Mesa is git master at bf611cef29d664c2c2b56d2175c3a966c720e8b8. Using the r600 driver and 2.6.32+drm-radeon-testing with KMS enabled, but this happens even when I set LIBGL_SOFTWARE_RENDERING=true. progs/perf/genmipmaps segfaults on the very first call to glGenerateMipmap(). Here is the backtrace: #0 0x0000000000000000 in ?? () #1 0x000000000042b9df in GenMipmap (count=2) at genmipmap.c:84 #2 0x000000000042bc07 in PerfMeasureRate (f=0x42b970 <GenMipmap>) at common.c:76 #3 0x000000000042b910 in PerfDraw () at genmipmap.c:128 #4 0x000000000042be69 in Draw () at glmain.c:186 #5 0x00007ffff7bbc79d in fghRedrawWindow (window=0x660310, enumerator=0x7fffffffdf40) at freeglut_main.c:212 #6 fghcbDisplayWindow (window=0x660310, enumerator=0x7fffffffdf40) at freeglut_main.c:229 #7 0x00007ffff7bbfb79 in fgEnumWindows ( enumCallback=0x7ffff7bbc6f0 <fghcbDisplayWindow>, enumerator=0x7fffffffdf40) at freeglut_structure.c:388 #8 0x00007ffff7bbcc51 in fghDisplayAll () at freeglut_main.c:251 #9 glutMainLoopEvent () at freeglut_main.c:998 #10 0x00007ffff7bbd4a0 in glutMainLoop () at freeglut_main.c:1046 #11 0x000000000042be42 in main (argc=1, argv=0x7fffffffe1b8) at glmain.c:266 Interestingly, when I replace the call to glGenerateMipmap() with glGenerateMipmapEXT(), the segfault does not happen and the program seems to work.
It does segfault here with R600 but not with software renderer. The environment variable to force the software renderer is LIBGL_ALWAYS_SOFTWARE not LIBGL_SOFTWARE_RENDERING. Are you sure you were running the program with the software renderer when it segfaulted? If not, you should change it to R600 bug.
On second thought, the radeonGenerateMipmap function isn't called at all. So maybe it is a mesa bug? When run normal: ------------------------------------ 0x08067fe3 84 glGenerateMipmap(GL_TEXTURE_2D); (gdb) si 0x00000000 in ?? () When forced to software renderer: ------------------------------------ 0x08067fe3 84 glGenerateMipmap(GL_TEXTURE_2D); (gdb) si 0xb7f1b290 in glGenerateMipmapEXT () from /usr/lib/libGL.so.1 (gdb) si 0xb7f1b296 in glGenerateMipmapEXT () from /usr/lib/libGL.so.1 (gdb) si _mesa_GenerateMipmapEXT (target=3553) at main/fbobject.c:1944
The bug is that the test uses the GL_ARB_framebuffer_object version of the function, but the driver doesn't support GL_ARB_framebuffer_object. It doesn't work. *shock* GLEW sees that the driver doesn't support GL_ARB_framebuffer_object, so it correctly never calls glXGetProcAddress on glGenerateMipmap. Since the driver does support GL_EXT_framebuffer_object, it does call glXGetProcAddress on glGenerateMipmapEXT.
I've added a test for GL_ARB_framebuffer_object in genmipmap.c on the 7.7 branch. If you want to submit a patch for using GL_EXT_framebuffer_object, feel free.
Hello, I can reproduce this bug every time, both with Fedora 13 and 14, latest KDE. Just enabling desktop effects causes the error. Please see the following bug reports, already posted to bugs.kde.org for backtraces (two, actually, one for fc13 and one for fc14, both are usable.) https://bugs.kde.org/show_bug.cgi?id=257492 https://bugs.kde.org/show_bug.cgi?id=257508 BR, Hristo
(In reply to comment #5) > Hello, > > I can reproduce this bug every time, both with Fedora 13 and 14, latest KDE. > Just enabling desktop effects causes the error. Please see the following bug > reports, already posted to bugs.kde.org for backtraces (two, actually, one for > fc13 and one for fc14, both are usable.) > > https://bugs.kde.org/show_bug.cgi?id=257492 > https://bugs.kde.org/show_bug.cgi?id=257508 These appear to be a completely different bug. They're also marked a duplicates of another bug. That bug (https://bugs.kde.org/show_bug.cgi?id=243767) is marked as "fixed upstream." Perhaps you need updated Mesa bits. In any case, I'm closing this bug.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.