SpeechInputExtensionManager now interface (exclusively) with...

SpeechInputExtensionManager now interface (exclusively) with SpeechRecognitionManagerDelegate (Speech CL1.11).

- The tray icon and balloon handing has been moved from speech_input_extensions_manager to chrome_speech_recognition_manager_delegate, since that code (tray icon) will be used also for continuous recognition.
- Removed the SpeechRecognizer interface from /content/public (thus de-virtualized and refcounted the SpeechRecognizerImpl)

BUG=116954
TEST=none

Review URL: https://chromiumcodereview.appspot.com/10377082

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@138498 0039d316-1c4b-4281-b951-d872f2087c98

chrome/browser/speech/chrome_speech_recognition_manager_delegate.cc

@@ -13,6 +13,7 @@
 #include "chrome/browser/browser_process.h"
 #include "chrome/browser/prefs/pref_service.h"
 #include "chrome/browser/profiles/profile_manager.h"
+#include "chrome/browser/speech/speech_recognition_tray_icon_controller.h"
 #include "chrome/browser/tab_contents/tab_util.h"
 #include "chrome/browser/view_type_utils.h"
 #include "chrome/common/pref_names.h"
@@ -36,10 +37,20 @@
 using content::BrowserThread;
 using content::SpeechRecognitionManager;
 using content::WebContents;
+using content::SpeechRecognitionSessionContext;
 
 namespace {
 const int kNoActiveBubble =
     content::SpeechRecognitionManager::kSessionIDInvalid;
+
+bool RequiresBubble(int session_id) {
+  return SpeechRecognitionManager::GetInstance()->
+      GetSessionContext(session_id).requested_by_page_element;
+}
+
+bool RequiresTrayIcon(int session_id) {
+  return !RequiresBubble(session_id);
+}
 }  // namespace
 
 namespace speech {
@@ -108,13 +119,17 @@ class ChromeSpeechRecognitionManagerDelegate::OptionalRequestInfo
 };
 
 ChromeSpeechRecognitionManagerDelegate::ChromeSpeechRecognitionManagerDelegate()
-    : bubble_controller_(new SpeechRecognitionBubbleController(
-          ALLOW_THIS_IN_INITIALIZER_LIST(this))),
-      active_bubble_session_id_(kNoActiveBubble) {
+    : active_bubble_session_id_(kNoActiveBubble) {
 }
 
 ChromeSpeechRecognitionManagerDelegate::
     ~ChromeSpeechRecognitionManagerDelegate() {
+  if (tray_icon_controller_.get())
+    tray_icon_controller_->Hide();
+  if (active_bubble_session_id_ != kNoActiveBubble) {
+    DCHECK(bubble_controller_.get());
+    bubble_controller_->CloseBubble(active_bubble_session_id_);
+  }
 }
 
 void ChromeSpeechRecognitionManagerDelegate::InfoBubbleButtonClicked(
@@ -122,8 +137,11 @@ void ChromeSpeechRecognitionManagerDelegate::InfoBubbleButtonClicked(
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   DCHECK_EQ(active_bubble_session_id_, session_id);
 
+  // Note, the session might have been destroyed, therefore avoid calls to the
+  // manager which imply its existance (e.g., GetSessionContext()).
+
   if (button == SpeechRecognitionBubble::BUTTON_CANCEL) {
-    bubble_controller_->CloseBubble(session_id);
+    GetBubbleController()->CloseBubble(session_id);
     last_session_config_.reset();
     active_bubble_session_id_ = kNoActiveBubble;
 
@@ -131,7 +149,7 @@ void ChromeSpeechRecognitionManagerDelegate::InfoBubbleButtonClicked(
     // the manager's public methods are reliable and will handle it properly.
     SpeechRecognitionManager::GetInstance()->AbortSession(session_id);
   } else if (button == SpeechRecognitionBubble::BUTTON_TRY_AGAIN) {
-    bubble_controller_->CloseBubble(session_id);
+    GetBubbleController()->CloseBubble(session_id);
     active_bubble_session_id_ = kNoActiveBubble;
     RestartLastSession();
   }
@@ -142,7 +160,10 @@ void ChromeSpeechRecognitionManagerDelegate::InfoBubbleFocusChanged(
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   DCHECK_EQ(active_bubble_session_id_, session_id);
 
