Tom Lane [Wed, 7 Feb 2001 20:30:20 +0000 (20:30 +0000)]
Use explicit path to libpgtcl.so, instead of relying on LD_LIBRARY_PATH
or local equivalent. Also, honor --with-pgport configure option for
default port number, and allow PGPORT environment variable to override
this.
Only pass the -L* portions of LDFLAGS to the Python build environment.
Other flags can have unpredictable effects when Python uses different
commands to build than we do.
Tom Lane [Wed, 7 Feb 2001 17:59:58 +0000 (17:59 +0000)]
Need to specify DYNAMIC_PATH flag to shl_load, else HPUX's dld will not
honor library search path that we so carefully installed into the
executable ...
Hiroshi Inoue [Tue, 6 Feb 2001 02:21:12 +0000 (02:21 +0000)]
Improved version handling introduced by Dave Page.
The driver version is 07.01.0002 now.
1) initialized pg_version by DSN's protocol info
so that we could always use pg_version info
once a connection is established (pg_version()
didn't exist before 6.4). PROTOCOL_XX() macros
are removed(except from connection.[ch]).
2) provided a few macros to encapsulate connection's
version info and replaced existent comparison
stuff by those macros.
3) change SQLTables() so that 7.1 servers could show
views.
In addtion, the following patch from Dave Page is applied.
This patch fixes a bug in SQLGetInfo for SQL_DBMS_VER which corrupted the
driver version string. The driver version number has also been incremented
to 07.01.0002.
Tom Lane [Tue, 6 Feb 2001 01:53:53 +0000 (01:53 +0000)]
Out-of-bounds memory allocation request sizes should be treated as just
elog(ERROR) not an Assert trap, since we've downgraded out-of-memory to
elog(ERROR) not a fatal error. Also, change the hard boundary from 256Mb
to 1Gb, just so that anyone who's actually got that much memory to spare
can play with TOAST objects approaching a gigabyte.
Dump/display 'timestamp' as 'timestamp with time zone', to make room for a
future 'timestamp without time zone', which SQL claims is equivalent to
plain 'timestamp'.
When making xrefs to refsects, the linkend should point to the refsect and
the endterm to the title element, not both to the title. At least
docbook2man prefers it that way.
Bruce Momjian [Thu, 1 Feb 2001 18:31:10 +0000 (18:31 +0000)]
Please replace with the one attached to this email message. The
previous version depended on the below patch, this one does not. It is
also a bit cleaner and has a few more helpful comments.
Peter Mount [Wed, 31 Jan 2001 09:23:45 +0000 (09:23 +0000)]
Wed Jan 31 08:46:00 GMT 2001 peter@retep.org.uk
- Some minor additions to Statement to make our own extensions more
portable.
- Statement.close() will now call ResultSet.close() rather than just
dissasociating with it.
Peter Mount [Wed, 31 Jan 2001 08:26:02 +0000 (08:26 +0000)]
Tue Jan 30 22:24:00 GMT 2001 peter@retep.org.uk
- Fixed bug where Statement.setMaxRows() was a global setting. Now
limited to just itself.
- Changed LargeObject.read(byte[],int,int) to return the actual number
of bytes read (used to be void).
- LargeObject now supports InputStream's!
- PreparedStatement.setBinaryStream() now works!
- ResultSet.getBinaryStream() now returns an InputStream that doesn't
copy the blob into memory first!
- Connection.isClosed() now tests to see if the connection is still alive
rather than if it thinks it's alive.
Tom Lane [Mon, 29 Jan 2001 00:39:20 +0000 (00:39 +0000)]
Clean up handling of tuple descriptors so that result-tuple descriptors
allocated by plan nodes are not leaked at end of query. This doesn't
really matter for normal queries, but it sure does for queries invoked
repetitively inside SQL functions. Clean up some other grotty code
associated with tupdescs, and fix a few other memory leaks exposed by
tests with simple SQL functions.
