Widgets get masked with separate remove views when suspended. Need to
update them with the correct pending intent. Moved around the logic
creating the suspended alert activity to make it reusable.
Test: Manually, "pm suspend <app-with-widget>", followed by tap on
widget
Existing tests pass:
atest GtsSuspendAppsTestCases
atest com.android.server.am.ActivityStartInterceptorTest
Bug: 79474038
Change-Id: Ic75ee762434bf92c54bda307e9cb6018f94a914c
Fixes splitscreen from the right for launching splitscreen from
fullscreen app including pixel launcher and 3rd party launchers. With
3rd party launchers, the experience is the same as Android O. With pixel
launcher since launcher and recents are merged, it will split to
minimized mode. When pixel launcher is used, the recents stack is
ignored.
Fixes: 73250406
Fixes: 78247419
Test: use split screen; invoke command with supportapp.apk in b/78247419
adb shell am broadcast -a launcher-acc --ei cmd 1
Change-Id: I70b0551c517465b3a059ff7d0641c27f71e8114e
- Stop using individual lock instances and use use the power manager lock.
- Make sure not to access external components with the lock held,
*except for settings provider*, which is already touched by the power manager
and the battery service
(I'm planning to re-organize things for Q: b/79580230)
Fixes: 79486713
Test: atest /android/pi-dev/frameworks/base/services/tests/servicestests/src/com/android/server/power/batterysaver/Battery*Test.java /android/pi-dev/frameworks/base/services/tests/servicestests/src/com/android/server/power/*Test.java
Test: Manual test with ./development/scripts/battery_simulator.py with toggling
battery saver on and off and then watch
adb logcat -b all | grep -P '(battery_saving_stats|battery_saver_mode)'
Change-Id: I8ecdfefe978de348ee979d765c82e3d8f0280082
Fallback would not be registered/drawn if background was explicitly set before calling
setContentView and the area that was supposed to be covered by fallback ended up black.
mBackgroundDrawable does not control fallback, so setting fallback should not be dependent on
it's existence.
Change-Id: I7da85e8c61dababa3a7ce930238ea0f283d19b32
Fixes: 68822553
Test: atest PhoneWindowTest
Fix some bugs, add the ability to monitor state changes, improve
dumpsys output to help debugging, add a new check API that allows
the caller to get the real state.
Bug: 78480444
Test: atest FrameworksServicesTests:AppOpsServiceTest
Test: atest CtsPermissionTestCases:AppOpsTest
Change-Id: I3d41be9968c1d95a1456f4052da958ea64aa068d
Pass an explicit flag to the runtime when we need to profile the system
server. This ensures that we only start the ProfileSaver and allocate the
code cache when needed.
Test: manual, on a device with system server profiling enabled
Bug: 74081010
(cherry picked from commit 8eb891be6f)
Merged-In: I756682a9644796afb53f0c010d2bd2764ea5a828
Change-Id: I756682a9644796afb53f0c010d2bd2764ea5a828
In minimized mode there is only 1 split snap target compared to regular
portrait that has 30%, 50%, 70% targets. Update the accessibility options
for minimized mode to reflect the correct targets. Also removed the
middle target if there is only 1 split snap target.
Test: using talkback, see custom options on divider handle in minimized
mode
Fixes: 78304300
Change-Id: I8421499616cf0fe503c1f6a081a2151098640a9c
Also remove all references to old loadSafeLabel method to prevent
proliferation of old method via copy+paste.
The implementation assumes that there are three cases:
- Short labels that don't have anything wrong with them
- Labels that are fine, but are a little too long
- Intentionally bad label that try to break stuff and slow things down.
In the first two cases no characters are marked for removal, in the
third case we have an implementation that does not use a lot of memory
and has linear performance when there are a lot of bad characters.
Test: gts-tradefed run gts-dev -m GtsContentTestCases
Bug: 77964730
Change-Id: I3feb17b2a12018cd5407c88fe3603f2ebbc9d14e
Test: manual, applied password length restriction via TestDPC and
set a new password via settings (see the other CL in the topic)
Bug: 30558331
Change-Id: I51302bad02604c8fc0def37dc33a5837f29ab858
Current limitation is that we cannot distinguish between key events
from Window#injectInputEvent() and key presses from physical devices.
Bug: 79185725
Test: Manually verified with the test app, and Cts
Change-Id: I1c3da4a963238512eec8b57fbc2d825c7436b140
* changes:
Revert "Rotate IMEs (subtypes) by Meta+Space."
Revert "Show toast when subtype is rotated by Meta-Space."
Revert "Shift+Meta+Space should reverse-rotate subtypes."
Revert "Shift+Meta+Space should reverse-rotate subtypes part 2."
Revert "Make sure that Toast is always shown by Meta-Space."
Certain apps use empty sender names, which looks very
broken. Let me do this one last workaround...
Change-Id: I958319d1e4f9387e5e05afb304c05bef55c3e2fb
Fixes: 73139872
Test: receive a message with empty sender, observe visually
Update the logic for system server profiles to work without disabling
SELinux policies.
Test: check that system_server profile is created without disabling
SElinux
Bug: 74081010
Change-Id: I8aa1e57da0f8059bcfb149fe7366d8278b4bf516
Updating a remoteInputHistory message to non-history
was generating a weird animation because we would
not reuse the view if the remoteInputHistory state
changed.
Bug: 77957392
Test: visual, observe normal animations
Change-Id: Iac869159411cc57c90476ef684a05e8547c14aa4
The historic message groups are now also gone,
not just the messages
Change-Id: Ib7a454809425c18f5c2fd5905b3dae9ec5ec846f
Fixes: 72556979
Test: visually, add historic messages, observe group is gone too
Pass an explicit flag to the runtime when we need to profile the system
server. This ensures that we only start the ProfileSaver and allocate the
code cache when needed.
Test: manual, on a device with system server profiling enabled
Bug: 74081010
Change-Id: I5647e0bfb47d31c0e39bb3f59650cc956f49c4db
