Stopping the Firefox Window Recorder takes over 30 seconds
Categories
(Testing :: Raptor, defect, P2)
Tracking
(Not tracked)
People
(Reporter: sparky, Unassigned)
References
(Blocks 1 open bug)
Details
In our CI tests, it takes a very long amount of time to stop the Firefox window recorder (~30s). This results in about 25 extra minutes of testing time because each browser iteration stops the recorder twice resulting in 1 minute (or more) extra per browser iteration.
See this test for example which takes 51 minutes to complete: https://treeherder.mozilla.org/jobs?repo=autoland&group_state=expanded&selectedTaskRun=B1pj7XuXSmKTsdWoWJOWNw.0&resultStatus=pending%2Crunning%2Csuccess%2Ctestfailed%2Cbusted%2Cexception%2Crunnable&searchStr=imgur%2Cbrowsertime&revision=39fecab5c852d2ed3bd4ef452b3c063863766327
On windows, with the FFMPEG recorder, it takes about 1-2 seconds to stop recording.
Reporter | ||
Comment 1•2 years ago
|
||
This might not be valid. I think the logs there are misleading because they don't print anything when we do this wait: https://searchfox.org/mozilla-central/source/testing/raptor/browsertime/browsertime_pageload.js#401-405
We're going to undo those changes because they caused a lot of variance in our measurements, and I think they're part of the source of this runtime issue: https://bugzilla.mozilla.org/show_bug.cgi?id=1773629
I have a new try run here that adds some logging so we can see: https://treeherder.mozilla.org/jobs?repo=try&revision=0a4835180a9ad201d6562d010203e34beee53400
Reporter | ||
Comment 2•2 years ago
|
||
Closing as invalid since it looks like the wait above the majority of the issue. I've seen a small handful of cases where stopping is longer than usual but it's not as bad as I originally thought.
Description
•