Philip Warner [Sun, 28 Jan 2001 02:57:06 +0000 (02:57 +0000)]
Applied (slightly modified) patches from Tatsuo:
Ok. I have made patches for fixing some of pg_dump problems(see
attached patches). The patches address the problem with user defined
functions, operators and aggregates.
Bruce Momjian [Sat, 27 Jan 2001 21:49:59 +0000 (21:49 +0000)]
Here is an update on the Win32 patch. Modified files are 'config.h.win32'
and two 'win32.mak'. Addresses the following:
1) Oops. Spelled fcntl.h wrong in the last one. D'uh.
2) PG_VERSION changed to be defined with " around it. psql/command.c failed
to compile without that.
3) Changed makefiles to use "/MD" and link both psql and libpq.dll against
MSVCRT.DLL instead of a static library. This takes care of the
crash-upon-free in psql.
I *think* this is what is on the "Open 7.1 Items" list as "Magnus Hagander
ODBC Issues?". It has nothing to do with ODBC, but it's the only issue I've
been involved with...
Tom Lane [Sat, 27 Jan 2001 04:42:32 +0000 (04:42 +0000)]
Don't crash if subquery appears multiple times in jointree. This should
not happen anyway, but let's try not to get completely confused if it does
(due to rewriter bugs or whatever).
Tom Lane [Sat, 27 Jan 2001 04:40:59 +0000 (04:40 +0000)]
Repair bug reported by Huxton, 1/24/01. We need to include a rule's
original table ('OLD' table) in its join tree if OLD is referenced by
either the rule action, the rule qual, or the original query qual that
will be added to the rule action. However, we only want one instance
of the original table to be included; so beware of the possibility that
the rule action already has a jointree entry for OLD.
Bruce Momjian [Sat, 27 Jan 2001 04:26:37 +0000 (04:26 +0000)]
Just a quick patch to make the geometry test on Sparc/Linux
regression tests for Pgsql 7.1beta3 pass. This is very similr to the one I
submitted back in July for Linux/Alpha. Apparently non-x86 Linux machines
like to compute nth place float point digits like Sun/Solaris does?
Otherwise, 7.1beta3 runs without problems (i.e. all other
regression tests pass) on my Sparc 20 running Debian GNU/Linux 2.2.
Tom Lane [Sat, 27 Jan 2001 01:44:20 +0000 (01:44 +0000)]
Refuse to try to attach a condition to a NOTIFY or other utility statement,
rather than coredumping (as prior 7.1 code did) or silently dropping the
condition (as 7.0 did). This is annoying but there doesn't seem to be
any good way to fix it, short of a major querytree restructuring.
Tom Lane [Sat, 27 Jan 2001 00:05:31 +0000 (00:05 +0000)]
Re-read Unix-socket lock file every so often (every CheckPoint interval,
actually) to ensure that its file access time doesn't get old enough to
tempt a /tmp directory cleaner to remove it. Still another reason we
should never have put the sockets in /tmp in the first place ...
Tom Lane [Fri, 26 Jan 2001 22:50:26 +0000 (22:50 +0000)]
Change float8-to-int8 conversion to round to nearest, rather than
truncating to integer. Remove regress test that checks whether 4567890123456789 can be converted to float without loss; since that's
52 bits, it's on the hairy edge of failing with IEEE float8s, and indeed
rint seems to give platform-dependent results for it.
Bruce Momjian [Fri, 26 Jan 2001 22:41:59 +0000 (22:41 +0000)]
odbc1.diff changes the text on the Protocol Radio buttons on the driver
dialogue from '6.4/6.5' to '6.5+' and removes some C++ comments from
resource.h (which VC++ insists on putting there).
odbc2.diff adds code to query the PostgreSQL version upon connection. This
is then used to determine what values to return for from SQLGetInfo for
SQL_DBMS_VER, SQL_MAX_ROW_SIZE, SQL_MAX_STATEMENT_LEN, SQL_OJ_CAPABILITIES
and SQL_OUTER_JOINS. The version string as returned by SELECT vERSION() (as
a char array) and the major.minor version number (as a flost) have been
added to the ConnectionClass structure.