]> granicus.if.org Git - postgresql/log
postgresql
20 years agoTranslation updates
Peter Eisentraut [Thu, 15 Apr 2004 08:15:09 +0000 (08:15 +0000)]
Translation updates

20 years agoTranslation updates
Peter Eisentraut [Thu, 15 Apr 2004 08:04:05 +0000 (08:04 +0000)]
Translation updates

20 years agoClarify meaning of second DateStyle value.
Bruce Momjian [Thu, 15 Apr 2004 04:46:18 +0000 (04:46 +0000)]
Clarify meaning of second DateStyle value.

20 years agoUpdate name:
Bruce Momjian [Wed, 14 Apr 2004 19:15:27 +0000 (19:15 +0000)]
Update name:

< * Alvaro Herrera <alvherre@dcc.uchile.cl>
> * Alvaro is Alvaro Herrera <alvherre@dcc.uchile.cl>

20 years agoUpdate owners of items:
Bruce Momjian [Wed, 14 Apr 2004 18:33:03 +0000 (18:33 +0000)]
Update owners of items:

< * Add the concept of dataspaces/tablespaces [tablespaces]
> * Add the concept of dataspaces/tablespaces [tablespaces] (Gavin)
167c167
< * Allow savepoints / nested transactions [transactions] (Bruce)
> * Allow savepoints / nested transactions [transactions] (Alvaro)
240c240
<  o Allow Java server-side programming [java]
>  o Allow Java server-side programming [java] (Dave)
496a497
> * Alvaro Herrera <alvherre@dcc.uchile.cl>

20 years agoHere's an attempt at new socket and signal code for win32.
Bruce Momjian [Mon, 12 Apr 2004 16:19:18 +0000 (16:19 +0000)]
Here's an attempt at new socket and signal code for win32.

It works on the principle of turning sockets into non-blocking, and then
emulate blocking behaviour on top of that, while allowing signals to
run. Signals are now implemented using an event instead of APCs, thus
getting rid of the issue of APCs not being compatible with "old style"
sockets functions.

It also moves the win32 specific code away from pqsignal.h/c into
port/win32, and also removes the "thread style workaround" of the APC
issue previously in place.

In order to make things work, a few things are also changed in pgstat.c:

1) There is now a separate pipe to the collector and the bufferer. This
is required because the pipe will otherwise only be signalled in one of
the processes when the postmaster goes down. The MS winsock code for
select() must have some kind of workaround for this behaviour, but I
have found no stable way of doing that. You really are not supposed to
use the same socket from more than one process (unless you use
WSADuplicateSocket(), in which case the docs specifically say that only
one will be flagged).

2) The check for "postmaster death" is moved into a separate select()
call after the main loop. The previous behaviour select():ed on the
postmaster pipe, while later explicitly saying "we do NOT check for
postmaster exit inside the loop".
The issue was that the code relies on the same select() call seeing both
the postmaster pipe *and* the pgstat pipe go away. This does not always
happen, and it appears that useing WSAEventSelect() makes it even more
common that it does not.
Since it's only called when the process exits, I don't think using a
separate select() call will have any significant impact on how the stats
collector works.

Magnus Hagander

20 years agoChange psql \copy stdin/stdout to read from command input/output.
Bruce Momjian [Mon, 12 Apr 2004 15:58:52 +0000 (15:58 +0000)]
Change psql \copy stdin/stdout to read from command input/output.

Add pstdin/pstdout to read from psql's stdin/stdout.

BACKWARD INCOMPATIBLE CHANGE

20 years ago> > This update fixes a few small typos in names,
Bruce Momjian [Mon, 12 Apr 2004 03:22:20 +0000 (03:22 +0000)]
> > This update fixes a few small typos in names,
> > pronouns and formatting in the Russian FAQ.

Serguei Mokhov

20 years agoExit backend from SIGTERM or FATAL by simulating client EOF, rather than
Bruce Momjian [Sun, 11 Apr 2004 00:54:45 +0000 (00:54 +0000)]
Exit backend from SIGTERM or FATAL by simulating client EOF, rather than
calling proc_exit() directly.  This should make SIGTERM more reliable.

