OSDN Git Service

pg-rex/syncrep.git
21 years agosync
Bruce Momjian [Wed, 25 Jun 2003 21:47:24 +0000 (21:47 +0000)]
sync

21 years agosync
Bruce Momjian [Wed, 25 Jun 2003 21:46:10 +0000 (21:46 +0000)]
sync

21 years agosync
Bruce Momjian [Wed, 25 Jun 2003 21:45:40 +0000 (21:45 +0000)]
sync

21 years agosync
Bruce Momjian [Wed, 25 Jun 2003 21:45:06 +0000 (21:45 +0000)]
sync

21 years agoAdd:
Bruce Momjian [Wed, 25 Jun 2003 21:42:07 +0000 (21:42 +0000)]
Add:

> * Allow the identifier length to be increased via a configure option

21 years agoAdd:
Bruce Momjian [Wed, 25 Jun 2003 21:40:48 +0000 (21:40 +0000)]
Add:

> * Promote debug_query_string into a server-side function current_query()

21 years agoBack out rserv changes that would allow limit of replicated tables ---
Bruce Momjian [Wed, 25 Jun 2003 21:37:53 +0000 (21:37 +0000)]
Back out rserv changes that would allow limit of replicated tables ---
not ready yet.

21 years agoBack out array mega-patch.
Bruce Momjian [Wed, 25 Jun 2003 21:30:34 +0000 (21:30 +0000)]
Back out array mega-patch.

Joe Conway

21 years agoIn ISO datestyle, never emit just HH:MM, always emit HH:MM:SS or
Tom Lane [Wed, 25 Jun 2003 21:14:15 +0000 (21:14 +0000)]
In ISO datestyle, never emit just HH:MM, always emit HH:MM:SS or
HH:MM:SS.SSS... when there is a nonzero part-of-a-day field in an
interval value.  The seconds part used to be suppressed if zero,
but there's no equivalent behavior for timestamp, and since we're
modeling this format on timestamp it's probably wrong.  Per complaint
and patch from Larry Rosenman.

21 years agosync
Bruce Momjian [Wed, 25 Jun 2003 21:08:06 +0000 (21:08 +0000)]
sync

21 years agosync
Bruce Momjian [Wed, 25 Jun 2003 21:07:28 +0000 (21:07 +0000)]
sync

21 years agosync
Bruce Momjian [Wed, 25 Jun 2003 21:06:29 +0000 (21:06 +0000)]
sync

21 years agoDon't generate 'zero' typeids in the output from gen_cross_product.
Tom Lane [Wed, 25 Jun 2003 20:07:39 +0000 (20:07 +0000)]
Don't generate 'zero' typeids in the output from gen_cross_product.
This is no longer necessary or appropriate since we don't use zero typeid
as a wildcard anymore, and it fixes a nasty performance problem with
functions with many parameters.  Per recent example from Reuven Lerner.

21 years ago'third' -> 'fourth' where needed, per Joe Conway.
Tom Lane [Wed, 25 Jun 2003 18:13:50 +0000 (18:13 +0000)]
'third' -> 'fourth' where needed, per Joe Conway.

21 years agoThe remaining files ...
Michael Meskes [Wed, 25 Jun 2003 17:55:50 +0000 (17:55 +0000)]
The remaining files ...

21 years agoremove including of postgres_fe.h
Michael Meskes [Wed, 25 Jun 2003 16:12:27 +0000 (16:12 +0000)]
remove including of postgres_fe.h

21 years agosync.
Bruce Momjian [Wed, 25 Jun 2003 15:01:14 +0000 (15:01 +0000)]
sync.

21 years agoSome systems need another header file included.
Michael Meskes [Wed, 25 Jun 2003 11:51:18 +0000 (11:51 +0000)]
Some systems need another header file included.

21 years agoImplemented Informix special way to treat NULLs, removed warnings, synced.
Michael Meskes [Wed, 25 Jun 2003 10:44:21 +0000 (10:44 +0000)]
Implemented Informix special way to treat NULLs, removed warnings, synced.

