[PBE-4085] Difficulty setting up automatic audio output switching for…

… calls on android (#1115)
GetStream · Jun 13, 2024 · 92aefe9 · 92aefe9
1 parent 5b5f112
commit 92aefe9
Show file tree

Hide file tree

Showing 7 changed files with 70 additions and 22 deletions.
diff --git a/stream-video-android-core/api/stream-video-android-core.api b/stream-video-android-core/api/stream-video-android-core.api
@@ -409,10 +409,12 @@ public final class io/getstream/video/android/core/LocalStats {
 }
 
 public final class io/getstream/video/android/core/MediaManagerImpl {
-	public fun <init> (Landroid/content/Context;Lio/getstream/video/android/core/Call;Lkotlinx/coroutines/CoroutineScope;Lorg/webrtc/EglBase$Context;)V
+	public fun <init> (Landroid/content/Context;Lio/getstream/video/android/core/Call;Lkotlinx/coroutines/CoroutineScope;Lorg/webrtc/EglBase$Context;I)V
+	public synthetic fun <init> (Landroid/content/Context;Lio/getstream/video/android/core/Call;Lkotlinx/coroutines/CoroutineScope;Lorg/webrtc/EglBase$Context;IILkotlin/jvm/internal/DefaultConstructorMarker;)V
 	public final fun cleanup ()V
 	public final fun getAudioSource ()Lorg/webrtc/AudioSource;
 	public final fun getAudioTrack ()Lorg/webrtc/AudioTrack;
+	public final fun getAudioUsage ()I
 	public final fun getCall ()Lio/getstream/video/android/core/Call;
 	public final fun getContext ()Landroid/content/Context;
 	public final fun getEglBaseContext ()Lorg/webrtc/EglBase$Context;
@@ -476,10 +478,12 @@ public final class io/getstream/video/android/core/MemberState {
 }
 
