OSDN Git Service

pg-rex/syncrep.git
21 years agoTranslation updates
Peter Eisentraut [Mon, 2 Sep 2002 22:22:12 +0000 (22:22 +0000)]
Translation updates

21 years agoAvoid shell syntax error if list of languages is empty.
Peter Eisentraut [Mon, 2 Sep 2002 22:19:42 +0000 (22:19 +0000)]
Avoid shell syntax error if list of languages is empty.

21 years agoFix typos.
Peter Eisentraut [Mon, 2 Sep 2002 22:18:56 +0000 (22:18 +0000)]
Fix typos.

21 years agoConvert control status codes to human-readable strings.
Peter Eisentraut [Mon, 2 Sep 2002 22:18:26 +0000 (22:18 +0000)]
Convert control status codes to human-readable strings.

21 years agopg_on_connection_loss command for libpgtcl. Patch from
Tom Lane [Mon, 2 Sep 2002 21:51:47 +0000 (21:51 +0000)]
pg_on_connection_loss command for libpgtcl.  Patch from
Gerhard Hintermayer, revised and documented by Tom Lane.

This patch also fixes a 'must fix' bug: libpgtcl's LISTEN/NOTIFY
support was broken by the recent changes to the PGnotify structure.
Guess that change wasn't quite so safe as we thought.

21 years agoFix obsolete description of PGnotify struct.
Tom Lane [Mon, 2 Sep 2002 20:21:32 +0000 (20:21 +0000)]
Fix obsolete description of PGnotify struct.

21 years agoDepartment of second thoughts: make checks for replacing a view slightly
Tom Lane [Mon, 2 Sep 2002 20:04:40 +0000 (20:04 +0000)]
Department of second thoughts: make checks for replacing a view slightly
more flexible, and improve the error reporting.  Also, add documentation
for REPLACE RULE/VIEW.

21 years agoNo more need for 'privileges' file.
Bruce Momjian [Mon, 2 Sep 2002 19:02:02 +0000 (19:02 +0000)]
No more need for 'privileges' file.

21 years agoImprove CVS log grouping.
Bruce Momjian [Mon, 2 Sep 2002 17:34:57 +0000 (17:34 +0000)]
Improve CVS log grouping.

21 years agoUpdated from Tom:
Bruce Momjian [Mon, 2 Sep 2002 16:21:30 +0000 (16:21 +0000)]
Updated from Tom:

<  o -ALTER TABLE ALTER COLUMN SET/DROP NOT NULL (Christopher Kings-Lynne)
<  o ALTER TABLE ADD COLUMN column SERIAL doesn't create sequence
>  o -ALTER TABLE ALTER COLUMN SET/DROP NOT NULL (Christopher)
200a200,201
>  o ALTER TABLE ADD COLUMN column SERIAL doesn't create sequence because
>           of the item above
232c233
<  o -Disallow missing columns in INSERT ... VALUES, per ANSI
>  o -Disallow missing columns in INSERT ... (col) VALUES, per ANSI
335,336d335
< * Have SERIAL generate non-colliding sequence names when we have
<   auto-destruction

21 years agoAdd check for Ant version >= 1.5.
Bruce Momjian [Mon, 2 Sep 2002 16:14:01 +0000 (16:14 +0000)]
Add check for Ant version >= 1.5.

21 years agoUpdates from Tom:
Bruce Momjian [Mon, 2 Sep 2002 15:30:50 +0000 (15:30 +0000)]
Updates from Tom:

