GNOME Bugzilla – Bug 334408
Yelp crashes
Last modified: 2006-12-18 08:10:07 UTC
Distribution: Ubuntu 6.04 (dapper) Package: Yelp Severity: blocker Version: GNOME2.13.92 unspecified Gnome-Distributor: Ubuntu Synopsis: Yelp crashes Bugzilla-Product: Yelp Bugzilla-Component: general Bugzilla-Version: unspecified BugBuddy-GnomeVersion: 2.0 (2.13.90) Description: Description of the crash: Yelp crashes all the time. Steps to reproduce the crash: 1. Start yelp 2. Wait a few seconds. 3. Crash. Additional Information: This is on an amd64. Debugging Information: Backtrace was generated from '/usr/bin/yelp' (no debugging symbols found) Using host libthread_db library "/lib/libthread_db.so.1". (no debugging symbols found) [Thread debugging using libthread_db enabled] [New Thread 46912600610384 (LWP 13506)] [New Thread 1124096352 (LWP 13519)] [New Thread 1115703648 (LWP 13518)] [New Thread 1107310944 (LWP 13517)] [New Thread 1098918240 (LWP 13516)] [New Thread 1090525536 (LWP 13508)] [New Thread 1082132832 (LWP 13507)] 0x00002aaab00a80ca in waitpid () from /lib/libpthread.so.0
+ Trace 66892
Thread 1 (Thread 46912600610384 (LWP 13506))
------- Bug created by bug-buddy at 2006-03-13 10:34 -------
*** Bug 335058 has been marked as a duplicate of this bug. ***
*** Bug 335094 has been marked as a duplicate of this bug. ***
*** Bug 335118 has been marked as a duplicate of this bug. ***
Seems like this is only a 64bit issue, and occurs due to the a11y patch since the "children_changed::add" signal is showing up (see stacktrace in bug 335094)
Ubuntu bug about that: https://launchpad.net/distros/ubuntu/+source/yelp/+bug/34384
I would love to get a stacktrace with full debugging symbols for this, especially for at-spi and Bonobo/CORBA. I don't think Ubuntu has a package with debugging symbols for at-spi, so you will probably have to compile this by hand.
Confirming on the number of bug reports. CCing Bill Haneman as the stack traces seem to be in libatk-bridge. Maybe he can help here...
No, notice that accessibility_module_shutdown() has already been called here (it should not have been called in these threads as far as I can see). Looks like a mozilla/gecko problem to me.
To expand on that thought - note that in some threads, gnome_accessibility_module_shutdown() appears in the stack frames three times! That seems quite pathological, and in fact that method should never get called unless a thread is exiting (even then I am not sure it is correct). But as Brent implied, we can really do nothing without full stacktrace info here. We have no idea when signal is getting emitted, why shutdown is being called, or what the codepath is at all.
I don't know if this helps any, but I can reproduce this bug. This is my stack trace:
+ Trace 67120
Thread 1 (Thread 46912600544848 (LWP 8235))
bug 327668 has a full stacktrace for this.
Ok, bug 327668 is not the exact same, but it crashes in the same CORBA function
hmm, those two bugs don't look similar in their traces to me (at least in the interesting parts, which come well before the CORBA_free stuff).
Starting program: /usr/bin/yelp [Thread debugging using libthread_db enabled] [New Thread 46912602193040 (LWP 4405)] Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 46912602193040 (LWP 4405)] 0x00002aaaae4d0cc0 in ORBit_freekids_via_TypeCode_T (mem=0x7fffffe7ebb0, tc=0x0) at allocators.c:32 32 switch (tc->kind) {
+ Trace 67177
Thread 1 (Thread 46912602193040 (LWP 4405))
*** Bug 336385 has been marked as a duplicate of this bug. ***
please see https://launchpad.net/distros/ubuntu/+source/at-spi/+bug/37760
Please look at bug 331924.
Apologies for spam-- marking AP1 to reflect accessibility impact.
Apologies for spam... ensuring Sun a11y folks are cc'ed on all current accessibility bugs.
