Fixes a race in BluetoothControllerImpl that could lead
to a ConcurrentModificationException, by moving the manipulation
of the callbacks array to the handler.
Change-Id: I7c3be74d239d4706f55357f54e539b7b6439e76d
Fixes: 29390132
This will restart the BT stack when it detects a transition
into OFF state while the user enable flag (mEnable) is set.
Bug: 29363429
Change-Id: I9839119b34c4694ad92e96240c6989008b2f8d52
Fixes: 29072773
By using computeFrameTime AnimationContext would
potentially end up modifying the latest vsync if
a very-slow frame was received from the UI thread.
This could potentially desync animations that were
RT & UI thread 'synchronized', but more significantly
it would confuse the swap chain which tries to only
draw one frame per vsync causing unneccessary frame
drops.
Change-Id: Ibd2ec3157ce32fee1eec8d56837c45a35e622895
When the heads up disappeared from the shade, it's layout would
switch to the expanded state. We're now delaying this until
the animation is done.
Change-Id: I6c5c970f6b471fb10b197922dfea520b07c84abc
Fixes: 29318819
currently with one data point the estimator is biased high. If we don't
have any data we want to be biased low to allow the first interaction
to be an isolated burst.
Also add metrics for enqueue rate to make it easier to debug this
code in the future (currently we only have metrics after the
rate limiter).
Bug: 29379789
Change-Id: I7b1d379290a55f120d039fd29d1bc0abbc6d4932
When we're looking for a task during adjacent launch, we need to
perform full intent filter comparison instead of just comparing
components.
Bug: 29424577
Change-Id: I10f271986a8dc405e28a053e4cb122c1e9aa6ee0
It was possible for the state to be updated in the background
but before the callback was actually registered and never propagate
to the UI. Call the callback directly after it is added (from the
QS thread) to ensure that the current state gets sent out on tiles
being added.
Change-Id: Idd76dce5cbc9dc9c99f9b3dc13998751fd170582
Fixes: 29419435
We depend on a layout listener for when we ask for focus, but layout
sometimes doesn't happen. Make it always happen.
Change-Id: I7c805470c6ce9863bf7eac627f59d49191ce8946
Fixes: 29126464
ag/1037916 added logic to only report window movement to the client
if the window doesn't have a task or its stack is bounds animating.
This causes problems on the client side for window that don't fall
into this category as they will have the wrong information of where
they are no screen.
Bug: 29093176
Change-Id: I958174af430f2c4003a1c0a74956964d209c0e4a
We might have a pending MSG_RESIZED_REPORT, but if it's executed after
relayoutWindow, mPendingInsets will already be the new value and it'll
not forceLayout. So we need to forceLayout here to make sure the measure
cache is cleared.
bug: 29391054
Change-Id: I73793b1427b89e75700369ec3b37053a6a732f0d
Make the activity manager look at the "killed" state of a process
before deciding it is in a state where a new client can attach to
a content provider. Hopefully this will get rid of some races
around attaching to providers and processes being killed.
Also add a little more information when killing a client of a
provider describing the oom adj of that provider's process.
Change-Id: I4efe5daf0a96e4663d5df88827ef401b1311de3f