< * Allow logging of query durations
> * -Allow logging of query durations
51,52d50
< * Make single-user local access permissions the default by limiting
<   permissions on the socket file (Peter E)
72,73c70,71
< * Reserve last few process slots for super-user if max_connections reached
< * Add GUC parameter to print queries that generate errors
> * -Reserve last few process slots for super-user if max_connections reached
> * -Add GUC parameter to print queries that generate errors
82,83c80,81
< * Declare typein/out functions in pg_proc with a special "C string" data type
< * Functions returning sets do not totally work
> * -Declare typein/out functions in pg_proc with a special "C string" data type
> * -Functions returning sets do not totally work
90c88
< * Allow bytea to handle LIKE with non-TEXT patterns
> * -Allow bytea to handle LIKE with non-TEXT patterns
94c92
<  o Store binary-compatible type information in the system
>  o -Store binary-compatible type information in the system
97d94
<  o -SELECT col FROM tab WHERE numeric_col = 10.1 fails, requires quotes
102c99
<  o Ensure we have array-eq operators for every built-in array type
>  o -Ensure we have array-eq operators for every built-in array type
119d115
< * Allow setting database character set without multibyte enabled
152d147
< * Have UPDATE/DELETE clean out indexes
198,199d192
<  o ALTER TABLE ADD COLUMN to inherited table put column in wrong place
<    [inheritance]
201d193
<  o Add ALTER FUNCTION
203,204c195,196
<  o -ALTER TABLE ADD PRIMARY KEY (Tom)
<  o -ALTER TABLE ADD UNIQUE (Tom)
>  o -ALTER TABLE ADD PRIMARY KEY (Christopher)
>  o -ALTER TABLE ADD UNIQUE (Christopher)
248c240
<  o -Remove SET KSQO option now that OR processing is improved (Tom)
>  o -Remove SET KSQO option now that OR processing is improved (Bruce)
280c272
< * Have pg_dump use LEFT OUTER JOIN in multi-table SELECTs
> * -Have pg_dump use LEFT OUTER JOIN in multi-table SELECTs
287d278
< * Add config file check for $ODBCINI, $HOME/.odbc.ini, installpath/etc/odbc.ini
318,322d308
< * ODBC
<  o ODBC 3.0 support
<  o Unicode(UCS-2) support
<  o Updatable cursors support
<
337c323
< * Fix foreign key constraints to not error on intermediate db states (Stephan)
> * -Fix foreign key constraints to not error on intermediate db states (Stephan)
352c338
< * Propagate column or table renaming to foreign key constraints
> * -Propagate column or table renaming to foreign key constraints
447c433
< * Remove wal_files postgresql.conf option because WAL files are now recycled
> * -Remove wal_files postgresql.conf option because WAL files are now recycled
460c446
< * Improve dynamic memory allocation by introducing tuple-context memory
> * -Improve dynamic memory allocation by introducing tuple-context memory
463c449
< * Nested FULL OUTER JOINs don't work (Tom)
> * -Nested FULL OUTER JOINs don't work (Tom)

21 years agoFix breakage introduced by evidently-completely-untested snprintf patch.
Tom Lane [Mon, 2 Sep 2002 14:43:14 +0000 (14:43 +0000)]
Fix breakage introduced by evidently-completely-untested snprintf patch.

21 years agoFix broken markup.
Tom Lane [Mon, 2 Sep 2002 13:45:30 +0000 (13:45 +0000)]
Fix broken markup.

21 years agoFix conversion regressin test failure.
Tatsuo Ishii [Mon, 2 Sep 2002 13:27:20 +0000 (13:27 +0000)]
Fix conversion regressin test failure.

21 years agoDone:
Bruce Momjian [Mon, 2 Sep 2002 06:44:22 +0000 (06:44 +0000)]
Done:

> * -Add OR REPLACE clauses to non-FUNCTION object creation
> * -Allow autocommit so always in a transaction block
> * -Cache most recent query plan(s) (Neil) [prepare]

21 years agoAdd missing dblink files.
Bruce Momjian [Mon, 2 Sep 2002 06:32:41 +0000 (06:32 +0000)]
Add missing dblink files.

21 years agoFixes for intagg regression tests to match new array internal handling.
Bruce Momjian [Mon, 2 Sep 2002 06:28:06 +0000 (06:28 +0000)]
Fixes for intagg regression tests to match new array internal handling.

mlw

21 years agoIn case Florian and I don't finish his changes to this contrib before
Bruce Momjian [Mon, 2 Sep 2002 06:27:04 +0000 (06:27 +0000)]
In case Florian and I don't finish his changes to this contrib before
beta, at least get this stuff in.

ftipatch.txt - Updates to docs and scripts.  Run in the fulltextindexdir
WARNING - Add to fulltextindex dir
uninstall.sql - Add to fulltextindex dir

21 years agoUpdate catversion so we are sure everyone gets the bytea LIKE fix.
Bruce Momjian [Mon, 2 Sep 2002 06:24:15 +0000 (06:24 +0000)]
Update catversion so we are sure everyone gets the bytea LIKE fix.

21 years ago> Okay. When you get back to the original issue, the gold is hidden in
Bruce Momjian [Mon, 2 Sep 2002 06:22:20 +0000 (06:22 +0000)]
> Okay.  When you get back to the original issue, the gold is hidden in
> src/backend/optimizer/path/indxpath.c; see the "special indexable
> operators" stuff near the bottom of that file.  (It's a bit of a crock
> that this code is hardwired there, and not somehow accessed through a
> system catalog, but it's what we've got at the moment.)

