[Mesa-dev] [Bug 96765] BindFragDataLocationIndexed on array fragment shader output.

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=96765

--- Comment #3 from Ilia Mirkin <imirkin at alum.mit.edu> ---
(In reply to Corentin Wallez from comment #2)
> Thanks for taking a look, indeed my understanding was wrong but I found a
> Khronos bug related to this that doesn't have a formal resolution but
> indicates that while you can't bind array elements individually, "array[0]"
> should be equivalent to "array". See
> https://cvs.khronos.org/bugzilla/show_bug.cgi?id=7829
> 
> Unfortunately Chrome has been testing this on few platforms, so far it only
> tested NVIDIA proprietary (passes), fglrx (doesn't pass). It also fails on
> the OSX drivers, seemingly for the same reason as for Mesa it seems.
> 
> Given all the above, I will fix Chromium's tests to not use indices, can you
> still consider making the [0] equivalent to no subscript?
> 
> Thank you for your time.

That does seem reasonable. As I'm not one of the cool kids with KHR access,
mind letting me know what that bug says about what to do when someone binds
both "array" and "array[0]"? Right now we store these in a map, and resolve
locations at link time (as per the spec). We don't know what's an array and
what's not at bind time.

Also, does the bug make any comments about AoA? i.e.

out vec4 foo[2][2][2]

Should you be able to bind with "foo[0][0][0]"? What about "foo[0][0]" and
"foo[0]" in that case? [Perhaps the AoA spec says something about this, I'm not
up on all the details myself.]

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160706/4c0913f3/attachment.html>