This change ensures that applications producing surface texture updates
at a steady rate will display the updates on screen at the same rate,
v-synced.
Change-Id: I663685b34d12809fbf8945c44851e30a2052ce74
Removing the code that delays a surface destruction when
WindowManager.FLAG_KEEP_SURFACE_WHILE_ANIMATING is set. The lock
screen that continued to animate after destroySurfaceLocked is no
longer used and this code was causing problems.
Also mDrawState was being set to NO_SURFACE in destroySurfaceLocked
even if the surface ended up not being destroyed. Later when it was
reused the false value of mDrawState was messing things up.
The screen lock bug referenced below no longer levaes the user stuck
with a black lockscreen. However it occasionally powers back up in the
launcher screen rather than the lock screen.
Fixes bug 6485955.
Change-Id: I684104c7e7c39c161a5118aa890889fbae92e635
1. Some accessibility actions should not be performed on disabled
views. For example, scrolling should not be permitted while
accessibility focus should be. Made a quick pass over the
actions we expose now.
Change-Id: I36626dfbc0d2f480309a910f58f1de64e9e05675
When using stable layouts, you are typically expected to hide and
show the status bar through the system UI fullscreen flag. This hides
both the status bar and the action bar. The stable layout assumed
that when not hiding the status bar through the system UI flags, that
the status bar would be visible.
This change makes things a little smarter, also looking at the
window's fullscreen flag (which only hides the status bar). If this
flag is set on the window, then the stable layout now assumes that
the status bar will never be shown. This allows us to position the
action bar correctly in the situation where the application has set
the window to fullscreen and requested a stable layout, instead of
always leaving room for the status bar above it.
Change-Id: I757072ae99cd3741753af7210dbf51afe94d3db5
This makes a noticeable improvement in cases where applications
post messages that need to be processed between animation frames.
Bug: 6418353
Change-Id: If225742e37aeaf3f0ca9710f9bf43dbb03bcde12
1. View is checking if the accessibility focus is its
descendant it clears the accessibility focus state
in ViewRootImpl. The check in View was missing the
case that the descendant may be the view itself. In
such a case we want the normal clearing code to run.
2. The check whether a view has iterable text for
accessibility was inverted and text nav was not
working.
Change-Id: I1a13b6809fb7f205fff76ca09cd449179d06e530
Bug #6485955
If an invalidate gets scheduled right before the EGL surface is destroyed,
the next draw pass is done in software. This causes the software renderer
to connect to the surface forever which prevents the hardware renderer
from coming back when the screen is turned back on.
The fix here is to ignore the draw request when hw acceleration is requested
but not yet available. Proper software fallback will still happen when an
error is encountered with hardware rendering (in which case hw acceleration
will not be marked as requested anymore.)
Change-Id: I1edc4a51c8dd38240aa2345092a18a081a756fc1
1. The prefetcher of accessibility nodes infos was not folloing
the childForAccessibility relationship when finding the views
whose node infos to prefetch.
2. NumberPicker was not reporting the correct parent.
bug:6471710
Change-Id: Ia7ad5dd031fb4b3816dfe630d5212201cfafa236
1. Finished the implementation of support for maintaining
accessibility focus in view with virtual descendants.
2. Finished the NumberPicker implementation of virtual
subtree such that all requred attributes are reported
and ensuring that it support accessibility focus in
its virtual descentants.
3. Fixed a bug where if a predecessor of the view that is
accessiiblity focused is removed the accessibliity focus
host in ViewRootImpl is not cleared leading to a crash
when trying to draw the accessibility focus highlight.:
bug:6472646
bug:6433864
Change-Id: I3645642b87b4a26025c0b2ba9dfaad92d11a48f1
1. Now the views considered during the accessibility focus search
are the ones that would get accessibility focus when thovered
over. This way the user will get the same items i.e. feedback
if he touch explores the screen and uses focus traversal. This
is imperative for a good user experience.
2. Updated which focusables are considered when searching for access
focus in ViewGroup. Generally accessibility focus ignores focus
before/after descendants.
3. Implemented focus search strategy in AbsListView that will traverse
the items of the current list (and the stuff withing one item
before moving to the next) before continuing the search if
forward and backward accessibility focus direction.
4. View focus search stops at root namespace. This is not the right
way to prevent some stuff that is not supposed to get a focus in
a container for a specific state. Actually the addFocusables
for that container has to be overriden. Further this approach
leads to focus getting stuck. The accessibility focus ignores
root names space since we want to traverse the entire screen.
5. Fixed an bug in AccessibilityInteractionController which was not
starting to search from the root of a virtual node tree.
