Switch Speech Recognition to asynchronous callback-based AudioManager interactions.

content/browser/speech/speech_recognizer_impl_unittest.cc

 // Copyright (c) 2013 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 <stddef.h>
 #include <stdint.h>
 
 #include <vector>
 
 #include "base/run_loop.h"
+#include "base/synchronization/waitable_event.h"
 #include "base/sys_byteorder.h"
+#include "base/threading/thread.h"
 #include "base/threading/thread_task_runner_handle.h"
 #include "content/browser/speech/proto/google_streaming_api.pb.h"
 #include "content/browser/speech/speech_recognition_engine.h"
 #include "content/browser/speech/speech_recognizer_impl.h"
 #include "content/public/browser/speech_recognition_event_listener.h"
 #include "content/public/test/test_browser_thread_bundle.h"
 #include "media/audio/audio_device_description.h"
+#include "media/audio/audio_system_impl.h"
 #include "media/audio/fake_audio_input_stream.h"
 #include "media/audio/fake_audio_output_stream.h"
 #include "media/audio/mock_audio_manager.h"
 #include "media/audio/test_audio_input_controller_factory.h"
 #include "media/base/audio_bus.h"
+#include "media/base/test_helpers.h"
 #include "net/base/net_errors.h"
 #include "net/url_request/test_url_fetcher_factory.h"
 #include "net/url_request/url_request_status.h"
 #include "testing/gtest/include/gtest/gtest.h"
 
 using media::AudioInputController;
 using media::AudioInputStream;
 using media::AudioOutputStream;
 using media::AudioParameters;
 using media::TestAudioInputController;
 using media::TestAudioInputControllerFactory;
 
 namespace content {
 
 class SpeechRecognizerImplTest : public SpeechRecognitionEventListener,
                                  public testing::Test {
  public:
   SpeechRecognizerImplTest()
-      : recognition_started_(false),
+      : audio_thread_("AudioThread"),

[tommi (sloooow) - chröme, 2017/02/05 20:14:54]

nit: SpeechAudioThread?  Nice improvement to really have a separate thread btw.

[o1ka, 2017/02/06 12:06:08]

Done.

+        recognition_started_(false),
         recognition_ended_(false),
         result_received_(false),
         audio_started_(false),
         audio_ended_(false),
         sound_started_(false),
         sound_ended_(false),
         error_(SPEECH_RECOGNITION_ERROR_NONE),
         volume_(-1.0f) {
     // SpeechRecognizer takes ownership of sr_engine.
     SpeechRecognitionEngine* sr_engine =
         new SpeechRecognitionEngine(NULL /* URLRequestContextGetter */);
     SpeechRecognitionEngine::Config config;
     config.audio_num_bits_per_sample =
         SpeechRecognizerImpl::kNumBitsPerAudioSample;
     config.audio_sample_rate = SpeechRecognizerImpl::kAudioSampleRate;
     config.filter_profanities = false;
     sr_engine->SetConfig(config);
 
     const int kTestingSessionId = 1;
+
+    audio_thread_.StartAndWaitForTesting();
+    audio_manager_.reset(
+        new media::MockAudioManager(audio_thread_.task_runner()));
+    audio_manager_->SetInputStreamParameters(
+        media::AudioParameters::UnavailableDeviceParams());
+    audio_system_ = media::AudioSystemImpl::Create(audio_manager_.get());
     recognizer_ = new SpeechRecognizerImpl(
-        this, kTestingSessionId, false, false, sr_engine);
-    audio_manager_.reset(
-        new media::MockAudioManager(base::ThreadTaskRunnerHandle::Get().get()));
-    recognizer_->SetAudioManagerForTesting(audio_manager_.get());
+        this, audio_system_.get(), kTestingSessionId, false, false, sr_engine);
 
     int audio_packet_length_bytes =
         (SpeechRecognizerImpl::kAudioSampleRate *
          SpeechRecognitionEngine::kAudioPacketIntervalMs *
          ChannelLayoutToChannelCount(SpeechRecognizerImpl::kChannelLayout) *
          SpeechRecognizerImpl::kNumBitsPerAudioSample) / (8 * 1000);
     audio_packet_.resize(audio_packet_length_bytes);
 
     const int channels =
         ChannelLayoutToChannelCount(SpeechRecognizerImpl::kChannelLayout);
     bytes_per_sample_ = SpeechRecognizerImpl::kNumBitsPerAudioSample / 8;
     const int frames = audio_packet_length_bytes / channels / bytes_per_sample_;
     audio_bus_ = media::AudioBus::Create(channels, frames);
     audio_bus_->Zero();
   }
 