*** Bug 341629 has been marked as a duplicate of this bug. ***
*** Bug 343888 has been marked as a duplicate of this bug. ***
Bug 343888 has a teeny bit more info in the stack trace than this one. A question: why is this NEEDINFO - what other info needs to be added to this bug?
The bug was originally marked NEEDINFO due to the lack of good stacktrace (around comment #6). Since then, a full stacktrace was provided, but no-one reopened the bug. Now, the bug is in kind-of a limbo. The problem seems to have been tracked to the patch from bug #331924. (The way Ubuntu solved it was to revert that patch, at least). It isn't a yelp problem, but no-one (or at least, I) am unsure where the bug actually belongs (either mozilla, ORBit or CORBA) Reopening until someone can point to where the bug actually belongs (Anyone?)
*** Bug 344710 has been marked as a duplicate of this bug. ***
*** Bug 345433 has been marked as a duplicate of this bug. ***
*** Bug 348313 has been marked as a duplicate of this bug. ***
see bug #337239 - we are seeing this in Gentoo, and can reliably reproduce it in gnome apps if accessibility support is turned on and at-spi is installed. If we use the patch on #337239 ( that fixes the & problem on the CORBA_free line) , the problem is "fixed" for 64 bit, but produces a different error on 32 bit machines. If we don't use the patch, it seems to work just fine with 32 bit machines, but then you see the error on 64 bit machines. I can provide full backtraces with orbit / at-spi / whatever you want with full debugging symbols. Let me know if you need more information. Thanks.
If you can come up with a stack trace that is better than the one in comment #14, then yes, please provide it.
(In reply to comment #27) > see bug #337239 - we are seeing this in Gentoo, and can reliably reproduce it > in gnome apps if accessibility support is turned on and at-spi is installed. Thank you for this tip, deinstalling at-spi and libatspi1.0-0 has solved the problem for me. My environment : arch : AMD64 kernel : 2.6.16-1-amd64-k8 distro : Debian SID
I can't reproduce this. AMD64 Dapper 2.6.15-26-amd64-k8. Start: I don't have at-spi installed nor much else for accessibility (slow connection :]). Yelp runs fine. Install at-spi, logout and in; >$ yelp >GTK Accessibility Module initialized > >(yelp:10042): Gnome-WARNING **: Accessibility: failed to find module >'libgail-gnome' which is needed to make this application accessible No crash. Install libgail-gnome-module, logout and in; >$ yelp >GTK Accessibility Module initialized >Bonobo accessibility support initialized No crash.
Is this fixed by my patch in bug 337239?
(In reply to comment #31) > Is this fixed by my patch in bug 337239? > The patch did not apply cleanly to at-spi 1.7.10 here. Is it against cvs with the first patch already applied?
(In reply to comment #30) > I can't reproduce this. AMD64 Dapper 2.6.15-26-amd64-k8. Ubuntu added a patch to at-spi to hide the crash (after many bug reports to launchpad). This results in a fair-sized memory leak though: * debian/patches/revert-memory-leak-fix.patch: - reverted patch from http://bugzilla.gnome.org/show_bug.cgi?id=331924 - we rather live with small memory leaks than GNOME exploding - thanks for leading the way: <dhuggins@fastmail.us> (In reply to comment #31) > Is this fixed by my patch in bug 337239? > Unfortunately, I can't test as I have no access to a 64-bit machine and under 32-bit it seems to work okay. Anyone able to test this?
*** Bug 351051 has been marked as a duplicate of this bug. ***
Can anyone running recent (i.e. >= gnome-2.16.0) gnome still see this bug? Stack trace is still short some info, unless I missed somet6hing - in particular we need symbols from modules at-spi and gail.
We had no more complaints in Ubuntu about this.
i believe this bug was caused by at-spi on 64-bit systems. it has since been fixed. see the ubuntu bug.
Marking bug as fixed, please reopen if the problem still occurs.
*** Bug 383388 has been marked as a duplicate of this bug. ***