I just comitted a bit of code for a new ARB program parser, which should
handle both vertex and fragment programs. Thanks to Michal Krol for his
hard work in putting together the parsing code and associated grammars.
This hasn't been tested too rigoursly, but it is stable enough to run
things like the arbfplight demo.
At this point, relative offsets and generic vertex attributes for vertex
programs are broken (not a parse problem, but a 'fill in the mesa structs
blanks'..), but everything for fragment programs should be in place.