! Please note that this is a snapshot of our old Bugzilla server, which is read only since May 29, 2020. Please go to gitlab.xfce.org for our new server !
crash when burning iso file
Status:
RESOLVED: FIXED

Comments

Description Lionel Le Folgoc 2009-08-16 19:07:07 CEST
Hi,

(bug originally reported at https://bugs.launchpad.net/bugs/410716)

"
Program crashed after it finished burning ISO file to disk.

xfburn crashed with SIGSEGV in g_closure_invoke()
StacktraceTop:check_burn_button (dialog=0x12eab10) at xfburn-burn-image-dialog.c:436
IA__g_closure_invoke (closure=0x18782d0,
signal_emit_unlocked_R (node=0x149cb80,
IA__g_signal_emit_valist (instance=0x1499800,
IA__g_signal_emit (instance=0x8, signal_id=0,
"

libburn4 0.6.4-1
libisofs6 0.6.16-1

Full stacktrace available at http://launchpadlibrarian.net/30056520/Stacktrace.txt

Thanks!
Comment 1 David Mohr 2009-08-16 23:07:13 CEST
Thanks for the report. Probably a bug introduced with the rewritten internals. I'll fix it very soon.
Comment 2 David Mohr 2009-08-18 06:43:35 CEST
Hmm, so what's that retracing service? The report on launchpad says it crashed after finishing the burn, but the retracing backtrace shows a location before the burning process was started. I'm confused...
Comment 3 Lionel Le Folgoc 2009-09-13 10:37:05 CEST
Created attachment 2537 
gdb log

I can confirm the crash: it occurs *after* the burning is successful (maybe when the "close" button is ungreyed out?). I attached a backtrace, but it seems to be very similar to the original one.
Comment 4 David Mohr 2009-09-17 02:30:17 CEST
Hm, still odd. I'll have a second look and hope to post a development version for you to test in a couple of days.
Comment 5 David Mohr 2009-09-25 00:56:10 CEST
*** Bug 5787 has been marked as a duplicate of this bug. ***
Comment 6 David Mohr 2009-09-26 05:20:54 CEST
One more question on this one: Is hal compiled in? Is hal running in the system, or is it already using devicekit?

Thanks.
Comment 7 Lionel Le Folgoc 2009-09-26 11:17:58 CEST
Yes, hal support is enabled (http://launchpadlibrarian.net/32357584/buildlog_ubuntu-karmic-amd64.xfburn_0.4.2-1ubuntu1_FULLYBUILT.txt.gz for a build log), and hal is running.
Although all gnome components in ubuntu are already using devicekit, I don't use many of them, so devicekit-disks package is installed, but not running (at least I can't see it with ps).
Comment 8 David Mohr 2009-10-13 03:39:25 CEST
Ok, this bug was most likely caused the introduction of the device list object. I believe it has been fixed in master. Please test and let me know if it works, thanks!

commit 0dffd44e19b60ab395e5353497de53d0dedf3c3b
Author: David Mohr <squisher@xfce.org>
Date:   Mon Oct 12 21:36:53 2009 -0600

    Disconnect from the volume change signal of the device box
Comment 9 Lionel Le Folgoc 2009-10-13 08:52:51 CEST
Created attachment 2602 
gdb log

Thanks for the patch. Unfortunately, I can still reproduce this issue (and the trace seems to be the same).
Comment 10 David Mohr 2009-10-14 05:22:33 CEST
Thanks for testing!

I'm not sure what the issue is now. Let's see if some debugging messages can't point to what's going on. Could you please:
- check out the branch 'burnimage' from git,
- compile with --enable-debug=full, 
- and send me the console output

You should be getting the same crash as before, since there was only one minor change aside from the debugging messages.
Comment 11 Lionel Le Folgoc 2009-10-18 09:55:31 CEST
(In reply to comment #10)
> Thanks for testing!
> 
> I'm not sure what the issue is now. Let's see if some debugging messages can't
> point to what's going on. Could you please:
> - check out the branch 'burnimage' from git,
> - compile with --enable-debug=full, 
> - and send me the console output
> 
> You should be getting the same crash as before, since there was only one minor
> change aside from the debugging messages.

Now the burn process always fails... I can see "DBG[xfburn-burn-image-dialog.c:241] xfburn_burn_image_dialog_finalize(): bug: Disconnected signal handler" being displayed at the very beginning of the burning process.
Comment 12 Lionel Le Folgoc 2009-10-18 09:56:37 CEST
Created attachment 2611 
GDB log with the 'burnimage' branch.
Comment 13 Lionel Le Folgoc 2009-10-18 09:59:52 CEST
Created attachment 2612 
hacky patch

Moreover, it wouldn't compile in full debug mode on my system without this patch (fix old-style declarations, mixed declarations and code, and variables shadowing). But I may have broken something when fixing that, so maybe that's why the burning failed... :/
Comment 14 David Mohr 2009-10-20 07:59:38 CEST
(In reply to comment #13)
> Created an attachment (id=2612) [details]
> hacky patch
> 
> Moreover, it wouldn't compile in full debug mode on my system without this
> patch (fix old-style declarations, mixed declarations and code, and variables
> shadowing). But I may have broken something when fixing that, so maybe that's
> why the burning failed... :/

Ah, yes, xdt 4.7 entered Debian. After an update, I had the same errors, and fixed them.