20 years agoPlease find a small patch to fix the brain damage "century" and
Bruce Momjian [Sat, 10 Apr 2004 18:02:59 +0000 (18:02 +0000)]
Please find a small patch to fix the brain damage "century" and
"millennium" date part implementation in postgresql, both in the code
and the documentation, so that it conforms to the official definition.
If you do not agree with the official definition, please send your
complaint to "pope@vatican.org". I'm not responsible for them;-)

With the previous version, the centuries and millenniums had a wrong
number and started the wrong year. Moreover century number 0, which does
not exist in reality, lasted 200 years. Also, millennium number 0 lasted
2000 years.

If you want postgresql to have it's own definition of "century" and
"millennium" that does not conform to the one of the society, just give
them another name. I would suggest "pgCENTURY" and "pgMILLENNIUM";-)

IMO, if someone may use the options, it means that postgresql is used for
historical data, so it make sense to have an historical definition. Also,
I just want to divide the year by 100 or 1000, I can do that quite easily.

BACKWARD INCOMPATIBLE CHANGE

Fabien Coelho - coelho@cri.ensmp.fr

20 years agoRemove libpgeasy and odbc doc reference in filelist.sgml --- removed to
Bruce Momjian [Fri, 9 Apr 2004 18:03:13 +0000 (18:03 +0000)]
Remove libpgeasy and odbc doc reference in filelist.sgml --- removed to
gborg.

20 years agoTranslation updates
Dennis Bjorklund [Fri, 9 Apr 2004 06:49:35 +0000 (06:49 +0000)]
Translation updates

20 years agoTranslation updates
Dennis Bjorklund [Fri, 9 Apr 2004 05:54:01 +0000 (05:54 +0000)]
Translation updates

20 years agoMagnus says new Win32 timezone library will not need special horology:
Bruce Momjian [Thu, 8 Apr 2004 23:00:55 +0000 (23:00 +0000)]
Magnus says new Win32 timezone library will not need special horology:

so remove regression result map line:

horology/win32=horology-no-DST-before-1970

20 years agoAdd SQL 2003 standards.
Bruce Momjian [Wed, 7 Apr 2004 23:32:30 +0000 (23:32 +0000)]
Add SQL 2003 standards.

20 years agoRemove bad URL.
Bruce Momjian [Wed, 7 Apr 2004 22:02:03 +0000 (22:02 +0000)]
Remove bad URL.

20 years agoUpdate:
Bruce Momjian [Wed, 7 Apr 2004 19:14:21 +0000 (19:14 +0000)]
Update:

< * Allow LOCALE on a per-column basis, default to ASCII
> * Allow locale to be set at database creation
> * Allow locale on a per-column basis, default to ASCII
> * Optimize locale to have minimal performance impact when not used (Peter E)
105d106
< * Optimize locale to have minimal performance impact when not used (Peter E)
111d111
< * Allow locale to be set at database creation

20 years agoPush superuser check farther down in the if() statements to fix startup
Bruce Momjian [Wed, 7 Apr 2004 18:52:26 +0000 (18:52 +0000)]
Push superuser check farther down in the if() statements to fix startup
crash with debug in log_statement patch.

20 years agoStill another place to make the world safe for zero-column tables.
Tom Lane [Wed, 7 Apr 2004 18:46:12 +0000 (18:46 +0000)]
Still another place to make the world safe for zero-column tables.
Per example from Jiang Wei.

20 years agoAdd:
Bruce Momjian [Wed, 7 Apr 2004 18:34:44 +0000 (18:34 +0000)]
Add:

> * Allow locale to be set at database creation

20 years agoUpdate Darwin thread compile flags, and OS version info.
Bruce Momjian [Wed, 7 Apr 2004 18:31:59 +0000 (18:31 +0000)]
Update Darwin thread compile flags, and OS version info.