21 years ago Fix up JOIN .. USING with domains
Bruce Momjian [Wed, 25 Jun 2003 04:32:03 +0000 (04:32 +0000)]
Fix up JOIN .. USING with domains

The attached fixes select_common_type() to support the below case:

create table t1( c1 int);
create domain dom_c1 int;
create table t2(c1 dom_c1);
select * from t1 join t2 using( c1 );

I didn't see a need for maintaining the domain as the preferred type. A
simple getBaseType() call on all elements of the list seems to be
enough.

--
Rod Taylor <rbt@rbt.ca>

21 years agoDone:
Bruce Momjian [Wed, 25 Jun 2003 04:23:59 +0000 (04:23 +0000)]
Done:

> * -Allow CIDR format to be used in pg_hba.conf

21 years agoDone:
Bruce Momjian [Wed, 25 Jun 2003 04:19:35 +0000 (04:19 +0000)]
Done:

> * -Allow UPDATE to use SET col = DEFAULT

21 years agoUPDATE ... SET <col> = DEFAULT
Bruce Momjian [Wed, 25 Jun 2003 04:19:24 +0000 (04:19 +0000)]
UPDATE ... SET <col> = DEFAULT

Rod Taylor

21 years agoIf they're not, the below causes problems, as the foreign key is added
Bruce Momjian [Wed, 25 Jun 2003 04:08:19 +0000 (04:08 +0000)]
If they're not, the below causes problems, as the foreign key is added
after the CHECK.  Cluster depends on the index name, so I thought it
wise to ensure all names are available, rather than leaving off the
CONSTRAINT "$n" portion for internally named constraints.

CREATE TABLE jkey (col integer primary key);
CREATE TABLE j (col integer REFERENCES jkey);
ALTER TABLE j ADD CHECK(col > 5);

This is a problem in 7.3 series as well as -Tip.

Rod Taylor <rbt@rbt.ca>

21 years agoUpdated the pg_get_constraintdef() to use conbin. Update pg_dump to use
Bruce Momjian [Wed, 25 Jun 2003 03:56:31 +0000 (03:56 +0000)]
Updated the pg_get_constraintdef() to use conbin.  Update pg_dump to use
pg_get_constraintdef() for >= 70400.

Rod Taylor <rbt@rbt.ca>

21 years agoAdd missing </row>'s.
Bruce Momjian [Wed, 25 Jun 2003 03:50:52 +0000 (03:50 +0000)]
Add missing </row>'s.

21 years agoIncludes:
Bruce Momjian [Wed, 25 Jun 2003 03:40:19 +0000 (03:40 +0000)]
Includes:

- LIKE <subtable> [ INCLUDING DEFAULTS | EXCLUDING DEFAULTS ]
- Quick cleanup of analyze.c function prototypes.
- New non-reserved keywords (INCLUDING, EXCLUDING, DEFAULTS), SQL 200X

Opted not to extend for check constraints at this time.

As per the definition that it's user defined columns, OIDs are NOT
inherited.

Doc and Source patches attached.

--
Rod Taylor <rbt@rbt.ca>

21 years agoBack out commit.
Bruce Momjian [Wed, 25 Jun 2003 03:02:37 +0000 (03:02 +0000)]
Back out commit.

21 years agoAdjust expected output for new functions.
Bruce Momjian [Wed, 25 Jun 2003 03:01:35 +0000 (03:01 +0000)]
Adjust expected output for new functions.

21 years agoAdd:
Bruce Momjian [Wed, 25 Jun 2003 02:43:54 +0000 (02:43 +0000)]
Add:

> * Allow creation of a libpq-only tarball
>

21 years agoSeems the runtime.sgml and perform.sgml patches combined are
Bruce Momjian [Wed, 25 Jun 2003 02:07:33 +0000 (02:07 +0000)]
Seems the runtime.sgml and perform.sgml patches combined are
incompatible.

I believe the attached xref meets the intent of the perform.sgml change.

