Ensure alpha channel in backbuffer is correct with DirectComposition.

Ensure alpha channel in backbuffer is correct with DirectComposition.

gles2_conform_test runs with the command buffer emulating an RGB backbuffer on top of RGBA. The backbuffer was initially being cleared correctly to 1.0 alpha, but after the first swap a buffer with alpha 0.0 was returned, which caused the test to fail.

Instead in this case after the swap the command buffer should set a flag that the backbuffer needs to be cleared before the next draw; the backbuffer clearing code can handle setting alpha to 1.0. Resize also needs to force a clear, but swaps afterwards will return a backbuffer that's been used before and has been cleared correctly.

BUG=524838
Committed: https://crrev.com/2634dd83a2ba68f4dd33c5e5bbc7ffde3ba87845
Cr-Commit-Position: refs/heads/master@{#369912}

[jbauman, 2016-01-15 03:00:04]

jbauman@chromium.org changed reviewers:
+ kbr@chromium.org

[Ken Russell (switch to Gerrit), 2016-01-15 03:38:34]

I strongly recommend adding a bug ID. It's clear that introducing
DirectComposition support has needed a series of patches, and they should all be
linked to the same bug.

LGTM with a couple of comments.

https://codereview.chromium.org/1583283004/diff/20001/gpu/command_buffer/serv...
File gpu/command_buffer/service/gles2_cmd_decoder.cc (right):

https://codereview.chromium.org/1583283004/diff/20001/gpu/command_buffer/serv...
gpu/command_buffer/service/gles2_cmd_decoder.cc:3648: glClearColor(0, 0, 0,
BackBufferHasAlpha() ? 0 : 1.f);
You're sure this change won't affect visible behavior?

https://codereview.chromium.org/1583283004/diff/20001/ui/gl/gl_surface_egl.cc
File ui/gl/gl_surface_egl.cc (right):

https://codereview.chromium.org/1583283004/diff/20001/ui/gl/gl_surface_egl.cc...
ui/gl/gl_surface_egl.cc:768: return g_use_direct_composition;
Is this correct in all cases? Is there really no situation where the underlying
EGL implementation uses buffer flipping and needs this kind of workaround?

[Ken Russell (switch to Gerrit), 2016-01-15 03:39:21]

In fact
https://build.chromium.org/p/tryserver.chromium.android/builders/linux_androi...
shows that some more thought is needed on other EGL implementations.

[jbauman, 2016-01-15 20:54:34]

Description was changed from

==========
Ensure alpha channel in backbuffer is correct with DirectComposition.

gles2_conform_test runs with the command buffer emulating an RGB backbuffer on
top of RGBA. The backbuffer was initially being cleared correctly to 1.0 alpha,
but after the first swap a buffer with alpha 0.0 was returned, which caused the
test to fail.

Instead in this case after the swap the command buffer should set a flag that
the backbuffer needs to be cleared before the next draw; the backbuffer clearing
code can handle setting alpha to 1.0. Resize also needs to force a clear, but
swaps afterwards will return a backbuffer that's been used before and has been
cleared correctly.


BUG=none
==========

to

==========
Ensure alpha channel in backbuffer is correct with DirectComposition.

gles2_conform_test runs with the command buffer emulating an RGB backbuffer on
top of RGBA. The backbuffer was initially being cleared correctly to 1.0 alpha,
but after the first swap a buffer with alpha 0.0 was returned, which caused the
test to fail.

Instead in this case after the swap the command buffer should set a flag that
the backbuffer needs to be cleared before the next draw; the backbuffer clearing
code can handle setting alpha to 1.0. Resize also needs to force a clear, but
swaps afterwards will return a backbuffer that's been used before and has been
cleared correctly.


BUG=524838
==========

[jbauman, 2016-01-15 21:50:37]

On 2016/01/15 03:38:34, Ken Russell wrote:
> I strongly recommend adding a bug ID. It's clear that introducing
> DirectComposition support has needed a series of patches, and they should all
be
> linked to the same bug.
> 
> LGTM with a couple of comments.
> 
>
https://codereview.chromium.org/1583283004/diff/20001/gpu/command_buffer/serv...
> File gpu/command_buffer/service/gles2_cmd_decoder.cc (right):
> 
>
https://codereview.chromium.org/1583283004/diff/20001/gpu/command_buffer/serv...
> gpu/command_buffer/service/gles2_cmd_decoder.cc:3648: glClearColor(0, 0, 0,
> BackBufferHasAlpha() ? 0 : 1.f);
> You're sure this change won't affect visible behavior?