-  bubble_controller_->CloseBubble(session_id);
+  // Note, the session might have been destroyed, therefore avoid calls to the
+  // manager which imply its existance (e.g., GetSessionContext()).
+
+  GetBubbleController()->CloseBubble(session_id);
   last_session_config_.reset();
   active_bubble_session_id_ = kNoActiveBubble;
 
@@ -164,6 +185,10 @@ void ChromeSpeechRecognitionManagerDelegate::RestartLastSession() {
 
 void ChromeSpeechRecognitionManagerDelegate::OnRecognitionStart(
     int session_id) {
+  const content::SpeechRecognitionSessionContext& context =
+      SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
+
+  if (RequiresBubble(session_id)) {
     // Copy the configuration of the session (for the "try again" button).
     last_session_config_.reset(new content::SpeechRecognitionSessionConfig(
         SpeechRecognitionManager::GetInstance()->GetSessionConfig(session_id)));
@@ -171,18 +196,25 @@ void ChromeSpeechRecognitionManagerDelegate::OnRecognitionStart(
     // Create and show the bubble.
     DCHECK_EQ(active_bubble_session_id_, kNoActiveBubble);
     active_bubble_session_id_ = session_id;
-  const content::SpeechRecognitionSessionContext& context =
-      SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  bubble_controller_->CreateBubble(session_id,
+    GetBubbleController()->CreateBubble(session_id,
                                         context.render_process_id,
                                         context.render_view_id,
                                         context.element_rect);
-  // TODO(primiano) Why not create directly the bubble in warmup mode?
-  bubble_controller_->SetBubbleWarmUpMode(session_id);
+
+    // TODO(primiano) Why not creating directly the bubble in warmup mode?
+    GetBubbleController()->SetBubbleWarmUpMode(session_id);
+  }
 }
 
 void ChromeSpeechRecognitionManagerDelegate::OnAudioStart(int session_id) {
-  bubble_controller_->SetBubbleRecordingMode(session_id);
+  if (RequiresBubble(session_id)) {
+    GetBubbleController()->SetBubbleRecordingMode(session_id);
+  } else if (RequiresTrayIcon(session_id)) {
+    const content::SpeechRecognitionSessionContext& context =
+        SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
+    GetTrayIconController()->Show(context.context_name,
+                                  context.is_first_request_for_context);
+  }
 }
 
 void ChromeSpeechRecognitionManagerDelegate::OnEnvironmentEstimationComplete(
@@ -198,16 +230,19 @@ void ChromeSpeechRecognitionManagerDelegate::OnSoundEnd(int session_id) {
 void ChromeSpeechRecognitionManagerDelegate::OnAudioEnd(int session_id) {
   // OnAudioEnd can be also raised after an abort, when the bubble has already
   // been closed.
-  if (active_bubble_session_id_ == session_id)
-    bubble_controller_->SetBubbleRecognizingMode(session_id);
+  if (RequiresBubble(session_id) && active_bubble_session_id_ == session_id) {
+    GetBubbleController()->SetBubbleRecognizingMode(session_id);
+  } else if (RequiresTrayIcon(session_id)) {
+    GetTrayIconController()->Hide();
+  }
 }
 
 void ChromeSpeechRecognitionManagerDelegate::OnRecognitionResult(
     int session_id, const content::SpeechRecognitionResult& result) {
   // A result can be dispatched when the bubble is not visible anymore (e.g.,
   // lost focus while waiting for a result, thus continuing in background).
-  if (active_bubble_session_id_ == session_id) {
-    bubble_controller_->CloseBubble(session_id);
+  if (RequiresBubble(session_id) && active_bubble_session_id_ == session_id) {
+    GetBubbleController()->CloseBubble(session_id);
     last_session_config_.reset();
     active_bubble_session_id_ = kNoActiveBubble;
   }
@@ -218,6 +253,7 @@ void ChromeSpeechRecognitionManagerDelegate::OnRecognitionError(
   // An error can be dispatched when the bubble is not visible anymore.
   if (active_bubble_session_id_ != session_id)
     return;
+  DCHECK(RequiresBubble(session_id));
 
