]> granicus.if.org Git - postgresql/log
postgresql
23 years agoChange SearchSysCache coding conventions so that a reference count is
Tom Lane [Thu, 16 Nov 2000 22:30:52 +0000 (22:30 +0000)]
Change SearchSysCache coding conventions so that a reference count is
maintained for each cache entry.  A cache entry will not be freed until
the matching ReleaseSysCache call has been executed.  This eliminates
worries about cache entries getting dropped while still in use.  See
my posting to pg-hackers of even date for more info.

23 years agoPrint bit values as binary strings (not hex)
Peter Eisentraut [Thu, 16 Nov 2000 21:43:28 +0000 (21:43 +0000)]
Print bit values as binary strings (not hex)

23 years agoAdd support for casting bit string constants.
Peter Eisentraut [Thu, 16 Nov 2000 17:27:10 +0000 (17:27 +0000)]
Add support for casting bit string constants.

23 years agoClean files after installing, since those files will most likely be root
Peter Eisentraut [Thu, 16 Nov 2000 16:21:55 +0000 (16:21 +0000)]
Clean files after installing, since those files will most likely be root
owned, which can be annoying.

23 years agoThat variable I removed broke XLOG, that part of the delta
Bruce Momjian [Thu, 16 Nov 2000 06:16:00 +0000 (06:16 +0000)]
That variable I removed broke XLOG, that part of the delta
should have read:

Alfred Perlstein

23 years agoI think in datatype.sgml line 506 or so:
Bruce Momjian [Thu, 16 Nov 2000 06:03:40 +0000 (06:03 +0000)]
I think in datatype.sgml line 506 or so:
    <entry>varchar(n)</entry>
        <entry>(4+x) bytes</entry>

should be
   <entry>varchar(n)</entry>
        <entry>(4+n) bytes</entry>

or
   <entry>varchar(x)</entry>
        <entry>(4+x) bytes</entry>

Regards

Laser Henry

23 years agoMake pgsql compile on FreeBSD-alpha.
Bruce Momjian [Thu, 16 Nov 2000 05:51:07 +0000 (05:51 +0000)]
Make pgsql compile on FreeBSD-alpha.

Context diff this time.

Remove -m486 compile args for FreeBSD-i386, compile -O2 on i386.

Compile with only -O on alpha for codegen safety.

Make the port use the TEST_AND_SET for alpha and i386 on FreeBSD.

Fix a lot of bogus string formats for outputting pointers (cast to int
and %u/%x replaced with no cast and %p), and 'Size'(size_t) are now
cast to 'unsigned long' and output with %lu/

Remove an unused variable.

Alfred Perlstein

23 years agoSyslog Facility Patch
Bruce Momjian [Thu, 16 Nov 2000 05:50:01 +0000 (05:50 +0000)]
Syslog Facility Patch

Here is one with a stray character removed.

Larry Rosenman

23 years agoRepair markup
Peter Eisentraut [Wed, 15 Nov 2000 19:43:39 +0000 (19:43 +0000)]
Repair markup

23 years agoRename parameter "hostname" to "virtual_host". Seemed very ambiguous...
Peter Eisentraut [Wed, 15 Nov 2000 18:36:06 +0000 (18:36 +0000)]
Rename parameter "hostname" to "virtual_host".  Seemed very ambiguous...

23 years agoClean up array-dimensions parser a bit.
Tom Lane [Tue, 14 Nov 2000 23:28:13 +0000 (23:28 +0000)]
Clean up array-dimensions parser a bit.
This code still needs a lot of love, however ...

23 years agoFix return type
Peter Eisentraut [Tue, 14 Nov 2000 21:26:21 +0000 (21:26 +0000)]
Fix return type

23 years agospaces to tab
Peter Eisentraut [Tue, 14 Nov 2000 21:23:47 +0000 (21:23 +0000)]
spaces to tab

23 years agoGive GCC on Irix a (theoretical) chance.
Peter Eisentraut [Tue, 14 Nov 2000 21:12:00 +0000 (21:12 +0000)]
Give GCC on Irix a (theoretical) chance.

