451985 - Crash at every shutdown with completely blank stack trace

Status: RESOLVED FIXED

(Core :: Widget: Gtk, defect, P3)

Description

[dolphinling]

Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.1a2pre) Gecko/20080824021213 Minefield/3.1a2pre ID:20080824021213

Every time I quit firefox, I get the crash reporter coming up. This happens in a completely new profile, and with a completely new installation.

The stack trace is completely blank. There is a correct time, build id, any comments I entered, and a few other things, but no stack trace at all.

Example trace: http://crash-stats.mozilla.com/report/index/cb9e24de-7217-11dd-8257-001cc45a2ce4?p=1

Other crashes do give a stack trace, so it's not a problem with submission.

This is a regression from between 2008-07-13-02 and 2008-07-14-02.

Comment 1

[dolphinling]

http://hg.mozilla.org/mozilla-central/index.cgi/log/98d39d1a3246 regression range. (If there's a better way to do that with the hg changelog, sorry).

I don't see anything at all that looks suspicious.

Comment 2

[Boris Zbarsky [:bzbarsky]]

You can enter dates for a regression range at http://hg.mozilla.org/mozilla-central/pushloghtml

Do you have either content or chrome jit enabled?

Comment 3

[dolphinling]

Regression range: http://hg.mozilla.org/mozilla-central/pushloghtml?startdate=July+13+2008+00%3A00%3A00&enddate=July+14+2008+03%3A00%3A00

There is nothing in there that jumps out at me, though a few things that look possible.

Neither content nor chrome jit is enabled (and the problem started before they were even available, even).

Comment 4

[Boris Zbarsky [:bzbarsky]]

Hmm...  there are various gtk changes there....

Comment 5

[dolphinling]

I am planning to work on bisecting this. So far, I have verified that it does crash on my own builds as well as the official builds. The crash reporter does not appear (do I need to configure that in?) but I get the following terminal output:

./run-mozilla.sh: line 131: 19334 Segmentation fault      "$prog" ${1+"$@"}

Also, I should note that I see this on two systems, both Gentoo Linux, one x86 and one x86_64 with no 32-bit libraries (and therefore self-compiled 64-bit firefox).

Comment 6

[dolphinling]

Using hg bisect, I get

The first bad revision is:
changeset:   15923:8be867e8b341
user:        L. David Baron <[e-mail]>
date:        Sun Jul 13 20:20:27 2008 -0700
summary:     Don't leak the result of PR_LoadLibrary for the xinerama library.  (Bug 403706)  r+sr=roc

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

If you run firefox inside gdb (./firefox -g, then type "run"), what stack does gdb show you when it crashes (type "bt" at the "(gdb)" prompt)?

Comment 8

[dolphinling]

Done three times with the following three results:

#0  0xb23031b2 in ?? ()
#1  0xb74e54ea in snd_seq_delete_port () from /usr/lib/libasound.so.2
#2  0xb6b23800 in ?? ()
#3  0xafba6884 in ?? ()
#4  0xb6b3f100 in ?? ()
#5  0xb77f8ff4 in gdk_drag_get_selection () from /usr/lib/libgdk-x11-2.0.so.0
#6  0x00000004 in ?? ()
#7  0x00000001 in ?? ()
#8  0xbfa48a48 in ?? ()
#9  0xb77af859 in ?? () from /usr/lib/libgdk_pixbuf-2.0.so.0
#10 0xb6b23800 in ?? ()
#11 0xb75f3a54 in ?? () from /usr/lib/libX11.so.6
#12 0x00000000 in ?? ()

#0  0xafd041b2 in ?? ()
#1  0xb75364ea in _XimGetAttributeID () from /usr/lib/libX11.so.6
#2  0xb6b23800 in ?? ()
#3  0xafca4884 in ?? ()
#4  0xb6b3f100 in ?? ()
#5  0xb7849ff4 in ?? () from /usr/lib/libgtk-x11-2.0.so.0
#6  0x00000004 in ?? ()
#7  0x00000001 in ?? ()
#8  0xbff99f98 in ?? ()
#9  0xb7800859 in ?? () from /usr/lib/libatk-1.0.so.0
#10 0xb6b23800 in ?? ()
#11 0xb7644a54 in ?? () from /usr/lib/libglib-2.0.so.0
#12 0x00000000 in ?? ()

#0  0xb02e71b2 in ?? ()
#1  0xb75674ea in ?? ()
#2  0xb6b23800 in ?? ()
#3  0xb04a4884 in ?? ()
#4  0xb6b3f100 in ?? ()
#5  0xb787aff4 in ?? () from /usr/lib/libgdk-x11-2.0.so.0
#6  0x00000004 in ?? ()
#7  0x00000001 in ?? ()
#8  0xbfccacc8 in ?? ()
#9  0xb7831859 in gdk_gc_set_rgb_fg_color () from /usr/lib/libgdk-x11-2.0.so.0
#10 0xb6b23800 in ?? ()
#11 0xb7675a54 in g_intern_static_string () from /usr/lib/libglib-2.0.so.0
#12 0x00000000 in ?? ()

Comment 9

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

If there's some bad interaction with the xinerama lib unloading, can we just leak it anyway?  That seems like the simplest fix..

Comment 10

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Yeah.  That's fine with me, assuming it's been observed on released versions of software (i.e., more than gentoo-if-built-on-particular-day).

Comment 11

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I think the patch in bug 448512 should fix this.

Comment 12

[dolphinling]

Indeed, it did.

Thanks!