rvokal (rvokal) wrote,

Bug reporting in Fedora and bug stats

Just few days ago Richard Marko send a notice to fedora-devel about updates in the bug reporting process for Fedora. To start from its origin ABRT server runs on Fedora now for 9 months and collects anonymous reports from Fedora users and generates nice stats about crashes. It can filter out components, distributions, hot issues (the top priority crashes within last couple days) etc. If you agree after your first crash on Fedora, the reports are uploaded automatically. ABRT uses micro-reports, a simplified backtraces with a hashtag. There are very small but still allow to connect duplicates (though this heurestics will never be perfect it has still a very high probability to find duplicates). For example the top issue for last couple days that many users experienced is a crash in libreoffice thumbnailer. You can see that it has been reported 35935 times! It's crashing on F18, F19 and in rawhide and on all archs (well, the two main ones where people are actually using libreoffice). Anyway, I find this very useful for developers. For years we were wondering how many users are actualy experiencing same crash, whether there is someone who has seen this crash with a different reproducer, with different inputs etc.

So how does reporting work now in Fedora?
Here's my quick summary of Richard's email and my observation. User experiences a crash and gets notified about it. Than he/she hits report button and that's where abrt creates the micro-report, sends it to the abrt server and waits for the reponse. At this point, the issue is either already known and user gets a link to the server where all the above described infomation can be found including a bug number if there is one already. The bug is created when at least one user who experienced the crash hits Report in gnome-abrt to provide aditional data. At this point abrt creates a full report, which includes the traceback, few stats about your system. All data can be reviewed before submitting and needs to be checked by user. After uploading these data, a bug gets created, linked to the abrt server report and maintainer can decide whether he'll fix the issue, talks to upstream about fixing that particular problem or provides a new version that fixes the problem.

What's new with the recent abrt server update is that if the number of crashes reaches certain treshold (originally 20 but I think Richard raised that number already) the bug gets opened even without additional data from user. This is a try to get atention from maintainers early enough to start investigating before even more users hit the same problem. On the other hand, maintainers get a very little data at this point and without a reporoducer and full backtrace, there's not much they can do with it. That said, if you report an issue, please fire up gnome-abrt, input the additional data, try to find reproducer, try to be the nice community member who helps others - developers, maintainer and users :-)

Thank you
  • Post a new comment


    default userpic

    Your reply will be screened

    Your IP address will be recorded 

  • 1 comment