]> granicus.if.org Git - postgresql/log
postgresql
24 years agoEnhance multibyte support.
Tatsuo Ishii [Sat, 20 May 2000 13:12:26 +0000 (13:12 +0000)]
Enhance multibyte support.
SJIS UDC (NEC selection IBM kanji) support contributed by Eiji Tokuya

24 years agoAdd KEEPALIVE option to the socket of backend. This will automatically
Tatsuo Ishii [Sat, 20 May 2000 13:10:54 +0000 (13:10 +0000)]
Add KEEPALIVE option to the socket of backend. This will automatically
terminate the backend that has no frontend anymore.

24 years agoUpdate rules man pages.
Bruce Momjian [Sat, 20 May 2000 11:38:25 +0000 (11:38 +0000)]
Update rules man pages.

24 years agoClean up sql functions examples.
Bruce Momjian [Sat, 20 May 2000 11:24:37 +0000 (11:24 +0000)]
Clean up sql functions examples.

24 years agoUpdate TODO list.
Bruce Momjian [Sat, 20 May 2000 03:45:20 +0000 (03:45 +0000)]
Update TODO list.

24 years agopg_dump barfs on negative values for index column numbers --- like, say,
Tom Lane [Fri, 19 May 2000 23:00:00 +0000 (23:00 +0000)]
pg_dump barfs on negative values for index column numbers --- like, say,
an index on a table's OID column.  Mea maxima culpa ... but how'd we get
through beta with no one noticing this?

24 years agoUpdate TODO list.
Bruce Momjian [Fri, 19 May 2000 19:27:56 +0000 (19:27 +0000)]
Update TODO list.

24 years agoRevise FlushRelationBuffers/ReleaseRelationBuffers per discussion with
Tom Lane [Fri, 19 May 2000 03:22:31 +0000 (03:22 +0000)]
Revise FlushRelationBuffers/ReleaseRelationBuffers per discussion with
Hiroshi.  ReleaseRelationBuffers now removes rel's buffers from pool,
instead of merely marking them nondirty.  The old code would leave valid
buffers for a deleted relation, which didn't cause any known problems
but can't possibly be a good idea.  There were several places which called
ReleaseRelationBuffers *and* FlushRelationBuffers, which is now
unnecessary; but there were others that did not.  FlushRelationBuffers
no longer emits a warning notice if it finds dirty buffers to flush,
because with the current bufmgr behavior that's not an unexpected
condition.  Also, FlushRelationBuffers will flush out all dirty buffers
for the relation regardless of block number.  This ensures that
pg_upgrade's expectations are met about tuple on-row status bits being
up-to-date on disk.  Lastly, tweak BufTableDelete() to clear the
buffer's tag so that no one can mistake it for being a still-valid
buffer for the page it once held.  Formerly, the buffer would not be
found by buffer hashtable searches after BufTableDelete(), but it would
still be thought to belong to its old relation by the routines that
sequentially scan the shared-buffer array.  Again I know of no bugs
caused by that, but it still can't be a good idea.

24 years agoCorrect busted syntax of CC invocation for tas.s.
Tom Lane [Fri, 19 May 2000 02:50:32 +0000 (02:50 +0000)]
Correct busted syntax of CC invocation for tas.s.

24 years agoUnset PGHOST
Bruce Momjian [Thu, 18 May 2000 16:11:39 +0000 (16:11 +0000)]
Unset PGHOST

24 years agoChange remove function warning for built-ins.
Bruce Momjian [Thu, 18 May 2000 16:09:38 +0000 (16:09 +0000)]
Change remove function warning for built-ins.

24 years agoUse postmaster -i only on certain platforms.
Bruce Momjian [Thu, 18 May 2000 15:58:43 +0000 (15:58 +0000)]
Use postmaster -i only on certain platforms.

