Bugzilla – Bug 3753
xfce-mcs-manager recieves SIGABRT on startup due to mouse_plugin.so from xfce-mcs-plugins
Last modified: 2009-07-14 08:31:05 UTC
You need to log in before you can comment on or make changes to this bug.
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en; rv:1.8.1.11) Gecko/20071208 Epiphany/2.20 Firefox/2.0.0.11 Build Identifier: When starting xfce-mcs-manager daemon, it comes up for a while - loads gtk theme etc, but then upon loading mouse mcs plugin it recieves SIGABRT. Reproducible: Always Steps to Reproduce: 1. Compile xfce-mcs-plugins and dependant software 2. Start xfce-mcs-manager 3. xfce-mcs-manager crashes Actual Results: xfce-mcs-manager recieved SIGABRT on startup most likely because of /usr/lib/xfce4/mcs-plugins/mouse_plugin.so from xfce-mcs-plugins package. See attachment for details. Expected Results: xfce-mcs-manager starts glib version: 1.2.10 gtk+ version: 1.2.10 While in 4.4.1 this worked ok, in 4.4.2 I can't get xfce-mcs-manager running. I've tried to recompile every package which is this dependant on (gentoo here), but with no luck. Also when removing /usr/lib/xfce4/mcs-plugins/mouse_plugin.so, mcs-manager starts with no problems. Bad thing is that I didn't find anyone to reproduce (asked some gentooists and it just runs fine for them), but for me it's always reproducible. Tell me if you need some further info, I will gladly try to provide it.
Created an attachment (id=1469) [details] Console output from xfce-mcs-manager --no-daemon-debug
Can you run in gdb and provide a backtrace?
Created an attachment (id=1470) [details] backtrace xfce-mcs-manager xfce-mcs-plugins and libxfce4mcs were compiled with --enable-debug configure option, -g CFLAG and were not stripped. If I need to compile more packages with some debugging options please let me know.
Sorry, this bt was with -fomit-frame-pointer I'll provide a new one in a minute
Jan: Do you use a tablet device? I have a report against the fedora packages that looks very much like this report. In that case disabling the wacom tablet they use gets things working. See: https://bugzilla.redhat.com/show_bug.cgi?id=425925 I've asked the reporter for a more detailed backtrace.
Created an attachment (id=1475) [details] more meaningful bt with debugging info
(In reply to comment #5) > Jan: Do you use a tablet device? I don't use tablet device at all, just common usb mouse.
One thing which comes on my mind: I use dual seat config, so I wonder if this mouse plugin doesn't touch something which it should not. I mean it touches something related to the mouse on the other seat.
Created an attachment (id=1477) [details] fix-bad-alloca.diff No, it's a bad use of g_alloca() which should be g_realloc() (current code both leaks memory and causes a crash if the device reports more than a certain number of buttons). Can you try this patch? Apply to the root of xfce-mcs-plugins with 'patch -p0 <fix-bad-alloca.diff' and rebuild and reinstall.
(In reply to comment #9) > Can you try this patch? Apply to the root of xfce-mcs-plugins with 'patch -p0 > <fix-bad-alloca.diff' and rebuild and reinstall. I've tried the patch, but I still recieve the same results (crash and same backtrace) as before.
Created an attachment (id=1478) [details] fix-bad-alloca-and-size.diff First of all: make sure you're installing your self-built package in the right location. By default, tarball installs go to /usr/local, while your distro's packages are probably set up to install to /usr. Anyhow, try this patch too (need to remove first patch with 'patch -p0 -R <fix-bad-alloc.diff' first). Looks like there's another potential problem with that function...
Wait, that patch isn't completely good either. Just a sec...
Created an attachment (id=1479) [details] fix-bad-alloca-and-size-v2.diff Ok, give this one a try. Applies to a clean 4.4.x tree.
Great, the second patch works. mcs-manager comes up and seems like everything is ok so far.
Great!
This patch fixes the aforementioned fedora bug as well. Thanks for the speedy patch! :)
Excellent, good to hear.