Support Espeak tts on linux Chrome OS Chrome builds

This change enables Chrome OS Chrome builds running on non-cros linux to
work with Espeak-ng.

On linux builds, dcheck always triggers an error in mojo.
Consequently, in TtsService, where the service makes mojo calls on a
Remote within the audio thread, a DCHECK was hit:

2021-01-11T22:46:32.666960Z FATAL chrome[2743839:2743847]: [sequence_checker.h(120)] Check failed: checker.CalledOnValidSequence().
#0 0x7f7ee68c681f base::debug::CollectStackTrace()
#1 0x7f7ee665fd8a base::debug::StackTrace::StackTrace()
#2 0x7f7ee665fd45 base::debug::StackTrace::StackTrace()
#3 0x7f7ee66a9057 logging::LogMessage::~LogMessage()
#4 0x7f7ee66a9769 logging::LogMessage::~LogMessage()
#5 0x7f7ee6620c1b logging::CheckError::~CheckError()
#6 0x7f7ee564e09c base::ScopedValidateSequenceChecker::ScopedValidateSequenceChecker()
#7 0x7f7ee566c068 mojo::InterfaceEndpointClient::SendMessage()
#8 0x7f7ee566c99f mojo::InterfaceEndpointClient::Accept()
#9 0x565159b53c29 chromeos::tts::mojom::TtsEventObserverProxy::OnStart()
#10 0x56515c4e3a5a chromeos::tts::TtsService::Render()
#11 0x7f7ed8648f02 media::AudioOutputDeviceThreadCallback::Process()
#12 0x7f7ed861fe9d media::AudioDeviceThread::ThreadMain()
#13 0x7f7ee68fb1ab base::(anonymous namespace)::ThreadFunc()
#14 0x7f7ea303fea7 start_thread
#15 0x7f7ea2baad8f clone

Fix this by queueing up already rendered audio buffers and processing
them on the main thread's task runner.

Test: manually on a linux chromeos build. Verify that web speech produces speech through Espeak.

Change-Id: I4b1e5b0a6e72e4e571b5a26fde731acc57936894
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2621983Reviewed-by: Katie Dektar <katie@chromium.org>
Commit-Queue: David Tseng <dtseng@chromium.org>
Cr-Commit-Position: refs/heads/master@{#842622}

chrome/browser/speech/extension_api/tts_engine_extension_api.cc

@@ -10,6 +10,7 @@
 
 #include "base/json/json_writer.h"
 #include "base/strings/utf_string_conversions.h"
+#include "base/system/sys_info.h"
 #include "base/values.h"
 #include "build/build_config.h"
 #include "build/chromeos_buildflags.h"
@@ -366,7 +367,12 @@ bool TtsExtensionEngine::IsBuiltInTtsEngineInitialized(
     saw_espeak |=
         voice.engine_id == extension_misc::kEspeakSpeechSynthesisExtensionId;
   }
-  return saw_google_tts && saw_espeak;
+
+  // When running on a real Chrome OS environment, require both Google tts and
+  // Espeak to be initialized; otherwise, only check for Espeak (i.e. on a
+  // non-Chrome OS linux system running the CHrome OS variant of Chrome).
+  return base::SysInfo::IsRunningOnChromeOS() ? (saw_google_tts && saw_espeak)
+                                              : saw_espeak;
 #else
   // Vacuously; no built in engines on other platforms yet. TODO: network tts?
   return true;

chromeos/services/tts/tts_service.cc

@@ -22,7 +22,9 @@ constexpr int kDefaultBufferSize = 512;
 }  // namespace
 
 TtsService::TtsService(mojo::PendingReceiver<mojom::TtsService> receiver)
-    : service_receiver_(this, std::move(receiver)), tts_stream_factory_(this) {
+    : service_receiver_(this, std::move(receiver)),
+      tts_stream_factory_(this),
+      task_runner_(base::ThreadTaskRunnerHandle::Get()) {
   if (setpriority(PRIO_PROCESS, 0, -10 /* real time audio */) != 0) {
     PLOG(ERROR) << "Unable to request real time priority; performance will be "
                    "impacted.";
@@ -122,7 +124,6 @@ int TtsService::Render(base::TimeDelta delay,
                        int prior_frames_skipped,
                        media::AudioBus* dest) {
   size_t frames_in_buf = 0;
-  int32_t status = -1;
   {
     base::AutoLock al(state_lock_);
     if (buffers_.empty())
@@ -130,34 +131,6 @@ int TtsService::Render(base::TimeDelta delay,
 
     const AudioBuffer& buf = buffers_.front();
 