The attached patch re-enables a bytea right hand argument (as compared
to a text right hand argument), and enables index usage, for bytea LIKE

Joe Conway

21 years agoThis patch fixes a minor inaccuracy in the documentation: NOT NULL is
Bruce Momjian [Mon, 2 Sep 2002 06:20:53 +0000 (06:20 +0000)]
This patch fixes a minor inaccuracy in the documentation: NOT NULL is
not synonymous with CHECK (xxx IS NOT NULL) -- for example, consider
ALTER TABLE ADD PRIMARY KEY, which checks for 'NOT NULL', not a check
constraint.

Neil Conway

21 years agoTab completion in psql is missing the REINDEX command.
Bruce Momjian [Mon, 2 Sep 2002 06:19:38 +0000 (06:19 +0000)]
Tab completion in psql is missing the REINDEX command.

Alvaro Herrera

21 years agoAttached is a fairly sizeable update to contrib/dblink. I'd love to get
Bruce Momjian [Mon, 2 Sep 2002 06:13:31 +0000 (06:13 +0000)]
Attached is a fairly sizeable update to contrib/dblink. I'd love to get
review/feedback if anyone is interested and can spend the time. But I'd
also love to get this committed and address changes as incremental
patches ;-), so if there are no objections, please apply.

Below I'll give a synopsis of the changes. More detailed descriptions
are now in a new doc directory under contrib/dblink. There is also a new

dblink.test.sql file which will give a pretty good overview of the
functions and their use.

Joe Conway

21 years agoI checked all the previous string handling errors and most of them were
Bruce Momjian [Mon, 2 Sep 2002 06:11:43 +0000 (06:11 +0000)]
I checked all the previous string handling errors and most of them were
already fixed by You. However there were a few left and attached patch
should fix the rest of them.

I used StringInfo only in 2 places and both of them are inside debug
ifdefs. Only performance penalty will come from using strlen() like all
the other code does.

I also modified some of the already patched parts by changing
snprintf(buf, 2 * BUFSIZE, ... style lines to
snprintf(buf, sizeof(buf), ... where buf is an array.

Jukka Holappa

21 years agoI guess the intention was to test incomplete SELECT statements, not
Bruce Momjian [Mon, 2 Sep 2002 06:05:16 +0000 (06:05 +0000)]
I guess the intention was to test incomplete SELECT statements, not
missing semicolons.

I also added a SELECT statement without a target list.

Manfred Koizar

21 years agoOkay, attached is a patch that moves the alter table regression test
Bruce Momjian [Mon, 2 Sep 2002 05:55:43 +0000 (05:55 +0000)]
Okay, attached is a patch that moves the alter table regression test
into a separate parallel group. I can't confirm that this fixes the
problem for me (since it only occurs about one in ten or one in twenty
runs), but I think it's a good bet that it does...

Neil Conway

21 years agoHere's a doc patch for the new string functions recently committed --
Bruce Momjian [Mon, 2 Sep 2002 05:53:23 +0000 (05:53 +0000)]
Here's a doc patch for the new string functions recently committed --
replace, split, and to_hex. The patch also moves encode and decode into
alphabetical order (since everything else in the table was).

Joe Conway

21 years agoMove pg_settings doc into the right section.
Bruce Momjian [Mon, 2 Sep 2002 05:52:34 +0000 (05:52 +0000)]
Move pg_settings doc into the right section.

21 years agoHere is a documentation patch for the pg_settings virtual table. If
Bruce Momjian [Mon, 2 Sep 2002 05:44:43 +0000 (05:44 +0000)]
Here is a documentation patch for the pg_settings virtual table. If
there are no objections, please apply.

Joe Conway

21 years agoAttached is an update to contrib/tablefunc. It introduces a new
Bruce Momjian [Mon, 2 Sep 2002 05:44:05 +0000 (05:44 +0000)]
Attached is an update to contrib/tablefunc. It introduces a new
function, connectby(), which can serve as a reference implementation for

the changes made in the last few days -- namely the ability of a
function to return an entire tuplestore, and the ability of a function
to make use of the query provided "expected" tuple description.

Description:

   connectby(text relname, text keyid_fld, text parent_keyid_fld,
     text start_with, int max_depth [, text branch_delim])
   - returns keyid, parent_keyid, level, and an optional branch string
   - requires anonymous composite type syntax in the FROM clause. See
     the instructions in the documentation below.

Joe Conway

21 years agoAdd GUC variable to print original query to the server logs when there
Bruce Momjian [Mon, 2 Sep 2002 05:42:54 +0000 (05:42 +0000)]
Add GUC variable to print original query to the server logs when there
is an error, warning, etc.

