[ Thanks to Robert Treat for this
link. ]
== PostgreSQL Weekly News – September 4th 2003 ==
Our first week of beta2 seemed to keep things moving along
briskly. There was some confusion as to what had been included in
this beta2, so some steps are being made to give everyone a better
heads up when the next beta build is tagged. For those concerned
that a specific fix may not be in yet, rest assured that the
PostgreSQL development process will usually have several betas to
get everything in, and then usually several release candidates
before the code is deemed ready for production. With that said lets
review some of the changes made this past week.
A lot of discussion took place regarding how best to support
thread safety on the variety of platforms that PostgreSQL supports.
Some of the changes that came out of that were add a test for
thread-safeness of libc functions and reporting of specific
threading failure functions. Unixware was marked as not requiring
*_r functions and BSD/OS was verified not needing reentrant
function names. Also the casting of function call pointers for
threads was tightened.
Date/time input parsing code was adjusted to correctly
distinguish the four SQLSTATE error codes required by SQL99
(invalid format, datetime field overflow, interval field overflow,
invalid time zone displacement value). Subplan clauses are now
pushed to the back in qual lists for join plans. We now ensure that
FK constraints are dumped after primary/unique constraints, rather
than relying on OID order of pg_constraint entries. The -r logic
for pg_dump and pg_restore was centralized to prevent any
divergence in the future. A couple bugs in ecpg pgtypes library
were fixed. The last remnants of the old pygresql code were cleaned
out and some outdated CLI bits were removed. Work around ENOTEMPTY
being an alias for EEXIST on some platforms. Repair problems with
to_char() overrunning its input string (from Karel Zak). Pass
session_authorization to the client and make psql update its prompt
accordingly. Allow non-super users to set log_duration to true,
then false, when the administrator has not turned it on, and fix
other PGC_USERLIMIT variables. And last but not least standalone
backends (including bootstrap case) now read the GUC config file if
it exists. This was already discussed as being a good idea (and in
fact knocks off a listed TODO item), and now seems the cleanest way
to deal with initdb-time failures on machines with small
SHMMAX.
A number of documentation updates went in again this week. Some,
like that addition of an for the JADEFLAGS make variable and
setting of LC_ALL=C when running collateindex.pl, affected the
documentation build process. Others changes were aimed at content
including some major cleanup of the SPI chapter, the addition and
editing of a number of the index entries, some improvement for
pg_restore’s docs regarding use of the -N, -o, and -r options, some
information regarding OID wraparound, and a note for plpgsql to
explain that a loop where the lower bound is greater than the upper
bound is valid but does nothing.
Last but not least, Tom Lane has been doing quite an overhaul in
the hash index code, addressing some serious issues recently raised
on the -hackers mailing list. Some of the work included rewriting
hashbulkdelete() to make it amenable to new bucket locking scheme
(with the pleasant side effect that it is *much* faster when
deleting a large fraction of the indexed tuples), widening
hashm_ntuples to double so that hash space management doesn’t get
confused by more than 4G entries, enlarging the allowed number of
free-space-bitmap pages, replacing the useless bshift field with a
useful bmshift field, eliminating 4 bytes of wasted space in the
per-page special area, and fixing a buffer reference leak and
incorrect computation of bit indexes when using multiple bitmap
pages. The changes do change the on-disk index layout, so a reindex
would be necessary for folks upgrading between beta releases. He
also fixed a few typos and added some more comments to the
documentation. Since this is a rather large change late in the
development cycle, users of hash indexes are strongly encourage to
test their applications on the latest beta release.
== PostgreSQL Product News ==
PolarLake Adds Support For Eclipse
http://www.ebizq.net/news/2678.html
== PostgreSQL In the News ==
PostgreSQL, Inc Releases Replication Product under BSD
License
http://www.osnews.com/story.php?news_id=4375
Random thoughts on GPL and BSD (blog)
http://ertw.com/blog/archives/000398.html#000398
No respect for the old Elephant ? (blog)
http://low.teacentral.net/index.php?p=4&c=1
== Upcoming Events ==
Gnu/Linux Congress: Veracruz, Mexico: September 17-19
Bruce Momjian has been asked to give a keynote address and he will
also be presenting tutorials
http://congreso.gulev.org.mx
LinuxWorld Conference & Expo 2003 : Frankfurt, Germany :
October 27-29
PostgreSQL will have a booth at the LinuxWorld Conference
http://www.linuxworldexpo.de/
== PostgreSQL Weekly News – September 4th 2003 ==
Don’t forget to read Elein Mustain’s Weekly Summary of the
PostgreSQL General Mailing List http://www.varlena.com/GeneralBits/
On the Web:
http://www.postgresql.org
http://advocacy.postgresql.org