apport 0.45 source package in Ubuntu

Changelog

apport (0.45) feisty; urgency=low

  New feature: Infrastructure for package install/upgrade failures:

  * Add bin/package_hook: Script for creating a report for a package
    installation/upgrade failure. It receives a package name, a number of log
    files, and an ErrorMessage: from stdin. This will be called from e.g.
    dist-upgrader.
  * setup.py, debian/apport.install: Ship package_hook.
  * apport/ui.py: If ProblemType is 'Package', call a new function
    self.ui_present_package_error() instead of presenting a crash. Add test
    suite checks for the package error report workflow.
  * apport/ui.py, create_crash_bug_title(): Create default bug title for
    package reports. Add various test cases.
  * gtk/apport-gtk{,.glade}: GTK implementation of ui_present_package_error().

  New feature: Maintain a per-binary blacklist to inhibit apport crash reports
  until the binary changes. Closes: LP#79408

  * apport/report.py: Add new Report methods check_ignored() and mark_ignore()
    to check for/set ignore list entries. Add test cases.
  * apport/ui.py: Add another return value of ui_present_crash() to specify
    whether or not to blacklist the current crash's executable. Check workflow
    of both responses in the test suite.
  * gtk/apport-gtk{,.glade}: Add a blacklist checkbox to the crash
    notification dialogs.
  * bin/apport: Do nothing if the current crash is blacklisted.
  * test-apport: Test blacklisting.

  Bug fixes:

  * gtk/apport-gtk: Fix return code for restarting the application ('reopen' ->
    'restart'). Closes: LP#81422
  * test-apport: Adapt to new core_pattern kernel interface mode:
    - Check core_pattern instead of the obsolete crashdump sysctl to determine
      whether or not apport is running.
    - Give apport max. 10 seconds to complete. The current kernel reaps the
      crashed process as soon as writing the core dump to the pipe is
      finished, but apport still needs to write the report file.
    - Do not EXFAIL the test for crashes in nonwriteable cwd any more, since
      it is now supposed to work (we do not write a core dump to the disk any
      more).
  * run-tests, use-local: Adapt to new core_pattern kernel interface.
  * apport: Improve logging of exceptions, include environment variables.
  * apport/report.py test suite: Use gdb to generate a test core dump, do not
    rely on kill(SIGSEGV) and the kernel to do it (since we now use a pipe in
    core_pattern).
  * backends/packaging-dpkg.py: Fix return value of get_modified_files() if
    dpkg .list file is missing.
  * apport/report.py, add_package_info(): Do not produce stray empty lines for
    uninstalled alternative dependencies.
  * apport/report.py: Fix test_add_gdb_info_script() to not leave behind a
    stray gzip process which randomly blocks stdin. Closes: LP#78421
  * backends/packaging-dpkg.py: Do not read the dpkg status in the
    constructor, but lazily initialize it when actually calling a query
    function. This avoids imposing the dpkg-query overhead for programs that
    import the apport package without doing package queries (such as any
    Python program under Ubuntu, due to the Python crash hook).
  * apport/ui.py, create_crash_bug_title():
    - Do not crash on an empty StacktraceTop. Closes: LP#81677
    - Do not mention an unknown function name ('??') in the bug title;
      instead, use the topmost function with a known name, or leave it out
      at all.
    - Add test cases for these situations.
  * apport/report.py, _get_ignore_dom(): Do not throw an error for an empty
    ignore list file.

  Code cleanups:

  * apport/report.py test suite: Refactorize generation of test crash program
    and core dump generation.
  * Consistently use 'in'/'not in' instead of find() for substring searches.
  * Changed the packaging backend import, so that its methods can now be
    accessed at apport.packaging instead of apport.packging.impl.

 -- Martin Pitt <email address hidden>   Sun, 28 Jan 2007 12:34:05 +0100