Refactor of various layout tests to use RAF

Refactor of various layout tests to use RAF

Refactor of various layout tests to use RAF because they break/become
flaky when the 1ms minimum delay in DOMTimer.cpp is removed by patch
https://codereview.chromium.org/467153003/

BUG=402694

[alexclarke, 2014-09-30 15:02:27]

alexclarke@google.com changed reviewers:
+ jamesr@chromium.org, skyostil@chromium.org

[jamesr, 2014-09-30 21:05:02]

jamesr@chromium.org changed reviewers:
- jamesr@chromium.org

[jamesr, 2014-09-30 21:05:03]

Thanks for the patch, but I'm afraid I won't be able to go through these.

[alexclarke, 2014-10-01 09:43:53]

alexclarke@google.com changed reviewers:
+ jchaffraix@chromium.org

[alexclarke, 2014-10-01 09:48:05]

alexclarke@google.com changed reviewers:
+ alexclarke@google.com

[alexclarke, 2014-10-01 09:48:06]

+jchaffraix@chromium.org

Hi Julien, I'm looking for a reviewer for these test refactors, would you be
able to do that? If not any idea who might be a good reviewer?

(Motivation for this refactor is https://codereview.chromium.org/467153003/)

Thanks!
Alex Clarke

[Julien - ping for review, 2014-10-03 22:22:32]

https://codereview.chromium.org/615033002/diff/1/LayoutTests/fast/events/scro...
File LayoutTests/fast/events/scroll-event-raf-timing.html (right):

https://codereview.chromium.org/615033002/diff/1/LayoutTests/fast/events/scro...
LayoutTests/fast/events/scroll-event-raf-timing.html:35: }, 1);
Those delay changes are separated from the rAF changes and should go in a
separate patch for easier blaming.

https://codereview.chromium.org/615033002/diff/1/LayoutTests/svg/custom/pendi...
File LayoutTests/svg/custom/pending-resource-after-removal.xhtml (right):

https://codereview.chromium.org/615033002/diff/1/LayoutTests/svg/custom/pendi...
LayoutTests/svg/custom/pending-resource-after-removal.xhtml:51: // Nested RAF to
ensure enough time elapsed for repaint logging to occur
runAfterDisplay would probably solve this issue but it would re-introduce the
previous setTimeout(..., 0).

IMO it would be cleaner than a nested rAF.

[alex clarke (OOO till 29th), 2014-10-10 17:46:02]

Sorry for the very slow response, this fell through the cracks.  PTAL

https://codereview.chromium.org/615033002/diff/1/LayoutTests/fast/events/scro...
File LayoutTests/fast/events/scroll-event-raf-timing.html (right):

https://codereview.chromium.org/615033002/diff/1/LayoutTests/fast/events/scro...
LayoutTests/fast/events/scroll-event-raf-timing.html:35: }, 1);
On 2014/10/03 22:22:32, Julien Chaffraix - PST wrote:
> Those delay changes are separated from the rAF changes and should go in a
> separate patch for easier blaming.

Done.

https://codereview.chromium.org/615033002/diff/1/LayoutTests/svg/custom/pendi...
File LayoutTests/svg/custom/pending-resource-after-removal.xhtml (right):

https://codereview.chromium.org/615033002/diff/1/LayoutTests/svg/custom/pendi...
LayoutTests/svg/custom/pending-resource-after-removal.xhtml:51: // Nested RAF to
ensure enough time elapsed for repaint logging to occur
On 2014/10/03 22:22:32, Julien Chaffraix - PST wrote:
> runAfterDisplay would probably solve this issue but it would re-introduce the
> previous setTimeout(..., 0).
> 
> IMO it would be cleaner than a nested rAF.

Done.

[alex clarke (OOO till 29th), 2014-10-14 10:20:33]

PTAL

[eseidel, 2014-10-14 16:13:28]

eseidel@chromium.org changed reviewers:
+ eseidel@chromium.org

[eseidel, 2014-10-14 16:13:28]

You can see how flaky they are:
http://test-results.appspot.com/dashboards/flakiness_dashboard.html

[eseidel, 2014-10-14 16:15:58]

https://codereview.chromium.org/615033002/diff/210001/LayoutTests/resources/j...
File LayoutTests/resources/js-test.js (right):

https://codereview.chromium.org/615033002/diff/210001/LayoutTests/resources/j...
LayoutTests/resources/js-test.js:694: requestAnimationFrame(callback);
I might leave a comment to explain why RAF works beter.  Fore xample it always
goes through the event loop?

[Julien - ping for review, 2014-10-15 20:37:03]

On 2014/10/14 at 16:15:58, eseidel wrote:
>
https://codereview.chromium.org/615033002/diff/210001/LayoutTests/resources/j...
> File LayoutTests/resources/js-test.js (right):
> 
>
https://codereview.chromium.org/615033002/diff/210001/LayoutTests/resources/j...
> LayoutTests/resources/js-test.js:694: requestAnimationFrame(callback);
> I might leave a comment to explain why RAF works beter.  Fore xample it always
goes through the event loop?

The idea is that we are removing the clamping for setTimeout(..., 0) to 1ms and
that *will* cause tests to become flaky.

My biggest concern about this change is that rAF is fired at 60 Hz, which is
potentially a 16ms delay vs 1ms without them. That's a 16x slowdown on these
tests, just because we are changing setTimeout's behavior.

I am not really involved in the original change which makes it hard for me to
understand the trade-off for this behavior change. The fact that setTimeout(...,
0) behaves flakily makes me uneasy as some applications must be using it.
Ideally you would want the same reviewer to look at both changes as they are
closely related.

[alex clarke (OOO till 29th), 2014-10-21 15:52:03]

Sorry again for the slow response.  Iv'e been working on some other stuff.
Anyway Eric suggested I change all the affected setTimeout 0s to 1's with a
fixme.  I've actually all of them in https://codereview.chromium.org/644093003/