Replace MediaStreamUIController with MediaStreamUIProxy.

content/browser/speech/speech_recognition_manager_impl.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 "content/browser/speech/speech_recognition_manager_impl.h"
 
 #include "base/bind.h"
 #include "content/browser/browser_main_loop.h"
 #include "content/browser/renderer_host/media/media_stream_manager.h"
+#include "content/browser/renderer_host/media/media_stream_ui_proxy.h"
 #include "content/browser/speech/google_one_shot_remote_engine.h"
 #include "content/browser/speech/google_streaming_remote_engine.h"
 #include "content/browser/speech/speech_recognition_engine.h"
 #include "content/browser/speech/speech_recognizer_impl.h"
 #include "content/public/browser/browser_thread.h"
 #include "content/public/browser/content_browser_client.h"
 #include "content/public/browser/resource_context.h"
 #include "content/public/browser/speech_recognition_event_listener.h"
 #include "content/public/browser/speech_recognition_manager_delegate.h"
 #include "content/public/browser/speech_recognition_session_config.h"
@@ skipping 41 matching lines @@
       delegate_(GetContentClient()->browser()->
                     GetSpeechRecognitionManagerDelegate()),
       weak_factory_(this) {
   DCHECK(!g_speech_recognition_manager_impl);
   g_speech_recognition_manager_impl = this;
 }
 
 SpeechRecognitionManagerImpl::~SpeechRecognitionManagerImpl() {
   DCHECK(g_speech_recognition_manager_impl);
   g_speech_recognition_manager_impl = NULL;
-  // Recognition sessions will be aborted by the corresponding destructors.
+
+  for (SessionsTable::iterator it = sessions_.begin(); it != sessions_.end();
+       ++it) {
+    // MediaStreamUIProxy must be deleted on the IO thread.
+    BrowserThread::DeleteSoon(BrowserThread::IO, FROM_HERE,
+                              it->second->ui.release());
+    delete it->second;
+  }
   sessions_.clear();
 }
 
 int SpeechRecognitionManagerImpl::CreateSession(
     const SpeechRecognitionSessionConfig& config) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 
   const int session_id = GetNextSessionID();
   DCHECK(!SessionExists(session_id));
   // Set-up the new session.
-  Session& session = sessions_[session_id];
-  session.id = session_id;
-  session.config = config;
-  session.context = config.initial_context;
+  Session* session = new Session();
+  sessions_[session_id] = session;
+  session->id = session_id;
+  session->config = config;
+  session->context = config.initial_context;
 
   std::string hardware_info;
   bool can_report_metrics = false;
   if (delegate_)
     delegate_->GetDiagnosticInformation(&can_report_metrics, &hardware_info);
 
   // The legacy api cannot use continuous mode.
   DCHECK(!config.is_legacy_api || !config.continuous);
 
 #if !defined(OS_ANDROID)
@@ skipping 23 matching lines @@
   if (config.is_legacy_api) {
     google_remote_engine =
         new GoogleOneShotRemoteEngine(config.url_request_context_getter.get());
   } else {
     google_remote_engine = new GoogleStreamingRemoteEngine(
         config.url_request_context_getter.get());
   }
 
   google_remote_engine->SetConfig(remote_engine_config);
 
-  session.recognizer = new SpeechRecognizerImpl(
+  session->recognizer = new SpeechRecognizerImpl(
       this,
       session_id,
       !config.continuous,
       google_remote_engine);
 #else
   // TODO(janx): Implement a SpeechRecognizerImplAndroid with a JNI interface
   // forwarding calls to Android's platform speech recognition service (see
   // crbug.com/222352).
-  session.recognizer = NULL;
+  session->recognizer = NULL;
 #endif
   return session_id;
 }
 
 void SpeechRecognitionManagerImpl::StartSession(int session_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   if (!SessionExists(session_id))
     return;
 
   // If there is another active session, abort that.
@@ skipping 21 matching lines @@
     return;
 
