AudioUnit: Don't rely on category switch for mic indicator to turn off (

#52)

* progress

* tweak

* clean

* simplify audio unit restart

call to SetupAudioBuffersForActiveAudioSession() might not be needed since sample rate won't change during restart. This might help reduce the unwanted noise when restarting audio unit.

* clean

sdk/objc/components/audio/RTCAudioSession+Private.h

@@ -35,8 +35,6 @@ NS_ASSUME_NONNULL_BEGIN
  */
 @property(nonatomic, assign) BOOL isInterrupted;
 
-@property(nonatomic, strong) NSString *activeCategory;
-
 /** Adds the delegate to the list of delegates, and places it at the front of
  *  the list. This delegate will be notified before other delegates of
  *  audio events.

sdk/objc/components/audio/RTCAudioSession.h

@@ -102,9 +102,6 @@ RTC_OBJC_EXPORT
 - (void)audioSession:(RTC_OBJC_TYPE(RTCAudioSession) *)audioSession
     audioUnitStartFailedWithError:(NSError *)error;
 
-/** Called when audio session changed from output-only to input & output */
-- (void)audioSessionDidChangeRecordingEnabled:(RTC_OBJC_TYPE(RTCAudioSession) *)audioSession;
-
 @end
 
 /** This is a protocol used to inform RTCAudioSession when the audio session

sdk/objc/components/audio/RTCAudioSession.mm

@@ -114,8 +114,6 @@ - (instancetype)initWithAudioSession:(id)audioSession {
                   options:NSKeyValueObservingOptionNew | NSKeyValueObservingOptionOld
                   context:(__bridge void *)RTC_OBJC_TYPE(RTCAudioSession).class];
 
-    _activeCategory = _session.category;
-
     RTCLog(@"RTC_OBJC_TYPE(RTCAudioSession) (%p): init.", self);
   }
   return self;
@@ -543,13 +541,6 @@ - (void)handleRouteChangeNotification:(NSNotification *)notification {
       break;
     case AVAudioSessionRouteChangeReasonCategoryChange:
       RTCLog(@"Audio route changed: CategoryChange to :%@", self.session.category);
-      {
-        if (![_session.category isEqualToString:_activeCategory]) {
-          _activeCategory = _session.category;
-          RTCLog(@"Audio route changed: Restarting Audio Unit");
-          [self notifyDidChangeAudioSessionRecordingEnabled];
-        }
-      }
       break;
     case AVAudioSessionRouteChangeReasonOverride:
       RTCLog(@"Audio route changed: Override");
@@ -1005,13 +996,4 @@ - (void)notifyFailedToSetActive:(BOOL)active error:(NSError *)error {
   }
 }
 
-- (void)notifyDidChangeAudioSessionRecordingEnabled {
-  for (auto delegate : self.delegates) {
-    SEL sel = @selector(audioSessionDidChangeRecordingEnabled:);
-    if ([delegate respondsToSelector:sel]) {
-      [delegate audioSessionDidChangeRecordingEnabled:self];
-    }
-  }
-}
-
 @end

sdk/objc/components/audio/RTCNativeAudioSessionDelegateAdapter.mm

@@ -86,9 +86,4 @@ - (void)audioSession:(RTC_OBJC_TYPE(RTCAudioSession) *)audioSession
   _observer->OnChangedOutputVolume();
 }
 
-- (void)audioSessionDidChangeRecordingEnabled:(RTC_OBJC_TYPE(RTCAudioSession) *)session {
-  // re-trigger audio unit init, by using interrupt ended callback
-  _observer->OnChangedRecordingEnabled();
-}
-
 @end

sdk/objc/native/src/audio/audio_device_ios.h

@@ -149,7 +149,6 @@ class AudioDeviceIOS : public AudioDeviceGeneric,
   void OnValidRouteChange() override;
   void OnCanPlayOrRecordChange(bool can_play_or_record) override;
   void OnChangedOutputVolume() override;
-  void OnChangedRecordingEnabled() override;
 
   // VoiceProcessingAudioUnitObserver methods.
   OSStatus OnDeliverRecordedData(AudioUnitRenderActionFlags* flags,
@@ -174,7 +173,8 @@ class AudioDeviceIOS : public AudioDeviceGeneric,
   void HandleSampleRateChange();
   void HandlePlayoutGlitchDetected();
   void HandleOutputVolumeChange();
-  void HandleAudioSessionRecordingEnabledChange();
+
+  bool RestartAudioUnit(bool enable_input);
 
