mokhov

21 years agoUse only two-part shared library version numbers, for better portability
Peter Eisentraut [Fri, 1 Aug 2003 16:18:04 +0000 (16:18 +0000)]
Use only two-part shared library version numbers, for better portability
and consistency.

21 years agoWhile having a parallel-make-safe genbki.sh is good, it's better not to
Peter Eisentraut [Fri, 1 Aug 2003 16:12:32 +0000 (16:12 +0000)]
While having a parallel-make-safe genbki.sh is good, it's better not to
uselessly invoke it in parallel in the first place.

21 years agoMake ecpg SQLSTATE-aware. Map existing SQLCODE assignments to SQLSTATEs,
Peter Eisentraut [Fri, 1 Aug 2003 13:53:36 +0000 (13:53 +0000)]
Make ecpg SQLSTATE-aware.  Map existing SQLCODE assignments to SQLSTATEs,
rather than parsing the message.  Add some documentation about embedded
SQL.

21 years agoUpdate comments.
Bruce Momjian [Fri, 1 Aug 2003 13:48:58 +0000 (13:48 +0000)]
Update comments.

21 years agoUpdate comments.
Bruce Momjian [Fri, 1 Aug 2003 13:48:25 +0000 (13:48 +0000)]
Update comments.

21 years agoAdded missing TO keyword.
Michael Meskes [Fri, 1 Aug 2003 11:25:55 +0000 (11:25 +0000)]
Added missing TO keyword.

21 years ago- Added some Informix error codes in Informix mode.
Michael Meskes [Fri, 1 Aug 2003 08:21:04 +0000 (08:21 +0000)]
- Added some Informix error codes in Informix mode.
- Added just another pgtypeslib function.

21 years agocvs log used -b for head, not -rHEAD.
Bruce Momjian [Fri, 1 Aug 2003 05:52:44 +0000 (05:52 +0000)]
cvs log used -b for head, not -rHEAD.

21 years agopython needs to be removed from here
Marc G. Fournier [Fri, 1 Aug 2003 04:28:13 +0000 (04:28 +0000)]
python needs to be removed from here

21 years agoclean python out of Makefile
Marc G. Fournier [Fri, 1 Aug 2003 04:24:57 +0000 (04:24 +0000)]
clean python out of Makefile

21 years agoremove python module, as its moved to http://www.pygresql.org
Marc G. Fournier [Fri, 1 Aug 2003 04:19:06 +0000 (04:19 +0000)]
remove python module, as its moved to http://www.pygresql.org

21 years agoFix for Alpha portability.
Tom Lane [Fri, 1 Aug 2003 03:55:05 +0000 (03:55 +0000)]
Fix for Alpha portability.

21 years ago--enable-thread-safeness -> --enable-thread-safety
Bruce Momjian [Fri, 1 Aug 2003 03:10:04 +0000 (03:10 +0000)]
--enable-thread-safeness -> --enable-thread-safety

21 years agoUpdate contrib regression tests for recent error message editing.
Tom Lane [Fri, 1 Aug 2003 02:38:09 +0000 (02:38 +0000)]
Update contrib regression tests for recent error message editing.

21 years agoAdd missing #include.
Tom Lane [Fri, 1 Aug 2003 02:25:31 +0000 (02:25 +0000)]
Add missing #include.

21 years agoFix some minor portability issues, per Chris K-L.
Tom Lane [Fri, 1 Aug 2003 02:21:17 +0000 (02:21 +0000)]
Fix some minor portability issues, per Chris K-L.

21 years agoFix my own mistake in GUC variable annotation.
Tom Lane [Fri, 1 Aug 2003 01:23:11 +0000 (01:23 +0000)]
Fix my own mistake in GUC variable annotation.

21 years agoImprove description of restoring pg_dumpall backups, per gripe from
Tom Lane [Fri, 1 Aug 2003 01:01:52 +0000 (01:01 +0000)]
Improve description of restoring pg_dumpall backups, per gripe from
Havard Eidnes.

21 years agoAdjust 'permission denied' messages to be more useful and consistent.
Tom Lane [Fri, 1 Aug 2003 00:15:26 +0000 (00:15 +0000)]
Adjust 'permission denied' messages to be more useful and consistent.

21 years agoVery minor editing.
Tom Lane [Thu, 31 Jul 2003 22:17:11 +0000 (22:17 +0000)]
Very minor editing.

21 years agoFix mistaken error code assignments, per Peter's note 7/28/03.
Tom Lane [Thu, 31 Jul 2003 21:34:18 +0000 (21:34 +0000)]
Fix mistaken error code assignments, per Peter's note 7/28/03.

21 years agoWhen using a temp installation, unset PGUSER and other variables we
Tom Lane [Thu, 31 Jul 2003 19:20:41 +0000 (19:20 +0000)]
When using a temp installation, unset PGUSER and other variables we
might have inherited from the environment that would possibly cause
psql to fail to connect to the temp installation properly.  Per trouble
report from Markus Bertheau 7/1/03.

21 years agoCause library-preload feature to report error if specified initialization
Tom Lane [Thu, 31 Jul 2003 18:36:46 +0000 (18:36 +0000)]
Cause library-preload feature to report error if specified initialization
function is not found.  Also, make all the PL libraries have initialization
functions with standard names.  Patch from Joe Conway.

21 years agoUpgrade parsing code for ACLs to be less hokey and more cognizant of
Tom Lane [Thu, 31 Jul 2003 17:21:57 +0000 (17:21 +0000)]
Upgrade parsing code for ACLs to be less hokey and more cognizant of
the actual logical structure and quoting rules being used.  Fixes bug
reported by Chris K-L on 7/8/03.