 #if defined(OS_IOS)
   // On iOS, voice search can only be initiated by clear user action and thus
   // it is always allowed.
   DCHECK(!ask_user && is_allowed);
 #else
   if (ask_user) {
     SessionsTable::iterator iter = sessions_.find(session_id);
     DCHECK(iter != sessions_.end());
-    SpeechRecognitionSessionContext& context = iter->second.context;
+    SpeechRecognitionSessionContext& context = iter->second->context;
     context.label =
         BrowserMainLoop::GetMediaStreamManager()->MakeMediaAccessRequest(
             context.render_process_id,
             context.render_view_id,
             StreamOptions(MEDIA_DEVICE_AUDIO_CAPTURE, MEDIA_NO_SERVICE),
             GURL(context.context_name),
             base::Bind(
                 &SpeechRecognitionManagerImpl::MediaRequestPermissionCallback,
-                weak_factory_.GetWeakPtr()));
-
+                weak_factory_.GetWeakPtr(), session_id));
     return;
   }
 #endif  // defined(OS_IOS)
 
   if (is_allowed) {
     base::MessageLoop::current()->PostTask(
         FROM_HERE,
         base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
                    weak_factory_.GetWeakPtr(),
                    session_id,
                    EVENT_START));
   } else {
     OnRecognitionError(session_id, SpeechRecognitionError(
         SPEECH_RECOGNITION_ERROR_NOT_ALLOWED));
     base::MessageLoop::current()->PostTask(
         FROM_HERE,
         base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
                    weak_factory_.GetWeakPtr(),
                    session_id,
                    EVENT_ABORT));
   }
 }
 
 void SpeechRecognitionManagerImpl::MediaRequestPermissionCallback(
-    const std::string& label, const MediaStreamDevices& devices) {
+    int session_id,
+    const MediaStreamDevices& devices,
+    scoped_ptr<MediaStreamUIProxy> stream_ui) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  for (SessionsTable::iterator iter = sessions_.begin();
-       iter != sessions_.end(); ++iter) {
-    if (iter->second.context.label == label) {
-      bool is_allowed = false;
-      if (!devices.empty()) {
-        // Copy the approved devices array to the context for UI indication.
-        iter->second.context.devices = devices;
-        is_allowed = true;
-      }
 
-      // Clear the label to indicate the request has been done.
-      iter->second.context.label.clear();
+  SessionsTable::iterator iter = sessions_.find(session_id);
+  if (iter == sessions_.end())
+    return;
 
-      // Notify the recognition about the request result.
-      RecognitionAllowedCallback(iter->first, false, is_allowed);
-      break;
-    }
+  bool is_allowed = !devices.empty();
+  if (is_allowed) {
+    // Copy the approved devices array to the context for UI indication.
+    iter->second->context.devices = devices;
+
+    // Save the UI object.
+    iter->second->ui = stream_ui.Pass();
   }
+
+  // Clear the label to indicate the request has been done.
+  iter->second->context.label.clear();
+
+  // Notify the recognition about the request result.
+  RecognitionAllowedCallback(iter->first, false, is_allowed);
 }
 
 void SpeechRecognitionManagerImpl::AbortSession(int session_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   if (!SessionExists(session_id))
     return;
 
-#if !defined(OS_IOS)
-  const SpeechRecognitionSessionContext& context =
-      GetSessionContext(session_id);
-  if (!context.label.empty())
-    BrowserMainLoop::GetMediaStreamManager()->CancelRequest(context.label);
-#endif  // !defined(OS_IOS)
+  SessionsTable::iterator iter = sessions_.find(session_id);
+  iter->second->ui.reset();
 
   base::MessageLoop::current()->PostTask(
       FROM_HERE,
       base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
                  weak_factory_.GetWeakPtr(),
                  session_id,
                  EVENT_ABORT));
 }
 
 void SpeechRecognitionManagerImpl::StopAudioCaptureForSession(int session_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   if (!SessionExists(session_id))
     return;
 
-#if !defined(OS_IOS)
-  const SpeechRecognitionSessionContext& context =
-      GetSessionContext(session_id);
-  if (!context.label.empty())
-    BrowserMainLoop::GetMediaStreamManager()->CancelRequest(context.label);
-#endif  // !defined(OS_IOS)
+  SessionsTable::iterator iter = sessions_.find(session_id);
+  iter->second->ui.reset();
 
   base::MessageLoop::current()->PostTask(
       FROM_HERE,
       base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
                  weak_factory_.GetWeakPtr(),
                  session_id,
                  EVENT_STOP_CAPTURE));
 }
 
 // Here begins the SpeechRecognitionEventListener interface implementation,
 // which will simply relay the events to the proper listener registered for the
 // particular session (most likely InputTagSpeechDispatcherHost) and to the
 // catch-all listener provided by the delegate (if any).
 
