]> granicus.if.org Git - postgresql/log
postgresql
23 years ago- Use subselect when enabling triggers,
Philip Warner [Tue, 13 Mar 2001 13:18:44 +0000 (13:18 +0000)]
- Use subselect  when enabling triggers,
- Use exact table names when enabling/disabling triggers

23 years agoThe attached file contains the registry settings required to install the
Hiroshi Inoue [Tue, 13 Mar 2001 09:03:24 +0000 (09:03 +0000)]
The attached file contains the registry settings required to install the
ODBC driver on Windows 9X/ME/NT/2K when using the later versions of the
driver that don't have the Installshield installation:

1) Install psqlodbc.dll in to C:\Windows\System or C:\Winnt\System32
2) Add the registry settings in the attached file using regedit.

A useful addition to src/interfaces/odbc perhaps?

Regards, Dave.

23 years agoUpdate for new pg_control format.
Tom Lane [Tue, 13 Mar 2001 01:17:40 +0000 (01:17 +0000)]
Update for new pg_control format.

23 years agoXLOG (and related) changes:
Tom Lane [Tue, 13 Mar 2001 01:17:06 +0000 (01:17 +0000)]
XLOG (and related) changes:

* Store two past checkpoint locations, not just one, in pg_control.
  On startup, we fall back to the older checkpoint if the newer one
  is unreadable.  Also, a physical copy of the newest checkpoint record
  is kept in pg_control for possible use in disaster recovery (ie,
  complete loss of pg_xlog).  Also add a version number for pg_control
  itself.  Remove archdir from pg_control; it ought to be a GUC
  parameter, not a special case (not that it's implemented yet anyway).

* Suppress successive checkpoint records when nothing has been entered
  in the WAL log since the last one.  This is not so much to avoid I/O
  as to make it actually useful to keep track of the last two
  checkpoints.  If the things are right next to each other then there's
  not a lot of redundancy gained...

* Change CRC scheme to a true 64-bit CRC, not a pair of 32-bit CRCs
  on alternate bytes.  Polynomial borrowed from ECMA DLT1 standard.

* Fix XLOG record length handling so that it will work at BLCKSZ = 32k.

* Change XID allocation to work more like OID allocation.  (This is of
  dubious necessity, but I think it's a good idea anyway.)

* Fix a number of minor bugs, such as off-by-one logic for XLOG file
  wraparound at the 4 gig mark.

* Add documentation and clean up some coding infelicities; move file
  format declarations out to include files where planned contrib
  utilities can get at them.

* Checkpoint will now occur every CHECKPOINT_SEGMENTS log segments or
  every CHECKPOINT_TIMEOUT seconds, whichever comes first.  It is also
  possible to force a checkpoint by sending SIGUSR1 to the postmaster
  (undocumented feature...)

* Defend against kill -9 postmaster by storing shmem block's key and ID
  in postmaster.pid lockfile, and checking at startup to ensure that no
  processes are still connected to old shmem block (if it still exists).

* Switch backends to accept SIGQUIT rather than SIGUSR1 for emergency
  stop, for symmetry with postmaster and xlog utilities.  Clean up signal
  handling in bootstrap.c so that xlog utilities launched by postmaster
  will react to signals better.

* Standalone bootstrap now grabs lockfile in target directory, as added
  insurance against running it in parallel with live postmaster.

23 years agoAvoid O(N^2) behavior in deferredTriggerAddEvent() for large numbers of
Tom Lane [Mon, 12 Mar 2001 23:02:00 +0000 (23:02 +0000)]
Avoid O(N^2) behavior in deferredTriggerAddEvent() for large numbers of
tuples inserted/deleted/updated in a single transaction.  On my machine,
this reduced the time to delete 80000 tuples in a foreign-key-referencing
table from ~15min to ~8sec.

23 years agoUpdate TODO list.
Bruce Momjian [Sun, 11 Mar 2001 19:26:58 +0000 (19:26 +0000)]
Update TODO list.