24 years agoThrow NOTICE on built-in function removal
Bruce Momjian [Thu, 18 May 2000 15:45:00 +0000 (15:45 +0000)]
Throw NOTICE on built-in function removal

24 years agoAdd QNX localhost flag to regression tests
Bruce Momjian [Thu, 18 May 2000 15:37:56 +0000 (15:37 +0000)]
Add QNX localhost flag to regression tests

24 years agoFix WARN->NOTICE in docs. Change libpgeasy connection parameters to use
Bruce Momjian [Thu, 18 May 2000 14:24:38 +0000 (14:24 +0000)]
Fix WARN->NOTICE in docs.  Change libpgeasy connection parameters to use
PQconnectdb() style connections.

24 years agoReduce COPY IN lock from AccessExclusive to a more reasonable
Tom Lane [Thu, 18 May 2000 01:52:45 +0000 (01:52 +0000)]
Reduce COPY IN lock from AccessExclusive to a more reasonable
RowExclusive (my fault).  Also, install a check to prevent people
from trying COPY BINARY to stdout/from stdin.  No way that will
work unless we redesign the frontend COPY protocol ... which is
not worth the trouble in the near future ...

24 years agoUpdate trigger.sgml
Bruce Momjian [Wed, 17 May 2000 18:30:02 +0000 (18:30 +0000)]
Update trigger.sgml

24 years agoClean up trigger sgml docs.
Bruce Momjian [Wed, 17 May 2000 17:44:13 +0000 (17:44 +0000)]
Clean up trigger sgml docs.

24 years agoThe check of class string in configure (line 1732) fails because this class
Bruce Momjian [Wed, 17 May 2000 16:57:41 +0000 (16:57 +0000)]
The check of class string in configure (line 1732) fails because this class
is in <string> and not in <string.h> on QNX4/egcs-2.91.60.
Probably this can be changed for all platforms. The test in line 1705 uses
<string> as well. Because I am not sure, I havn't this included into the
patch.

doc/Makefile has to be sligthly  modified as it has been done for
src/backend/Makefile due to a QNX4 problem (patch attached)

Furthermore src/test/regress/run_check.sh needs to be patched as it has been
done for regress.sh (patch attached). Please note that in the patch the
postmaster is started always with the -i option.

run_check.sh reports the test "limit" as failed, but in reallity it is OK.
regress.sh reports it as OK.

Andreas Kardos

24 years agoBack out odbc changes until 7.1.
Bruce Momjian [Wed, 17 May 2000 15:36:37 +0000 (15:36 +0000)]
Back out odbc changes until 7.1.

24 years ago*** empty log message ***
Michael Meskes [Wed, 17 May 2000 06:03:14 +0000 (06:03 +0000)]
*** empty log message ***

24 years agoChange plpgsql example from SELECT * INTO rec to SELECT INTO rec *.
Bruce Momjian [Wed, 17 May 2000 00:15:48 +0000 (00:15 +0000)]
Change plpgsql example from SELECT * INTO rec to SELECT INTO rec *.

24 years agoFix typo
Bruce Momjian [Tue, 16 May 2000 21:16:12 +0000 (21:16 +0000)]
Fix typo

24 years agoSeveral compilation and run-time problems occur when building on SGI
Bruce Momjian [Tue, 16 May 2000 20:48:52 +0000 (20:48 +0000)]
Several compilation and run-time problems occur when building on SGI
IRIX systems using the native compilers.  A summary is:
- Various files use "//" as a comment delimiter in c files.
- Problems caused by assuming "char" is signed.
  cash.in: building -signed the rules regression test fails as described
    in FAQ_QNX4.  If CHAR_MAX is "255U" then ((signed char)CHAR_MAX) is -1.
  postmaster.c: random number regression test failed without this change.
- Some generic build issues and warning message cleanup.