 void SpeechRecognitionManagerImpl::OnRecognitionStart(int session_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   if (!SessionExists(session_id))
     return;
 
-#if !defined(OS_IOS)
-  const SpeechRecognitionSessionContext& context =
-      GetSessionContext(session_id);
-  if (!context.devices.empty()) {
-    // Notify the UI the devices are being used.
-    BrowserMainLoop::GetMediaStreamManager()->NotifyUIDevicesOpened(
-        context.label);
+  SessionsTable::iterator iter = sessions_.find(session_id);
+  if (iter->second->ui) {
+    // Notify the UI that the devices are being used.
+    iter->second->ui->OnStarted(base::Closure());
   }
-#endif  // !defined(OS_IOS)
 
   DCHECK_EQ(primary_session_id_, session_id);
   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
     delegate_listener->OnRecognitionStart(session_id);
   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
     listener->OnRecognitionStart(session_id);
 }
 
 void SpeechRecognitionManagerImpl::OnAudioStart(int session_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
@@ skipping 93 matching lines @@
   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
     delegate_listener->OnAudioLevelsChange(session_id, volume, noise_volume);
   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
     listener->OnAudioLevelsChange(session_id, volume, noise_volume);
 }
 
 void SpeechRecognitionManagerImpl::OnRecognitionEnd(int session_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   if (!SessionExists(session_id))
     return;
-#if !defined(OS_IOS)
-  const SpeechRecognitionSessionContext& context =
-        GetSessionContext(session_id);
-  if (!context.devices.empty()) {
-    // Notify the UI the devices has been closed.
-     BrowserMainLoop::GetMediaStreamManager()->NotifyUIDevicesClosed(
-         context.label);
-  }
-#endif  // !defined(OS_IOS)
 
   if (SpeechRecognitionEventListener* delegate_listener = GetDelegateListener())
     delegate_listener->OnRecognitionEnd(session_id);
   if (SpeechRecognitionEventListener* listener = GetListener(session_id))
     listener->OnRecognitionEnd(session_id);
   base::MessageLoop::current()->PostTask(
       FROM_HERE,
       base::Bind(&SpeechRecognitionManagerImpl::DispatchEvent,
                  weak_factory_.GetWeakPtr(),
                  session_id,
                  EVENT_RECOGNITION_ENDED));
 }
 
 int SpeechRecognitionManagerImpl::GetSession(
     int render_process_id, int render_view_id, int request_id) const {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   SessionsTable::const_iterator iter;
   for(iter = sessions_.begin(); iter != sessions_.end(); ++iter) {
     const int session_id = iter->first;
-    const SpeechRecognitionSessionContext& context = iter->second.context;
+    const SpeechRecognitionSessionContext& context = iter->second->context;
     if (context.render_process_id == render_process_id &&
         context.render_view_id == render_view_id &&
         context.request_id == request_id) {
       return session_id;
     }
   }
   return kSessionIDInvalid;
 }
 
 SpeechRecognitionSessionContext
 SpeechRecognitionManagerImpl::GetSessionContext(int session_id) const {
-  return GetSession(session_id).context;
+  return GetSession(session_id)->context;
 }
 
 void SpeechRecognitionManagerImpl::AbortAllSessionsForListener(
     SpeechRecognitionEventListener* listener) {
   // This method gracefully destroys sessions for the listener. However, since
   // the listener itself is likely to be destroyed after this call, we avoid
   // dispatching further events to it, marking the |listener_is_active| flag.
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   for (SessionsTable::iterator it = sessions_.begin(); it != sessions_.end();
        ++it) {
-    Session& session = it->second;
-    if (session.config.event_listener == listener) {
-      AbortSession(session.id);
-      session.listener_is_active = false;
+    Session* session = it->second;
+    if (session->config.event_listener == listener) {
+      AbortSession(session->id);
+      session->listener_is_active = false;
     }
   }
 }
 
 void SpeechRecognitionManagerImpl::AbortAllSessionsForRenderView(
     int render_process_id,
     int render_view_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   for (SessionsTable::iterator it = sessions_.begin(); it != sessions_.end();
        ++it) {
-    Session& session = it->second;
-    if (session.context.render_process_id == render_process_id &&
-        session.context.render_view_id == render_view_id) {
-      AbortSession(session.id);
+    Session* session = it->second;
+    if (session->context.render_process_id == render_process_id &&
+        session->context.render_view_id == render_view_id) {
+      AbortSession(session->id);
     }
   }
 }
 
