Closed GoogleCodeExporter closed 9 years ago
This is after r1001 rolled into chromium.
Original comment by z...@google.com
on 9 Apr 2012 at 8:24
The fragment shader doesn't reach the ANGLE DLLs, so it must be failing at the
validator. This is with Canary 20.0.1096.1. It works fine in a stand-alone ES2
application.
Original comment by nicolas....@gmail.com
on 9 Apr 2012 at 11:37
Did you just have r1001 rolled in, or were you using at least r1016?
Original comment by dan...@transgaming.com
on 11 Apr 2012 at 4:00
The test is failing in current top of tree Chromium, which is using ANGLE
revision 1018 per about:gpu.
Original comment by kbr@chromium.org
on 11 Apr 2012 at 5:11
I found the issue. The ESSL translator is turning this:
do int k; while(...)
into:
doint k;while(...)
So there should be a newline or space after "do". Should be an easy fix...
Original comment by nicolas....@gmail.com
on 13 Apr 2012 at 5:14
Sorry, it actually does output a newline after "do". I didn't correctly copy
the output. Anyway, feeding this output from the ESSL translator into the HLSL
translator didn't cause any errors at all.
So is seems to me that the issue must be in Chrome. When I attach a debugger
and run the conformance test, there is no call to glShaderSource for the
fragment shader (only for the vertex shader). There's a call to glCreateShader
for both though.
Original comment by nicolas....@gmail.com
on 13 Apr 2012 at 5:51
OK, I'll have a look.
Original comment by z...@google.com
on 13 Apr 2012 at 5:58
After further debugging, it is in ValidateLimitations.cpp that while-loop and
do-while-loop are rejected. In the GLSL ES spec, Appendix A-4, "Support for
while and do-while is not mandated".
Therefore, our implementation is correct, and the test was wrong. I am going
to fix the test.
Original comment by z...@google.com
on 18 Apr 2012 at 5:38
Ah thanks for tracking that down. I realize what we did wrong in testing with
ESSL translator -- we were using the ESSL spec, not WebGL spec option.
Original comment by dan...@transgaming.com
on 18 Apr 2012 at 6:12
Original issue reported on code.google.com by
zmo@chromium.org
on 9 Apr 2012 at 8:24