Different devices have different precision, leading to different pixels
being touched during rendering operations. We need to ensure that the
dirty rect we draw with (and which gets erased on the following frame)
encompasses all possible pixels instead of some ideal rounded rectangle.
The bug from this code led to dropped-pixels artifacts on some devices,
where we'd scale a view, drawing it into some pixels, then invalidate
that same area on the next frame, but the invalidation rectangle didn't
cover the same pixels as the device drew into.
The fix is to floor() the left/top pixels and ceil() the right/bottom
pixels of the transformed invalidation rectangle.
Issue #8971348 dropped pixel artifacts during some scaling operations
Change-Id: Iedb1afd5621dff43bf7a3919bdbd8d2251647fd2
The predicate for detecting whether text was right aligned was missing
some cases - it was correctly sensing RIGHT or OPPOSITE aligned text
in the LTR case, but was missing NORMAL for rtl. This patch just adds
that case.
Change-Id: I06691e7122e1ca91d66299cf0d7ef1f8a476a44b
We added APIs to allow an accessibility service to extend the
selection while moving the cursor at a given granularity such
as word, character, etc. The problem is that the traversal was
extending only the end of the selection while moving forward
and the start of the selection while moving backward. This leads
to a case in which the user cannot shrink/extend the selection
because for example instead of shrinking the end of the selection
the implementation was extending the start.
Now extending the selection moves only the selection end. This is
the same behavior as text view using a keyboard.
Tests: https://googleplex-android-review.googlesource.com/#/c/307062
bug:8839844
Change-Id: Id6965b102647df909f61301fcc8ec05458dd5881
This lets apps get the audio session id of the video being played, so
they can apply effects to the audio track.
b/8767565
Change-Id: Iaa39d97d0b6fb528ed04b52d579afa58444ebcfe
Loading the avatar icons and drawing them into the sized bitmap
turns out to be quite expensive and the cost increases with number
of users. Caching them shaves off several hundred milliseconds from
Keyguard inflation time during user switching on the lockscreen.
For instance, 15ms vs. 750ms with 3 avatars on a certain 7" tablet.
Bug: 7986933
Change-Id: I3e2065bfa25aa263133ba204ca364c3b04d7c0ff
AbsListView is backed by an adapter. After the adapter data changes
the view sets a flag that its state is dirty and requests a layout.
If an accessibility service asks for the state of a list item at
this point, it may incur an error since the views and the adapter
are not in sync.
Now if an accessibility service queries for a list item when the
data set is changed and the item views are dirty, we pretend the
children do not exist. After the layout happens, we notify the
accessibility layer that the screen content changed so it can
refetch the views if desired (this notification mechanism is
already in place in AbsListView#handleDataChanged()).
bug:8433433
Change-Id: I4287a0ac2ef6bb33f1f988d5ddad973556c305ca
If an app mistakenly puts some Parcelable implementation of
CharSequence in here, Bundle will try to marshal the derived
class, which is going to be bad news when we try to unparcel
it in another process.
Additionally add a proper stack trace to unparceling
failures (otherwise you have no idea where the bad parcel
was being used).
Bug: 8725271
Change-Id: Ide19b986d65c7f47929cf2dde4e62b762c469fee
