Bugzilla

Hm, though I've just had an idea. I think the thing that is actually making this fail on windows is the following:

1. we hang on shutdown
2. the harness kills us after 5 seconds (cf. https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=292530855&repo=try&lineNumber=4293 ). That kills the launcher process but nothing else.
3. the harness restarts Firefox for the next run, with the same profile
4. we show the "but this profile is in use" error, because the previous process hasn't gone away yet and so the lockfile is in use -- this is what hangs the test forever.

So I imagine that as-is this won't ever reproduce on linux in the "the test is orange" sense, but it might reproduce in the "we hang on shutdown" sense, and if we could detect this and make the job error then that might make our job easier...

Hm, though I've just had an idea. I think the thing that is actually making this fail on windows is the following:

1. we hang on shutdown
2. the harness kills us after 5 seconds (cf. https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=292530855&repo=try&lineNumber=4293 ). That kills the launcher process but nothing else.
3. the harness restarts Firefox for the next run, with the same profile
4. we show the "but this profile is in use" error, because the previous process hasn't gone away yet and so the lockfile is in use -- this is what hangs the test forever.

So I imagine that as-is this won't ever reproduce on linux in the "the test is orange" sense (because there's no launcher process), but it might reproduce in the "we hang on shutdown" sense, and if we could detect this and make the job error then that might make our job easier...