Gavin Sherry

21 years agoRemove documentation that says debug_query_string is only used by
Bruce Momjian [Mon, 2 Sep 2002 05:25:37 +0000 (05:25 +0000)]
Remove documentation that says debug_query_string is only used by
pgmonitor.

Now log_min_error_statement uses it.

21 years agoFix ordering regression problem on freebsd/alpha by adding ORDER BY.
Bruce Momjian [Mon, 2 Sep 2002 05:20:56 +0000 (05:20 +0000)]
Fix ordering regression problem on freebsd/alpha by adding ORDER BY.

21 years agoRemoved because we don't use cs() anymore on AIX:
Bruce Momjian [Mon, 2 Sep 2002 04:43:17 +0000 (04:43 +0000)]
Removed because we don't use cs() anymore on AIX:

< * Evaluate AIX cs() spinlock macro for performance optimizations (Tatsuo)

21 years agoI tried to build PostgreSQL with the following step to see backends hung
Bruce Momjian [Mon, 2 Sep 2002 04:42:52 +0000 (04:42 +0000)]
I tried to build PostgreSQL with the following step to see backends hung
during the regression test. The problem has been reproduced on two machine
but both of these are the same type of hardware and software. I also tried
to recreate the problem on other machines, on older version of AIX but I
couldn't.

After looked through pgsql-hackers mailing list, I focused on spin lock
issue to solve the problem. The easiest and may not be the best solution
for the problem is to give up HAS_TEST_AND_SET. This actually works.

One another and better solution for the problem is to use _check_lock() and
_clear_lock() as spin lock.  Important thing here is to define S_UNLOCK()
with _clear_lock().  This will solve the so called "Compiler bug" issue
someone wrote on the mailing list.

We have some other API such as cs(), compare_and_swap() and fetch_and_or()
to do test and set on AIX, but any of these didn't solve my problem.  I
wrote tiny testing program to see if we have any bug of these API of AIX,
but I couldn't see any problem except for compare_and_swap(). It seems that
you can not use compare_and_swap() for the purpose, as it would not work as
spin lock on any SMP machines I tested.  I don't know the reason why cs()
nor fetch_and_or()/fetch_and_and() will not work with PostgreSQL on p690.
These worked with my testing program on all machines I tested.

Tomoyuki Niijima

21 years agoJDBC checkin fixing the following bugs:
Barry Lind [Mon, 2 Sep 2002 03:07:36 +0000 (03:07 +0000)]
JDBC checkin fixing the following bugs:
  Fixed support in the driver for notifications (added PGConnection.getNotifications()) - problem reported by Benjamin.Feinstein@guardent.com
  Worked around server problems with int8/int2 and constants; quote values when they are intended to bind to an int8/int2 column - reported by many
  Fixed bug in the Array interface with string parsing not handling escaped characters correctly - reported by devajx@yahoo.com
  Added workaround to support 'infinity' and '-infinity' for dates - reported bydmitry@openratings.com
  Fixed some performance issues with setBlob - reported by d.wall@computer.org
  Added support for using new prepared statements functionality in 7.3 (added PGStatement.setUseServerPrepare() and isUseServerPrepare() methods)

 Modified Files:
  jdbc/org/postgresql/PGConnection.java
  jdbc/org/postgresql/PGStatement.java
  jdbc/org/postgresql/core/QueryExecutor.java
  jdbc/org/postgresql/jdbc1/AbstractJdbc1Connection.java
  jdbc/org/postgresql/jdbc1/AbstractJdbc1ResultSet.java
  jdbc/org/postgresql/jdbc1/AbstractJdbc1Statement.java
  jdbc/org/postgresql/jdbc2/AbstractJdbc2Statement.java
  jdbc/org/postgresql/jdbc2/Array.java
 Added Files:
  jdbc/org/postgresql/PGNotification.java
  jdbc/org/postgresql/core/Notification.java

21 years agoRemove sys/types.h in files that include postgres.h, and hence c.h,
Bruce Momjian [Mon, 2 Sep 2002 02:47:07 +0000 (02:47 +0000)]
Remove sys/types.h in files that include postgres.h, and hence c.h,
because c.h has sys/types.h.

21 years agoCREATE OR REPLACE VIEW, CREATE OR REPLACE RULE.
Tom Lane [Mon, 2 Sep 2002 02:13:02 +0000 (02:13 +0000)]
CREATE OR REPLACE VIEW, CREATE OR REPLACE RULE.
Gavin Sherry, Neil Conway, and Tom Lane all got their hands dirty
on this one ...