   int error_message_id = 0;
   switch (error.code) {
@@ -250,17 +286,29 @@ void ChromeSpeechRecognitionManagerDelegate::OnRecognitionError(
       NOTREACHED() << "unknown error " << error.code;
       return;
   }
-  bubble_controller_->SetBubbleMessage(
+  GetBubbleController()->SetBubbleMessage(
       session_id, l10n_util::GetStringUTF16(error_message_id));
 }
 
 void ChromeSpeechRecognitionManagerDelegate::OnAudioLevelsChange(
     int session_id, float volume, float noise_volume) {
-  if (active_bubble_session_id_ == session_id)
-    bubble_controller_->SetBubbleInputVolume(session_id, volume, noise_volume);
+  if (active_bubble_session_id_ == session_id) {
+    DCHECK(RequiresBubble(session_id));
+    GetBubbleController()->SetBubbleInputVolume(session_id,
+                                                volume, noise_volume);
+  } else if (RequiresTrayIcon(session_id)) {
+    GetTrayIconController()->SetVUMeterVolume(volume);
+  }
 }
 
 void ChromeSpeechRecognitionManagerDelegate::OnRecognitionEnd(int session_id) {
+  // No need to remove the bubble here, since either one of the following events
+  // must have happened prior to this callback:
+  // - A previous OnRecognitionResult event already closed the bubble.
+  // - An error occurred, so the bubble is showing the error and will be closed
+  //   when it will lose focus (by InfoBubbleFocusChanged()).
+  // - The bubble lost focus or the user pressed the Cancel button, thus it has
+  //   been closed by InfoBubbleFocusChanged(), which triggered an AbortSession.
 }
 
 void ChromeSpeechRecognitionManagerDelegate::GetDiagnosticInformation(
@@ -286,11 +334,20 @@ void ChromeSpeechRecognitionManagerDelegate::CheckRecognitionIsAllowed(
     int session_id,
     base::Callback<void(int session_id, bool is_allowed)> callback) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  const content::SpeechRecognitionSessionContext& context =
-      SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
 
+  // We don't need any particular check for sessions not using a bubble. In such
+  // cases, we just notify it to the manager (calling-back synchronously, since
+  // we remain in the IO thread).
+  if (RequiresTrayIcon(session_id)) {
+    callback.Run(session_id, true /* is_allowed */);
+    return;
+  }
+
+  // Sessions using bubbles, conversely, need a check on the renderer view type.
   // The check must be performed in the UI thread. We defer it posting to
   // CheckRenderViewType, which will issue the callback on our behalf.
+  const content::SpeechRecognitionSessionContext& context =
+      SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
   BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
                           base::Bind(&CheckRenderViewType,
                                      session_id,
@@ -330,4 +387,19 @@ void ChromeSpeechRecognitionManagerDelegate::CheckRenderViewType(
                           base::Bind(callback, session_id, allowed));
 }
 
+SpeechRecognitionBubbleController*
+ChromeSpeechRecognitionManagerDelegate::GetBubbleController() {
+  if (!bubble_controller_.get())
+    bubble_controller_ = new SpeechRecognitionBubbleController(this);
+  return bubble_controller_.get();
+}
+
+SpeechRecognitionTrayIconController*
+ChromeSpeechRecognitionManagerDelegate::GetTrayIconController() {
+  if (!tray_icon_controller_.get())
+    tray_icon_controller_ = new SpeechRecognitionTrayIconController();
+  return tray_icon_controller_.get();
+}
+
+
 }  // namespace speech

chrome/browser/speech/chrome_speech_recognition_manager_delegate.h

@@ -13,6 +13,8 @@
 #include "content/public/browser/speech_recognition_manager_delegate.h"
 #include "content/public/browser/speech_recognition_session_config.h"
 
+class SpeechRecognitionTrayIconController;
+
 namespace speech {
 
 // This is Chrome's implementation of the SpeechRecognitionManagerDelegate
@@ -69,7 +71,12 @@ class ChromeSpeechRecognitionManagerDelegate
   // (which is copied into |last_session_config_|). Used for "try again".
   void RestartLastSession();
 