--
Rod Taylor <rbt@rbt.ca>

21 years agoUpdate catversion for ipv6 addition.
Bruce Momjian [Wed, 25 Jun 2003 01:26:16 +0000 (01:26 +0000)]
Update catversion for ipv6 addition.

21 years agoDocumentation about using CIDR addresses in pg_hba.conf.
Bruce Momjian [Wed, 25 Jun 2003 01:20:50 +0000 (01:20 +0000)]
Documentation about using CIDR addresses in pg_hba.conf.

Andrew Dunstan

21 years ago> This change (I'm sure this will wrap poorly -- sorry):
Bruce Momjian [Wed, 25 Jun 2003 01:19:47 +0000 (01:19 +0000)]
> This change (I'm sure this will wrap poorly -- sorry):
> http://developer.postgresql.org/cvsweb.cgi/pgsql-server/src/include/libpq/pqcomm.h.diff?r1=1.85&r2=1.86
>
> modified SockAddr, but no corresponding change was made here
> (fe-auth.c:612):
>
>   case AUTH_REQ_KRB5:
> #ifdef KRB5
>   if (pg_krb5_sendauth(PQerrormsg, conn->sock, &conn->laddr.in,
>                        &conn->raddr.in,
>                        hostname) != STATUS_OK)
>
> It's not obvious to me what the change ought to be though.

This patch should hopefully fix both kerberos 4 and 5.

Kurt Roeckx

21 years ago> I've worked with the Pl/Python code in the past and will see about removing
Bruce Momjian [Wed, 25 Jun 2003 01:18:58 +0000 (01:18 +0000)]
> I've worked with the Pl/Python code in the past and will see about removing
> rexec and making it an untrusted language.  Last time I looked, it didn't
> look particularly difficult.  I've set aside some time next week, so stay
> tuned.

Attached is a patch that removes all of the RExec code from plpython from
the current PostgreSQL CVS.  In addition, plpython needs to be changed to an
untrusted language in createlang.  Please let me know if there are any
problems.

Kevin Jacobs

21 years agoAttached is a patch that provides *VERY* limited support for multiple
Bruce Momjian [Wed, 25 Jun 2003 01:17:44 +0000 (01:17 +0000)]
Attached is a patch that provides *VERY* limited support for multiple
slave
servers.  I haven't tested it very well, so use at your own risk (and I
recommend against using it in production).

Basically, I have a central database server that has 4 summary tables
inside
it replicated to a remote slave (these database tables are for my mail
server
authentication, so these are replicated to another server tuned for many
connections, and so I don't have postgres connections opened straight to
my
back-end database server).

Unfortunately, I also wanted to implement a replication database server
for
hot-backups.  I realized, too late, that the replication process is
pretty
greedy and will try to replicate all tables marked as a
"MasterAddTable".

To make a long story, I made a patch to RServ.pm and Replicate that
allows you
to specify, on the command line, a list of tables that you want to
replicate...it'll ignore all others.

I haven't finished, since this has to be integrated with CleanLog for
instance, but this should (and does) suffice for the moment.

I have yet to test it with two slaves, but at least my mail server
replication
database now works (it was failing every time it tried to replicate, for
a
variety of reasons).

Anyone have any suggestions on how to improve on this?  (or, if someone
more
familiar with this code wants to take the ball and run with it, you're
welcome to).

--
Michael A Nachbaur <mike@nachbaur.com>

21 years agoHere's a small patch to pg_hba.conf.sample that explains the use of CIDR
Bruce Momjian [Wed, 25 Jun 2003 01:15:24 +0000 (01:15 +0000)]
Here's a small patch to pg_hba.conf.sample that explains the use of CIDR
addresses.

Andrew Dunstan

21 years agoCreate <link> entries for the foreign keys in the system catalog
Bruce Momjian [Wed, 25 Jun 2003 01:14:37 +0000 (01:14 +0000)]
Create <link> entries for the foreign keys in the system catalog
chapter.

I got tired of hitting the 'up' link, then finding the table in the
table of contents, and following it back down.