-    status = buf.status;
-    // Done, 0, or error, -1.
-    if (status <= 0) {
-      if (status == -1)
-        tts_event_observer_->OnError();
-      else
-        tts_event_observer_->OnEnd();
-
-      StopLocked();
-      return 0;
-    }
-
-    if (buf.is_first_buffer) {
-      start_playback_time_ = base::Time::Now();
-      tts_event_observer_->OnStart();
-    }
-
-    // Implicit timepoint.
-    if (buf.char_index != -1)
-      tts_event_observer_->OnTimepoint(buf.char_index);
-
-    // Explicit timepoint(s).
-    base::TimeDelta start_to_now = base::Time::Now() - start_playback_time_;
-    while (!timepoints_.empty() && timepoints_.front().second <= start_to_now) {
-      tts_event_observer_->OnTimepoint(timepoints_.front().first);
-      timepoints_.pop();
-    }
-
     frames_in_buf = buf.frames.size();
     const float* frames = nullptr;
     if (!buf.frames.empty())
@@ -165,7 +138,16 @@ int TtsService::Render(base::TimeDelta delay,
     float* channel = dest->channel(0);
     for (size_t i = 0; i < frames_in_buf; i++)
       channel[i] = frames[i];
+
+    rendered_buffers_.push(std::move(buffers_.front()));
     buffers_.pop();
+
+    if (!process_rendered_buffers_posted_) {
+      process_rendered_buffers_posted_ = true;
+      task_runner_->PostTask(FROM_HERE,
+                             base::BindOnce(&TtsService::ProcessRenderedBuffers,
+                                            weak_factory_.GetWeakPtr()));
+    }
   }
 
   return frames_in_buf;
@@ -175,6 +157,7 @@ void TtsService::OnRenderError() {}
 
 void TtsService::StopLocked(bool clear_buffers) {
   output_device_->Pause();
+  rendered_buffers_ = std::queue<AudioBuffer>();
   if (clear_buffers) {
     buffers_ = std::queue<AudioBuffer>();
     timepoints_ = std::queue<Timepoint>();
@@ -191,6 +174,41 @@ void TtsService::ProcessPendingTtsStreamFactories() {
   tts_stream_factory_.Bind(std::move(factory));
 }
 
+void TtsService::ProcessRenderedBuffers() {
+  base::AutoLock al(state_lock_);
+  process_rendered_buffers_posted_ = false;
+  for (; !rendered_buffers_.empty(); rendered_buffers_.pop()) {
+    const auto& buf = rendered_buffers_.front();
+    int status = buf.status;
+    // Done, 0, or error, -1.
+    if (status <= 0) {
+      if (status == -1)
+        tts_event_observer_->OnError();
+      else
+        tts_event_observer_->OnEnd();
+
+      StopLocked();
+      return;
+    }
+
+    if (buf.is_first_buffer) {
+      start_playback_time_ = base::Time::Now();
+      tts_event_observer_->OnStart();
+    }
+
+    // Implicit timepoint.
+    if (buf.char_index != -1)
+      tts_event_observer_->OnTimepoint(buf.char_index);
+  }
+
+  // Explicit timepoint(s).
+  base::TimeDelta start_to_now = base::Time::Now() - start_playback_time_;
+  while (!timepoints_.empty() && timepoints_.front().second <= start_to_now) {
+    tts_event_observer_->OnTimepoint(timepoints_.front().first);
+    timepoints_.pop();
+  }
+}
+
 TtsService::AudioBuffer::AudioBuffer() = default;
 
 TtsService::AudioBuffer::~AudioBuffer() = default;

chromeos/services/tts/tts_service.h

@@ -92,6 +92,10 @@ class TtsService : public mojom::TtsService,
   // Satisfies any pending tts stream factory receivers.
   void ProcessPendingTtsStreamFactories();
 
+  // Do any processing (e.g. sending start/end events) on buffers that have just
+  // been rendered on the audio thread.
+  void ProcessRenderedBuffers();
+
   // Connection to tts in the browser.
   mojo::Receiver<mojom::TtsService> service_receiver_;
 
@@ -117,6 +121,7 @@ class TtsService : public mojom::TtsService,
 
   // The queue of audio buffers to be played by the audio thread.
   std::queue<AudioBuffer> buffers_ GUARDED_BY(state_lock_);
+  std::queue<AudioBuffer> rendered_buffers_;
 
   // An explicit list of increasing time delta sorted timepoints to be fired
   // while rendering audio at the specified |delay| from start of audio
@@ -126,6 +131,14 @@ class TtsService : public mojom::TtsService,
 
   // The time at which playback of the current utterance started.
   base::Time start_playback_time_;
+
+  // Whether a task to process rendered audio buffers has been posted.
+  bool process_rendered_buffers_posted_ GUARDED_BY(state_lock_) = false;
+
+  // The main thread's task runner handle.
+  scoped_refptr<base::SingleThreadTaskRunner> task_runner_;
+
+  base::WeakPtrFactory<TtsService> weak_factory_{this};
 };
 
 }  // namespace tts