David Kaelbling

24 years agoCleanup plpgsql man page typo
Bruce Momjian [Tue, 16 May 2000 19:29:43 +0000 (19:29 +0000)]
Cleanup plpgsql man page typo

24 years agoThe old init file won't run under bash on my box.
Bruce Momjian [Tue, 16 May 2000 03:18:43 +0000 (03:18 +0000)]
The old init file won't run under bash on my box.
Joseph Shraibman

24 years agoRemove configure check for how to abbreviate 'tr A-Z a-z', and instead
Tom Lane [Tue, 16 May 2000 02:14:15 +0000 (02:14 +0000)]
Remove configure check for how to abbreviate 'tr A-Z a-z', and instead
just use the portable form,
tr ABCDEFGHIJKLMNOPQRSTUVWXYZ abcdefghijklmnopqrstuvwxyz
There were a bunch of places that weren't paying attention to configure's
result anyway (including configure itself!?); clean them up too.

24 years agoFixed the message Makefile produces after compiling. It still said
Peter Mount [Mon, 15 May 2000 21:32:51 +0000 (21:32 +0000)]
Fixed the message Makefile produces after compiling. It still said
about the old Driver class, not the new package. Spotted by
Joseph Shraibman <jks@p1.selectacast.net>

24 years agocleanup
Bruce Momjian [Mon, 15 May 2000 16:20:45 +0000 (16:20 +0000)]
cleanup

24 years agocleanup
Bruce Momjian [Mon, 15 May 2000 16:16:23 +0000 (16:16 +0000)]
cleanup

24 years agoFix createlang -l dbname so it works, update sgml.
Bruce Momjian [Mon, 15 May 2000 16:12:39 +0000 (16:12 +0000)]
Fix createlang -l dbname so it works, update sgml.

24 years agoAdd mention of lo_unlink to manual
Bruce Momjian [Mon, 15 May 2000 12:42:23 +0000 (12:42 +0000)]
Add mention of lo_unlink to manual

24 years ago*** empty log message ***
Michael Meskes [Mon, 15 May 2000 09:36:06 +0000 (09:36 +0000)]
*** empty log message ***

24 years agoFix atan3-> atan2.
Bruce Momjian [Mon, 15 May 2000 01:41:18 +0000 (01:41 +0000)]
Fix atan3-> atan2.

24 years agoThat psql option should be --no-readline (as it used to be), not --noreadline.
Peter Eisentraut [Sun, 14 May 2000 18:05:05 +0000 (18:05 +0000)]
That psql option should be --no-readline (as it used to be), not --noreadline.

24 years agoRemove array file, per Tom Lane.
Bruce Momjian [Sun, 14 May 2000 03:47:27 +0000 (03:47 +0000)]
Remove array file, per Tom Lane.

24 years agoRemove arrays file, already have array file.
Bruce Momjian [Sun, 14 May 2000 03:26:38 +0000 (03:26 +0000)]
Remove arrays file, already have array file.

24 years agoFinally fix LISTEN problem.
Bruce Momjian [Sun, 14 May 2000 03:18:35 +0000 (03:18 +0000)]
Finally fix LISTEN problem.

24 years agoBack out listen fix.
Bruce Momjian [Sun, 14 May 2000 02:53:49 +0000 (02:53 +0000)]
Back out listen fix.

24 years agoUpdate TODO list.
Bruce Momjian [Sun, 14 May 2000 01:29:07 +0000 (01:29 +0000)]
Update TODO list.

24 years agoUpdate bug template.
Bruce Momjian [Sun, 14 May 2000 01:14:01 +0000 (01:14 +0000)]
Update bug template.

24 years agoFix create user for pgaccess.
Bruce Momjian [Sun, 14 May 2000 00:50:16 +0000 (00:50 +0000)]
Fix create user for pgaccess.

24 years agoTweak selectivity for area-based operators. Still a crock...
Tom Lane [Sat, 13 May 2000 06:04:46 +0000 (06:04 +0000)]
Tweak selectivity for area-based operators.  Still a crock...