 public final class io/getstream/video/android/core/MicrophoneManager {
-	public fun <init> (Lio/getstream/video/android/core/MediaManagerImpl;Z)V
+	public fun <init> (Lio/getstream/video/android/core/MediaManagerImpl;ZI)V
+	public final fun canHandleDeviceSwitch ()Z
 	public final fun cleanup ()V
 	public final fun disable (Z)V
 	public static synthetic fun disable$default (Lio/getstream/video/android/core/MicrophoneManager;ZILjava/lang/Object;)V
+	public final fun getAudioUsage ()I
 	public final fun getDevices ()Lkotlinx/coroutines/flow/StateFlow;
 	public final fun getMediaManager ()Lio/getstream/video/android/core/MediaManagerImpl;
 	public final fun getPreferSpeakerphone ()Z
@@ -817,7 +821,8 @@ public final class io/getstream/video/android/core/StreamVideoBuilder {
 	public fun <init> (Landroid/content/Context;Ljava/lang/String;Lio/getstream/video/android/core/GEO;Lio/getstream/video/android/model/User;Ljava/lang/String;Lkotlin/jvm/functions/Function2;Lio/getstream/video/android/core/logging/LoggingLevel;Lio/getstream/video/android/core/notifications/NotificationConfig;Lkotlin/jvm/functions/Function1;JZLjava/lang/String;ZLjava/lang/String;Lio/getstream/video/android/core/sounds/Sounds;)V
 	public fun <init> (Landroid/content/Context;Ljava/lang/String;Lio/getstream/video/android/core/GEO;Lio/getstream/video/android/model/User;Ljava/lang/String;Lkotlin/jvm/functions/Function2;Lio/getstream/video/android/core/logging/LoggingLevel;Lio/getstream/video/android/core/notifications/NotificationConfig;Lkotlin/jvm/functions/Function1;JZLjava/lang/String;ZLjava/lang/String;Lio/getstream/video/android/core/sounds/Sounds;Z)V
 	public fun <init> (Landroid/content/Context;Ljava/lang/String;Lio/getstream/video/android/core/GEO;Lio/getstream/video/android/model/User;Ljava/lang/String;Lkotlin/jvm/functions/Function2;Lio/getstream/video/android/core/logging/LoggingLevel;Lio/getstream/video/android/core/notifications/NotificationConfig;Lkotlin/jvm/functions/Function1;JZLjava/lang/String;ZLjava/lang/String;Lio/getstream/video/android/core/sounds/Sounds;ZLio/getstream/video/android/core/permission/android/StreamPermissionCheck;)V
-	public synthetic fun <init> (Landroid/content/Context;Ljava/lang/String;Lio/getstream/video/android/core/GEO;Lio/getstream/video/android/model/User;Ljava/lang/String;Lkotlin/jvm/functions/Function2;Lio/getstream/video/android/core/logging/LoggingLevel;Lio/getstream/video/android/core/notifications/NotificationConfig;Lkotlin/jvm/functions/Function1;JZLjava/lang/String;ZLjava/lang/String;Lio/getstream/video/android/core/sounds/Sounds;ZLio/getstream/video/android/core/permission/android/StreamPermissionCheck;ILkotlin/jvm/internal/DefaultConstructorMarker;)V
+	public fun <init> (Landroid/content/Context;Ljava/lang/String;Lio/getstream/video/android/core/GEO;Lio/getstream/video/android/model/User;Ljava/lang/String;Lkotlin/jvm/functions/Function2;Lio/getstream/video/android/core/logging/LoggingLevel;Lio/getstream/video/android/core/notifications/NotificationConfig;Lkotlin/jvm/functions/Function1;JZLjava/lang/String;ZLjava/lang/String;Lio/getstream/video/android/core/sounds/Sounds;ZLio/getstream/video/android/core/permission/android/StreamPermissionCheck;I)V
+	public synthetic fun <init> (Landroid/content/Context;Ljava/lang/String;Lio/getstream/video/android/core/GEO;Lio/getstream/video/android/model/User;Ljava/lang/String;Lkotlin/jvm/functions/Function2;Lio/getstream/video/android/core/logging/LoggingLevel;Lio/getstream/video/android/core/notifications/NotificationConfig;Lkotlin/jvm/functions/Function1;JZLjava/lang/String;ZLjava/lang/String;Lio/getstream/video/android/core/sounds/Sounds;ZLio/getstream/video/android/core/permission/android/StreamPermissionCheck;IILkotlin/jvm/internal/DefaultConstructorMarker;)V
 	public final fun build ()Lio/getstream/video/android/core/StreamVideo;
 	public final fun getScope ()Lkotlinx/coroutines/CoroutineScope;
 }
@@ -1024,7 +1029,8 @@ public final class io/getstream/video/android/core/call/connection/StreamPeerCon
 }
 
 public final class io/getstream/video/android/core/call/connection/StreamPeerConnectionFactory {
-	public fun <init> (Landroid/content/Context;)V
+	public fun <init> (Landroid/content/Context;I)V
+	public synthetic fun <init> (Landroid/content/Context;IILkotlin/jvm/internal/DefaultConstructorMarker;)V
 	public final fun getEglBase ()Lorg/webrtc/EglBase;
 	public final fun makeAudioSource (Lorg/webrtc/MediaConstraints;)Lorg/webrtc/AudioSource;
 	public static synthetic fun makeAudioSource$default (Lio/getstream/video/android/core/call/connection/StreamPeerConnectionFactory;Lorg/webrtc/MediaConstraints;ILjava/lang/Object;)Lorg/webrtc/AudioSource;

diff --git a/stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/Call.kt b/stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/Call.kt
@@ -214,6 +214,7 @@ public class Call(
                 this,
                 scope,
                 clientImpl.peerConnectionFactory.eglBase.eglBaseContext,
+                clientImpl.audioUsage,
             )
         }
     }

diff --git a/stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/MediaManager.kt b/stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/MediaManager.kt
@@ -30,6 +30,7 @@ import android.os.IBinder
 import androidx.core.content.ContextCompat
 import androidx.core.content.getSystemService
 import io.getstream.log.taggedLogger
+import io.getstream.video.android.core.audio.AudioHandler
 import io.getstream.video.android.core.audio.AudioSwitchHandler
 import io.getstream.video.android.core.audio.StreamAudioDevice
 import io.getstream.video.android.core.audio.StreamAudioDevice.Companion.fromAudio
@@ -328,11 +329,12 @@ class ScreenShareManager(
 class MicrophoneManager(
     val mediaManager: MediaManagerImpl,
     val preferSpeakerphone: Boolean,
+    val audioUsage: Int,
 ) {
     // Internal data
     private val logger by taggedLogger("Media:MicrophoneManager")
 
