Make capture timestamp optional in ADM.

This is to avoid using 0 as a default value.

Also fix a bug in audio_device_buffer where the timestamp aligner used the wrong input timestamp.

Bug: webrtc:13609
Change-Id: I00016e68ab50d052990c2b9f80aa1e2d7e167b93
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/291118
Reviewed-by: Olov Brändström <brandstrom@google.com>
Commit-Queue: Jakob Ivarsson‎ <jakobi@webrtc.org>
Reviewed-by: Henrik Andreassson <henrika@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#39177}

audio/audio_transport_impl.cc

@@ -107,35 +107,35 @@ AudioTransportImpl::~AudioTransportImpl() {}
 
 int32_t AudioTransportImpl::RecordedDataIsAvailable(
     const void* audio_data,
-    const size_t number_of_frames,
-    const size_t bytes_per_sample,
-    const size_t number_of_channels,
-    const uint32_t sample_rate,
-    const uint32_t audio_delay_milliseconds,
-    const int32_t clock_drift,
-    const uint32_t volume,
-    const bool key_pressed,
+    size_t number_of_frames,
+    size_t bytes_per_sample,
+    size_t number_of_channels,
+    uint32_t sample_rate,
+    uint32_t audio_delay_milliseconds,
+    int32_t clock_drift,
+    uint32_t volume,
+    bool key_pressed,
     uint32_t& new_mic_volume) {  // NOLINT: to avoid changing APIs
   return RecordedDataIsAvailable(
       audio_data, number_of_frames, bytes_per_sample, number_of_channels,
       sample_rate, audio_delay_milliseconds, clock_drift, volume, key_pressed,
-      new_mic_volume, /* estimated_capture_time_ns */ 0);
+      new_mic_volume, /*estimated_capture_time_ns=*/absl::nullopt);
 }
 
 // Not used in Chromium. Process captured audio and distribute to all sending
 // streams, and try to do this at the lowest possible sample rate.
 int32_t AudioTransportImpl::RecordedDataIsAvailable(
     const void* audio_data,
-    const size_t number_of_frames,
-    const size_t bytes_per_sample,
-    const size_t number_of_channels,
-    const uint32_t sample_rate,
-    const uint32_t audio_delay_milliseconds,
-    const int32_t /*clock_drift*/,
-    const uint32_t /*volume*/,
-    const bool key_pressed,
+    size_t number_of_frames,
+    size_t bytes_per_sample,
+    size_t number_of_channels,
+    uint32_t sample_rate,
+    uint32_t audio_delay_milliseconds,
+    int32_t /*clock_drift*/,
+    uint32_t /*volume*/,
+    bool key_pressed,
     uint32_t& /*new_mic_volume*/,
-    const int64_t
+    absl::optional<int64_t>
         estimated_capture_time_ns) {  // NOLINT: to avoid changing APIs
   RTC_DCHECK(audio_data);
   RTC_DCHECK_GE(number_of_channels, 1);
@@ -166,8 +166,11 @@ int32_t AudioTransportImpl::RecordedDataIsAvailable(
   ProcessCaptureFrame(audio_delay_milliseconds, key_pressed,
                       swap_stereo_channels, audio_processing_,
                       audio_frame.get());
-  audio_frame->set_absolute_capture_timestamp_ms(estimated_capture_time_ns /
+
+  if (estimated_capture_time_ns) {
+    audio_frame->set_absolute_capture_timestamp_ms(*estimated_capture_time_ns /
                                                    1000000);
+  }
 
   RTC_DCHECK_GT(audio_frame->samples_per_channel_, 0);
   if (async_audio_processing_)

audio/audio_transport_impl.h

@@ -52,7 +52,8 @@ class AudioTransportImpl : public AudioTransport {
                                   bool keyPressed,
                                   uint32_t& newMicLevel) override;
 
-  int32_t RecordedDataIsAvailable(const void* audioSamples,
+  int32_t RecordedDataIsAvailable(
+      const void* audioSamples,
       size_t nSamples,
       size_t nBytesPerSample,
       size_t nChannels,
@@ -62,7 +63,7 @@ class AudioTransportImpl : public AudioTransport {
       uint32_t currentMicLevel,
       bool keyPressed,
       uint32_t& newMicLevel,
-                                  int64_t estimated_capture_time_ns) override;
+      absl::optional<int64_t> estimated_capture_time_ns) override;
 
   int32_t NeedMorePlayData(size_t nSamples,
                            size_t nBytesPerSample,

modules/audio_device/audio_device_buffer.cc

@@ -55,7 +55,6 @@ AudioDeviceBuffer::AudioDeviceBuffer(TaskQueueFactory* task_queue_factory)
       typing_status_(false),
       play_delay_ms_(0),
       rec_delay_ms_(0),
-      capture_timestamp_ns_(0),
       num_stat_reports_(0),
       last_timer_task_time_(0),
       rec_stat_count_(0),
@@ -231,12 +230,13 @@ void AudioDeviceBuffer::SetVQEData(int play_delay_ms, int rec_delay_ms) {
 
 int32_t AudioDeviceBuffer::SetRecordedBuffer(const void* audio_buffer,
                                              size_t samples_per_channel) {
-  return SetRecordedBuffer(audio_buffer, samples_per_channel, 0);
+  return SetRecordedBuffer(audio_buffer, samples_per_channel, absl::nullopt);
 }
 
