apport 1.91-0ubuntu1 source package in Ubuntu

Changelog

apport (1.91-0ubuntu1) precise; urgency=low

  * New upstream release:
    - crashdb.py, check_duplicate(): If a crash has a signature but no
      existing duplicate in the DB, also check for an existing address
      signature duplicate in the DB.
    - apport-retrace: Use DistroRelease specific subdirectory of the cache dir
      for mapping a file to a package, as these maps are release specific.
    - packaging-apt-dpkg.py: Refresh Contents.gz cache if it is older than one
      day.
    - crashdb.py: Ensure that address_signature duplicate db table does not
      have multiple identical signatures by making it a primary key. Bump the
      db format to "3". Existing databases need to be migrated manually as
      SQLite does not allow adding a "PRIMARY KEY" constraint to existing
      tables.
    - crashdb.py: Do not add a new address signature entry if one already
      exists.
    - apport-cli: Fix UnicodeDecodeError on unicode report values.
      (LP: #275972)
    - launchpad.py: Only set bug task importance if it is undecided.
    - apport-retrace: Fix "an useful" typo. (LP: #911437)
    - report.py: Filter out frames which are internal kernel/glibc
      implementation details and not stable across duplicates. In particular,
      filter out __kernel-syscall() and the SSE stubs.
    - crashdb.py: Remove debugging leftover which completely disabled bug
      pattern checking.
    - report.py: Update reading AssertionMessage. Current (e)glibc turned
      __abort_msg from a simple static string into a struct.
    - Change permissions of .crash files from 0600 to 0640, so that /var/crash
      can be made g+s and crash handling daemons can access those.
    - Python exceptions: Blacklist DBus.Error.NoReply. It does not help to get
      these traces from the client-side application, you need the actual
      exception in the D-Bus server backend instead. (LP: #914220)
    - Support /etc/apport/whitelist.d/ similarly to /etc/apport/blacklist.d/,
      for cases like installer environments where only crashes of a few selected
      programs should be reported.
 -- Martin Pitt <email address hidden>   Wed, 18 Jan 2012 09:56:00 +0100