I made a seemingly unrelated change to the reftest harness, but it causes a segfault on shutdown (after calling nsIAppStartup.quit()) in indexedDB code: https://treeherder.mozilla.org/logviewer.html#?job_id=14176596&repo=try Scrolling down through the gecko output a little later on yields a little bit more information: > JavaScript Error: "IndexedDB UnknownErr: IDBDatabase.cpp:664" > JavaScript Error: "UnknownError: The operation failed for reasons unrelated to the database itself and not covered by any other error code." {file: "resource://gre/modules/IndexedDBHelper.jsm" line: 138} > I/GonkMemoryPressure( 748): Observed XPCOM shutdown. > F/libc ( 748): Fatal signal 11 (SIGSEGV) at 0x00000000 (code=1) This happens on all b2g emulator reftests with the patch from bug 1034290 applied. I have no idea how that patch has any affect on this, but it is 100% reproducible. It is blocking me from landing bug 1034290. Any help is much appreciated!

Actually, on further examination this isn't triggered by bug 1034290.. it happens in every b2g emulator reftest run as seen by this log from a green job on central: http://mozilla-releng-blobs.s3.amazonaws.com/blobs/mozilla-central/sha512/2df2cc1a237c768a632073b244b136ad6b7b7bee62f5ae814d395a20a8d4311dbdfb501ca044d4007fd4fff341a169888fb08358b7cf481c361eac09665acd7e (search for "Fatal signal 11") The difference is that prior to bug 1034290 this wasn't turning the jobs red.

Do you have a stack?

Nope. Just verified manually that no minidump gets generated (also why none show up in the logs). Could it be because reftest does a force quite? https://dxr.mozilla.org/mozilla-central/source/layout/tools/reftest/reftest.js#1389

I don't know why that would affect it since we're crashing anyways. Without a stack there's nothing we can do here. We don't even know if it's crashing in IndexedDB.

When I filed this bug, I thought it was turning my jobs red in bug 1034290. But upon further investigation, this was a red herring and the jobs were turning red for another reason. So if you aren't worried about it (given that it's on shutdown and emulator only), I'm happy WONTFIX'ing.

Well I would like to fix it because crashing is bad ... but that requires a stack.

I don't know what to do about that. Here's a log from Friday that proves crash detection is still working with b2g emulator reftest: https://treeherder.mozilla.org/logviewer.html#?job_id=17947028&repo=mozilla-inbound I guess it could be a bug in breakpad? Ted, do you have any ideas? There's (always) a shutdown crash with b2g emulator reftests after calling nsIAppStartup.quit(), though no minidump is generated. Crash detection still seems to be working in other cases, so I don't think it's a configuration issue.

Hm. My best guess would be that it's the content process crashing, and the chrome process is too far into shutdown to handle the crash properly. Do we get tombstone files in the emulator? You could try modifying the harness to pull any files from /data/tombstones and upload them: http://bytesthink.com/blog/?p=133 If there's any data there we could get a stack.