23 years agoThrow error if Ant is not found and Java is requested. Remove redundant
Peter Eisentraut [Sun, 11 Mar 2001 11:24:59 +0000 (11:24 +0000)]
Throw error if Ant is not found and Java is requested.  Remove redundant
AC_SUBST that messed up the diversions and thus the configure output.

23 years agoAdd uninstall target to Java build.
Peter Eisentraut [Sun, 11 Mar 2001 11:07:01 +0000 (11:07 +0000)]
Add uninstall target to Java build.
Respect default port setting in JDBC driver.
Pick up version number from Makefile.global.
Change installation directory to share/java/.
Document.

23 years agoTurns out the HPUX linker likes -Bsymbolic too. Without this, ODBC
Tom Lane [Sat, 10 Mar 2001 23:15:20 +0000 (23:15 +0000)]
Turns out the HPUX linker likes -Bsymbolic too.  Without this, ODBC
driver does not work because its internal cross-references get bound
to similarly named functions in unixODBC shared library.

23 years agoProvide a hack to let initialization happen on platforms for which
Tom Lane [Sat, 10 Mar 2001 23:12:28 +0000 (23:12 +0000)]
Provide a hack to let initialization happen on platforms for which
psqlodbc.c's constructor-making techniques do not work.

23 years agoDo not accept values from sections following the specified section.
Tom Lane [Sat, 10 Mar 2001 22:30:57 +0000 (22:30 +0000)]
Do not accept values from sections following the specified section.

23 years agoBOX documentation disagreed with code about corner order.
Tom Lane [Sat, 10 Mar 2001 21:16:32 +0000 (21:16 +0000)]
BOX documentation disagreed with code about corner order.

23 years agoUpdate TODO list.
Bruce Momjian [Sat, 10 Mar 2001 18:18:26 +0000 (18:18 +0000)]
Update TODO list.

23 years agoDocument --with-java.
Peter Eisentraut [Sat, 10 Mar 2001 16:57:18 +0000 (16:57 +0000)]
Document --with-java.

23 years agoEliminate some hackery when creating text files (INSTALL) with a few lines
Peter Eisentraut [Sat, 10 Mar 2001 16:05:35 +0000 (16:05 +0000)]
Eliminate some hackery when creating text files (INSTALL) with a few lines
of DSSSL.

23 years agoUse install-sh unconditionally.
Peter Eisentraut [Sat, 10 Mar 2001 10:38:59 +0000 (10:38 +0000)]
Use install-sh unconditionally.

23 years agoFATAL errors should cause exit with nonzero status if we are not running
Tom Lane [Sat, 10 Mar 2001 04:21:51 +0000 (04:21 +0000)]
FATAL errors should cause exit with nonzero status if we are not running
under the postmaster --- specifically, if we are a standalone backend
running under the initdb script, this is critical!

23 years ago1)Allow the access to indexes with up to 16 keys.
Hiroshi Inoue [Sat, 10 Mar 2001 03:34:57 +0000 (03:34 +0000)]
1)Allow the access to indexes with up to 16 keys.
2)Fix some memory leaks.
3)Change some bogus error messages.

23 years agoDon't choke on superuser names containing random punctuation.
Tom Lane [Fri, 9 Mar 2001 22:10:13 +0000 (22:10 +0000)]
Don't choke on superuser names containing random punctuation.

23 years agoRecursive distclean shouldn't depend on recursive clean. Subdirectories
Peter Eisentraut [Fri, 9 Mar 2001 21:50:27 +0000 (21:50 +0000)]
Recursive distclean shouldn't depend on recursive clean.  Subdirectories
handle this.

23 years agoIntegrate "Porting from Oracle PL/SQL" HOWTO from Roberto Mello.
Peter Eisentraut [Fri, 9 Mar 2001 19:09:00 +0000 (19:09 +0000)]
Integrate "Porting from Oracle PL/SQL" HOWTO from Roberto Mello.

23 years agohold interupts during LockWaitCancel().
Hiroshi Inoue [Fri, 9 Mar 2001 06:36:32 +0000 (06:36 +0000)]
hold interupts during LockWaitCancel().