 // -----------------------  Core FSM implementation ---------------------------
 void SpeechRecognitionManagerImpl::DispatchEvent(int session_id,
                                                  FSMEvent event) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 
   // There are some corner cases in which the session might be deleted (due to
   // an EndRecognition event) between a request (e.g. Abort) and its dispatch.
   if (!SessionExists(session_id))
     return;
 
-  const Session& session = GetSession(session_id);
+  Session* session = GetSession(session_id);
   FSMState session_state = GetSessionState(session_id);
   DCHECK_LE(session_state, SESSION_STATE_MAX_VALUE);
   DCHECK_LE(event, EVENT_MAX_VALUE);
 
   // Event dispatching must be sequential, otherwise it will break all the rules
   // and the assumptions of the finite state automata model.
   DCHECK(!is_dispatching_event_);
   is_dispatching_event_ = true;
   ExecuteTransitionAndGetNextState(session, session_state, event);
   is_dispatching_event_ = false;
 }
 
 // This FSM handles the evolution of each session, from the viewpoint of the
 // interaction with the user (that may be either the browser end-user which
 // interacts with UI bubbles, or JS developer intracting with JS methods).
 // All the events received by the SpeechRecognizer instances (one for each
 // session) are always routed to the SpeechRecognitionEventListener(s)
 // regardless the choices taken in this FSM.
 void SpeechRecognitionManagerImpl::ExecuteTransitionAndGetNextState(
-    const Session& session, FSMState session_state, FSMEvent event) {
+    Session* session, FSMState session_state, FSMEvent event) {
   // Note: since we're not tracking the state of the recognizer object, rather
   // we're directly retrieving it (through GetSessionState), we see its events
   // (that are AUDIO_ENDED and RECOGNITION_ENDED) after its state evolution
   // (e.g., when we receive the AUDIO_ENDED event, the recognizer has just
   // completed the transition from CAPTURING_AUDIO to WAITING_FOR_RESULT, thus
   // we perceive the AUDIO_ENDED event in WAITING_FOR_RESULT).
   // This makes the code below a bit tricky but avoids a lot of code for
   // tracking and reconstructing asynchronously the state of the recognizer.
   switch (session_state) {
     case SESSION_STATE_IDLE:
       switch (event) {
         case EVENT_START:
-          return SessionStart(session);
+          return SessionStart(*session);
         case EVENT_ABORT:
-          return SessionAbort(session);
+          return SessionAbort(*session);
         case EVENT_RECOGNITION_ENDED:
           return SessionDelete(session);
         case EVENT_STOP_CAPTURE:
-          return SessionStopAudioCapture(session);
+          return SessionStopAudioCapture(*session);
         case EVENT_AUDIO_ENDED:
           return;
       }
       break;
     case SESSION_STATE_CAPTURING_AUDIO:
       switch (event) {
         case EVENT_STOP_CAPTURE:
-          return SessionStopAudioCapture(session);
+          return SessionStopAudioCapture(*session);
         case EVENT_ABORT:
-          return SessionAbort(session);
+          return SessionAbort(*session);
         case EVENT_START:
           return;
         case EVENT_AUDIO_ENDED:
         case EVENT_RECOGNITION_ENDED:
-          return NotFeasible(session, event);
+          return NotFeasible(*session, event);
       }
       break;
     case SESSION_STATE_WAITING_FOR_RESULT:
       switch (event) {
         case EVENT_ABORT:
-          return SessionAbort(session);
+          return SessionAbort(*session);
         case EVENT_AUDIO_ENDED:
-          return ResetCapturingSessionId(session);
+          return ResetCapturingSessionId(*session);
         case EVENT_START:
         case EVENT_STOP_CAPTURE:
           return;
         case EVENT_RECOGNITION_ENDED:
-          return NotFeasible(session, event);
+          return NotFeasible(*session, event);
       }
       break;
   }
-  return NotFeasible(session, event);
+  return NotFeasible(*session, event);
 }
 
 SpeechRecognitionManagerImpl::FSMState
 SpeechRecognitionManagerImpl::GetSessionState(int session_id) const {
-  const Session& session = GetSession(session_id);
-  if (!session.recognizer.get() || !session.recognizer->IsActive())
+  Session* session = GetSession(session_id);
+  if (!session->recognizer.get() || !session->recognizer->IsActive())
     return SESSION_STATE_IDLE;
-  if (session.recognizer->IsCapturingAudio())
+  if (session->recognizer->IsCapturingAudio())
     return SESSION_STATE_CAPTURING_AUDIO;
   return SESSION_STATE_WAITING_FOR_RESULT;
 }
 
