I can’t see anything answering this in the ARB_multisample spec (correct me if I’m wrong), so does anyone know if ‘alpha to coverage’ should work when outputing to multiple colour buffers (COLOR0/1/2/3). If so, does anyone know the official line on whether the coverage mask calculation is based on the COLOR0 alpha of the fragment output or something else?
Hmm… I wouldn’t rely on this. Do you have multiple multisample buffers?
I’m not 100% sure if that’s mentioned anywhere in the spec, but I would expect it to use glFragData.a
This is specified in ARB_draw_buffers:
If there are multiple output colors being written to multiple buffers, the alpha used in alpha to coverage and alpha test is the alpha of output color 0.
That’s exactly what I wanted to know.
(sorry for delay in acknowledgement. I hadn’t subscribed to email updates).