24 years agoFix the off by one errors in ResultSet from 6.5.3, and more.
Bruce Momjian [Fri, 12 May 2000 20:54:22 +0000 (20:54 +0000)]
Fix the off by one errors in ResultSet from 6.5.3, and more.

I'm including a diff of
postgresql-7.0/src/interfaces/jdbc/org/postgresql/jdbc2/ResultSet.java.
I've clearly marked all the fixes I did. Would *someone* who has access
to the cvs please put this in?

Joseph Shraibman

24 years agoThis is the second time I've answered this exact same problem in two
Bruce Momjian [Fri, 12 May 2000 18:51:59 +0000 (18:51 +0000)]
This is the second time I've answered this exact same problem in two
days.  It seems to be a FAQ, and I think I know why. When creating a 'c'
language function, CREATE FUNCTION is fed the shared object filename,
and seems to succeed. Only when trying to use the function is an error
thrown, by which time the coder thinks something's wrong with executing
the code, not with loading it.

I think I once saw it proposed to load shared objects at function creation
time, but that idea was shot down on the grounds of resident memory bloat,
ISTR. Here's a patch for a compromise: all it does is stat() the file,
just like the loader code does, so that the errors caused by non existent
files, and no directory 'x' permissions (the most common ones, it seems),
get caught while the developer is still thinking about code loading. It
doesn't catch all errors (like the code not being readable by the postgres
user) but seems to catch the most common, without actually opening the file.

What do you think?

Ross

24 years agoUpdate TODO list.
Bruce Momjian [Fri, 12 May 2000 17:20:14 +0000 (17:20 +0000)]
Update TODO list.

24 years agoRemove cluster TODO e-mail file
Bruce Momjian [Fri, 12 May 2000 16:33:29 +0000 (16:33 +0000)]
Remove cluster TODO e-mail file

24 years agoBack out -\?. Didn't look good to Peter.
Bruce Momjian [Fri, 12 May 2000 16:28:31 +0000 (16:28 +0000)]
Back out -\?.  Didn't look good to Peter.

24 years ago/home/peter/commit-msg
Peter Eisentraut [Fri, 12 May 2000 16:13:44 +0000 (16:13 +0000)]
/home/peter/commit-msg

24 years agoSquash some more CLUSTER bugs. Never has worked on multiple-column
Tom Lane [Fri, 12 May 2000 16:10:09 +0000 (16:10 +0000)]
Squash some more CLUSTER bugs.  Never has worked on multiple-column
indexes, apparently, nor on functional indexes with more than one input
column (force of natts = 1 was in the wrong branch of IF statement).
Coredumped if source relation contained any uncommitted tuples, due to
failure to test for success return from heap_fetch.  Fetched tuple
was passed directly to heap_insert, which clobbers the TID and commit
status in the tuple header it's given, which meant that the source
relation's tuples all got trashed as the copy proceeded.  Abort partway
through, and you're left with a lot of missing tuples.
I wonder what else is lurking here ...

24 years agothis fixes the bug where setting the entry in he process table no longer works
Marc G. Fournier [Fri, 12 May 2000 14:33:08 +0000 (14:33 +0000)]
this fixes the bug where setting the entry in he process table no longer works
under FreeBSD ... basically, if setproctitle() exists, use it ...

the draw back right now is the PS_SET_STATUS stuff doesn't work, but am looking
into that one right now ... at lesat now you can see who is connecting where
and from where ...

24 years agoAdd two checks ... one for setproctitle and one for -lutil ...
Marc G. Fournier [Fri, 12 May 2000 13:58:25 +0000 (13:58 +0000)]
Add two checks ... one for setproctitle and one for -lutil ...

Don't do anything with them at this time, but am working on that ...

24 years agoUpdate TODO list.
Bruce Momjian [Fri, 12 May 2000 12:33:48 +0000 (12:33 +0000)]
Update TODO list.

