Bug 1439624, bug 1440884, bug 1444557, bug 1450712, bug 1450838, bug 1456349, bug 1458743, bug 1471446, bug 1485737, bug 1496693, bug 1516284, bug 1516457, bug 1516971, bug 1526689, bug 1530385, bug 1539041. There are all intermittent failures of test_profile_worker.html or test_profile_worker_bug_1428076.html that have been happening sporadically.

With bug 1529774, it seems things are blowing up more reliably, and the tests are perma-failing (with very variable crash locations).

I've confirmed that things don't go bad without the profiler enabled in those tests, but since they are profiler tests, they are not useful as tests anymore when this is done.

It could be a range of things, from a kernel bug (the emulator images we use for tests are outdated and use a kernel version that is not even available on actual devices using the lowest version of android we support), to a bug in the profiler signal handler itself, etc. It could be related to bug 1007824.

Old kernels being involved, it might not even be an actual issue on real hardware, but the fact is they currently fail on automation. Well, until I disable them to unblock bug 1529774.

This bug is for the underlying issue (rather than disabling the tests, which I'll do in bug 1529774).

This is a bit weird that this was happening in chrome tests only, and not browser tests.

Is this possible that this is the "console.log" of the captured profile that breaks this?