-int32_t AudioDeviceBuffer::SetRecordedBuffer(const void* audio_buffer,
+int32_t AudioDeviceBuffer::SetRecordedBuffer(
+    const void* audio_buffer,
     size_t samples_per_channel,
-                                             int64_t capture_timestamp_ns) {
+    absl::optional<int64_t> capture_timestamp_ns) {
   // Copy the complete input buffer to the local buffer.
   const size_t old_size = rec_buffer_.size();
   rec_buffer_.SetData(static_cast<const int16_t*>(audio_buffer),
@@ -247,17 +247,13 @@ int32_t AudioDeviceBuffer::SetRecordedBuffer(const void* audio_buffer,
     RTC_LOG(LS_INFO) << "Size of recording buffer: " << rec_buffer_.size();
   }
 
-  // If the timestamp is less then or equal to zero, it's not valid and are
-  // ignored. If we do antimestamp alignment on them they might accidentally
-  // become greater then zero, and will be handled as if they were a correct
-  // timestamp.
+  if (capture_timestamp_ns) {
     capture_timestamp_ns_ =
-      (capture_timestamp_ns > 0)
-          ? rtc::kNumNanosecsPerMicrosec *
+        rtc::kNumNanosecsPerMicrosec *
         timestamp_aligner_.TranslateTimestamp(
-                    capture_timestamp_ns_ / rtc::kNumNanosecsPerMicrosec,
-                    rtc::TimeMicros())
-          : capture_timestamp_ns;
+            *capture_timestamp_ns / rtc::kNumNanosecsPerMicrosec,
+            rtc::TimeMicros());
+  }
   // Derive a new level value twice per second and check if it is non-zero.
   int16_t max_abs = 0;
   RTC_DCHECK_LT(rec_stat_count_, 50);

modules/audio_device/audio_device_buffer.h

@@ -102,9 +102,10 @@ class AudioDeviceBuffer {
   virtual int32_t SetRecordedBuffer(const void* audio_buffer,
                                     size_t samples_per_channel);
 
-  virtual int32_t SetRecordedBuffer(const void* audio_buffer,
+  virtual int32_t SetRecordedBuffer(
+      const void* audio_buffer,
       size_t samples_per_channel,
-                                    int64_t capture_timestamp_ns);
+      absl::optional<int64_t> capture_timestamp_ns);
   virtual void SetVQEData(int play_delay_ms, int rec_delay_ms);
   virtual int32_t DeliverRecordedData();
   uint32_t NewMicLevel() const;
@@ -194,7 +195,7 @@ class AudioDeviceBuffer {
   int rec_delay_ms_;
 
   // Capture timestamp.
-  int64_t capture_timestamp_ns_;
+  absl::optional<int64_t> capture_timestamp_ns_;
 
   // Counts number of times LogStats() has been called.
   size_t num_stat_reports_ RTC_GUARDED_BY(task_queue_);

modules/audio_device/audio_device_data_observer.cc

@@ -55,14 +55,15 @@ class ADMWrapper : public AudioDeviceModule, public AudioTransport {
                                   uint32_t currentMicLevel,
                                   bool keyPressed,
                                   uint32_t& newMicLevel) override {
-    return RecordedDataIsAvailable(audioSamples, nSamples, nBytesPerSample,
-                                   nChannels, samples_per_sec, total_delay_ms,
-                                   clockDrift, currentMicLevel, keyPressed,
-                                   newMicLevel, /*capture_timestamp_ns*/ 0);
+    return RecordedDataIsAvailable(
+        audioSamples, nSamples, nBytesPerSample, nChannels, samples_per_sec,
+        total_delay_ms, clockDrift, currentMicLevel, keyPressed, newMicLevel,
+        /*capture_timestamp_ns=*/absl::nullopt);
   }
 
   // AudioTransport methods overrides.
-  int32_t RecordedDataIsAvailable(const void* audioSamples,
+  int32_t RecordedDataIsAvailable(
+      const void* audioSamples,
       size_t nSamples,
       size_t nBytesPerSample,
       size_t nChannels,
@@ -72,7 +73,7 @@ class ADMWrapper : public AudioDeviceModule, public AudioTransport {
       uint32_t currentMicLevel,
       bool keyPressed,
       uint32_t& newMicLevel,
-                                  int64_t capture_timestamp_ns) override {
+      absl::optional<int64_t> capture_timestamp_ns) override {
     int32_t res = 0;
     // Capture PCM data of locally captured audio.
     if (observer_) {

modules/audio_device/include/audio_device_defines.h

@@ -56,7 +56,7 @@ class AudioTransport {
       uint32_t currentMicLevel,
       bool keyPressed,
       uint32_t& newMicLevel,
-      int64_t estimatedCaptureTimeNS) {  // NOLINT
+      absl::optional<int64_t> estimatedCaptureTimeNS) {  // NOLINT
     // TODO(webrtc:13620) Make the default behaver of the new API to behave as
     // the old API. This can be pure virtual if all uses of the old API is
     // removed.

modules/audio_device/include/mock_audio_transport.h

@@ -48,7 +48,7 @@ class MockAudioTransport : public AudioTransport {
                uint32_t currentMicLevel,
                bool keyPressed,
                uint32_t& newMicLevel,
-               int64_t estimated_capture_time_ns),
+               absl::optional<int64_t> estimated_capture_time_ns),
               (override));
 
   MOCK_METHOD(int32_t,