Hi Lee, Thanks for your help with debugging this problem Please CC: [EMAIL PROTECTED] always when we discuss this bug. That way people with better knowledge are able to follow our discussion and give some help. I dscussed this bug with Guido Günther <[EMAIL PROTECTED]>. Please Cc him, so can tell us his opinion. I tried to set a header on the bug to send him all messages going to the bug address, but I don't know if this works correctly.
On Fri, Apr 08, 2005 at 08:59:16AM +0100, Lee Braiden wrote: > Hi Wolfram, > > On Thursday 07 April 2005 18:17, you wrote: > > Hm, can you send me the full output of "strace inkscape" and "ltrace > > inkscape"? Perhaps also the core file, but for that it is better to > > compile inkscape wit debugging symbols. Do you want to do this yourself > > or should I provide a suitable package? > > I've attached an strace, and a core file from the sid package built with > debugging. Thanks for this. If I try to get a bt from your core file, I get [EMAIL PROTECTED]:/tmp $ gdb -c inkscape_ppc_segfault.core /usr/bin/inkscape GNU gdb 6.3-debian [...snip...] Core was generated by `inkscape'. Program terminated with signal 11, Segmentation fault. #0 0x30009da0 in ?? () (gdb) bt #0 0x30009da0 in ?? () #1 0x30009e7c in ?? () Previous frame identical to this frame (corrupt stack?) From your report I assume you get the same? > ltrace outputs NOTHING, however, and never finishes, even though > I left it running all night. It seems to work fine for other programs. I > tried running it with a higher debug level, but just got endless lines like > the following, which seemed useless :( > > Here's what I *do* get, anyway... > > $ ltrace -d inkscape > DEBUG: read_config_file.c:172: read_config_file(): Reading config file > `/etc/ltrace.conf'... > DEBUG: read_config_file.c:172: read_config_file(): Reading config file > `/home/jel/.ltrace.conf'... > DEBUG: elf.c:32: do_init_elf(): Reading ELF from /usr/bin/inkscape... > DEBUG: elf.c:216: do_init_elf(): /usr/bin/inkscape 1392 PLT relocations > DEBUG: execute_program.c:75: execute_program(): Executing > `/usr/bin/inkscape'... > DEBUG: execute_program.c:89: execute_program(): PID=6746 > DEBUG: process_event.c:70: process_event(): event: none > ... This is a very strange behaviour. Do you use prelink on your system? It is known to have some issues on ppc. If you use it, can you please try to "unprelink" inkscape and check if it works then? > > > Is your system plain unstable > > or do you have some libs from experimental installed? I got it > > right that you start inkscape without any parameters? Which window > > manager do you use and which terminal emulator? > > Running sid, but with Michel Dänzer's DRI trunk X Server packages > (http://people.debian.org/~daenzer/dri-trunk-unstable/), mplayer > (http://honk.physik.uni-konstanz.de/~agx/linux-ppc/debian/), KDE 3.4 > installed from pre-sid packages from the maintainers (and now from alioth). > I generally use KDE's default stuff: kwin, Konsole, etc. And libexpat is the current package from unstable, right? What happens if you uninstall libexpat1-dev? libexpat is the lib which was loaded shortly before the segfault. However, inkscape does not depend on this library and it is not linked against it (ldd /usr/bin/inkscape doesn't show libexpat here and I can run inkscape without having it installed). > > However, inkscape crashes just the same in a GNOME session (started via kdm), > and I'm fairly certain it worked in this xserver before, so I'm thinking > that's not the problem. I could try installing the standard sid X server and > running gnome via gdm, if you want to be sure. No, the way you start your session shouldn't be relevant here. > > I've also tried renaming the .inkscape directory so that inkscape would be > starting with a default config, but it didn't help. > > > I have no real idea right know, especially since inkscape quits without > > saying anything. Normally it catches sigsegv, does an emergency save and > > asks you to inform the developers. > > Maybe it's not much help either way, but doesn't this narrow the segfault > down > to somewhere before those safeguards are set up? Yes, I guess so, or to a library used by inkscape. Another thing which came to my mind is, do you have CONFIG_PREEMPT=y in your kernel configuration? > > -- > Lee. With best wishes, Wolfi
signature.asc
Description: Digital signature