23 years agoAdd Japanese FAQ>
Bruce Momjian [Thu, 8 Mar 2001 17:00:28 +0000 (17:00 +0000)]
Add Japanese FAQ>

23 years agoUpdate TODO list.
Bruce Momjian [Thu, 8 Mar 2001 16:53:07 +0000 (16:53 +0000)]
Update TODO list.

23 years agoUpdate FAQ.
Bruce Momjian [Thu, 8 Mar 2001 16:52:10 +0000 (16:52 +0000)]
Update FAQ.

23 years agoA subplan invoked within an aggregate function's argument should
Tom Lane [Thu, 8 Mar 2001 01:49:01 +0000 (01:49 +0000)]
A subplan invoked within an aggregate function's argument should
be allowed to receive ungrouped variables of the current query level.
Curious that no one reported this bug before...

23 years agoModify wchar conversion routines to not fetch the next byte past the end
Tom Lane [Thu, 8 Mar 2001 00:24:34 +0000 (00:24 +0000)]
Modify wchar conversion routines to not fetch the next byte past the end
of a counted input string.  Marinos Yannikos' recent crash report turns
out to be due to applying pg_ascii2wchar_with_len to a TEXT object that
is smack up against the end of memory.  This is the second just-barely-
reproducible bug report I have seen that traces to some bit of code
fetching one more byte than it is allowed to.  Let's be more careful
out there, boys and girls.
While at it, I changed the code to not risk a similar crash when there
is a truncated multibyte character at the end of an input string.  The
output in this case might not be the most reasonable output possible;
if anyone wants to improve it further, step right up...

23 years agoRepair a number of places that didn't bother to check whether PageAddItem
Tom Lane [Wed, 7 Mar 2001 21:20:26 +0000 (21:20 +0000)]
Repair a number of places that didn't bother to check whether PageAddItem
succeeds or not.  Revise rtree page split algorithm to take care about
making a feasible split --- ie, will the incoming tuple actually fit?
Failure to make a feasible split, combined with failure to notice the
failure, account for Jim Stone's recent bug report.  I suspect that
hash and gist indices may have the same type of bug, but at least now
we'll get error messages rather than silent failures if so.  Also clean
up rtree code to use Datum rather than char* where appropriate.

23 years ago> Applied. Thanks.
Bruce Momjian [Wed, 7 Mar 2001 16:18:08 +0000 (16:18 +0000)]
> Applied.  Thanks.

One more :)) It's for improper function argumets for
PLTCL_UNKNOWN_SUPPORT code

I'm not an autoconf expert, but is it possible to enable unknown
support in pltcl with configure option ?
This support is really handy for real life usage of pl/tcl.

seva@sevasoft.kiev.ua

23 years agopltcl_loadmod has problems with big (>4k) modules because of missing incr
Bruce Momjian [Wed, 7 Mar 2001 16:09:09 +0000 (16:09 +0000)]
pltcl_loadmod has problems with big (>4k) modules because of missing incr
in splitting code:

seva@sevasoft.kiev.ua

23 years agoUpdate my2pg, new version.
Bruce Momjian [Tue, 6 Mar 2001 22:46:50 +0000 (22:46 +0000)]
Update my2pg, new version.

23 years agoBelieve $POSTGRES_LIB and $POSTGRES_INCLUDE only if they name actual
Tom Lane [Tue, 6 Mar 2001 22:07:09 +0000 (22:07 +0000)]
Believe $POSTGRES_LIB and $POSTGRES_INCLUDE only if they name actual
directories, per suggestion from Robert Creager.

23 years agoSync rule for making INSTALL file.
Peter Eisentraut [Tue, 6 Mar 2001 20:41:06 +0000 (20:41 +0000)]
Sync rule for making INSTALL file.

23 years agoHelp message s/dump/restore/ (must have been copy&pasted from pg_dump).
Peter Eisentraut [Tue, 6 Mar 2001 19:58:27 +0000 (19:58 +0000)]
Help message s/dump/restore/ (must have been copy&pasted from pg_dump).