Rod Taylor

21 years agoclient-auth.sgml references a renamed ID in runtime.sgml
Bruce Momjian [Wed, 25 Jun 2003 01:14:05 +0000 (01:14 +0000)]
client-auth.sgml references a renamed ID in runtime.sgml

Josh Berkus

21 years agoAttached is the fully corrected version of the re-ording patch for
Bruce Momjian [Wed, 25 Jun 2003 01:13:24 +0000 (01:13 +0000)]
Attached is the fully corrected version of the re-ording patch for
Runtime.sgml and dependant files.

Josh Berkus

21 years ago- Corrections for tables, columns with uppercase characters
Bruce Momjian [Wed, 25 Jun 2003 01:11:09 +0000 (01:11 +0000)]
- Corrections for tables, columns with uppercase characters
- Don't attempt to convert partial or expressional unique indexes
- Don't attempt to convert unique indexes based on a non-default
opclasses

- Untested prevention of conversion of non-btree indexes unique
indexes.  Untested as postgresql doesn't allow hash, gist, or rtree
based indexes to be unique.

rbt=# create unique index t on a using hash (col);
ERROR:  DefineIndex: access method "hash" does not support UNIQUE
indexes
rbt=# create unique index t on a using gist (col);
ERROR:  DefineIndex: access method "gist" does not support UNIQUE
indexes
rbt=# select version();
                                version
------------------------------------------------------------------------
 PostgreSQL 7.4devel on i386-unknown-freebsd4.8, compiled by GCC 2.95.4

Rod Taylor

21 years agoPlease apply attached patch to contrib/dblink. It adds named persistent
Bruce Momjian [Wed, 25 Jun 2003 01:10:15 +0000 (01:10 +0000)]
Please apply attached patch to contrib/dblink. It adds named persistent
connections to dblink.

Shridhar Daithanka

21 years agoThis is a bug in python interface module,
Bruce Momjian [Wed, 25 Jun 2003 01:09:24 +0000 (01:09 +0000)]
This is a bug in python interface module,
postgresql-7.3.3/src/interfaces/python/pg.py.

_quote() function fails due to integer overflow if input d is larger
than max integer.

In the case where the column type is "BIGINT", the input d may very well
be larger than max integer while its type, t, is labeled 'int'.
The conversion on line 19, return "%d" % int(d), will fail due to
"OverflowError: long int too large to convert to int".

Please describe a way to repeat the problem.   Please try to provide a
concise reproducible example, if at all possible:
----------------------------------------------------------------------

[1] create a table with a column type 'BIGINT'.
[2] use pg.DB.insert() to insert a value that is larger than max integer

If you know how this problem might be fixed, list the solution below:
---------------------------------------------------------------------

Just changing the conversion at line 19 of pg.py to long(d) instead of
int(d) should fix it. The following is a patch:

Chih-Hao Huang

21 years ago>> If a transaction marks a tuple for update and later commits without
Bruce Momjian [Wed, 25 Jun 2003 01:08:13 +0000 (01:08 +0000)]
>> If a transaction marks a tuple for update and later commits without
>> actually having updated the tuple, [...] can we simply
>> set the HEAP_XMAX_INVALID hint bit of the tuple?
>
>AFAICS this is a reasonable thing to do.

Thanks for the confirmation.  Here's a patch which also contains some
more noncritical changes to tqual.c:
 .  make code more readable by introducing local variables for xvac
 .  no longer two separate branches for aborted and crashed.
    The actions were the same in all cases.

Manfred Koizar

21 years agoMost of the synopsis areas for DROP commands use "name" as the
Bruce Momjian [Tue, 24 Jun 2003 23:29:25 +0000 (23:29 +0000)]
Most of the synopsis areas for DROP commands use "name" as the
identifier, while some areas do not.

The attached converts be below to "name":
conversion_name
index_name

The below have an existing, initdb supplied, entity named "name".  As
such, it could be confusing for the reader to see that identifier used
in the example.

domainname
typename

Rod Taylor

21 years agoRevert ORDBMS change, per Peter.
Bruce Momjian [Tue, 24 Jun 2003 23:27:24 +0000 (23:27 +0000)]
Revert ORDBMS change, per Peter.

