]> granicus.if.org Git - postgresql/log
postgresql
22 years agooid is needed, it is added at the end of the struct (after the null
Bruce Momjian [Sat, 20 Jul 2002 05:16:59 +0000 (05:16 +0000)]
oid is needed, it is added at the end of the struct (after the null
bitmap, if present).

Per Tom Lane's suggestion the information whether a tuple has an oid
or not is carried in the tuple descriptor.  For debugging reasons
tdhasoid is of type char, not bool.  There are predefined values for
WITHOID, WITHOUTOID and UNDEFOID.

This patch has been generated against a cvs snapshot from last week
and I don't expect it to apply cleanly to current sources.  While I
post it here for public review, I'm working on a new version against a
current snapshot.  (There's been heavy activity recently; hope to
catch up some day ...)

This is a long patch;  if it is too hard to swallow, I can provide it
in smaller pieces:

Part 1:  Accessor macros
Part 2:  tdhasoid in TupDesc
Part 3:  Regression test
Part 4:  Parameter withoid to heap_addheader
Part 5:  Eliminate t_oid from HeapTupleHeader

Part 2 is the most hairy part because of changes in the executor and
even in the parser;  the other parts are straightforward.

Up to part 4 the patched postmaster stays binary compatible to
databases created with an unpatched version.  Part 5 is small (100
lines) and finally breaks compatibility.

Manfred Koizar

22 years agoThe attached patch fixes a build problem with GEQO when using the
Bruce Momjian [Sat, 20 Jul 2002 04:59:10 +0000 (04:59 +0000)]
The attached patch fixes a build problem with GEQO when using the
PX recombination operator, changes some elog() messages from LOG
to DEBUG1, puts some debugging functions inside the appropriate
#ifdef (not enabled by default), and makes a few other minor
cleanups.

BTW, the elog() change is motivated by at least one user who
has sent a concerned email to -general asking exactly what the
"ERX recombination operator" is, and what it is doing to their
DBMS.

Neil Conway

22 years agoAdd new vacuum regression test files.
Bruce Momjian [Sat, 20 Jul 2002 04:58:14 +0000 (04:58 +0000)]
Add new vacuum regression test files.

22 years agoThis patch fixes a regression caused by my recent changes to heap
Bruce Momjian [Sat, 20 Jul 2002 04:57:13 +0000 (04:57 +0000)]
This patch fixes a regression caused by my recent changes to heap
tuple header.  The fix is based on the thought that HEAP_MOVED_IN is
not needed any more as soon as HEAP_XMIN_COMMITTED has been set.  So
in tqual.c and vacuum.c the HEAP_MOVED bits are cleared when
HEAP_XMIN_COMMITTED is set.

Vacuum robustness is enhanced by rearranging ifs, so that we have a
chance to elog(ERROR, ...) before an assertion fails.

A new regression test is included.

Manfred Koizar

22 years agoApply patches from Neil Conway.
Tatsuo Ishii [Sat, 20 Jul 2002 03:02:01 +0000 (03:02 +0000)]
Apply patches from Neil Conway.

> Hi Tatsuo,
>
> I've attached a patch for the version of pgbench in CVS. It includes the
> following changes:
>
>     - fix some spelling mistakes, indentation stuff, etc.
>
>     - minor code cleanup -- (void) args instead of (), etc.
>
>     - allocate the state array dynamically, so that it is only as
>     large as needed. This reduces the memory consumption of pgbench
>     slightly, and makes a larger MAXCLIENTS setting possible
>
>     - (the only controversial change) add an option "-l" to log
>     transaction latencies to a file. The "transaction latency"
>     is the time between when the BEGIN is issued and the transaction
>     commits. This is written to a file, along with the client #
>     and the transaction #. The data in the file can then be used
>     for things like:
>
>         - consistency analysis: is the TPS the same through the
>         entire run of pgbench, or does it change?
>
>         - more detailed stats: what is the average latency, worse-case
>         latency, best-case latency?
>
>         - graphs: feed the data to gnuplot, graph latency versus. time
>
>         - etc.
>
>     I was going to store this data in memory and write it to disk
>     at the end of the pgbench run, but that isn't feasible because
>     the data can be very large: for example, ~70MB if benchmarking
>     128 clients doing 100,000 transactions each.
>
> Cheers,
>
> Neil

22 years agoSmall performance improvement --- use indexscan not heapscan to find
Tom Lane [Fri, 19 Jul 2002 22:21:17 +0000 (22:21 +0000)]
Small performance improvement --- use indexscan not heapscan to find
pg_statistic rows to delete while dropping a relation.