23 years agoSome editing, enhance markup, move description section before options list.
Peter Eisentraut [Tue, 6 Mar 2001 18:55:57 +0000 (18:55 +0000)]
Some editing, enhance markup, move description section before options list.

23 years agoShorten TOAST mention
Bruce Momjian [Tue, 6 Mar 2001 18:11:39 +0000 (18:11 +0000)]
Shorten TOAST mention

23 years agoUpdate/correct/refine.
Peter Eisentraut [Tue, 6 Mar 2001 17:43:56 +0000 (17:43 +0000)]
Update/correct/refine.

23 years agoTue Mar 06 12:05:00 GMT 2001 peter@retep.org.uk
Peter Mount [Tue, 6 Mar 2001 12:04:46 +0000 (12:04 +0000)]
Tue Mar 06 12:05:00 GMT 2001 peter@retep.org.uk
        - Removed org.postgresql.xa.Test from the JDBC EE driver as it's an old
          test class and prevented it from compiling.

23 years ago- Reflect change of option -U to -L
Philip Warner [Tue, 6 Mar 2001 05:22:50 +0000 (05:22 +0000)]
- Reflect change of option -U to -L

23 years ago- Minor correction to usage
Philip Warner [Tue, 6 Mar 2001 05:22:18 +0000 (05:22 +0000)]
- Minor correction to usage

23 years ago- Dump relevant parts of sequences only when doing schemaOnly & dataOnly
Philip Warner [Tue, 6 Mar 2001 04:53:28 +0000 (04:53 +0000)]
- Dump relevant parts of sequences only when doing schemaOnly & dataOnly
- Prevent double-dumping of sequences when dataOnly.

23 years ago- Only disable triggers in DataOnly (or implied data-only) restores.
Philip Warner [Tue, 6 Mar 2001 04:08:04 +0000 (04:08 +0000)]
- Only disable triggers in DataOnly (or implied data-only) restores.
- Change -U option to -L to allow -U to specify username in future.\7f (pg_restore)

23 years agoAdd missing space, change some markup.
Peter Eisentraut [Mon, 5 Mar 2001 19:00:39 +0000 (19:00 +0000)]
Add missing space, change some markup.

23 years agoSorry, that change was not correct.
Peter Eisentraut [Mon, 5 Mar 2001 19:00:01 +0000 (19:00 +0000)]
Sorry, that change was not correct.

23 years agoI'm attaching those diffs for the Reference Guide in a tar file, as
Bruce Momjian [Mon, 5 Mar 2001 18:42:57 +0000 (18:42 +0000)]
I'm attaching those diffs for the Reference Guide in a tar file, as
not all of them attached properly in the post I made a few minutes
ago. Please disregard those earlier files. The diffs in the tar file
replace them.

Pierce Tyler

23 years agoUpdate email addresses.
Bruce Momjian [Mon, 5 Mar 2001 18:09:55 +0000 (18:09 +0000)]
Update email addresses.

23 years agoHere is the chinese_big5 patch for PgAccess. I've tested under Chinese
Bruce Momjian [Mon, 5 Mar 2001 17:20:37 +0000 (17:20 +0000)]
Here is the chinese_big5 patch for PgAccess.  I've tested under Chinese
Windows 2000 without any problem.

Have fun.

LM.Liu

23 years agoHmmm, a conflict with V1.104 of configure.in remained and broke configure
Peter Mount [Mon, 5 Mar 2001 10:18:48 +0000 (10:18 +0000)]
Hmmm, a conflict with V1.104 of configure.in remained and broke configure
(said redirection required when run).

After checking using cvsweb, removed the offending conflict.
Rebuilt configure using autoconf, and it now works fine.

23 years agoThere's always 1 file missed out ;-)
Peter Mount [Mon, 5 Mar 2001 10:02:35 +0000 (10:02 +0000)]
There's always 1 file missed out ;-)

Found while testing against a full checkout. Peter

23 years agoOk, I've split todays commit into three, the first two already done had some
Peter Mount [Mon, 5 Mar 2001 09:40:02 +0000 (09:40 +0000)]
Ok, I've split todays commit into three, the first two already done had some
bits in JDBC & the first set of tools into contrib.