24 years agoUpdate TODO list.
Bruce Momjian [Fri, 12 May 2000 12:24:45 +0000 (12:24 +0000)]
Update TODO list.

24 years agoRepair list-vs-node confusion that resulted in failure for INNER JOIN ON.
Tom Lane [Fri, 12 May 2000 01:33:56 +0000 (01:33 +0000)]
Repair list-vs-node confusion that resulted in failure for INNER JOIN ON.
Make it behave correctly when there are more than two tables being
joined, also.  Update regression test expected outputs.

24 years agoSmall cleanup of file.
Bruce Momjian [Fri, 12 May 2000 00:54:53 +0000 (00:54 +0000)]
Small cleanup of file.

24 years agoMore psql help cleanup
Bruce Momjian [Thu, 11 May 2000 18:41:00 +0000 (18:41 +0000)]
More psql help cleanup

24 years agoMakefile CFLAGS cleanups.
Bruce Momjian [Thu, 11 May 2000 17:46:35 +0000 (17:46 +0000)]
Makefile CFLAGS cleanups.

24 years agoUpdate SGML for FETCH and url's.
Bruce Momjian [Thu, 11 May 2000 17:32:33 +0000 (17:32 +0000)]
Update SGML for FETCH and url's.

24 years agoFix url reference with missing fields.
Thomas G. Lockhart [Thu, 11 May 2000 15:18:13 +0000 (15:18 +0000)]
Fix url reference with missing fields.
 Reported by Adrian Oboroc <aoboroc@btr.md>.

24 years agoOops, plpgsql didn't have the datetime->timestamp and timespan->interval
Bruce Momjian [Thu, 11 May 2000 04:00:00 +0000 (04:00 +0000)]
Oops, plpgsql didn't have the datetime->timestamp and timespan->interval
mappings.  In fact, it had them backward because it was using the 6.5.*
code.  Copied them from parser/gram.y, so it is fixed now.  Looks like
our first 7.0.1 fix.  Oops, seems Tom has beat me to it as I was typing
this.

24 years agoFix CLUSTER ... or at least undo the bit-rot it's suffered since 6.5.
Tom Lane [Thu, 11 May 2000 03:54:18 +0000 (03:54 +0000)]
Fix CLUSTER ... or at least undo the bit-rot it's suffered since 6.5.
It's still pretty fundamentally bogus :-(.
Freebie side benefit: ALTER TABLE RENAME works on indexes now.

24 years agoDisplay -? as -\? under unix for psql.
Bruce Momjian [Thu, 11 May 2000 03:14:19 +0000 (03:14 +0000)]
Display -? as -\? under unix for psql.

24 years agoAdd options and cleanup psql \? and -? help displays
Bruce Momjian [Thu, 11 May 2000 01:37:54 +0000 (01:37 +0000)]
Add options and cleanup psql \? and -? help displays

24 years agoUpdate psql \? to show file rather than "fname".
Bruce Momjian [Tue, 9 May 2000 19:08:36 +0000 (19:08 +0000)]
Update psql \? to show file rather than "fname".

24 years agoFix psql quotes in docs.
Bruce Momjian [Tue, 9 May 2000 18:55:56 +0000 (18:55 +0000)]
Fix psql quotes in docs.

24 years agoMore single-quote fixes.
Bruce Momjian [Tue, 9 May 2000 18:30:43 +0000 (18:30 +0000)]
More single-quote fixes.

24 years agoAdd single quote.
Bruce Momjian [Tue, 9 May 2000 18:27:25 +0000 (18:27 +0000)]
Add single quote.

24 years agoChange HISTORY to show outer joins in 7.1 or 7.2. REL7_0
Bruce Momjian [Mon, 8 May 2000 23:56:50 +0000 (23:56 +0000)]
Change HISTORY to show outer joins in 7.1 or 7.2.

