It turns out that my previous CL [1] unexpectedly changed the behavior
of InputMethodUtils#getImplicitlyApplicableSubtypesLocked() in terms of
when "EnabledWhenDefaultIsNotAsciiCapable" extra value is taken into
account.
Suppose if an IME X has the following three subtypes:
A. locale: en_US
mode: handwriting
extraValue:
B. locale: hi
mode: keyboard
extraValue:
C. locale: en_US
mode: keyboard
extraValue: AsciiCapable
D. locale: zz
mode: keyboard
extraValue: AsciiCapable, EnabledWhenDefaultIsNotAsciiCapable
Given the above subtypes, here are results of what subtypes are enabled
by InputMethodUtils#getImplicitlyApplicableSubtypesLocked() I) before
the CL [1] and II) after the CL [1].
- system language: hi:
I: B, D
II: B, D
- system language: hi-IN:
I: B, D
II: B, D
- system language: en-US
I: A, C
II: A, C
- system language: en-GB
I: A, C
II: A, C
- system language: ja-JP
I: B
II: D
What my previous CL actually broke is the the last one, and it's broken
because we accidentally started using
"EnabledWhenDefaultIsNotAsciiCapable" even when there is no subtype that
matches to the requested language. Previously that attribute was used
if and only if 1) there is a subtype that matches the requested language
and 2) that subtype is not marked to be AsciiCapable.
If there there is no subtype that matches to the requested language,
what we had relied on is actually the result of
InputMethodUtils#findLastResortApplicableSubtypeLocked() called with
canIgnoreLocaleAsLastResort = true,
which means that we had just picked up the first keyboard subtype as the
last fallback candidate regardless of it's locale. This is why the
subtype B should be picked up in the last case where system language is
ja-JP.
This CL fixes the above unexpected behavior change regarding
"EnabledWhenDefaultIsNotAsciiCapable" so that the previous behavior can
be preserved.
[1] Iaf179d60c12b9a98b4f097e2449471c4184e049b
e985c240e3
Bug: 27129703
Bug: 27425459
Change-Id: Icd86cad955bf827a1eb41255f57fdf4ec315b93b
From the runtime perspective extract-only/profile-guide-compiled/fully-
compiled oat files are up-to-date and don't need dex2oat. However,
wihout knowing the exact "class" of the aot file we are not able to do a
full compilation of something previously only-extracted, or limit
profile-guide compilation to only previously profile-guide or extract-
only oat files.
GetDexOptNeeded now accepts a mask of desired types of compilation
that the runtime takes into account when advising what's needed.
This CL adds the necessary handling in DexOptimizer.
Also:
- removes the contraint to use extract-only oat files only when
profiles are enabled.
- removes mDexOptPerformed which interfere with the different types of
compilation and only saved a call to GetDexOptNeeded.
Bug: 27189430
Change-Id: Iced2bdcc3aa7866ff888a3bace43f4b71fffb353
Code that uses WakeupMessage uses the AlarmManager. Testing such
code is slow because AlarmManager.MIN_FUTURITY ensures that
alarms must wait at least 5 seconds before firing.
This change makes WakeupMessage's fields protected so that test
code can subclass from it and override schedule() and cancel()
with implementations that do not use AlarmManager, for example
by making schedule() call sendEmptyMessageDelayed and making
cancel() call removeMessages.
Change-Id: I51096b182d9eb87cc7bd46c3c91906f18356b354
To allow static initialization of a number of View classes based
on TextView, refactor the initialization of the font cache to be
explicit from the zygote.
Bug: 27265238
Change-Id: I1b71086d3f49d8b3e72eea2bf8359351d25fc0fd
Similar to first patch, but now using new "rethrowFromSystemServer()"
method which internally translates DeadObjectException into
DeadSystemException. New logic over in Log.printlns() now
suppresses the DeadSystemException stack traces, since they're
misleading and just added pressure to the precious log buffer space.
Add some extra RuntimeInit checks to suppress logging-about-logging
when the system server is dead.
Bug: 27364859
Change-Id: I05316b3e8e42416b30a56a76c09cd3113a018123
The registered shortcut will be called from PhoneWindowManager,
before dispatching
Change-Id: If26128939b45a639c8895719a7a23ca433f39fd9
(cherry picked from commit 4da863c5a8872dcabb179a978a2b2157d9081679)
Both bytes transmitted and received must be 0 if we are to short circuit and
ignore the UID's network delta.
Change-Id: Ia69f605c7d5954b8f2c3671bab596dc9925f1d02
Encapsulating the logic to toggle multiwindow mode from recents,
and plumbing it through to accessibility global actions. Sending
accessibility events when windows bounds change. Exposing the
dock divider window type to accessibility services.
Bug: 27250995
Change-Id: Ib7491f1f853dc7f01bf5c5a4ac1f914f55d0608a
Configuration values are not persistent.
To make system locale persistent, use Settings.System entry.
During start up, the configuration is filled from Settings.
When the system locale is changed by Settings, the locale list in
configuration is copied to Settings value.
Bug: 26732134
Change-Id: If57e13cc4179ff6299c71b4ab030f9d856748dea
There are two major changes in this CL:
1. Now IMMS resets its internal state whenever the system locale list
is changed, rather than just checking the primary system locale.
2. For software keyboard subtypes,
InputMethodUtils#getImplicitlyApplicableSubtypesLocked() now takes
the entire system locale list into account when determining what
subtypes should be enabled by default when the user does not
explicitly enable one or more subtypes.
