On 02/02/2014 04:13 AM, Kenneth Graunke wrote:
While attempting to bump GL_MAX_3D_TEXTURE_SIZE from 256 to 2048 on i965,
I ran into failures in the max-texture-size test. This wasn't because
anything was actually broken, but because the test reported things as
"fail" when it really meant "I couldn't allocate enough memory to actually
tell you one way or another." Which is not what "fail" means.
So I started looking into fixing the test, and...ended up rewriting a lot
of it in the process. It now uses the subtest mechanism to report all 30
cases independently, and reports "skip" for subcases that couldn't run
successfully due to a lack of memory.
The test could probably use some more work, but I think I've done enough
damage for the time being.
Thanks for doing this. I've made half-hearted attempts to rewrite this
test too but never carried through on it.
Reviewed-by: Brian Paul <[email protected]>
One issue I recall is the width*height*depth calculation overflowing a
32-bit int/size_t somewhere, causing invalid results on 32-bit systems.
That might still need to be addressed.
-Brian
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit
