we've bisected a bug in i915_dri.so which was caused by this commit:
commit c8a86f717f8e30204c615d06bcec159410ac06c5
Author: Brian Paul <email address hidden>
Date: Fri Aug 31 08:33:31 2012 -0600
mesa: move IS_NEGATIVE() and DIFFERENT_SIGNS() to macros.h
and it was discussed briefly on #intel-gfx:
< ickle> let me guess: GLboolean is a char
< ickle> yup.
< ickle> DIFFERENT_SIGNS() always return 0 as (unsigned char)(1<<31) == 0
< ickle> also note that fi_type isn't well defined when sizeof(int)!=4
we've bisected a bug in i915_dri.so which was caused by this commit:
commit c8a86f717f8e302 04c615d06bcec15 9410ac06c5
Author: Brian Paul <email address hidden>
Date: Fri Aug 31 08:33:31 2012 -0600
mesa: move IS_NEGATIVE() and DIFFERENT_SIGNS() to macros.h
and it was discussed briefly on #intel-gfx:
< ickle> let me guess: GLboolean is a char
< ickle> yup.
< ickle> DIFFERENT_SIGNS() always return 0 as (unsigned char)(1<<31) == 0
< ickle> also note that fi_type isn't well defined when sizeof(int)!=4
the downstream bug is
https:/ /bugs.launchpad .net/ubuntu/ quantal/ +source/ mesa/+bug/ 1047306
causing unity/compiz segfaults with hw using the dri driver. Reverting just that commit was verified to work.