Limit struct size to size of full shader constant.

Limit struct size to size of full shader constant.

The HLSL compiler can optimize out the last elements of a struct, which caused ANGLE to overwrite constants right after the end of the struct. Each element in a struct should be limited by the size of the entire struct.

BUG=

Committed: https://code.google.com/p/angleproject/source/detail?r=1394

[kbr1, 2012-11-27 17:07:41]

Any possibility of a test case for this (maybe something in the WebGL
conformance suite)?

Is there an associated bug? How was the issue discovered?

[nicolas, 2012-11-27 17:16:16]

On 2012/11/27 17:07:41, kbr1 wrote:
> Any possibility of a test case for this (maybe something in the WebGL
> conformance suite)?
> 
> Is there an associated bug? How was the issue discovered?

This is the From Dust issue.

[kbr1, 2012-11-27 17:22:05]

On 2012/11/27 17:16:16, nicolas wrote:
> On 2012/11/27 17:07:41, kbr1 wrote:
> > Any possibility of a test case for this (maybe something in the WebGL
> > conformance suite)?
> > 
> > Is there an associated bug? How was the issue discovered?
> 
> This is the From Dust issue.

Thanks. Still any possibility of creating a regression test?

[John Bauman, 2012-11-27 18:36:33]

On 2012/11/27 17:22:05, kbr1 wrote:
> On 2012/11/27 17:16:16, nicolas wrote:
> > On 2012/11/27 17:07:41, kbr1 wrote:
> > > Any possibility of a test case for this (maybe something in the WebGL
> > > conformance suite)?
> > > 
> > > Is there an associated bug? How was the issue discovered?
> > 
> > This is the From Dust issue.
> 
> Thanks. Still any possibility of creating a regression test?

I'll see what I can do.

[apatrick1, 2012-11-27 20:11:16]

Thanks for fixing this. I was tearing my hair out! Agree with Ken that a WebGL
regression test would be good. LGTM for the ANGLE change if none of the existing
WebGL conformance tests regressed.

[John Bauman, 2012-11-28 00:22:54]

On 2012/11/27 20:11:16, apatrick1 wrote:
> Thanks for fixing this. I was tearing my hair out! Agree with Ken that a WebGL
> regression test would be good. LGTM for the ANGLE change if none of the
existing
> WebGL conformance tests regressed.
Ok, here's the new conformance test:
https://github.com/jbauman2/WebGL/commit/a2456978aee51ecc93885d6831a0e6a18c58...

[kbr1, 2012-11-28 01:29:43]

On 2012/11/28 00:22:54, John Bauman wrote:
> On 2012/11/27 20:11:16, apatrick1 wrote:
> > Thanks for fixing this. I was tearing my hair out! Agree with Ken that a
WebGL
> > regression test would be good. LGTM for the ANGLE change if none of the
> existing
> > WebGL conformance tests regressed.
> Ok, here's the new conformance test:
>
https://github.com/jbauman2/WebGL/commit/a2456978aee51ecc93885d6831a0e6a18c58...

Awesome. Thanks very much.