388038 - crash in NS_CopyUnicodeToNative

Status: VERIFIED INCOMPLETE

(Core :: XPCOM, defect)

Description

[Robert Kaiser]

Attachment: stack trace

I'm not sure what my trunk build was doing, as I switches desktops to my chat window and when I went back it had crashed. Thankfully I'm using a debug build, so I could attach gdb and get a stack trace.

Build identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9a7pre) Gecko/200707111313 SeaMonkey/2.0a1pre

Comment 1

[Jesse Ruderman]

peterv, is this stack trace likely to be useful, or should we mark the bug as incomplete?

Comment 2

[Peter Van der Beken [:peterv]]

The stack trace looks usable, but we don't have enough to go on to figure this out.

Comment 3

[timeless]

I claim that frame 6 NS_CopyUnicodeToNative is bogus. frame 7 is calling foo::Release which could probably tail call to foo::~foo. I can't imagine any foo::~foo which would tail call NS_CopyUnicodeToNative.

Kairo, if you hit something like this, could you get locals for the crashing frame (windbg: |dv|, gdb: |info locals|), and also please provide hg rev ids when filing (so that people can try to figure out source lines)

Comment 4

[Robert Kaiser]

Sure, can provide hg revs when fining stuff like that in the future, but we still were on cvs back when this was filed.

timeless, I'm not sure how I could reproduce this, so this stack is all we have right now - if that's not enough to find an actual problem or if it looks too fishy, we should really close this down as incomplete.

I filed a few such bugs back then because I for myself am unable to tell anything from the stack and thought someone looking at the stack could ptotentially get a hold of a real problem, but if that's not the case on such a stack filed by me, feel free to close it down as INCOMPLETE.

Comment 5

[timeless]

i'm just offering nuggets of hopefully useful information (which someday should be distilled somewhere, but at least for now, hopefully you'll benefit from them).