21 years agoAlter query.sgml to mention PostgreSQL as an ORDBMS. This falls in line
Bruce Momjian [Tue, 24 Jun 2003 23:26:46 +0000 (23:26 +0000)]
Alter query.sgml to mention PostgreSQL as an ORDBMS.  This falls in line
with advocacy and 'portal' websites.

Link to createdb / dropdb from the tutorial page about create / dropdb.
A pair of notes were asking about more info...

Rod Taylor

21 years agoIn an attempt to simplify my life I'm submitting this patch that
Bruce Momjian [Tue, 24 Jun 2003 23:25:44 +0000 (23:25 +0000)]
In an attempt to simplify my life I'm submitting this patch that
restructures the deferred trigger queue.  The fundamental change is to
put all the static variables to hold the deferred triggers in a single
structure.

Alvaro Herrera

21 years agoAdd x86_64 support for spinlocks.
Bruce Momjian [Tue, 24 Jun 2003 23:20:08 +0000 (23:20 +0000)]
Add x86_64 support for spinlocks.

Jeffrey W. Baker

21 years agoSmall patch to link to the proper place in the "runtime" file,
Bruce Momjian [Tue, 24 Jun 2003 23:19:11 +0000 (23:19 +0000)]
Small patch to link to the proper place in the "runtime" file,
and to add the "schemaname" column to the description of the
pg_stats view.

Greg Sabino Mullane

21 years agoArray mega-patch.
Bruce Momjian [Tue, 24 Jun 2003 23:14:49 +0000 (23:14 +0000)]
Array mega-patch.

Joe Conway

21 years agoDone:
Bruce Momjian [Tue, 24 Jun 2003 23:05:20 +0000 (23:05 +0000)]
Done:

> * -Add IPv6 capability to INET/CIDR types

21 years agoJim C. Nasby wrote:
Bruce Momjian [Tue, 24 Jun 2003 22:59:46 +0000 (22:59 +0000)]
Jim C. Nasby wrote:
> Second argument to metaphone is suposed to set the limit on the
> number of characters to return, but it breaks on some phrases:
>
> usps=# select metaphone(a,3),metaphone(a,4),metaphone(a,20) from
> (select 'Hello world'::varchar AS a) a;
> HLW       | HLWR      | HLWRLT
>
> usps=# select metaphone(a,3),metaphone(a,4),metaphone(a,20) from
> (select 'A A COMEAUX MEMORIAL'::varchar AS a) a;
  > AKM       | AKMKS     | AKMKSMMRL
>
> In every case I've found that does this, the 4th and 5th letters are
> always 'KS'.

Nice catch.

There was a bug in the original metaphone algorithm from CPAN. Patch
attached (while I was at it I updated my email address, changed the
copyright to PGDG, and removed an unnecessary palloc). Here's how it
looks now:

regression=# select metaphone(a,4) from (select 'A A COMEAUX
MEMORIAL'::varchar AS a) a;
   metaphone
-----------
   AKMK
(1 row)

regression=# select metaphone(a,5) from (select 'A A COMEAUX
MEMORIAL'::varchar AS a) a;
   metaphone
-----------
   AKMKS
(1 row)

Joe Conway

21 years agoPrevent compiler warning from sprintf in recent ipv6 patch.
Bruce Momjian [Tue, 24 Jun 2003 22:42:42 +0000 (22:42 +0000)]
Prevent compiler warning from sprintf in recent ipv6 patch.

21 years ago Add ipv6 address parsing support to 'inet' and 'cidr' data types.
Bruce Momjian [Tue, 24 Jun 2003 22:21:24 +0000 (22:21 +0000)]
    Add ipv6 address parsing support to 'inet' and 'cidr' data types.

        Regression tests for IPv6 operations added.

        Documentation updated to document IPv6 bits.

