Closed Bug 846225 Opened 7 years ago Closed 3 years ago
Run monkey/orangutan tests in automated infrastructure
Bug 832328 provides an orangutan harness to send random input gestures to content. This kind of infrastructure is like UI fuzzing, and turns up tons of valuable bugs. We should run this on automation. The problem to be solved here is - what's a pass/fail? Maybe pass means "X hours without 'problems'", see below. - what's a "problem"? Crashes are always problems. Hangs are too. It's a bit hard to define "hang" from within the orangutan harness, but we should be able to figure something out. Once we get these issues sorted out, we just need to install this code on infra.
I don't know that pass/fail is useful terminology here. I think you want this to act just like our existing fuzzing setups, where you simply get a report of crashes and other bad things that it turns up.
Crashes are definitely the first priority here. Next up would be assertions and then hangs. We should clear most of these out first then worry about the rest later (misc. UI issues, or incorrect JS error messages, maybe?) I know Christoph was trying out marionette stuff too via Peach, so cc'ing some other fuzzer folks.
We caught up briefly on IRC, but you can run the monkey tests with a script that's part of the b2g "distribution" checked out $ ./scripts/run-monkey.sh
Q: Is it already possible to get crash dumps on B2G other than monitoring logcat? If yes, where do I find those?
Yes, they're saved to /data/b2g/mozilla/Crash Reports/pending.
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.