Revert of Revert of ozone: evdev: Fix spurious key repeats during UI jank (patchset #1 id:1 of https://codereview.chromium.org/1121233008/)

Reason for revert:
Can't be the cause of the test failure

It is not even compiled for Linux ChromiumOS GN (dbg)

Original issue's description:
> Revert of ozone: evdev: Fix spurious key repeats during UI jank (patchset #1 id:1 of https://codereview.chromium.org/1121293002/)
>
> Reason for revert:
> A couple of Linux ChromeOS interactive_ui_tests started failing here:
>
> http://build.chromium.org/p/chromium.chromiumos/builders/Linux%20ChromiumOS%20Tests%20%281%29/builds/2341
>
> They are timing out on a touch event, so this seemed like the most likely cause. I have been unable to reproduce locally and the failures are flaky/inconsistent (but pretty common).
>
> I apologize for reverting this if it is not the cause.
>
> Original issue's description:
> > ozone: evdev: Fix spurious key repeats during UI jank
> >
> > There's no guarantee that a delayed task runs after other pending tasks
> > with lower delayed_run_time, if those tasks were posted later. The
> > reason for this is that there are two queues in the message loop: an
> > incoming queue and a work queue. We drain the incoming queue as little
> > as possible, to avoid synchronization overhead (see ReloadWorkQueue()).
> >
> > This breaks some assumptions in the key repeat code. Fix this by doing a
> > second non-delayed post in OnRepeatTimeout(). Those are definitely
> > guaranteed to run in FIFO order, so it should fix spurious repeats once
> > and for all.
> >
> > BUG=481780
> > TEST=Use multi-profile switch hotkeys with repeat delay set to 200ms.
> >   Pressing the sequence once does not occasionally switch twice.
> >
> > Committed: https://crrev.com/67182c7958f5af3be5151e5efb4422240895f09b
> > Cr-Commit-Position: refs/heads/master@{#328166}
>
> TBR=kpschoedel@chromium.org,spang@chromium.org
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> BUG=481780
>
> Committed: https://crrev.com/6f3e50b69083c51c3d6eca4913cb164e390fded6
> Cr-Commit-Position: refs/heads/master@{#328266}

TBR=kpschoedel@chromium.org,stevenjb@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=481780

Review URL: https://codereview.chromium.org/1128463005

Cr-Commit-Position: refs/heads/master@{#328323}

2 files changed, 13 insertions, 0 deletions

diff --git a/ui/events/ozone/evdev/keyboard_evdev.cc b/ui/events/ozone/evdev/keyboard_evdev.cc
index 2c14f600..bc33354 100644
--- a/ui/events/ozone/evdev/keyboard_evdev.cc
+++ b/ui/events/ozone/evdev/keyboard_evdev.cc
@@ -176,6 +176,18 @@ void KeyboardEvdev::OnRepeatTimeout(unsigned int sequence) {
   if (repeat_sequence_ != sequence)
     return;
 
+  // Post a task behind any pending key releases in the message loop
+  // FIFO. This ensures there's no spurious repeats during periods of UI
+  // thread jank.
+  base::ThreadTaskRunnerHandle::Get()->PostTask(
+      FROM_HERE, base::Bind(&KeyboardEvdev::OnRepeatCommit,
+                            weak_ptr_factory_.GetWeakPtr(), repeat_sequence_));
+}
+
+void KeyboardEvdev::OnRepeatCommit(unsigned int sequence) {
+  if (repeat_sequence_ != sequence)
+    return;
+
   DispatchKey(repeat_key_, true /* down */, true /* repeat */,
               EventTimeForNow(), repeat_device_id_);
 
diff --git a/ui/events/ozone/evdev/keyboard_evdev.h b/ui/events/ozone/evdev/keyboard_evdev.h
index 2dd0792..8a5108e 100644
--- a/ui/events/ozone/evdev/keyboard_evdev.h
+++ b/ui/events/ozone/evdev/keyboard_evdev.h
@@ -58,6 +58,7 @@ class EVENTS_OZONE_EVDEV_EXPORT KeyboardEvdev {
   void StopKeyRepeat();
   void ScheduleKeyRepeat(const base::TimeDelta& delay);
   void OnRepeatTimeout(unsigned int sequence);
+  void OnRepeatCommit(unsigned int sequence);
   void DispatchKey(unsigned int key,
                    bool down,
                    bool repeat,