22 years agoMake note that ALTER TABLE can change table owner these days.
Tom Lane [Fri, 19 Jul 2002 18:53:50 +0000 (18:53 +0000)]
Make note that ALTER TABLE can change table owner these days.

22 years agoComplete TODO item:
Bruce Momjian [Fri, 19 Jul 2002 17:35:11 +0000 (17:35 +0000)]
Complete TODO item:

* -Add BSD-licensed qsort() for Solaris

22 years agoDone:
Bruce Momjian [Fri, 19 Jul 2002 17:26:35 +0000 (17:26 +0000)]
Done:

> * -Add BSD-licensed qsort() for Solaris

22 years agoAnother "$@" cleanup with new syntax, ${1+"$@"}.
Bruce Momjian [Fri, 19 Jul 2002 15:31:43 +0000 (15:31 +0000)]
Another "$@" cleanup with new syntax, ${1+"$@"}.

22 years agoFix "$@" to ${1:+"$@"} for older shells.
Bruce Momjian [Fri, 19 Jul 2002 13:50:06 +0000 (13:50 +0000)]
Fix "$@" to ${1:+"$@"} for older shells.

22 years agoOops. Too much ifdef out.
Tatsuo Ishii [Fri, 19 Jul 2002 11:09:25 +0000 (11:09 +0000)]
Oops. Too much ifdef out.

22 years agoTemporary ifdef out migrating functions to avoid compiler warnings.
Tatsuo Ishii [Fri, 19 Jul 2002 00:22:24 +0000 (00:22 +0000)]
Temporary ifdef out migrating functions to avoid compiler warnings.

22 years agoDone:
Bruce Momjian [Fri, 19 Jul 2002 00:17:53 +0000 (00:17 +0000)]
Done:

> * -HOLDER/HOLDERTAB rename to PROCLOCK/PROCLOCKTAG (Bruce)

22 years agoComplete TODO item:
Bruce Momjian [Fri, 19 Jul 2002 00:17:40 +0000 (00:17 +0000)]
Complete TODO item:

* -HOLDER/HOLDERTAB rename to PROCLOCK/PROCLOCKTAG

22 years agopg_cast table, and standards-compliant CREATE/DROP CAST commands, plus
Peter Eisentraut [Thu, 18 Jul 2002 23:11:32 +0000 (23:11 +0000)]
pg_cast table, and standards-compliant CREATE/DROP CAST commands, plus
extension to create binary compatible casts.  Includes dependency tracking
as well.

pg_proc.proimplicit is now defunct, but will be removed in a separate
commit.

pg_dump provides a migration path from the previous scheme to declare
casts.  Dumping binary compatible casts is currently impossible, though.

22 years agoDone:
Bruce Momjian [Thu, 18 Jul 2002 23:07:09 +0000 (23:07 +0000)]
Done:

> * -Merge LockMethodCtl and LockMethodTable into one shared structure (Bruce)

22 years agoComplete TODO item:
Bruce Momjian [Thu, 18 Jul 2002 23:06:20 +0000 (23:06 +0000)]
Complete TODO item:

* Merge LockMethodCtl and LockMethodTable into one shared structure (Bruce)

22 years agoCreate directory before installing files.
Peter Eisentraut [Thu, 18 Jul 2002 22:58:08 +0000 (22:58 +0000)]
Create directory before installing files.

22 years agoNew depend code doesn't need sequence drop.
Bruce Momjian [Thu, 18 Jul 2002 22:31:44 +0000 (22:31 +0000)]
New depend code doesn't need sequence drop.

22 years agoSic transit src/utils.
Tom Lane [Thu, 18 Jul 2002 22:15:10 +0000 (22:15 +0000)]
Sic transit src/utils.

