[PATCH 11/11] ozone: evdev: Add dedicated events thread

ozone: evdev: Add dedicated events thread

This creates a new thread for polling evdev on, and moves the
InputDeviceFactoryEvdev object onto that thread. We'll use this new
thread to avoid UI-thread jank affecting the cursor.

BUG=449710
TEST=boot link_freon & move cursor with simulated UI jank.
     cursor moves while UI thread is stalled.

Committed: https://crrev.com/be9342d2b87b41a865fa9672577b98d1ef8e2816
Cr-Commit-Position: refs/heads/master@{#313981}

[spang, 2015-01-23 22:02:21]

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

[spang, 2015-01-27 19:31:36]

spang@chromium.org changed reviewers:
+ piman@chromium.org

[spang, 2015-01-27 19:31:37]

alexst@ for in-depth review

+piman@ for high-level review
 (+CC for prep work)

This series finishes the work moving cursor latency off of UI. The general idea
is moving I/O & early event processing (incl. cursor movement) to a new events
thread, with additional processing (modifiers, keyboard layout, etc) on UI. Most
of the series is preparatory work separating our state into two distinct pieces.

[piman, 2015-01-28 00:07:26]

One thing, but otherwise LGTM in principle

https://codereview.chromium.org/873833002/diff/20001/ui/events/ozone/evdev/ev...
File ui/events/ozone/evdev/event_thread_evdev.cc (right):

https://codereview.chromium.org/873833002/diff/20001/ui/events/ozone/evdev/ev...
ui/events/ozone/evdev/event_thread_evdev.cc:33: ~EvdevThread() override {}
You need to call Stop() here to avoid subtle races.

~Thread will call Stop(), joining the thread, but that will only happen after
fields in this class are destructed. So there is a potential for the thread to
still execute tasks during the destruction of this class. (see warning in
base::Thread).

[spang, 2015-01-28 00:20:14]

https://codereview.chromium.org/873833002/diff/20001/ui/events/ozone/evdev/ev...
File ui/events/ozone/evdev/event_thread_evdev.cc (right):

https://codereview.chromium.org/873833002/diff/20001/ui/events/ozone/evdev/ev...
ui/events/ozone/evdev/event_thread_evdev.cc:33: ~EvdevThread() override {}
On 2015/01/28 00:07:26, piman (Very slow to review) wrote:
> You need to call Stop() here to avoid subtle races.
> 
> ~Thread will call Stop(), joining the thread, but that will only happen after
> fields in this class are destructed. So there is a potential for the thread to
> still execute tasks during the destruction of this class. (see warning in
> base::Thread).

Done.

Oops, missed the warning in all caps :(

[piman, 2015-01-28 01:06:28]

lgtm

[alexst (slow to review), 2015-01-28 21:06:13]

lgtm, modulo rebases and nitty comments from previous patches.

https://codereview.chromium.org/873833002/diff/40001/ui/events/ozone/evdev/ev...
File ui/events/ozone/evdev/event_factory_evdev.cc (right):

https://codereview.chromium.org/873833002/diff/40001/ui/events/ozone/evdev/ev...
ui/events/ozone/evdev/event_factory_evdev.cc:302:
scoped_ptr<DeviceEventDispatcherEvdev> dispatcher(
per comments in previous patches, can this be dispatcher_proxy for clarity about
it's cross thread nature?

[spang, 2015-01-28 21:37:32]

https://codereview.chromium.org/873833002/diff/40001/ui/events/ozone/evdev/ev...
File ui/events/ozone/evdev/event_factory_evdev.cc (right):

https://codereview.chromium.org/873833002/diff/40001/ui/events/ozone/evdev/ev...
ui/events/ozone/evdev/event_factory_evdev.cc:302:
scoped_ptr<DeviceEventDispatcherEvdev> dispatcher(
On 2015/01/28 21:06:12, alexst wrote:
> per comments in previous patches, can this be dispatcher_proxy for clarity
about
> it's cross thread nature?

Done.

[spang, 2015-01-30 19:59:14]

The CQ bit was checked by spang@chromium.org

[commit-bot: I haz the power, 2015-01-30 20:00:13]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/873833002/80001

[commit-bot: I haz the power, 2015-01-30 20:38:48]

Committed patchset #5 (id:80001)

[commit-bot: I haz the power, 2015-01-30 20:39:57]

Patchset 5 (id:??) landed as
https://crrev.com/be9342d2b87b41a865fa9672577b98d1ef8e2816
Cr-Commit-Position: refs/heads/master@{#313981}