retracer should set importance of bugs

[Brian Murray]

Binary package hint: apport

After marking a bug 'need-i386-retrace' I wait around for the bug to get retraced and then set the importance of the bug to medium if the retrace is useful. I believe dholbach and seb128 do the same. It would be nice if the retracer could set the importance for us. My criteria for the retrace being useful is quite low, I just make sure it isn't "StacktraceTop:?? ()".

[Martin Pitt (pitti)]

This is currently blocked by a bug in python-lauchpad-bugs:

from launchpadBugs.HTMLOperations import Bug
b = Bug('124354', cookie_file='/home/martin/txt/lp-apport.cookie')
b.set_status(importance='Medium')

(adapt the cookie_file for your system, of course). This does not actually change the importance of the bug.

[Martin Pitt (pitti)]

PEBCAK. apport is not in ubuntu-qa, thus cannot set importance, and I used a cookie for apport.

[Martin Pitt (pitti)]

Fixed in bzr head.

[Martin Pitt (pitti)]

apport (0.92) gutsy; urgency=low

  * bin/crash-digger: Do not crash if duplicate db is locked when attempting
    to consolidate it. This happens often because in the DC we have two
    parallel instances (for amd64 and i386).
  * Move ubuntu-fat-chroot from bin/ to debian/local/, since it is so heavily
    Ubuntu specific.
  * debian/local/ubuntu-fat-chroot: Use diversions for the binaries we want to
    disable, so that chroot upgrades do not trash the modifications.
  * debian/local/setup-apport-retracer: launchpad-crash-digger ->
    crash-digger.
  * bin/crash-digger: Add option -i/--arch-indep-dupcheck to explicitly enable
    duplicate checking of arch-independent crashes like Python exceptions. We
    only want to process them on one architecture to avoid scattering the
    duplicate database.
  * apport/crashdb_impl/launchpad.py, get_unfixed(): Search for 'apport-crash'
    tag, not 'apport'.
  * bin/apport-unpack: Fix format string in error message.
  * apport/ui.py, __init__(): Intercept ImportError, which can happen for
    crashes during system upgrades. (LP: #124354)
  * Add general-hooks/automatix.py: Refuse to send problem reports if
    automatix is installed.
  * doc/package-hooks.txt: Do not document UnsupportableReason, since it does
    not make sense to set it in package hooks (it is checked before calling
    the hooks). Hooks should use UnreportableReason only.
  * apport/ui.py, test_run_crash_package(): Check that 'Package' problem
    reports collect additional information, too.
  * apport/ui.py, collect_info(): Collect additional information for 'Package'
    problem reports, too.
  * Revive preloadlib/:
    - Remove PIPE_CORE #ifdefs and make them the default. We do not need to
      support the Edgy kernel patches in this version any more.
    - Install signal handler for SIGABRT, too.
    - Read core ulimit, pass it to apport in CORE_REAL_RLIM, and set it to
      zero for the program, since we do not actually want the kernel to write
      core files when we pipe the core dump to apport.
    - test-apport: Pass APPORT_REPORT_DIR to the manually called apport
      instance in the memory clipping test; otherwise it'll write into
      /var/crash/, which we do not consider in library mode.
  * apport/crashdb_impl/launchpad.py, __init__(): Only do the "download bug
    #2" hack if we actually have an authentication cookie. Thus, do it only on
    the retracing servers, not on the client side. (LP: #125142)
  * apport/report.py, crash_signature(): Generate a signature for one-line
    Python tracebacks, too. This sometimes seems to happen, e. g. LP#124588.
    (LP: #125020)
  * apport/crashdb_impl/launchpad.py, update(): Set bug importance to Medium
    if retracing was successful. (LP: #106379)

 -- Martin Pitt <email address hidden> Tue, 24 Jul 2007 21:50:34 +0200