20 years agoUpdate darwin OS version.
Bruce Momjian [Wed, 7 Apr 2004 18:30:59 +0000 (18:30 +0000)]
Update darwin OS version.

20 years agoExtend set-operation planning to keep track of the sort ordering induced
Tom Lane [Wed, 7 Apr 2004 18:17:25 +0000 (18:17 +0000)]
Extend set-operation planning to keep track of the sort ordering induced
by the set operation, so that redundant sorts at higher levels can be
avoided.  This was foreseen a good while back, but not done.  Per request
from Karel Zak.

20 years agobuild_subquery_pathkeys() was examining wrong copy of subquery target list,
Tom Lane [Wed, 7 Apr 2004 17:42:28 +0000 (17:42 +0000)]
build_subquery_pathkeys() was examining wrong copy of subquery target list,
causing it to fail to recognize the output ordering of subqueries that
contain set operations (UNION/INTERSECT/EXPECT).  Per example from Karel Zak.

20 years agoDone:
Bruce Momjian [Wed, 7 Apr 2004 17:04:45 +0000 (17:04 +0000)]
Done:

> * -Allow logging of only data definition(DDL), or DDL and modification statements

Add:
> * Add a session mode to warn about non-standard SQL usage

20 years agoUpdate FAQ_DEV.
Bruce Momjian [Wed, 7 Apr 2004 14:23:32 +0000 (14:23 +0000)]
Update FAQ_DEV.

20 years agoUpdate Darwin as thread-safe.
Bruce Momjian [Wed, 7 Apr 2004 14:21:39 +0000 (14:21 +0000)]
Update Darwin as thread-safe.

20 years ago> >>1. change the type of "log_statement" option from boolean to string,
Bruce Momjian [Wed, 7 Apr 2004 05:05:50 +0000 (05:05 +0000)]
> >>1. change the type of "log_statement" option from boolean to string,
> >>with allowed values of "all, mod, ddl, none" with default "none".

OK, here is a patch that implements #1.  Here is sample output:

        test=> set client_min_messages = 'log';
        SET
        test=> set log_statement = 'mod';
        SET
        test=> select 1;
         ?column?
        ----------
                1
        (1 row)

        test=> update test set x=1;
        LOG:  statement: update test set x=1;
        ERROR:  relation "test" does not exist
        test=> update test set x=1;
        LOG:  statement: update test set x=1;
        ERROR:  relation "test" does not exist
        test=> copy test from '/tmp/x';
        LOG:  statement: copy test from '/tmp/x';
        ERROR:  relation "test" does not exist
        test=> copy test to  '/tmp/x';
        ERROR:  relation "test" does not exist
        test=> prepare xx as select 1;
        PREPARE
        test=> prepare xx as update x set y=1;
        LOG:  statement: prepare xx as update x set y=1;
        ERROR:  relation "x" does not exist
        test=> explain analyze select 1;;
                                             QUERY PLAN
        ------------------------------------------------------------------------------------
         Result  (cost=0.00..0.01 rows=1 width=0) (actual time=0.006..0.007 rows=1 loops=1)
         Total runtime: 0.046 ms
        (2 rows)

        test=> explain analyze update test set x=1;
        LOG:  statement: explain analyze update test set x=1;
        ERROR:  relation "test" does not exist
        test=> explain update test set x=1;
        ERROR:  relation "test" does not exist

It checks PREPARE and EXECUTE ANALYZE too.  The log_statement values are
'none', 'mod', 'ddl', and 'all'.  For 'all', it prints before the query
is parsed, and for ddl/mod, it does it right after parsing using the
node tag (or command tag for CREATE/ALTER/DROP), so any non-parse errors
will print after the log line.

20 years agoSupport FULL JOIN with no join clauses, such as X FULL JOIN Y ON TRUE.
Tom Lane [Tue, 6 Apr 2004 18:46:03 +0000 (18:46 +0000)]
Support FULL JOIN with no join clauses, such as X FULL JOIN Y ON TRUE.
That particular corner case is not exactly compelling, but given 7.4's
ability to discard redundant join clauses, it is possible for the situation
to arise from queries that are not so obviously silly.  Per bug report
of 6-Apr-04.