21 years agoModify:
Bruce Momjian [Thu, 31 Jul 2003 15:07:17 +0000 (15:07 +0000)]
Modify:

> * Implement dirty reads or shared row locks and use them in RI triggers

21 years agoModify:
Bruce Momjian [Thu, 31 Jul 2003 04:49:38 +0000 (04:49 +0000)]
Modify:

> * Implement dirty reads or shared locks and use them in RI triggers

21 years agocontraints -> constraints
Bruce Momjian [Thu, 31 Jul 2003 04:48:15 +0000 (04:48 +0000)]
contraints -> constraints

Rod Taylor

21 years agoUpdate Turkish FAQ, Devrim GUNDUZ
Bruce Momjian [Thu, 31 Jul 2003 04:41:56 +0000 (04:41 +0000)]
Update Turkish FAQ, Devrim GUNDUZ

21 years agoAdd fflush of stdout when outputing query.
Bruce Momjian [Thu, 31 Jul 2003 04:23:40 +0000 (04:23 +0000)]
Add fflush of stdout when outputing query.

21 years ago dbf2pg - Insert xBase-style .dbf-files into a PostgreSQL-table
Bruce Momjian [Thu, 31 Jul 2003 02:12:43 +0000 (02:12 +0000)]
  dbf2pg  -  Insert  xBase-style  .dbf-files  into  a PostgreSQL-table
  There is an option "-s oldname=newname", which changes the old field name of
  the dbf-file to the newname in PostgeSQL. If the length of the new name is 0,
  the field is skiped. If you want to skip the first field of the dbf-file,
  you get the wildest error-messages from the backend.
  dbf2pg load the dbf-file via "COPY tablename FROM STDIN". If you skip the
  first field, it is an \t to much in STDIN.

A fix could be an counter j=0, which increments only, if a field is imported
(IF (strlen(fields[h].db_name)> 0) j++. And only if j > 1 (if an other field is
imported) the \t is printed.

An other small bug in the README:
-s start
              Specify  the  first record-number in the xBase-file
              we will insert.
should be
-e start
              Specify  the  first record-number in the xBase-file
              we will insert.

Thomas Behr

21 years agoAdd pretty-printing variants of pg_get_viewdef and related functions.
Tom Lane [Wed, 30 Jul 2003 22:56:24 +0000 (22:56 +0000)]
Add pretty-printing variants of pg_get_viewdef and related functions.
Patch from Andreas Pflug.

21 years agoAdd:;
Bruce Momjian [Wed, 30 Jul 2003 20:23:32 +0000 (20:23 +0000)]
Add:;
> * Allow server log information to be output as INSERT statements

21 years agoFix numeric_smaller, numeric_larger, float4smaller, float4larger,
Tom Lane [Wed, 30 Jul 2003 19:48:41 +0000 (19:48 +0000)]
Fix numeric_smaller, numeric_larger, float4smaller, float4larger,
float8smaller, float8larger (and thereby the MIN/MAX aggregates on these
datatypes) to agree with the datatypes' comparison operations as
regards NaN handling.  In all these datatypes, NaN is arbitrarily
considered larger than any normal value ... but MIN/MAX had not gotten
the word.  Per recent discussion on pgsql-sql.

21 years agoModify:
Bruce Momjian [Wed, 30 Jul 2003 19:16:18 +0000 (19:16 +0000)]
Modify:
> * Allow log lines to include session-level information, like database and user

21 years agoCause ARRAY[] construct to return a NULL array, rather than raising an
Tom Lane [Wed, 30 Jul 2003 19:02:18 +0000 (19:02 +0000)]
Cause ARRAY[] construct to return a NULL array, rather than raising an
error, if any input element is NULL.  This is not what we ultimately want,
but until arrays can have NULL elements, it will have to do.  Patch from
Joe Conway.

21 years agoAdd:
Bruce Momjian [Wed, 30 Jul 2003 18:52:06 +0000 (18:52 +0000)]
Add:

> * Add database name and user name to logging options

21 years agoWhen shutting down the regression test postmaster after 'make check',
Tom Lane [Wed, 30 Jul 2003 17:08:47 +0000 (17:08 +0000)]
When shutting down the regression test postmaster after 'make check',
wait for the postmaster to actually exit.  Otherwise running repeated
'make check's tends to misbehave, because we try to remove and recreate
the data directory while the old PM is still alive.

21 years agoFix longstanding error in _bt_search(): should moveright at top of loop not
Tom Lane [Tue, 29 Jul 2003 22:18:38 +0000 (22:18 +0000)]
Fix longstanding error in _bt_search(): should moveright at top of loop not
bottom.  Otherwise we fail to moveright when the root page was split while
we were "in flight" to it.  This is not a significant problem when the root
is above the leaf level, but if the root was also a leaf (ie, a single-page
index just got split) we may return the wrong leaf page to the caller,
resulting in failure to find a key that is in fact present.  Bug has existed
at least since 7.1, probably forever.

21 years agoCoerce unknown-literal-constant default values to the column type during
Tom Lane [Tue, 29 Jul 2003 17:21:27 +0000 (17:21 +0000)]
Coerce unknown-literal-constant default values to the column type during
CREATE TABLE (or ALTER TABLE SET DEFAULT), rather than postponing it to
the time that the default is inserted into an INSERT command by the
rewriter.  This reverses an old decision that was intended to make the
world safe for writing
f1 timestamp default 'now'
but in fact merely made the failure modes subtle rather than obvious.
Per recent trouble report and followup discussion.

initdb forced since there is a chance that stored default expressions
will change.