Add mouseLeave API in eventSender.

Add mouseLeave API in eventSender.

This patch makes the test can trigger a mouseLeave event.

BUG=450631
R=rbyers@chromium.org

Committed: https://crrev.com/12c7fea23d6077af7d6a2ea6f8fc4680cfaf23d1
Cr-Commit-Position: refs/heads/master@{#319393}

[Miyoung Shin, 2015-02-16 07:42:08]

@rbyers, PTAL.

I verified it by the test uloaded at http://crrev.com/930803002 [1]. 
After merging this patch, we can upload the patch[1].

[Rick Byers, 2015-02-20 20:48:54]

On 2015/02/16 07:42:08, Miyoung Shin wrote:
> @rbyers, PTAL.
> 
> I verified it by the test uloaded at http://crrev.com/930803002 [1]. 
> After merging this patch, we can upload the patch[1].

This is unnecessarily complicated and overly specific.

At the blink layer we don't care why the OS may choose to tell us that the mouse
has left our window (whether it's because a new window was created on top, or
because our window suddenly got sent to another monitor, or because the mouse
was unplugged).  The key issue is that the OS (via the embedder - chromium) sent
us an event saying the mouse had left our window, and blink needs to honor that
by clearing any hover state and dispatching any mouseout/mouseleave events to
JS.  The bug is that even when chromium was telling us the mouse had left our
window, we were sending only a mousemove event to the page when the co-ordinates
appeared to us to be inside the window still.  That decision isn't up to us -
the OS decides which window the mouse targets based on it's co-ordinates (and
perhaps other things), if the OS says the mouse is leaving for whatever reason,
then we should tell the page that.

So you can test this more thoroughly and more simply by removing all the logic
about "creating a new window".  Just expose an API so a test can trigger a
mouseLeave event (similar to what you've added to event_sender.cc already but
exposed to layout tests the same way eventSender.mouseMoveTo is).  Then your
test doesn't need to create any new windows, just send a mouseleave event whose
co-ordinates happen to still be inside the window.  Without your earlier fix
you'll surprisingly get a mousemove event in such a case.  Now that you've fixed
it we should see both mouseout and mouseleave fired to the window.

[Miyoung Shin, 2015-02-23 02:38:37]

Thank you for your kind comment.
I might miss the point you asked. I uploaded the patch to add
eventSender.mouseLeave.
PTAL when you come back. :)

[Miyoung Shin, 2015-03-05 04:59:06]

On 2015/02/23 02:38:37, Miyoung Shin wrote:
> Thank you for your kind comment.
> I might miss the point you asked. I uploaded the patch to add
> eventSender.mouseLeave.
> PTAL when you come back. :)

Rick@, could you take a look?

[Rick Byers, 2015-03-05 14:42:06]

LGTM

[Rick Byers, 2015-03-05 14:43:50]

On 2015/03/05 14:42:06, Rick Byers wrote:
> LGTM

I kicked off a couple blink tryjobs (since the chromium CQ doesn't currently run
layouttests), but I don't see how this patch could cause any issues (just trying
to get in the habit of running layoutests on all eventsender changes since we've
had some painful regressions recently).

[Miyoung Shin, 2015-03-06 01:02:39]

On 2015/03/05 14:43:50, Rick Byers wrote:
> On 2015/03/05 14:42:06, Rick Byers wrote:
> > LGTM
> 
> I kicked off a couple blink tryjobs (since the chromium CQ doesn't currently
run
> layouttests), but I don't see how this patch could cause any issues (just
trying
> to get in the habit of running layoutests on all eventsender changes since
we've
> had some painful regressions recently).

Can I land it now ? we got the success result from tryjobs.

[Rick Byers, 2015-03-06 01:08:24]

On 2015/03/06 01:02:39, Miyoung Shin wrote:
> On 2015/03/05 14:43:50, Rick Byers wrote:
> > On 2015/03/05 14:42:06, Rick Byers wrote:
> > > LGTM
> > 
> > I kicked off a couple blink tryjobs (since the chromium CQ doesn't currently
> run
> > layouttests), but I don't see how this patch could cause any issues (just
> trying
> > to get in the habit of running layoutests on all eventsender changes since
> we've
> > had some painful regressions recently).
> 
> Can I land it now ? we got the success result from tryjobs.

Yep go ahead

[Miyoung Shin, 2015-03-06 01:57:02]

The CQ bit was checked by myid.shin@samsung.com

[commit-bot: I haz the power, 2015-03-06 01:58:16]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/928973002/20001

[commit-bot: I haz the power, 2015-03-06 02:31:36]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-03-06 02:32:41]

Patchset 2 (id:??) landed as
https://crrev.com/12c7fea23d6077af7d6a2ea6f8fc4680cfaf23d1
Cr-Commit-Position: refs/heads/master@{#319393}