]> granicus.if.org Git - postgresql/log
postgresql
20 years agoDone:
Bruce Momjian [Wed, 19 May 2004 21:09:53 +0000 (21:09 +0000)]
Done:

> * -Make pg_restore continue after errors, so it acts more like pg_dump scripts

20 years agoOnly do find_my_exec if it doesn't come from the postmaster.
Bruce Momjian [Wed, 19 May 2004 19:39:27 +0000 (19:39 +0000)]
Only do find_my_exec if it doesn't come from the postmaster.

20 years agoAdd get_pkglib_path to postmaster.c
Bruce Momjian [Wed, 19 May 2004 19:11:25 +0000 (19:11 +0000)]
Add get_pkglib_path to postmaster.c

20 years agoMove find_my_exec lower so elog() works, per Tom.
Bruce Momjian [Wed, 19 May 2004 18:58:44 +0000 (18:58 +0000)]
Move find_my_exec lower so elog() works, per Tom.

20 years agoRe-add elog() to exec.c, per Tom.
Bruce Momjian [Wed, 19 May 2004 17:15:21 +0000 (17:15 +0000)]
Re-add elog() to exec.c, per Tom.

20 years agoAdd MinGW to cases where libdir should be used for regression tests:
Bruce Momjian [Wed, 19 May 2004 05:20:27 +0000 (05:20 +0000)]
Add MinGW to cases where libdir should be used for regression tests:

    case $host_platform in
        *-*-cygwin*|*-*-mingw)
            PATH=$libdir:$PATH
            export PATH
            ;;
    esac

20 years agoWin32 needs -lpgport for dll creation of ecpglib.
Bruce Momjian [Wed, 19 May 2004 05:14:58 +0000 (05:14 +0000)]
Win32 needs -lpgport for dll creation of ecpglib.

20 years agoSupress verbose stderr messages while trying to find my exec path.
Bruce Momjian [Wed, 19 May 2004 04:36:33 +0000 (04:36 +0000)]
Supress verbose stderr messages while trying to find my exec path.

Per report from Christopher.

20 years agoMinor correction for previous SQLSTATE patch: I changed dsqrt() to emit the
Neil Conway [Wed, 19 May 2004 04:32:26 +0000 (04:32 +0000)]
Minor correction for previous SQLSTATE patch: I changed dsqrt() to emit the
right error code previously, and this patch applies an analogous change
to numeric_sqrt().

20 years agoWin32 can't have the same function coming from two library object files,
Bruce Momjian [Wed, 19 May 2004 04:21:49 +0000 (04:21 +0000)]
Win32 can't have the same function coming from two library object files,
so we make is_absolute_path a macro so libpq doesn't use path.o.

20 years agoAllow zic to compile by defining a "" value for my_exec_path. It isn't
Bruce Momjian [Wed, 19 May 2004 01:17:42 +0000 (01:17 +0000)]
Allow zic to compile by defining a "" value for my_exec_path.  It isn't
used by zic anyway.

20 years agosetRuleCheckAsUser has to be applied to any subqueries appearing in a
Tom Lane [Tue, 18 May 2004 22:49:51 +0000 (22:49 +0000)]
setRuleCheckAsUser has to be applied to any subqueries appearing in a
rule's event_qual, not only to the rule's action.  Per example from
Arturs Zoldners.

20 years agoTrim down relocatable docs, per Peter.
Bruce Momjian [Tue, 18 May 2004 21:46:25 +0000 (21:46 +0000)]
Trim down relocatable docs, per Peter.

20 years agoAdd some documentation on relocatable installs.
Bruce Momjian [Tue, 18 May 2004 20:34:01 +0000 (20:34 +0000)]
Add some documentation on relocatable installs.

20 years agoMove get_pkglib_path up into main.c too.
Bruce Momjian [Tue, 18 May 2004 20:27:25 +0000 (20:27 +0000)]
Move get_pkglib_path up into main.c too.

20 years agoMove find_my_exec() way up into main.c so it is available to the
Bruce Momjian [Tue, 18 May 2004 20:18:59 +0000 (20:18 +0000)]
Move find_my_exec() way up into main.c so it is available to the
timezone code and other places.

Remove elog() calls from find_my_exec;  do fprintf(stderr) instead.  We
can then remove the exec.c handling in the makefile because it doesn't
have to be built to suppress elog calls.

