Test that main thread frames cease to fire during composited animation

Test that main thread frames cease to fire during composited animation

This change adds a WebKitBrowserTest that runs a composited transform
animation and verifies that main thread frames don't execute during
the animation.

BUG=704763
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel

[alancutter (OOO until 2018), 2017-04-06 08:12:55]

Description was changed from

==========
+testMainFrameBadBad

BUG=704763
==========

to

==========
+testMainFrameBadBad

BUG=704763
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel
==========

[alancutter (OOO until 2018), 2017-04-06 09:42:45]

The CQ bit was checked by alancutter@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-04-06 09:43:10]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[alancutter (OOO until 2018), 2017-04-06 09:46:30]

Description was changed from

==========
+testMainFrameBadBad

BUG=704763
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel
==========

to

==========
Test that main thread frames cease to fire during composited animation

This change adds a WebKitBrowserTest that runs a composited transform
animation and verifies that main thread frames don't execute during
the animation.

BUG=704763
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel
==========

[alancutter (OOO until 2018), 2017-04-06 09:46:30]

alancutter@chromium.org changed reviewers:
+ sunnyps@chromium.org

[alancutter (OOO until 2018), 2017-04-06 09:48:11]

Please let me know if this test isn't going about this in the right way. I'm new
to CC and browser tests.

[alancutter (OOO until 2018), 2017-04-07 00:22:56]

The CQ bit was checked by alancutter@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-04-07 00:23:54]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2017-04-07 02:19:48]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-04-07 02:19:49]

Dry run: This issue passed the CQ dry run.

[sunnyps, 2017-04-11 23:35:32]

sunnyps@chromium.org changed reviewers:
+ enne@chromium.org

[sunnyps, 2017-04-11 23:35:32]

I like the idea of testing this. However, I don't think we want a new member in
CompositorFrameMetadata that's used only for this test. Is there some kind of
frame count on the main thread that can be queried from the browser? enne@ might
have ideas on how to test this.

[enne (OOO), 2017-04-12 17:22:36]

I'm not super keen on browser tests.

Can this be a webkit unit test, maybe something in AnimationSimTest.cpp?

[alancutter (OOO until 2018), 2017-04-13 01:46:32]

On 2017/04/12 at 17:22:36, enne wrote:
> I'm not super keen on browser tests.
> 
> Can this be a webkit unit test, maybe something in AnimationSimTest.cpp?

Reproducing the original bug requires executing cc/ code that calls into Blink.
SimTests are restricted to Blink only code.
This bug cannot be caught by Blink test frameworks alone, it requires the
end-to-end coverage of the interaction between Blink and cc.

[enne (OOO), 2017-04-13 18:26:42]

One other suggestion might be to use tracing to check the number of begin
frames? There is a RenderWidget::willBeginCompositorFrame trace event.  You
could do something like what TracingbrowserTest does to see how many begin
frames happened.  That'd require less plumbing.

sunnyps: what do you think of that?

[sunnyps, 2017-04-13 21:17:31]

On 2017/04/13 18:26:42, enne wrote:
> One other suggestion might be to use tracing to check the number of begin
> frames? There is a RenderWidget::willBeginCompositorFrame trace event.  You
> could do something like what TracingbrowserTest does to see how many begin
> frames happened.  That'd require less plumbing.
> 
> sunnyps: what do you think of that?

Using tracing sounds good to me. You should be able to count BeginMainFrame
events and check that.

[alancutter (OOO until 2018), 2017-04-19 00:23:20]

On 2017/04/13 at 18:26:42, enne wrote:
> One other suggestion might be to use tracing to check the number of begin
frames? There is a RenderWidget::willBeginCompositorFrame trace event.  You
could do something like what TracingbrowserTest does to see how many begin
frames happened.  That'd require less plumbing.
> 
> sunnyps: what do you think of that?

This is an excellent idea, thank you!