Thanks for going through the trouble of patching it yourself! It is perfectly ok to let me know about these issues without providing one :-). Since you said it crashed, I ended up going through the errors myself, since the context is important. Can you try the burnimage branch again please?
Comment 15 Lionel Le Folgoc 2009-10-20 09:30:01 CEST
Created attachment 2621 
new gdb log

Thanks, it builds fine now. ;)

Unfortunately, I won't be able to test with a real cdr for some days, but I think I managed to reproduce without burning: I inserted a random (full) cd, started xfburn, opened the 'burn image' dialog, which showed that the disc was full (as expected), hit cancel, and ejected the disc without closing xfburn => crash, 100% reproducible here.

Maybe it's not the same bug, but backtraces are very similar.
Comment 16 David Mohr 2009-10-20 17:02:27 CEST
Thanks for the new backtrace and console output. It looks useful, I'll get back to you soon with some new code to try.
Comment 17 David Mohr 2009-10-24 03:07:41 CEST
Yup, I'm pretty sure that it's the same bug.

It turns out I actually had a typo in my fix before, which is why the bug remained. I was able to reproduce the issue, thanks for finding the second way to show the bug, and at least for me it's gone now.

Please test the burnimage branch again:
commit 18e1859ba86c33ae03231f70e93aaa757f2e96ed
Author: David Mohr <squisher@xfce.org>
Date:   Fri Oct 23 20:58:31 2009 -0600

    Keep the devlist around, disconnect from the propper object
Comment 18 Lionel Le Folgoc 2009-10-24 10:19:35 CEST
Yeah, it's gone for me too, thanks!
Comment 19 David Mohr 2009-10-24 16:55:30 CEST
Great. If anyone else still has this issue, please reopen this bug.
Comment 20 David Mohr 2009-10-24 19:15:53 CEST
Just for completeness, the fix is now merged into master.

commit	1a88317ebcf26f7854c0b7a107ffa3f410961505
Comment 21 Christoph Wickert editbugs 2009-10-24 19:44:25 CEST
Is this all I need to fix the issue? I don't want to update to git in the stable Fedora releases, so I prefer cherry picking.
Comment 22 David Mohr 2009-10-24 21:24:10 CEST
No, unfortunately the fix is spread over a couple of commits.

It should work by applying the following ones, in reverse order:

84a6ff9475375b0eeee9c71e091e9027dc22a99a
18e1859ba86c33ae03231f70e93aaa757f2e96ed
cfd9906c051c63cb9e1fac6cd447d45c15b563f4
0dffd44e19b60ab395e5353497de53d0dedf3c3b

I wanted to try if this really does work, and found git format-patch. If you use $ git format-patch 0dffd44e19b60ab395e5353497de53d0dedf3c3b
on master, it should generate ~10 patches. Delete the ones you don't want to apply, i.e. i10n updates, and the ones which fix the compiler warnings (they don't apply cleanly against 0.4.2). You should be left with something like

0001-Disconnect-from-the-volume-change-signal-of-the-devi.patch
0002-Add-debugging-messages-chain-up-finalize-call.patch
0009-Keep-the-devlist-around-disconnect-from-the-propper-.patch
0010-Remove-debugging-messages.patch

Then apply with 
$ git am *.patch

I hope that helps, please let me know if this worked, or if you ended up using some other way to create the patch (trying to learn about git here :-).
Comment 23 Christoph Wickert editbugs 2009-10-29 23:23:32 CET
With 
$ git format-patch 0dffd44e19b60ab395e5353497de53d0dedf3c3b
I did not get the 0001-Disconnect-from-the-volume-change-signal-of-the-devi.patch, so i looked up the parent at http://git.xfce.org/apps/xfburn/commit/?id=0dffd44e19b60ab395e5353497de53d0dedf3c3b and used 
$ git format-patch 9415787601c953ae9db39365cc01185aaa72a5c5
instead. I applied the 4 patches you mentioned on xfburn-0.4.2. There were a couple of hunks and rpmbuild doesn't accept fuzzy patches, so I diffed against 0.4.2 again to get a single, non-fuzzy patch.

For me the crash is fixed, however I'm still waiting for feedback from the people who reported the bug in Red Hat's bugzilla (rhbz #525518).
Comment 24 David Mohr 2009-10-30 00:18:06 CET
Thanks for cooking up a working patch. If it's not too much trouble, report back here if you get feedback on the RH bug.

Bug #5678

Reported by:
Lionel Le Folgoc
Reported on: 2009-08-16
Last modified on: 2009-10-30
Duplicates (1):
  • 5787 Crash detected in Xfburn 0.4.2

People

Assignee:
Xfburn Bug Triage
CC List:
2 users

Version

Attachments

gdb log (23.20 KB, text/plain)
2009-09-13 10:37 CEST , Lionel Le Folgoc
no flags
gdb log (9.96 KB, patch)
2009-10-13 08:52 CEST , Lionel Le Folgoc
no flags
GDB log with the 'burnimage' branch. (18.37 KB, text/plain)
2009-10-18 09:56 CEST , Lionel Le Folgoc
no flags
hacky patch (28.37 KB, patch)
2009-10-18 09:59 CEST , Lionel Le Folgoc
no flags
new gdb log (15.83 KB, text/plain)
2009-10-20 09:30 CEST , Lionel Le Folgoc
no flags

Additional information