20 years agoAdd:
Bruce Momjian [Tue, 18 May 2004 16:06:08 +0000 (16:06 +0000)]
Add:

> * Allow binaries to be statically linked so they are more easily relocated

20 years agoUpdate Russian FAQ.
Bruce Momjian [Tue, 18 May 2004 13:18:30 +0000 (13:18 +0000)]
Update Russian FAQ.

Viktor Vislobokov

20 years agoRun autoconf for link update.
Bruce Momjian [Tue, 18 May 2004 04:12:00 +0000 (04:12 +0000)]
Run autoconf for link update.

20 years agoImprove win32 link test. Previous commit added most of the code, but
Bruce Momjian [Tue, 18 May 2004 04:11:29 +0000 (04:11 +0000)]
Improve win32 link test.  Previous commit added most of the code, but
this commit enables only for Mingw.

20 years agoUpdate Makefile dependencies for Win32 timezones, per Claudio.
Bruce Momjian [Tue, 18 May 2004 04:10:33 +0000 (04:10 +0000)]
Update Makefile dependencies for Win32 timezones, per Claudio.

20 years agoClean up some relative path install issues with Claudio's help.
Bruce Momjian [Tue, 18 May 2004 03:36:45 +0000 (03:36 +0000)]
Clean up some relative path install issues with Claudio's help.

20 years agoImprove configure win32 link test to run from config.status.
Bruce Momjian [Mon, 17 May 2004 19:14:47 +0000 (19:14 +0000)]
Improve configure win32 link test to run from config.status.

Andrew Dunstan

20 years agoAdd documentation that installations are relocatable unless configure
Bruce Momjian [Mon, 17 May 2004 16:06:25 +0000 (16:06 +0000)]
Add documentation that installations are relocatable unless configure
locations are overridden.

20 years agoReformat sgml doc paragraph.
Bruce Momjian [Mon, 17 May 2004 15:00:50 +0000 (15:00 +0000)]
Reformat sgml doc paragraph.

20 years agoReorganize code to allow path-relative installs.
Bruce Momjian [Mon, 17 May 2004 14:35:34 +0000 (14:35 +0000)]
Reorganize code to allow path-relative installs.

Create new get_* functions to access compiled-in paths and adjust if
relative installs are to be used.

Clean up substitute_libpath_macro() code.

20 years agoRename backendbin to backend_exec in initdb.c.
Bruce Momjian [Mon, 17 May 2004 13:17:29 +0000 (13:17 +0000)]
Rename backendbin to backend_exec in initdb.c.

20 years agoAdd documentation for the new "dollar quoting" feature, and update existing
Neil Conway [Sun, 16 May 2004 23:22:08 +0000 (23:22 +0000)]
Add documentation for the new "dollar quoting" feature, and update existing
examples to use dollar quoting when appropriate. Original patch from David
Fetter, additional work and editorializing by Neil Conway.

20 years agoChange ln(), log(), power(), and sqrt() to emit the correct SQLSTATE
Neil Conway [Sun, 16 May 2004 23:18:55 +0000 (23:18 +0000)]
Change ln(), log(), power(), and sqrt() to emit the correct SQLSTATE
error codes for certain error conditions, as specified by SQL2003.

20 years agoRestore one mention of logrotate, per Peter.
Bruce Momjian [Sun, 16 May 2004 19:34:46 +0000 (19:34 +0000)]
Restore one mention of logrotate, per Peter.

20 years agoRemove an unused (and empty) header file.
Neil Conway [Fri, 14 May 2004 22:21:43 +0000 (22:21 +0000)]
Remove an unused (and empty) header file.

20 years agoImplement the width_bucket() function, per SQL2003. This commit only adds
Neil Conway [Fri, 14 May 2004 21:42:30 +0000 (21:42 +0000)]
Implement the width_bucket() function, per SQL2003. This commit only adds
a variant of the function for the 'numeric' datatype; it would be possible
to add additional variants for other datatypes, but I haven't done so yet.

This commit includes regression tests and minimal documentation; if we
want developers to actually use this function in applications, we'll
probably need to document what it does more fully.

20 years agoProperly document rotatelogs, and add mention of it to pg_ctl manual page.
Bruce Momjian [Fri, 14 May 2004 20:01:19 +0000 (20:01 +0000)]
Properly document rotatelogs, and add mention of it to pg_ctl manual page.

