I'm not sure how/when this case occurs (perhaps during some transition/state
change), but manually removing the floating view matches the symptoms in the
bug.
Bug: 72996404
Change-Id: I1e7c1a338fcd16c8e07b3c49fb9c9b2097eb2708
When bouncing between top/bottom overscroll, mDistance calculation
allows it to increase without bounds. This solution resets
mDistance to 0 for the edge that is not being overscrolled.
Bug: 78364220
Change-Id: I8371b3aee37e82d40aaa6adc2473a87f2b69931c
- In the case where an app auto-enters into PiP, we don't run the
predefined remote animation, which means that the force-invisible flag
is never reset. With ag/3876550, we can rely on enter animation complete
to be called after the transition, and we can reset the flag here as well.
Bug: 77730018
Change-Id: Ifa4cbe518047eed7ab327721ff6d17db1a2b8cd5
Some animation might be running from a previous orientation, which can cuase property changes
to get skipped.
Bug: 77848165
Bug: 77774619
Change-Id: I3e198196192746abdd72a1970ff2ef407bf4aff9
The shape detection logic was setting a transform on the mask path, which
overrides the original transform set by the drawable, which causes it
to give different transforms depending on how many times the method is called.
Bug: 77875383
Change-Id: I3054b8f48ba698a6c7866ed27e1129665b79c447
WallpaperManagerCompatVOMR1 is targeting API 27 (O MR1), but the
api guard is checking, if we are running API 26 (O) or above.
Correct the check to prevent using unavailable Wallpaper APIs.
Change-Id: I83cceee8eb61b2d74e647d473d7d560d00be14d3
(cherry picked from commit Iaf510bb79010e640db04adbbc2ee44de9abd6913)
Signed-off-by: Alexander Martinz <amartinz@shiftphones.com>
It’s an initial implementation, i.e. no fancy effects.
It shares a parent (LauncherRecentsViewContainer) with RecentsView.
The button is centered in clear_all_button_container, which gets
positioned programmatically to the right of the last task. (RTL polish
will be a separate CL as well).
Bug: 72222505
Change-Id: Ia912908a93a30c2f51450ccf0f97c7495e7916d5
Test: Manual
onStop() is a confirmed signal of user leaving Launcher. In most use
cases, we use onUserLeaveHint() to move PIP. However, in some cases,
onUserLeaveHint() does not get triggered. It's still a good signal for
those use cases because it happens sooner than onStop() and is not
interrupted by focus change. In order to address cases that
onUserLeaveHint() is not triggered (e.g. cancelling swipe up), onStop()
pushes PIP to the right position. In other cases, we simply set it
twice, which does not hurt.
Change-Id: Ia206590d10d673b16196bbab927a24aaae3ba46b
Fixes: 77648293
Test: manual test
We were requesting unbind in onCreate(), but the NotificationListenerService
documentation for requestUnbind() clearly states "The service should wait for the
{@link #onListenerConnected()} event before performing this operation. I know it's
tempting, but you must wait." I was tempted, and I did not wait. :(
The fact that the notification listener was binding even though the setting was off
was not only inefficient, but also had at least one user-visible bug: because
secure settings are set per user, the global badging setting actually only applies
canShowBadge = false for user 0; other users such as work profile still show badges.
Repro steps:
1. Have a work profile
2. Get a notification on work profile app and normal app
3. Turn off global badging setting ("Allow notification dots" from home settings)
4. Reboot the device
In this case, we get onCreate, call requestUnbind() which is ignored since we aren't
bound, then get onBind() and onListenerConnected() etc. Thus the work profile app has
a notification dot and other apps don't.
Bug: 71545493
Change-Id: I7f7dc219b25c28257f8b98fba7e362b99d3cba45
The current page is announced upon switching to it.
Is should be done this way for all future page implementations with
indicators.
Bug: 77827828
Change-Id: I8ba3bd9d068954df51d6de45da79133ede12c96d
Test: Manual
Merged-In: I8f2f163a261dd7098939674d39e014236fc66d5d
There seems to be an edge case where we can get a null notification in
onNotificationRemoved(); there's nothing actionable about that, so just
ignore it to prevent NPE. Also add null check to onNotificationPosted()
for good measure.
Bug: 69140873
Change-Id: I3586bf435d05aee38b99dffd3d01315b433e8476
Otherwise we'd be stuck using the old touch controllers until DragLayer
is setup again (e.g. launcher is killed).
Bug: 77921826
Change-Id: I8aac6fc453839902cb2d99279a6bd1549ee17d79
> If launcher already started, creating the state transition only after threshold crossed, so that previous animations are not cancelled
> Not posting animaiton callbacks at the front of the queue, as that sometimes causes it get executed before onNewIntent
> Farking the activity as forceInvisible while launching an opaque app, so that quickly pressing home/back runs the reverse animation
> Not running state animations when force-invisible is true
Bug: 77830325
Bug: 77898806
Change-Id: I50a7e915ca35fd6aeb284c8f321ecca74396fe98
Jon Miranda
TreeHugger Robot
Winson Chung
Sunny Goyal
Tracy Zhou
Vadim Tryshev
Tony Wickham