]> granicus.if.org Git - postgresql/log
postgresql
21 years agoApply the proper version of Christopher Kings-Lynne's describe patch
Tom Lane [Tue, 7 Jan 2003 20:56:07 +0000 (20:56 +0000)]
Apply the proper version of Christopher Kings-Lynne's describe patch
(ie, the one with describe-schema support).  Minor code review.
Adjust display of casts to use standard type names.

21 years agoAdd connection service file.
Bruce Momjian [Tue, 7 Jan 2003 20:29:30 +0000 (20:29 +0000)]
Add connection service file.

21 years agoSilence compiler warning caused by removal of netinet/in.h
Bruce Momjian [Tue, 7 Jan 2003 18:48:13 +0000 (18:48 +0000)]
Silence compiler warning caused by removal of netinet/in.h

21 years agoDocument that psql \encoding doesn't see SET CLIENT_ENCODING changes.
Bruce Momjian [Tue, 7 Jan 2003 18:46:52 +0000 (18:46 +0000)]
Document that psql \encoding doesn't see SET CLIENT_ENCODING changes.

21 years agoFix missing tab in file.
Bruce Momjian [Tue, 7 Jan 2003 15:12:17 +0000 (15:12 +0000)]
Fix missing tab in file.

21 years agoSmall syntax cleanup.
Bruce Momjian [Tue, 7 Jan 2003 06:43:20 +0000 (06:43 +0000)]
Small syntax cleanup.

21 years agoCheck for sockaddr_in6 for IPv6 code, rather than ip6.h.
Bruce Momjian [Tue, 7 Jan 2003 06:37:51 +0000 (06:37 +0000)]
Check for sockaddr_in6 for IPv6 code, rather than ip6.h.

21 years agoDocument libpq service capability, and add sample file.
Bruce Momjian [Tue, 7 Jan 2003 04:25:29 +0000 (04:25 +0000)]
Document libpq service capability, and add sample file.

21 years agoTweak mdnblocks() to avoid doing lseek() on segments that it has
Tom Lane [Tue, 7 Jan 2003 01:19:12 +0000 (01:19 +0000)]
Tweak mdnblocks() to avoid doing lseek() on segments that it has
previously determined not to be the last segment of a relation.
This reduces the expected cost to one seek, rather than one seek per
segment.  We can get away with this because truncation of a relation
will cause a relcache flush and so the md.c file descriptor will be
closed; when it is re-opened we will re-determine the last segment.

21 years agoEnable IPv6 libpq 'hostaddr' addresses. Update docs.
Bruce Momjian [Mon, 6 Jan 2003 22:48:16 +0000 (22:48 +0000)]
Enable IPv6 libpq 'hostaddr' addresses.  Update docs.

21 years agoUse our own version of getopt_long() if the OS doesn't have one.
Peter Eisentraut [Mon, 6 Jan 2003 18:53:25 +0000 (18:53 +0000)]
Use our own version of getopt_long() if the OS doesn't have one.

21 years agoFix for systems that don't have INET_ADDRSTRLEN.
Peter Eisentraut [Mon, 6 Jan 2003 09:58:36 +0000 (09:58 +0000)]
Fix for systems that don't have INET_ADDRSTRLEN.

21 years agoUpdate comments on IPv6 #define.
Bruce Momjian [Mon, 6 Jan 2003 06:07:20 +0000 (06:07 +0000)]
Update comments on IPv6  #define.

21 years agoAdd test for netinet/ip6.h and getaddrinfo, to enable IPv6.
Bruce Momjian [Mon, 6 Jan 2003 06:06:04 +0000 (06:06 +0000)]
Add test for netinet/ip6.h and getaddrinfo, to enable IPv6.

21 years agoUpdate ipv6 comment, move configure test to the right place
Bruce Momjian [Mon, 6 Jan 2003 05:46:18 +0000 (05:46 +0000)]
Update ipv6 comment, move configure test to the right place

21 years agoAdd prototype for getaddrinfo().
Bruce Momjian [Mon, 6 Jan 2003 05:33:45 +0000 (05:33 +0000)]
Add prototype for getaddrinfo().

21 years agoEnable IPv6 connections to the server, and add pg_hba.conf IPv6 entries
Bruce Momjian [Mon, 6 Jan 2003 03:18:27 +0000 (03:18 +0000)]
Enable IPv6 connections to the server, and add pg_hba.conf IPv6 entries
if the OS supports it.  Code will still compile on non-IPv6-aware
machines (feature added by Bruce).