22 years agoBack out BETWEEN node patch, was causing initdb failure.
Bruce Momjian [Thu, 18 Jul 2002 17:14:20 +0000 (17:14 +0000)]
Back out BETWEEN node patch, was causing initdb failure.

22 years ago New depend code auto-drops sequence, fix copy2.out.
Bruce Momjian [Thu, 18 Jul 2002 17:11:49 +0000 (17:11 +0000)]
 New depend code auto-drops sequence, fix copy2.out.

22 years agoImplement DROP SCHEMA. It lacks support for dropping conversions and
Tom Lane [Thu, 18 Jul 2002 16:47:26 +0000 (16:47 +0000)]
Implement DROP SCHEMA.  It lacks support for dropping conversions and
operator classes, both of which are schema-local and so should really
be droppable.

22 years agoFix copy-and-pasteo (CASCADE/RESTRICT in wrong place).
Tom Lane [Thu, 18 Jul 2002 15:49:08 +0000 (15:49 +0000)]
Fix copy-and-pasteo (CASCADE/RESTRICT in wrong place).

22 years agoFix typo (PG_UNICODE -> PG_UTF8)
Tatsuo Ishii [Thu, 18 Jul 2002 05:07:30 +0000 (05:07 +0000)]
Fix typo (PG_UNICODE -> PG_UTF8)

22 years agoDone:
Bruce Momjian [Thu, 18 Jul 2002 05:01:51 +0000 (05:01 +0000)]
Done:

> * -Make sure all block numbers are unsigned to increase maximum table size

22 years agoUnique and primary key constraints are both dumped using ALTER TABLE
Bruce Momjian [Thu, 18 Jul 2002 04:50:51 +0000 (04:50 +0000)]
Unique and primary key constraints are both dumped using ALTER TABLE
statements.  Unique indexes with CREATE INDEX.

Basically, pg_constraint left outer'd to pg_index.

Rod Taylor

22 years agoREVOKE ALL ON FUNCTION nonexistant() FROM PUBLIC;
Bruce Momjian [Thu, 18 Jul 2002 04:50:10 +0000 (04:50 +0000)]
REVOKE ALL ON FUNCTION nonexistant() FROM PUBLIC;

Used to report that GRANT could not find function nonexistant.

Rod Taylor

22 years agoFix for PgTransaction class to make these visible to C apps:
Bruce Momjian [Thu, 18 Jul 2002 04:49:30 +0000 (04:49 +0000)]
Fix for PgTransaction class to make these visible to C apps:

   ExecStatusType BeginTransaction();
   ExecStatusType EndTransaction();

Piotr Klaban

22 years agoHere (finally ;-)) is a doc patch covering the Table Function C API. It
Bruce Momjian [Thu, 18 Jul 2002 04:47:17 +0000 (04:47 +0000)]
Here (finally ;-)) is a doc patch covering the Table Function C API. It
reflects the changes in the tablefunc-fix patch that I sent in the other
day. It also refers to "see contrib/tablefunc for more examples", which
is next on my list of things to finish and submit.

Joe Conway

22 years agoThe attached patch fixes a tiny memory leak in psql, when using
Bruce Momjian [Thu, 18 Jul 2002 04:46:24 +0000 (04:46 +0000)]
The attached patch fixes a tiny memory leak in psql, when using
the 'expanded' output mode (\x).

Neil Conway

22 years agohere are the copy2.sql and copy2.out files for the new regression
Bruce Momjian [Thu, 18 Jul 2002 04:45:51 +0000 (04:45 +0000)]
here are the copy2.sql and copy2.out files for the new regression
tests

Brent Verner

22 years agoThe attached patch (against HEAD) implements
Bruce Momjian [Thu, 18 Jul 2002 04:43:51 +0000 (04:43 +0000)]
The attached patch (against HEAD) implements

  COPY x (a,d,c,b) from stdin;
  COPY x (a,c) to stdout;

as well as the corresponding changes to pg_dump to use the new
functionality.  This functionality is not available when using
the BINARY option.  If a column is not specified in the COPY FROM
statement, its default values will be used.

In addition to this functionality, I tweaked a couple of the
error messages emitted by the new COPY <options> checks.