23 years agoMinor cleanup of tableOid-related coding.
Tom Lane [Tue, 14 Nov 2000 21:04:32 +0000 (21:04 +0000)]
Minor cleanup of tableOid-related coding.

23 years agoClean up comments.
Tom Lane [Tue, 14 Nov 2000 20:47:34 +0000 (20:47 +0000)]
Clean up comments.

23 years agoSmall polishing of syslog facility and ident settings. Only allow setting
Peter Eisentraut [Tue, 14 Nov 2000 19:13:27 +0000 (19:13 +0000)]
Small polishing of syslog facility and ident settings.  Only allow setting
at postmaster start, rename syslog_progid to syslog_ident, since syslog
itself uses that term, fix doc markup.

23 years agoExtend CREATE DATABASE to allow selection of a template database to be
Tom Lane [Tue, 14 Nov 2000 18:37:49 +0000 (18:37 +0000)]
Extend CREATE DATABASE to allow selection of a template database to be
cloned, rather than always cloning template1.  Modify initdb to generate
two identical databases rather than one, template0 and template1.
Connections to template0 are disallowed, so that it will always remain
in its virgin as-initdb'd state.  pg_dumpall now dumps databases with
restore commands that say CREATE DATABASE foo WITH TEMPLATE = template0.
This allows proper behavior when there is user-added data in template1.
initdb forced!

23 years agoUpdate and polish postmaster and postgres help output and man pages.
Peter Eisentraut [Tue, 14 Nov 2000 18:11:32 +0000 (18:11 +0000)]
Update and polish postmaster and postgres help output and man pages.

23 years agoFix illegal combination of SearchSysCacheTuple() and heap_freetuple()
Hiroshi Inoue [Tue, 14 Nov 2000 01:57:30 +0000 (01:57 +0000)]
Fix illegal combination of SearchSysCacheTuple() and heap_freetuple()
in AlterTableOwner().

23 years agoAdd mention of new leading-slash handling of hostname to libpq docs.
Bruce Momjian [Tue, 14 Nov 2000 01:53:37 +0000 (01:53 +0000)]
Add mention of new leading-slash handling of hostname to libpq docs.

23 years agoRename PortName to PortNumber.
Bruce Momjian [Tue, 14 Nov 2000 01:15:06 +0000 (01:15 +0000)]
Rename PortName to PortNumber.

23 years agoAdd mention of hostname leading slash handling to all manual pages that use -h.
Bruce Momjian [Mon, 13 Nov 2000 23:57:20 +0000 (23:57 +0000)]
Add mention of hostname leading slash handling to all manual pages that use -h.

23 years agoRemove -k unix socketpath option from client side, allow hostname with
Bruce Momjian [Mon, 13 Nov 2000 23:37:54 +0000 (23:37 +0000)]
Remove -k unix socketpath option from client side, allow hostname with
leading slash to behave as a unix socket path.

23 years agoOk, You guys are probably tired of me, BUT, here is another one, that
Bruce Momjian [Mon, 13 Nov 2000 21:35:03 +0000 (21:35 +0000)]
Ok, You guys are probably tired of me, BUT, here is another one, that
adds the facility to set the program name used in syslog.
(this includes the other ones).

One gotcha, the parser doesn't like special characters in strings.
For example, i tried to use pg-test, and if failed the parse coming
from the postgresql.conf file.

I don't think it's a showstopper..

Larry Rosenman

23 years agoFix for no hterror().
Bruce Momjian [Mon, 13 Nov 2000 21:31:46 +0000 (21:31 +0000)]
Fix for no hterror().

