Reland "Passing audio capture timestamp to WebRTC."

This is a reland of 00134099

Original change's description:
> Passing audio capture timestamp to WebRTC.
> 
> Bug: 1054403
> Change-Id: I6349687ce55535a0cdc405db67687d967a04fea1
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2066777
> Commit-Queue: Minyue Li <minyue@chromium.org>
> Reviewed-by: Xianzhu Wang <wangxianzhu@chromium.org>
> Reviewed-by: Guido Urdaneta <guidou@chromium.org>
> Reviewed-by: Henrik Andreasson <henrika@chromium.org>
> Reviewed-by: Yuri Wiitala <miu@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#749126}

Bug: 1054403
Change-Id: Icf0af14fd91f2390c933d4b97b8cdbe06dc34ada
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2093446Reviewed-by: Xianzhu Wang <wangxianzhu@chromium.org>
Commit-Queue: Minyue Li <minyue@chromium.org>
Cr-Commit-Position: refs/heads/master@{#749332}

third_party/blink/renderer/platform/BUILD.gn

@@ -1886,6 +1886,7 @@ jumbo_source_set("blink_platform_unittests_sources") {
     "peerconnection/task_queue_factory_test.cc",
     "peerconnection/transmission_encoding_info_handler_test.cc",
     "peerconnection/two_keys_adapter_map_unittest.cc",
+    "peerconnection/webrtc_audio_sink_test.cc",
     "peerconnection/webrtc_video_track_source_test.cc",
     "text/bidi_resolver_test.cc",
     "text/bidi_test_harness.h",

third_party/blink/renderer/platform/peerconnection/webrtc_audio_sink.cc

@@ -12,6 +12,7 @@
 #include "base/logging.h"
 #include "base/stl_util.h"
 #include "base/strings/stringprintf.h"
+#include "media/base/audio_timestamp_helper.h"
 #include "third_party/blink/public/platform/modules/webrtc/webrtc_logging.h"
 #include "third_party/blink/renderer/platform/scheduler/public/post_cross_thread_task.h"
 #include "third_party/blink/renderer/platform/wtf/cross_thread_functional.h"
@@ -98,6 +99,16 @@ void WebRtcAudioSink::OnData(const media::AudioBus& audio_bus,
                              base::TimeTicks estimated_capture_time) {
   // No thread check: OnData might be called on different threads (but not
   // concurrently).
+
+  // TODO(crbug.com/1054769): Better to let |fifo_| handle the estimated capture
+  // time and let it return a corrected interpolated capture time to
+  // DeliverRebufferedAudio(). Current, similar treatment is used at different
+  // places where |AudioPushFifo| is applied. So a update to |AudioPushFifo|
+  // will be a joint effort, and should be carefully carried out.
+  last_estimated_capture_time_ = estimated_capture_time;
+
+  adapter_->UpdateTimestampAligner(estimated_capture_time);
+
   // The following will result in zero, one, or multiple synchronous calls to
   // DeliverRebufferedAudio().
   fifo_.Push(audio_bus);
@@ -130,9 +141,14 @@ void WebRtcAudioSink::DeliverRebufferedAudio(const media::AudioBus& audio_bus,
                 "ToInterleaved expects 2 bytes.");
   audio_bus.ToInterleaved<media::SignedInt16SampleTypeTraits>(
       audio_bus.frames(), interleaved_data_.get());
+
+  const base::TimeTicks estimated_capture_time =
+      last_estimated_capture_time_ + media::AudioTimestampHelper::FramesToTime(
+                                         frame_delay, params_.sample_rate());
+
   adapter_->DeliverPCMToWebRtcSinks(interleaved_data_.get(),
                                     params_.sample_rate(), audio_bus.channels(),
-                                    audio_bus.frames());
+                                    audio_bus.frames(), estimated_capture_time);
 }
 
 namespace {
@@ -171,11 +187,20 @@ void WebRtcAudioSink::Adapter::DeliverPCMToWebRtcSinks(
     const int16_t* audio_data,
     int sample_rate,
     size_t number_of_channels,
-    size_t number_of_frames) {
+    size_t number_of_frames,
+    base::TimeTicks estimated_capture_time) {
   base::AutoLock auto_lock(lock_);
+
+  // This use |timestamp_aligner_| to transform |estimated_capture_timestamp| to
+  // rtc::TimeMicros(). See the comment at UpdateTimestampAligner() for more
+  // details.
+  const int64_t capture_timestamp_us = timestamp_aligner_.TranslateTimestamp(
+      estimated_capture_time.since_origin().InMicroseconds());
+
   for (webrtc::AudioTrackSinkInterface* sink : sinks_) {
     sink->OnData(audio_data, sizeof(int16_t) * 8, sample_rate,
-                 number_of_channels, number_of_frames);
+                 number_of_channels, number_of_frames,
+                 capture_timestamp_us / rtc::kNumMicrosecsPerMillisec);
   }
 }
 