Nigel Kukard

21 years agoAdd note explaining that a mergejoinable equality operator is now
Tom Lane [Mon, 6 Jan 2003 01:20:40 +0000 (01:20 +0000)]
Add note explaining that a mergejoinable equality operator is now
required if a datatype is to be accepted by GROUP BY, DISTINCT, or
ORDER BY.  This is documentation for code changes made pursuant to
pgsql-hackers discussion around 29-Nov-02.

21 years agoALTER DOMAIN OWNER, from Rod Taylor.
Tom Lane [Mon, 6 Jan 2003 00:31:45 +0000 (00:31 +0000)]
ALTER DOMAIN OWNER, from Rod Taylor.

21 years agoRevert rpath-mangling patch. See discussion on -patches around Nov 29 for
Peter Eisentraut [Sun, 5 Jan 2003 13:45:47 +0000 (13:45 +0000)]
Revert rpath-mangling patch.  See discussion on -patches around Nov 29 for
the rationale.

21 years agoFix inherited UPDATE for cases where child column numbering doesn't
Tom Lane [Sun, 5 Jan 2003 00:56:40 +0000 (00:56 +0000)]
Fix inherited UPDATE for cases where child column numbering doesn't
match parent table.  This used to work, but was broken in 7.3 by
rearrangement of code that handles targetlist sorting.  Add a regression
test to catch future breakage.

21 years agoPartial code review for ALTER DOMAIN patch. Incorporates Rod Taylor's
Tom Lane [Sat, 4 Jan 2003 00:46:08 +0000 (00:46 +0000)]
Partial code review for ALTER DOMAIN patch.  Incorporates Rod Taylor's
patches of 9-Dec (permissions fix) and 13-Dec (performance) as well as
a partial fix for locking issues: concurrent DROP COLUMN should not
create trouble anymore.  But concurrent DROP TABLE is still a risk, and
there is no protection at all against creating a column of a domain while
we are altering the domain.

21 years agoAdjust lo type in contrib during pg_restore so that pg_restore could
Hiroshi Inoue [Fri, 3 Jan 2003 18:05:02 +0000 (18:05 +0000)]
Adjust lo type in contrib during pg_restore so that pg_restore could
reload the type.

21 years agoflockfile/funlockfile is POSIX, so NetBSD has them too --- use them.
Bruce Momjian [Thu, 2 Jan 2003 23:22:49 +0000 (23:22 +0000)]
flockfile/funlockfile is POSIX, so NetBSD has them too --- use them.

21 years agoEnforces NOT NULL constraints to be applied against new PRIMARY KEY
Tom Lane [Thu, 2 Jan 2003 19:29:22 +0000 (19:29 +0000)]
Enforces NOT NULL constraints to be applied against new PRIMARY KEY
columns in DefineIndex.  So, ALTER TABLE ... PRIMARY KEY will now
automatically add the NOT NULL constraint.  It appeared the alter_table
regression test wanted this to occur, as after the change the regression
test better matched in inline 'fails'/'succeeds' comments.

Rod Taylor

21 years agoFix BSD/OS thread file locking bug in new fseeko code.
Bruce Momjian [Thu, 2 Jan 2003 06:00:33 +0000 (06:00 +0000)]
Fix BSD/OS thread file locking bug in new fseeko code.

Backpatch to 7.3.X.

21 years agofastpath code neglected to check whether user has privileges to call the
Tom Lane [Wed, 1 Jan 2003 21:57:05 +0000 (21:57 +0000)]
fastpath code neglected to check whether user has privileges to call the
target function.  Also, move SetQuerySnapshot() call to avoid assert
failure when a fastpath call is attempted in an aborted transaction.

21 years agoAwhile back I wrote that freebsd.h was probably broken in the places where
Tom Lane [Wed, 1 Jan 2003 20:35:39 +0000 (20:35 +0000)]
Awhile back I wrote that freebsd.h was probably broken in the places where
it diverged from netbsd.h and openbsd.h.  This has now been confirmed.
Accordingly, make all three exactly alike.