   // Uses current `playout_parameters_` and `record_parameters_` to inform the
   // audio device buffer (ADB) about our internal audio parameters.
@@ -204,7 +204,7 @@ class AudioDeviceIOS : public AudioDeviceGeneric,
 
   // Activates our audio session, creates and initializes the voice-processing
   // audio unit and verifies that we got the preferred native audio parameters.
-  bool InitPlayOrRecord();
+  bool InitPlayOrRecord(bool enable_input);
 
   // Closes and deletes the voice-processing I/O unit.
   void ShutdownPlayOrRecord();
@@ -264,19 +264,21 @@ class AudioDeviceIOS : public AudioDeviceGeneric,
   // will be changed dynamically to account for this behavior.
   rtc::BufferT<int16_t> record_audio_buffer_;
 
+  // Set to 1 when recording is initialized and 0 otherwise.
+  volatile int recording_is_initialized_;
+
   // Set to 1 when recording is active and 0 otherwise.
   std::atomic<int> recording_;
 
+  // Set to 1 when playout is initialized and 0 otherwise.
+  volatile int playout_is_initialized_;
+
   // Set to 1 when playout is active and 0 otherwise.
   std::atomic<int> playing_;
 
   // Set to true after successful call to Init(), false otherwise.
   bool initialized_ RTC_GUARDED_BY(thread_);
 
-  // Set to true after successful call to InitRecording() or InitPlayout(),
-  // false otherwise.
-  bool audio_is_initialized_;
-
   // Set to true if audio session is interrupted, false otherwise.
   bool is_interrupted_;
 

sdk/objc/native/src/audio/audio_device_ios.mm

@@ -68,7 +68,6 @@
   kMessageTypeCanPlayOrRecordChange,
   kMessageTypePlayoutGlitchDetected,
   kMessageOutputVolumeChange,
-  kMessageTypeRecordingEnabledChange,
 };
 
 using ios::CheckAndLogError;
@@ -104,10 +103,11 @@ static void LogDeviceInfo() {
     : bypass_voice_processing_(bypass_voice_processing),
       audio_device_buffer_(nullptr),
       audio_unit_(nullptr),
+      recording_is_initialized_(0),
       recording_(0),
+      playout_is_initialized_(0),
       playing_(0),
       initialized_(false),
-      audio_is_initialized_(false),
       is_interrupted_(false),
       has_configured_session_(false),
       num_detected_playout_glitches_(0),
@@ -194,7 +194,9 @@ static void LogDeviceInfo() {
       return -1;
     }
   }
-  audio_is_initialized_ = true;
+
+  rtc::AtomicOps::ReleaseStore(&playout_is_initialized_, 1);
+
   return 0;
 }
 
@@ -219,8 +221,13 @@ static void LogDeviceInfo() {
       RTC_LOG_F(LS_ERROR) << "InitPlayOrRecord failed for InitRecording!";
       return -1;
     }
+  } else {
+    // playout already initialized, restart audio unit with input
+    RestartAudioUnit(true);
   }
-  audio_is_initialized_ = true;
+
+  rtc::AtomicOps::ReleaseStore(&recording_is_initialized_, 1);
+
   return 0;
 }
 
@@ -257,7 +264,8 @@ static void LogDeviceInfo() {
   }
   if (!recording_.load()) {
     ShutdownPlayOrRecord();
-    audio_is_initialized_ = false;
+
+    rtc::AtomicOps::ReleaseStore(&recording_is_initialized_, 0);
   }
   playing_.store(0, std::memory_order_release);
 
@@ -311,7 +319,11 @@ static void LogDeviceInfo() {
   }
   if (!playing_.load()) {
     ShutdownPlayOrRecord();
-    audio_is_initialized_ = false;
+
+    rtc::AtomicOps::ReleaseStore(&playout_is_initialized_, 0);
+  } else if (playout_is_initialized_) {
+    // restart audio unit with no input
+    RestartAudioUnit(false);
   }
   recording_.store(0, std::memory_order_release);
   return 0;
@@ -370,11 +382,6 @@ static void LogDeviceInfo() {
   thread_->PostTask(SafeTask(safety_, [this] { HandleOutputVolumeChange(); }));
 }
 