+  // Lazy initializers for bubble and tray icon controller.
+  SpeechRecognitionBubbleController* GetBubbleController();
+  SpeechRecognitionTrayIconController* GetTrayIconController();
+
   scoped_refptr<SpeechRecognitionBubbleController> bubble_controller_;
+  scoped_refptr<SpeechRecognitionTrayIconController> tray_icon_controller_;
   scoped_refptr<OptionalRequestInfo> optional_request_info_;
   scoped_ptr<content::SpeechRecognitionSessionConfig> last_session_config_;
 

chrome/browser/speech/speech_input_extension_apitest.cc

@@ -23,6 +23,10 @@ namespace net {
 class URLRequestContextGetter;
 }
 
+namespace {
+const int kSessionIDForTests = 0;
+}
+
 // Mock class used to test the extension speech input API.
 class SpeechInputExtensionApiTest : public ExtensionApiTest,
                                     public SpeechInputExtensionInterface {
@@ -77,10 +81,11 @@ class SpeechInputExtensionApiTest : public ExtensionApiTest,
   virtual void StartRecording(
       content::SpeechRecognitionEventListener* listener,
       net::URLRequestContextGetter* context_getter,
-      int session_id,
+      const string16& extension_name,
       const std::string& language,
       const std::string& grammar,
-      bool filter_profanities) OVERRIDE;
+      bool filter_profanities,
+      bool show_notification) OVERRIDE;
 
   virtual void StopRecording(bool recognition_failed) OVERRIDE;
 
@@ -107,7 +112,7 @@ class SpeechInputExtensionApiTest : public ExtensionApiTest,
   };
 
  private:
-  void ProvideResults(int session_id);
+  void ProvideResults();
 
   bool recording_devices_available_;
   bool recognizer_is_valid_;
@@ -131,10 +136,11 @@ SpeechInputExtensionApiTest::~SpeechInputExtensionApiTest() {
 void SpeechInputExtensionApiTest::StartRecording(
       content::SpeechRecognitionEventListener* listener,
       net::URLRequestContextGetter* context_getter,
-      int session_id,
+      const string16& extension_name,
       const std::string& language,
       const std::string& grammar,
-      bool filter_profanities) {
+      bool filter_profanities,
+      bool show_notification) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   recognizer_is_valid_ = true;
 
@@ -143,7 +149,7 @@ void SpeechInputExtensionApiTest::StartRecording(
       FROM_HERE,
       base::Bind(&SpeechInputExtensionManager::OnAudioStart,
                  GetManager(),
-                 session_id),
+                 kSessionIDForTests),
       base::TimeDelta());
 
   // Notify sound start in the input device.
@@ -151,16 +157,14 @@ void SpeechInputExtensionApiTest::StartRecording(
       FROM_HERE,
       base::Bind(&SpeechInputExtensionManager::OnSoundStart,
                  GetManager(),
-                 session_id),
+                 kSessionIDForTests),
       base::TimeDelta());
 
   if (result_delay_ms_ != kDontDispatchCall) {
     // Dispatch the recognition results.
     MessageLoop::current()->PostDelayedTask(
         FROM_HERE,
-        base::Bind(&SpeechInputExtensionApiTest::ProvideResults,
-                   this,
-                   session_id),
+        base::Bind(&SpeechInputExtensionApiTest::ProvideResults, this),
         base::TimeDelta::FromMilliseconds(result_delay_ms_));
   }
 }
@@ -170,18 +174,19 @@ void SpeechInputExtensionApiTest::StopRecording(bool recognition_failed) {
   recognizer_is_valid_ = false;
 }
 
-void SpeechInputExtensionApiTest::ProvideResults(int session_id) {
+void SpeechInputExtensionApiTest::ProvideResults() {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 
   if (next_error_ != content::SPEECH_RECOGNITION_ERROR_NONE) {
     GetManager()->OnRecognitionError(
-        session_id, content::SpeechRecognitionError(next_error_));
+        kSessionIDForTests, content::SpeechRecognitionError(next_error_));
     return;
   }
 
-  GetManager()->OnSoundEnd(session_id);
-  GetManager()->OnAudioEnd(session_id);
-  GetManager()->OnRecognitionResult(session_id, next_result_);
+  GetManager()->OnSoundEnd(kSessionIDForTests);
+  GetManager()->OnAudioEnd(kSessionIDForTests);
+  GetManager()->OnRecognitionResult(kSessionIDForTests, next_result_);
+  GetManager()->OnRecognitionEnd(kSessionIDForTests);
 }
 