@@ -249,4 +274,16 @@ webrtc::AudioSourceInterface* WebRtcAudioSink::Adapter::GetSource() const {
   return source_.get();
 }
 
+void WebRtcAudioSink::Adapter::UpdateTimestampAligner(
+    base::TimeTicks capture_time) {
+  // The |timestamp_aligner_| stamps an audio frame as if it is captured 'now',
+  // taking rtc::TimeMicros as the reference clock. It does not provide the time
+  // that the frame was originally captured, Using |timestamp_aligner_| rather
+  // than calling rtc::TimeMicros is to take the advantage that it aligns its
+  // output timestamps such that the time spacing in the |capture_time| is
+  // maintained.
+  timestamp_aligner_.TranslateTimestamp(
+      capture_time.since_origin().InMicroseconds(), rtc::TimeMicros());
+}
+
 }  // namespace blink

third_party/blink/renderer/platform/peerconnection/webrtc_audio_sink.h

@@ -24,6 +24,8 @@
 #include "third_party/blink/renderer/platform/platform_export.h"
 #include "third_party/webrtc/api/media_stream_interface.h"
 #include "third_party/webrtc/pc/media_stream_track.h"
+#include "third_party/webrtc/rtc_base/time_utils.h"
+#include "third_party/webrtc/rtc_base/timestamp_aligner.h"
 
 namespace blink {
 
@@ -100,7 +102,8 @@ class PLATFORM_EXPORT WebRtcAudioSink : public WebMediaStreamAudioSink {
     void DeliverPCMToWebRtcSinks(const int16_t* audio_data,
                                  int sample_rate,
                                  size_t number_of_channels,
-                                 size_t number_of_frames);
+                                 size_t number_of_frames,
+                                 base::TimeTicks estimated_capture_time);
 
     std::string label() const { return label_; }
 
@@ -116,6 +119,8 @@ class PLATFORM_EXPORT WebRtcAudioSink : public WebMediaStreamAudioSink {
         override;
     webrtc::AudioSourceInterface* GetSource() const override;
 
+    void UpdateTimestampAligner(base::TimeTicks capture_time);
+
    protected:
     ~Adapter() override;
 
@@ -151,6 +156,11 @@ class PLATFORM_EXPORT WebRtcAudioSink : public WebMediaStreamAudioSink {
     // receive the audio data.
     std::vector<webrtc::AudioTrackSinkInterface*> sinks_;
 
+    // Used for getting capture timestamps referenced on the rtc::TimeMicros()
+    // clock. See the comment at the implementation of UpdateTimestampAligner()
+    // for more details.
+    rtc::TimestampAligner timestamp_aligner_;
+
     DISALLOW_COPY_AND_ASSIGN(Adapter);
   };
 
@@ -182,6 +192,8 @@ class PLATFORM_EXPORT WebRtcAudioSink : public WebMediaStreamAudioSink {
   // interleaved samples.
   std::unique_ptr<int16_t[]> interleaved_data_;
 
+  base::TimeTicks last_estimated_capture_time_;
+
   // In debug builds, check that WebRtcAudioSink's public methods are all being
   // called on the main render thread.
   THREAD_CHECKER(thread_checker_);