Brent Verner

22 years agoYou made a tiny little type in a comment in parsenodes.h (abotu instead
Bruce Momjian [Thu, 18 Jul 2002 04:42:29 +0000 (04:42 +0000)]
You made a tiny little type in a comment in parsenodes.h (abotu instead
of about).

Christopher Kings-Lynne

22 years agoFinished the Between patch Christopher started.
Bruce Momjian [Thu, 18 Jul 2002 04:41:46 +0000 (04:41 +0000)]
Finished the Between patch Christopher started.

Implements between (symmetric / asymmetric) as a node.

Executes the left or right expression once, makes a Const out of the
resulting Datum and executes the >=, <= portions out of the Const sets.

Of course, the parser does a fair amount of preparatory work for this to
happen.

Rod Taylor

22 years agoHere is a patch for the Table Function API. It fixes a bug found by Neil
Bruce Momjian [Thu, 18 Jul 2002 04:40:30 +0000 (04:40 +0000)]
Here is a patch for the Table Function API. It fixes a bug found by Neil

Conway (BuildTupleFromCStrings sets NULL for pass-by-value types when
intended value is 0). It also implements some other improvements
suggested by Neil.

Joe Conway

22 years agoContrib port/ usage is same as others so no need for rules.
Bruce Momjian [Thu, 18 Jul 2002 04:33:39 +0000 (04:33 +0000)]
Contrib port/ usage is same as others so no need for rules.

22 years agoAdd comment to file about port files.
Bruce Momjian [Thu, 18 Jul 2002 04:30:36 +0000 (04:30 +0000)]
Add comment to file about port files.

22 years agoFix breakage for pltcl modules. pg_get_enconv_by_encoding() has been
Tatsuo Ishii [Thu, 18 Jul 2002 04:16:07 +0000 (04:16 +0000)]
Fix breakage for pltcl modules. pg_get_enconv_by_encoding() has been
changed since CREATE CONVERSION supported.

22 years agoMove libc replacement files from src/backend/port to src/port.
Bruce Momjian [Thu, 18 Jul 2002 04:13:59 +0000 (04:13 +0000)]
Move libc replacement files from src/backend/port to src/port.

22 years agoMake src/backend/port/*.c file location dependent only on configure.in.
Bruce Momjian [Thu, 18 Jul 2002 03:59:49 +0000 (03:59 +0000)]
Make src/backend/port/*.c file location dependent only on configure.in.

22 years agoI have committed many support files for CREATE CONVERSION. Default
Tatsuo Ishii [Thu, 18 Jul 2002 02:02:30 +0000 (02:02 +0000)]
I have committed many support files for CREATE CONVERSION. Default
conversion procs and conversions are added in initdb. Currently
supported conversions are:

UTF-8(UNICODE) <--> SQL_ASCII, ISO-8859-1 to 16, EUC_JP, EUC_KR,
    EUC_CN, EUC_TW, SJIS, BIG5, GBK, GB18030, UHC,
    JOHAB, TCVN

EUC_JP <--> SJIS
EUC_TW <--> BIG5
MULE_INTERNAL <--> EUC_JP, SJIS, EUC_TW, BIG5

Note that initial contents of pg_conversion system catalog are created
in the initdb process. So doing initdb required is ideal, it's
possible to add them to your databases by hand, however. To accomplish
this:

psql -f your_postgresql_install_path/share/conversion_create.sql your_database

So I did not bump up the version in cataversion.h.

TODO:
Add more conversion procs
Add [CASCADE|RESTRICT] to DROP CONVERSION
Add tuples to pg_depend
Add regression tests
Write docs
Add SQL99 CONVERT command?
--
Tatsuo Ishii

22 years agoDone, not sure when, reported by Neil Conway:
Bruce Momjian [Wed, 17 Jul 2002 22:15:20 +0000 (22:15 +0000)]
Done, not sure when, reported by Neil Conway:

> * -Report failure to find readline or zlib at end of configure run

22 years agoImprove getopt_long search, per Peter:
Bruce Momjian [Wed, 17 Jul 2002 02:54:01 +0000 (02:54 +0000)]
Improve getopt_long search, per Peter:

AC_SEARCH_LIBS(getopt_long, [getopt])

22 years agoAdd more dependency insertions --- this completes the basic pg_depend
Tom Lane [Tue, 16 Jul 2002 22:12:20 +0000 (22:12 +0000)]
Add more dependency insertions --- this completes the basic pg_depend
functionality.  Of note: dropping a table that has a SERIAL column
defined now drops the associated sequence automatically.