-    private lateinit var audioHandler: AudioSwitchHandler
+    private lateinit var audioHandler: AudioHandler
     private var setupCompleted: Boolean = false
     internal var audioManager: AudioManager? = null
     internal var priorStatus: DeviceStatus? = null
@@ -434,6 +436,8 @@ class MicrophoneManager(
         setupCompleted = false
     }
 
+    fun canHandleDeviceSwitch() = audioUsage != AudioAttributes.USAGE_MEDIA
+
     // Internal logic
     internal fun setup() {
         if (setupCompleted) {
@@ -445,14 +449,18 @@ class MicrophoneManager(
             audioManager?.allowedCapturePolicy = AudioAttributes.ALLOW_CAPTURE_BY_ALL
         }
 
-        audioHandler =
-            AudioSwitchHandler(mediaManager.context, preferSpeakerphone) { devices, selected ->
-                logger.i { "audio devices. selected $selected, available devices are $devices" }
-                _devices.value = devices.map { it.fromAudio() }
-                _selectedDevice.value = selected?.fromAudio()
-            }
+        if (canHandleDeviceSwitch()) {
+            audioHandler =
+                AudioSwitchHandler(mediaManager.context, preferSpeakerphone) { devices, selected ->
+                    logger.i { "audio devices. selected $selected, available devices are $devices" }
+                    _devices.value = devices.map { it.fromAudio() }
+                    _selectedDevice.value = selected?.fromAudio()
+                }
 
-        audioHandler.start()
+            audioHandler.start()
+        } else {
+            logger.d { "[MediaManager#setup] usage is MEDIA, cannot handle device switch" }
+        }
         setupCompleted = true
     }
 
@@ -463,7 +471,7 @@ class MicrophoneManager(
 
     private fun ifAudioHandlerInitialized(then: (audioHandler: AudioSwitchHandler) -> Unit) {
         if (this::audioHandler.isInitialized) {
-            then(this.audioHandler)
+            then(this.audioHandler as AudioSwitchHandler)
         } else {
             logger.e { "Audio handler not initialized. Ensure calling setup(), before using the handler." }
         }
@@ -802,6 +810,7 @@ class MediaManagerImpl(
     val call: Call,
     val scope: CoroutineScope,
     val eglBaseContext: EglBase.Context,
+    val audioUsage: Int = defaultAudioUsage,
 ) {
     private val filterVideoProcessor =
         FilterVideoProcessor({ call.videoFilter }, { camera.surfaceTextureHelper })
@@ -838,7 +847,7 @@ class MediaManagerImpl(
     )
 
     internal val camera = CameraManager(this, eglBaseContext)
-    internal val microphone = MicrophoneManager(this, preferSpeakerphone = true)
+    internal val microphone = MicrophoneManager(this, preferSpeakerphone = true, audioUsage)
     internal val speaker = SpeakerManager(this, microphone)
     internal val screenShare = ScreenShareManager(this, eglBaseContext)
 

diff --git a/...-video-android-core/src/main/kotlin/io/getstream/video/android/core/StreamVideoBuilder.kt b/...-video-android-core/src/main/kotlin/io/getstream/video/android/core/StreamVideoBuilder.kt
@@ -71,6 +71,7 @@ import kotlinx.coroutines.launch
  * @property sounds Overwrite the default SDK sounds. See [Sounds].
  * @property crashOnMissingPermission if [permissionCheck] returns false there will be an exception.
  * @property permissionCheck used to check for system permission based on call capabilities. See [StreamPermissionCheck].
+ * @property audioUsage used to signal to the system how to treat the audio tracks (voip or media).
  */
 public class StreamVideoBuilder @JvmOverloads constructor(
     context: Context,
@@ -90,6 +91,7 @@ public class StreamVideoBuilder @JvmOverloads constructor(
     private val sounds: Sounds = Sounds(),
     private val crashOnMissingPermission: Boolean = true,
     private val permissionCheck: StreamPermissionCheck = DefaultStreamPermissionCheck(),
+    private val audioUsage: Int = defaultAudioUsage,
 ) {
     private val context: Context = context.applicationContext
 