        Stop treating IPv4 as an "unsigned int" and IPv6 as an array of
        characters.  Instead, always use the array of characters so we
        can have one function fits all.  This makes bitncmp(), addressOK(),
        and several other functions "just work" on both address families.

        add family() function which returns integer 4 or 6 for IPv4 or
        IPv6.  (See examples below)  Note that to add this new function
        you will need to dump/initdb/reload or find the correct magic
        to add the function to the postgresql function catalogs.

        IPv4 addresses always sort before IPv6.

        On disk we use AF_INET for IPv4, and AF_INET+1 for IPv6 addresses.
        This prevents the need for a dump and reload, but lets IPv6 parsing
        work on machines without AF_INET6.

        To select all IPv4 addresses from a table:

                select * from foo where family(addr) = 4 ...

        Order by and other bits should all work.

Michael Graff

21 years agoUpdate:
Bruce Momjian [Tue, 24 Jun 2003 18:45:38 +0000 (18:45 +0000)]
Update:

< * Have SELECT '13 minutes'::interval display zero seconds
> * Have SELECT '13 minutes'::interval display zero seconds in ISO datestyle

21 years agoAdd -L../ecpglib -lecpg for Cygwin compile.
Bruce Momjian [Tue, 24 Jun 2003 14:45:46 +0000 (14:45 +0000)]
Add  -L../ecpglib -lecpg for Cygwin compile.

21 years agoAdd:
Bruce Momjian [Tue, 24 Jun 2003 03:42:44 +0000 (03:42 +0000)]
Add:

> * Have SELECT '13 minutes'::interval display zero seconds

21 years agoAdd:
Bruce Momjian [Tue, 24 Jun 2003 03:40:09 +0000 (03:40 +0000)]
Add:

> * Allow the regression tests to start postmaster with -i so the tests
>   can be run on systems that don't support unix-domain sockets

21 years agoAdd:
Bruce Momjian [Tue, 24 Jun 2003 02:54:23 +0000 (02:54 +0000)]
Add:

> * Rename /scripts directory because they are all C programs now

21 years agoAdd:
Bruce Momjian [Tue, 24 Jun 2003 02:53:36 +0000 (02:53 +0000)]
Add:

<
>  o Allow ALTER TABLE to modify column lengths and change to binary
>    compatible types

21 years agoAdd:
Bruce Momjian [Tue, 24 Jun 2003 02:39:49 +0000 (02:39 +0000)]
Add:

> * Modify pg_get_triggerdef() to take a boolean to pretty-print,
>  and use that as part of pg_dump along with psql
>

21 years agoUpdate:
Bruce Momjian [Tue, 24 Jun 2003 01:54:57 +0000 (01:54 +0000)]
Update:

* Allow UPDATE, DELETE to handle table aliases for self-joins [delete]

21 years agotypedef sa_family_t for cygwin.
Bruce Momjian [Tue, 24 Jun 2003 01:49:22 +0000 (01:49 +0000)]
typedef sa_family_t for cygwin.

21 years agoAdd:
Bruce Momjian [Tue, 24 Jun 2003 01:08:20 +0000 (01:08 +0000)]
Add:

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

21 years agoDo not include:
Bruce Momjian [Tue, 24 Jun 2003 00:44:29 +0000 (00:44 +0000)]
Do not include:

#include <pwd.h>
#include <netdb.h>

in MS CC/Borland builds.

21 years agoRemove references to sa_family_t, except when SOCKADDR_STORAGE requires
Bruce Momjian [Mon, 23 Jun 2003 23:52:00 +0000 (23:52 +0000)]
Remove references to sa_family_t, except when SOCKADDR_STORAGE requires
it.

Also handle __ss_family as a synonym for ss_family.

Kurt Roeckx

21 years agoAdd with '?':
Bruce Momjian [Mon, 23 Jun 2003 19:33:40 +0000 (19:33 +0000)]
Add with '?':

> * Have initdb set DateStyle based on locale?