This is the third, and deals with enabling JDBC to be compiled with the main
source.

What it does is add a new option to configure: --with-java

This option tells configure to look for ant (our build tool of choice) and
if found, it then compiles both the JDBC driver and the new tools as part
of the normal make.

Also, when the postgresql install is done, all the .jar files are also
installed into the ${PGLIB}/java directory (thought best to keep then separate)

Now I had some conflicts when this applied so could someone please double check
that everything is ok?

Peter

23 years agoMinor fixes...
Peter Mount [Mon, 5 Mar 2001 09:17:43 +0000 (09:17 +0000)]
Minor fixes...

23 years agoFirst batch of the tools merged in...
Peter Mount [Mon, 5 Mar 2001 09:15:38 +0000 (09:15 +0000)]
First batch of the tools merged in...

23 years agoPyGreSQL documentation. Based on the README, converted to DocBook by
Peter Eisentraut [Sun, 4 Mar 2001 18:54:07 +0000 (18:54 +0000)]
PyGreSQL documentation.  Based on the README, converted to DocBook by
Mikhail Terekhov <terekhov@emc.com>.

23 years agoUpdate mysql converter, new version released.
Bruce Momjian [Sun, 4 Mar 2001 15:43:33 +0000 (15:43 +0000)]
Update mysql converter, new version released.

23 years agoDocument TEMP option.
Tom Lane [Sat, 3 Mar 2001 22:11:40 +0000 (22:11 +0000)]
Document TEMP option.

23 years agoAdd configure check for -lunix, for QNX.
Peter Eisentraut [Sat, 3 Mar 2001 15:53:41 +0000 (15:53 +0000)]
Add configure check for -lunix, for QNX.

Recode test for equality of source and build directory using 'test -ef',
because even using pwd you might not get equal strings.  Thanks, QNX.

23 years agoUpdate the version number. We may change this to 7.1 if we align it with
D'Arcy J.M. Cain [Sat, 3 Mar 2001 14:00:56 +0000 (14:00 +0000)]
Update the version number.  We may change this to 7.1 if we align it with
PostgreSQL.

Add notice that development has moved into the PostgreSQL tree.

23 years agoAdd oid to list of keys cached.
D'Arcy J.M. Cain [Sat, 3 Mar 2001 13:58:24 +0000 (13:58 +0000)]
Add oid to list of keys cached.

Add a test to avoid an exception in certain cases.

23 years agoIncrementing version number in preparation for next release. Note that I
D'Arcy J.M. Cain [Sat, 3 Mar 2001 13:54:35 +0000 (13:54 +0000)]
Incrementing version number in preparation for next release.  Note that I
am talking with Thomas Lockhart about the idea of bringing the PyGreSQL
version number into alignment with PostgreSQL so this may change to 7.1
before the release.

I have added to the copyright to indicate that from now on the PostgreSQL
copyright will apply.  If someone wants to make that clearer please do.
The existing copyrights need to stay there for now but if necessary I can
ask Pascal Andre if he agrees to a different wording.

Added reference to the Python DB-API 2.0 compliant API wrapper.

Added reference to the PyGreSQL mailing list.

23 years agoAdded postgres.h header for more type checking.
D'Arcy J.M. Cain [Sat, 3 Mar 2001 13:42:37 +0000 (13:42 +0000)]
Added postgres.h header for more type checking.

Changed the way that OID is retrieved on inserts.  PQoidStatus appears
to be deprecated so I am using PQoidValue instead.

23 years agoUpdate TODO list.
Bruce Momjian [Fri, 2 Mar 2001 16:10:48 +0000 (16:10 +0000)]
Update TODO list.

23 years agoAdd missing include.
Peter Eisentraut [Thu, 1 Mar 2001 19:03:44 +0000 (19:03 +0000)]
Add missing include.

