...rotating to landscape
When doing spell checking in the same process as the spell checker, we
need to make sure it is still done asynchronously.
Putting this in I noticed quite a few threading issues in this code, so
I also addressed those (which became very obviously a problem with the
async stuff here now).
Also tweaked the service side to run spell checking at background priority.
Change-Id: I01bafe3bec6bceeca911d6bf2f61a486a2fd4c48
Bug 5755099
In extracted text mode, we spell check the background and the foreground
edit text. All changes in the background are already propagated to the
foreground, and updates on the foreground may trigger a spell check
of the entire text since we use setText.
Change-Id: Ie44c3d9113258082ac6df5890851d0e21c702ea2
Bug 5753061
https://android-git.corp.google.com/g/#/c/154756/1 revealed an
other bug in SpanSet. The fitered (non empty) spans were added
in their original position instead of being indexed by count.
The nullation on recycle hence left null holes in the array.
Change-Id: If5c1435cee9a2cb88a608aa8e5f4f2f23382154c
We need to work more like before in determining whether the menu
key is needed -- in some cases look back in the window list to
determine this if we don't know the value from the current window.
This requires adding a new private flag indicating whether the
compat menu state is known for a window, which is set by
PhoneWindow as part of its existing process of computing the flag
for its own windows.
Now we can have a new API on WindowState to determine the value
of this flag for a window, which if needed walks back in the window list
to find a window the value is known for (or stops at what the policy
has determined is the top full-screen window, so we stop like we used
to at things like the lock screen or the bottom of an application).
Change-Id: I829de6d629b5af8bcb422cb85249ee4041c7205e
Update some of the platform documentation to directly link to
relevent support lib docs. Yay!
Also improve BroadcastReceiver documentation to more clearly
discussion security around receivers, and how the support
lib's LocalBroadcastManager can help.
Change-Id: I563c7516d5fbf91ab884c86bc411aff726249e42
Bug 5740435
The new SpanSet cached objects should be recycled when the TextLine is
recycled to prevent a reference to their span, which may lead to the
entire activity.
Change-Id: I4469e418a228bd71eb8be1d43ed86de49baba165
Bug 5738416
This problem was introduced in
https://android-git.corp.google.com/g/#/c/152599
The invalidated cursor region is no longer expanded to include the
cursor drawable (different from the handles' drawables).
Added that code back. Added a flag, since invalidateRegion is also
used to invalidate text span regions in spell check (assumes the decorated
span bounds are not bigger that text boundaries), which is fine in case
of underline.
When the cursor is moved on a single line, invalidate the whole line
(same as what was done before). This is sub-optimal, will file a bug
to fix this. The core problem is that we should invalidate the previous
and the new cursor's positions. We only have one of these.
Change-Id: I9ada9340fb52aad3d80c39efd021fd3f9ec0cc4d
CTS test revealed that persistSwitchSetting could take more than 1 second.
STATE_TURNING_ON intent missed the test timing. We move the broadcast
before persistSwitchSetting
bug 5740715
Change-Id: I54a0a0afc8d1c4fc28f84e6122809af2f1456ca7
Limit each parse to batches of a few words, to keep the UI thread
responsive.
Possible optimizations for the future:
- SpellCheck in a thread, but that requires some locking mecanism
- Only spell check what is visible on screen. Will require additional
spans to tag the pieces of text.
This is a cherry pick of 145656 into ICS-MR1
Patch Set 2: Make the Runnable shared and stop it when detached.
Change-Id: Ibf8e98274bda84b7176aac181ff267fc1f1fa4cb
TextLine objects are already stored in a pool to limit allocation.
Associate SpanSet objects to each TextLine to further limit the
allocation of these and re-use their internal arrays (re-sized if needed).
Drastically reduces new objects allocation during rendering.
Priority code removed in init: priorities are already handled by getSpans.
This is a duplicate of the already accepted
https://android-git.corp.google.com/g/#/c/153970/
but without the dependency on an other CL.
Change-Id: Iaa7e2f7a8ea4337c8d60c3a9a620e9e3e60caf12
The ICS navigation guidelines state that the OK action should be to the
right of the cancel action in UI dialogs. Update the SIM PIN unlock and
SIM PUK unlock layouts to swap the OK and Cancel buttons (for soft keypad)
and the OK and Emergency Call buttons for the physical keyboard landscape
layout.
Bug: 5730977
Change-Id: Iaf565bc2b584292a34f5a72a7a763003f9149f27
Fixed a problem where the key up for the ALT or META key was not
delivered to the task switcher dialog because it was deemed
to be inconsistent with the window's observed state. Consequently
the dialog would not be dismissed when the key was released.
Moved global hotkey handling for META+* shortcuts and ALT/META-TAB
into the window manager policy's interceptKeyBeforeDispatching
method. This change prevents applications from hijacking these
keys.
The original idea was that these shortcuts would be handled only
if the application did not handle them itself. That way certain
applications, such as remote desktop tools, could deliberately
override some of these less important system shortcuts.
Unfortunately, that does make the behavior inconsistent across
applications. What's more, bugs in the onKeyDown handler of
applications can cause the shortcuts to not work at all, for
no good reason.
Perhaps we can add an opt-in feature later to enable specific
applications to repurpose these keys when it makes sense.
Bug: 5720358
Change-Id: I22bf17606d12dbea6549c60d20763e6608576cf7
1. If the number picker was not wrapping the selector wheel
it was incorrectly initializing the fling scroller and
the error was proportional to the difference of the
current value and the max size. As a result if the number
picker gets stuck for a very long time.
2. The alpha of the selector wheel paint was not reset on
every subsequent fling so if the wheel was fading out
and a new fling is performed the selector wheel is
dimmed during the fling, rather bright.
bug:5728363
Change-Id: Ia90a6af86f162bda1b28f43f7ba7e4a9f5051b78
Reduce likelihood of crash when state machine has quit and someone
sends a message using one of the public functions.
Bug: 5724844
Change-Id: I6582a1d19113e6ed545c8ab20adb0a414d8784a7