20 years agoImprove documentation for SQLSTATE error codes, per recent thread on
Neil Conway [Fri, 14 May 2004 18:04:02 +0000 (18:04 +0000)]
Improve documentation for SQLSTATE error codes, per recent thread on
-patches.

20 years agoAdjust find_my_exec/find_other_exec() so that the return parameter is
Bruce Momjian [Fri, 14 May 2004 17:04:48 +0000 (17:04 +0000)]
Adjust find_my_exec/find_other_exec() so that the return parameter is
last, not first.  This fits our style better.

20 years agoRemove extra 'then'.
Bruce Momjian [Fri, 14 May 2004 16:29:12 +0000 (16:29 +0000)]
Remove extra 'then'.

20 years agoTighten parsing of boolean options to CREATE TYPE and related functions,
Tom Lane [Fri, 14 May 2004 16:11:25 +0000 (16:11 +0000)]
Tighten parsing of boolean options to CREATE TYPE and related functions,
so as to deliver more useful error messages for mistakes like
'PASSEDBYVALUE = f'.  Per gripe from Gaetano Mendola.

20 years agoUpdate docs from Andrew J. Kopciuch <akopciuch@bddf.ca>
Teodor Sigaev [Fri, 14 May 2004 10:39:25 +0000 (10:39 +0000)]
Update docs from Andrew J. Kopciuch <akopciuch@bddf.ca>

20 years agoPropagate enlargeStringInfo() fixes into the equivalent code in
Tom Lane [Fri, 14 May 2004 00:20:38 +0000 (00:20 +0000)]
Propagate enlargeStringInfo() fixes into the equivalent code in
pqexpbuffer.c.  While a client-side failure doesn't seem like a
security issue, it's still a bug.

20 years agoProperly quote variable for makefile use.
Bruce Momjian [Fri, 14 May 2004 00:10:52 +0000 (00:10 +0000)]
Properly quote variable for makefile use.

20 years agoMove unixware thread stuff into Makefile.unixware.
Bruce Momjian [Fri, 14 May 2004 00:03:07 +0000 (00:03 +0000)]
Move unixware thread stuff into Makefile.unixware.

20 years agoForce thread flags for all Unixware builds if threading is requested.
Bruce Momjian [Thu, 13 May 2004 23:05:54 +0000 (23:05 +0000)]
Force thread flags for all Unixware builds if threading is requested.

This is required because once you link with a library that uses threads,
all references to that library have to use thread flags.

20 years agoRefactor code to warn about configure link failures on MinGW.
Bruce Momjian [Thu, 13 May 2004 22:59:14 +0000 (22:59 +0000)]
Refactor code to warn about configure link failures on MinGW.

20 years agoReorganize backend code to more cleanly manage executable names and
Bruce Momjian [Thu, 13 May 2004 22:45:04 +0000 (22:45 +0000)]
Reorganize backend code to more cleanly manage executable names and
backend startup.

20 years agoNo need for gcc thread stuff anymore, only cc.
Bruce Momjian [Thu, 13 May 2004 15:44:05 +0000 (15:44 +0000)]
No need for gcc thread stuff anymore, only cc.

20 years agoRemove warning about Unixware.
Bruce Momjian [Thu, 13 May 2004 14:07:19 +0000 (14:07 +0000)]
Remove warning about Unixware.

20 years agoFix Win32 bug with missing errno for strerror().
Bruce Momjian [Thu, 13 May 2004 01:47:12 +0000 (01:47 +0000)]
Fix Win32 bug with missing errno for strerror().

20 years agoWarn when MinGW fails to create symlinks during configure. Report
Bruce Momjian [Thu, 13 May 2004 01:45:02 +0000 (01:45 +0000)]
Warn when MinGW fails to create symlinks during configure.  Report
already made to MinGW maintainers.

Andrew Dunstan

20 years agoTighten up overflow check in path_recv, pursuant to code review inspired
Tom Lane [Wed, 12 May 2004 22:38:44 +0000 (22:38 +0000)]
Tighten up overflow check in path_recv, pursuant to code review inspired
by Ken Ashcraft's report.  I think there is no actual bug here since if
the int32 value does wrap a little bit, palloc will still reject it.
Still it's better that the code be obviously correct.