+  ~SpeechRecognizerImplTest() override {
+    // Deleting |audio_manager_| on audio thread.
+    audio_manager_.reset();
+    audio_thread_.Stop();
+  }
+
   void CheckEventsConsistency() {
     // Note: "!x || y" == "x implies y".
     EXPECT_TRUE(!recognition_ended_ || recognition_started_);
     EXPECT_TRUE(!audio_ended_ || audio_started_);
     EXPECT_TRUE(!sound_ended_ || sound_started_);
     EXPECT_TRUE(!audio_started_ || recognition_started_);
     EXPECT_TRUE(!sound_started_ || audio_started_);
     EXPECT_TRUE(!audio_ended_ || (sound_ended_ || !sound_started_));
     EXPECT_TRUE(!recognition_ended_ || (audio_ended_ || !audio_started_));
   }
@@ skipping 89 matching lines @@
     }
     CopyPacketToAudioBus();
   }
 
   void OnData(media::AudioBus* data) {
     auto* writer =
         static_cast<AudioInputController::SyncWriter*>(recognizer_.get());
     writer->Write(data, 0.0, false, 0);
   }
 
+  void WaitForAudioThreadToPostDeviceInfo() {
+    media::WaitableMessageLoopEvent event;
+    audio_thread_.task_runner()->PostTaskAndReply(
+        FROM_HERE, base::Bind(&base::DoNothing), event.GetClosure());
+    // Runs the loop and waits for the |audio_thread_| to call event's closure,
+    // which means AudioSystem reply containing device parameters is already
+    // queued on the main thread.
+    event.RunAndWait();
+  }
+
  protected:
   TestBrowserThreadBundle thread_bundle_;
   scoped_refptr<SpeechRecognizerImpl> recognizer_;
-  media::ScopedAudioManagerPtr audio_manager_;
+  base::Thread audio_thread_;
+  media::MockAudioManager::UniquePtr audio_manager_;
+  std::unique_ptr<media::AudioSystem> audio_system_;
   bool recognition_started_;
   bool recognition_ended_;
   bool result_received_;
   bool audio_started_;
   bool audio_ended_;
   bool sound_started_;
   bool sound_ended_;
   SpeechRecognitionErrorCode error_;
   net::TestURLFetcherFactory url_fetcher_factory_;
   TestAudioInputControllerFactory audio_input_controller_factory_;
   std::vector<uint8_t> audio_packet_;
   std::unique_ptr<media::AudioBus> audio_bus_;
   int bytes_per_sample_;
   float volume_;
   float noise_volume_;
 };
 
+TEST_F(SpeechRecognizerImplTest, StartNoInputDevices) {
+  // Check for callbacks when stopping record before any audio gets recorded.
+  audio_manager_->SetHasInputDevices(false);
+  recognizer_->StartRecognition(
+      media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
+  // recognizer_->StopAudioCapture();

[tommi (sloooow) - chröme, 2017/02/05 20:14:54]

remove?

[o1ka, 2017/02/06 12:06:08]

Done.

