[Mesa-dev] Last big core UBO support dump.

[Eric Anholt]

I've managed to get the lowering pass working sufficiently to get the same set
of tests passing on i965, and the problems seem tractable enough to go with
that.  I'm hoping to finish up UBOs by the end of the month.  Here are my
remaining core changes other than that lowering pass.

There are a couple of oglconform failures related to the core support that I
haven't dealt with yet.  It expects to be able to declare a structure type
without declaring a variable.  I need to spend a bit more time with the spec,
and maybe just some experimenting on other drivers, to see if this is correct.
Also, it complains about the offset of variable 13 in one of the awful
make-random-variables-in-a-uniform-block tests.  We also don't check block
size against limits yet, but it's not in an oglconform or piglit test currently.

As usual, the WIP stuff is in the "ubo" branch of my mesa tree.