21 years agoCode review for HeapTupleHeader changes. Add version number to page headers
Tom Lane [Mon, 2 Sep 2002 01:05:06 +0000 (01:05 +0000)]
Code review for HeapTupleHeader changes.  Add version number to page headers
(overlaying low byte of page size) and add HEAP_HASOID bit to t_infomask,
per earlier discussion.  Simplify scheme for overlaying fields in tuple
header (no need for cmax to live in more than one place).  Don't try to
clear infomask status bits in tqual.c --- not safe to do it there.  Don't
try to force output table of a SELECT INTO to have OIDs, either.  Get rid
of unnecessarily complex three-state scheme for TupleDesc.tdhasoids, which
has already caused one recent failure.  Improve documentation.

21 years agopatch from Thomas O'Dowd to ensure that connections are closed
Dave Cramer [Sun, 1 Sep 2002 23:56:13 +0000 (23:56 +0000)]
patch from Thomas O'Dowd to ensure that connections are closed

21 years agotesting for insertRow
Dave Cramer [Sun, 1 Sep 2002 23:45:20 +0000 (23:45 +0000)]
testing for insertRow

21 years agoUpdate MEMSET_LOOP_LIMIT to 1024 because it is the best value for most
Bruce Momjian [Sun, 1 Sep 2002 23:41:47 +0000 (23:41 +0000)]
Update MEMSET_LOOP_LIMIT to 1024 because it is the best value for most
platforms.

21 years agopatch to allow insertRow on an empty resultSet
Dave Cramer [Sun, 1 Sep 2002 23:40:36 +0000 (23:40 +0000)]
patch to allow insertRow on an empty resultSet

21 years agoChanged to reflect the current reality, ie we do support updateable resultsets
Dave Cramer [Sun, 1 Sep 2002 23:38:53 +0000 (23:38 +0000)]
Changed to reflect the current reality, ie we do support updateable resultsets

21 years agoImprove footer length computation for psql.
Bruce Momjian [Sun, 1 Sep 2002 23:30:46 +0000 (23:30 +0000)]
Improve footer length computation for psql.

Greg Sabino Mullan

21 years agoAdd log_duration to GUC/postgresql.conf.
Bruce Momjian [Sun, 1 Sep 2002 23:26:06 +0000 (23:26 +0000)]
Add log_duration to GUC/postgresql.conf.

Rename debug_print_query to log_statement and rename show_query_stats to
show_statement_stats.

21 years agoAdd:
Bruce Momjian [Sun, 1 Sep 2002 22:36:11 +0000 (22:36 +0000)]
Add:

>  o Add table function support to pltcl, plperl, plpython

21 years agoplpgsql functions can return RECORD, per Neil Conway.
Tom Lane [Sun, 1 Sep 2002 16:28:06 +0000 (16:28 +0000)]
plpgsql functions can return RECORD, per Neil Conway.

21 years agoMinor copy-editing.
Tom Lane [Sun, 1 Sep 2002 02:37:02 +0000 (02:37 +0000)]
Minor copy-editing.

21 years agoCode cleanups: make non-implicit WITHOUT FUNCTION casts work, avoid
Tom Lane [Sun, 1 Sep 2002 02:27:32 +0000 (02:27 +0000)]
Code cleanups: make non-implicit WITHOUT FUNCTION casts work, avoid
redundant pg_cast searches, fix obsolete comments.

21 years agoRemove mention of wal_files in a SHOW example.
Bruce Momjian [Sun, 1 Sep 2002 01:59:48 +0000 (01:59 +0000)]
Remove mention of wal_files in a SHOW example.

21 years agoRemove code that suggested increasing wal_files.
Bruce Momjian [Sun, 1 Sep 2002 01:58:42 +0000 (01:58 +0000)]
Remove code that suggested increasing wal_files.

21 years agoDone:
Bruce Momjian [Sun, 1 Sep 2002 01:32:21 +0000 (01:32 +0000)]
Done:

>  o -SELECT col FROM tab WHERE numeric_col = 10.1 fails, requires quotes

21 years agoMark the float8 -> int8 cast as implicit. This resolves the problem
Tom Lane [Sun, 1 Sep 2002 00:58:07 +0000 (00:58 +0000)]
Mark the float8 -> int8 cast as implicit.  This resolves the problem
pointed out by Barry Lind: UPDATE bigintcol = 10000000000 fails because
the constant is initially taken as float8.  We really need a better way,
but it's not gonna happen for 7.3.

Also, remove int4reltime() function, which is redundant with the
existing binary-compatibility coercion path from int4 to reltime,
and probably has been unreachable code for a long while.

