WebGL: optimize webgl.texSubImage2D(video) path. (Blink side)

WebGL: optimize webgl.texSubImage2D(video) path. (Blink side)

Currently, texSubImage2D(video) requires glReadPixels to copy video texture to
target texture. This CL makes WebGL copy video texture to target texture
directly.

In addition, it will reduce WebGL video test flakiness significantly.

TEST=video texture test on webgl_conformance
./content/test/gpu/run_gpu_test.py webgl_conformance --browser=release --story-filter=WebglConformance.conformance_textures_video_tex_image_and_sub_image_2d_with_video
BUG=349871, 504773

[dshwang, 2015-09-04 11:01:37]

dongseong.hwang@intel.com changed reviewers:
+ kbr@chromium.org

[dshwang, 2015-09-04 11:04:13]

kbr, could you review? This reduce WebGL video test flakiness significantly
(>80%).
It depends on chromium CL; https://codereview.chromium.org/1315323006/

As I mentioned early in
https://code.google.com/p/chromium/issues/detail?id=504773#c17, to fix it
completely, we need fix Android video framework, which doesn't draws the first
frame on the stream texture in time.

[Ken Russell (switch to Gerrit), 2015-10-01 05:06:55]

I'm sorry for taking so long to get to this. In general I've been concerned that
sidestepping the existing code paths with these GPU-to-GPU blits is masking some
deep, underlying bug. However, it seems it may be the case that some of the
problems are caused by attempting to do ReadPixels from a framebuffer backed by
a GL_TEXTURE_EXTERNAL_OES texture, which apparently is forbidden. So perhaps
this will be the best path forward after all.

At this point Blink is merged into the Chromium workspace. Would you combine
this with your other Chromium-side CL and re-upload it?

Thank you.

[dshwang, 2015-10-21 16:06:13]

On 2015/10/01 05:06:55, Ken Russell wrote:
> I'm sorry for taking so long to get to this. In general I've been concerned
that
> sidestepping the existing code paths with these GPU-to-GPU blits is masking
some
> deep, underlying bug. However, it seems it may be the case that some of the
> problems are caused by attempting to do ReadPixels from a framebuffer backed
by
> a GL_TEXTURE_EXTERNAL_OES texture, which apparently is forbidden. So perhaps
> this will be the best path forward after all.
> 
> At this point Blink is merged into the Chromium workspace. Would you combine
> this with your other Chromium-side CL and re-upload it?
> 
> Thank you.

I'm sorry for taking so long to get to this too. I had long vacation and in
another project. Now let me get this job done.

I merge this into https://codereview.chromium.org/1315323006/

BTW, I don't think Android flakiness is due to ReadPixels from a framebuffer
backed by a GL_TEXTURE_EXTERNAL_OES
It's because WebMediaPlayerAndroid::paint&copyVideoTextureToPlatformTexture use
temporary GL_TEXTURE_2D texture in the case.

Close as duplicated to https://codereview.chromium.org/1315323006/