3b

Looks like the only correct use of `VertexAttribEnumNV` was for the `pname` of `GetProgramParameterfvNV` and `GetProgramParameterdvNV`, which I switched to use `ProgramParameterPNameNV` instead. None of the other places using it...

Removed changes to deprecated group tags, and renamed some groups to old names where they still matched reasonably well. `VertexArrayPNameAPPLE` was pretty much useless before (it contained enums for `param`...

OK, will add the old group names back in addition to the merged ones, and rename the apple one. any suggestions for the name of the enum that actually gets...

@frederikja163 : reports that at least some of it is actively useful, and that the rest doesn't break things for existing code is helpful. I'll push this back up towards...

hmm, looks like a few already have a group, like `"CompressedTextureARB"` in ```xml void glCompressedTexImage1D const void *data ``` not sure how to handle that.

Yeah, that matches my expectations. The API shouldn't retain pointers to user data unless it is explicit in the spec (or otherwise obvious, like callback functions and their `user data`...

`--script` doesn't create a `.fasl` file, but by default sbcl compiles all non-trivial forms as part of evaluation, so it is running compiled code for the benchmark. That does mean...

to compile separately, remove the `(main)` call from `lisprb.lisp` and do something like ``` sbcl --load lisprb.lisp --eval '(sb-ext:save-lisp-and-die "lisprb" :executable t :toplevel (function main))' or sbcl --load lisprb.lisp --eval...