The exception handler should handle DebugBreak exceptions: we use these in release code (mozalloc_abort): see bug 624809 for STR. Ted says we should be calling ExceptionHandler::set_handle_debug_exceptions(true).

I believe this should block, we might be hiding intentional aborts which affect our perceived crashiness.

This appears to work, but I don't have a test. Ted, do you think that matters?

Comment on attachment 502931 [details] [diff] [review] Catch ::DebugBreak, rev. 1 You could add an xpcshell test, you'd have to add to the C++ component to call DebugBreak though, probably.

[hardblocker] or [softblocker], bsmedberg?

(In reply to comment #3) > (From update of attachment 502931 [details] [diff] [review]) > You could add an xpcshell test, you'd have to add to the C++ component to call > DebugBreak though, probably. ctypes?

http://hg.mozilla.org/mozilla-central/rev/dab2f429fe20 (tests) http://hg.mozilla.org/mozilla-central/rev/0e44149ece98 (patch) I had to add Windows #ifdefs because set_handle_debug_exceptions is Windows-only.

(In reply to comment #6) > http://hg.mozilla.org/mozilla-central/rev/dab2f429fe20 (tests) > http://hg.mozilla.org/mozilla-central/rev/0e44149ece98 (patch) Ftr, the former is patch and the latter is test. *** +++ b/toolkit/crashreporter/test/unit/test_crash_runtimeabort.js + dump("INFO | test_crash_purevirtual.js | Can't test crashreporter in a non-libxul build.\n"); Nit: s/test_crash_purevirtual.js/test_crash_runtimeabort.js/