@@ -171,6 +173,7 @@ public class StreamVideoBuilder @JvmOverloads constructor(
             testSfuAddress = localSfuAddress,
             sounds = sounds,
             permissionCheck = permissionCheck,
+            audioUsage = audioUsage,
         )
 
         if (user.type == UserType.Guest) {

diff --git a/stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/StreamVideoImpl.kt b/stream-video-android-core/src/main/kotlin/io/getstream/video/android/core/StreamVideoImpl.kt
@@ -17,6 +17,7 @@
 package io.getstream.video.android.core
 
 import android.content.Context
+import android.media.AudioAttributes
 import androidx.lifecycle.Lifecycle
 import io.getstream.android.push.PushDevice
 import io.getstream.log.taggedLogger
@@ -125,6 +126,7 @@ import kotlin.coroutines.Continuation
 import kotlin.coroutines.resumeWithException
 
 internal const val WAIT_FOR_CONNECTION_ID_TIMEOUT = 5000L
+internal const val defaultAudioUsage = AudioAttributes.USAGE_VOICE_COMMUNICATION
 
 /**
  * @param lifecycle The lifecycle used to observe changes in the process
@@ -145,6 +147,7 @@ internal class StreamVideoImpl internal constructor(
     internal val sounds: Sounds,
     internal val crashOnMissingPermission: Boolean = true,
     internal val permissionCheck: StreamPermissionCheck = DefaultStreamPermissionCheck(),
+    internal val audioUsage: Int = defaultAudioUsage,
 ) : StreamVideo, NotificationHandler by streamNotificationManager {
 
     private var locationJob: Deferred<Result<String>>? = null
@@ -165,7 +168,7 @@ internal class StreamVideoImpl internal constructor(
     private lateinit var connectContinuation: Continuation<Result<ConnectedEvent>>
 
     @InternalStreamVideoApi
-    public var peerConnectionFactory = StreamPeerConnectionFactory(context)
+    public var peerConnectionFactory = StreamPeerConnectionFactory(context, audioUsage)
     public override val userId = user.id
 
     private val logger by taggedLogger("Call:StreamVideo")

diff --git a/...ain/kotlin/io/getstream/video/android/core/call/connection/StreamPeerConnectionFactory.kt b/...ain/kotlin/io/getstream/video/android/core/call/connection/StreamPeerConnectionFactory.kt
@@ -17,9 +17,11 @@
 package io.getstream.video.android.core.call.connection
 
 import android.content.Context
+import android.media.AudioAttributes
 import android.os.Build
 import io.getstream.log.taggedLogger
 import io.getstream.video.android.core.call.video.FilterVideoProcessor
+import io.getstream.video.android.core.defaultAudioUsage
 import io.getstream.video.android.core.model.IceCandidate
 import io.getstream.video.android.core.model.StreamPeerType
 import kotlinx.coroutines.CoroutineScope
@@ -44,8 +46,13 @@ import java.nio.ByteBuffer
  * Builds a factory that provides [PeerConnection]s when requested.
  *
  * @param context Used to build the underlying native components for the factory.
+ * @param audioUsage signal to the system how the audio tracks are used.
+ * Set this to [AudioAttributes.USAGE_MEDIA] if you want the audio track to behave like media, useful for livestreaming scenarios.
  */
-public class StreamPeerConnectionFactory(private val context: Context) {
+public class StreamPeerConnectionFactory(
+    private val context: Context,
+    private val audioUsage: Int = defaultAudioUsage,
+) {
 