 // Every test should leave the manager in the idle state when finished.

chrome/browser/speech/speech_input_extension_manager.cc

@@ -15,7 +15,6 @@
 #include "chrome/browser/profiles/profile_dependency_manager.h"
 #include "chrome/browser/profiles/profile_keyed_service.h"
 #include "chrome/browser/profiles/profile_keyed_service_factory.h"
-#include "chrome/browser/speech/speech_recognition_tray_icon_controller.h"
 #include "chrome/common/chrome_notification_types.h"
 #include "chrome/common/extensions/extension.h"
 #include "chrome/common/pref_names.h"
@@ -23,9 +22,11 @@
 #include "content/public/browser/notification_registrar.h"
 #include "content/public/browser/notification_service.h"
 #include "content/public/browser/speech_recognition_manager.h"
-#include "content/public/browser/speech_recognizer.h"
+#include "content/public/browser/speech_recognition_session_config.h"
+#include "content/public/browser/speech_recognition_session_context.h"
 #include "content/public/common/speech_recognition_error.h"
 #include "content/public/common/speech_recognition_result.h"
+#include "net/url_request/url_request_context_getter.h"
 
 using content::BrowserThread;
 using content::SpeechRecognitionHypothesis;
@@ -55,12 +56,6 @@ const char kOnResultEvent[] = "experimental.speechInput.onResult";
 const char kOnSoundStartEvent[] = "experimental.speechInput.onSoundStart";
 const char kOnSoundEndEvent[] = "experimental.speechInput.onSoundEnd";
 
-// Session id provided to the speech recognizer. Since only one extension can
-// be recording on the same time a constant value is enough as id.
-// TODO(primiano) this will not be valid anymore once speech input extension
-// will use the SpeechRecognitionManager and not the SpeechRecognizer directly.
-static const int kSpeechInputSessionId = 1;
-
 // Wrap an SpeechInputExtensionManager using scoped_refptr to avoid
 // assertion failures on destruction because of not using release().
 class SpeechInputExtensionManagerWrapper : public ProfileKeyedService {
@@ -157,7 +152,10 @@ SpeechInputExtensionManager::SpeechInputExtensionManager(Profile* profile)
     : profile_(profile),
       state_(kIdle),
       registrar_(new content::NotificationRegistrar),
-      speech_interface_(NULL) {
+      speech_interface_(NULL),
+      is_recognition_in_progress_(false),
+      speech_recognition_session_id_(
+          SpeechRecognitionManager::kSessionIDInvalid) {
   registrar_->Add(this, chrome::NOTIFICATION_EXTENSION_UNLOADED,
                   content::Source<Profile>(profile_));
 }
@@ -194,11 +192,17 @@ void SpeechInputExtensionManager::ShutdownOnUIThread() {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
   VLOG(1) << "Profile shutting down.";
 
+  // Note: Unretained(this) is safe, also if we are freed in the meanwhile.
+  // It is used by the SR manager just for comparing the raw pointer and remove
+  // the associated sessions.
+  BrowserThread::PostTask(BrowserThread::IO, FROM_HERE,
+      base::Bind(&SpeechRecognitionManager::AbortAllSessionsForListener,
+                 base::Unretained(SpeechRecognitionManager::GetInstance()),
+                 base::Unretained(this)));
+
   base::AutoLock auto_lock(state_lock_);
   DCHECK(state_ != kShutdown);
   if (state_ != kIdle) {
-    DCHECK(notification_.get());
-    notification_->Hide();
     BrowserThread::PostTask(BrowserThread::IO, FROM_HERE,
         base::Bind(&SpeechInputExtensionManager::ForceStopOnIOThread, this));
   }
@@ -245,7 +249,7 @@ void SpeechInputExtensionManager::OnRecognitionResult(
     int session_id,
     const content::SpeechRecognitionResult& result) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
+  DCHECK_EQ(session_id, speech_recognition_session_id_);
 
