If a subsequent screenshot is taken after we bind to the screenshot
service, but before we get the onServiceConnected callback, we
leak the connection to the service, preventing it from ever getting
shut down. Fix this by unbinding the old connection before connecting
with the new one.
Bug: 242862442
Fix: 242862442
Test: forced DisplayPolicy to take two consecutive screenshots
Change-Id: Iffb62170266a2c1bb332bfb416f0f5e19c5236fb
TrustAgents currently show the string "Device was locked manually" if
the trustagent times out, or if lockUser() was called. Changed to show a
more generic string, informing the user that they need to authenticate
with whatever credentials they have set up (instead of using the
trustagent).
Test: Test trustagent timeout, lockUser, verify string appears
Bug: 243710694
Change-Id: I96b113ab3ae63123782517339b0bdb3c6cdec2e2
Disables pooling of SimpleIconFactory objects as it caches
density which could be changed in runtime in tests.
Test: presubmit
Bug: 245354130
Change-Id: I124e02e078ba7e6ed3a42b543698c2758e5d0c0e
This CL consists of following parts:
- TEST_MAPPING files for presubmit.
- Correct CUJ constant definitions in LatencyTracker.
- Use auto-generated enum constants instead of hardcoded
numbers in LatencyTracker.
- Create LatencyTrackerTest to test:
- The constant names of CUJ definitions should be the same with
corresponding enum names.
- The amount of CUJ definitions should be the same with
the amount of enums, means both must be added at the same time.
- The static getNameOfXX() methods return just the name of the
CUJ constant definitions.
- The CUJ constant definitions has 1:1 mapping onto the enums.
Bug: 245250904
Test: check TH if FrameworksCoreTests is triggered
Test: atest FrameTrackerTest InteractionJankMonitorTest
Test: atest LatencyTrackerTest
Change-Id: I0a6f403694039aec6b9ce44254f6c5cab531f765
Merged-In: Ibddd36410a0d4a909522de011f23a337b53d6889
This reverts commit 50c258ca37.
Reason for revert: Relanding after fixes for the failures Fixes:
1. Mark the com.android.perftests.core as profileable from shell so
that we would be able to read /proc/self/io
2. Fix a JDWP test to expect the new behaviour on userdebug builds.
We can no longer attach debugger on profileable apps on userdebug builds
Bug: 240410400, 246616689
Test: System boots
Change-Id: I6fff6f1a683893244bc2d40f930044747befb2ce
Constructors for boxed primitives (Byte, Short, Integer, etc) have
long been deprecated. The preferred pattern is to use valueOf(). d8
fails to convert these instances in TypedProperties, most likely due
to the method complexity.
Bug: N/A
Test: Treehugger
Change-Id: Ib58ecc853e2bd1e5d53e9ec256ef0467aa051d9f
Previously we only auto enable factory system trust agents when the
devices is booted for the first time. This change added logic to detect
newly installed/enabled trust agents, and enable them if appropriate.
Fixes: b/243705470
Test: atest FrameworksMockingServicesTests:TrustManagerServiceTest
Change-Id: I3c345a44cfcfccc6f7e2507ba37e378b6b16bd98
We would like to have userdebug builds as close to user builds as
possible so don't enable jdwp by default on userdebug builds. This
means that we won't be able to attach a debugger to non-debuggable
builds in userdebug builds by default. However, this support can
be enabled by setting persist.debuggable.dalvik.vm.jdwp.enabled
property. Eng builds still enable jdwp by default.
Bug: 240410400
Test: System boots
Change-Id: Iad52254be9b9bc8c9deac8b3ad46e9db1b610160
This is a tm-qpr re-write of http://ag/19779936 and http://ag/19806033. There was large refactor of BatteryStatsHistory that prevent the changes from being cherry-picked directly. Instead, this change copies the relevant sections and stitches together the rest.
Things that are identical
* The whole TraceDelegate
* recordTraceCounters
* recordTraceEvents
Things that are different
* Record calls are moved to BatteryStatsImpl
* Class setup (mTracer isn't changeable)
* No tests (can be added, but hard prior to the refactor)
Bug: 245749764
Test: presubmit and local perfetto tracing
Merged-In: Iaae4e9796a44be3ffdaee31b956bbb3e87735ead
Change-Id: I62fc2e93b662b89176cc26bdff8b3e88d216028a
Talkback can now correctly focus the collapse button before going to the action list when expanded. To know why mExpandButton is dynamically placed, check commit: 54fe9ad2dc
Bug: 237505828
Test: Visual, see the STEPS TO REPRODUCE section of bug 237505828, the focus order of talkback should be collapse button then actions.
Change-Id: Id0cf8730b00e0e7636a163fb5d51e37e8e0a3e84
When we calculate dismiss target, it will deduce nav bar size from
total height. Remove this rule to make divider always move to end
of screen.
Fix: 232858513
Test: manual
Test: pass existing tests
Change-Id: I34cf12f242daa31bd4dd7d4b202fc43ac0d752b4