third_party/blink/renderer/platform/peerconnection/webrtc_audio_sink_test.cc

+// Copyright 2020 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "third_party/blink/renderer/platform/peerconnection/webrtc_audio_sink.h"
+
+#include "media/base/fake_single_thread_task_runner.h"
+#include "testing/gmock/include/gmock/gmock.h"
+#include "testing/gtest/include/gtest/gtest.h"
+
+using testing::_;
+
+namespace blink {
+
+namespace {
+
+class MockAudioSink : public webrtc::AudioTrackSinkInterface {
+ public:
+  MockAudioSink() = default;
+  ~MockAudioSink() override = default;
+  MOCK_METHOD6(OnData,
+               void(const void* audio_data,
+                    int bits_per_sample,
+                    int sample_rate,
+                    size_t number_of_channels,
+                    size_t number_of_samples,
+                    absl::optional<int64_t> absolute_capture_timestamp_ms));
+};
+
+class ScopedFakeClock : public rtc::ClockInterface {
+ public:
+  explicit ScopedFakeClock(int64_t init_time_ms)
+      : prev_clock_(rtc::SetClockForTesting(this)),
+        time_ns_(init_time_ms * rtc::kNumNanosecsPerMillisec) {}
+
+  ~ScopedFakeClock() override { rtc::SetClockForTesting(prev_clock_); }
+
+  int64_t TimeNanos() const override { return time_ns_; }
+
+  void AdvanceTimeMilliseconds(int64_t time_ms) {
+    time_ns_ += time_ms * rtc::kNumNanosecsPerMillisec;
+  }
+
+ private:
+  ClockInterface* const prev_clock_;
+  int64_t time_ns_;
+};
+
+}  // namespace
+
+TEST(WebRtcAudioSinkTest, CaptureTimestamp) {
+  MockAudioSink sink_1;
+  MockAudioSink sink_2;
+  base::SimpleTestTickClock dummy_clock;
+  std::unique_ptr<WebRtcAudioSink> webrtc_audio_sink(
+      new WebRtcAudioSink("test_sink", nullptr,
+                          /*signaling_task_runner=*/
+                          new media::FakeSingleThreadTaskRunner(&dummy_clock),
+                          /*main_task_runner=*/
+                          new media::FakeSingleThreadTaskRunner(&dummy_clock)));
+
+  // |web_media_stream_audio_sink| is to access methods that are privately
+  // inherited by WebRtcAudioSink.
+  WebMediaStreamAudioSink* const web_media_stream_audio_sink =
+      static_cast<WebMediaStreamAudioSink*>(webrtc_audio_sink.get());
+
+  webrtc_audio_sink->webrtc_audio_track()->AddSink(&sink_1);
+  webrtc_audio_sink->webrtc_audio_track()->AddSink(&sink_2);
+
+  constexpr int kInputChannels = 2;
+  constexpr int kInputFramesPerBuffer = 96;
+  constexpr int kSampleRateHz = 8000;
+  constexpr int kOutputFramesPerBuffer = kSampleRateHz / 100;
+  constexpr int kEnqueueFrames = kInputFramesPerBuffer - kOutputFramesPerBuffer;
+
+  constexpr int64_t kStartRtcTimestampMs = 87654321;
+  constexpr int64_t kStartCaptureTimestampMs = 12345678;
+  constexpr int64_t kCaptureIntervalMs = 567;
+
+  web_media_stream_audio_sink->OnSetFormat(media::AudioParameters(
+      media::AudioParameters::AUDIO_PCM_LINEAR, media::CHANNEL_LAYOUT_STEREO,
+      kSampleRateHz, kOutputFramesPerBuffer));
+  std::unique_ptr<media::AudioBus> bus =
+      media::AudioBus::Create(kInputChannels, kInputFramesPerBuffer);
+  bus->Zero();
+
+  {
+    ScopedFakeClock clock(kStartRtcTimestampMs);
+
+    base::TimeTicks capture_time =
+        base::TimeTicks() +
+        base::TimeDelta::FromMilliseconds(kStartCaptureTimestampMs);
+
+    // The first time to the call OnData(), the TimestampAligner should have no
+    // effect work. So expected capture timestamp is from fake_clock.
+    EXPECT_CALL(
+        sink_1,
+        OnData(_, _, kSampleRateHz, kInputChannels, kOutputFramesPerBuffer,
+               absl::make_optional<int64_t>(kStartRtcTimestampMs)));
+    EXPECT_CALL(
+        sink_2,
+        OnData(_, _, kSampleRateHz, kInputChannels, kOutputFramesPerBuffer,
+               absl::make_optional<int64_t>(kStartRtcTimestampMs)));
+
+    web_media_stream_audio_sink->OnData(*bus, capture_time);
+
+    capture_time += base::TimeDelta::FromMilliseconds(kCaptureIntervalMs);
+    clock.AdvanceTimeMilliseconds(kCaptureIntervalMs);
+
+    constexpr int64_t kExpectedTimestampMs =
+        kStartRtcTimestampMs + kCaptureIntervalMs -
+        kEnqueueFrames * 1000 / kSampleRateHz;
+    EXPECT_CALL(
+        sink_1,
+        OnData(_, _, kSampleRateHz, kInputChannels, kOutputFramesPerBuffer,
+               absl::make_optional<int64_t>(kExpectedTimestampMs)));
+    EXPECT_CALL(
+        sink_2,
+        OnData(_, _, kSampleRateHz, kInputChannels, kOutputFramesPerBuffer,
+               absl::make_optional<int64_t>(kExpectedTimestampMs)));
+
+    web_media_stream_audio_sink->OnData(*bus, capture_time);
+  }
+}
+
+}  // namespace blink