21 years agoCode review for domain-constraints patch. Use a new ConstraintTest node
Tom Lane [Sat, 31 Aug 2002 22:10:48 +0000 (22:10 +0000)]
Code review for domain-constraints patch.  Use a new ConstraintTest node
type for runtime constraint checks, instead of misusing the parse-time
Constraint node for the purpose.  Fix some damage introduced into type
coercion logic; in particular ensure that a coerced expression tree will
read out the correct result type when inspected (patch had broken some
RelabelType cases).  Enforce domain NOT NULL constraints against columns
that are omitted from an INSERT.

21 years agoWups, didn't mean to commit that just yet.
Tom Lane [Sat, 31 Aug 2002 19:10:08 +0000 (19:10 +0000)]
Wups, didn't mean to commit that just yet.

21 years agoThe UNDEFOID later causes an assertion failure in heap_formtuple when
Tom Lane [Sat, 31 Aug 2002 19:09:27 +0000 (19:09 +0000)]
The UNDEFOID later causes an assertion failure in heap_formtuple when
you try to use the tupdesc to build a tuple.

Joe Conway

21 years agoCode review for pg_locks feature. Make shmemoffset of PROCLOCK structs
Tom Lane [Sat, 31 Aug 2002 17:14:28 +0000 (17:14 +0000)]
Code review for pg_locks feature.  Make shmemoffset of PROCLOCK structs
available (else there's no way to interpret the list links).  Change
pg_locks view to show transaction ID locks separately from ordinary
relation locks.  Avoid showing N duplicate rows when the same lock is
held multiple times (seems unlikely that users care about exact hold
count).  Improve documentation.

21 years ago*** empty log message ***
Bruce Momjian [Sat, 31 Aug 2002 01:36:33 +0000 (01:36 +0000)]
*** empty log message ***

21 years agoAdd expected tuple descriptor to ReturnSetInfo information for table
Tom Lane [Fri, 30 Aug 2002 23:59:46 +0000 (23:59 +0000)]
Add expected tuple descriptor to ReturnSetInfo information for table
functions, per suggestion from John Gray and Joe Conway.  Also, fix
plpgsql RETURN NEXT to verify that returned values match the expected
tupdesc.

21 years agoUpdates to reflect availability of autocommit option.
Tom Lane [Fri, 30 Aug 2002 22:45:25 +0000 (22:45 +0000)]
Updates to reflect availability of autocommit option.

21 years agoAUTOCOMMIT mode is now an available backend GUC variable; setting it
Tom Lane [Fri, 30 Aug 2002 22:18:07 +0000 (22:18 +0000)]
AUTOCOMMIT mode is now an available backend GUC variable; setting it
to false provides more SQL-spec-compliant behavior than we had before.
I am not sure that setting it false is actually a good idea yet; there
is a lot of client-side code that will probably be broken by turning
autocommit off.  But it's a start.

Loosely based on a patch by David Van Wie.

21 years agoFix for breakage of C-coded SRFs, from Joe Conway.
Tom Lane [Fri, 30 Aug 2002 19:56:49 +0000 (19:56 +0000)]
Fix for breakage of C-coded SRFs, from Joe Conway.

21 years agoAdd attisinherited column to pg_attribute; use it to guard against
Tom Lane [Fri, 30 Aug 2002 19:23:20 +0000 (19:23 +0000)]
Add attisinherited column to pg_attribute; use it to guard against
column additions, deletions, and renames that would let a child table
get out of sync with its parent.  Patch by Alvaro Herrera, with some
kibitzing by Tom Lane.

21 years agoBack out unintented tab change.
Bruce Momjian [Fri, 30 Aug 2002 18:15:23 +0000 (18:15 +0000)]
Back out unintented tab change.

21 years agoRemove pgaccess from Makefile.
Bruce Momjian [Fri, 30 Aug 2002 18:14:21 +0000 (18:14 +0000)]
Remove pgaccess from Makefile.

21 years agoRun autoconf.
Bruce Momjian [Fri, 30 Aug 2002 17:16:42 +0000 (17:16 +0000)]
Run autoconf.

21 years agoUpdate documentation for --with-perl.
Bruce Momjian [Fri, 30 Aug 2002 17:14:30 +0000 (17:14 +0000)]
Update documentation for --with-perl.

21 years agoImprove wording after wal_files removal.
Bruce Momjian [Fri, 30 Aug 2002 16:53:34 +0000 (16:53 +0000)]
Improve wording after wal_files removal.

21 years agoComplete TODO item:
Bruce Momjian [Fri, 30 Aug 2002 16:50:50 +0000 (16:50 +0000)]
Complete TODO item:

* Remove wal_files postgresql.conf option because WAL files are
  now recycled

21 years agoRemove pgaccess from CVS. It has its own home now.
Bruce Momjian [Fri, 30 Aug 2002 16:48:44 +0000 (16:48 +0000)]
Remove pgaccess from CVS.  It has its own home now.