+  base::RunLoop().RunUntilIdle();
+  EXPECT_TRUE(recognition_started_);
+  EXPECT_FALSE(audio_started_);
+  EXPECT_FALSE(result_received_);
+  EXPECT_EQ(SPEECH_RECOGNITION_ERROR_AUDIO_CAPTURE, error_);
+  CheckFinalEventsConsistency();
+}
+
+TEST_F(SpeechRecognizerImplTest, StopBeforeDeviceInfoReceived) {
+  // Check for callbacks when stopping record before reply is received from
+  // AudioSystem.
+  base::WaitableEvent event(base::WaitableEvent::ResetPolicy::AUTOMATIC,
+                            base::WaitableEvent::InitialState::NOT_SIGNALED);
+
+  // Block audio thread.
+  audio_thread_.task_runner()->PostTask(
+      FROM_HERE,
+      base::Bind(&base::WaitableEvent::Wait, base::Unretained(&event)));
+
+  recognizer_->StartRecognition(
+      media::AudioDeviceDescription::kDefaultDeviceId);
+  recognizer_->StopAudioCapture();
+  base::RunLoop().RunUntilIdle();
+
+  // Release audio thread and receive a callback from it.
+  event.Signal();
+  WaitForAudioThreadToPostDeviceInfo();
+  base::RunLoop().RunUntilIdle();
+
+  EXPECT_TRUE(recognition_started_);
+  EXPECT_FALSE(audio_started_);
+  EXPECT_FALSE(result_received_);
+  EXPECT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_);
+  CheckFinalEventsConsistency();
+}
+
+TEST_F(SpeechRecognizerImplTest, CancelBeforeDeviceInfoReceived) {
+  // Check for callbacks when stopping record before reply is received from
+  // AudioSystem.
+  base::WaitableEvent event(base::WaitableEvent::ResetPolicy::AUTOMATIC,
+                            base::WaitableEvent::InitialState::NOT_SIGNALED);
+
+  // Block audio thread.
+  audio_thread_.task_runner()->PostTask(
+      FROM_HERE,
+      base::Bind(&base::WaitableEvent::Wait, base::Unretained(&event)));
+
+  recognizer_->StartRecognition(
+      media::AudioDeviceDescription::kDefaultDeviceId);
+  recognizer_->AbortRecognition();
+  base::RunLoop().RunUntilIdle();
+
+  // Release audio thread and receive a callback from it.
+  event.Signal();
+  WaitForAudioThreadToPostDeviceInfo();
+  base::RunLoop().RunUntilIdle();
+
+  EXPECT_TRUE(recognition_started_);
+  EXPECT_FALSE(audio_started_);
+  EXPECT_FALSE(result_received_);
+  EXPECT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_);
+  CheckFinalEventsConsistency();
+}
+
 TEST_F(SpeechRecognizerImplTest, StopNoData) {
   // Check for callbacks when stopping record before any audio gets recorded.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   recognizer_->StopAudioCapture();
   base::RunLoop().RunUntilIdle();
   EXPECT_TRUE(recognition_started_);
   EXPECT_FALSE(audio_started_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, CancelNoData) {
   // Check for callbacks when canceling recognition before any audio gets
   // recorded.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   recognizer_->AbortRecognition();
   base::RunLoop().RunUntilIdle();
   EXPECT_TRUE(recognition_started_);
   EXPECT_FALSE(audio_started_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_ABORTED, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, StopWithData) {
   // Start recording, give some data and then stop. This should wait for the
   // network callback to arrive before completion.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
 
   // Try sending 5 chunks of mock audio data and verify that each of them
   // resulted immediately in a packet sent out via the network. This verifies
   // that we are streaming out encoded data as chunks without waiting for the
   // full recording to complete.
   const size_t kNumChunks = 5;
@@ skipping 42 matching lines @@
   EXPECT_TRUE(recognition_ended_);
   EXPECT_TRUE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, CancelWithData) {
   // Start recording, give some data and then cancel.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   OnData(audio_bus_.get());
   base::RunLoop().RunUntilIdle();
   recognizer_->AbortRecognition();
   base::RunLoop().RunUntilIdle();
   ASSERT_TRUE(url_fetcher_factory_.GetFetcherByID(0));
   EXPECT_TRUE(recognition_started_);
   EXPECT_TRUE(audio_started_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_ABORTED, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, ConnectionError) {
   // Start recording, give some data and then stop. Issue the network callback
   // with a connection error and verify that the recognizer bubbles the error up
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   OnData(audio_bus_.get());
   base::RunLoop().RunUntilIdle();
   net::TestURLFetcher* fetcher = url_fetcher_factory_.GetFetcherByID(0);
   ASSERT_TRUE(fetcher);
 