20 years agoALTER SEQUENCE RESTART did the wrong thing if sequence last_value was
Tom Lane [Tue, 6 Apr 2004 16:39:30 +0000 (16:39 +0000)]
ALTER SEQUENCE RESTART did the wrong thing if sequence last_value was
equal to the desired restart value (must clear is_called, did not).
Per bug report #1127 from Piotr Konieczny.

20 years agoAdd thread program mention of errno thread-safety.
Bruce Momjian [Tue, 6 Apr 2004 13:55:17 +0000 (13:55 +0000)]
Add thread program mention of errno thread-safety.

20 years agoThe attached applied patch throws an error if the delimiter appears in
Bruce Momjian [Tue, 6 Apr 2004 13:21:33 +0000 (13:21 +0000)]
The attached applied patch throws an error if the delimiter appears in
the COPY NULL string:

        test=> copy pg_language to '/tmp/x' with delimiter '|';
        COPY
        test=> copy pg_language to '/tmp/x' with delimiter '|' null '|x';
        ERROR:  COPY delimiter must not appear in the NULL specification
        test=> copy pg_language from '/tmp/x' with delimiter '|' null '|x';
        ERROR:  COPY delimiter must not appear in the NULL specification

It also throws an error if it conflicts with the default NULL string:

        test=> copy pg_language to '/tmp/x' with delimiter '\\';
        ERROR:  COPY delimiter must not appear in the NULL specification
        test=> copy pg_language to '/tmp/x' with delimiter '\\' NULL 'x';
        COPY

20 years agoAdd CLUSTER tag to psql \d display.
Bruce Momjian [Tue, 6 Apr 2004 04:05:17 +0000 (04:05 +0000)]
Add CLUSTER tag to psql \d display.

20 years agoAdd comment that sched_yield might be a portability problem.
Bruce Momjian [Mon, 5 Apr 2004 14:13:45 +0000 (14:13 +0000)]
Add comment that sched_yield might be a portability problem.

20 years agoChange getpid() while loop to use sched_yield(), greatly speeds up test.
Bruce Momjian [Mon, 5 Apr 2004 14:12:32 +0000 (14:12 +0000)]
Change getpid() while loop to use sched_yield(), greatly speeds up test.

sched_yield is a requirement for the test but not for our PostgreSQL
code, so if it causes a portability problem, we will remove it.

20 years agoNew translation
Peter Eisentraut [Mon, 5 Apr 2004 09:34:11 +0000 (09:34 +0000)]
New translation

20 years agoTranslation updates
Peter Eisentraut [Mon, 5 Apr 2004 09:13:43 +0000 (09:13 +0000)]
Translation updates

20 years agoHave threads wait for parent to test thread-specific pointers before
Bruce Momjian [Mon, 5 Apr 2004 05:43:06 +0000 (05:43 +0000)]
Have threads wait for parent to test thread-specific pointers before
exiting.

20 years agoCorrects issues recently posted by Dann Corbit, allowing libpq/psql to
Bruce Momjian [Mon, 5 Apr 2004 03:16:21 +0000 (03:16 +0000)]
Corrects issues recently posted by Dann Corbit, allowing libpq/psql to
be built under VC++. Moves a pgstat win32 #def to port.h

Claudio Natoli

20 years agoThis is a cleanup patch for access/transam/xact.c. It only removes some
Bruce Momjian [Mon, 5 Apr 2004 03:11:39 +0000 (03:11 +0000)]
This is a cleanup patch for access/transam/xact.c.  It only removes some
#ifdef NOT_USED code, and adds a new TBLOCK state which signals the fact
that StartTransaction() has been executed.

Alvaro Herrera

20 years agoAllow LIKE/ILIKE to appear in more places in a query.
Bruce Momjian [Mon, 5 Apr 2004 03:07:26 +0000 (03:07 +0000)]
Allow LIKE/ILIKE to appear in more places in a query.