20 years agoHere's a patch that will allow external modules like pl<lang> to set
Bruce Momjian [Wed, 12 May 2004 21:56:27 +0000 (21:56 +0000)]
Here's a patch that will allow external modules like pl<lang> to set
additional flags for the dlltool and dllwrap commands. More info here:

http://archives.postgresql.org/pgsql-hackers/2004-05/msg00350.php

Thomas Hallgren

20 years agoRename find_my_binary/find_other_binary to
Bruce Momjian [Wed, 12 May 2004 13:38:49 +0000 (13:38 +0000)]
Rename find_my_binary/find_other_binary to
find_my_exec/find_other_exec().  Remove passing of progname to these
functions as they can find that out from argv[0], which they already
have.

Make get_progname return const char *, and update all progname variables
to be const char *.

20 years agoRemove:
Bruce Momjian [Wed, 12 May 2004 04:10:41 +0000 (04:10 +0000)]
Remove:

<  o Add PL/PHP (Joe)

20 years agoFix bug by passing arg[0] to find_my_binary().
Bruce Momjian [Wed, 12 May 2004 03:48:42 +0000 (03:48 +0000)]
Fix bug by passing arg[0] to find_my_binary().

20 years agoYawn ... still another place not quite ready for zero-column tables.
Tom Lane [Tue, 11 May 2004 22:43:55 +0000 (22:43 +0000)]
Yawn ... still another place not quite ready for zero-column tables.

20 years agoAs part of the work for making relocatable installs, I have re-factored
Bruce Momjian [Tue, 11 May 2004 21:57:15 +0000 (21:57 +0000)]
As part of the work for making relocatable installs, I have re-factored
all the code that looks for other binaries.  I move FindExec into
port/exec.c (and renamed it to find_my_binary()).  I also added
find_other_binary that looks for another binary in the same directory as
the calling program, and checks the version string.

The only behavior change was that initdb and pg_dump would look in the
hard-coded bindir directory if it can't find the requested binary in the
same directory as the caller.  The new code throws an error.  The old
behavior seemed too error prone for version mismatches.

20 years agoAdd DELETE:
Bruce Momjian [Tue, 11 May 2004 20:23:05 +0000 (20:23 +0000)]
Add DELETE:

> * Add MERGE command that does UPDATE/DELETE, or on failure, INSERT (rules, triggers?)

20 years agoAdd tests to enlargeStringInfo() to avoid possible buffer-overrun or
Tom Lane [Tue, 11 May 2004 20:07:26 +0000 (20:07 +0000)]
Add tests to enlargeStringInfo() to avoid possible buffer-overrun or
infinite-loop problems if a bogus data length is passed.

20 years agoRefactor low-level aclcheck code to provide useful interfaces for multi-bit
Tom Lane [Tue, 11 May 2004 17:36:13 +0000 (17:36 +0000)]
Refactor low-level aclcheck code to provide useful interfaces for multi-bit
permissions tests in about the same amount of code as before.  Exactly what
the GRANT/REVOKE code ought to be doing is still up for debate, but this
should be helpful in any case, and it already solves an efficiency problem
in executor startup.

20 years agoAdd mention of rules/triggers:
Bruce Momjian [Tue, 11 May 2004 13:44:05 +0000 (13:44 +0000)]
Add mention of rules/triggers:

> * Add MERGE command that does UPDATE, or on failure, INSERT (rules, triggers?)

20 years agoFix oversights in processing of LIMIT expressions during planning.
Tom Lane [Tue, 11 May 2004 13:15:15 +0000 (13:15 +0000)]
Fix oversights in processing of LIMIT expressions during planning.

20 years agoRepair recalculation failure for nested sub-SELECTs, per bug report from
Tom Lane [Tue, 11 May 2004 02:21:39 +0000 (02:21 +0000)]
Repair recalculation failure for nested sub-SELECTs, per bug report from
Didier Moens.  Bug is new in 7.4, and was caused by not updating everyplace
I should've when replacing locParam markers by allParam.
Add a regression test to catch related errors in future.

20 years agoClean up indenting of Makefile filenames.
Bruce Momjian [Mon, 10 May 2004 23:09:04 +0000 (23:09 +0000)]
Clean up indenting of Makefile filenames.

