The exact rooting analysis is not able to tell that the existing Vector<Holder> is not safe. Let's just switch this over to AutoValueVector now. There is a comment in the source about this approach triggering crashes on Android: is there a bug number for that? Hopefully the problem is gone now: if not we should track it down and fix the underlying issue so we can use the correct rooting primitives here. https://tbpl.mozilla.org/?tree=Try&rev=87580a1fd5a1

Comment on attachment 821119 [details] [diff] [review] hazard_event_dispatch-v0.diff Review of attachment 821119 [details] [diff] [review]: ----------------------------------------------------------------- Are you sure this works? I wrote almost the exact same patch in bug 814026... And I filed bug 835826 on it.

You should review the patch that removes all of this code instead :-P

Thanks for the bug references! I'm sad we never figured out what was going on there. Opt only crashes make me think it was a compiler issue and I believe we've upgraded since then, even on B2G. In any case, try appears to be quite green now, so it should be good to go. (In reply to Kyle Huey [:khuey] (khuey@mozilla.com) from comment #2) > You should review the patch that removes all of this code instead :-P What is the timeframe on that?

(In reply to Terrence Cole [:terrence] from comment #3) > (In reply to Kyle Huey [:khuey] (khuey@mozilla.com) from comment #2) > > You should review the patch that removes all of this code instead :-P > > What is the timeframe on that? Whenever bent and peterv review bug 928312. Ideally early next cycle.

https://hg.mozilla.org/integration/mozilla-inbound/rev/2dc38d7aba6f Okay, that's not too long to wait if this push fails horribly. Hopefully this will just work, however.