Fabien COELHO

20 years agoRemove 'syslog' GUC variable, and add more logical 'log_destination'
Bruce Momjian [Mon, 5 Apr 2004 03:02:11 +0000 (03:02 +0000)]
Remove 'syslog' GUC variable, and add more logical 'log_destination'
variable to control logoutput location on Unix and Win32.

Magnus Hagander

20 years agoImprove handling of GUC USERLIMIT variables by reorganizing code. Also,
Bruce Momjian [Mon, 5 Apr 2004 02:48:09 +0000 (02:48 +0000)]
Improve handling of GUC USERLIMIT variables by reorganizing code.  Also,
handle new postgresql.conf values with SIGHUP better by better enforcing
USERLIMIT settings on existing non-super-user backends.

20 years agoAdd volatile to thread-specific thread variables.
Bruce Momjian [Mon, 5 Apr 2004 02:22:14 +0000 (02:22 +0000)]
Add volatile to thread-specific thread variables.

20 years agoUpdate thread test to only test for 'localhost' and local machine name.
Bruce Momjian [Mon, 5 Apr 2004 01:27:58 +0000 (01:27 +0000)]
Update thread test to only test for 'localhost' and local machine name.

20 years agoUpdate thread test to do getpid() in while loop, also use weather.com
Bruce Momjian [Sun, 4 Apr 2004 17:23:54 +0000 (17:23 +0000)]
Update thread test to do getpid() in while loop, also use weather.com
rather than slashdot.org for testing.

20 years agocheck_sql_fn_retval has always thought that we supported doing
Tom Lane [Fri, 2 Apr 2004 23:14:08 +0000 (23:14 +0000)]
check_sql_fn_retval has always thought that we supported doing
'SELECT foo()' in a SQL function returning a rowtype, to simply pass
back the results of another function returning the same rowtype.
However, that hasn't actually worked in many years.  Now it works again.

20 years agoFix poor choice of error message in corner cases.
Tom Lane [Fri, 2 Apr 2004 21:30:44 +0000 (21:30 +0000)]
Fix poor choice of error message in corner cases.

20 years agoClean up some code that had gotten a bit ugly through repeated revisions.
Tom Lane [Fri, 2 Apr 2004 21:05:32 +0000 (21:05 +0000)]
Clean up some code that had gotten a bit ugly through repeated revisions.

20 years agoGet rid of crocky use of RangeVar nodes in parser to represent partially
Tom Lane [Fri, 2 Apr 2004 19:07:02 +0000 (19:07 +0000)]
Get rid of crocky use of RangeVar nodes in parser to represent partially
transformed whole-row variables.  Cleaner to use regular whole-row Vars.

20 years agoTranslation updates
Dennis Bjorklund [Fri, 2 Apr 2004 13:47:56 +0000 (13:47 +0000)]
Translation updates

20 years agoTranslation updates
Dennis Bjorklund [Fri, 2 Apr 2004 13:42:26 +0000 (13:42 +0000)]
Translation updates

20 years agoFix some more compatibility issues (ctype.h macros must never be passed
Tom Lane [Fri, 2 Apr 2004 00:41:18 +0000 (00:41 +0000)]
Fix some more compatibility issues (ctype.h macros must never be passed
signed chars...)

20 years agoFix some portability issues with new float input code (didn't work on
Tom Lane [Thu, 1 Apr 2004 23:52:18 +0000 (23:52 +0000)]
Fix some portability issues with new float input code (didn't work on
HPUX 11 ...)

20 years agoFix some portability issues (reliance on gcc-isms).
Tom Lane [Thu, 1 Apr 2004 23:44:38 +0000 (23:44 +0000)]
Fix some portability issues (reliance on gcc-isms).

20 years agoFix portability issues with functions that don't match their declaration.
Tom Lane [Thu, 1 Apr 2004 23:35:35 +0000 (23:35 +0000)]
Fix portability issues with functions that don't match their declaration.

20 years agoAdd missing casts to unsigned char in recently-added isspace() calls.
Tom Lane [Thu, 1 Apr 2004 22:51:31 +0000 (22:51 +0000)]
Add missing casts to unsigned char in recently-added isspace() calls.

