apport stucks in a loop

Bug #61078 reported by magilus
2
Affects Status Importance Assigned to Milestone
apport (Ubuntu)
Fix Released
Medium
Martin Pitt
libapt-front (Ubuntu)
Confirmed
Undecided
Michael Vogt

Bug Description

I moved my computer. After connecting my computer to my network again, I found out that there were several apport processes running consuming all the cpu usage. I then wanted to use apt but it hung reading the package list. I then chrootet in my system and found out that apt-get segfaults. I solved that by removing /var/cache/apt/*.bin.

Apport should not consume the whole cpu when apt has got that problem. This may be caused by apt-index-watcher.

Revision history for this message
magilus (magilus) wrote :
Revision history for this message
magilus (magilus) wrote :
Revision history for this message
Martin Pitt (pitti) wrote :

The apt-index-watcher crash should be fixed in any case.

Changed in libapt-front:
assignee: nobody → mvo
status: Unconfirmed → Confirmed
Revision history for this message
Martin Pitt (pitti) wrote :

This is tricky to handle for apport. It will never create more than 5 reports for the same binary in 24 hours, and it runs with nice level 5, so it already has some boundaries. OTOH we do want reports about apt-get crashing.

I think I will further narrow down the number of reports that may happen in a given timeframe.

Changed in apport:
assignee: nobody → pitti
importance: Untriaged → Medium
status: Unconfirmed → Confirmed
Revision history for this message
Martin Pitt (pitti) wrote :

 apport (0.22) edgy; urgency=low
 .
   * apport_utils.py, report_add_proc_info(): Make 'interpreted script'
     detection more general to also work for mono programs.
   * test-apport: Check that non-packaged scripts do not generate a report.
   * apport: Call ic.collect_runtime_information() earlier and drop the local
     /proc/pid/exe examination, so that we get proper script detection. This
     avoids getting crash reports for non-packaged scripts (see test case
     change from above).
   * apport: Do not try to chmod the report file if we could not create it and
     output to stderr instead (this mainly affects local testing only).
   * apport_utils.py, find_file_package(): First grep the package lists whose
     names are a substring of the crashed binary name (or vice versa), to
     immensely speed up the package name determination in many cases.
   * apport: Drop the maximum number of consecutive crashes per executable
     from 5 to 2. 5 creates a too bad user experience and creates the
     impression that it will never stop. Closes: LP#61078

Changed in apport:
status: Confirmed → Fix Released
Revision history for this message
magilus (magilus) wrote :

great, thanks..

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.