20 years agoPromote row expressions to full-fledged citizens of the expression syntax,
Tom Lane [Mon, 10 May 2004 22:44:49 +0000 (22:44 +0000)]
Promote row expressions to full-fledged citizens of the expression syntax,
rather than allowing them only in a few special cases as before.  In
particular you can now pass a ROW() construct to a function that accepts
a rowtype parameter.  Internal generation of RowExprs fixes a number of
corner cases that used to not work very well, such as referencing the
whole-row result of a JOIN or subquery.  This represents a further step in
the work I started a month or so back to make rowtype values into
first-class citizens.

20 years agoFix typo.
Neil Conway [Mon, 10 May 2004 21:08:28 +0000 (21:08 +0000)]
Fix typo.

20 years agoAdd variables names to static prototypes in initdb.c.
Bruce Momjian [Mon, 10 May 2004 20:51:58 +0000 (20:51 +0000)]
Add variables names to static prototypes in initdb.c.

20 years ago- Argh, just another bug in adjust_informix.
Michael Meskes [Mon, 10 May 2004 13:46:06 +0000 (13:46 +0000)]
- Argh, just another bug in adjust_informix.
- Added "extern C" flags for C++ compiler. Hopefully I got that one right.

20 years agoAdd:
Bruce Momjian [Mon, 10 May 2004 03:58:19 +0000 (03:58 +0000)]
Add:

> * Add MERGE command that does UPDATE, or on failure, INSERT

20 years agoFix a couple of oversights in new ALTER TABLE code that broke
Tom Lane [Sat, 8 May 2004 22:46:29 +0000 (22:46 +0000)]
Fix a couple of oversights in new ALTER TABLE code that broke
ALTER SET STATISTICS for functional indexes.

20 years agoAlter string format used for integer and OID lists in stored rules.
Tom Lane [Sat, 8 May 2004 21:21:18 +0000 (21:21 +0000)]
Alter string format used for integer and OID lists in stored rules.
This simplifies and speeds up the reader by letting it get the representation
right the first time, rather than correcting it after-the-fact.  Also,
after int and OID lists become separate node types per Neil's pending
patch, this will let us treat these lists as just plain Nodes instead
of requiring separate read/write macros the way we have now.

20 years agoGet rid of rd_nblocks field in relcache entries. Turns out this was
Tom Lane [Sat, 8 May 2004 19:09:25 +0000 (19:09 +0000)]
Get rid of rd_nblocks field in relcache entries.  Turns out this was
costing us lots more to maintain than it was worth.  On shared tables
it was of exactly zero benefit because we couldn't trust it to be
up to date.  On temp tables it sometimes saved an lseek, but not often
enough to be worth getting excited about.  And the real problem was that
we forced an lseek on every relcache flush in order to update the field.
So all in all it seems best to lose the complexity.

20 years agoFix typo.
Bruce Momjian [Sat, 8 May 2004 02:13:31 +0000 (02:13 +0000)]
Fix typo.

20 years agoImprove style of new log_statement_stats error message, per Peter.
Bruce Momjian [Sat, 8 May 2004 02:11:46 +0000 (02:11 +0000)]
Improve style of new log_statement_stats error message, per Peter.

20 years agoGet rid of cluster.c's apparatus for rebuilding a relation's indexes
Tom Lane [Sat, 8 May 2004 00:34:49 +0000 (00:34 +0000)]
Get rid of cluster.c's apparatus for rebuilding a relation's indexes
in favor of using the REINDEX TABLE apparatus, which does the same thing
simpler and faster.  Also, make TRUNCATE not use cluster.c at all, but
just assign a new relfilenode and REINDEX.  This partially addresses
Hartmut Raschick's complaint from last December that 7.4's TRUNCATE is
an order of magnitude slower than prior releases.  By getting rid of
a lot of unnecessary catalog updates, these changes buy back about a
factor of two (on my system).  The remaining overhead seems associated
with creating and deleting storage files, which we may not be able to
do much about without abandoning transaction safety for TRUNCATE.

20 years agoRefactor CheckDropPermissions() to move some initialization code for
Neil Conway [Fri, 7 May 2004 19:12:26 +0000 (19:12 +0000)]
Refactor CheckDropPermissions() to move some initialization code for
printing the proper error message out of the common path.

