Comment 5 for bug 655293

Revision history for this message
Launchpad Janitor (janitor) wrote :

This bug was fixed in the package postgresql-8.4 - 8.4.5-0ubuntu10.04

---------------
postgresql-8.4 (8.4.5-0ubuntu10.04) lucid-security; urgency=low

  * New upstream security/bug fix update: (LP: #655293)
    - Use a separate interpreter for each calling SQL userid in PL/Perl
      and PL/Tcl.
      This change prevents security problems that can be caused by
      subverting Perl or Tcl code that will be executed later in the same
      session under another SQL user identity (for example, within a
      SECURITY DEFINER function). Most scripting languages offer numerous
      ways that that might be done, such as redefining standard functions
      or operators called by the target function. Without this change,
      any SQL user with Perl or Tcl language usage rights can do
      essentially anything with the SQL privileges of the target
      function's owner.
      The cost of this change is that intentional communication among
      Perl and Tcl functions becomes more difficult. To provide an escape
      hatch, PL/PerlU and PL/TclU functions continue to use only one
      interpreter per session. This is not considered a security issue
      since all such functions execute at the trust level of a database
      superuser already.
      It is likely that third-party procedural languages that claim to
      offer trusted execution have similar security issues. We advise
      contacting the authors of any PL you are depending on for
      security-critical purposes.
      Our thanks to Tim Bunce for pointing out this issue
      (CVE-2010-3433).
    - Prevent possible crashes in pg_get_expr() by disallowing it from
      being called with an argument that is not one of the system catalog
      columns it's intended to be used with.
    - Fix incorrect placement of placeholder evaluation.
      This bug could result in query outputs being non-null when they
      should be null, in cases where the inner side of an outer join is a
      sub-select with non-strict expressions in its output list.
    - Fix possible duplicate scans of UNION ALL member relations.
    - Fix "cannot handle unplanned sub-select" error.
      This occurred when a sub-select contains a join alias reference
      that expands into an expression containing another sub-select.
    - Fix mishandling of whole-row Vars that reference a view or
      sub-select and appear within a nested sub-select.
    - Fix mishandling of cross-type IN comparisons.
      This could result in failures if the planner tried to implement an
      IN join with a sort-then-unique-then-plain-join plan.
    - Fix computation of "ANALYZE" statistics for tsvector columns.
      The original coding could produce incorrect statistics, leading to
      poor plan choices later.
    - Improve planner's estimate of memory used by array_agg(),
      string_agg(), and similar aggregate functions.
      The previous drastic underestimate could lead to out-of-memory
      failures due to inappropriate choice of a hash-aggregation plan.
    - Fix failure to mark cached plans as transient.
      If a plan is prepared while "CREATE INDEX CONCURRENTLY" is in
      progress for one of the referenced tables, it is supposed to be
      re-planned once the index is ready for use. This was not happening
      reliably.
    - Reduce PANIC to ERROR in some occasionally-reported btree failure
      cases, and provide additional detail in the resulting error
      messages.
      This should improve the system's robustness with corrupted indexes.
    - Fix incorrect search logic for partial-match queries with GIN
      indexes.
      Cases involving AND/OR combination of several GIN index conditions
      didn't always give the right answer, and were sometimes much slower
      than necessary.
    - Prevent show_session_authorization() from crashing within
      autovacuum processes.
    - Defend against functions returning setof record where not all the
      returned rows are actually of the same rowtype.
    - Fix possible corruption of pending trigger event lists during
      subtransaction rollback.
      This could lead to a crash or incorrect firing of triggers.
    - Fix possible failure when hashing a pass-by-reference function
      result.
    - Improve merge join's handling of NULLs in the join columns.
      A merge join can now stop entirely upon reaching the first NULL, if
      the sort order is such that NULLs sort high.
    - Take care to fsync the contents of lockfiles (both "postmaster.pid"
      and the socket lockfile) while writing them.
      This omission could result in corrupted lockfile contents if the
      machine crashes shortly after postmaster start. That could in turn
      prevent subsequent attempts to start the postmaster from
      succeeding, until the lockfile is manually removed.
    - Avoid recursion while assigning XIDs to heavily-nested
      subtransactions.
      The original coding could result in a crash if there was limited
      stack space.
    - Avoid holding open old WAL segments in the walwriter process.
      The previous coding would prevent removal of no-longer-needed
      segments.
    - Fix log_line_prefix's %i escape, which could produce junk early in
      backend startup.
    - Prevent misinterpretation of partially-specified relation options
      for TOAST tables.
      In particular, fillfactor would be read as zero if any other
      reloption had been set for the table, leading to serious bloat.
    - Fix inheritance count tracking in "ALTER TABLE ... ADD CONSTRAINT"
    - Fix possible data corruption in "ALTER TABLE ... SET TABLESPACE"
      when archiving is enabled.
    - Allow "CREATE DATABASE" and "ALTER DATABASE ... SET TABLESPACE" to
      be interrupted by query-cancel.
    - Improve "CREATE INDEX"'s checking of whether proposed index
      expressions are immutable.
    - Fix "REASSIGN OWNED" to handle operator classes and families.
    - Fix possible core dump when comparing two empty tsquery values.
    - Fix LIKE's handling of patterns containing % followed by _.
      We've fixed this before, but there were still some
      incorrectly-handled cases.
    - Re-allow input of Julian dates prior to 0001-01-01 AD.
      Input such as 'J100000'::date worked before 8.4, but was
      unintentionally broken by added error-checking.
    - Fix PL/pgSQL to throw an error, not crash, if a cursor is closed
      within a FOR loop that is iterating over that cursor.
    - In PL/Python, defend against null pointer results from
      PyCObject_AsVoidPtr and PyCObject_FromVoidPtr.
    - In libpq, fix full SSL certificate verification for the case where
      both host and hostaddr are specified.
    - Make psql recognize "DISCARD ALL" as a command that should not be
      encased in a transaction block in autocommit-off mode.
    - Fix some issues in pg_dump's handling of SQL/MED objects.
      Notably, pg_dump would always fail if run by a non-superuser, which
      was not intended.
    - Improve pg_dump and pg_restore's handling of non-seekable archive
      files.
      This is important for proper functioning of parallel restore.
    - Improve parallel pg_restore's ability to cope with selective
      restore (-L option).
      The original code tended to fail if the -L file commanded a
      non-default restore ordering.
    - Fix ecpg to process data from RETURNING clauses correctly.
    - Fix some memory leaks in ecpg.
    - Improve "contrib/dblink"'s handling of tables containing dropped
      columns.
    - Fix connection leak after "duplicate connection name" errors in
      "contrib/dblink".
    - Fix "contrib/dblink" to handle connection names longer than 62
      bytes correctly.
    - Add hstore(text, text) function to "contrib/hstore".
      This function is the recommended substitute for the now-deprecated
      => operator. It was back-patched so that future-proofed code can be
      used with older server versions. Note that the patch will be
      effective only after "contrib/hstore" is installed or reinstalled
      in a particular database. Users might prefer to execute the "CREATE
      FUNCTION" command by hand, instead.
    - Update build infrastructure and documentation to reflect the source
      code repository's move from CVS to Git.
 -- Martin Pitt <email address hidden> Tue, 05 Oct 2010 22:05:37 +0200