This is a follow up CL to my previous CL [1], which introduced
RegisterStatusBarResult to consolidate return values from
StatusBarManagerService#registerStatusBar().
It turns out that the following fields are not correctly unmarshalled.
- RegisterStatusBarResult#mFullscreenStackBounds
- RegisterStatusBarResult#mDockedStackBounds
This is because those two fields were written with
Parcel#writeParcelable() while we were reading them directly with
Rect.CREATOR#createFromParcel(). Parcel#writeParcelable() needs to be
used with Parcel#readParcelable() for null handling and custom
class-loader handling.
With this CL, we now read/write these two fields with the following
methods, which can also be used to write/read Parcelable objects in a
bit more simpler style unless as long as we can rely on the default
class-loader.
- Parcel#writeTypedObject()
- Parcel#readTypedObject()
This CL also adds unit tests for StatusBarIcon and
RegisterStatusBarResult to make sure that these classes can be sent
over Parcel.
[1]: Ib1c0ae8f591ca09d0bce7a39f85ba57aad386e47
49efa718a8
Bug: 122439339
Test: atest FrameworksCoreTests:StatusBarIconTest
Test: atest FrameworksCoreTests:RegisterStatusBarResultTest
Change-Id: I0f41cbcb4c2f1222f18e1e34e3cf40bb0dfdfdb3
Set the default value for the USAP Pool to true for devices that don't
receive DeviceConfig profiles.
Bug: 128851983
Test: m & boot & check log for USAP creation
Change-Id: Id171336671a4fb8b8ad59c5b0e2e725657361107
(cherry picked from commit 0f7bce31a8)
There needs to be a single source of truth about whether DWB is
enabled or disabled.
Bug: 123930917
Test: atest FrameworksServicesTests:DisplayWhiteBalanceTintControllerTest
Change-Id: If634b46c2d0da123901bd5833e2114d958957540
(cherry picked from commit ff6770d669)
This is for statsd to log new metrics, as statsd has moved to use
thermal service in framework instead of connecting to HAL directly.
Bug: 119688911
Test: Build and dumpsys thermalservice
Test: atest $ANDROID_BUILD_TOP/frameworks/base/services/tests/servicestests/src/com/android/server/power/ThermalManagerServiceTest.java
Change-Id: Ib334c448c3615bf9d1cb0f1b6c2dd8a83d44f371
1. Add ripple feedback to clipboard copy even though the sharesheet
disappears really fast and you may not see it.
2. Make sure file icon is gray and not influence by the overlay.
3. Default multiple file sharing to file preview area and not image
preview area when there are mixed file types
Bug: 129979196
Test: Visual inspection
Change-Id: Ia86d418ccaca9a122f00dbc3e9c3b30785fb2abd
restorecon_recursive updates the SELinux label of the files in the
filesystem, and then attempts to write the xattr "security.sehash" as an
optimization for future restorecons. Writing security.* extended
attributes requires CAP_SYS_ADMIN, which system_server doesn't have (and
shouldn't have).
Suppress the computation and writing of the hash value. It's not
needed.
This bug has been around for a long time, but due to the fix for
bug 62302954, the error message is being generated more frequently
now.
TODO: It would be better if the default for restorecon was to suppress
the hash computation, since otherwise it encourages programs to be
overprivileged with CAP_SYS_ADMIN. I'll plan on doing that in a followup
commit.
Bugs where this error message has been called out:
Bug: 129766333
Bug: 129271240
Bug: 128700692
Bug: 129925723
Test: install an APK and ensure that no "SELinux: setxattr failed"
error messages are generated.
(cherry picked from commit cb1dddad27)
Change-Id: Ifc5be24d14029cb616d5564366fc10a0b93c9939
With the change in I4f13638598037acaeb30d61c8d5178f45882fcba
to separate the PackageWatchdog package expiry deadline from the explicit
health check deadline. It would be cleaner for ExtServices to supply
this deadline per-package. We now do that as a field in
PackageInfo.
Bug: 120598832
Test: Builds
Change-Id: I29e2d619a5296716c29893ab3aa2f35f69bfb4d7
The objective is to allow us to push model parameters by using a settings flag,
without the need of pushing a new model file.
