GNOME Bugzilla – Bug 98308
freeze and segfault when drawing from inside to outside window with pen tablet
Last modified: 2004-12-22 21:47:04 UTC
gimp cvs as of nov 12 wm is metacity 2.4.1 (debian) wacom intuos pen Does not occur when using mouse to draw. freeze: 1. create new image of medium size (should fit entirely on screen) 2. ctrl-e (image is entirely within window bounds) 3. select paintbrush 4. drag a stroke with the pen from inside the image out to the desktop. result: gimp freezes. I tried this with most of the drawing tools (including smudge and convolve) and got the same result. Stack trace after hitting ctrl-c in gdb below. segfault: 1. create a new image 2. ctrl-e, but then resize the window smaller than the image it is showing 3. select paintbrush 4. draw a stroke with the pen from inside the image off result: segfault. trace attached. both traces end in mallopt in libc, every time i repeat them. 600x600 paintbrush drag from on canvas to out of window with pen (mouse works) image fully visible (from SIGINT)
+ Trace 30207
I can't reproduce this bug from current CVS (5th Nov)
I was tempted to say that I cannot reproduce the problem but after a lot of successful strokes, gimp-1.3.11 crashed on me when trying to reproduce the segfault. As described by the reporter, the segfault traces back to libc memory allocation routines so I suspect the segfault is caused by an earlier problem, most probably a double free(). This might be a duplicate of bug #91041.
It would be very helpful if the original reporter could include information about the version of GTK+ he was using. I was able to reproduce this with gtk+-2.0.6 (debian testing).
Would be nice to get some feedback if the problem can still be reproduced using newer versions of GTK+ (2.2.x).
I am not able to reproduce the problem with GIMP-1.3.15 and GTK+-2.2.2. Unless we get some feedback very soon, we will assume that the xinput fix in gtk+-2.2.2 cured the problem and close this report as FIXED.
Don't know if anyone cares anymore, but for future reference if anyone runs across the bug, the stack trace looks similar to the ones in bug 112548 (which had some potentially interesting comments at the end), bug 98630, and bug 115666. Each of these bug reports was on a Debian testing/unstable system.