22 years agofixed bug in support for timestamp without time zone reported by Yuva Chandolu (ychan...
Barry Lind [Tue, 16 Jul 2002 21:05:17 +0000 (21:05 +0000)]
fixed bug in support for timestamp without time zone reported by Yuva Chandolu (ychandolu@ebates.com)

22 years agoOops, sometimes strtol isn't called in pg_atoi, so we do need that badp
Bruce Momjian [Tue, 16 Jul 2002 18:34:16 +0000 (18:34 +0000)]
Oops, sometimes strtol isn't called in pg_atoi, so we do need that badp
check.

22 years agoFix tid to in/out as unsigned.
Bruce Momjian [Tue, 16 Jul 2002 17:55:25 +0000 (17:55 +0000)]
Fix tid to in/out as unsigned.

22 years agoAdd initdb code to set up initial contents of pg_depend, pinning all
Tom Lane [Tue, 16 Jul 2002 17:48:46 +0000 (17:48 +0000)]
Add initdb code to set up initial contents of pg_depend, pinning all
objects created during initdb (except for the system views, which I
think do not need to be pinned).

22 years ago> pgsql-bugs@postgresql.org wrote:
Bruce Momjian [Tue, 16 Jul 2002 17:05:46 +0000 (17:05 +0000)]
> pgsql-bugs@postgresql.org wrote:
>  > David Clark (dclarknospam@opsi.co.za) reports a bug with a severity
>  > Table 3-7 SQL Literal escaped octets shows the input escape
>  > representation for a single quote as '\\'' , but the third paragraph
>  > below table 3-8 SQL Output Escaped Octets says that the single quote
>  > must be input as '\''
>
> Nice catch. '\'' is correct as shown in the example in Table 3-7.
>
>  >
>  > Also in the same paragraph mentioned above it says input for the
>  > single quote must be '\'' (or '\\134') shouldn't this be (or '\\047')
>
> Also a bug. Should be '\\047', as you pointed out.
>

Here's a patch to fix the binary string doc errors.

Joe Conway

22 years agoAdd conversion procs for CREATE CONVERSION
Tatsuo Ishii [Tue, 16 Jul 2002 09:25:06 +0000 (09:25 +0000)]
Add conversion procs for CREATE CONVERSION

22 years agoFix typo
Tatsuo Ishii [Tue, 16 Jul 2002 06:58:44 +0000 (06:58 +0000)]
Fix typo

22 years agoAdd FindDefaultConversionProc
Tatsuo Ishii [Tue, 16 Jul 2002 06:58:14 +0000 (06:58 +0000)]
Add FindDefaultConversionProc

22 years agoAdd code to extract dependencies from an expression tree, and use it
Tom Lane [Tue, 16 Jul 2002 05:53:34 +0000 (05:53 +0000)]
Add code to extract dependencies from an expression tree, and use it
to build dependencies for rules, constraint expressions, and default
expressions.  Repair some problems in the original design of
recursiveDeletion() exposed by more complex dependency sets.  Fix
regression tests that were deleting things in illegal sequences.

22 years agoAdd DLLINIT mention in port/Makefile.
Bruce Momjian [Tue, 16 Jul 2002 05:49:38 +0000 (05:49 +0000)]
Add DLLINIT mention in port/Makefile.

22 years agoMove few remaining src/utils files to backend/port so everything is in
Bruce Momjian [Tue, 16 Jul 2002 05:46:36 +0000 (05:46 +0000)]
Move few remaining src/utils files to backend/port so everything is in
one place.  Everything may be moved to src/utils eventually.

Add DLLINIT variable to simplify makfiles.