20 years agoThe timezone() family of functions (a/k/a AT TIME ZONE construct) can
Tom Lane [Fri, 7 May 2004 16:57:16 +0000 (16:57 +0000)]
The timezone() family of functions (a/k/a AT TIME ZONE construct) can
safely be marked immutable, since their results don't depend on the
TimeZone GUC variable.  Per recent discussion.

20 years ago- Added additional test case.
Michael Meskes [Fri, 7 May 2004 13:42:49 +0000 (13:42 +0000)]
- Added additional test case.
- Fixed bug that reversed string length in typedefs.
- Added portability file to pgtypeslib.

20 years agoSuppress 'uninitialized variable' warning emitted by some (not all)
Tom Lane [Fri, 7 May 2004 13:09:12 +0000 (13:09 +0000)]
Suppress 'uninitialized variable' warning emitted by some (not all)
versions of gcc.  The code is correct AFAICS, but it requires slightly
more analysis than usual to see that the variable can't be used uninitialized.

20 years agoUse regprocedure type instead of oid. Usefull for human read and dump/restore
Teodor Sigaev [Fri, 7 May 2004 11:19:06 +0000 (11:19 +0000)]
Use regprocedure type instead of oid. Usefull for human read and dump/restore

20 years agoNATURAL CROSS JOIN is a contradiction in terms, not to mention disallowed
Tom Lane [Fri, 7 May 2004 03:19:44 +0000 (03:19 +0000)]
NATURAL CROSS JOIN is a contradiction in terms, not to mention disallowed
by the SQL spec and by our parser.  Thanks to Jonathan Scott for finding
this longstanding error.

20 years agoUppercase keyword for ecpg.
Bruce Momjian [Fri, 7 May 2004 02:40:43 +0000 (02:40 +0000)]
Uppercase keyword for ecpg.

20 years agoRemove crude test for log_statement_stats in startup code now that we
Bruce Momjian [Fri, 7 May 2004 01:53:41 +0000 (01:53 +0000)]
Remove crude test for log_statement_stats in startup code now that we
have a more proper GUC based test.

Also change error return code to ERRCODE_INVALID_PARAMETER_VALUE so it
matches the old error return code.

20 years agoThrow error if log_statement_stats is used with confliction options for
Bruce Momjian [Fri, 7 May 2004 01:34:08 +0000 (01:34 +0000)]
Throw error if log_statement_stats is used with confliction options for
per-query stage stats.

20 years agoSolve the 'Turkish problem' with undesirable locale behavior for case
Tom Lane [Fri, 7 May 2004 00:24:59 +0000 (00:24 +0000)]
Solve the 'Turkish problem' with undesirable locale behavior for case
conversion of basic ASCII letters.  Remove all uses of strcasecmp and
strncasecmp in favor of new functions pg_strcasecmp and pg_strncasecmp;
remove most but not all direct uses of toupper and tolower in favor of
pg_toupper and pg_tolower.  These functions use the same notions of
case folding already developed for identifier case conversion.  I left
the straight locale-based folding in place for situations where we are
just manipulating user data and not trying to match it to built-in
strings --- for example, the SQL upper() function is still locale
dependent.  Perhaps this will prove not to be what's wanted, but at
the moment we can initdb and pass regression tests in Turkish locale.

20 years agoDone:
Bruce Momjian [Thu, 6 May 2004 19:32:09 +0000 (19:32 +0000)]
Done:

>  o -ALTER TABLE ADD COLUMN column SERIAL doesn't create sequence because

20 years agosysv_shmem.c patch is to correct a bug that prevents the postmaster
Bruce Momjian [Thu, 6 May 2004 19:23:25 +0000 (19:23 +0000)]
sysv_shmem.c patch is to correct a bug that prevents the postmaster
recovering from an unexpected backend termination.

The remaining patches are to allow whitespace in db/install directory
names.

Claudio Natoli

20 years agoErase MD5 user passwords when a user is renamed because the username is
Bruce Momjian [Thu, 6 May 2004 16:59:16 +0000 (16:59 +0000)]
Erase MD5 user passwords when a user is renamed because the username is
used as salt for the MD5 password.

