If we intersect each window in the application with
our screenshot bounds we will end up reduced to the bounds
of the smallest window. Modify logic to just check for
intersection without modifying the bounds.
Bug: 27996291
Change-Id: Ib0ebb6bd02a313f67f1562dabcfb394efa8f16cf
- When SysUI crashed, make sure state is up-to-date
- When divider gets invisible, update state because animations
won't finish
- Make sure to keep mMinimizedDock up-to-date in all cases, even
if docked stack is already gone.
- If docked stack is going away, reset minimized state
- Disallow undocking when minimized.
Bug: 27588187
Bug: 27486464
Change-Id: I6ea48562fae3a8df038a690a488580d758681370
Add a shell command for switching webview implementation so that this
can be done automatically - without interacting with a Dev Setting.
Bug: 27540566
Change-Id: I3c4acaa1b8f4ba24c637ef8a1d48cc7b2616bd91
If a DimLayer becomes fullscreen while it is animating out,
then it will not be the shared fullscreen dim layer. However,
state.continueDimming will be false (as we are ending the Dim),
so around DimLayerController.cpp L191 we will identify it as
fullscreen, skip animating it for now, and not set fullScreenAndDimming
to the index (because continueDimming is false). fullScreenAndDimming
will be -1 so we will now animate the dimmer at index "fullscreen"
this will simply be the last fullscreen state we found, which would be
fine if the state we had skipped was the shared fullscreen dim layer,
but it wasn't. Ensure that we identify the layer as fullscreen and the
shared fullscreen dim layer.
Bug: 27775427
Change-Id: I49d4f59bb09af6a7e5bd666e499be1c7e11ee766
We need to leave space WINDOW_LAYER_MULTIPLIER between Z-order
adjacent normal windows, for things such as Dim Layers, etc. The
existing logic was breaking DIM_UNDER DimLayers in DockedStack. Consider
two normal windows, the second on top and with a DIM_UNDER. If we are
docked we will have restacked all the docked windows at HighestLayer+1
HighestLayer+2, etc...and so our second normal window is only one higher
than our first one. So when we put a dim under the second window
we end up with the same layer and the output is undefined.
Bug: 27721903
Change-Id: Ic27e0a10da2d1ab4a6058e0755ad4409b29527eb
This follows up to my previous CL [1], with which we changed global
hardware key combinations regarding how multilingual user can manage
multiple keyboard layouts / input methods. Now we offer Meta-Space to
rotate input method (subtype), while we offered Shift-Space to rotate
hardware keyboard layouts previously.
One thing that is lost during above transition is an indication when the
such a key combination takes effect. Actually there was a toast that
shows new keyboard layout name, which is now lost.
With this CL, we bring back a toast so that we can show the new input
method (subtype) name. Note that the toast will be shown if all of the
following conditions are met.
- The input method (subtype) is rotated by a hardware key combination.
- IME window does not have IME_VISIBLE bit. Otherwise, showing a toast
is likely to cause UI overlap with the current IME's window.
[1]: I4005692215edfcf8bed3e86b1e07000148f986f5
ae61f7118a92e097e854c840d5726c0920f5db0e
Bug: 27547054
Change-Id: If3fe17adbffe4c1125783fc77fed3cfe78fc7933
Move more code from WebViewUpdateService to utility classes (methods
handling settings and uninstalling/enabling/disabling packages) to be
overridden during tests.
Also rename system utility class.
Bug: 27635535
Change-Id: If49999fba4fd0962f103f389898fa5ddf19365bd
When the system is shutting down or the screen is going to be off, <Standby>
message will be broadcasted if and only if the setting variable,
Settings.Global#HDMI_CONTROL_AUTO_DEVICE_OFF_ENABLED, is enabled. This change
will make the shutdown case to be checked with the setting properly, too.
Bug: 27781559
Change-Id: Ia6605d6b5627ab1ff67285d72846b618a09548bc
Add shell commands to check on current FBE status and system ready
status. Mark variables without first-class locking as volatile.
Fix bug where UI automation would crash while device was locked by
marking it as forced direct-boot aware.
Bug: 26498834
Change-Id: Ib4dfb9350925e5413f93a09baacf84c62f2ba0ea
Reverting the revert.
The original CL(commit a666d74d4bc7e1298314c516d1309571fb87c212) had a
bug in it. It was calling accountExistsCacheLocked(), while holding
accounts.cacheLock. The function accountExistsCacheLocked, was in turn
calling into getUserAccounts, which acquires mUsers. And this causes
trouble. mUsers is a a lock on all accounts, and hence calling it after
holding accounts.cachelock is calling for trouble. Since the locks are
acquired in other order, it causes a potential deadlock issue, which we
discovered later on.
That bug was fixed by commit 0e592733ecbde2b7e7f2aa92001656dfbcae9641.
We already have useraccount object and should reuse it.
This reverts commit 27d0e1fd660a7e92644ab6a2893ac1149f0c0488. This
commit was unneeded as the bug had been fixed by that time.
Change-Id: I5328c31fd485fd2c1c652cd0e7c2c4bded38a5fd
When the transistion starts and we setTokenVisibilityLocked, mark all
relevant window flags on a starting window that's running exit animation.
This makes sure the starting window is removed when animation finishes.
bug: 27747191
Change-Id: I917ca18583e31db9a3e49ea70aae932c242877b8
Set bounds to null when docked stack is removed and other stacks are
resized to occupy whole display area. It ensures that stacks will have
correct fullscreen status.
Bug: 27857485
Change-Id: Ifbb8de54e6e5af3da1266b60d53656e4278fb75d
When the device was locked the fullscreen stack was laid out
fulllscreen, even though the configuration was half-screen, which
lead to a race between the app drawing and the unlock animation
so sometimes you could see the fullscreen frame when the unlock
animation started.
Instead, only layout in fullscreen if the docked stack is fully
"hidden".
Bug: 27154882
Change-Id: I4ba0c396eb0312c2bf2d911903b68c88d28aae8c
The behavior of this command was not well specified in case of errors,
so it's better to remove it and let clients use 'list wifi-networks'
instead, checking for the value of the 'metered' column.
BUG: 27808364
Change-Id: I3f4216c143b4b2a803b99002585fafe0e3885665
This may or may not happen, but putting in some preventative measure and
extra logging to help debug crashes related to null mSurfaceController.
bug: 27533667
Change-Id: I010147da819402c48efd26a7cc631776052a702d
Also:
- Fix cross profile pinning. Now ShortcutLauncher correctly keeps
track of target package user IDs.
- Add performance stats for package manager calls and show them on
dumpsys.
Bug 27548047
Change-Id: I2472980b2955b6c70947aaaa9b76f75d0616db6e
Update configuration with WM first and check if the stacks need to be
resized due to the update. If so, let activity manager resize the stacks
inline, instead of letting WM schedule another pass of resizeStack. This
way the configuration will be updated to the latest before ensureActivity-
ConfigurationLocked, and we don't need another relaunch there.
bug: 27834014
Change-Id: Ib761a96cada0c3247b0480f18370670c593159da
