Indeed, a mere touch would not actually change the executable; a package
upgrade would change the inode, so they don't test the same thing.

** Description changed:

  According to Sebastian we often get crash reports for packages which
  just have been upgraded, but the old version of the executable is still
  running. (e. g. bug 983697). Apport already prevents filing a report
  when the executable changed between the time the crash happens and the
  time the user reports the bug, but this is not sufficient here: In that
  example, evince crashed right _after_ installing the new version (the
  running process got confused about the new data on disk presumably), so
  that check doesn't help.
  
  We need to check if the binary was modified after the process started.
  
  FIX: http://bazaar.launchpad.net/~apport-
  hackers/apport/trunk/revision/2296
  
  SRU TEST CASE:
-  - Start "evince &" in a terminal (or any other program, really)
-  - Pretend the package would get an upgrade, either by actually installing a 
newer package version, or doing "sudo touch /usr/bin/evince"
-  - Let it crash with "killall -SEGV evince"
-  - In the precise version, apport either displays the bug, or 
/var/log/apport.og gets an exception about self['ExecutablePath'] not existing 
any more.
-  - In this fixed version, apport.log just gets "executable was modified after 
program start, ignoring" and does not produce a report in /var/crash/.
+  - Install an older version of gnomine, e. g. the version from precise final.
+  - Start "gnomine &" in a terminal (or any other program, really)
+  - upgrade gnomine to precise-updates.
+  - Let it crash with "killall -SEGV gnomine"
+  - In the precise version, apport either displays the bug, or 
/var/log/apport.log gets an exception about self['ExecutablePath'] not existing 
any more.
+  - In this fixed version, apport.log just gets "executable was modified after 
program start, ignoring" and does not produce a report in /var/crash/.
  
  REGRESSION POTENTIAL: Low, the change has been tested in quantal for a
  while already, is covered by a test case, and the other dozens of test
  cases cover the "normal" case when a binary is older than the process
  start time.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/984944

Title:
  Reject crashes that happen right after upgrade

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/apport/+bug/984944/+subscriptions

-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to