23 years agoFix memory leak.
Peter Eisentraut [Thu, 1 Mar 2001 18:52:50 +0000 (18:52 +0000)]
Fix memory leak.

23 years agoEnsure previous setting of pset.queryFout is restored after a failed
Tom Lane [Thu, 1 Mar 2001 18:34:29 +0000 (18:34 +0000)]
Ensure previous setting of pset.queryFout is restored after a failed
backslash-g command.

23 years agoAdd missing semicolon required by QNX shell.
Peter Eisentraut [Thu, 1 Mar 2001 16:17:53 +0000 (16:17 +0000)]
Add missing semicolon required by QNX shell.
from "Tegge, Bernd" <tegge@repas-aeg.de>

23 years agoRemove HAVE_OPTARG per discussion in hackers list.
Tatsuo Ishii [Thu, 1 Mar 2001 05:05:29 +0000 (05:05 +0000)]
Remove HAVE_OPTARG per discussion in hackers list.

23 years agoUpdate TODO list.
Bruce Momjian [Thu, 1 Mar 2001 03:30:19 +0000 (03:30 +0000)]
Update TODO list.

23 years agoUpdate TODO list.
Bruce Momjian [Thu, 1 Mar 2001 02:17:32 +0000 (02:17 +0000)]
Update TODO list.

23 years agoDo not strip whitespace within quotes.
Peter Eisentraut [Wed, 28 Feb 2001 20:39:42 +0000 (20:39 +0000)]
Do not strip whitespace within quotes.

23 years agoUpdate TODO list.
Bruce Momjian [Wed, 28 Feb 2001 18:36:26 +0000 (18:36 +0000)]
Update TODO list.

23 years agoUpdate TODO list.
Bruce Momjian [Wed, 28 Feb 2001 18:07:15 +0000 (18:07 +0000)]
Update TODO list.

23 years agoTweak portal (cursor) code so that it will not call the executor again
Tom Lane [Tue, 27 Feb 2001 22:07:34 +0000 (22:07 +0000)]
Tweak portal (cursor) code so that it will not call the executor again
when user does another FETCH after reaching end of data, or another
FETCH backwards after reaching start.  This is needed because some plan
nodes are not very robust about being called again after they've already
returned NULL; for example, MergeJoin will crash in some states but not
others.  While the ideal approach would be for them all to handle this
correctly, it seems foolish to assume that no such bugs would creep in
again once cleaned up.  Therefore, the most robust answer is to prevent
the situation from arising at all.

23 years agoMark new text<->date, text<->time, text<->timetz conversion functions as
Tom Lane [Tue, 27 Feb 2001 20:34:10 +0000 (20:34 +0000)]
Mark new text<->date, text<->time, text<->timetz conversion functions as
noncachable, so that CURRENT_DATE and CURRENT_TIME work as functions
again, rather than being collapsed to constants immediately.  Marking the
reverse conversions noncachable might be overkill, but I'm not sure;
do these datatypes have the notion of a CURRENT value?  Better safe than
sorry, for now.

23 years agostart of an admin tool for FreeBSD to help configure shared memory for
Marc G. Fournier [Tue, 27 Feb 2001 15:37:39 +0000 (15:37 +0000)]
start of an admin tool for FreeBSD to help configure shared memory for
that machine

I kept the name as generic as possible though, as other OS should have
similar methods, so this can be extended "as appropriate" ...

23 years agoMassive commits for SunOS4 port.
Tatsuo Ishii [Tue, 27 Feb 2001 08:13:31 +0000 (08:13 +0000)]
Massive commits for SunOS4 port.

23 years agoFix vacuum analyze error.
Tatsuo Ishii [Tue, 27 Feb 2001 07:07:00 +0000 (07:07 +0000)]
Fix vacuum analyze error.

vacuum analyze on pg_type fails if bogus entries remain in pg_operator.
Here is a sample script to reproduce the problem.

drop table t1;
create table t1(i int);
drop function foo(t1,t1);
create function foo(t1,t1) returns bool as 'select true' language 'sql';
create operator = (
leftarg = t1,
rightarg = t1,
commutator = =,
procedure = foo
);
drop table t1;
vacuum analyze;

23 years agoFixed variable handling in preproc.y.
Michael Meskes [Mon, 26 Feb 2001 14:42:54 +0000 (14:42 +0000)]
Fixed variable handling in preproc.y.