22 years agoAdd mention of TOAST storage for character columns.
Bruce Momjian [Tue, 16 Jul 2002 04:45:59 +0000 (04:45 +0000)]
Add mention of TOAST storage for character columns.

22 years agoMark 'line' as 'not implemented' in SGML and psql \dT, per Thomas
Bruce Momjian [Tue, 16 Jul 2002 03:30:27 +0000 (03:30 +0000)]
Mark 'line' as 'not implemented' in SGML and psql \dT, per Thomas
Lockhart.  initdb not forced.

22 years agoThis fixes 2 inaccuracies in the recently added SQL99 feature list docs.
Bruce Momjian [Tue, 16 Jul 2002 00:51:37 +0000 (00:51 +0000)]
This fixes 2 inaccuracies in the recently added SQL99 feature list docs.
UNIQUE and DISTINCT predicates are both listed as implemented -- AFAIK,
neither is.

I also included another trivial patch which adds the default location
of the DSSSL stylesheets on my system (Debian unstable, docbook-dsssl
1.76) to the list of paths that configure looks for.

Neil Conway

22 years agoAdd comment for isbn,issn data type, from Pete St. Onge
Bruce Momjian [Tue, 16 Jul 2002 00:48:30 +0000 (00:48 +0000)]
Add comment for isbn,issn data type, from Pete St. Onge

22 years agoItem not needed:
Bruce Momjian [Mon, 15 Jul 2002 23:34:34 +0000 (23:34 +0000)]
Item not needed:

< * Use our own getopt() for FreeBSD/OpenBSD to allow --xxx flags (Bruce)

22 years agoRemove indenting of Makefile comments, per Peter.\
Bruce Momjian [Mon, 15 Jul 2002 23:32:28 +0000 (23:32 +0000)]
Remove indenting of Makefile comments, per Peter.\

22 years agoRemove utils/getopt.c file, not needed, it only allowed "--flag"
Bruce Momjian [Mon, 15 Jul 2002 22:48:54 +0000 (22:48 +0000)]
Remove utils/getopt.c file, not needed, it only allowed "--flag"
warnings, and we now look for getopt_long in libgetopt.a.

22 years agoAdd search for libgetopt.a in hopes of finding getopt_long().
Bruce Momjian [Mon, 15 Jul 2002 22:41:45 +0000 (22:41 +0000)]
Add search for libgetopt.a in hopes of finding getopt_long().

22 years agoRemove certain Makefile dependencies by using full pathnames in
Bruce Momjian [Mon, 15 Jul 2002 21:34:05 +0000 (21:34 +0000)]
Remove certain Makefile dependencies by using full pathnames in
configure.in.

22 years agoUse the dependency mechanism to manage column defaults. We need this
Tom Lane [Mon, 15 Jul 2002 16:33:32 +0000 (16:33 +0000)]
Use the dependency mechanism to manage column defaults.  We need this
so that dependencies in default expressions (on operators, functions,
etc) can be expressed properly.

22 years agoFix make_ctags for exhuberant tags.
Bruce Momjian [Mon, 15 Jul 2002 14:45:51 +0000 (14:45 +0000)]
Fix make_ctags for exhuberant tags.

22 years agoFix a bug about the handling of '.' in parse.c.
Hiroshi Inoue [Mon, 15 Jul 2002 02:56:39 +0000 (02:56 +0000)]
Fix a bug about the handling of '.' in parse.c.

22 years agoImprove relcache.c error reporting for the next guy who has to debug
Bruce Momjian [Mon, 15 Jul 2002 01:57:51 +0000 (01:57 +0000)]
Improve relcache.c error reporting for the next guy who has to debug
this thing.

22 years agoFix \? and \pset pager handling. \? wasn't honoring pager before.
Bruce Momjian [Mon, 15 Jul 2002 01:56:25 +0000 (01:56 +0000)]
Fix \? and \pset pager handling.  \? wasn't honoring pager before.

22 years agoAdd COMMENT ON CONSTRAINT facility (from Rod Taylor's pg_constraint patch).
Tom Lane [Sun, 14 Jul 2002 23:38:13 +0000 (23:38 +0000)]
Add COMMENT ON CONSTRAINT facility (from Rod Taylor's pg_constraint patch).
Fix comment.c to not depend on parser token values, per discussion awhile
back.