23 years agoUUNET is looking into offering PostgreSQL as a part of a managed web
Bruce Momjian [Mon, 13 Nov 2000 15:18:15 +0000 (15:18 +0000)]
UUNET is looking into offering PostgreSQL as a part of a managed web
hosting product, on both shared and dedicated machines.  We currently
offer Oracle and MySQL, and it would be a nice middle-ground.
However, as shipped, PostgreSQL lacks the following features we need
that MySQL has:

1. The ability to listen only on a particular IP address.  Each
   hosting customer has their own IP address, on which all of their
   servers (http, ftp, real media, etc.) run.
2. The ability to place the Unix-domain socket in a mode 700 directory.
   This allows us to automatically create an empty database, with an
   empty DBA password, for new or upgrading customers without having
   to interactively set a DBA password and communicate it to (or from)
   the customer.  This in turn cuts down our install and upgrade times.
3. The ability to connect to the Unix-domain socket from within a
   change-rooted environment.  We run CGI programs chrooted to the
   user's home directory, which is another reason why we need to be
   able to specify where the Unix-domain socket is, instead of /tmp.
4. The ability to, if run as root, open a pid file in /var/run as
   root, and then setuid to the desired user.  (mysqld -u can almost
   do this; I had to patch it, too).

The patch below fixes problem 1-3.  I plan to address #4, also, but
haven't done so yet.  These diffs are big enough that they should give
the PG development team something to think about in the meantime :-)
Also, I'm about to leave for 2 weeks' vacation, so I thought I'd get
out what I have, which works (for the problems it tackles), now.

With these changes, we can set up and run PostgreSQL with scripts the
same way we can with apache or proftpd or mysql.

In summary, this patch makes the following enhancements:

1. Adds an environment variable PGUNIXSOCKET, analogous to MYSQL_UNIX_PORT,
   and command line options -k --unix-socket to the relevant programs.
2. Adds a -h option to postmaster to set the hostname or IP address to
   listen on instead of the default INADDR_ANY.
3. Extends some library interfaces to support the above.
4. Fixes a few memory leaks in PQconnectdb().

The default behavior is unchanged from stock 7.0.2; if you don't use
any of these new features, they don't change the operation.

David J. MacKenzie

23 years agoSearchSysCacheTupleCopy() instead of SearchSysCache()
Hiroshi Inoue [Mon, 13 Nov 2000 09:16:55 +0000 (09:16 +0000)]
SearchSysCacheTupleCopy() instead of SearchSysCache()

23 years agoprevent open failure of VIEW
Hiroshi Inoue [Mon, 13 Nov 2000 09:06:36 +0000 (09:06 +0000)]
prevent open failure of VIEW

23 years agoKeep track of the last active slot in the shared ProcState array, so
Tom Lane [Sun, 12 Nov 2000 20:51:52 +0000 (20:51 +0000)]
Keep track of the last active slot in the shared ProcState array, so
that search loops only have to scan that far and not through all maxBackends
entries.  This eliminates a performance penalty for setting maxBackends
much higher than the average number of active backends.  Also, eliminate
no-longer-used 'backend tag' concept.  Remove setting of environment
variables at backend start (except for CYR_RECODE), since none of them
are being examined by the backend any longer.

23 years agoAdd new 7.0.3 item to all places.
Bruce Momjian [Sun, 12 Nov 2000 14:26:48 +0000 (14:26 +0000)]
Add new 7.0.3 item to all places.

23 years agoRestructure handling of inheritance queries so that they work with outer
Tom Lane [Sun, 12 Nov 2000 00:37:02 +0000 (00:37 +0000)]
Restructure handling of inheritance queries so that they work with outer
joins, and clean things up a good deal at the same time.  Append plan node
no longer hacks on rangetable at runtime --- instead, all child tables are
given their own RT entries during planning.  Concept of multiple target
tables pushed up into execMain, replacing bug-prone implementation within
nodeAppend.  Planner now supports generating Append plans for inheritance
sets either at the top of the plan (the old way) or at the bottom.  Expanding
at the bottom is appropriate for tables used as sources, since they may
appear inside an outer join; but we must still expand at the top when the
target of an UPDATE or DELETE is an inheritance set, because we actually need
a different targetlist and junkfilter for each target table in that case.
Fortunately a target table can't be inside an outer join...  Bizarre mutual
recursion between union_planner and prepunion.c is gone --- in fact,
union_planner doesn't really have much to do with union queries anymore,
so I renamed it grouping_planner.