23 years agoAllow pgaccess to input Japanese. See included mail.
Tatsuo Ishii [Mon, 26 Feb 2001 05:15:48 +0000 (05:15 +0000)]
Allow pgaccess to input Japanese. See included mail.

Subject: [HACKERS] pgaccess Japanese input capability patch
From: Tatsuo Ishii <t-ishii@sra.co.jp>
To: teo@flex.ro
Cc: pgsql-hackers@postgresql.org, pgsql-interfaces@postgresql.org
Date: Sat, 24 Feb 2001 21:41:14 +0900

Hi Teodorescu,

I have made patches which enable pgaccess to input Japanese characters
in the table editing window. As you might know, to input Japanese
characters, we first type in "hiragana" then convert it to "kanji". To
make this proccess transparent to tcl application programs, libraries
are provided with localized version of Tcl/Tk. The patches bind
certain keys to initiate a function (kanjiInput) that is responsible
for the conversion process. If the function is not available, those
keys will not be binded.

Comments?
--
Tatsuo Ishii

23 years agoImplement COMMIT_SIBLINGS parameter to allow pre-commit delay to occur
Tom Lane [Mon, 26 Feb 2001 00:50:08 +0000 (00:50 +0000)]
Implement COMMIT_SIBLINGS parameter to allow pre-commit delay to occur
only if at least N other backends currently have open transactions.  This
is not a great deal of intelligence about whether a delay might be
profitable ... but it beats no intelligence at all.  Note that the default
COMMIT_DELAY is still zero --- this new code does nothing unless that
setting is changed.
Also, mark ENABLEFSYNC as a system-wide setting.  It's no longer safe to
allow that to be set per-backend, since we may be relying on some other
backend's fsync to have synced the WAL log.

23 years agoEven better example for operator precedence mis-parsing.
Peter Eisentraut [Sun, 25 Feb 2001 16:05:21 +0000 (16:05 +0000)]
Even better example for operator precedence mis-parsing.

23 years agoDoesn't need libpq.
Peter Eisentraut [Sun, 25 Feb 2001 15:57:45 +0000 (15:57 +0000)]
Doesn't need libpq.

23 years agoRemove binary files.
Peter Eisentraut [Sun, 25 Feb 2001 15:55:05 +0000 (15:55 +0000)]
Remove binary files.

23 years agoAt least on HPUX, select with delay.tv_sec = 0 and delay.tv_usec = 1000000
Tom Lane [Sat, 24 Feb 2001 22:42:45 +0000 (22:42 +0000)]
At least on HPUX, select with delay.tv_sec = 0 and delay.tv_usec = 1000000
does not lead to a one-second delay, but to an immediate EINVAL failure.
This causes CHECKPOINT to crash with s_lock_stuck much too quickly :-(.
Fix by breaking down the requested wait div/mod 1e6.

23 years agoChoose a more suitable example for the operator precedence mis-parsing
Peter Eisentraut [Sat, 24 Feb 2001 18:09:51 +0000 (18:09 +0000)]
Choose a more suitable example for the operator precedence mis-parsing
example.

23 years agoMarkup, spell check, refine JDBC documentation.
Peter Eisentraut [Sat, 24 Feb 2001 12:40:27 +0000 (12:40 +0000)]
Markup, spell check, refine JDBC documentation.

23 years agoWhen under postmaster, bogus arguments should cause proc_exit(0) not
Tom Lane [Sat, 24 Feb 2001 02:04:51 +0000 (02:04 +0000)]
When under postmaster, bogus arguments should cause proc_exit(0) not
proc_exit(1).  Unless you think a system-wide restart is an appropriate
response to bogus PGOPTIONS, that is.