22 years agoRemove no-longer-appropriate notes about lack of CASCADE/RESTRICT behavior.
Tom Lane [Sun, 14 Jul 2002 22:47:56 +0000 (22:47 +0000)]
Remove no-longer-appropriate notes about lack of CASCADE/RESTRICT behavior.

22 years agoIn DeleteAttributeTuples, use a single indexscan instead of the multiple
Tom Lane [Sun, 14 Jul 2002 21:08:08 +0000 (21:08 +0000)]
In DeleteAttributeTuples, use a single indexscan instead of the multiple
scans that will most likely be caused by SearchSysCache probes.  Also,
share some code between index deletion and table deletion.

22 years agoFix testing of partial-index predicates to work correctly in cases where
Tom Lane [Sat, 13 Jul 2002 19:20:34 +0000 (19:20 +0000)]
Fix testing of partial-index predicates to work correctly in cases where
varno of index's relation is not 1.  This embarrassing oversight pointed
out by Dmitry Tkach 12-Jul-02.

22 years agoMark as done, per Rod:
Bruce Momjian [Sat, 13 Jul 2002 02:42:16 +0000 (02:42 +0000)]
Mark as done, per Rod:

> o -Add ALTER TABLE DROP non-CHECK CONSTRAINT
> * -Allow psql \d to show foreign keys
> * -Auto-destroy sequence on DROP of table with SERIAL; perhaps a separate
> * -Prevent column dropping if column is used by foreign key
> * -Automatically drop constraints/functions when object is dropped
> * -Make foreign key constraints clearer in dump file
> * -Make foreign keys easier to identify

22 years agoBoth done:
Bruce Momjian [Sat, 13 Jul 2002 02:21:15 +0000 (02:21 +0000)]
Both done:

o -Add SET or BEGIN timeout parameter to cancel query
> * -Add pg_depend table for dependency recording; use sysrelid, oid,

22 years agoAdd SET statement_timeout capability. Timeout is in ms. A value of
Bruce Momjian [Sat, 13 Jul 2002 01:02:14 +0000 (01:02 +0000)]
Add SET statement_timeout capability.  Timeout is in ms.  A value of
zero turns off the timer.

22 years agoRe-add -I documentation, per Tom.
Bruce Momjian [Sat, 13 Jul 2002 00:55:53 +0000 (00:55 +0000)]
Re-add -I documentation, per Tom.

22 years agoSecond phase of committing Rod Taylor's pg_depend/pg_constraint patch.
Tom Lane [Fri, 12 Jul 2002 18:43:19 +0000 (18:43 +0000)]
Second phase of committing Rod Taylor's pg_depend/pg_constraint patch.
pg_relcheck is gone; CHECK, UNIQUE, PRIMARY KEY, and FOREIGN KEY
constraints all have real live entries in pg_constraint.  pg_depend
exists, and RESTRICT/CASCADE options work on most kinds of DROP;
however, pg_depend is not yet very well populated with dependencies.
(Most of the ones that are present at this point just replace formerly
hardwired associations, such as the implicit drop of a relation's pg_type
entry when the relation is dropped.)  Need to add more logic to create
dependency entries, improve pg_dump to dump constraints in place of
indexes and triggers, and add some regression tests.

22 years agoChanged getCrossReference to return information about each composite key seperately
Dave Cramer [Fri, 12 Jul 2002 13:07:48 +0000 (13:07 +0000)]
Changed getCrossReference to return information about each composite key seperately
It used to return them as a,b in one row, and now returns
a in one row, and b in one row

22 years agoAn improvement of User Interface(Setup dialog) by Dave Page.
Hiroshi Inoue [Fri, 12 Jul 2002 02:02:26 +0000 (02:02 +0000)]
An improvement of User Interface(Setup dialog) by Dave Page.

22 years ago1) Fix a memory leak in use declare/fetch mode.
Hiroshi Inoue [Fri, 12 Jul 2002 01:41:25 +0000 (01:41 +0000)]
1) Fix a memory leak in use declare/fetch mode.
2) Change default build mode to multibyte(Windows).