21 years agoUpdate:
Bruce Momjian [Wed, 1 Jan 2003 19:04:07 +0000 (19:04 +0000)]
Update:

< * Have DEFAULT dependency track use of sequence, for DROP DEFAULT check
> * Disallow changing default expression of a SERIAL column

21 years agoAdd:
Bruce Momjian [Tue, 31 Dec 2002 18:06:03 +0000 (18:06 +0000)]
Add:

> * Have DEFAULT dependency track use of sequence, for DROP DEFAULT check

21 years agoAdd casts between lo and oid.
Hiroshi Inoue [Tue, 31 Dec 2002 10:22:03 +0000 (10:22 +0000)]
Add casts between lo and oid.

21 years agoAdjust Tcl-related code to compile cleanly with Tcl 8.4 (add const modifiers as
Tom Lane [Mon, 30 Dec 2002 22:10:54 +0000 (22:10 +0000)]
Adjust Tcl-related code to compile cleanly with Tcl 8.4 (add const modifiers as
needed).  Some desultory const-ification of SPI interface to support this.

21 years agoAttached is a patch to provide makefiles, etc. to allow the compilation
Bruce Momjian [Mon, 30 Dec 2002 21:07:26 +0000 (21:07 +0000)]
Attached is a patch to provide makefiles, etc. to allow the compilation
of the libpq interface static and dynamic libraries with the freely
downloadable Borland C++ compiler version 5.5 and/or C++ Builder.

Lester Godwin

21 years agoCode review for transaction-safe-TRUNCATE patch: minor cleanups.
Tom Lane [Mon, 30 Dec 2002 19:45:17 +0000 (19:45 +0000)]
Code review for transaction-safe-TRUNCATE patch: minor cleanups.

21 years agoCode review for CLUSTER ALL patch. Fix bogus locking, incorrect transaction
Tom Lane [Mon, 30 Dec 2002 18:42:17 +0000 (18:42 +0000)]
Code review for CLUSTER ALL patch.  Fix bogus locking, incorrect transaction
stop/start nesting, other infelicities.

21 years agoMake use of TCL_INCLUDE_SPEC if available (it's new in Tcl 8.4, too bad
Tom Lane [Mon, 30 Dec 2002 17:19:54 +0000 (17:19 +0000)]
Make use of TCL_INCLUDE_SPEC if available (it's new in Tcl 8.4, too bad
it took 'em this long to realize it's needed...)

21 years agoCause FETCH 1 to return the current cursor row, or zero if at
Bruce Momjian [Mon, 30 Dec 2002 15:31:51 +0000 (15:31 +0000)]
Cause FETCH 1 to return the current cursor row, or zero if at
beginning/end of cursor.

Have MOVE return 0/1 depending on cursor position.

Matches SQL spec.

Pass cursor counter from parser as a long rather than int.

Doc updates.

21 years agoBetter solution to integer overflow problem in hash batch-number
Tom Lane [Mon, 30 Dec 2002 15:21:23 +0000 (15:21 +0000)]
Better solution to integer overflow problem in hash batch-number
computation: reduce the bucket number mod nbatch.  This changes the
association between original bucket numbers and batches, but that
doesn't matter.  Minor other cleanups in hashjoin code to help
centralize decisions.

21 years agoAdd missing slash to python install path.
Bruce Momjian [Mon, 30 Dec 2002 05:06:50 +0000 (05:06 +0000)]
Add missing slash to python install path.

21 years agoAdd:
Bruce Momjian [Mon, 30 Dec 2002 04:48:57 +0000 (04:48 +0000)]
Add:

> * Allow infinite dates just like infinite timestamps

21 years agoPoint does not return a center of lseg.
Tom Lane [Mon, 30 Dec 2002 02:18:29 +0000 (02:18 +0000)]
Point does not return a center of lseg.

Kenji Sugita

21 years agoAdjust hash table sizing algorithm to avoid integer overflow in
Tom Lane [Sun, 29 Dec 2002 22:28:50 +0000 (22:28 +0000)]
Adjust hash table sizing algorithm to avoid integer overflow in
ExecHashJoinGetBatch().  Fixes core dump on large hash joins, as in
example from Rae Stiening.

21 years agoClean up quoting in configure.in. configure did not change.
Bruce Momjian [Sun, 29 Dec 2002 03:56:35 +0000 (03:56 +0000)]
Clean up quoting in configure.in.  configure did not change.

21 years agoDeliver better error message when a relation name is used in an expression.
Tom Lane [Fri, 27 Dec 2002 20:06:19 +0000 (20:06 +0000)]
Deliver better error message when a relation name is used in an expression.
Per report from Ian Barwick.