23 years agoRepair markup
Peter Eisentraut [Sat, 11 Nov 2000 23:27:48 +0000 (23:27 +0000)]
Repair markup

23 years agoMarkup enhancements, some factual corrections.
Peter Eisentraut [Sat, 11 Nov 2000 23:01:45 +0000 (23:01 +0000)]
Markup enhancements, some factual corrections.

23 years agoMake sure shell scripts send error messages to stderr.
Peter Eisentraut [Sat, 11 Nov 2000 22:59:48 +0000 (22:59 +0000)]
Make sure shell scripts send error messages to stderr.
Improve help output.
Teach droplang to remove 'pltclu'.

23 years agoAllow more timezone-like interpretation of INTERVALs.
Thomas G. Lockhart [Sat, 11 Nov 2000 19:57:03 +0000 (19:57 +0000)]
Allow more timezone-like interpretation of INTERVALs.
Fix up labeling of some new test cases.

23 years agoAllow more timezone-like interpretation of INTERVALs.
Thomas G. Lockhart [Sat, 11 Nov 2000 19:56:43 +0000 (19:56 +0000)]
Allow more timezone-like interpretation of INTERVALs.

23 years agoCheck for existence of the results directory before invoking mkdir.
Thomas G. Lockhart [Sat, 11 Nov 2000 19:56:20 +0000 (19:56 +0000)]
Check for existence of the results directory before invoking mkdir.

23 years agoDefine conversions to and from text for date, time, and timetz.
Thomas G. Lockhart [Sat, 11 Nov 2000 19:55:39 +0000 (19:55 +0000)]
Define conversions to and from text for date, time, and timetz.

23 years agoAllow interpretation of INTERVALs with more timezone-like syntax.
Thomas G. Lockhart [Sat, 11 Nov 2000 19:55:19 +0000 (19:55 +0000)]
Allow interpretation of INTERVALs with more timezone-like syntax.
Define conversions to and from text for date, time, and timetz.
Have millisecond and microsecond return full # of seconds in those units.
 Previously, only returned full fractional part in those units.

23 years agoMention new AT TIME ZONE feature.
Thomas G. Lockhart [Sat, 11 Nov 2000 19:50:31 +0000 (19:50 +0000)]
Mention new AT TIME ZONE feature.

23 years agoFix bug in recent improvement to type resolution code. Forgot to retain
Thomas G. Lockhart [Sat, 11 Nov 2000 19:49:26 +0000 (19:49 +0000)]
Fix bug in recent improvement to type resolution code. Forgot to retain
 "best choice" type category when resolving UNKNOWN function and operator
 arguments. Thanks to Tom Lane for finding test case.

23 years agoEliminate word rot
Peter Eisentraut [Sat, 11 Nov 2000 13:53:49 +0000 (13:53 +0000)]
Eliminate word rot

23 years agoUpdate expected output to track recent code changes.
Tom Lane [Fri, 10 Nov 2000 23:19:46 +0000 (23:19 +0000)]
Update expected output to track recent code changes.

23 years agoUpdate to PyGreSQL 3.1:
Bruce Momjian [Fri, 10 Nov 2000 22:29:21 +0000 (22:29 +0000)]
Update to PyGreSQL 3.1:

Fix some quoting functions. In particular handle NULLs better.

Use a method to add primary key information rather than direct
manipulation of the class structures.

Break decimal out in _quote (in pg.py) and treat it as float.

Treat timestamp like date for quoting purposes.

Remove a redundant SELECT from the get method speeding it, and
insert since it calls get, up a little.

Add test for BOOL type in typecast method to pgdbTypeCache class.
(tv@beamnet.de)

Fix pgdb.py to send port as integer to lower level function
(dildog@l0pht.com)

Change pg.py to speed up some operations

Allow updates on tables with no primary keys.

