Comment # 2 on bug 68503 from Vadim Girlin

Please attach the output with R600_DEBUG=sb,ps,vs.

Could you also record the GL trace that reproduces the issue using apitrace -
http://apitrace.github.io/

You can just upload the trace if it's not too big, then it might help me to
reproduce the bug if it's not hw-specific, but also you can use recorded trace
to find the exact failing shader as described in "Regression debugging" section
here - http://people.freedesktop.org/~vadimg/r600-sb.html - you'll need to
replay it with R600_DEBUG=sb,sbstat first to figure out initial range of shader
indices used by the app, and then apply R600_SB_DSKIP_* env vars when replaying
the trace to bisect the range as described in that section. After locating
broken shader just add "ps,vs" to R600_DEBUG and attach the output. You'll have
a cmd line like this in the end (with some other shader index instead of 37):
R600_DEBUG=sb,ps,vs R600_SB_DSKIP_MODE=2 R600_SB_DSKIP_START=37
R600_SB_DSKIP_END=37 <cmd to replay the trace> &> log_file

I can try to do everything myself if you can send me the key for SS3 (to
vadimgirlin at gmail dot com), this can save some time, but the bug may be
specific to your GPU or system configuration and in such case possibly I won't
be able to reproduce it on my hardware, then I'll still need your help with
debugging as described above.