21 years agoRemove overenthusiastic free'ing of comment dependencies; could lead to
Tom Lane [Fri, 27 Dec 2002 17:10:45 +0000 (17:10 +0000)]
Remove overenthusiastic free'ing of comment dependencies; could lead to
core dump in pg_dump when dumping views having comments.  See bug #855.

21 years agoSmall fix in documentation and some examples of usage. Please, apply to
Bruce Momjian [Fri, 27 Dec 2002 14:10:29 +0000 (14:10 +0000)]
Small fix in documentation and some examples of usage. Please, apply to
7.3 and current CVS

Oleg Bartunov

21 years agoTrivial patch to increase max_fsm_pages as per earlier discussion.
Bruce Momjian [Fri, 27 Dec 2002 14:06:34 +0000 (14:06 +0000)]
Trivial patch to increase max_fsm_pages as per earlier discussion.

Philip Warner

21 years agoAdded:
Bruce Momjian [Fri, 27 Dec 2002 05:56:53 +0000 (05:56 +0000)]
Added:

> * Two-phase commit to implement distributed transactions

21 years agoClamp the output of estimate_hash_bucketsize() to a sane range;
Tom Lane [Thu, 26 Dec 2002 23:38:42 +0000 (23:38 +0000)]
Clamp the output of estimate_hash_bucketsize() to a sane range;
per example from Bruno Wolff in which it produced a silly result.

21 years agoDon't try to free executor state of an InitPlan early --- this breaks
Tom Lane [Thu, 26 Dec 2002 22:37:42 +0000 (22:37 +0000)]
Don't try to free executor state of an InitPlan early --- this breaks
EXPLAIN ANALYZE.  (Premature optimization is the root of all evil?)

21 years agoAdd developer organizations:
Bruce Momjian [Tue, 24 Dec 2002 22:06:53 +0000 (22:06 +0000)]
Add developer organizations:

< * Bruce is Bruce Momjian <pgman@candle.pha.pa.us>
< * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au>
< * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net>
> * Bruce is Bruce Momjian <pgman@candle.pha.pa.us> of Software Research Assoc.
> * Christopher is Christopher Kings-Lynne <chriskl@familyhealth.com.au> of
>     Family Health Network
> * D'Arcy is D'Arcy J.M. Cain <darcy@druid.net> of The Cain Gang Ltd.
460,461c461,462
< * Fernando Nasser <fnasser@redhat.com>
< * Gavin Sherry <swm@linuxworld.com.au>
> * Fernando Nasser <fnasser@redhat.com> of Red Hat
> * Gavin Sherry <swm@linuxworld.com.au> of Alcove Systems Engineering
464,466c465,467
< * Jan is Jan Wieck <wieck@sapserv.debis.de>
< * Liam is Liam Stewart <liams@redhat.com>
< * Marc is Marc Fournier <scrappy@hub.org>
> * Jan is Jan Wieck <JanWieck@Yahoo.com> of PeerDirect Corp.
> * Liam is Liam Stewart <liams@redhat.com> of Red Hat
> * Marc is Marc Fournier <scrappy@hub.org> of PostgreSQL, Inc.
468,469c469
< * Marko is Marko Kreen <marko@l-t.ee>
< * Michael is Michael Meskes <meskes@postgresql.org>
> * Michael is Michael Meskes <meskes@postgresql.org> of Credativ
472c472
< * Peter M is Peter T Mount <peter@retep.org.uk>
> * Peter M is Peter T Mount <peter@retep.org.uk> of Retep Software
474c474
< * Philip is Philip Warner <pjw@rhyme.com.au>
> * Philip is Philip Warner <pjw@rhyme.com.au> of Albatross Consulting Pty. Ltd.
477d476
< * Ryan is Ryan Bradetich <rbrad@hpb50023.boi.hp.com>
479,483c478,481
< * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp>
< * Thomas is Thomas Lockhart <lockhart@fourpalms.org>
< * Tom is Tom Lane <tgl@sss.pgh.pa.us>
< * TomH is Tom I Helbekkmo <tih@Hamartun.Priv.no>
< * Vadim is Vadim B. Mikheev <vadim4o@email.com>
> * Tatsuo is Tatsuo Ishii <t-ishii@sra.co.jp> of Software Research Assoc.
> * Thomas is Thomas Lockhart <lockhart@fourpalms.org> of Jet Propulsion Labratory
> * Tom is Tom Lane <tgl@sss.pgh.pa.us> of Red Hat
> * Vadim is Vadim B. Mikheev <vadim4o@email.com> of Sector Data

21 years agoUpdate SSL error message for improper permissions. Backpatch to 7.3.X.
Bruce Momjian [Mon, 23 Dec 2002 22:19:00 +0000 (22:19 +0000)]
Update SSL error message for improper permissions.  Backpatch to 7.3.X.