20 years agoAdjust expected regression outputs for plan changes caused by recent
Tom Lane [Thu, 1 Apr 2004 21:59:45 +0000 (21:59 +0000)]
Adjust expected regression outputs for plan changes caused by recent
'fuzzy cost comparison' patch.  Mea culpa for not having noticed this
when I committed the patch.

20 years agoReplace TupleTableSlot convention for whole-row variables and function
Tom Lane [Thu, 1 Apr 2004 21:28:47 +0000 (21:28 +0000)]
Replace TupleTableSlot convention for whole-row variables and function
results with tuples as ordinary varlena Datums.  This commit does not
in itself do much for us, except eliminate the horrid memory leak
associated with evaluation of whole-row variables.  However, it lays the
groundwork for allowing composite types as table columns, and perhaps
some other useful features as well.  Per my proposal of a few days ago.

20 years agoImprove comments on USERLIMIT GUC processing.
Bruce Momjian [Thu, 1 Apr 2004 14:25:47 +0000 (14:25 +0000)]
Improve comments on USERLIMIT GUC processing.

20 years agoAdd description of log_statement_stats's function (total statement stats
Bruce Momjian [Wed, 31 Mar 2004 19:59:22 +0000 (19:59 +0000)]
Add description of log_statement_stats's function (total statement stats
rather than per-stage stats).

20 years agoBack out tutorial changes:
Bruce Momjian [Wed, 31 Mar 2004 16:20:53 +0000 (16:20 +0000)]
Back out tutorial changes:

---------------------------------------------------------------------------

1.  In keeping with the recent discussion that there should be more
said about views, stored procedures, and triggers, in the tutorial, I
have added a bit of verbiage to that end.

2.  Some formatting changes to the datetime discussion, as well as
addition of a citation of a relevant book on calendars.

Christopher Browne

20 years agoAdd psql backslash command discussion.
Bruce Momjian [Wed, 31 Mar 2004 01:05:54 +0000 (01:05 +0000)]
Add psql backslash command discussion.

20 years agoAdd:
Bruce Momjian [Wed, 31 Mar 2004 01:04:49 +0000 (01:04 +0000)]
Add:

* Move psql backslash database information into the backend, use
nmumonic commands? [psql]

20 years agoRemove:
Bruce Momjian [Wed, 31 Mar 2004 01:01:38 +0000 (01:01 +0000)]
Remove:

* Prevent unneeded quoting in psql \d output using fmtId()

20 years agoFix markup errors from recent patch.
Bruce Momjian [Tue, 30 Mar 2004 22:08:50 +0000 (22:08 +0000)]
Fix markup errors from recent patch.

20 years ago1. In keeping with the recent discussion that there should be more
Bruce Momjian [Tue, 30 Mar 2004 21:58:20 +0000 (21:58 +0000)]
1.  In keeping with the recent discussion that there should be more
said about views, stored procedures, and triggers, in the tutorial, I
have added a bit of verbiage to that end.

2.  Some formatting changes to the datetime discussion, as well as
addition of a citation of a relevant book on calendars.

Christopher Browne

20 years agoAdd:
Bruce Momjian [Tue, 30 Mar 2004 21:51:41 +0000 (21:51 +0000)]
Add:

> * Add GUC variable to allow output of interval values in ISO8601 format

20 years agoAdd mention of how to use \df to find functions using/returning specific
Bruce Momjian [Tue, 30 Mar 2004 15:54:33 +0000 (15:54 +0000)]
Add mention of how to use \df to find functions using/returning specific
data types.

20 years agoFix to_char for 1 BC. Previously it returned 1 AD.
Bruce Momjian [Tue, 30 Mar 2004 15:53:18 +0000 (15:53 +0000)]
Fix to_char for 1 BC.  Previously it returned 1 AD.

Fix to_char(year) for BC dates.  Previously it returned one less than
the current year.

Add documentation mentioning that there is no 0 AD.

