This reverts the previous commit [1] since the keyboard shows up without
such a workaround. Presumably someone might have fixed something in
somewhere.
[1]: Ifd588bf7db76f9ea5d6a90854f8e92f86f4c2910
250faf9243.
Bug: 25942488
Bug: 26190615
Change-Id: Ie0e0ead960473a3504fc88a4f7d4d18108ef0e18
With ag/899084 TextView no longer automatically sets the
LinkMovementMethod when setText is called with a ClickableSpan.
So in LinkTextView we do that manually instead.
Bug: 27973982
Change-Id: I86626bd71bf2ca41cb7e73d479ba764e809b0190
MVNO type is judged as none by comparing key string to "None",
but this logic does not work when language setting is non English
because key is not "None" due to translation.
Bug: 27267663
Change-Id: Ic39a084cd77f661f833b13106dc28444926487d8
- Added "wifi-only" option to the list of wifi calling modes.
- Added code in settings initialization to check whether the carrier
supports including the wifi-only option. If it does not, changes the
entries/values to a ser which does not include the wifi-only option.
Bug: 27858149
Change-Id: Ib02a97afcf6f2722c28b7a4d9c241774944c52d0
The root cause of crash bug #27749932 is that the state mismatch between
when a Loader is created and when the Loader object finishes background
task. We can easily reproduce this crash by:
1. Pair two hardware keyboard A and B.
2. Open Physical Keyboard settings.
3. Press the power button to turn off the display.
4. Move keyboard A far away so that it is unpaired.
5. Press the power button to turn on the display.
6. Unlock the device.
One of the reasons PhysicalKeyboardFragment was unstable is that loader
ID reuse. PhysicalKeyboardFragment starts background data loading
because of many events such as #onResume() and #onInputDeviceAdded() but
there are chances that loader ID was reused because we specified
hardware keyboard device index as the loader ID. This was dangerous
also because device index can change when a device is added and removed.
With his CL each loader object has an unique ID and
PhysicalKeyboardFragment keeps tracking the list of active Loader IDs
only from which PhysicalKeyboardFragment should accept data.
Also, this CL removes dependencies on PhysicalKeyboardFragment from each
loader object so that we can have a clear boundary of responsibility
between data loader and data consumer.
Bug: 27749932
Change-Id: I53fcb2426d028a492c775bb2b4ec6a5419e33bb4
InputMethodAndSubtypeUtil#saveInputMethodSubtypeList() has a bug that
it saves implicitly enabled subtypes when "Use system languages" is
checked. Implicitly enabled subtypes are transient data and the system
should have only a null data (0) in the persistent strage. The root
cause of this bug is that the method in question has not checked whether
the preference item is in enabled (not grayed-out). If it is
grayed-out, its checked state does not mean that the user manually
checked that subtype but it is just an indicator for the user.
The strange UI jank when dismissing InputMethodAndSubtypeEnabler is one
of the victim of the above bug because we have worked around it by
actually changing checked state before calling the method in question.
With this CL we no longer need to update preference items in
InputMethodAndSubtypeEnabler#onPause().
Bug: 27867966
Change-Id: Ifc291d77ea41a988438765b9ba16bc5d18a15e1b
All settings preferences related to credentials of any kind should be
stopped by this user restriction.
Bug: 26879958
Change-Id: I983c6e58081bd4022bb006942499cba4b74954e7
- Notification icon is Settings icon
- VpnName is obtained as the same way as AppPreference
- Post notification as the current user with PendingIntent to VpnSettings to parent user
- Auto-cancel when user taps on the notification
- This implementation posts notification only when the failure happens sychronously
(Assume ConnectivityService only unset always-on package immediately after calling setAlwaysOnVpnPackageForUser)
Bug:27374485
Change-Id: I0aee38498c8cc300dd8eb9687adcae5f9dc4f8af