21 years agocase sensitive updates by Kris Jurka
Dave Cramer [Mon, 23 Dec 2002 16:12:36 +0000 (16:12 +0000)]
case sensitive updates by Kris Jurka

21 years agopg_dump should consider information_schema to be a system schema.
Tom Lane [Sat, 21 Dec 2002 22:45:09 +0000 (22:45 +0000)]
pg_dump should consider information_schema to be a system schema.
Also, tweak -C option (emit CREATE DATABASE command) to emit encoding
name rather than encoding number, for consistency with pg_dumpall
and better cross-version portability.

21 years agoFix possible buffer overrun in \d command: substr(..., 128) produces
Tom Lane [Sat, 21 Dec 2002 01:07:07 +0000 (01:07 +0000)]
Fix possible buffer overrun in \d command: substr(..., 128) produces
a result of at most 128 chars, but that could be more than 128 bytes.
Also ensure we don't try to pfree uninitialized pointers during error
cleanup.

21 years agomike beachy's patch for statement handling
Dave Cramer [Fri, 20 Dec 2002 14:10:34 +0000 (14:10 +0000)]
mike beachy's patch for statement handling

21 years agopatch for null table in getPrimaryKeys
Dave Cramer [Fri, 20 Dec 2002 13:15:53 +0000 (13:15 +0000)]
patch for null table in getPrimaryKeys

21 years agoUpdate 7.3.1 to give proper attribution to pg_dump item.
Bruce Momjian [Fri, 20 Dec 2002 00:24:00 +0000 (00:24 +0000)]
Update 7.3.1 to give proper attribution to pg_dump item.

21 years agoCause planner to include costs of InitPlans in startup cost of parent
Tom Lane [Thu, 19 Dec 2002 23:25:01 +0000 (23:25 +0000)]
Cause planner to include costs of InitPlans in startup cost of parent
plan node.  Not sure why this oversight has persisted so long ...

21 years agopgindent fe-connect.c --- done to make IPv6 patch easier to apply.
Bruce Momjian [Thu, 19 Dec 2002 19:30:24 +0000 (19:30 +0000)]
pgindent fe-connect.c --- done to make IPv6 patch easier to apply.

21 years agoMVCC doc improvements:
Bruce Momjian [Wed, 18 Dec 2002 20:40:24 +0000 (20:40 +0000)]
MVCC doc improvements:

> I'm not objecting to improving the text.  I am objecting to deleting it
> outright...

Ok, fair enough. I've attached a revised version of the patch -- let me
know you think it needs further improvements.

Neil Conway

21 years agoUpdate Russian FAQ. Update all FAQ's for 7.3.X branch.
Bruce Momjian [Wed, 18 Dec 2002 20:38:48 +0000 (20:38 +0000)]
Update Russian FAQ.  Update all FAQ's for 7.3.X branch.

21 years agoUpdate RELEASE checklist.
Bruce Momjian [Wed, 18 Dec 2002 20:07:32 +0000 (20:07 +0000)]
Update RELEASE checklist.

21 years agoAdd some items to development-branch release notes.
Tom Lane [Wed, 18 Dec 2002 19:01:25 +0000 (19:01 +0000)]
Add some items to development-branch release notes.

21 years agoFrom the SSL_CTX_new man page:
PostgreSQL Daemon [Wed, 18 Dec 2002 13:15:15 +0000 (13:15 +0000)]
From the SSL_CTX_new man page:

"SSLv23_method(void), SSLv23_server_method(void), SSLv23_client_method(void)

 A TLS/SSL connection established with these methods will understand the SSLv2,
 SSLv3, and TLSv1 protocol. A client will send out SSLv2 client hello messages
 and will indicate that it also understands SSLv3 and TLSv1. A server will
 understand SSLv2, SSLv3, and TLSv1 client hello messages. This is the best
 choice when compatibility is a concern."

