Bug 17936593
Instead of calling setLeft(), setTop(), setRight(), setBottom()
separately, make one call that does all at the same time.
Change-Id: I986274f3a98b3136e71204501ffc272986ad31dd
For now, only animators use it but we can consider migrating
drawable cache to it as well.
Bug: 17456416
Change-Id: I571b96856805edb171f0fc52e6bff5a365f46b70
In explore by touch mode the user performs a double tap to click on
an item. In this case the system sends down and up events at the
location of accessibility focus. The accessibility focused view may
be partially covered. In order to click in this view we compute a
point where to send the down and up events. This clicking strategy
is a bridge-gap and we will switch to accessibility actions in the
future.
When computing the point to click we were taking into account whether
the view was covered by a clickable sibling or a clickable sibling of
a predecessor. Despite our expectation cases in which this is not
enough happen in practice. In particular, the focused view may be
covered by a clickable descendant of a non-clickable sibling of a
predecessor that covers the focused view. This change takes care
of handling this case. Note that computing the click point is a fair
amount of work but this happens very rarely and on demand. Also the
code is short circuiting where possible.
Change-Id: I4d3cd8b67a7baf0bcc12f370ea7ba1b04c42c355
An accessibility service may register to observe the interactive windows
on the primary display. These windows are the one that has input focus and
ones a sighted user can touch. It is sometimes beneficial for an
accessibility service to overlay a window to intercept user interaction
and based on that introspect and perform an action on the windows that
are on the screen. This is problematic as overlaying a full screen window
that is touchable prevents the accessibility service to introspect the
content under this window.
This change adds a special type of window that only an accessibility service
can place which does not affect what an accessibility service can "see" on
the screen. Hence, even putting such a window full screen the service will
be able to interact with the other interactive windows it covers.
Change-Id: I053ccc3a5c6360a98dc40bdb172b54dab35d8b31
The Lollipop release introduced a feature that allowed
apps to extend under the navigation bar. This also means
any popup window that is anchored to the bottom of its
parent window will overlap with the navigation bar if the
parent window is extending underneath the navigation bar.
This change introduces a new window flag
(FLAG_LAYOUT_ATTACHED_IN_DECOR) that allows the app to
specify if the popup window should be attached to the decor
frame of the parent window thereby avoiding an overlap
with the screen decorations.
By default the flag is set on SDK version LOLLIPOP_MR1 or
greater and cleared on lesser SDK versions.
Also, replaced flags FLAG_NEEDS_MENU_KEY and
PRIVATE_FLAG_NEEDS_MENU_KEY_SET with needsMenuKey state
variable to make room for the new
FLAG_LAYOUT_ATTACHED_IN_DECOR flag.
Bug: 17789629
Change-Id: I2150e0c6ac688c966c0e8f7e54d42fd20285bea6
Updating the accessibility layer behavior to reflect the new
model where accessibility no longer overrides strong encryption.
Now enabling an accessibility service lowers the encryption
level but the user can bump it up in settings if desired.
bug:17881324
Change-Id: Ic60d760c267d3f934040a42e1963b179bd8b9f5f
First, this CL removes the need to decompose the DrawFilters
in Java and instead passes the SkDrawFilter to HWUI directly.
This also allows the removal of duplicated logic between HWUI
and other Canvas implementations regarding Paint filter levels.
Second, the DrawFilter is now stored in the DisplayListRenderer
where we apply it to every paint BEFORE it is stored in the
DisplayList. This eliminates the need to filter all Paints on
playback and removes additional complexity at playback.
Finally, as a result of storing the filtered paint we can now
do a better job caching the paints. This takes advantage of
recent changes in Skia to quickly enable quick hashing and
comparison of paint objects.
Change-Id: Iec507a2d894827975cc4f1d22241542bb0534b4e
If a view hasn't explicitly requested layout and it's asked to measure
with MeasureSpec.EXACTLY in both dimensions and sizes that match its
current measured size, the measure operation is a no-op.
This helps out a number of ViewGroups that perform initial speculative
measurements with AT_MOST or UNSPECIFIED followed by looping over
child views and measuring EXACTLY to lock in the final measurement
with perhaps some extra leftover space distributed. In practice this
happens a fair bit, especially for views high up in the view
hierarchy. This optimization allows ViewGroup measurement code to be a
little cleaner in not having to keep track of this on its own.
Change-Id: I88ff46a7d37aeda7a4cd16204b68cab0d051b341