21 years agoAdd mention that the grep's have to be adjusted in pgtest.
Bruce Momjian [Mon, 23 Jun 2003 19:33:25 +0000 (19:33 +0000)]
Add mention that the grep's have to be adjusted in pgtest.

21 years agoRemove '?':
Bruce Momjian [Mon, 23 Jun 2003 19:22:30 +0000 (19:22 +0000)]
Remove '?':

<   from making invalid dates valid?
>   from making invalid dates valid

21 years agoChange pqInternalNotice to accept a format string and args instead of
Tom Lane [Mon, 23 Jun 2003 19:20:25 +0000 (19:20 +0000)]
Change pqInternalNotice to accept a format string and args instead of
just a preformatted message; per suggestion by Sean Chittenden.

21 years agoRemove crypt.h from fe-connect.c --- not needed, and caused problems on
Bruce Momjian [Mon, 23 Jun 2003 17:03:19 +0000 (17:03 +0000)]
Remove crypt.h from fe-connect.c --- not needed, and caused problems on
Solaris with Open SSL version 0.9.7b

21 years agoUpdate Russian version of FAQ.
Bruce Momjian [Mon, 23 Jun 2003 16:58:21 +0000 (16:58 +0000)]
Update Russian version of FAQ.

21 years agoAdd:
Bruce Momjian [Mon, 23 Jun 2003 04:31:25 +0000 (04:31 +0000)]
Add:

> * Exit postmaster if postgresql.conf can not be opened

21 years agoAdd:
Bruce Momjian [Mon, 23 Jun 2003 04:00:10 +0000 (04:00 +0000)]
Add:

* Allow SSL-enabled clients to turn off SSL transfers

21 years agoAdd:
Bruce Momjian [Mon, 23 Jun 2003 03:52:24 +0000 (03:52 +0000)]
Add:

> * Change CVS $Id$ to $PostgreSQL$

21 years agoRevise hash join and hash aggregation code to use the same datatype-
Tom Lane [Sun, 22 Jun 2003 22:04:55 +0000 (22:04 +0000)]
Revise hash join and hash aggregation code to use the same datatype-
specific hash functions used by hash indexes, rather than the old
not-datatype-aware ComputeHashFunc routine.  This makes it safe to do
hash joining on several datatypes that previously couldn't use hashing.
The sets of datatypes that are hash indexable and hash joinable are now
exactly the same, whereas before each had some that weren't in the other.

21 years agoFix some index entries.
Tom Lane [Sun, 22 Jun 2003 16:17:01 +0000 (16:17 +0000)]
Fix some index entries.

21 years agoAdjust chapter ordering in Internals part to something that seems more
Tom Lane [Sun, 22 Jun 2003 16:16:44 +0000 (16:16 +0000)]
Adjust chapter ordering in Internals part to something that seems more
reasonable.

21 years agoAdded missing terminating '\0' char for data put into char *.
Michael Meskes [Sun, 22 Jun 2003 11:00:48 +0000 (11:00 +0000)]
Added missing terminating '\0' char for data put into char *.

21 years agoRemove a lot of desperately obsolete material (which was all out of sight,
Tom Lane [Sun, 22 Jun 2003 05:48:26 +0000 (05:48 +0000)]
Remove a lot of desperately obsolete material (which was all out of sight,
out of mind, because it'd been commented out years ago).  Try to bring the
remains up to a reasonable level of currency, and give it all approximately
the same high level of abstraction.

21 years agoAdd question mark to item:
Bruce Momjian [Sun, 22 Jun 2003 05:04:52 +0000 (05:04 +0000)]
Add question mark to item:

<   from making invalid dates valid
>   from making invalid dates valid?

21 years agoAdd:
Bruce Momjian [Sun, 22 Jun 2003 05:01:17 +0000 (05:01 +0000)]
Add:

> * Allow current datestyle to restrict dates;  prevent month/day swapping
>   from making invalid dates valid
> * Prevent month/day swapping of ISO dates to make invalid dates valid

21 years agoAdd pgtest script.
Bruce Momjian [Sun, 22 Jun 2003 04:22:57 +0000 (04:22 +0000)]
Add pgtest script.

