ozone: evdev: Make cancellation of repeats on key up more robust

ozone: evdev: Make cancellation of repeats on key up more robust

Key releases enqueue a task on KeyboardEvdev that dispatches the release
ui::Event as well as cancels any scheduled repeats for that key. The
release itself is enqueued from the dedicated evdev thread and so can be
expected to arrive in a timely manner.

Unfotunately, due to the way base::Timer works, this is not robust when
the UI thread is very busy. The base::Timer object posts a delayed task,
and upon running checks to see if enough real time has passed (see
Timer::RunScheduledTask). If so, it runs the timer callback immediately.
This is fine, except that base::Timer also reuses scheduled tasks from
previous calls to Start() if they are still in the future (see
Timer::Reset). This can mean the timeout callback runs from a task that
was scheduled before the current timer was started.

The net effect of this is that the timer task may run first even if
(initial timer start + delay) > (key release task timestamp). The message
loop queue itself would have run the tasks in the correct order.

So, if we simply use PostDelayedTask() ourselves, there's no cleverness
in base::Timer to worry about and the tasks run in the needed order.

BUG=473446
TEST=Press Ctrl-W to close a tab with lots of activity (multiple copies
  of gmail loading, etc). Only one tab closes. Also tested with
  simulated jank (1s delay in tab closure).

Committed: https://crrev.com/50e43d96539246fe55b282e5e161b12ed39c42eb
Cr-Commit-Position: refs/heads/master@{#324081}

[spang, 2015-04-07 18:03:07]

spang@chromium.org changed reviewers:
+ alexst@chromium.org, kpschoedel@chromium.org

[alexst (slow to review), 2015-04-07 18:20:58]

lgtm with a nit.

https://codereview.chromium.org/1052273006/diff/1/ui/events/ozone/evdev/keybo...
File ui/events/ozone/evdev/keyboard_evdev.cc (right):

https://codereview.chromium.org/1052273006/diff/1/ui/events/ozone/evdev/keybo...
ui/events/ozone/evdev/keyboard_evdev.cc:173: DispatchKeyRepeat(repeat_key_);
Probably worth just moving DispatchKey from DispatchKeyRepeat here since it's
only used in one place.

[kpschoedel, 2015-04-07 18:24:14]

LGTM 2.0

[spang, 2015-04-07 18:28:50]

https://codereview.chromium.org/1052273006/diff/1/ui/events/ozone/evdev/keybo...
File ui/events/ozone/evdev/keyboard_evdev.cc (right):

https://codereview.chromium.org/1052273006/diff/1/ui/events/ozone/evdev/keybo...
ui/events/ozone/evdev/keyboard_evdev.cc:173: DispatchKeyRepeat(repeat_key_);
On 2015/04/07 18:20:58, alexst wrote:
> Probably worth just moving DispatchKey from DispatchKeyRepeat here since it's
> only used in one place.

Done.

[spang, 2015-04-07 18:31:05]

The CQ bit was checked by spang@chromium.org

[spang, 2015-04-07 18:31:05]

The patchset sent to the CQ was uploaded after l-g-t-m from alexst@chromium.org,
kpschoedel@chromium.org
Link to the patchset: https://codereview.chromium.org/1052273006/#ps20001
(title: "merge DispatchKeyRepeat")

[spang, 2015-04-07 18:31:17]

The CQ bit was unchecked by spang@chromium.org

[spang, 2015-04-07 18:31:17]

The CQ bit was checked by spang@chromium.org

[commit-bot: I haz the power, 2015-04-07 18:31:46]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1052273006/20001

[commit-bot: I haz the power, 2015-04-07 18:56:21]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-04-07 18:57:53]

Patchset 2 (id:??) landed as
https://crrev.com/50e43d96539246fe55b282e5e161b12ed39c42eb
Cr-Commit-Position: refs/heads/master@{#324081}