KDE is now evaluating Sentry, a crash monitoring system.
Who can get entry? Everybody with a KDE developer account.
However what’s it?
Since perpetually we’ve got used Bugzilla to handle crash studies however this has quite a few challenges that haven’t made any enhancements in a minimum of 10 years:
- Discovering duplicates crashes is difficult and in our case entails a human discovering them
- When debug symbols are lacking we have to ask the consumer to recreate the issue, which isn’t at all times attainable
- Customers want to fret about debug symbols (that is partially improved by the rise of debuginfod – yay!)
- We now have no simply consumed graphs on how prevalent a selected crash is, and by extension we’ve got a tough time judging the significance
- The consumer wants to really write a report for us to study of the crash (spoiler: most crashes by no means get this far)
- …
All in all it’s a reasonably dissatisfactory scenario we’re in at present. Enter Sentry.
Sentry is a purpose-built crash monitoring system. It receives crash studies through API ingestion factors and traces lacking frames with the assistance of debuginfod, can detect duplicates mechanically and thus present us notably aggressive crashes, and far more. Finest but, it helps many various programming languages which permits us to not solely enhance the standard of our software program but in addition our infrastructure providers.
The present analysis occasion is already superb and helped repair quite a few issues, and the present setup just isn’t even utilizing all options but and we’ve got hampered rollout a bit: solely git builds at present submit information. If all goes effectively and we discover it to be superb I hope we’ll ultimately be capable of roll this out to manufacturing releases.
Let’s have a look at a crash I’ve mounted not too long ago.
Right here’s what Sentry acquired from the consumer:
Not terribly helpful. So with the facility of debuginfod it turned it into this:
I then utilized some mind energy to create a repair and consequently the crash has disappeared, as we are able to see on this neat graphic right here:
Right here’s a whole crash info web page from a latest infrastructure drawback in our bugzilla bot:
Additionally try my Akademy discuss:
