From bug 508128, comment 84: >I talked to jst about this, and he sounded ok with just making the reference counting threadsafe in this case. > >http://hg.mozilla.org/mozilla-central/file/3a5b84a290d7/xpcom/base/nsExceptionService.cpp#l95 >already documents that it can be released on a different thread during shutdown.

This patch makes the failures in bug 508128 disappear.

Comment on attachment 485418 [details] [diff] [review] Make nsExceptionService practice threadsafe refcounting. This is needed to land bug 508128, which is a righteous goal.

Isn't this the same problem that we ran into when trying to move WeaveCrypto off the main thread? The real problem is that nsExeceptionService shuts down too early. This might make the asserts go away, but what happens if we trigger an exception after the nsExceptionService is dead?

Comment on attachment 485418 [details] [diff] [review] Make nsExceptionService practice threadsafe refcounting. Looks like this was never landed. It can't land now.

Is this about nsExceptionManager or nsExceptionService?

Hmm, I never ended up deciding whether this patch or the one in bug 450812 would be the better choice. I guess we can always back out in the future.

(In reply to comment #8) > Hmm, I never ended up deciding whether this patch or the one in bug 450812 > would be the better choice. I guess we can always back out in the future. Yes, let me know if you want me to backout, please, and sorry for stepping on your toes.