21 years agoRe-enable perl configure tests for psql help file generation.
Bruce Momjian [Fri, 30 Aug 2002 16:23:21 +0000 (16:23 +0000)]
Re-enable perl configure tests for psql help file generation.

21 years agoBack out password packet length check.
Bruce Momjian [Fri, 30 Aug 2002 16:00:41 +0000 (16:00 +0000)]
Back out password packet length check.

Improve wording of pre-7.3 syntax mention.

21 years agoperl5 interface moved to gborg
Marc G. Fournier [Fri, 30 Aug 2002 13:06:22 +0000 (13:06 +0000)]
perl5 interface moved to gborg

21 years agolibpgeasy moved to gborg ...
Marc G. Fournier [Fri, 30 Aug 2002 13:03:11 +0000 (13:03 +0000)]
libpgeasy moved to gborg ...

21 years agoFix off-by-one allocation error in PasswordFromFile(), per Gordon Runkle.
Tom Lane [Fri, 30 Aug 2002 05:28:50 +0000 (05:28 +0000)]
Fix off-by-one allocation error in PasswordFromFile(), per Gordon Runkle.

21 years agoAdd comment on old sytax for SELECT FOR UPDATE/LIMIT and COPY.
Bruce Momjian [Fri, 30 Aug 2002 03:18:23 +0000 (03:18 +0000)]
Add comment on old sytax for SELECT FOR UPDATE/LIMIT and COPY.

21 years agoRevert _LARGEFILE64_SOURCE addition; good try but doesn't work,
Tom Lane [Fri, 30 Aug 2002 02:01:34 +0000 (02:01 +0000)]
Revert _LARGEFILE64_SOURCE addition; good try but doesn't work,
at least not on HPUX 10.20, and there's no reason to think it
is needed on later versions.

21 years agointagg subdirectory was missing.
Tom Lane [Fri, 30 Aug 2002 01:44:00 +0000 (01:44 +0000)]
intagg subdirectory was missing.

21 years agoGroup-manipulating code neglected to fill in array element type,
Tom Lane [Fri, 30 Aug 2002 01:01:02 +0000 (01:01 +0000)]
Group-manipulating code neglected to fill in array element type,
which is now required.

21 years agoPL/pgSQL functions can return sets. Neil Conway's patch, modified so
Tom Lane [Fri, 30 Aug 2002 00:28:41 +0000 (00:28 +0000)]
PL/pgSQL functions can return sets.  Neil Conway's patch, modified so
that the functionality is available to anyone via ReturnSetInfo, rather
than hard-wiring it to PL/pgSQL.

21 years agoFix a bug introduced in 7.2.
Hiroshi Inoue [Thu, 29 Aug 2002 23:39:05 +0000 (23:39 +0000)]
Fix a bug introduced in 7.2.

21 years ago> > > > If you want to put in security restrictions that are actually useful,
Bruce Momjian [Thu, 29 Aug 2002 23:06:32 +0000 (23:06 +0000)]
> > > > If you want to put in security restrictions that are actually useful,
> > > > where is the code to verify that PGPASSWORDFILE points at a
> > > > non-world-readable file?  That needs to be there now, not later, or
> > > > we'll have people moaning about backward compatibility when we finally
> > > > do plug that hole.

Alvaro Herrera

21 years agoSir Mordred The Traitor <mordred@s-mail.com> writes:
Bruce Momjian [Thu, 29 Aug 2002 23:05:44 +0000 (23:05 +0000)]
Sir Mordred The Traitor <mordred@s-mail.com> writes:
> Upon invoking a polygon(integer, circle) function a
> src/backend/utils/adt/geo_ops.c:circle_poly() function will gets
> called, which suffers from a buffer overflow.
>
> 2) A src/backend/adt/utils/geo_ops.c:path_encode() fails to detect a
> buffer overrun condition. It is called in multiple places, the most
> interesting are path_out() and poly_out() functions.

> 5) A src/backend/utils/adt/geo_ops.c:path_add() also fails to detect
> a simple buffer overrun.

I've attached a patch which should fix these problems.

Neil Conway

21 years agoAdd cube changes file.
Bruce Momjian [Thu, 29 Aug 2002 23:05:03 +0000 (23:05 +0000)]
Add cube changes file.

