The quest of automating crash handling at Unity
summary
Unity has >1M registered users and gets ~6,000 bug reports/month so a custom crash pipeline is built because existing systems like Sakura and Breakpad prove impractical for Mono/C# stacks. The system parses logs, forms a gist (stripped call stack), and buckets reports; 8,000 input reports map to ~20,000 unique crashes and ~4,000 buckets. Slack alerts, duplicate collapsing and version metrics give immediate visibility; bucketing cuts duplicate triage but still needs manual split/ML for tricky driver/environment cases.
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:08Toolsmiths make sure things run smoothly
- 1:26Reporting bugs in Unity
- 2:16When the old tricks stopped working
- 4:04Automation to the rescue!
- 4:28Socorro: Mozilla's Backend plus Reporting UI
- 5:44Breakpad: Google's cross platform crash collecting
- 8:46Full stack matching vs top frame matching
- 10:08Middle ground the notion of GIST
- 12:13Easy navigation to/from FogBugz
- 14:09The notion of the duplicate and repro
- 15:22Other things to consider or memo to ourselves
- 19:29Conclusions & takeaways
related talks
30mHarnessing the Unreal Engine Automation Framework for Performance Measurement
17mHow to crash your app in production and still delight your users
55mSo You're Trying to Support a Medium Sized Studio....
19mHow to handle incidents at scale
57mThe Curious Incident in your Software in the Day Time
39mOops!… We F*ed It Up Again: The Art of Staying Calm When Systems Crash
14mWhat’s new for Firebase LiveOps
44mEfficient Android app monitoring