Bugzilla – Bug 134604
gimp-2.2.8-6 crashes while loading image
Last modified: 2006-02-04 17:01:56 UTC
Loading a 1269x1753 greyscale image in any format (png, gif, bmp, jpg tested) results in gimp throwing core: (gimp:25682): GLib-GObject-CRITICAL **: g_object_unref: assertion `G_IS_OBJECT (object)' failed (script-fu:25685): LibGimpBase-WARNING **: script-fu: wire_read(): error [2]+ Segmentation fault (core dumped) gimp ttt.gif Apparently, this happens with all images available on this machine.
Created attachment 57822 [details] list of packages installed on the machine
I cannot reproduce and list of packages does not help for debugging. Could you install gimp-debuginfo package (in OpenSuSE mirrors, e.g. ftp://ftp.linux.cz/pub/linux/opensuse/distribution/SL-10.0-OSS/inst-source/suse/i586/gimp-debuginfo-2.2.8-6.i586.rpm) and then provide backtrace: gdb gimp run ... gimp crashes bt (or debug a core: gdb --core=core bt (Core is a name if your core dump file.) What gtk2 theme you are using. Does theme change help?
Sorry, backtrace is apparently endless and gives no sense IMHO. I did several (strace, ltrace) tracing attempts, but all I can see that there is some indefinite nesting in some library. Please close bug, I have lost interest.
This looks like http://bugzilla.gnome.org/show_bug.cgi?id=162778
stanislav, will your mmx workaround fix this as well?
Probably not. My work-around keeps 2.2.10 code intact, only compiles two affected files with -O0 and the work-around was applied to CVS after 2.2.10.
2.2.10, which is what we ship, has different code than that in the patch in bugzilla.gnome.org. It looks like the code changed and the ADDITION mode *is* #ifdef-ed out. I can't reproduce this as I have fairly new Intel Pentiums. Could we get someone to test this on an AMD machine? You have to use a grayscale JPEG (real grayscale, not RGB with only gray colors).
Klaus-J, do you think you could test this again with the gimp-2.2.10 RPM?
Not seeing it with 2.2.10 on amd opteron 64bit.