Consistently generate INVALID_VALUE on NaN line width

Consistently generate INVALID_VALUE on NaN line width

GL drivers are allowed to generate errors when NaN is supplied to
functions under GLES 3.0 section 2.1.1 or OpenGL 4.5 section 2.3.4.1.
NVIDIA OpenGL driver generates an error when NaN is supplied to
glLineWidth. Make this behavior consistent for Chromium, which removes
the possibility of drawing with an undefined line width, and makes sure
that errors don't leak from the GL driver.

BUG=402847
TEST=WebGL conformance tests on debug build with virtual contexts,
gpu_unittests

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=289594

[oetuaho-nv, 2014-08-13 10:46:44]

Issue that https://codereview.chromium.org/458333003/ was supposed to fix was
not a driver bug after all. Please review this alternate fix.

[Zhenyao Mo, 2014-08-13 19:47:59]

On 2014/08/13 10:46:44, oetuaho-nv wrote:
> Issue that https://codereview.chromium.org/458333003/ was supposed to fix was
> not a driver bug after all. Please review this alternate fix.

LGTM

I still suggest we update the WebGL spec so we can write a conformance test to
force the behavior - undefiend behaviors are not OK in WebGL.

[oetuaho-nv, 2014-08-14 08:29:19]

The CQ bit was checked by oetuaho@nvidia.com

[commit-bot: I haz the power, 2014-08-14 08:32:43]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/oetuaho@nvidia.com/469673003/1

[oetuaho-nv, 2014-08-14 08:38:05]

On 2014/08/13 19:47:59, Zhenyao Mo wrote:
> On 2014/08/13 10:46:44, oetuaho-nv wrote:
> > Issue that https://codereview.chromium.org/458333003/ was supposed to fix
was
> > not a driver bug after all. Please review this alternate fix.
> 
> LGTM
> 
> I still suggest we update the WebGL spec so we can write a conformance test to
> force the behavior - undefiend behaviors are not OK in WebGL.

Yeah, it would be nice if we could get rid of undefined behavior altogether, but
it's not feasible to specify the handling of NaN in OpenGL shaders in general.
It's a bit tricky where to draw the line. The need to address the behavior of
this function in particular has risen more out of practical issues than trying
to make the spec as rigorous as possible. I do agree that it is still good to
spec this nevertheless, pull request here:
https://github.com/KhronosGroup/WebGL/pull/662

[commit-bot: I haz the power, 2014-08-14 15:21:01]

Committed patchset #1 (1) as 289594