   // Stopping will start the disassociation with the extension.
   // Make a copy to report the results to the proper one.
@@ -288,13 +292,13 @@ void SpeechInputExtensionManager::SetRecognitionResultOnUIThread(
 }
 
 void SpeechInputExtensionManager::OnRecognitionStart(int session_id) {
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
+  DCHECK_EQ(session_id, speech_recognition_session_id_);
 }
 
 void SpeechInputExtensionManager::OnAudioStart(int session_id) {
   VLOG(1) << "OnAudioStart";
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
+  DCHECK_EQ(session_id, speech_recognition_session_id_);
 
   BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
       base::Bind(&SpeechInputExtensionManager::DidStartReceivingAudioOnUIThread,
@@ -302,11 +306,16 @@ void SpeechInputExtensionManager::OnAudioStart(int session_id) {
 }
 
 void SpeechInputExtensionManager::OnAudioEnd(int session_id) {
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
 }
 
 void SpeechInputExtensionManager::OnRecognitionEnd(int session_id) {
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
+  // In the very exceptional case in which we requested a new recognition before
+  // the previous one ended, don't clobber the speech_recognition_session_id_.
+  if (speech_recognition_session_id_ == session_id) {
+    is_recognition_in_progress_ = false;
+    speech_recognition_session_id_ =
+        SpeechRecognitionManager::kSessionIDInvalid;
+  }
 }
 
 void SpeechInputExtensionManager::DidStartReceivingAudioOnUIThread() {
@@ -320,21 +329,9 @@ void SpeechInputExtensionManager::DidStartReceivingAudioOnUIThread() {
   VLOG(1) << "State changed to recording";
   state_ = kRecording;
 
-  const extensions::Extension* extension = profile_->GetExtensionService()->
-      GetExtensionById(extension_id_in_use_, true);
-  DCHECK(extension);
-
-  bool show_notification = !profile_->GetPrefs()->GetBoolean(
-      prefs::kSpeechInputTrayNotificationShown);
-
-  if (!notification_.get())
-    notification_ = new SpeechRecognitionTrayIconController();
-  notification_->Show(UTF8ToUTF16(extension->name()), show_notification);
-
-  if (show_notification) {
+  DCHECK(profile_);
   profile_->GetPrefs()->SetBoolean(
       prefs::kSpeechInputTrayNotificationShown, true);
-  }
 
   VLOG(1) << "Sending start notification";
   content::NotificationService::current()->Notify(
@@ -346,20 +343,13 @@ void SpeechInputExtensionManager::DidStartReceivingAudioOnUIThread() {
 void SpeechInputExtensionManager::OnRecognitionError(
     int session_id, const content::SpeechRecognitionError& error) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
-
-  // Simply return in case of an ERROR_ABORTED, since it is not contemplated
-  // in the speech input extensions architecture.
-  if (error.code == content::SPEECH_RECOGNITION_ERROR_ABORTED)
-      return;
-
+  DCHECK_EQ(session_id, speech_recognition_session_id_);
   VLOG(1) << "OnRecognitionError: " << error.code;
 
   base::AutoLock auto_lock(state_lock_);
   if (state_ == kShutdown)
     return;
 
-  // Release the recognizer object.
   GetSpeechInputExtensionInterface()->StopRecording(true);
 
   std::string event_error_code;
@@ -369,6 +359,15 @@ void SpeechInputExtensionManager::OnRecognitionError(
     case content::SPEECH_RECOGNITION_ERROR_NONE:
       break;
 
+    case content::SPEECH_RECOGNITION_ERROR_ABORTED:
+      // ERROR_ABORTED is received whenever AbortSession is called on the
+      // manager. However, we want propagate the error only if it is triggered
+      // by an external cause (another recognition started, aborting us), thus
+      // only if it occurs while we are capturing audio.
+      if (state_ == kRecording)
+        event_error_code = kErrorCaptureError;
+      break;
+
     case content::SPEECH_RECOGNITION_ERROR_AUDIO:
       if (state_ == kStarting) {
         event_error_code = kErrorUnableToStart;
@@ -411,12 +410,12 @@ void SpeechInputExtensionManager::OnRecognitionError(
 
 void SpeechInputExtensionManager::OnEnvironmentEstimationComplete(
     int session_id) {
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
+  DCHECK_EQ(session_id, speech_recognition_session_id_);
 }
 
 void SpeechInputExtensionManager::OnSoundStart(int session_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
+  DCHECK_EQ(session_id, speech_recognition_session_id_);
   VLOG(1) << "OnSoundStart";
 
   std::string json_args;
@@ -428,7 +427,6 @@ void SpeechInputExtensionManager::OnSoundStart(int session_id) {
 
 void SpeechInputExtensionManager::OnSoundEnd(int session_id) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
   VLOG(1) << "OnSoundEnd";
 
   std::string json_args;
@@ -471,11 +469,6 @@ void SpeechInputExtensionManager::DispatchError(
     if (state_ == kShutdown)
       return;
 
