We recently added the width/height <0 check to prevent
passing unresolved widths in to tryFitVertical or
tryFitHorizontal and produce strange behaviors. This made
more sense than the code before it but still not that much
sense. We need to treat MATCH_PARENT and WRAP_CONTENT differently
as the documentation suggests.
Bug: 28267951
Change-Id: I6f2f248c8d1f63d31bdd75c24f7250c28381c712
* changes:
Hiding clear-all button on drag start.
Adding hint text for drop targets.
Fixing animation spec animation problem.
Making the dismiss animation feel smoother.
Fixing issue with gap between task views.
Adding desired undocking animation.
- We were always using the stack layout to get the task transforms, but
when we are animating to a given task position (ie. hitting recents
while the home->recents transition is happening), we would start the
animation with the final task rect, instead of the current task rect.
Instead, use the current transform of the task view, and transform
that to screen coordinates and use it.
Bug: 27154882
Change-Id: I25af8fad97ee78669c952dd81708e6cee5dfab05
- Now that we changed the way the landscape tasks clipped, we no longer
have to worry about the aspect ratio scaled height.
Bug: 27921659
Change-Id: I86b10805e92decb9995caa3790649f9f7d211784
Prior to Android N we were simply checking if a fragment had a
FragmentManager set before we would trigger a deferred
start. Unfortunately this also gets set before a fragment transaction
is committed, so if setUserVisibleHint was called before a transaction
commit, we would start the fragment way too
early. FragmentPagerAdapter triggers this situation.
Unfortunately some apps relied on this timing in overrides of
setUserVisibleHint on their own fragments, and expected, however
erroneously, that after a call to super.setUserVisibleHint their
onStart methods had been run.
Bug 28184671
Change-Id: Ie40d5f6963d312c2fad4a48fb4f992d33e65c83b
Force focus to the fullscreen stack so that its activity is resumed
regardless of it current state when moving all tasks from the docked
stack to the fullscreen stack and the fullscreen stack should be on
top.
Bug: 28247643
Bug: 28196862
Change-Id: Ic60c3794db765e9ee31b0facbbc4f461d65f30c2
The statusbar is now behaving nicer and other
parts are behaving similar as if they had been inflated
that way. The signalclusterview is now properly recreated
ensuring that it's scaled correclty.
Fixes: 26844819
Change-Id: Id1aab629204d03c079815358e7a5eb12d2a4e557
Reinflated the brightnessmirror to have the propper layouting.
Also fixed the height of the heads up scrim.
Bug: 26844819
Change-Id: Idb24dbd40f071a2a722c59cc202ae46c5ace2da8
The Quick settings page indicator was trying to access views that
don't exist yet.
Fixes: 27927374
Change-Id: I1600ce8432b6550268da20443d32193b1d2debe4
Use the override configuration for the task the app is contained
in to generate resources for the starting window.
Bug: 28220001
Change-Id: I6fdf39a6d6de41287b44b25861a76f58fe58dd53
In normal circumstances, StorageVolume can only be obtained through
StorageManager and hence it will contain a valid path internally. But
the path could be invalid in at least 2 occations:
- App maliciously changed it using reflection.
- Mounted volume was ejected.
This change improves OpenExternalDirectoryActivity so it returns a
RESULT_CANCELED upon receiveving an invalid path.
BUG: 27962875
Change-Id: Ide43968babaa37961c7b704bd289f44eac952e94
- Use thinner divider when IME adjusted
- Dim non-focused stack
- Restrict IME adjustment so that top pane is shrunk to 30% at most
- Move top stack up together with IME adjust
- Move bottom stack up even when not focused for thinner divider
bug: 28175599
Change-Id: I83556535e5a171f0ec28d88429186fa1315b910c
Bug: 28242626
ScriptGroup finalizer clears out the list of closures, which is
uninitialized for the old API.
Need to check null first, before accessing the list.
Change-Id: Ibf914e17a0878c8c561f823c5f6f6f6619594de1
This fixes a bug where string expression was being evaluated
and then compared to null. The result is a constant expression
which always passes, resulting in inadvertently accessing the
null variable.
Fixes bug 28237450
Change-Id: I755b4466fd1985bd7f8984c17e06839ac4a30642
When restoring a task when starting it from recents, we need to set
all the relevant state for the transition before we actually restore
the task and attach the docked stack, as all kinds of weird things
might go wrong otherwise.
Bug: 28194162
Bug: 27607141
Change-Id: I21d173a43c62d425e507e8cc3a1a576f95cfe048
When introducing the fitToDisplay variable, it should have been
fitToDisplay = task != null -> (implies) !task.isFloating();
but was written as task != null && !task.isFloating.
Bug: 28182018
Change-Id: If0be86f1ed8bb88914ce167e9f5273b6b3dc2571
We use the application context in DecorContext to avoid holding a reference
to the Activity. However, the resources don't match the Activity's in a multi-window
world, so we use the resources from the Activity, which get updated as the Activity
goes between multi-window states. This avoids holding a reference to the Activity.
Bug:28096052
Change-Id: I2804f2420afad6dcecc56e5dece008c390540d72
One of the changes in the 'interactive bugreport' bugreport workflow
introduced on N is that the initial screenshot was taken right away (by
Shell, not dumpstate).
Unfortunately, such initial screenshot is often delayed when the system
is overload. Also, if the user is not interested in a screenshot, it
would be adding more load on the system unnecessarily.
Given these issues, and the fact that the user can still easily take an
initial screenhsot by selecting the proper notification action, the
initial screenshot is being removed.
BUG: 28167977
Change-Id: I2cf6616ce3124102b62ec9a36dc5a0ce6455a909