This will maintain backwards compatibility for those us that don't use
TLS connections ...

21 years ago*** empty log message ***
Bruce Momjian [Wed, 18 Dec 2002 04:56:30 +0000 (04:56 +0000)]
*** empty log message ***

21 years agoincorporates latest changes in the English original.
Bruce Momjian [Wed, 18 Dec 2002 04:46:17 +0000 (04:46 +0000)]
incorporates latest changes in the English original.

Ian Barwick

21 years agoStamp 7.3.1
Bruce Momjian [Wed, 18 Dec 2002 04:39:48 +0000 (04:39 +0000)]
Stamp 7.3.1

21 years agoStamp 7.3.1.
Bruce Momjian [Wed, 18 Dec 2002 02:06:23 +0000 (02:06 +0000)]
Stamp 7.3.1.

21 years agoUpdate EvalPlanQual() to work with new executor memory management method.
Tom Lane [Wed, 18 Dec 2002 00:14:47 +0000 (00:14 +0000)]
Update EvalPlanQual() to work with new executor memory management method.
It doesn't leak memory anymore ...

21 years agoPrevent core dump when ExecMarkPos is called before any tuples have been
Tom Lane [Wed, 18 Dec 2002 00:14:24 +0000 (00:14 +0000)]
Prevent core dump when ExecMarkPos is called before any tuples have been
retrieved.  This cannot happen in ordinary execution, but it can happen
under EvalPlanQual().

21 years agoImprovement to CREATE TRIGGER and catalog documentation.
Bruce Momjian [Tue, 17 Dec 2002 17:41:30 +0000 (17:41 +0000)]
Improvement to CREATE TRIGGER and catalog documentation.

Neil Conway

21 years agoSkip unnecessary plan-copying now that plan trees are read-only in the
Tom Lane [Tue, 17 Dec 2002 15:51:59 +0000 (15:51 +0000)]
Skip unnecessary plan-copying now that plan trees are read-only in the
executor.

21 years agoFix typo.
Tom Lane [Tue, 17 Dec 2002 15:45:01 +0000 (15:45 +0000)]
Fix typo.

21 years agoTo suppress memory leakage in long-lived Lists, lremove() should pfree
Tom Lane [Tue, 17 Dec 2002 01:18:35 +0000 (01:18 +0000)]
To suppress memory leakage in long-lived Lists, lremove() should pfree
the cons cell it's deleting from the list.  Do this, and fix a few callers
that were bogusly assuming it wouldn't free the cons cell.

21 years agoFix GEQO to work again in CVS tip, by being more careful about memory
Tom Lane [Mon, 16 Dec 2002 21:30:30 +0000 (21:30 +0000)]
Fix GEQO to work again in CVS tip, by being more careful about memory
allocation in best_inner_indexscan().  While at it, simplify GEQO's
interface to the main planner --- make_join_rel() offers exactly the
API it really wants, whereas calling make_rels_by_clause_joins() and
make_rels_by_clauseless_joins() required jumping through hoops.
Rewrite gimme_tree for clarity (sometimes iteration is much better than
recursion), and approximately halve GEQO's runtime by recognizing that
tours of the forms (a,b,c,d,...) and (b,a,c,d,...) are equivalent
because of symmetry in make_join_rel().

21 years agoDocument that zero-column tables are allowed (and non-standard).
Tom Lane [Mon, 16 Dec 2002 19:08:25 +0000 (19:08 +0000)]
Document that zero-column tables are allowed (and non-standard).

21 years agoUpdate release to mark 7.3.
Bruce Momjian [Mon, 16 Dec 2002 18:40:32 +0000 (18:40 +0000)]
Update release to mark 7.3.

21 years agoFix ALTER TABLE ADD COLUMN to disallow the same column types that are
Tom Lane [Mon, 16 Dec 2002 18:39:22 +0000 (18:39 +0000)]
Fix ALTER TABLE ADD COLUMN to disallow the same column types that are
disallowed by CREATE TABLE (eg, pseudo-types); also disallow these types
from being introduced by the range-function syntax.  While at it, allow
CREATE TABLE to create zero-column tables, per recent pghackers discussion.
I am back-patching this into 7.3 since failure to disallow pseudo-types
is arguably a security hole.