-void AudioDeviceIOS::OnChangedRecordingEnabled() {
-  RTC_DCHECK(thread_);
-  thread_->Post(RTC_FROM_HERE, this, kMessageTypeRecordingEnabledChange);
-}
-
 OSStatus AudioDeviceIOS::OnDeliverRecordedData(AudioUnitRenderActionFlags* flags,
                                                const AudioTimeStamp* time_stamp,
                                                UInt32 bus_number,
@@ -505,9 +512,6 @@ static void LogDeviceInfo() {
     case kMessageOutputVolumeChange:
       HandleOutputVolumeChange();
       break;
-    case kMessageTypeRecordingEnabledChange:
-      HandleAudioSessionRecordingEnabledChange();
-      break;
   }
 }
 
@@ -623,7 +627,7 @@ static void LogDeviceInfo() {
   SetupAudioBuffersForActiveAudioSession();
 
   // Initialize the audio unit again with the new sample rate.
-  if (!audio_unit_->Initialize(playout_parameters_.sample_rate())) {
+  if (!audio_unit_->Initialize(playout_parameters_.sample_rate(), recording_is_initialized_)) {
     RTCLogError(@"Failed to initialize the audio unit with sample rate: %d",
                 playout_parameters_.sample_rate());
     return;
@@ -677,59 +681,44 @@ static void LogDeviceInfo() {
   last_output_volume_change_time_ = rtc::TimeMillis();
 }
 
-void AudioDeviceIOS::HandleAudioSessionRecordingEnabledChange() {
+bool AudioDeviceIOS::RestartAudioUnit(bool enable_input) {
   RTC_DCHECK_RUN_ON(&thread_checker_);
 
-  LOGI() << "HandleAudioSessionRecordingEnabledChange";
+  LOGI() << "RestartAudioUnit";
 
   // If we don't have an audio unit yet, or the audio unit is uninitialized,
   // there is no work to do.
   if (!audio_unit_ || audio_unit_->GetState() < VoiceProcessingAudioUnit::kInitialized) {
-    return;
+    return false;
   }
 
-  // The audio unit is already initialized or started.
-  // Check to see if the sample rate or buffer size has changed.
-  RTC_OBJC_TYPE(RTCAudioSession)* session = [RTC_OBJC_TYPE(RTCAudioSession) sharedInstance];
-  const double session_sample_rate = session.sampleRate;
-
-  // Extra sanity check to ensure that the new sample rate is valid.
-  if (session_sample_rate <= 0.0) {
-    RTCLogError(@"Sample rate is invalid: %f", session_sample_rate);
-    LOGI() << "Sample rate is invalid " << session_sample_rate;
-    return;
-  }
-  // We need to adjust our format and buffer sizes.
-  // The stream format is about to be changed and it requires that we first
-  // stop and uninitialize the audio unit to deallocate its resources.
-  RTCLog(@"Stopping and uninitializing audio unit to adjust buffers.");
   bool restart_audio_unit = false;
   if (audio_unit_->GetState() == VoiceProcessingAudioUnit::kStarted) {
     audio_unit_->Stop();
-    restart_audio_unit = true;
     PrepareForNewStart();
+    restart_audio_unit = true;
   }
+
   if (audio_unit_->GetState() == VoiceProcessingAudioUnit::kInitialized) {
     audio_unit_->Uninitialize();
   }
 
-  // Allocate new buffers given the new stream format.
-  SetupAudioBuffersForActiveAudioSession();
+  // Initialize the audio unit again with the same sample rate.
+  const double sample_rate = playout_parameters_.sample_rate();
 
-  // Initialize the audio unit again with the new sample rate.
-  RTC_DCHECK_EQ(playout_parameters_.sample_rate(), session_sample_rate);
-  if (!audio_unit_->Initialize(session_sample_rate)) {
-    RTCLogError(@"Failed to initialize the audio unit with sample rate: %f", session_sample_rate);
-    return;
+  if (!audio_unit_->Initialize(sample_rate, enable_input)) {
+    RTCLogError(@"Failed to initialize the audio unit with sample rate: %f", sample_rate);
+    return false;
   }
 
   // Restart the audio unit if it was already running.
   if (restart_audio_unit && !audio_unit_->Start()) {
-    RTCLogError(@"Failed to start audio unit with sample rate: %f", session_sample_rate);
-    return;
+    RTCLogError(@"Failed to start audio unit with sample rate: %f", sample_rate);
+    return false;
   }
 
   LOGI() << "Successfully enabled audio unit for recording.";
+  return true;
 }
 
 void AudioDeviceIOS::UpdateAudioDeviceBuffer() {
@@ -825,7 +814,7 @@ static void LogDeviceInfo() {
 