     private val webRtcLogger by taggedLogger("Call:WebRTC")
     private val audioLogger by taggedLogger("Call:AudioTrackCallback")
@@ -151,7 +158,15 @@ public class StreamPeerConnectionFactory(private val context: Context) {
                     .builder(context)
                     .setUseHardwareAcousticEchoCanceler(
                         Build.VERSION.SDK_INT >= Build.VERSION_CODES.Q,
-                    )
+                    ).apply {
+                        if (audioUsage != defaultAudioUsage) {
+                            setAudioAttributes(
+                                AudioAttributes.Builder().setUsage(audioUsage)
+                                    .build(),
+                            )
+                            audioLogger.d { "[setAudioAttributes] usage: $audioUsage" }
+                        }
+                    }
                     .setUseHardwareNoiseSuppressor(Build.VERSION.SDK_INT >= Build.VERSION_CODES.Q)
                     .setAudioRecordErrorCallback(object :
                         JavaAudioDeviceModule.AudioRecordErrorCallback {
@@ -297,7 +312,10 @@ public class StreamPeerConnectionFactory(private val context: Context) {
      * @return [VideoSource] that can be used to build tracks.
      */
 
-    internal fun makeVideoSource(isScreencast: Boolean, filterVideoProcessor: FilterVideoProcessor): VideoSource =
+    internal fun makeVideoSource(
+        isScreencast: Boolean,
+        filterVideoProcessor: FilterVideoProcessor,
+    ): VideoSource =
         factory.createVideoSource(isScreencast).apply {
             setVideoProcessor(filterVideoProcessor)
         }

diff --git a/...deo-android-core/src/test/kotlin/io/getstream/video/android/core/MicrophoneManagerTest.kt b/...deo-android-core/src/test/kotlin/io/getstream/video/android/core/MicrophoneManagerTest.kt
@@ -17,6 +17,7 @@
 package io.getstream.video.android.core
 
 import android.content.Context
+import android.media.AudioAttributes
 import android.media.AudioManager
 import io.mockk.every
 import io.mockk.mockk
@@ -28,11 +29,13 @@ import org.junit.Test
 
 class MicrophoneManagerTest {
 
+    private val audioUsage = AudioAttributes.USAGE_VOICE_COMMUNICATION
+
     @Test
     fun `Ensure setup is called prior to any action onto the microphone manager`() = runTest {
         // Given
         val mediaManager = mockk<MediaManagerImpl>(relaxed = true)
-        val actual = MicrophoneManager(mediaManager, false)
+        val actual = MicrophoneManager(mediaManager, false, audioUsage)
         val context = mockk<Context>(relaxed = true)
         val microphoneManager = spyk(actual)
         every { mediaManager.context } returns context
@@ -63,7 +66,12 @@ class MicrophoneManagerTest {
     fun `Don't crash when accessing audioHandler prior to setup`() {
         // Given
         val mediaManager = mockk<MediaManagerImpl>(relaxed = true)
-        val actual = MicrophoneManager(mediaManager, false)
+        val actual =
+            MicrophoneManager(
+                mediaManager,
+                false,
+                audioUsage = AudioAttributes.USAGE_VOICE_COMMUNICATION,
+            )
         val context = mockk<Context>(relaxed = true)
         val microphoneManager = spyk(actual)
         every { mediaManager.context } returns context
@@ -83,7 +91,7 @@ class MicrophoneManagerTest {
     fun `Ensure setup if ever the manager was cleaned`() {
         // Given
         val mediaManager = mockk<MediaManagerImpl>(relaxed = true)
-        val actual = MicrophoneManager(mediaManager, false)
+        val actual = MicrophoneManager(mediaManager, false, audioUsage)
         val context = mockk<Context>(relaxed = true)
         val microphoneManager = spyk(actual)
         every { mediaManager.context } returns context
@@ -111,7 +119,7 @@ class MicrophoneManagerTest {
     fun `Resume will call enable only if prior status was DeviceStatus#enabled`() {
         // Given
         val mediaManager = mockk<MediaManagerImpl>(relaxed = true)
-        val actual = MicrophoneManager(mediaManager, false)
+        val actual = MicrophoneManager(mediaManager, false, audioUsage)
         val context = mockk<Context>(relaxed = true)
         val microphoneManager = spyk(actual)
         every { mediaManager.context } returns context