Regularly poll the X11 event queue.

Regularly poll the X11 event queue.

This change ensures that we deliver events which have been enqueued
in memory in a timely fashion.  Previously they would just sit there
and rot until something happened to kick the display connection file
descriptor resulting in greyed out windows and delayed input events.

Added the "--dispatch-one-event-per-loop" switch to experiment with
a mode where the event loop only dispatches one X11 event each
iteration.  By default, we dispatch all pending events to minimize
input dispatch latency.

Bug: 661

R=abarth@google.com, jamesr@chromium.org

Committed: https://chromium.googlesource.com/external/mojo/+/6f37494c4bbcdc32e0a869bf0f08c6f1e12fdabc

[jeffbrown, 2016-04-26 02:43:14]

Description was changed from

==========
Poll the X11 event queue on each loop of the message pump.

This change ensures that we deliver events which have been enueued
in memory in a timely fashion.  Previously they would just sit there
and rot until something happened to kick the display connection file
descriptor resulting in greyed out windows and delayed input events.

Bug: 661

BUG=
==========

to

==========
Regularly poll the X11 event queue.

This change ensures that we deliver events which have been enueued
in memory in a timely fashion.  Previously they would just sit there
and rot until something happened to kick the display connection file
descriptor resulting in greyed out windows and delayed input events.

Bug: 661

BUG=
==========

[jeffbrown, 2016-04-26 02:43:37]

jeffbrown@google.com changed reviewers:
+ abarth@google.com, jamesr@chromium.org

[abarth, 2016-04-26 16:54:35]

lgtm

[jamesr, 2016-04-28 21:08:25]

Here's a proposal - implement both the drain-everything (or drain-all-pending)
and drain-one behaviors and configure a switch to toggle between the two
behaviors so that if we observe bad behavior of the system we can switch to the
other behavior without needing to recompile (and thus roll and redeploy binaries
elsewhere).  Command line switches are pretty easy to wire through our system so
this should provide enough control for folks attempting to give a demo or
whatnot to try both things out.

How does that sound?

[jeffbrown, 2016-05-04 20:30:30]

Description was changed from

==========
Regularly poll the X11 event queue.

This change ensures that we deliver events which have been enueued
in memory in a timely fashion.  Previously they would just sit there
and rot until something happened to kick the display connection file
descriptor resulting in greyed out windows and delayed input events.

Bug: 661

BUG=
==========

to

==========
Regularly poll the X11 event queue.

This change ensures that we deliver events which have been enqueued
in memory in a timely fashion.  Previously they would just sit there
and rot until something happened to kick the display connection file
descriptor resulting in greyed out windows and delayed input events.

Added the "--dispatch-one-event-per-loop" switch to experiment with
a mode where the event loop only dispatches one X11 event each
iteration.  By default, we dispatch all pending events to minimize
input dispatch latency.

Bug: 661
==========

[jeffbrown, 2016-05-17 00:25:46]

@jamesr, did you want to take another look at this now that I implemented your
proposal?  Thanks!

[jamesr, 2016-05-17 00:27:45]

Sorry, I thought this landed ages ago.  I must have confused this with the
dependent patch.

lgtm

[jeffbrown, 2016-05-18 00:26:41]

Description was changed from

==========
Regularly poll the X11 event queue.

This change ensures that we deliver events which have been enqueued
in memory in a timely fashion.  Previously they would just sit there
and rot until something happened to kick the display connection file
descriptor resulting in greyed out windows and delayed input events.

Added the "--dispatch-one-event-per-loop" switch to experiment with
a mode where the event loop only dispatches one X11 event each
iteration.  By default, we dispatch all pending events to minimize
input dispatch latency.

Bug: 661
==========

to

==========
Regularly poll the X11 event queue.

This change ensures that we deliver events which have been enqueued
in memory in a timely fashion.  Previously they would just sit there
and rot until something happened to kick the display connection file
descriptor resulting in greyed out windows and delayed input events.

Added the "--dispatch-one-event-per-loop" switch to experiment with
a mode where the event loop only dispatches one X11 event each
iteration.  By default, we dispatch all pending events to minimize
input dispatch latency.

Bug: 661

R=abarth@google.com, jamesr@chromium.org

Committed:
https://chromium.googlesource.com/external/mojo/+/6f37494c4bbcdc32e0a869bf0f0...
==========

[jeffbrown, 2016-05-18 00:26:43]

Committed patchset #5 (id:80001) manually as
6f37494c4bbcdc32e0a869bf0f08c6f1e12fdabc (presubmit successful).