Fixed framebuffer-object-attachment WebGL conformance test....

Fixed framebuffer-object-attachment WebGL conformance test.

We updated it with WebGL conformance test r14153 to exercise zero-size FBO attachments, which failed with ANGLE. The new conformance test passes with OpenGL and Mesa.

See this Chromium bug:
http://code.google.com/p/chromium/issues/detail?id=75666

D3D fails if you try to create a zero size depth buffer but OpenGL ES2 allows it. D3D / drivers seem to sometimes crash rather than fail normally, though this might just be because some users have enabled the D3D debug runtime and break on error and we're getting the reports.

It was also returning unexpected results when calling GetRenderbufferParameter for parameters that do not apply to a particular buffer. For example, RED_SIZE for a zero sized depth buffer should be 0.

Tested by running WebGL conformance test with retail D3D runtime (passes) and debug D3D runtime (passes and does not assert in D3D or ANGLE).
Committed: http://code.google.com/p/angleproject/source/detail?r=577

[apatrick1, 2011-03-16 21:10:05]

Hi Daniel,

Please review.

Thanks.

[dgkoch, 2011-03-17 04:07:16]

one minor nit.   

LGTM after that is fixed.

Daniel

http://codereview.appspot.com/4284053/diff/1/src/libGLESv2/utilities.cpp
File src/libGLESv2/utilities.cpp (right):

http://codereview.appspot.com/4284053/diff/1/src/libGLESv2/utilities.cpp#newc...
src/libGLESv2/utilities.cpp:724: return 0;
this is unreachable, might as well delete it like the others.