[Android] Properly guard selection updates in RWHVAndroid

[Android] Properly guard selection updates in RWHVAndroid

One external invariant of RWHVAndroid is that if it has a valid attached
ContentViewCore, its TouchSelectionController will be non-null. However,
there are no guarantees that these lifetimes hold for methods *internal*
to RWHVAndroid. One such violation came when the frame had a non-zero
lock count when the attached ContentViewCore was changed; releasing the
surface lock caused a frame metadata update, in turn accessing the
TouchSelectionController that had already been reset.

As a workaround, guard against this and possible future violations by
always checking the controller for validity before using it.

BUG=419619
Committed: https://crrev.com/a9ee8a34791e66dc77341b69220868c5f90b3f50
Cr-Commit-Position: refs/heads/master@{#299792}

[jdduke (slow), 2014-10-15 22:04:53]

jdduke@chromium.org changed reviewers:
+ aelias@chromium.org

[jdduke (slow), 2014-10-15 22:04:53]

aelias@: PTAL, thanks. There's a fix which changes the order of the

    selection_controller_.reset();
    ReleaseLocksOnSurface();

calls in RWHVAndroid::SetContentViewCore, however, it seems a little fragile to
assume that the frame metadata will *never* be updated as a side effect of other
operations that take place in that method.

[aelias_OOO_until_Jul13, 2014-10-15 22:09:24]

lgtm

[jdduke (slow), 2014-10-15 22:10:54]

The CQ bit was checked by jdduke@chromium.org

[commit-bot: I haz the power, 2014-10-15 22:12:04]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/653883003/20001

[commit-bot: I haz the power, 2014-10-16 00:31:16]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2014-10-16 00:31:52]

Patchset 2 (id:??) landed as
https://crrev.com/a9ee8a34791e66dc77341b69220868c5f90b3f50
Cr-Commit-Position: refs/heads/master@{#299792}