Description of problem: Starting mythtv on the latest FC 6 with all updates results in a segfault, kicking me out of X. Version-Release number of selected component (if applicable): mesa-libGL-6.5.1-8.fc6 How reproducible: Always Steps to Reproduce: 1. Voodoo 3500 2. FC 6 with all updates applied 3. Build mythtv, and attempt to start it Actual results: Backtrace: 0: X(xf86SigHandler+0x81) [0x80c9bd1] 1: [0xb69420] 2: /usr/lib/dri/tdfx_dri.so [0xf2c9fa] 3: /usr/lib/xorg/modules/extensions/libglx.so [0xcdf16a] 4: /usr/lib/xorg/modules/extensions/libglx.so(DoMakeCurrent+0x3bc) [0xcb880c] 5: /usr/lib/xorg/modules/extensions/libglx.so(__glXMakeCurrent+0x39) [0xcb89b9] 6: /usr/lib/xorg/modules/extensions/libglx.so [0xcbbd6a] 7: X(Dispatch+0x19a) [0x80880da] 8: X(main+0x485) [0x806fa65] 9: /lib/libc.so.6(__libc_start_main+0xdc) [0x16ff2c] 10: X(FontFileCompleteXLFD+0x1e9) [0x806eda1] Fatal server error: Caught signal 11. Server aborting Expected results: I want my mythtv fix. Additional info: See attached
Created attachment 146074 [details] Xorg.0.log
Created attachment 146075 [details] Output of lspci I _NEVER_ use any non-free binary blobs. I'm using only the drivers that come with X.org
Can we get your /etc/X11/xorg.conf as well, please? And you can also try to restart X without any such file whatsoever, and send us report (with /var/log/Xorg.*.log files, of course) what happens. Thanks a lot.
Created attachment 146674 [details] My xorg.conf My basic xorg.conf. Nothing fancy.
Created attachment 146675 [details] Xorg.0.log, after renaming tdfx_dri.so Looks not much different than with tdfx_dri, except for the actual complaint that the file is missing. After renaming tdfx_dri and starting X, Myth appears to be working. Looks like this module is not absolutely necessary for hardware acceleration, at least it is not required by mythtv.
Making an attachment did not change this bug's status, checking if a plain comment does the trick. Ping.
Mmm, looks like a real bug in the tdfx dri driver. Really need to hook up some addr2line action on the backtrace output though. Will take a look.
fwiw -- tdfx_dri in the last FC4 x.org errata worked fine. Shouldn't be too many changes in the driver for these old cards, since then.
Fixed in rawhide. tdfx driver probes for Glide3 during CreateContext, not during CreateScreen, and then the DRI core assumes the context private exists. It doesn't when glide3 isn't present, and you segfault.