D'Arcy J.M. Cain

23 years agothe bug was not fixed in the snapshot of November 5th. Also the enterprise
Bruce Momjian [Fri, 10 Nov 2000 22:06:26 +0000 (22:06 +0000)]
the bug was not fixed in the snapshot of November 5th. Also the enterprise
edition of the driver did not compile. I have fixed both issues again. I have
attached the modified files to this email, maybe you can check them into the
repository. (Fixes are marked with //FIXME). Enterprise edition driver now
compiles and seems to work.

Jan Thomae

23 years agoAdjust INET/CIDR display conventions and reimplement some INET/CIDR
Tom Lane [Fri, 10 Nov 2000 20:13:27 +0000 (20:13 +0000)]
Adjust INET/CIDR display conventions and reimplement some INET/CIDR
functions, per recent discussions on pghackers.  For now, I have called
the verbose-display formatting function text(), but will reconsider if
enough people object.
initdb forced.

23 years agoUpdate TODO list.
Bruce Momjian [Fri, 10 Nov 2000 19:42:35 +0000 (19:42 +0000)]
Update TODO list.

23 years agoAdd update of HISTORY file for 7.0.3.
Bruce Momjian [Fri, 10 Nov 2000 19:41:26 +0000 (19:41 +0000)]
Add update of HISTORY file for 7.0.3.

23 years agoFix typos.
Tom Lane [Fri, 10 Nov 2000 16:32:09 +0000 (16:32 +0000)]
Fix typos.

23 years agoUse SearchSysCacheTupleCopy() instead of SearchSysCacheTuple() in
Hiroshi Inoue [Fri, 10 Nov 2000 09:38:21 +0000 (09:38 +0000)]
Use SearchSysCacheTupleCopy() instead of SearchSysCacheTuple() in
order to continue to access the tuple more than now.
This would resolve a segmentation fault error.

23 years agopg_fsync is fsync in WAL version.
Vadim B. Mikheev [Fri, 10 Nov 2000 03:53:45 +0000 (03:53 +0000)]
pg_fsync is fsync in WAL version.

23 years agoClean up syscache so that recursive invocation is safe, and remove error
Tom Lane [Fri, 10 Nov 2000 00:33:12 +0000 (00:33 +0000)]
Clean up syscache so that recursive invocation is safe, and remove error
message about recursive use of a syscache.  Also remove most of the
specialized indexscan routines in indexing.c --- it turns out that
catcache.c is perfectly able to perform the indexscan for itself,
in fact has already looked up all the information needed to do so!
This should be faster as well as needing far less boilerplate code.

23 years agoReal dynamic loader code
Peter Eisentraut [Thu, 9 Nov 2000 19:00:50 +0000 (19:00 +0000)]
Real dynamic loader code

23 years agoAlso test for 'void *' as third arg of accept() (as on Solaris 7 and 8),
Peter Eisentraut [Thu, 9 Nov 2000 18:18:42 +0000 (18:18 +0000)]
Also test for 'void *' as third arg of accept() (as on Solaris 7 and 8),
but take it as 'int *' instead.

Add real test for whether ld -R works on Unixware.

Rename --enable-uniconv to --enable-unicode-conversion.

Install shlibs mode 755 by default, since 644 causes gratuitous complaints
from ldd et al. on some systems.

23 years agoExecEndAppend() neglected to close indices on appended result rels,
Tom Lane [Thu, 9 Nov 2000 18:12:53 +0000 (18:12 +0000)]
ExecEndAppend() neglected to close indices on appended result rels,
and improperly prevented the main result rel from being closed if it
wasn't one of the Append's own result rels.  Per report from Hiroshi.

23 years agoApplied yet another patch by Christof. Thanks Cristof!
Michael Meskes [Thu, 9 Nov 2000 14:06:57 +0000 (14:06 +0000)]
Applied yet another patch by Christof. Thanks Cristof!
Synced parser.

23 years agoAuto checkpoint creation.
Vadim B. Mikheev [Thu, 9 Nov 2000 11:26:00 +0000 (11:26 +0000)]
Auto checkpoint creation.