24 years agoMinor fixups for markup and wording.
Thomas G. Lockhart [Mon, 8 May 2000 16:19:56 +0000 (16:19 +0000)]
Minor fixups for markup and wording.

24 years agoUpdate TODO list.
Bruce Momjian [Sat, 6 May 2000 20:21:08 +0000 (20:21 +0000)]
Update TODO list.

24 years agoUpdate TODO list.
Bruce Momjian [Sat, 6 May 2000 16:05:17 +0000 (16:05 +0000)]
Update TODO list.

24 years agoUpdate TODO list.
Bruce Momjian [Sat, 6 May 2000 01:05:10 +0000 (01:05 +0000)]
Update TODO list.

24 years agoUpdate TODO list.
Bruce Momjian [Fri, 5 May 2000 22:25:20 +0000 (22:25 +0000)]
Update TODO list.

24 years agoForgot that dumpall's output script should 'delete from pg_group' before
Tom Lane [Fri, 5 May 2000 17:50:38 +0000 (17:50 +0000)]
Forgot that dumpall's output script should 'delete from pg_group' before
loading new data, for consistency with its handling of pg_shadow.

24 years agopsql: suppress warnings about too many arguments if the command is not valid in the...
Peter Eisentraut [Fri, 5 May 2000 09:38:40 +0000 (09:38 +0000)]
psql: suppress warnings about too many arguments if the command is not valid in the first place

24 years agoSpello SERIALIZED -> SERIALIZABLE in psql tab completion
Peter Eisentraut [Fri, 5 May 2000 08:44:27 +0000 (08:44 +0000)]
Spello SERIALIZED -> SERIALIZABLE in psql tab completion

24 years agoImageViewer transaction fixes
Peter Mount [Fri, 5 May 2000 07:35:29 +0000 (07:35 +0000)]
ImageViewer transaction fixes

24 years agoUpdate TODO list.
Bruce Momjian [Fri, 5 May 2000 04:06:18 +0000 (04:06 +0000)]
Update TODO list.

24 years agoPlease apply the following patch to the FAQ_AIX
Bruce Momjian [Fri, 5 May 2000 03:40:20 +0000 (03:40 +0000)]
Please apply the following patch to the FAQ_AIX

Thanks
Andreas

24 years agoBecause we ended up forcing an initdb for 7.0 final, we aren't going
Tom Lane [Fri, 5 May 2000 03:11:24 +0000 (03:11 +0000)]
Because we ended up forcing an initdb for 7.0 final, we aren't going
to need this updatepgproc.sql script after all...

24 years agoForce initdb because of pg_group index fix.
Tom Lane [Fri, 5 May 2000 03:10:24 +0000 (03:10 +0000)]
Force initdb because of pg_group index fix.

24 years agoMake the indexes on pg_group be shared system relations.
Tom Lane [Fri, 5 May 2000 03:09:43 +0000 (03:09 +0000)]
Make the indexes on pg_group be shared system relations.

24 years agoAccept pg_group as well as pg_shadow data from dumpall script.
Tom Lane [Fri, 5 May 2000 03:08:20 +0000 (03:08 +0000)]
Accept pg_group as well as pg_shadow data from dumpall script.
Rearrange handling of VACUUMs so that they are certain to be executed
as superuser not some random user; also, do not forget to vacuum
template1 itself.

24 years agoDump contents of pg_group along with pg_shadow.
Tom Lane [Fri, 5 May 2000 03:04:00 +0000 (03:04 +0000)]
Dump contents of pg_group along with pg_shadow.

24 years agoDon't leak a file descriptor when updating pg_pwd file. Also, check for
Tom Lane [Thu, 4 May 2000 20:06:07 +0000 (20:06 +0000)]
Don't leak a file descriptor when updating pg_pwd file.  Also, check for
failure of rename() call.