-    if (state_ == kRecording) {
-      DCHECK(notification_.get());
-      notification_->Hide();
-    }
-
     extension_id = extension_id_in_use_;
     ResetToIdleState();
 
@@ -531,23 +524,37 @@ bool SpeechInputExtensionManager::Start(
       NOTREACHED();
   }
 
+  const extensions::Extension* extension = profile_->GetExtensionService()->
+      GetExtensionById(extension_id, true);
+  DCHECK(extension);
+  const string16& extension_name = UTF8ToUTF16(extension->name());
+
   extension_id_in_use_ = extension_id;
   VLOG(1) << "State changed to starting";
   state_ = kStarting;
 
+  // Checks if the security notification balloon has been already shown (only
+  // once for a profile). It is reset on DidStartReceivingAudioOnUIThread.
+  scoped_refptr<net::URLRequestContextGetter> url_request_context_getter =
+      profile_->GetRequestContext();
+  const bool show_notification = !profile_->GetPrefs()->GetBoolean(
+       prefs::kSpeechInputTrayNotificationShown);
+
   BrowserThread::PostTask(
       BrowserThread::IO, FROM_HERE,
       base::Bind(&SpeechInputExtensionManager::StartOnIOThread, this,
-                 profile_->GetRequestContext(), language, grammar,
-                 filter_profanities));
+                 url_request_context_getter, extension_name, language, grammar,
+                 filter_profanities, show_notification));
   return true;
 }
 
 void SpeechInputExtensionManager::StartOnIOThread(
-    net::URLRequestContextGetter* context_getter,
+    scoped_refptr<net::URLRequestContextGetter> context_getter,
+    const string16& extension_name,
     const std::string& language,
     const std::string& grammar,
-    bool filter_profanities) {
+    bool filter_profanities,
+    bool show_notification) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
   VLOG(1) << "Requesting start (IO thread)";
 
@@ -558,6 +565,12 @@ void SpeechInputExtensionManager::StartOnIOThread(
   if (state_ == kShutdown)
     return;
 
+  // TODO(primiano) These two checks below could be avoided, since they are
+  // already handled in the speech recognition classes. However, since the
+  // speech input extensions tests are bypassing the manager, we need them to
+  // pass the tests. A complete unit test which puts all the pieces together,
+  // mocking just the endpoints (the audio input controller and the URL fetcher)
+  // should be written.
   if (!GetSpeechInputExtensionInterface()->HasAudioInputDevices()) {
     BrowserThread::PostTask(
         BrowserThread::UI, FROM_HERE,
@@ -573,9 +586,13 @@ void SpeechInputExtensionManager::StartOnIOThread(
     return;
   }
 
-  GetSpeechInputExtensionInterface()->StartRecording(
-      this, context_getter, kSpeechInputSessionId, language, grammar,
-      filter_profanities);
+  GetSpeechInputExtensionInterface()->StartRecording(this,
+                                                     context_getter,
+                                                     extension_name,
+                                                     language,
+                                                     grammar,
+                                                     filter_profanities,
+                                                     show_notification);
 }
 
 bool SpeechInputExtensionManager::HasAudioInputDevices() {
@@ -602,6 +619,7 @@ void SpeechInputExtensionManager::IsRecordingOnIOThread(
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
 
