GNOME Bugzilla – Bug 141112
GIMP 2.0.1 segfault at startup under Solaris 2.8
Last modified: 2004-04-30 00:06:07 UTC
Using Solaris 2.8 build Gimp 2.0.1 using gtk-2.4 and disable the gimp-print stuff: ./configure --disable-print No problems encountered in the build. Upon startup it gets to the point where it is loading plug-ins and then segfaults. Here is a stack trace: Starting program: /usr/local/bin/gimp-2.0 Program received signal SIGSEGV, Segmentation fault. 0xfe8b3374 in strlen () from /usr/lib/libc.so.1 (gdb) where
+ Trace 46408
The stack trace doesn't give any useful information. The crash happens in an some code that is well tested and almost certainly correct. This is very likely a build problem. I don't know if anyone else tried GIMP with GTK+-2.4 on Solaris yet. Perhaps there's a problem on the GObject level? Do other applications using GTK+-2.4 work for you? Can you run the GObject test suite? What about the GTK+ tests?
Sorry about the stack trace. The instructions for reporting bugs said it would be nice to include one if you have a crash problem.... I'm also sorry if this is my fault because of something I did wrong during the build process, but I did not get any indication during the build that there might be a problem. I scrutinized the output from .configure pretty thoroughly but confess that I just typed 'make' and went off for a cup of coffee. GTK+-2.4 seems to work fine for me other than this. All the demos available when I run gtk-demo and tests in tests/testgtk seem to work OK. (At least they don't crash and *seem* to be working right. I don't really know if they are doing everything they are supposed to.) I'd be happy to run the GObject test suite if you can give me a hint where to find it.
In #3 of the stack trace, nick and blurb should both be NULL pointers. The fact that they are not suggests a compilation problem. A problem in GTK+ is possible but less likely. Could you specify what hardware you are running on, and what compiler you used?
GTK and GIMP were both compiled on a Sun Ultra 10 using gcc 2.8.1
Yeah. Either upgrading gcc or turning down the optimization (at least in the app/paint directory) would be reasonable things to try, I think. If you upgrade gcc, it would be best to recompile both gtk and gimp.
OK, here's what I did: Rebuilt the following packages, in this order: Gnu make-3.80 Gnu tar-1.13 gcc-3.4.0 pkgconfig-0.15.0 jpeg-6b tiff-v3.6.1 zlib-1.1.4 libpng-1.2.5 aalib-1.4.0 expat-1.95.7 libart_lgpl-2.3.16 render-0.8 xrender-0.8.3 libXft-2.1.6 freetype-2.1.8 fontconfig-2.2.94 gettext-0.14.1 libiconv-1.9.1 glib-2.4.0 atk-1.6.0 pango-1.4.0 gtk+-2.4.0 At this point I should have just tried to compile GIMP, but I decided to try to press on compiling all of the gnome stuff required to get gtkhtml2 to work.... libIDL-0.8.3 popt-1.9 ORBit2-2.10.0 At this point I hit a brick wall. So far I have not figured out how to get ORBit2 to compile. I'll figure it out later. I decided to just try compiling GIMP 2.0.1 without the help system. To get it to compile at all with gcc 3.4 I had to apply the following patch to gimp-2.0.1/app/paint-funcs/paint-funcs.h: diff -rup paint-funcs/paint-funcs.h paint-funcs-patched/paint-funcs.h --- paint-funcs/paint-funcs.h 2004-03-10 08:13:43.000000000 -0500 +++ paint-funcs-patched/paint-funcs.h 2004-04-28 14:10:19.790005000 -0400 @@ -350,13 +350,17 @@ void copy_component PixelRegion *dest, guint pixel); -void initial_region (PixelRegion *, PixelRegion *, - PixelRegion *, - guchar *, - gint, - GimpLayerModeEffects, - gboolean *, - CombinationMode); + +void +initial_region (PixelRegion *src, + PixelRegion *dest, + PixelRegion *mask, + guchar *data, + gint opacity, + GimpLayerModeEffects mode, + gboolean *affect, + InitialMode type); + void combine_regions (PixelRegion *, PixelRegion *, PixelRegion *, PixelRegion *, The resulting gimp-2.0 does not segfault (at least so far ;^) Steve
Congratulations. Let's hope it turns out to be worth the trouble.
Reopening and resolving with a proper resolution. Please don't close bugs with state NEEDINFO unless the reason is that no information was provided.
*** This bug has been marked as a duplicate of 140115 ***