23 years agoPlease apply this patch to current, to fix a problem with runcheck,
Bruce Momjian [Thu, 9 Nov 2000 04:17:53 +0000 (04:17 +0000)]
Please apply this patch to current, to fix a problem with runcheck,
that installs into a different path than is configured.

With this applied both postmaster and the shared libs are location
independent
for AIX 4.2 and up.

Thanks
Andreas

23 years agoArrange for CASE or UNION with only untyped literal constants as input
Tom Lane [Thu, 9 Nov 2000 04:14:32 +0000 (04:14 +0000)]
Arrange for CASE or UNION with only untyped literal constants as input
to resolve the unknown constants as type TEXT.

23 years agoAdd some more union/intersect/except test cases, per suggestions
Tom Lane [Thu, 9 Nov 2000 02:47:49 +0000 (02:47 +0000)]
Add some more union/intersect/except test cases, per suggestions
from Kevin O'Gorman.

23 years agoRepair some bugs in new union/intersect/except code.
Tom Lane [Thu, 9 Nov 2000 02:46:17 +0000 (02:46 +0000)]
Repair some bugs in new union/intersect/except code.
Thanks to Kevin O'Gorman for finding these...

23 years agoClean up compiler warnings.
Tom Lane [Wed, 8 Nov 2000 23:24:24 +0000 (23:24 +0000)]
Clean up compiler warnings.

23 years agoMake DROP TABLE rollback-able: postpone physical file delete until commit.
Tom Lane [Wed, 8 Nov 2000 22:10:03 +0000 (22:10 +0000)]
Make DROP TABLE rollback-able: postpone physical file delete until commit.
(WAL logging for this is not done yet, however.)  Clean up a number of really
crufty things that are no longer needed now that DROP behaves nicely.  Make
temp table mapper do the right things when drop or rename affecting a temp
table is rolled back.  Also, remove "relation modified while in use" error
check, in favor of locking tables at first reference and holding that lock
throughout the statement.

23 years agoAdd ANALYSE spelling of ANALYZE for vacuum.
Bruce Momjian [Wed, 8 Nov 2000 21:28:06 +0000 (21:28 +0000)]
Add ANALYSE spelling of ANALYZE for vacuum.

23 years agoOnly use ld -R flag in UW 7.1*, since it's new.
Peter Eisentraut [Wed, 8 Nov 2000 20:27:38 +0000 (20:27 +0000)]
Only use ld -R flag in UW 7.1*, since it's new.

23 years agoFix 'soname's for various platforms, so that libpq.so.2.0 is considered
Peter Eisentraut [Wed, 8 Nov 2000 20:18:49 +0000 (20:18 +0000)]
Fix 'soname's  for various platforms, so that libpq.so.2.0 is considered
compatible with libpq.so.2.1, etc.

23 years agoUpdate TODO list.
Bruce Momjian [Wed, 8 Nov 2000 20:04:23 +0000 (20:04 +0000)]
Update TODO list.

23 years agoAdd dependencies required for parallel make
Peter Eisentraut [Wed, 8 Nov 2000 18:56:25 +0000 (18:56 +0000)]
Add dependencies required for parallel make

23 years agoDump identity of alternative location, if any.
Peter Eisentraut [Wed, 8 Nov 2000 18:23:44 +0000 (18:23 +0000)]
Dump identity of alternative location, if any.

23 years agoAdd '-c name=val' flag for setting run-time parameters.
Peter Eisentraut [Wed, 8 Nov 2000 17:57:46 +0000 (17:57 +0000)]
Add '-c name=val' flag for setting run-time parameters.

23 years agoReimplement alternative database locations with symlinks. No changes in
Peter Eisentraut [Wed, 8 Nov 2000 16:59:50 +0000 (16:59 +0000)]
Reimplement alternative database locations with symlinks.  No changes in
user interface.

23 years agoCLUSTER did no permissions checking, forsooth ...
Tom Lane [Wed, 8 Nov 2000 16:31:06 +0000 (16:31 +0000)]
CLUSTER did no permissions checking, forsooth ...