21 years agoCode review for palloc0 patch --- avoid dangerous and unnecessary
Tom Lane [Mon, 16 Dec 2002 16:22:46 +0000 (16:22 +0000)]
Code review for palloc0 patch --- avoid dangerous and unnecessary
practice of evaluating MemSet's arguments multiple times, except for
the special case of newNode(), where we can assume the argument is
a constant sizeof() operator.
Also, add GetMemoryChunkContext() to mcxt.c's API, in preparation for
fixing recent GEQO breakage.

21 years agoTweak default memory context allocation policy so that a context is not
Tom Lane [Sun, 15 Dec 2002 21:01:34 +0000 (21:01 +0000)]
Tweak default memory context allocation policy so that a context is not
given any malloc block until something is first allocated in it; but
thereafter, MemoryContextReset won't release that first malloc block.
This preserves the quick-reset property of the original policy, without
forcing 8K to be allocated to every context whether any of it is ever
used or not.  Also, remove some more no-longer-needed explicit freeing
during ExecEndPlan.

21 years agoRevise executor APIs so that all per-query state structure is built in
Tom Lane [Sun, 15 Dec 2002 16:17:59 +0000 (16:17 +0000)]
Revise executor APIs so that all per-query state structure is built in
a per-query memory context created by CreateExecutorState --- and destroyed
by FreeExecutorState.  This provides a final solution to the longstanding
problem of memory leaked by various ExecEndNode calls.

21 years agoImprove comment.
Bruce Momjian [Sun, 15 Dec 2002 03:16:58 +0000 (03:16 +0000)]
Improve comment.

21 years agoAdd major/minor release changes info to RELEASE_CHANGES file.
Bruce Momjian [Sat, 14 Dec 2002 19:45:46 +0000 (19:45 +0000)]
Add major/minor release changes info to RELEASE_CHANGES file.

21 years agoThe actual segfault was caused by a double pfree(), but ISTM that
Bruce Momjian [Sat, 14 Dec 2002 18:49:37 +0000 (18:49 +0000)]
The actual segfault was caused by a double pfree(), but ISTM that
failing to find pg_hba.conf should be a fatal error anyway, so I
increased the priority of the elog() from LOG to FATAL and refactored
the code a little bit.

Neil Conway

21 years agoOk, I think I've gotten this figured out now. I saw this comment in
Bruce Momjian [Sat, 14 Dec 2002 18:39:14 +0000 (18:39 +0000)]
Ok, I think I've gotten this figured out now.  I saw this comment in
pqcomm.c, switched the ERROR logs to COMMERROR logs and it all works.
I've attached a patch to be-secure.c that fixes all my problems.

Nathan Mueller

21 years agoInitial version of the SQL information schema
Peter Eisentraut [Sat, 14 Dec 2002 00:24:35 +0000 (00:24 +0000)]
Initial version of the SQL information schema

21 years agoClean up plantree representation of SubPlan-s --- SubLink does not appear
Tom Lane [Sat, 14 Dec 2002 00:17:59 +0000 (00:17 +0000)]
Clean up plantree representation of SubPlan-s --- SubLink does not appear
in the planned representation of a subplan at all any more, only SubPlan.
This means subselect.c doesn't scribble on its input anymore, which seems
like a good thing; and there are no longer three different possible
interpretations of a SubLink.  Simplify node naming and improve comments
in primnodes.h.  No change to stored rules, though.

21 years agoIncrement libpq major number for 7.3.1 and minor for 7.4.
Bruce Momjian [Fri, 13 Dec 2002 22:17:57 +0000 (22:17 +0000)]
Increment libpq major number for 7.3.1 and minor for 7.4.

21 years agoAdmit defeat on the prospect of keeping straight exactly which platforms
Tom Lane [Fri, 13 Dec 2002 20:35:57 +0000 (20:35 +0000)]
Admit defeat on the prospect of keeping straight exactly which platforms
produce which output in the geometry test, even with the problem narrowed
down to only whether they print minus zero or not.  Instead, use
pg_regress' locale-variant mechanism to automatically consider the test
to pass if it matches either supplied comparison file.  geometry_1.out
replaces the former geometry-positive-zeros.out.

