Listen to focus change events instead of assuming they happen at next rAF

Listen to focus change events instead of assuming they happen at next rAF

focus-shadowhost-display-none.html is testing a case where a focused
element immediately unfocuses by hiding itself. The test assumes that
the unfocus completes by the next requestAnimationFrame, but because
Document uses a timer to implement unfocusing[1], this ordering isn't
guaranteed.

This patch makes the test more robust by listening to the blur event
explicitly instead of assuming this relationship with
requestAnimationFrame.

BUG=463143
[1] clearFocusedElementSoon at https://code.google.com/p/chromium/codesearch#chromium/src/third_party/WebKit/Source/core/dom/Document.cpp&rcl=1432882933&l=1888

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=196141

[Sami, 2015-05-29 12:25:32]

skyostil@chromium.org changed reviewers:
+ alexclarke@chromium.org, kochi@chromium.org

[alex clarke (OOO till 29th), 2015-05-29 12:27:15]

Looks fine to me.

[alex clarke (OOO till 29th), 2015-05-29 12:30:17]

lgtm

Since this is blocking the TimerBase refactoring I think we should land this
now. We can always come back to this if there are any issues.

[Sami, 2015-05-29 12:30:56]

The CQ bit was checked by skyostil@chromium.org

[commit-bot: I haz the power, 2015-05-29 12:31:48]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1148293010/1

[commit-bot: I haz the power, 2015-05-29 13:43:15]

Committed patchset #1 (id:1) as
https://src.chromium.org/viewvc/blink?view=rev&revision=196141

[kochi, 2015-06-01 02:37:55]

lgtm

Thanks for catching this.

FYI adding some information for supporting this change.

Actually I think the timing/order is not well defined on spec
as to requestAnimationFrame and focus->display:none
transition.

On the other hand,
display:none -> blur is discussed in whatwg ML:
https://lists.w3.org/Archives/Public/public-whatwg-archive/2015Jan/0023.html