Created attachment 392911 [details] [diff] [review] v1 nsGlobalModalWindow traverses mArguments, which is already traversed in its base class (nsGlobalWindow). This means the cycle collector might be collecting live objects. It doesn't traverse its own member mReturnValue but does unlink it. I found this from code inspection and I don't have a testcase, so no idea how bad it is. We should be nulling out pointers when unlinking so I don't think we'll end up with stale pointers.
Peter: do you think this should block the alpha? That's what P1 blockers mean at this time ...
Created attachment 415604 [details] [diff] [review] v1 (1.9.1 branch version) This fixes the leaks from bug 504862. Safe patch, has been on trunk since beginning of August.
Comment on attachment 415604 [details] [diff] [review] v1 (1.9.1 branch version) Approved for 18.104.22.168 and 22.214.171.124, a=dveditz for release-drivers Please land this before bug 504862