22 years agoNot a valid item:
Bruce Momjian [Thu, 11 Jul 2002 21:41:25 +0000 (21:41 +0000)]
Not a valid item:

< * Add version file format stamp to heap and other table types

22 years agoRemove duplicate:
Bruce Momjian [Thu, 11 Jul 2002 21:40:20 +0000 (21:40 +0000)]
Remove duplicate:

< * Use BlockNumber rather than int where appropriate

22 years agoChange error messages ExecAppend->ExecInsert and ExecReplace->ExecUpdate
Bruce Momjian [Thu, 11 Jul 2002 21:36:20 +0000 (21:36 +0000)]
Change error messages ExecAppend->ExecInsert and ExecReplace->ExecUpdate
as discussed on hackers.

22 years agoRemove references to pre-7.1; too old.
Bruce Momjian [Thu, 11 Jul 2002 16:38:08 +0000 (16:38 +0000)]
Remove references to pre-7.1;  too old.

22 years agoUpdate FAQ.
Bruce Momjian [Thu, 11 Jul 2002 16:35:56 +0000 (16:35 +0000)]
Update FAQ.

22 years agoAdd new CREATE CONVERSION/DROP CONVERSION command.
Tatsuo Ishii [Thu, 11 Jul 2002 07:39:28 +0000 (07:39 +0000)]
Add new CREATE CONVERSION/DROP CONVERSION command.
This is the first cut toward CREATE CONVERSION/DROP CONVERSION implementaion.
The commands can now add/remove tuples to the new pg_conversion system
catalog, but that's all. Still need work to make them actually working.
Documentations, regression tests also need work.

22 years agoDone, according to Rod Taylor:
Bruce Momjian [Thu, 11 Jul 2002 02:56:49 +0000 (02:56 +0000)]
Done, according to Rod Taylor:

> * -Add ALTER TRIGGER ... RENAME

22 years agoFix getopt flags. Man thing thing was messed up.
Bruce Momjian [Thu, 11 Jul 2002 02:00:14 +0000 (02:00 +0000)]
Fix getopt flags.  Man thing thing was messed up.

22 years agoFix another --help typo in pg_restore.
Bruce Momjian [Thu, 11 Jul 2002 01:57:12 +0000 (01:57 +0000)]
Fix another --help typo in pg_restore.

22 years agoMake NAMEDATALEN changeable at compile time.
Hiroshi Inoue [Thu, 11 Jul 2002 01:52:46 +0000 (01:52 +0000)]
Make NAMEDATALEN changeable at compile time.

22 years agoFix pg_restore flags in documentation.
Bruce Momjian [Wed, 10 Jul 2002 02:57:31 +0000 (02:57 +0000)]
Fix pg_restore flags in documentation.

22 years agofix bug in getTime() with fractional seconds reported by Laurette Cisneros (laurette...
Barry Lind [Wed, 10 Jul 2002 00:51:36 +0000 (00:51 +0000)]
fix bug in getTime() with fractional seconds reported by Laurette Cisneros (laurette@nextbus.com)

22 years agoOops, proper initialization for domainTypMod was none at all, not 0.
Bruce Momjian [Tue, 9 Jul 2002 13:52:14 +0000 (13:52 +0000)]
Oops, proper initialization for domainTypMod was none at all, not 0.

22 years agoFix compiler warning:
Bruce Momjian [Tue, 9 Jul 2002 05:21:35 +0000 (05:21 +0000)]
Fix compiler warning:

int32           domainTypMod = NULL;

should be:

int32           domainTypMod = 0;

22 years agoFix typo mentioned by Rich Morin.
Bruce Momjian [Tue, 9 Jul 2002 04:47:07 +0000 (04:47 +0000)]
Fix typo mentioned by Rich Morin.

22 years agoDone:
Bruce Momjian [Mon, 8 Jul 2002 02:01:02 +0000 (02:01 +0000)]
Done:

> * -Make one version of simple_prompt() in code (Bruce, Tom)

22 years agoFix case syntax for freebsd template.
Bruce Momjian [Mon, 8 Jul 2002 01:54:30 +0000 (01:54 +0000)]
Fix case syntax for freebsd template.