20 years agoCleanup vectors of GISTENTRY and eliminate problem with 64-bit strict-aligned
Teodor Sigaev [Tue, 30 Mar 2004 15:45:33 +0000 (15:45 +0000)]
Cleanup vectors of GISTENTRY and eliminate problem with 64-bit strict-aligned
boxes. Change interface to user-defined GiST support methods union and
picksplit. Now instead of bytea struct it used special GistEntryVector
structure.

20 years agoUse fuzzy comparison of path costs in add_path(), so that paths with the
Tom Lane [Mon, 29 Mar 2004 19:58:04 +0000 (19:58 +0000)]
Use fuzzy comparison of path costs in add_path(), so that paths with the
same path keys and nearly equivalent costs will be considered redundant.
The exact nature of the fuzziness may get adjusted later based on current
discussions, but no one has shot a hole in the basic idea yet ...

20 years agoFixed possible segfault in type.c (by Juergen Cappel)
Michael Meskes [Mon, 29 Mar 2004 15:11:58 +0000 (15:11 +0000)]
Fixed possible segfault in type.c (by Juergen Cappel)

20 years agoUpdate FAQ from Robert Treat
Bruce Momjian [Mon, 29 Mar 2004 05:07:20 +0000 (05:07 +0000)]
Update FAQ from Robert Treat

20 years agoSome further editorializing on README.CVS.
Tom Lane [Sun, 28 Mar 2004 06:09:08 +0000 (06:09 +0000)]
Some further editorializing on README.CVS.

20 years agoClean up thread test program.
Bruce Momjian [Sun, 28 Mar 2004 02:37:31 +0000 (02:37 +0000)]
Clean up thread test program.

20 years agoAdd test for thread-safe errno to thread test program.
Bruce Momjian [Sat, 27 Mar 2004 23:02:44 +0000 (23:02 +0000)]
Add test for thread-safe errno to thread test program.

20 years agoAdd comment about psql \s (history display):
Bruce Momjian [Sat, 27 Mar 2004 18:01:40 +0000 (18:01 +0000)]
Add comment about psql \s (history display):

        /* This scrolls off the screen when using /dev/tty */

20 years agoClean up function call arg appearance.
Bruce Momjian [Sat, 27 Mar 2004 17:59:35 +0000 (17:59 +0000)]
Clean up function call arg appearance.

20 years agoChange small 'if/else' test to use Max().
Bruce Momjian [Sat, 27 Mar 2004 17:32:42 +0000 (17:32 +0000)]
Change small 'if/else' test to use Max().

20 years agoFix brain-dead placement of global variable declaration.
Tom Lane [Sat, 27 Mar 2004 03:08:42 +0000 (03:08 +0000)]
Fix brain-dead placement of global variable declaration.

20 years agoNow that we are allowing index opclasses to contain operators that are
Tom Lane [Sat, 27 Mar 2004 00:24:28 +0000 (00:24 +0000)]
Now that we are allowing index opclasses to contain operators that are
only stable and not immutable, pred_test_simple_clause has to guard
against making invalid deductions.  Add a test for immutability of
the selected test_op.

20 years agoAdd -D_REENTRANT for Solaris threading.
Bruce Momjian [Fri, 26 Mar 2004 18:24:08 +0000 (18:24 +0000)]
Add -D_REENTRANT for Solaris threading.

20 years agoMore cross-reference work, other minor SGML fixes and improvements.
Neil Conway [Fri, 26 Mar 2004 03:18:28 +0000 (03:18 +0000)]
More cross-reference work, other minor SGML fixes and improvements.

20 years agoUpdate gcc solaris flags.
Bruce Momjian [Thu, 25 Mar 2004 22:12:37 +0000 (22:12 +0000)]
Update gcc solaris flags.

20 years agoFix for SSL compile.
Bruce Momjian [Thu, 25 Mar 2004 21:48:59 +0000 (21:48 +0000)]
Fix for SSL compile.

20 years agoConvert some GUC variable references to links.
Tom Lane [Thu, 25 Mar 2004 18:57:57 +0000 (18:57 +0000)]
Convert some GUC variable references to links.