 // ----------- Contract for all the FSM evolution functions below -------------
 //  - Are guaranteed to be executed in the IO thread;
 //  - Are guaranteed to be not reentrant (themselves and each other);
 
 void SpeechRecognitionManagerImpl::SessionStart(const Session& session) {
   DCHECK_EQ(primary_session_id_, session.id);
@@ skipping 12 matching lines @@
   DCHECK(session.recognizer.get());
   session.recognizer->StopAudioCapture();
 }
 
 void SpeechRecognitionManagerImpl::ResetCapturingSessionId(
     const Session& session) {
   DCHECK_EQ(primary_session_id_, session.id);
   primary_session_id_ = kSessionIDInvalid;
 }
 
-void SpeechRecognitionManagerImpl::SessionDelete(const Session& session) {
-  DCHECK(session.recognizer.get() == NULL || !session.recognizer->IsActive());
-  if (primary_session_id_ == session.id)
+void SpeechRecognitionManagerImpl::SessionDelete(Session* session) {
+  DCHECK(session->recognizer == NULL || !session->recognizer->IsActive());
+  if (primary_session_id_ == session->id)
     primary_session_id_ = kSessionIDInvalid;
-  sessions_.erase(session.id);
+  sessions_.erase(session->id);
+  delete session;
 }
 
 void SpeechRecognitionManagerImpl::NotFeasible(const Session& session,
                                                FSMEvent event) {
   NOTREACHED() << "Unfeasible event " << event
                << " in state " << GetSessionState(session.id)
                << " for session " << session.id;
 }
 
 int SpeechRecognitionManagerImpl::GetNextSessionID() {
   ++last_session_id_;
   // Deal with wrapping of last_session_id_. (How civilized).
   if (last_session_id_ <= 0)
     last_session_id_ = 1;
   return last_session_id_;
 }
 
 bool SpeechRecognitionManagerImpl::SessionExists(int session_id) const {
   return sessions_.find(session_id) != sessions_.end();
 }
 
-const SpeechRecognitionManagerImpl::Session&
+SpeechRecognitionManagerImpl::Session*
 SpeechRecognitionManagerImpl::GetSession(int session_id) const {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   SessionsTable::const_iterator iter = sessions_.find(session_id);
   DCHECK(iter != sessions_.end());
   return iter->second;
 }
 
 SpeechRecognitionEventListener* SpeechRecognitionManagerImpl::GetListener(
     int session_id) const {
-  const Session& session = GetSession(session_id);
-  return session.listener_is_active ? session.config.event_listener : NULL;
+  Session* session = GetSession(session_id);
+  return session->listener_is_active ? session->config.event_listener : NULL;
 }
 
 SpeechRecognitionEventListener*
 SpeechRecognitionManagerImpl::GetDelegateListener() const {
   return delegate_.get() ? delegate_->GetEventListener() : NULL;
 }
 
 const SpeechRecognitionSessionConfig&
 SpeechRecognitionManagerImpl::GetSessionConfig(int session_id) const {
-  return GetSession(session_id).config;
+  return GetSession(session_id)->config;
 }
 
 bool SpeechRecognitionManagerImpl::HasAudioInputDevices() {
   return BrowserMainLoop::GetAudioManager()->HasAudioInputDevices();
 }
 
 string16 SpeechRecognitionManagerImpl::GetAudioInputDeviceModel() {
   return BrowserMainLoop::GetAudioManager()->GetAudioInputDeviceModel();
 }
 
 void SpeechRecognitionManagerImpl::ShowAudioInputSettings() {
   // Since AudioManager::ShowAudioInputSettings can potentially launch external
   // processes, do that in the FILE thread to not block the calling threads.
   BrowserThread::PostTask(BrowserThread::FILE, FROM_HERE,
                           base::Bind(&ShowAudioInputSettingsOnFileThread));
 }
 
 SpeechRecognitionManagerImpl::Session::Session()
   : id(kSessionIDInvalid),
     listener_is_active(true) {
 }
 
 SpeechRecognitionManagerImpl::Session::~Session() {
 }
 
 }  // namespace content