Th wait() call in AMS.getContentProviderImpl() can cause a deadlock when it
starts a new process for a target provider and all system server binder threads
are reaching out to the same provider at the same time, because AMS won't be
able to receive a callback from the provider process.
Let's serialize access to AM.getContentProvider() to avoid this scenario.
Longer team, we should stop waiting in the system server and instead do so in
the client side.
Bug: 74523247
Test: Boot, add a google account, and let syncs run and finish.
Test: Watch free video on Coursera
Change-Id: If5be2dc4b6f22f72d1bb3aa1b5b4d49f20c8d94d
OpEntry.duration was being used to indicate that the
operation was still running if -1 is returned. A recent
change caused a regression.
Adding a new mRunning field in OpEntry to explicitly
hold the running state, even when partial duration is
updated.
Change-Id: Ib29f4c903f990aaa202e84f964959aedfc24abdb
Fixes: 80242152
Test: atest FrameworksServicesTests:AppOpsActiveWatcherTest
Test: Launch maps and verify the location icon is visible
in the status bar
- Add icons to the right and shortcut modes
- Add color
- Make sure dialogs are secure
Test: manual
Bug: 79960669
Change-Id: I6d251b0ea5bd3c1b3eed32b184e4aae22c52ffb5
Fixes an issue where the group conversation heuristic was still
applied even when the app is targeting P. We're now following
the isGroupConversation completely and only do the migration
for apps targeting < P.
Change-Id: I471d58e8a8f5e6270f0dcce8691d08a2bdd1c582
Fixes: 78450835
Test: add messaging notification, observe correct display
This reverts commit c50f47d970.
Fixes: 79465234
Reason for revert: Google still does it using private APIs and apps were relying on this behavior, not good for the ecosystem.
Change-Id: I62e2b4cd1e6e562fcdd89c97e599bcdade83381a
Exemption given to a sync request made by a foreground app (including
PROCESS_STATE_IMPORTANT_FOREGROUND).
At the schedule time, we promote the sync adapter app for a higher bucket:
- If the device is not dozing (so the sync will start right away)
promote to ACTIVE for 1 hour.
- If the device is dozing (so the sync *won't* start right away),
promote to WORKING_SET for 4 hours, so it'll get a higher chance to be started once the
device comes out of doze.
- When the sync actually starts, we promote the sync adapter app to ACTIVE for 10 minutes,
so it can schedule and start more syncs without getting throttled, even when the first
operation was canceled and now we're retrying.
Test: atest cts/tests/tests/syncmanager/
Test: Manual test with "requestsync -f" and "am set-standby-bucket", while checking
"dumpsys usagestats"
Test: settings put global app_idle_constants \
exempted_sync_scheduled_nd_duration=1,exempted_sync_scheduled_d_duration=2,exempted_sync_start_duration=3
and check "dumpsys usagestats" and make sure the constants are properly updated.
Fixes: 72443754
Change-Id: I233d8e4be85769150830bac798abc04810f4cc7b
For multipackage requests, caller must be system uid or hold
interact_across_users if not requesting fate for the same user.
For single package requests, caller must be system uid or the
same package and hold interact_across_users if not requesting data
for the current user.
In both cases the caller must have the usagestats permission.
Test: couldn't find automated tests for this file; manual
Bug: 79142791
Change-Id: I8d650f8e875e0e4578cb48c9ba42e821342852a0
Because those notifications now stay around but the user
has already seen them, were hiding the icon instead.
Change-Id: I6b8590e31295ed79218698b1ac54df33b273d38c
Fixes: 78486801
Test: add notification, reply, observe icon gone
This also fixed the squishing when heads-upped by only allowing
one remote input text to be visible when heads upped.
Change-Id: I0c77931233b452420dee4dcf4772092d96d214c9
Fixes: 78877946
Test: add inbox style with 6 entries, reply a few times, observe no squishing
If too many events are queried from UsageStats, the Parcel will be too
big for a transaction. If the data is written as a Blob, the Parcel will
store too large data in ashmem.
Test: atest android.app.usage.cts.UsageStatsTest
Test: manual (create 1MB in UsageEvents and queryEvents)
Bug: 79860717
Change-Id: Ieb4bd486c83a9cc11fdff55e8fa457a9bdca7395
Marking some API's as deprecated, so users avoid getting confused.
Test: Test that build works.
Bug: 80099023
Change-Id: I4b3d4e4fa1ee3d706e49b8180aa4d0ad0e7d6eeb
Fixes jank issues where trimMemory gets scheduled during an
animation. trimMemory likely hits code that is not in the
page cache anymore, leading to high execution time because of page
misses. Scheduling it after the next frame has been drawn
minimizes the risk that this page miss actually causes jank.
Test: Boot, switch between a couple of apps.
Bug: 78611607
Change-Id: Ia1fc411fbe6ca07861183ae484124406681118e5
If an app cancels a notification while we are
sending a smart reply then don't dismiss the
notification.
Bug: 79120026
Test: atest NotificationEntryManagerTest
Test: atest SmartReplyControllerTest
Test: atest statusbar.NotificationListenerTest
Test: atest NotificationRemoteInputManagerTest
Change-Id: Ibbc0014ae0aeb12b58be13c041d720903251490f
Apparently comparing Spannables is dangerous because
the various Span classes do not implement .equals() in any
meaningful way, so all CharSequences must be converted to
flat Strings before being compared.
Lots of additional debug code remains, for the next time we
don't understand why an innocuous notification update
appears to be interruptive.
Test: atest com.android.server.notification.NotificationManagerServiceTest
atest com.android.server.notification.NotificationTest
Bug: 78643290
Change-Id: I1c282238687f28b5b197e28a4b878dc697049f4d
Lets shell/root get a list of pkgs that have access to any slices
within an authority.
Test: manual
Bug: 78136892
Change-Id: Id33fd458ed5ad73b30f69b3a464676428344e446
The behavior that #setChoices didn't render on mobile devices from apps
that target SDK < P went undocumented. Because this is changing in P and
above, this documentation needs to point out this behavior.
Also updated formatting and style of docs to improve interop with
kotlin.
Test: none
Bug: 79244189
Change-Id: Iabfe3b2ddcc369e5ec15f9773481468cbc48dabb
- Only report that something has changed when a uid changes between
the restricted and unrestricted states. (Adds new constant that
defines this important division point.)
- Have different settle times for different key transitions: (a)
out of the top state, (b) out of any other unrestricted state,
(c) down to a lower background state.
- But, don't apply any settle time when transitioning from top to
another unrestricted state... there is no need to.
Bug: 78480444
Test: manual
Change-Id: Ife772858009a6da5c286e285a4059aea3ad3f6c7
Should use default appComponentFactory, not the actual one from the
app because we have no code.
Test: manual
Change-Id: I982967543c4cf5c28b19b36b6d183a30052bf602
Fixes: 78883500
ActivityThread#setHttpProxy uses a system context to get
a ConnectivityManager instance. If it's not already cached, the
system service registry will cache this instance initialized
with the system context, and subsequent calls to get this service
will return the bad instance.
This patch uses the application context instead, unless it's
not available for some unlikely reason.
A possible symptom of this issue is to have Bluetooth tethering
failing with a security exception, because it's requested by
the "android" package without system uid.
Bug: 73572062
Test: manual
Change-Id: I745707e4ffe564f6d201ec259fa6b5ee32fbeda6
