Route input-related IPCs through compositor thread filter to preserve order

Route input-related IPCs through compositor thread filter to preserve order

The compositor thread input filter redirects some IPC messages in the render
process to the compositor thread for handling in order to accelerate scrolling,
pinch zooming, and related gestures. IPC messages that the compositor thread
cannot handle are redirected to the main thread for normal processing. This
thread hop can result in the IPC messages being received on the render process
main thread in a different order than they were sent from the browser. This can
break things for input-related messages, for instance changing the order of a
MouseCaptureLost and MouseUp event can result in the mouse up not being
delivered to the capturing element.

This forces a compositor thread hop to preserve order for a set of IPC types
that appear (mostly by inspection and guessing) to need their order preserved
relative to input events.

BUG=233365, 231532
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=195329

[jamesr, 2013-04-19 00:34:48]

This fixes the referenced bug and appears to work, but the set of things
filtered is rather ad-hoc.

[piman, 2013-04-19 00:43:29]

lgtm

[darin (slow to review), 2013-04-19 06:13:27]

LGTM

https://codereview.chromium.org/14328027/diff/1/content/renderer/gpu/input_ev...
File content/renderer/gpu/input_event_filter.cc (right):

https://codereview.chromium.org/14328027/diff/1/content/renderer/gpu/input_ev...
content/renderer/gpu/input_event_filter.cc:148: // Save this message for later,
in case we need to bounce it back up to the
nit: new line above this comment

[darin (slow to review), 2013-04-19 06:15:19]

I think a separate IPC class might still be handy.  If those were defined in
their own file, then folks would be more likely to understand when they should
grow the set.  We could call them ViewInputMsg_Foo or something.  Just trying to
make it less errorprone when someone adds another input related message :-/

[commit-bot: I haz the power, 2013-04-19 19:51:10]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/jamesr@chromium.org/14328027/7001

[commit-bot: I haz the power, 2013-04-19 21:05:54]

Retried try job too often on mac_rel for step(s) base_unittests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

[commit-bot: I haz the power, 2013-04-19 22:57:29]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/jamesr@chromium.org/14328027/7001

[jamesr, 2013-04-19 22:57:38]

On 2013/04/19 21:05:54, I haz the power (commit-bot) wrote:
> Retried try job too often on mac_rel for step(s) base_unittests
>
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

Died due to some .isolate cruft that maruel cleaned up.  Trying again.

I'll make an attempt at making a separate IPC class as a follow-up but I want to
make sure this bug is fixed first.

[commit-bot: I haz the power, 2013-04-19 23:16:52]

Change committed as 195329