23 years agoEnable fallback to string type when argument(s) are of UNKNOWN type.
Thomas G. Lockhart [Tue, 7 Nov 2000 16:01:01 +0000 (16:01 +0000)]
Enable fallback to string type when argument(s) are of UNKNOWN type.
 Same code exactly as for function resolution.
 An obvious example is for
  select '1' = '01';
 which used to throw an error and which now resolves to two text strings.

23 years agoThis somehow gets moot. Is there a way to make gcc reject those comments?
Bruce Momjian [Tue, 7 Nov 2000 11:35:16 +0000 (11:35 +0000)]
This somehow gets moot. Is there a way to make gcc reject those comments?

Andreas

23 years agoUndo incorrectly committed Makefile.
Michael Meskes [Tue, 7 Nov 2000 08:52:22 +0000 (08:52 +0000)]
Undo incorrectly committed Makefile.

23 years agoThird try. Sorry, I had a wrong path in my copy statement.
Michael Meskes [Tue, 7 Nov 2000 08:46:27 +0000 (08:46 +0000)]
Third try. Sorry, I had a wrong path in my copy statement.

23 years agoAdded more patches by Cristif Petig.
Michael Meskes [Tue, 7 Nov 2000 08:41:15 +0000 (08:41 +0000)]
Added more patches by Cristif Petig.

23 years agoUpdate for new results.
Tom Lane [Tue, 7 Nov 2000 06:51:21 +0000 (06:51 +0000)]
Update for new results.

23 years agoDropErrorMsg() died with assert failure if try to drop a toast table...
Tom Lane [Tue, 7 Nov 2000 02:17:50 +0000 (02:17 +0000)]
DropErrorMsg() died with assert failure if try to drop a toast table...
mea culpa...

23 years agoUpdate geometry for FreeBSD 4.1 (same as 4.0)
Peter Eisentraut [Mon, 6 Nov 2000 22:40:30 +0000 (22:40 +0000)]
Update geometry for FreeBSD 4.1 (same as 4.0)

23 years agoAdd -V option to backend, to show version, since --version doesn't seem
Peter Eisentraut [Mon, 6 Nov 2000 22:18:10 +0000 (22:18 +0000)]
Add -V option to backend, to show version, since --version doesn't seem
to work everywhere.  Also, on FreeBSD you need to set the optreset variable
to 1 before parsing the command line a second time with getopt().

23 years agoFix GNU make detection.
Peter Eisentraut [Mon, 6 Nov 2000 18:18:22 +0000 (18:18 +0000)]
Fix GNU make detection.

(The test 'make --version' with FreeBSD make (and potentially others)
would just reinvoke make on the same Makefile, resulting in an infinite
loop.)

23 years agoAdd explicit ORDER BYs in a couple of cases --- seems to be necessary
Tom Lane [Mon, 6 Nov 2000 18:11:46 +0000 (18:11 +0000)]
Add explicit ORDER BYs in a couple of cases --- seems to be necessary
to get platform-independent results.

23 years agoI think Thomas missed an update here...
Tom Lane [Mon, 6 Nov 2000 18:11:03 +0000 (18:11 +0000)]
I think Thomas missed an update here...

23 years agoBack out aix makefile patch.
Bruce Momjian [Mon, 6 Nov 2000 18:06:58 +0000 (18:06 +0000)]
Back out aix makefile patch.

23 years agoMakefile.ports (copied from makefiles/Makefile.aix) is included before
Bruce Momjian [Mon, 6 Nov 2000 17:07:43 +0000 (17:07 +0000)]
Makefile.ports (copied from makefiles/Makefile.aix) is included before
the default target in GNUmakefile is reached.  This patch calls that
default target.

Pete Forman

23 years agoAdd support for AT TIME ZONE syntax.
Thomas G. Lockhart [Mon, 6 Nov 2000 16:05:25 +0000 (16:05 +0000)]
Add support for AT TIME ZONE syntax.

23 years agoAdd tests for more INTERVAL syntax.
Thomas G. Lockhart [Mon, 6 Nov 2000 16:03:47 +0000 (16:03 +0000)]
Add tests for more INTERVAL syntax.
Add more tests for JOIN syntax.
All tests pass on my Linux box (except for the usual couple of lines
 for geometry).

