While working on bug 1544147 I noticed a race in the test harness: if a frame is crashed with BrowserTestUtils.crashFrame() we register an ipc:content-shutdown handler to remove the resulting crash report files. However SpecialPowers is already doing that so the two handlers will race.

The race is harmless most of the time but I suspect that it might be behind some obscure intermittent failures we've been experiencing.

Sorry, edited the wrong bug.

I've got a working WIP for this. As a side note the code in BrowserTestUtils.crashFrame() always waits for the crash to be recorded and parses the resulting .extra file to return the annotations to the caller. AFAIK there is only one test using this functionality which makes it sound like it's overkill, I'll file a bug to simplify it and make it optional.

It seems like only dom/ipc/tests/test_process_error.xhtml is using both BrowserTestUtils and SpecialPowers at the same time, all the other browser mochitests aren't so this problem is smaller than I thought. Still I find it odd that we have two separate mechanisms to cleanup crashes depending if the test is a plain or a browser mochitest.

The priority flag is not set for this bug.
:ahal, could you have a look please?

For more information, please visit auto_nag documentation.