23 years agoFix pg_dump crashes caused by bogus use of va_start/va_end (only seen
Tom Lane [Fri, 23 Feb 2001 22:52:32 +0000 (22:52 +0000)]
Fix pg_dump crashes caused by bogus use of va_start/va_end (only seen
on some platforms, which is not too surprising considering how platform
specific these macros must be).

23 years agoI had a need to read such things as the backend locale and the catalog
Bruce Momjian [Fri, 23 Feb 2001 20:38:35 +0000 (20:38 +0000)]
I had a need to read such things as the backend locale and the catalog
version number from the current database, and couldn't find any existing
program to do that.

linda:~$ pg_controldata
Log file id:                          0
Log file segment:                     5
Last modified:                        Wed Feb  7 19:35:47 2001
Database block size:                  8192
Blocks per segment of large relation: 131072
Catalog version number:               200101061
LC_COLLATE:                           en_GB
LC_CTYPE:                             en_GB
Log archive directory:

Oliver Elphick                                Oliver.Elphick@lfix.co.uk

23 years agoAs long as we're fixing this space calculation, let's actually do it
Tom Lane [Fri, 23 Feb 2001 20:12:37 +0000 (20:12 +0000)]
As long as we're fixing this space calculation, let's actually do it
right.  We should MAXALIGN the individual items because we'll
allocate them individually, not as an array.

23 years agoUpdate TODO list.
Bruce Momjian [Fri, 23 Feb 2001 19:38:43 +0000 (19:38 +0000)]
Update TODO list.

23 years agoUpdate comments on locks.
Bruce Momjian [Fri, 23 Feb 2001 19:24:06 +0000 (19:24 +0000)]
Update comments on locks.

23 years agoUpdate TODO list.
Bruce Momjian [Fri, 23 Feb 2001 19:21:53 +0000 (19:21 +0000)]
Update TODO list.

23 years agoUpdate TODO list.
Bruce Momjian [Fri, 23 Feb 2001 19:14:52 +0000 (19:14 +0000)]
Update TODO list.

23 years agoBruce Momjian <pgman@candle.pha.pa.us> writes:
Bruce Momjian [Fri, 23 Feb 2001 18:28:46 +0000 (18:28 +0000)]
Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Is there one LOCKMETHODCTL for every backend?  I thought there was only
> one of them.
>>
>> You're right, that line is erroneous; it should read
>>
>> size += MAX_LOCK_METHODS * MAXALIGN(sizeof(LOCKMETHODCTL));
>>
>> Not a significant error but it should be changed for clarity ...

23 years agoupdate to beta5 ... I think I ran cvs2cl right this time? :)
Marc G. Fournier [Fri, 23 Feb 2001 18:09:59 +0000 (18:09 +0000)]
update to beta5 ... I think I ran cvs2cl right this time? :)

23 years agoupdate to beta5
Marc G. Fournier [Fri, 23 Feb 2001 18:09:43 +0000 (18:09 +0000)]
update to beta5

23 years agoResolve a compile error.
Hiroshi Inoue [Fri, 23 Feb 2001 09:31:52 +0000 (09:31 +0000)]
Resolve a compile error.

23 years agoReindex of shared system indexes must be overwrite mode.
Hiroshi Inoue [Fri, 23 Feb 2001 09:26:14 +0000 (09:26 +0000)]
Reindex of shared system indexes must be overwrite mode.

23 years agoEnhanced UTF-8/SJIS mapping generator, contributed by
Tatsuo Ishii [Fri, 23 Feb 2001 08:44:33 +0000 (08:44 +0000)]
Enhanced UTF-8/SJIS mapping generator, contributed by
Eiji Tokuya" <e-tokuya@Mail.Sankyo-Unyu.co.jp>

23 years agoFreshen with a few recent platform reports:
Thomas G. Lockhart [Fri, 23 Feb 2001 02:17:27 +0000 (02:17 +0000)]
Freshen with a few recent platform reports:
 Linux/Alpha, Linux/Sparc, MacOS-X.