Comment 5 for bug 866049

This bug was fixed in the package postgresql-8.4 - 8.4.9-0ubuntu0.11.04

---------------
postgresql-8.4 (8.4.9-0ubuntu0.11.04) natty-security; urgency=low

  * New upstream bug fix/security release: (LP: #866049)
     - Fix bugs in indexing of in-doubt HOT-updated tuples.
       These bugs could result in index corruption after reindexing a
       system catalog. They are not believed to affect user indexes.
     - Fix multiple bugs in GiST index page split processing.
       The probability of occurrence was low, but these could lead to
       index corruption.
     - Fix possible buffer overrun in tsvector_concat().
       The function could underestimate the amount of memory needed for
       its result, leading to server crashes.
     - Fix crash in xml_recv when processing a "standalone" parameter.
     - Make pg_options_to_table return NULL for an option with no value.
       Previously such cases would result in a server crash.
     - Avoid possibly accessing off the end of memory in "ANALYZE" and in
       SJIS-2004 encoding conversion.
       This fixes some very-low-probability server crash scenarios.
     - Prevent intermittent hang in interactions of startup process with
       bgwriter process.
       This affected recovery in non-hot-standby cases.
     - Fix race condition in relcache init file invalidation.
       There was a window wherein a new backend process could read a stale
       init file but miss the inval messages that would tell it the data
       is stale. The result would be bizarre failures in catalog accesses,
       typically "could not read block 0 in file ..." later during
       startup.
     - Fix memory leak at end of a GiST index scan.
       Commands that perform many separate GiST index scans, such as
       verification of a new GiST-based exclusion constraint on a table
       already containing many rows, could transiently require large
       amounts of memory due to this leak.
     - Fix incorrect memory accounting (leading to possible memory bloat)
       in tuplestores supporting holdable cursors and plpgsql's RETURN
       NEXT command.
     - Fix performance problem when constructing a large, lossy bitmap.
     - Fix join selectivity estimation for unique columns.
       This fixes an erroneous planner heuristic that could lead to poor
       estimates of the result size of a join.
     - Fix nested PlaceHolderVar expressions that appear only in
       sub-select target lists. This mistake could result in outputs of an
       outer join incorrectly appearing as NULL.
     - Allow nested EXISTS queries to be optimized properly.
     - Fix array- and path-creating functions to ensure padding bytes are
       zeroes. This avoids some situations where the planner will think that
       semantically-equal constants are not equal, resulting in poor
       optimization.
     - Fix "EXPLAIN" to handle gating Result nodes within inner-indexscan
       subplans. The usual symptom of this oversight was "bogus varno" errors.
     - Work around gcc 4.6.0 bug that breaks WAL replay. This could lead to
       loss of committed transactions after a server crash.
     - Fix dump bug for VALUES in a view.
     - Disallow SELECT FOR UPDATE/SHARE on sequences.
       This operation doesn't work as expected and can lead to failures.
     - Fix "VACUUM" so that it always updates pg_class.reltuples/relpages.
       This fixes some scenarios where autovacuum could make increasingly
       poor decisions about when to vacuum tables.
     - Defend against integer overflow when computing size of a hash table.
     - Fix cases where "CLUSTER" might attempt to access already-removed
       TOAST data.
     - Fix portability bugs in use of credentials control messages for
       "peer" authentication.
     - Fix SSPI login when multiple roundtrips are required.
       The typical symptom of this problem was "The function requested is
       not supported" errors during SSPI login.
     - Throw an error if "pg_hba.conf" contains hostssl but SSL is
       disabled. This was concluded to be more user-friendly than the
       previous behavior of silently ignoring such lines.
     - Fix typo in pg_srand48 seed initialization.
       This led to failure to use all bits of the provided seed. This
       function is not used on most platforms (only those without
       srandom), and the potential security exposure from a
       less-random-than-expected seed seems minimal in any case.
     - Avoid integer overflow when the sum of LIMIT and OFFSET values
       exceeds 2^63.
     - Add overflow checks to int4 and int8 versions of generate_series().
     - Fix trailing-zero removal in to_char(). In a format with FM and no
       digit positions after the decimal point, zeroes to the left of the
       decimal point could be removed incorrectly.
     - Fix pg_size_pretty() to avoid overflow for inputs close to 2^63.
     - Weaken plpgsql's check for typmod matching in record values.
       An overly enthusiastic check could lead to discarding length
       modifiers that should have been kept.
     - Fix pg_upgrade to preserve toast tables' relfrozenxids during an
       upgrade from 8.3. Failure to do this could lead to "pg_clog" files
       being removed too soon after the upgrade.
     - Fix psql's counting of script file line numbers during COPY from a
       different file.
     - Fix pg_restore's direct-to-database mode for
       standard_conforming_strings. pg_restore could emit incorrect commands
       when restoring directly to a database server from an archive file that
       had been made with standard_conforming_strings set to on.
     - Be more user-friendly about unsupported cases for parallel
       pg_restore. This change ensures that such cases are detected and
       reported before any restore actions have been taken.
     - Fix write-past-buffer-end and memory leak in libpq's LDAP service
       lookup code.
     - In libpq, avoid failures when using nonblocking I/O and an SSL
       connection.
     - Improve libpq's handling of failures during connection startup.
       In particular, the response to a server report of fork() failure
       during SSL connection startup is now saner.
     - Improve libpq's error reporting for SSL failures.
     - Fix PQsetvalue() to avoid possible crash when adding a new tuple to
       a PGresult originally obtained from a server query.
     - Make ecpglib write double values with 15 digits precision.
     - In ecpglib, be sure LC_NUMERIC setting is restored after an error.
     - Apply upstream fix for blowfish signed-character bug
       (CVE-2011-2483) (Closes: #631285)
       "contrib/pg_crypto"'s blowfish encryption code could give wrong
       results on platforms where char is signed (which is most), leading
       to encrypted passwords being weaker than they should be.
     - Fix memory leak in "contrib/seg".
     - Fix pgstatindex() to give consistent results for empty indexes.
     - Allow building with perl 5.14. (Closes: #628503)
 -- Martin Pitt <email address hidden> Tue, 04 Oct 2011 12:00:41 +0200