recent i965 bugs

[Richard Goedeken]

> And, yes, please do see if you can find the source of the alpha blending
> problem.

I did another test series with git-bisect, and the source of this problem was 
identified as:

b35811e1b38e8fd454f369aefc0b3b3c50ecb8bc is first bad commit
Date:   Wed Jan 2 13:20:00 2008 -0800
[965] Convert CC unit to use a cache key instead of brw_cache_data.

It may be related to the Z-buffer rather than alpha blending.  I went back to 
the tip and reverted the 2 commits with bugs.  This one had some conflicts, 
which I resolved in the obvious way without really knowing the code.  The 
resulting drivers did not exhibit the 2 bugs that I just reported, but there was 
an obvious Z-buffer problem visible -- things were painted in the wrong order.

Richard