21 years agoUse libpq's new logic to get the server version, instead of doing it ourselves.
Tom Lane [Sun, 22 Jun 2003 00:56:58 +0000 (00:56 +0000)]
Use libpq's new logic to get the server version, instead of doing it ourselves.

21 years agoBring the libpq example programs into the 21st century.
Tom Lane [Sun, 22 Jun 2003 00:29:29 +0000 (00:29 +0000)]
Bring the libpq example programs into the 21st century.

21 years agoGet rid of extraneous newline in PQendcopy error output (was causing
Tom Lane [Sat, 21 Jun 2003 23:25:38 +0000 (23:25 +0000)]
Get rid of extraneous newline in PQendcopy error output (was causing
regression test diffs...).

21 years agoUpdate libpq to make new features of FE/BE protocol available to
Tom Lane [Sat, 21 Jun 2003 21:51:35 +0000 (21:51 +0000)]
Update libpq to make new features of FE/BE protocol available to
client applications.  Some editorial work on libpq.sgml, too.

21 years agoFix some markup problems.
Tom Lane [Sat, 21 Jun 2003 19:33:36 +0000 (19:33 +0000)]
Fix some markup problems.

21 years agoFix for extended-query protocol: in event of error, backend was issuing
Tom Lane [Fri, 20 Jun 2003 21:58:02 +0000 (21:58 +0000)]
Fix for extended-query protocol: in event of error, backend was issuing
a ReadyForQuery (Z message) immediately and then another one after the
Sync message arrives.  Suppress the first one to make it work per spec.

21 years agoSynced with backend.
Michael Meskes [Fri, 20 Jun 2003 15:16:06 +0000 (15:16 +0000)]
Synced with backend.

21 years agoJust another Informix compatibility change. They uses "free" for cursors as wellafter...
Michael Meskes [Fri, 20 Jun 2003 13:36:34 +0000 (13:36 +0000)]
Just another Informix compatibility change. They uses "free" for cursors as wellafter closing them.

21 years agoJust another ecpg todo done.
Michael Meskes [Fri, 20 Jun 2003 13:20:09 +0000 (13:20 +0000)]
Just another ecpg todo done.

21 years agoSorry, missed a file.
Michael Meskes [Fri, 20 Jun 2003 12:01:46 +0000 (12:01 +0000)]
Sorry, missed a file.

21 years agoAllow constants in using clauses.
Michael Meskes [Fri, 20 Jun 2003 12:00:59 +0000 (12:00 +0000)]
Allow constants in using clauses.

21 years agoWe neglected to set conn->raddr.salen, leading to breakage of CANCEL
Tom Lane [Fri, 20 Jun 2003 04:09:12 +0000 (04:09 +0000)]
We neglected to set conn->raddr.salen, leading to breakage of CANCEL
and probably other stuff.

21 years agoDisallow dollar sign in operator names, instead allow it as a non-first
Tom Lane [Thu, 19 Jun 2003 23:22:40 +0000 (23:22 +0000)]
Disallow dollar sign in operator names, instead allow it as a non-first
character in identifiers.  The first change eliminates the current need
to put spaces around parameter references, as in "x<=$2".  The second
change improves compatibility with Oracle and some other RDBMSes.  This
was discussed and agreed to back in January, but did not get done.

21 years agoFixed fetch into char * and added missing prototype for an Informix function.
Michael Meskes [Thu, 19 Jun 2003 09:52:11 +0000 (09:52 +0000)]
Fixed fetch into char * and added missing prototype for an Informix function.

21 years agoMove thread checking code farther down in conflgure.
Bruce Momjian [Wed, 18 Jun 2003 16:04:15 +0000 (16:04 +0000)]
Move thread checking code farther down in conflgure.

21 years agoChange clusterdb and vacuumdb into C programs.
Peter Eisentraut [Wed, 18 Jun 2003 12:19:11 +0000 (12:19 +0000)]
Change clusterdb and vacuumdb into C programs.