There are two types of IMEs:
A. IMEs that have one or more subtypes
B. IMEs that have no subtype
The initial implementation to update hardware keyboard layout per
subtype change of layout (See Bug 25752812) has supported IMEs in the
category A only, and IMEs in the category B are just ignored in both
system and Settings app.
In order to support IMEs in the category B, InputMethodSubtypeHandle and
related methods need to accept null InputMethodSubtype. Technically
this is a straightforward change, because in InputMethodManagerService
we have already used InputMethodUtils.NOT_A_SUBTYPE_ID for those IMEs in
the category B. We also need to update Setting App, which will be done
by a different CL [1].
[1]: I46b9c5b018f08e3eaa4614a0893db0be91652f3c
Bug: 28182650
Change-Id: Ia013784a594ad3beaf30976d047f5ac0fa8185be
Print interval clamp message
only for periodic jobs having
period/flex less than the minimum value.
Bug: 28296128
Change-Id: Id0d7c3b56048582a490bb8214eca491ec2bbdc93
Test all A2DP type variants when checking if an A2DP device is connected
in isBluetoothA2dpOn().
Bug: 28286051
Change-Id: I756d632b12d584d8a27cc1890e758d8accff7120
scheduleCrash() calls into the app process to ask it to nicely show
the crash dialog. If that call fails, rather than just silently
stopping, kill the process right away. It might be out of control.
This also adds an additional check for killedByAm to the flow there
so if the activity manager already has killed it, the dialog won't
be shown.
Bug: 28196243
Change-Id: I979f01074e5890640e7b06e29eeb0d6c38803569
When we attach an activity to an existing
task, and that task is already on top,
disallow changing stacks altogether even
if launch stack id is set.
Bug: 28026847
Change-Id: Ie70f0585a29dc1b85a5093624fede32110be3c76
Code was checking whether the directory for recent_images existed in
many places unnecessarily causing StrictMode violations. Moved the
code to only attempt to check and create the directory before
actually writing to disk.
Bug: b/28195831
Change-Id: I05f77a10f1dafc8cc0b1836b62352d56549ac1ee
As a workaround for b/27870771, convert fatal exception into a
warning.
Duplicate calls to finishPendingSequence with the same sequenceId
are seen in user crash reports, but with no further context and
inability to reproduce after testing various hypothesis, just
warn about it happening.
Bug: 27870771
Change-Id: Icd7d408887e04b94092689ce61809d6c664d8e3a
BugreportProgressService do not persist the user-provided
information (like details and screenshot paths), so if it's killed by
the framework, that info is lost.
Running it as foreground mitigates the changes of it being killed.
BUG: 27431998
BUG: 28291423
Change-Id: I2f58507beb38309628f2f19d3f7f950d07eca16f
Calculate the baseline power usage for the device when it is in suspend and idle.
The device is drawing POWER_CPU_IDLE power at its lowest power state.
The device is drawing POWER_CPU_IDLE + POWER_CPU_AWAKE power when a wakelock is held.
The device is drawing POWER_CPU_IDLE + POWER_CPU_AWAKE + POWER_SCREEN_ON when the screen is on.
Bug:27533512
Change-Id: Idcb587390bc8159fcbd6625cca4cb1aca19976d6
In addition, now that the full uncropped wallpaper image is being
backed up, we now handle that via the full-data backup path instead
of key/value. Restore still knows about legacy data that gets
delivered via the older key/value mechanism.
This change also has the effect of removing the size limitations
around wallpaper restore acceptance. Any size source imagery is
valid, as crop & scale are rerun in a device-appropriate way
after the restore.
Bug 25453848
Bug 25727875
Change-Id: Idc64a2eaab97a8ecc9d2b8ca5dc011f29cab324d
When choosing what window parameters to pass on
to child windows, we need to respect the manually set
FLAG_HARDWARE_ACCELERATED from Window.setFlags() in
addition to the activity manifest value, and the value
from setWindowManager. Given that we pass the value
on to child windows from when setWindowManager is called,
I think it makes sense to pass it when setFlags is
called.
Bug: 27099358
Bug: 23036374
Change-Id: Ie018841aadd270910fe0f8bc5a5ddca3bfbee69b
The RemoteInputView leaked through unhandled touches. To prevent this,
we now make sure the send button covers the entire area, and we catch
any touches that did not go to either the send button or the text box.
Bug: 27357771
Change-Id: Idc08d98dc7970b99c735f80088ea1eec0ec2d831
Previously if we failed to load an asset path, we would return a
null AssetManager which would lead to an immediate NPE.
Throw a proper exception instead, detailing which path failed to load.
Bug:28216288
Change-Id: I491b9316b837bf333b5ae78d4eb1ee31741617ec
- cleaned up private API to ensure userId is distinct from groupId.
- fixed bug where we were sending the wrong userId when attempting to
- fix warning about wrong fingerId when receiving final id of 0.
Fixes bug 28268635
Change-Id: I9507723c1a763152775f2feff76c16762f23cf2d
For M, we ignored device type entirely, but with ag/760625 we started
paying attention to it, which rejects events when keys from different
devices are pressed simultaneously. Since the volume keys show up as
different devices, the new behavior disrupted the event stream when
both volume keys were pressed.
Now tracking each device separately, which restores the old behavior
but still takes the device id into account. Holding down two keys
when enabling an accessibility service, however, always has and still
can produce an invalid event stream. It doesn't seem worth the
overhead, however, to track each key separately.
Bug: 28091773
Change-Id: I8d30de1f5e05f779b6fe305856d42f209ff8b038
If we close recents when beginning the animation, we will
trigger a resume of the previous fullscreen app, which will
attempt to aniamte in at the same time we are animating
the PiP to fullscreen. These conflict causing flicker and
churn.
Bug: 27793381
Change-Id: I520181dadab938bbf62b25891f5ba0e4e9783967
