The problem turns out to have been a deep weirdness in the way that keyguard
and incall interact. Incall gets relaunched when the keyboard is opened/closed,
which transiently exposes keyguard with its nosensor orientation demands, and
that plus the long keyguard-hide animation was leaving incall in a bad state
from which the window manager didn't try to recover.
We now disregard animating-towards-hidden windows [i.e. keyguard] when running
through the app tokens to determine what orientation should be, and do not do
configuration calculations at all while the display is frozen. There can still
be a transient state in which incall is drawn in landscape, but things proceed
from there to relaunch it back into the proper portrait orientation, and it
ends up in the right state in the end.
Change-Id: I0d74ee19064b6d7f65600976f1b5b16b7ec36f31
onScrollChanged() is called so that the View system
can do what it needs to adjust.
As WebView's onScrollChanged will call sendOurVisibleRect(),
we don't need to call it explicitly any more.
Fix http://b/issue?id=2438457
Look for corrupted bytes and replace with ?
If the problem is in the parsing code, then this will not help.
Change-Id: If33643f126a4099d4113e37760c6ef74683fe97e
Now that wifi start is asynchronous
at the time of bring up, make sure
Wifi is not started if in airplane
mode.
Bug: 2567652
Change-Id: I947b7c8480029973bcbf028f6143aabbc88c9793
Allow the user to start scrolling back into content immediately
after dragging past the end of content.
Remove some dead code.
Change-Id: Ife8f60499ac58179ba1e2500ca8745916392a4bc
This fixes a problem where the pattern unlock area shrank on
mdpi devices.
This change is required since some devices (mdpi) don't have
enough vertical room to show a square LockPatternView and shrink
the area to match the minimum of width and height. To fix the
problem and minimize changes to existing platforms, the parameter
is configurable and has a default in values/donottranslate that
emulates the old behavior. New behavior for mdpi devices is
specified in values-port-mdpi/donottranslate, which is the
only place we currently see the problem.
Tested:
mdpi device in both orientations
hdpi device in both orientations
hdpi device in both orientations with keyboard open/closed
Change-Id: I3b9a5ef1d9b5e70f7b1823c9f48a3938f52cc7a4
Move all tests for android.* classes from tests/AndroidTests and
tests/CoreTests into framework/base/<core|graphics>/tests.
Consolidate all tests for java.* classes to tests/CoreTests.
Eventually hopefully these will be moved to dalvik/ somewhere.
Remove tests/AndroidTests entirely.
Change-Id: I86584d086ab7bd045bb38a10b699907805298a95
be sent in RecognizerIntent.EXTRA_CALLING_PACKAGE. This never should have
been using toShortString because that includes curly braces and is not
correctly parsed by ComponentName.unflattenFromString.
This was needed as part of http://b/2546173
Change-Id: If34a9273cbcae13a75347be0b6b7ef307134ed40
MediaMetadataRetriever uses a single static lock for all operations.
This effectively serializes all metadata retrieval operations in a
single process. This patch uses the object level lock for all normal
operations and only uses the static lock to serialize calls to
release.
Change-Id: I81c9f234c2f0007a26d18e1398c709b41a4dbbd7
will have a reference of the current window's main
WebView. It is only non-null if the WebView's window
has the focus.
Extract setActive() from onWindowFocusChanged() so
that onAttachedToWindow() can call it directly. The
old way has a mis-matching call to onWindowFocusChanged.
Fix http://b/issue?id=2559152
interfere with one another.
Root cause was not caching all the params that were needed.
This change fixes that bug as well as related bugs for remembering
the default engine and for making sure that the right engine is
loaded when checking for language availability.
Change-Id: I2a76da8faec8112036e68d27539db444c53a1509