23 years agoAdd functions to support AT TIME ZONE syntax.
Thomas G. Lockhart [Mon, 6 Nov 2000 15:58:46 +0000 (15:58 +0000)]
Add functions to support AT TIME ZONE syntax.
Update catalog version to today.

23 years agoFix INTERVAL output when year/month has different sign as day/hour etc.
Thomas G. Lockhart [Mon, 6 Nov 2000 15:57:00 +0000 (15:57 +0000)]
Fix INTERVAL output when year/month has different sign as day/hour etc.
 Previously, all fields were unsigned, with only a trailing "ago" to
 indicate negative intervals. Now, ISO format does not use "ago", and
 and the traditional PostgreSQL format has the first numeric field unsigned
 with "ago" supporting that field. So "1 month - 2 days ago" is two days
 less than a month in the past.
Fix interval arithmetic across daylight savings time boundaries.
 Previously, most math across boundaries introduced a one hour offset.
Allow some date/time functions to return NULL if called with NULL args.
Implement functions for AT TIME ZONE support.
Support "SAT" as an Australian time zone if USE_AUSTRALIAN_RULES
 is defined.

23 years agoImplement AT TIME ZONE SQL9x syntax.
Thomas G. Lockhart [Mon, 6 Nov 2000 15:47:17 +0000 (15:47 +0000)]
Implement AT TIME ZONE SQL9x syntax.
 AT is now a keyword but is not a reserved word.

23 years agoAllow type resolution for UNKNOWN arguments to functions to fall back to
Thomas G. Lockhart [Mon, 6 Nov 2000 15:42:30 +0000 (15:42 +0000)]
Allow type resolution for UNKNOWN arguments to functions to fall back to
 any available string type. Previously, all candidate choices must have
 fallen within the same "type category" for PostgreSQL to be willing to
 choose any of them.
Need to apply the same fixup to operator type resolution.

23 years agoUpdate TODO list.
Bruce Momjian [Mon, 6 Nov 2000 03:59:20 +0000 (03:59 +0000)]
Update TODO list.

23 years agoNew CHECKPOINT command.
Vadim B. Mikheev [Sun, 5 Nov 2000 22:50:21 +0000 (22:50 +0000)]
New CHECKPOINT command.
Auto removing of offline log files and creating new file
at checkpoint time.

23 years agoAdd some configure checks for DocBook and related tools. With a somewhat
Peter Eisentraut [Sun, 5 Nov 2000 21:04:07 +0000 (21:04 +0000)]
Add some configure checks for DocBook and related tools.  With a somewhat
standard installation layout it should be possible to build the HTML
and print documentation without additional manual setup.

23 years agoUpdate for recent developments (C++, shlib), minor fixes
Peter Eisentraut [Sun, 5 Nov 2000 20:52:59 +0000 (20:52 +0000)]
Update for recent developments (C++, shlib), minor fixes

23 years agoUNION select in a CREATE RULE caused a weird error, because transformRuleStmt
Tom Lane [Sun, 5 Nov 2000 01:42:07 +0000 (01:42 +0000)]
UNION select in a CREATE RULE caused a weird error, because transformRuleStmt
got confused by 'dummy' targetlist built for the UNION's toplevel query.
Fix by making dummy targetlist a little less cheesy.

23 years agoUpdate TODO list.
Bruce Momjian [Sun, 5 Nov 2000 00:22:02 +0000 (00:22 +0000)]
Update TODO list.

23 years agoAllow ORDER BY, LIMIT in sub-selects. Fix most (not all) cases where
Tom Lane [Sun, 5 Nov 2000 00:15:54 +0000 (00:15 +0000)]
Allow ORDER BY, LIMIT in sub-selects.  Fix most (not all) cases where
the grammar did not allow redundant parentheses around sub-selects.
Distinguish LIMIT ALL from LIMIT 0; make the latter behave as one would
expect.