This method is called by CarService when the system goes into suspend
state and the initial implementation completely disabled the service
(and unbound system server from it), which had 2 drawbacks:
- It adds extra work when the system resumes as it has to rebind.
- It can cause memory leaks if the service implementation doesn't
handle it well.
This CL changes how the API is implemented: rather than disabling the
service, it just ignores new sessions.
To test it:
1. Disable
$ adb shell cmd voiceinteraction disable true
2. Start a session
$ adb shell cmd voiceinteraction show
3. Check logcat
I VoiceInteractionManager: showSessionForActiveService(): ignored while temporarily disabled
Bug: 190051449
Test: manual verification (see above)
Change-Id: Idbc4881b92d9254fb07ea5af1370ea600fc10b80
HotwordDetectionService is an isolated service which ordinarily cannot
hold permissions. An isolated service has its own uid separate from the
owning package.
To allow it to access mic audio, we dynamically override the permissions
check based on the currently bound HotwordDetectionService.
Bug: 190011174
Test: manual - sample app can read audio (with a few other wip changes)
Test: atest CtsVoiceInteractionTestCases
Change-Id: I5abc809546184ef13cb9eb009f916afae6bdf1af
HotwordDetectionService is an isolated service which ordinarily cannot
access audioflinger due to selinux restrictions on isolated processes.
To bypass those restrictions, the audioflinger binder is passed from
the system server to the HotwordDetectionService. This follows the
existing pattern of providing access to the ContentCapture service.
In T we *may* instead solve this at the selinux layer. For now, this
simpler approach is used.
Bug: 190011174
Test: manual - sample app can read audio (with a few other wip changes)
Test: atest CtsVoiceInteractionTestCases
Change-Id: I5493a5c3addf68f2ae80d64b6971a8576484eb8e
For DSP-based detection, stopRecognition() disables detection at the
sound trigger layer. For software detection, the HotwordDetectionService
must be notified so it can stop detection.
Bug: 189236104
CTS-Coverage-Bug: 183425641
Test: atest CtsVoiceInteractionTestCases
Change-Id: I220d2255e5e500c22723f6e2fbc34860a529f216
The problem is that the assistant doesn't draw its own
UI in some cases so even though it's in the background the user
considers it as a foreground app. We can check if the app trying to
access the microphone currently has a voice interaction session which
should be a good indicator that the user intends to user the app.
Test: Use the assistant, observe dialog
Bug: 178243539
Change-Id: Ie17f6e47a002e75467cce23fd74f5fcbc92aa577
* Make the Hotword*Result payloads NonNull. HotwordDetectedResult on the
VoiceInteraction side is still @Nullable since the EventPayload is
used for other result types too.
* Remove @NonNull for a primitive return type.
Bug: 184962614
Test: atest CtsVoiceInteractionTestCases
Change-Id: Ib57368dab442f80805233864dc8fd86249f677cf
By API review feedback, it would be better to use a range
of numbers for OEM errors instead of just defining
CUSTOM_ERROR_1 / CUSTOM_ERROR_2 in HotwordDetectionService.
Bug: 184962614
Test: atest CtsVoiceInteractionTestCases
Test: atest CtsVoiceInteractionTestCases --instant
Change-Id: Id1e3a29e37ef566a1d8e30c703b7f2c4ae138f82
Due to the file descriptor of audio source was closed before
the attempt to parcel the file descriptor when passing the
file descriptor to the HotwordDetectionService.
It will cause the exception below:
java.lang.RuntimeException: Bad file descriptor
It would be better to close audio source when getting
callback from HotwordDetectionService
Bug: 185569471
Test: atest CtsVoiceInteractionTestCases
Test: atest CtsVoiceInteractionTestCases --instant
Change-Id: I10774fd29d559b5d2bedc21b0ebb2ef849a1b7c6
The Audio Framework now supports sharing audio history, so
VoiceInteractionManagerService no longer needs to handle audio streams
itself.
See I5beba6c1e489148a14ba86165b8ef2fdc78c802a.
Bug: 168305377
Test: atest CtsVoiceInteractionTestCases
Test: manual
Change-Id: Ia7c5830aa5af6c015ae26619e0a90967ddff74e5
This would allow pushing data into sandboxed ContentCaptureService.
Bug: 168305377
CTS-Coverage-Bug: 183425641
Test: atest CtsVoiceInteractionTestCases
Change-Id: I798b4f525a5707d4c97ae525ed69bd3864cb261d
Update setHotwordMode to setHotwordModeEnabled and set/clear the hotword
mode based on the boolean flag passed to setHotwordModeEnabled. This is
to meet the API requirements. Also update the references to this API.
BUG: 180419637
Test: Able to setHotwordModeEnabled from AGSA application. Confirmed
HAL receives the flag and audioSource correctly in open_input_stream.
Change-Id: Ia01e8b5f3edaeafc9617846efc247635005456a8
remove the ACCESS_VOICE_INTERACTION_SERVICE permission from
VoiceInteractionManagerService.getActiveServiceComponentName
Test: build and verify no stability impact
Bug: 173122149
Change-Id: I9e42b7b6e205ec352b3cfbac51022dd7594d1d0f
Use a session-based identity mechanism similar to SoundTriggerMiddleware.
Clear identity context within service to allow internal calls to
originate from this service.
Fixes: 184021066
Test: Manual verification of OKG / Now Playing on Sargo
Change-Id: Iccda9ef864f82eafbf3473e706bcffe695705953