Bug: 27129703
Change-Id: Iaf179d60c12b9a98b4f097e2449471c4184e049b
This is a preparation CL to take secondary system locales into
account in InputMethodUtils#getImplicitlyApplicableSubtypesLocked().
Suppose the following situation:
available subtypes:
en-US, en-IN, and en-GB, fr, fr-CA, fr-CH, fr (QWERTZ)
system locales:
en-GB, en-US, fr-MC
Basically we want to have at most one subtype for each language appears
in system locales. Hence the goal of this utility method is to filter
the above available subtypes into en-GB and fr. In other word, we do
not want to enable both en-GB and en-US subtypes in this scenario.
This CL introduces LocaleUtils#filterByLanguage() for this purpose, with
some unit tests. Note that that method is not used in production yet.
Bug: 27129703
Change-Id: I315cf3722a06e00bdbfac284c4949578da8fe78d
When an app requests SYSTEM_UI_FLAG_LAYOUT_HIDE_NAVIGATION but we
force show the navigation bar, we need to treat for the app like
there is no virtual navigation bar on the device. Because if you
combine it with FLAG_HIDE_NAVIGATION, you'd expect the navigation
bar gets hidden but it doesn't, so there could be content that
overlaps with the navigation bar.
Bug: 27157904
Change-Id: I088e02eae2e723c35e9cb4873de6b1325458533b
* Removed warning icons in locale list (b/26710677)
* Use getDisplayNameWithDialect() instead of getDisplayName()
for the locale names (b/27066642)
* Set proper locale for the language labels (b/27130516)
* Fixed build warning in LocaleHelper (b/27148053)
Bug: 26710677
Bug: 27066642
Bug: 27130516
Bug: 27148053
Change-Id: I149a3675262b286c8b740d2ad5b02f4df31316ec
- Add a Clocks interface to wrap SystemClock.uptimeMillis and
SystemClock.elapsedRealtime.
- Make the inner classes static so they can be created independently,
which meant passing explicit pointers to the BatteryStatsImpl and
the Uid objects in a lot of places.
- Make several fields protected so they can be accessed by the tests (they are
still @hidden)
- Add a test suite for the battery stats.
- Add the tests
Change-Id: I4e45afedc64f7050985c28a7eb4f5c1ddbaa63fa
This is a safe refactoring that changes nothing.
In order to improve the keyboard dismissal lags [1][2], we have used
IMMS#windowGainedFocus() as a combined event to do startInput() in
certain situations.
To make the intent of those CLs clear, this CL renames
IMMS#windowGainedFocus() to IMMS#startInputOrWindowGainedFocus(). Note
that these are @hide internal IPC protocols. Hence this change is never
observable to application developers.
[1] I8494cbd6e19e2ab6db03f2463d9906680dda058b
a82ba54b0b
[2] Icb58bef75ef4bf9979f3e2ba88cea20db2e2c3fb
7663d80f6b
Bug: 25373872
Change-Id: I56934f18e30d90fcdf77bcbb0c35a92a5feb1b82
Also fix a couple of issues in the code:
1. Check that fields are static before attempting to read them.
Currently, if a non-static field's name starts with one of the
specified constant prefixes (by default, "CMD_" and "EVENT_")
we'd get an NPE.
2. Only read final fields.
Change-Id: I0e07c4c27a66180cbaf7f2b1c235daafc080863b
- We keep the docked stack visible when home task is visible even
though it's not resizable.
- We introduce the a new concept called "minimizing" the docked stack,
which happens when going home. In this state, the docked stack is
clipped of almost completely.
- To achieve that, we introduce TaskStackBoundsAdjustController,
which adjusts the bounds of the docked stack when minimized. Also,
migrate the IME handling to this new class.
- We also need to inform SysUI that it is now minimized so it can
remove the drag affordance on the divider, and also make it a bit
smaller.
- When we detect an app transition, we check whether the home stack
gets visible/invisible. We then start an animation which runs in
sync with the normal app transition. For that we introduce
DockedStackDividerController.animate(), which performs the animation.
Bug: 27137961
Change-Id: I8623bc73cc6872bf28c5b1b8d5795974576811b2
This is useful when using the new AlarmManager direct callback
interface to wake up the system and request that an object whose
API consists of messages (such as a StateMachine) perform some
action.
In this situation, using AlarmManager.onAlarmListener by itself
will wake up the system to send the message, but does not
guarantee that the system will be awake until the target object
has processed it. This is because as soon as the onAlarmListener
sends the message and returns, the system is free to go to sleep
again.
Bug: 20157436
Bug: 25823676
Cherry-picked WakeupMessage.java from 9d3aadb247
Change-Id: Idff20029d287f26347441a2523b7fb20eda6a8b0
This class uses reflection to find accessible static integer
members in a specified list of classes and returns a SparseArray
mapping the integers to their names. This will allow us to
replace various 400-line switch statements with a simple
array access.
Change-Id: I3607e6389a423cde0bd83270c00b3c863ae1bb29
Make printlns in Log public-@hide so it can be used.
Print uncaught exceptions that will terminate the process in
RuntimeInit using printlns, so that long exception traces are not
being truncated.
Bug: 27245306
Change-Id: Ib24635f0ebdd80bd125e367302cab6a78e6a210a
And set the drawable's callback to null during drag-resizing, since
we use multi-threaded renderer, will do not want to schedule draws
to the ViewRootImpl's thread.
bug: 26729953
Change-Id: I6e5f94a5a6ba15edc2d391dd11d8fee3c657d337
(cherry picked from commit 1cc95075e8)