20 years agoUpdate todo/faq timestamps
Bruce Momjian [Thu, 6 May 2004 16:49:13 +0000 (16:49 +0000)]
Update todo/faq timestamps

20 years agoMake ALTER COLUMN TYPE preserve clustered status for indexes it doesn't
Tom Lane [Thu, 6 May 2004 16:10:57 +0000 (16:10 +0000)]
Make ALTER COLUMN TYPE preserve clustered status for indexes it doesn't
modify.  Also fix a passel of problems with ALTER TABLE CLUSTER ON:
failure to check that the index is safe to cluster on (or even belongs
to the indicated rel, or even exists), and failure to broadcast a relcache
flush event when changing an index's state.

20 years agoRewrite nodeRead() in a less obfuscated fashion, per discussion with
Tom Lane [Thu, 6 May 2004 14:01:33 +0000 (14:01 +0000)]
Rewrite nodeRead() in a less obfuscated fashion, per discussion with
Neil Conway.

20 years agoFix grammatical error in comment.
Neil Conway [Thu, 6 May 2004 06:11:01 +0000 (06:11 +0000)]
Fix grammatical error in comment.

20 years agoSecond try at a portable unsetenv().
Tom Lane [Wed, 5 May 2004 21:18:29 +0000 (21:18 +0000)]
Second try at a portable unsetenv().

20 years agoDon't assume that struct timeval's tv_sec field is the same datatype as
Tom Lane [Wed, 5 May 2004 17:28:46 +0000 (17:28 +0000)]
Don't assume that struct timeval's tv_sec field is the same datatype as
time_t; on some platforms they are not the same width.  Per Manfred Koizar.

20 years agoFix get_rels_with_domain to not do the wrong thing with views and
Tom Lane [Wed, 5 May 2004 17:06:56 +0000 (17:06 +0000)]
Fix get_rels_with_domain to not do the wrong thing with views and
composite types that use a domain.

20 years agoUse a more portable technique for unsetting environment variables,
Tom Lane [Wed, 5 May 2004 16:09:31 +0000 (16:09 +0000)]
Use a more portable technique for unsetting environment variables,
and unset PGCLIENTENCODING to prevent backend from dying if it's set
to something incompatible with the -E option.

20 years ago- Fixed bug in adjust_informix that treated arrays as simple variables.
Michael Meskes [Wed, 5 May 2004 15:03:04 +0000 (15:03 +0000)]
- Fixed bug in adjust_informix that treated arrays as simple variables.
- Synced parser again.
- Synced lexer.

20 years agoDone:
Bruce Momjian [Wed, 5 May 2004 14:36:20 +0000 (14:36 +0000)]
Done:

        o -ALTER TABLE ADD COLUMN does not honor DEFAULT and non-CHECK CONSTRAINT
        o -ALTER TABLE ADD COLUMN column DEFAULT should fill existing
          rows with DEFAULT value
        o -Allow ALTER TABLE to modify column lengths and change to binary
          compatible types

Remove:

        o Allow columns to be reordered using ALTER ... POSITION i col1 [,col2];
          have SELECT * and INSERT honor such ordering

20 years agoFix a typo in the documentation.
Neil Conway [Wed, 5 May 2004 09:33:38 +0000 (09:33 +0000)]
Fix a typo in the documentation.

20 years agoALTER TABLE rewrite. New cool stuff:
Tom Lane [Wed, 5 May 2004 04:48:48 +0000 (04:48 +0000)]
ALTER TABLE rewrite.  New cool stuff:

* ALTER ... ADD COLUMN with defaults and NOT NULL constraints works per SQL
spec.  A default is implemented by rewriting the table with the new value
stored in each row.

* ALTER COLUMN TYPE.  You can change a column's datatype to anything you
want, so long as you can specify how to convert the old value.  Rewrites
the table.  (Possible future improvement: optimize no-op conversions such
as varchar(N) to varchar(N+1).)

* Multiple ALTER actions in a single ALTER TABLE command.  You can perform
any number of column additions, type changes, and constraint additions with
only one pass over the table contents.

Basic documentation provided in ALTER TABLE ref page, but some more docs
work is needed.

Original patch from Rod Taylor, additional work from Tom Lane.

20 years agoMore PWD adjustments for win32.
Bruce Momjian [Mon, 3 May 2004 13:25:23 +0000 (13:25 +0000)]
More PWD adjustments for win32.

20 years agoFix obscure typo in 7.4.0 release notes.
Neil Conway [Mon, 3 May 2004 08:47:54 +0000 (08:47 +0000)]
Fix obscure typo in 7.4.0 release notes.