24 years agoPolishing for 7.0 release. Fix up inconsistant spacing in release notes.
Thomas G. Lockhart [Thu, 4 May 2000 16:15:08 +0000 (16:15 +0000)]
Polishing for 7.0 release. Fix up inconsistant spacing in release notes.

24 years agoInstallation guide for 7.0 release. From SGML sources.
Thomas G. Lockhart [Thu, 4 May 2000 16:12:05 +0000 (16:12 +0000)]
Installation guide for 7.0 release. From SGML sources.

24 years agoDo not try to build libpq++ if class 'string' is not defined in the
Tom Lane [Wed, 3 May 2000 18:30:17 +0000 (18:30 +0000)]
Do not try to build libpq++ if class 'string' is not defined in the
available C++ header files.

24 years agoMinor fixes ready for 7.0
Peter Mount [Wed, 3 May 2000 15:58:09 +0000 (15:58 +0000)]
Minor fixes ready for 7.0

24 years agoOops,I've forgotten to add an documentation about -P option of
Hiroshi Inoue [Wed, 3 May 2000 07:33:44 +0000 (07:33 +0000)]
Oops,I've forgotten to add an documentation about -P option of
postgres.

24 years agoUpdate sco faq.
Bruce Momjian [Wed, 3 May 2000 03:45:55 +0000 (03:45 +0000)]
Update sco faq.

24 years agoUpdate FAQ.
Bruce Momjian [Tue, 2 May 2000 21:57:59 +0000 (21:57 +0000)]
Update FAQ.

24 years agoRemove derived files from CVS. Now kept in
Thomas G. Lockhart [Tue, 2 May 2000 21:08:36 +0000 (21:08 +0000)]
Remove derived files from CVS. Now kept in
 hub.org:/home/projects/pgsql/ftp/www/html/users-lounge/7.0/docs/

24 years agoImprove section heading for date/time history.
Thomas G. Lockhart [Tue, 2 May 2000 20:36:21 +0000 (20:36 +0000)]
Improve section heading for date/time history.

24 years agoFixups in content and markup for 7.0 release.
Thomas G. Lockhart [Tue, 2 May 2000 20:02:03 +0000 (20:02 +0000)]
Fixups in content and markup for 7.0 release.

24 years agoUpdate HISTORY/sgml for new pgaccess version.
Bruce Momjian [Tue, 2 May 2000 17:06:10 +0000 (17:06 +0000)]
Update HISTORY/sgml for new pgaccess version.

24 years agoUpdate SCO FAQ. Billy G. Allie
Bruce Momjian [Tue, 2 May 2000 10:57:11 +0000 (10:57 +0000)]
Update SCO FAQ. Billy G. Allie

24 years agoModify getdatabaseencoding(), pg_encoding_to_char()
Tatsuo Ishii [Tue, 2 May 2000 08:13:08 +0000 (08:13 +0000)]
Modify getdatabaseencoding(), pg_encoding_to_char()
pg_char_to_encoding() in multibyte disbaled case so that it does not
throw an error, rather return HARD CODED default value (currently SQL_ASCII).
This would solve the "non-mb backend vs. mb-enabled frontend" problem.

24 years agoReset CurrentMemoryContext to TopMemoryContext at the beginning of error
Tom Lane [Sun, 30 Apr 2000 21:29:23 +0000 (21:29 +0000)]
Reset CurrentMemoryContext to TopMemoryContext at the beginning of error
cleanup, ie, as soon as we have caught the longjmp.  This ensures that
current context will be a valid context throughout error cleanup.  Before
it was possible that current context was pointing at a context that would
get deleted during cleanup, leaving any subsequent pallocs in deep
trouble.  I was able to provoke an Assert failure when compiled with
asserts + -DCLOBBER_FREED_MEMORY, if I did something that would cause
an error to be reported by the backend large-object code, because indeed
that code operates in a context that gets deleted partway through xact
abort --- and CurrentMemoryContext was still pointing at it!  Boo hiss.