Settings.Global#TEXT_CLASSIFIER_ACTION_MODEL_PARAMS stores a comma
separated string that contains these three fields:
1. required_model_version
2. required_locales
3. serialized_preconditions
To ensure serialized_preconditions is applied to the target model file,
TextClassifierImpl only applies the serialized_preconditions when
required_model_version and required_locales are both met.
Test: atest frameworks/base/core/tests/coretests/src/android/view/textclassifier/
Test: adb shell settings put global text_classifier_action_model_params 'required_model_version=0,required_locales=en,serialized_preconditions=FAAAAAAADgAIAAAAAAAAAAAABAAOAAAAAAAAAA=='
Observe that the flag is actually applied in the model.
Test: Ensure that finalize is called when the activity is dead.
BUG: 123616497
Change-Id: Ie42dcfeee705c83bbb693a5c1a0fedd0821df5e7
There are a few assumptions in the rollback manager that fail in the
multi-user case that need to be fixed:
* getAllSessions only returns sessions for the specific user.
* Session callbacks are only called on sessions associated with the
registered user.
* getPackageInfo only returns info for the specific user.
Fix these issues so that rollbacks will work properly, in particular
when the initial install session is owned by a non-system user.
Bug: 129809507
Bug: 129397974
Test: On single user device: atest RollbackTest StagedRollbackTest
Test: On primary user of multi-user device: atest RollbackTest StagedRollbackTest
Test: On multi user device manually:
adb install RollbackTestAppAv1.apk
adb install --user 10 --enable-rollback RollbackTestAppAv2.apk
-- verify the install succeeded --
adb shell pm rollback-app com.android.tests.rollback.testapp.A
-- verify the rollback succeeded --
Test: On multi user device manually:
adb install RollbackTestAppAv1.apk
adb install --staged --user 10 --enable-rollback RollbackTestAppAv2.apk
adb reboot
-- verify the install succeeded --
adb shell pm rollback-app com.android.tests.rollback.testapp.A
adb reboot
-- verify the rollback succeeded --
(cherry picked from commit d81ff97866)
Merged-In: I1a7cf101b3bc3575421629c4bf0ff63418eb8731
Change-Id: I1a7cf101b3bc3575421629c4bf0ff63418eb8731
StateMachine.toString() calls mSmHandler.getCurrentState(). In the
rare case when mStateStackTopIndex is -1 due to the StateMachine being
not yet initalized, this results in an ArrayIndexOutOfBoundsException.
Bug: 124858416
Test: none
Change-Id: Id558be95b6e12d8a923d325bca5f6f6d7f942b22
Calculating the reserved height was using the wrong figures. Expose
the correct reserved alwaysShow height, and use it to determine the
offset for scrolling.
Bug: 120419081
Test: manual, rotate device
Change-Id: I21fd9fa8494a258d17f6002f79967aacc0efde34
Unit tests for HashedStringCache that was commited in earlier CL
ag/6867725 . This is testing the various inputs and expected outputs.
Testing also revealed some vulnerability for invalid input so added
validation in the code under test.
Bug: b/129870147
Test: This is the test file
Change-Id: I7387f808df87a869f81339cd4aea99b23dfc06bd
This is a preparation to fix an issue that IME cannot enable light
navigation bar on direct-reply.
The initial state of StatusBar client is provided as return values of
IStatusBarService#registerStatusBar(). However, the way for the
client to receive the return values is to rely on "out" parameters in
the Binder method as follows.
void registerStatusBar(IStatusBar callbacks,
out List<String> iconSlots,
out List<StatusBarIcon> iconList,
out int[] switches,
out List<IBinder> binders,
out Rect fullscreenStackBounds,
out Rect dockedStackBounds);
The issue is that if I want to add a new parameter into
StatusBarManagerService to fix Bug 122439339, then it needs to be
notified back to the client via one of these out values, e.g. a new
element in "switches", which is hard to understand and maintain.
It'd be better to introduce a dedicated return object as follows.
RegisterStatusBarResult registerStatusBar(IStatusBar callbacks);
This is a purely mechanical code clean-up. Hence there should be no
behavior change.
Bug: 122439339
Test: presubmit
Change-Id: Ib1c0ae8f591ca09d0bce7a39f85ba57aad386e47
It needs to use the same STOPSHIP logic for consistency with the
two other enforcement sites across the OS.
Bug: 129487770
Test: atest android.appsecurity.cts.ExternalStorageHostTest
Change-Id: I7a3fa836e0795912c264aae58a55472ffae3d8c3