   // If we're not initialized we don't need to do anything. Audio unit will
   // be initialized on initialization.
-  if (!audio_is_initialized_) return;
+  if (!playout_is_initialized_ && !recording_is_initialized_) return;
 
   // If we're initialized, we must have an audio unit.
   RTC_DCHECK(audio_unit_);
@@ -863,7 +852,7 @@ static void LogDeviceInfo() {
     RTCLog(@"Initializing audio unit for UpdateAudioUnit");
     ConfigureAudioSession();
     SetupAudioBuffersForActiveAudioSession();
-    if (!audio_unit_->Initialize(playout_parameters_.sample_rate())) {
+    if (!audio_unit_->Initialize(playout_parameters_.sample_rate(), recording_is_initialized_)) {
       RTCLogError(@"Failed to initialize audio unit.");
       return;
     }
@@ -953,7 +942,7 @@ static void LogDeviceInfo() {
   RTCLog(@"Unconfigured audio session.");
 }
 
-bool AudioDeviceIOS::InitPlayOrRecord() {
+bool AudioDeviceIOS::InitPlayOrRecord(bool enable_input) {
   LOGI() << "InitPlayOrRecord";
   RTC_DCHECK_RUN_ON(thread_);
 
@@ -989,7 +978,7 @@ static void LogDeviceInfo() {
       return false;
     }
     SetupAudioBuffersForActiveAudioSession();
-    audio_unit_->Initialize(playout_parameters_.sample_rate());
+    audio_unit_->Initialize(playout_parameters_.sample_rate(), enable_input);
   }
 
   // Release the lock.

sdk/objc/native/src/audio/audio_session_observer.h

@@ -32,8 +32,6 @@ class AudioSessionObserver {
 
   virtual void OnChangedOutputVolume() = 0;
 
-  virtual void OnChangedRecordingEnabled() = 0;
-
  protected:
   virtual ~AudioSessionObserver() {}
 };

sdk/objc/native/src/audio/voice_processing_audio_unit.h

@@ -75,7 +75,7 @@ class VoiceProcessingAudioUnit {
   VoiceProcessingAudioUnit::State GetState() const;
 
   // Initializes the underlying audio unit with the given sample rate.
-  bool Initialize(Float64 sample_rate);
+  bool Initialize(Float64 sample_rate, bool enable_input);
 
   // Starts the underlying audio unit.
   OSStatus Start();

sdk/objc/native/src/audio/voice_processing_audio_unit.mm

@@ -180,7 +180,7 @@ static OSStatus GetAGCState(AudioUnit audio_unit, UInt32* enabled) {
   return state_;
 }
 
-bool VoiceProcessingAudioUnit::Initialize(Float64 sample_rate) {
+bool VoiceProcessingAudioUnit::Initialize(Float64 sample_rate, bool enable_input) {
   RTC_DCHECK_GE(state_, kUninitialized);
   RTCLog(@"Initializing audio unit with sample rate: %f", sample_rate);
 
@@ -191,19 +191,11 @@ static OSStatus GetAGCState(AudioUnit audio_unit, UInt32* enabled) {
   LogStreamDescription(format);
 #endif
 
-  // Enable input on the input scope of the input element.
-  // keep it disabled if audio session is configured for playback only
-  AVAudioSession* session = [AVAudioSession sharedInstance];
-  UInt32 enable_input = 0;
-  if ([session.category isEqualToString: AVAudioSessionCategoryPlayAndRecord] ||
-      [session.category isEqualToString: AVAudioSessionCategoryRecord]) {
-    enable_input = 1;
-  }
-  RTCLog(@"Initializing AudioUnit, category=%@, enable_input=%d", session.category, (int) enable_input);
-  // LOGI() << "Initialize" << session.category << ", enable_input=" << enable_input;
+  UInt32 _enable_input = enable_input ? 1 : 0;
+  RTCLog(@"Initializing AudioUnit, _enable_input=%d", (int) _enable_input);
   result = AudioUnitSetProperty(vpio_unit_, kAudioOutputUnitProperty_EnableIO,
-                                kAudioUnitScope_Input, kInputBus, &enable_input,
-                                sizeof(enable_input));
+                                kAudioUnitScope_Input, kInputBus, &_enable_input,
+                                sizeof(_enable_input));
   if (result != noErr) {
     DisposeAudioUnit();
     RTCLogError(@"Failed to enable input on input scope of input element. "