21 years agoProtect from null param on ecpg disconnect, verified by Roland Karch
Bruce Momjian [Fri, 13 Dec 2002 20:29:07 +0000 (20:29 +0000)]
Protect from null param on ecpg disconnect, verified by Roland Karch

21 years agoAdd a basic regression test for IS DISTINCT FROM, which has spent way too
Tom Lane [Fri, 13 Dec 2002 20:16:11 +0000 (20:16 +0000)]
Add a basic regression test for IS DISTINCT FROM, which has spent way too
much time in a broken state for lack of anyone noticing.

21 years agoPhase 3 of read-only-plans project: ExecInitExpr now builds expression
Tom Lane [Fri, 13 Dec 2002 19:46:01 +0000 (19:46 +0000)]
Phase 3 of read-only-plans project: ExecInitExpr now builds expression
execution state trees, and ExecEvalExpr takes an expression state tree
not an expression plan tree.  The plan tree is now read-only as far as
the executor is concerned.  Next step is to begin actually exploiting
this property.

21 years agoAdjust costsize calculations to avoid introducing unnecessary roundoff
Tom Lane [Fri, 13 Dec 2002 17:29:25 +0000 (17:29 +0000)]
Adjust costsize calculations to avoid introducing unnecessary roundoff
error.  This seems to explain the differing choice of plan that's been
causing geometry regress test to fail for the last few days.

21 years agoRemove strerror output for openssl SYSCALL error check.
Bruce Momjian [Fri, 13 Dec 2002 05:51:29 +0000 (05:51 +0000)]
Remove strerror output for openssl SYSCALL error check.

21 years agoThis patch fixes minor bugs in dictionary generator in contrib/tsearch
Bruce Momjian [Fri, 13 Dec 2002 05:50:18 +0000 (05:50 +0000)]
This patch fixes minor bugs in dictionary generator in contrib/tsearch
(contrib/tsearch/makedict/makedict.pl)

[ Backpatched to 7.3.]

Teodor Sigaev

21 years agoThis patch fixes a few typos in the documentation for the '~/.pgpass'
Bruce Momjian [Fri, 13 Dec 2002 05:49:10 +0000 (05:49 +0000)]
This patch fixes a few typos in the documentation for the '~/.pgpass'
libpq feature.

Neil Conway

21 years agoHere's the patch I mentioned on HACKERS to fix up ALTER command tab
Bruce Momjian [Fri, 13 Dec 2002 05:36:24 +0000 (05:36 +0000)]
Here's the patch I mentioned on HACKERS to fix up ALTER command tab
completion. Note that it's based on 7.3 tarball, not CVS HEAD, or 7.3rel
branch. Damn, looking at CVS, this will patch into 7.3rel (just tested,
it does) probably collide with Rod Taylor's patch adding ALTER TRIGGER
stuff. O.K, second patch attached against HEAD - not tested, hand
merged.

Ross Reedstrom

21 years ago> There is an ugly little problem with the DB wrapper class.
Bruce Momjian [Thu, 12 Dec 2002 22:49:27 +0000 (22:49 +0000)]
> There is an ugly little problem with the DB wrapper class.
>
> In pg.py the attributes of DB are defined as being the same as
> the attributes of the corresponding pgobject "db", using the following

...

> The problem is that the attributes of db (which are read only)
> are not static (they are actually function calls to PostgreSQL),
> especially "status" and "error", but those attributes are copied
> and this is done only once when initializing the DB object.
>
> So, in effect, only the attribute "db.error" of a DB instance
> will be updated, but not the attribute "error". Same with "status".

> Don't copy the (read only) attributes of the pgobject to the
> DB object, but only the methods, and all of them, like this:
>
> --------------- change in pg.py ------------------
> # Create convience methods, in a way that is still overridable.
> for e in self.db.__methods__:
>  setattr(self, e, getattr(self.db, e))
> ----------------------------------------------------
>
> Furthermore, make an addition to the documentation of the
> DB wrapper class (i.e. in pygresql-pg-db.html):
> After the sentence "All pgobject methods are included in this class also."
> add the following sentence "The pgobject read-only attributes can be
> accessed py adding the prefix 'db.' to them."

Christoph Zwerschke

21 years agoCheck SSL_get_error() value SSL_ERROR_SYSCALL to see if SSL_read()
Bruce Momjian [Thu, 12 Dec 2002 22:42:39 +0000 (22:42 +0000)]
Check SSL_get_error() value SSL_ERROR_SYSCALL to see if SSL_read()
returned -1, per SSL_get_error() documentation.

Nathan Mueller