This should fix the clear after a glDiscardFramebufferEXT with an onscreen
surface with an alpha channel.
> 
> https://codereview.chromium.org/1583283004/diff/20001/ui/gl/gl_surface_egl.cc
> File ui/gl/gl_surface_egl.cc (right):
> 
>
https://codereview.chromium.org/1583283004/diff/20001/ui/gl/gl_surface_egl.cc...
> ui/gl/gl_surface_egl.cc:768: return g_use_direct_composition;
> Is this correct in all cases? Is there really no situation where the
underlying
> EGL implementation uses buffer flipping and needs this kind of workaround?

Having this extension enabled is the only circumstance where we really know that
there's this sort of flipping behavior. We may need to do something more
elaborate than this if we add support for EGL_KHR_partial_update.

[jbauman, 2016-01-16 00:33:45]

The CQ bit was checked by jbauman@chromium.org

[jbauman, 2016-01-16 00:33:45]

The patchset sent to the CQ was uploaded after l-g-t-m from kbr@chromium.org
Link to the patchset: https://codereview.chromium.org/1583283004/#ps40001
(title: " ")

[commit-bot: I haz the power, 2016-01-16 00:33:56]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1583283004/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1583283004/40001

[commit-bot: I haz the power, 2016-01-16 01:44:04]

Description was changed from

==========
Ensure alpha channel in backbuffer is correct with DirectComposition.

gles2_conform_test runs with the command buffer emulating an RGB backbuffer on
top of RGBA. The backbuffer was initially being cleared correctly to 1.0 alpha,
but after the first swap a buffer with alpha 0.0 was returned, which caused the
test to fail.

Instead in this case after the swap the command buffer should set a flag that
the backbuffer needs to be cleared before the next draw; the backbuffer clearing
code can handle setting alpha to 1.0. Resize also needs to force a clear, but
swaps afterwards will return a backbuffer that's been used before and has been
cleared correctly.


BUG=524838
==========

to

==========
Ensure alpha channel in backbuffer is correct with DirectComposition.

gles2_conform_test runs with the command buffer emulating an RGB backbuffer on
top of RGBA. The backbuffer was initially being cleared correctly to 1.0 alpha,
but after the first swap a buffer with alpha 0.0 was returned, which caused the
test to fail.

Instead in this case after the swap the command buffer should set a flag that
the backbuffer needs to be cleared before the next draw; the backbuffer clearing
code can handle setting alpha to 1.0. Resize also needs to force a clear, but
swaps afterwards will return a backbuffer that's been used before and has been
cleared correctly.


BUG=524838
==========

[commit-bot: I haz the power, 2016-01-16 01:44:04]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-01-16 01:45:06]

Description was changed from

==========
Ensure alpha channel in backbuffer is correct with DirectComposition.

gles2_conform_test runs with the command buffer emulating an RGB backbuffer on
top of RGBA. The backbuffer was initially being cleared correctly to 1.0 alpha,
but after the first swap a buffer with alpha 0.0 was returned, which caused the
test to fail.

Instead in this case after the swap the command buffer should set a flag that
the backbuffer needs to be cleared before the next draw; the backbuffer clearing
code can handle setting alpha to 1.0. Resize also needs to force a clear, but
swaps afterwards will return a backbuffer that's been used before and has been
cleared correctly.


BUG=524838
==========

to

==========
Ensure alpha channel in backbuffer is correct with DirectComposition.

gles2_conform_test runs with the command buffer emulating an RGB backbuffer on
top of RGBA. The backbuffer was initially being cleared correctly to 1.0 alpha,
but after the first swap a buffer with alpha 0.0 was returned, which caused the
test to fail.

Instead in this case after the swap the command buffer should set a flag that
the backbuffer needs to be cleared before the next draw; the backbuffer clearing
code can handle setting alpha to 1.0. Resize also needs to force a clear, but
swaps afterwards will return a backbuffer that's been used before and has been
cleared correctly.

BUG=524838

Committed: https://crrev.com/2634dd83a2ba68f4dd33c5e5bbc7ffde3ba87845
Cr-Commit-Position: refs/heads/master@{#369912}
==========

[commit-bot: I haz the power, 2016-01-16 01:45:06]

Patchset 3 (id:??) landed as
https://crrev.com/2634dd83a2ba68f4dd33c5e5bbc7ffde3ba87845
Cr-Commit-Position: refs/heads/master@{#369912}