21 years agoThe changes I have made are described in CHANGES. This was based on
Bruce Momjian [Thu, 29 Aug 2002 23:03:58 +0000 (23:03 +0000)]
The changes I have made are described in CHANGES. This was based on
diffs to 7.3-devel and may not be applicable to 7.2. I have included a
change covered by a previous bugfix patch I submitted (the problem with
-.1 not being accepted by cube_in). It does not include a fix for the
potential buffer overrun issue I reported for cube_yyerror in
cubeparse.y.

Bruno Wolff III

21 years agoMake pg_resetxlog options parsing more standard and prepare messages for
Peter Eisentraut [Thu, 29 Aug 2002 22:19:03 +0000 (22:19 +0000)]
Make pg_resetxlog options parsing more standard and prepare messages for
translation.

21 years agoCreate "po" subdirectory before writing there.
Peter Eisentraut [Thu, 29 Aug 2002 22:13:01 +0000 (22:13 +0000)]
Create "po" subdirectory before writing there.

21 years agoWorkaround for broken large file support on HP-UX
Peter Eisentraut [Thu, 29 Aug 2002 22:09:22 +0000 (22:09 +0000)]
Workaround for broken large file support on HP-UX

21 years agoPrevent problem with extra-long password packets from allocating lots of
Bruce Momjian [Thu, 29 Aug 2002 21:50:36 +0000 (21:50 +0000)]
Prevent problem with extra-long password packets from allocating lots of
memory.

Neil Conway

21 years agoThis patch reserves the last superuser_reserved_connections slots for
Bruce Momjian [Thu, 29 Aug 2002 21:02:12 +0000 (21:02 +0000)]
This patch reserves the last superuser_reserved_connections slots for
connections by the superuser only.

This patch replaces the last patch I sent a couple of days ago.

It closes a connection that has not been authorised by a superuser if it would
leave less than the GUC variable ReservedBackends
(superuser_reserved_connections in postgres.conf) backend process slots free
in the SISeg. This differs to the first patch which only reserved the last
ReservedBackends slots in the procState array. This has made the free slot
test more expensive due to the use of a lock.

After thinking about a comment on the first patch I've also made it a fatal
error if the number of reserved slots is not less than the maximum number of
connections.

Nigel J. Andrews

21 years agoplease apply small patch for README.tsearch.
Bruce Momjian [Thu, 29 Aug 2002 19:55:26 +0000 (19:55 +0000)]
please apply small patch for README.tsearch.

I've documented space usage and using CLUSTER command

Oleg Bartunov

21 years agoThe unit for statement_timeout is s/1000, not s/1000000
Bruce Momjian [Thu, 29 Aug 2002 19:53:58 +0000 (19:53 +0000)]
The unit for statement_timeout is s/1000, not s/1000000

Manfred Koizar

21 years agoBack out patch, that was alpha, which I think still needs just -O.
Bruce Momjian [Thu, 29 Aug 2002 19:35:47 +0000 (19:35 +0000)]
Back out patch, that was alpha, which I think still needs just -O.

21 years agoChange FreeBSD to use -O2, important for MemSet.
Bruce Momjian [Thu, 29 Aug 2002 19:35:07 +0000 (19:35 +0000)]
Change FreeBSD to use -O2, important for MemSet.

21 years agoAdjust nodeFunctionscan.c to reset transient memory context between calls
Tom Lane [Thu, 29 Aug 2002 17:14:33 +0000 (17:14 +0000)]
Adjust nodeFunctionscan.c to reset transient memory context between calls
to the table function, thus preventing memory leakage accumulation across
calls.  This means that SRFs need to be careful to distinguish permanent
and local storage; adjust code and documentation accordingly.  Patch by
Joe Conway, very minor tweaks by Tom Lane.

21 years agoPush down outer qualification clauses into UNION and INTERSECT subqueries.
Tom Lane [Thu, 29 Aug 2002 16:03:49 +0000 (16:03 +0000)]
Push down outer qualification clauses into UNION and INTERSECT subqueries.
Per pghackers discussion from back around 1-August.

21 years agoCause REINDEX to regard TOAST tables as regular relations, not system
Tom Lane [Thu, 29 Aug 2002 15:56:20 +0000 (15:56 +0000)]
Cause REINDEX to regard TOAST tables as regular relations, not system
tables that need special defenses.  I believe this is okay even for
TOAST tables that belong to system tables.

21 years agochkpass_rout returns text so change PG_RETURN_CSTRING to PG_RETURN_TEXT_P.
D'Arcy J.M. Cain [Thu, 29 Aug 2002 12:18:20 +0000 (12:18 +0000)]
chkpass_rout returns text so change PG_RETURN_CSTRING to PG_RETURN_TEXT_P.
This is currently a cosmetic difference but I make the change now in case
the macros diverge one day.