scroll content out of view. Scrolling will slow down halfway to the
barrier point. API added in View. AbsListView, ScrollView,
HorizontalScrollView all use this API. Overscrolling uses haptic
feedback. Added scroll barrier pattern to config.xml.
This will allow Market and others to find out what the "real" name
of a package is, when it is currently running under the old name of
a previously installed version.
Set a 70ms debounce time - i.e., if you spend less than 70ms on the touch-up key,
which was less time than the time spent on the previous key (assuming they weren't the same),
then don't emit the last key. Use the previous key that you lingered on for longer.
Force the event-target bookkeeping to reset when an app explicitly
acknowledges event receipt. Furthermore, notify the event dispatcher
whenever a new window is created, even if there is nominally an
existing event target window.
These changes in tandem address a subtle race bug in which the event
dispatcher believes that there is an event recipient that it is still
waiting for even when the event in question has been acknowledged;
this results in a spurious ANR. There were checks in the existing code
that wound up doing the right thing in the case of windows becoming
hidden, but the case of non-fullscreen activities floating over them
[unavailable for input but still visible] exposed the race condition.
Bug: 2432828
Change-Id: I29fb741aace34736b029f9ba43f2c43184366258
The forgotten parts from Id72f718c / d72f718c9c. Whoops.
Tested by watching a device's logcat -b events and observing no huge
or negative values. And this time with the right system.img file,
even!
Previously, SQLiteDatabase was using Debug.threadCpuTimeNanos(), which
doesn't include I/O time (user-perceived latency), and ContentResolver
was using System.currentTimeMillis(), which didn't account for deep
sleeps.
Now both are consistently using SystemClock.uptimeMillis().
This is merged with force-lock. These both allow effectively the same thing,
so there is no reason to junk up the user experience with them as separate
entities.
The SSLSessionCache attempts to create a directory through the context
In MockContext, this throws an UnsupportedOperationExcetion
This CL changes MockContext2 in ProviderTestCase2 to allow getDir
to be called, but the directory is named so it will be seperated from
a directory created by a regular context
to be used by anyone implementing a voice recognition service. Also define
a new <recognition-service> styleable to be used in such service's metadata
xml.
Still to do: Change VoiceSearch's GoogleRecognitionService to respond to this
intent, and actually use this intent instead of ACTION_RECOGNIZE_SPEECH here
in RecognitionService.