   recognizer_->StopAudioCapture();
@@ skipping 16 matching lines @@
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_NETWORK, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, ServerError) {
   // Start recording, give some data and then stop. Issue the network callback
   // with a 500 error and verify that the recognizer bubbles the error up
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   OnData(audio_bus_.get());
   base::RunLoop().RunUntilIdle();
   net::TestURLFetcher* fetcher = url_fetcher_factory_.GetFetcherByID(0);
   ASSERT_TRUE(fetcher);
 
   recognizer_->StopAudioCapture();
@@ skipping 14 matching lines @@
   EXPECT_TRUE(recognition_ended_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_NETWORK, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, AudioControllerErrorNoData) {
   // Check if things tear down properly if AudioInputController threw an error.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller->event_handler()->OnError(controller,
       AudioInputController::UNKNOWN_ERROR);
   base::RunLoop().RunUntilIdle();
   EXPECT_TRUE(recognition_started_);
   EXPECT_FALSE(audio_started_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_AUDIO_CAPTURE, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, AudioControllerErrorWithData) {
   // Check if things tear down properly if AudioInputController threw an error
   // after giving some audio data.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   OnData(audio_bus_.get());
   controller->event_handler()->OnError(controller,
       AudioInputController::UNKNOWN_ERROR);
   base::RunLoop().RunUntilIdle();
   ASSERT_TRUE(url_fetcher_factory_.GetFetcherByID(0));
   EXPECT_TRUE(recognition_started_);
   EXPECT_TRUE(audio_started_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_AUDIO_CAPTURE, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, NoSpeechCallbackIssued) {
   // Start recording and give a lot of packets with audio samples set to zero.
   // This should trigger the no-speech detector and issue a callback.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
 
   int num_packets = (SpeechRecognizerImpl::kNoSpeechTimeoutMs) /
                      SpeechRecognitionEngine::kAudioPacketIntervalMs + 1;
   // The vector is already filled with zero value samples on create.
   for (int i = 0; i < num_packets; ++i) {
     OnData(audio_bus_.get());
   }
   base::RunLoop().RunUntilIdle();
   EXPECT_TRUE(recognition_started_);
   EXPECT_TRUE(audio_started_);
   EXPECT_FALSE(result_received_);
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_NO_SPEECH, error_);
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, NoSpeechCallbackNotIssued) {
   // Start recording and give a lot of packets with audio samples set to zero
   // and then some more with reasonably loud audio samples. This should be
   // treated as normal speech input and the no-speech detector should not get
   // triggered.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller = audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
 
   int num_packets = (SpeechRecognizerImpl::kNoSpeechTimeoutMs) /
                      SpeechRecognitionEngine::kAudioPacketIntervalMs;
 
@@ skipping 17 matching lines @@
   CheckFinalEventsConsistency();
 }
 
 TEST_F(SpeechRecognizerImplTest, SetInputVolumeCallback) {
   // Start recording and give a lot of packets with audio samples set to zero
   // and then some more with reasonably loud audio samples. Check that we don't
   // get the callback during estimation phase, then get zero for the silence
   // samples and proper volume for the loud audio.
   recognizer_->StartRecognition(
       media::AudioDeviceDescription::kDefaultDeviceId);
+  WaitForAudioThreadToPostDeviceInfo();
   base::RunLoop().RunUntilIdle();
   TestAudioInputController* controller =
       audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
   controller = audio_input_controller_factory_.controller();
   ASSERT_TRUE(controller);
 
   // Feed some samples to begin with for the endpointer to do noise estimation.
   int num_packets = SpeechRecognizerImpl::kEndpointerEstimationTimeMs /
                     SpeechRecognitionEngine::kAudioPacketIntervalMs;
@@ skipping 17 matching lines @@
 
   EXPECT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_);
   EXPECT_FALSE(audio_ended_);
   EXPECT_FALSE(recognition_ended_);
   recognizer_->AbortRecognition();
   base::RunLoop().RunUntilIdle();
   CheckFinalEventsConsistency();
 }
 
 }  // namespace content