6. Fixed a couple of bugs in FocusFinder where it was possible to
get index out of bounds exception if the focusables list is empty.
bug:5932640
Change-Id: Ic3bdd11767a7d40fbb21f35dcd79a4746af784d4
Whenever a stream type is muted, the progress bar in volume panel
is at 0.
If a stream is muted by ringer mode and does not control ringer mode,
the progress bar is disabled.
Pressing VOL- when in vibrate or silent mode resets the last audible
volume of ringtone stream (music strem on tablets) to 0.
VolumePanel implementation:
- Always prefer AudioManager APIs over AudioService APIs when available on both.
- Do not use AudioManager.shouldVibrate() (deprecated).
Change-Id: I57fcb19ada4e8d729b6b41d668496562ebe340c3
The window manager now performs the crop internally, evaluating
it every animation from, to be able to update it along with
the surface position.
Change-Id: I960a2161b9defb6fba4840fa35aee4e411c39b32
SurfaceView didn't recompute the transparent region
when the INVISIBLE state (as opposed to GONE) was involved.
Bug: 6467123
Change-Id: I05930bd568a345331840c1ad8d9123ef4904c04f
* moved View#isVisibleToUser to protected with @hide
* added a new View#isVisibleToUser(Rect), so that a portion of
the view can be tested for visibility
* NumberPicker will report its concrete class name
* code to append virtual children was at wrong place
* boundInScreen for increment and decrement buttons are reported
wrong
Change-Id: Ic5d644b3e1efa15b1f0537907c8cdd4ce43a97c1
This refactoring sets the stage for a follow-on change that
will make use additional functions of the power HAL.
Moved functionality from android.os.Power into PowerManagerService.
None of these functions make sense being called outside of the
system server. Moving them to the PowerManagerService makes it
easier to ensure that the power HAL is initialized exactly once.
Similarly, moved ShutdownThread out of the policy package and into
the services package where it can tie into the PowerManagerService
as needed.
Bug: 6435382
Change-Id: I958241bb124fb4410d96f5d5eb00ed68d60b29e5
1. Fixed errors in the accessibility node cache.
A. The cache was not catching the case when the current window changes as a
result the user touch exploring it. As a result the cache had nodes from
more that one window but the node ids are not unique thus causing a mess.
B. The node info tree was prefetched regardless if a prefetched node is root
name space (i.e. view ids - not accessibility ids - are namespaced) while
the prefetched nodes were taking this into account. As a result there can
get disconnected subtrees in the cache.
C. When an event for a property change such as focus was received the cache
we were removing the source node. As a result there may be disconnected nodes.
D. When a node was added to the cache and an older version exists there was
no check if it will point to the same children and parent. As a result if
the state of the node has fewer children the subtrees rooted at the no
longer present children will stay disconnected in the cache.
E. When a node got accessibility or input focus the old one in the cache was
not removed. As a result you may have a state with more than one access
or input focus.
2. Added integrity check enabled only on user builds when a specific flag is set
for the cache which checks whether:
A. All nodes are from the same window.
B. All nodes are connected.
C. There are no duplicates.
D. There is only one input focus.
E. There is only one accessibility focus.
3. The reported accessibility node info tree was stopping at the root namespace
boundary which is not correct. The reported tree has to reflect everything
on the screen that the user can see such a workspace with widgets. The root
namespace is added to avoid clash of view id but the accessibility ids are
unique no matter if the view is inflated from a remote view.
4. Added calls to notify the accessibility layer when a preoprty that is interesting
for accessibiliy has changed.
bug:6471710
Change-Id: I069470d91f209ba16313fa6539787a55efa3512e
Use View transient state tracking to allow selection to persist across
ListView-style item view recycling.
Fix some bugs with transient state tracking.
Bug 6110122
Change-Id: Ic084b8fc2289bff718b19478a37ce64459b3ed4c
When a parent draws its child views, each child's bounds is checked
against the current dirty region. If the view falls outside of that
region is is rejected and doesn't enter into the parent's DisplayList.
This works in general, for both transformed and untransformed views
(because we skip this check if the view is transformed). But it breaks down
when the transform properties of the view change later, since DisplayList
properties simply push these values down to the view's DisplayList without
invalidating the parent. If a view is rejected when untransformed, then there
is nothing to cause it to be considered again until something causes an
invalidate of the parent.
The fix is to note when a view is rejected and record that information.
Later, when one of the transform-related properties change, we invalidate the parent
to force the check to happen again, which will cause the view to get drawn
if it's visible.
Issue #6477730 quickRejected views must recreate their DisplayLists on transform changes
Change-Id: I70caf198005cd7e424a37bccc6ae050e09880a6c
