If Socket.connect() times out, the socket cannot be used any
more - any attempt to do so fails with EBADF. Use a new
socket for each IP address.
Bug: 16664129
Change-Id: If3616df86f7c2da0eabd30dca5db65d0da85cb17
Unsynchronizes the call into app code from setEnabledFeatures,
replacing it with a callback mechanism. Also makes this actually
work by fixing the check in TrustManagerService to take into account
whitelisting.
Change-Id: I0831752cd2d3158eda9c8404a5569498f11ac2ac
Adds a facility for trust agents to indicate if they
are ready to manage trust. Also adds an indication to
the lock icon on the lockscreen to show whether trust is
being managed.
Bug: 15518469
Bug: 16123013
Change-Id: Ie17f588aebeafe66c81dea4a69c733b0d2c72fd4
This change addresses the following API council feedback.
TvParentalControlManager:
-- Replace the listener interface with a broadcast Intent with well-known
action, something like ACTION_BLOCKED_RATINGS_CHANGED.
-- TvParentalControlManager goes away, and two remaining methods move to
TvInputManager.
== Implementation changes ==
-- The user-selected rating preferences should be stored in an XML file, and
loaded at boot by your internal system service. It should _not_ be stored in
a Secure/Global setting.
Bug: 16542165
Change-Id: I278df963576652ec9163a66c1ef99419fcdb36c2
Bitmaps added to TaskPersister were piling up in the queue.
- The mRecentsChanged flag was being modified without holding the
lock. There is no mRecentsChanged flag now. Everything to be
written goes into a queue.
- TaskPersister now runs until the queue is empty.
- Bitmaps being written to the same file were being added to the
end of the queue without replacing the earlier bitmap. Now we
search the queue for matching tasks and replace the bitmaps
if needed.
- Method notify() was renamed to wakeup() so IDE could find usages
quicker.
- Bitmaps that were being requested but were still in the queue
are now being fetched from the queue.
Fixes bug 16512870.
Change-Id: Idca1c712e5d2df8196e93faaf563a54405ee96bf
Previously input change listener is not invoked if the routing change
does not cause the TV port change, but this failed to trigger the change
from an active source connected to port 1 to a non-cec device also
connected to port 1. Do this by comparing paths not ports.
Also did some cleanup:
- Avoid allocating empty lists too often for getActions
- ActiveSource.invalidate()
Change-Id: Id79531e2552ef7fa38bd604796108b8650db69a1
- Accept the broadcast <Standby> message.
- Fix the bug that onStandyCompleted() is called mutliple times.
- Fix the issue that <Standby> message is not sent when going to
the standby state, since the local device was cleared already.
Bug: 16497304
Change-Id: Ib04ad2386621929814a801edd6b6c0e1a3902a7c
When signalled by an external CEC device about to become a new active
source, TIF(TvInputHardwareManager) sends an intent for TV app to tune to
the input mapped to the device. For non-CEC devices or MHL device
the device info comes with port ID only, which should be used to find
the right input.
Change-Id: I71c8454fead037d880ee440b0a8d64dd361ed558
This adds a new feature that allows a device admin to specify a
whitelist of features that are allowed for the given admin.
Change-Id: I83f853318efbcf72308532d0a997374f73fa9c10
Was previously removing by trying to match activities, but the
activity being matched would not have been in the stack yet. Since
it doesn't get added to the stack when it goes into the list
of pending activities. By removing all pending activities associated
with the stack we clean up all such pending activities.
Fixes bug 16045752.
Change-Id: I6b5981ffcce674139837b06362e573299bbc9e52
By finishing crashed activities right away the VirtualDisplays
associated with ActivityView are removed immediately rather than
being deferred. Also there is no waiting for Pause (0.5 sec) and
Destroy (10 sec) timeouts to expire.
Fixes bug 15092354.
Change-Id: Ie4941cb19dcc6f6fc61b2475162ea2794c9558a2
Client can jobFinished() before getting a cancel msg.
1) Do better clean up of JobServiceContext after client jobFinished()
to remove superfluous MSG_CANCELs
2) When processing MSG_CANCEL check whether the context is still active
3) Do JobServiceContext cleanup before calling back to JobSchedulerService
Client can get a cancel msg even after calling jobFinished() (opposite to above)
1) explicitly check whether there are any MSG_CALLBACKs in the queue before
processing a MSG_CANCEL. If there are we can throw away the cancel.
Bug: 16547638
Change-Id: I90644586c7895a9ce97de752a5d657faf7f74b78
Removing ConnectivityService.NetworkFactory. This requires disabling
the ConnectivityServiceTest, but that's been broken since we stopped
using NetworkStateTrackers anyway.
Change-Id: I9b86bd37eb9d018c40f60dca5b00d62c36d4e3ad
Bug 16626030
Getting the timing correct for when to clear returning activity options
was difficult when Activities can be destroyed and recreated. Moving
the returning ActivityOptions to the ActivityRecord forces the
ActivityOptions to only apply to the correct Activity.
Change-Id: Ib5400b157eb741ae484aae0cc83172a3db637a3e
Starting with startUsingNetworkFeature and stop.
Figure it's easier to code review incremental changes.
Change-Id: I19aee65e740858c3a9a2a1a785663f6fee094334
* changes:
Add context to MediaController constructor
Add getPackageName to MediaController
Expose a way to set a PendingIntent for restarting playback
Finish plumbing for launch intents in sessions
This removes the MediaSessionInfo class and adds a method on the
controller to get the package name. This also converts code that
depended on the session's UUID to use the token for comparison or
the Tag in the case of test code that needs a string identifier.
Change-Id: Icb8dc76c70e23e897da8fd07651c524b96372bf9