20 years ago1 Minimize memory allocation for void (but not null) value.
Teodor Sigaev [Thu, 25 Mar 2004 16:56:10 +0000 (16:56 +0000)]
1 Minimize memory allocation for void (but not null) value.
2 Add silly ordering for ts_vector to aim grouping, union, except etc. Don't use BTree opclass (tsvector_ops).

20 years agoAdd a more useful error message for the case where someone tries to pass
Tom Lane [Wed, 24 Mar 2004 23:38:49 +0000 (23:38 +0000)]
Add a more useful error message for the case where someone tries to pass
a whole row or record variable into a SQL function.  Eventually this case
should be made to actually work, but for now this is better than what it
did before.

20 years agoReplace max_expr_depth parameter with a max_stack_depth parameter that
Tom Lane [Wed, 24 Mar 2004 22:40:29 +0000 (22:40 +0000)]
Replace max_expr_depth parameter with a max_stack_depth parameter that
is measured in kilobytes and checked against actual physical execution
stack depth, as per my proposal of 30-Dec.  This gives us a fairly
bulletproof defense against crashing due to runaway recursive functions.

20 years agoAdd:
Bruce Momjian [Wed, 24 Mar 2004 18:15:56 +0000 (18:15 +0000)]
Add:

* Have pg_dump use multi-statement transactions for INSERT dumps

20 years agoAdjust error message wording per Andrew Dunstan's advice.
Tom Lane [Wed, 24 Mar 2004 15:20:54 +0000 (15:20 +0000)]
Adjust error message wording per Andrew Dunstan's advice.

20 years agoMake a cross-reference in the ALTER TABLE ref page into a <xref>.
Neil Conway [Wed, 24 Mar 2004 09:49:20 +0000 (09:49 +0000)]
Make a cross-reference in the ALTER TABLE ref page into a <xref>.

20 years agoUpdate description error text:
Bruce Momjian [Wed, 24 Mar 2004 04:07:31 +0000 (04:07 +0000)]
Update description error text:

        (errmsg("no socket configured for listening")));

20 years agoHere's a patch implementing the "thread method" to workaround the bug
Bruce Momjian [Wed, 24 Mar 2004 04:04:51 +0000 (04:04 +0000)]
Here's a patch implementing the "thread method" to workaround the bug
with socket calls in signal handlers (APC) on Win32. See details in mail
to pgsql-hackers-win32 a couple of minutes ago.

Magnus Hagander

20 years agoAllow unlink/rename of files open by another process on Win32, using a
Bruce Momjian [Wed, 24 Mar 2004 03:54:16 +0000 (03:54 +0000)]
Allow unlink/rename of files open by another process on Win32, using a
special Win32 open flag FILE_SHARE_DELETE.

Claudio Natoli

20 years agoFix a minor typo in the log_line_prefix docs and make a couple other
Neil Conway [Wed, 24 Mar 2004 03:48:41 +0000 (03:48 +0000)]
Fix a minor typo in the log_line_prefix docs and make a couple other
tiny improvements.

20 years agoAdd thread locking to SSL and Kerberos connections.
Bruce Momjian [Wed, 24 Mar 2004 03:45:00 +0000 (03:45 +0000)]
Add thread locking to SSL and Kerberos connections.

I have removed the docs mentioning that SSL and Kerberos are not
thread-safe.

Manfred Spraul

20 years agoDocument fallback to tcp/ip localhost connection from psql on ports
Bruce Momjian [Wed, 24 Mar 2004 03:27:46 +0000 (03:27 +0000)]
Document fallback to tcp/ip localhost connection from psql on ports
without unix domain sockets (win32).

20 years agoFix some whitespace formatting, and remove an overly-verbose
Neil Conway [Wed, 24 Mar 2004 03:22:35 +0000 (03:22 +0000)]
Fix some whitespace formatting, and remove an overly-verbose
parameter description: postgresql.conf is not the place for
documentating the functionality of a GUC var.