Code review comment for lp://staging/~kdub/mir/gbm-ext-v2

(5) I would ignore other extensions as prior art. Those haven't been around as long as the rest of the project where we do use "const" more correctly, so should use const correctly (at least on parameters).

(6) The new naming convention sounds confusing and looks weakly typed:
   typedef uint32_t (*MirBufferGbmStride)(MirBuffer* buffer);
What if 'buffer' is not a GBM buffer?

I'm starting to think that the client producing its own gbm_bo might be the right answer. We would have fewer functions, less confusion and stronger typing. Admittedly with some additional linkage and the client would need to specify a gbm_device. But maybe that's a good thing in a multi-GPU world. I also wonder how this plays with RAOF's multi-GPU work in progress...