   bool result = GetSpeechInputExtensionInterface()->IsCapturingAudio();
+
   BrowserThread::PostTask(
       BrowserThread::UI, FROM_HERE,
       base::Bind(&SpeechInputExtensionManager::IsRecordingOnUIThread,
@@ -618,20 +636,41 @@ void SpeechInputExtensionManager::IsRecordingOnUIThread(
 void SpeechInputExtensionManager::StartRecording(
     content::SpeechRecognitionEventListener* listener,
     net::URLRequestContextGetter* context_getter,
-    int session_id,
+    const string16& extension_name,
     const std::string& language,
     const std::string& grammar,
-    bool filter_profanities) {
+    bool filter_profanities,
+    bool show_notification) {
   DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  DCHECK(!recognizer_);
-  recognizer_ = content::SpeechRecognizer::Create(
-      listener, session_id, language, grammar, context_getter,
-      filter_profanities, "", "");
-  recognizer_->StartRecognition();
+
+  content::SpeechRecognitionSessionContext context;
+  context.requested_by_page_element = false;
+  context.is_first_request_for_context = show_notification;
+  context.context_name = extension_name;
+
+  content::SpeechRecognitionSessionConfig config;
+  // config.is_one_shot = true; // TODO(primiano) Uncomment when CL2.0 lands.
+  config.language = language;
+  config.grammars.push_back(content::SpeechRecognitionGrammar(grammar));
+  config.initial_context = context;
+  config.url_request_context_getter = context_getter;
+  config.filter_profanities = filter_profanities;
+  config.event_listener = listener;
+
+  DCHECK(!is_recognition_in_progress_);
+  SpeechRecognitionManager& manager = *SpeechRecognitionManager::GetInstance();
+  speech_recognition_session_id_ =
+      manager.CreateSession(config);
+  DCHECK_NE(speech_recognition_session_id_,
+            SpeechRecognitionManager::kSessionIDInvalid);
+  is_recognition_in_progress_ = true;
+  manager.StartSession(speech_recognition_session_id_);
 }
 
 bool SpeechInputExtensionManager::HasValidRecognizer() {
-  return !!recognizer_;
+  if (!is_recognition_in_progress_)
+    return false;
+  return SpeechRecognitionManager::GetInstance()->IsCapturingAudio();
 }
 
 bool SpeechInputExtensionManager::Stop(const std::string& extension_id,
@@ -692,13 +731,13 @@ void SpeechInputExtensionManager::ForceStopOnIOThread() {
 }
 
 void SpeechInputExtensionManager::StopRecording(bool recognition_failed) {
-  if (recognizer_) {
-    // Recognition is already cancelled in case of failure.
-    // Double-cancelling leads to assertion failures.
-    if (!recognition_failed)
-      recognizer_->AbortRecognition();
-    recognizer_.release();
-  }
+  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
+  if (!is_recognition_in_progress_)
+    return;
+  DCHECK_NE(speech_recognition_session_id_,
+            SpeechRecognitionManager::kSessionIDInvalid);
+  SpeechRecognitionManager::GetInstance()->AbortSession(
+      speech_recognition_session_id_);
 }
 
 void SpeechInputExtensionManager::StopSucceededOnUIThread() {
@@ -712,9 +751,6 @@ void SpeechInputExtensionManager::StopSucceededOnUIThread() {
   std::string extension_id = extension_id_in_use_;
   ResetToIdleState();
 
-  DCHECK(notification_.get());
-  notification_->Hide();
-
   content::NotificationService::current()->Notify(
       chrome::NOTIFICATION_EXTENSION_SPEECH_INPUT_RECORDING_STOPPED,
       // Guarded by the state_ == kShutdown check.
@@ -724,17 +760,4 @@ void SpeechInputExtensionManager::StopSucceededOnUIThread() {
 
 void SpeechInputExtensionManager::OnAudioLevelsChange(int session_id,
                                                       float volume,
-                                                       float noise_volume) {
-  DCHECK_EQ(session_id, kSpeechInputSessionId);
-  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::IO));
-  BrowserThread::PostTask(BrowserThread::UI, FROM_HERE,
-      base::Bind(&SpeechInputExtensionManager::SetInputVolumeOnUIThread,
-      this, volume));
-}
-
-void SpeechInputExtensionManager::SetInputVolumeOnUIThread(
-    float volume) {
-  DCHECK(BrowserThread::CurrentlyOn(BrowserThread::UI));
-  DCHECK(notification_.get());
-  notification_->SetVUMeterVolume(volume);
-}
+                                                      float noise_volume) {}

chrome/browser/speech/speech_input_extension_manager.h

@@ -11,18 +11,18 @@
 #include "base/callback_forward.h"
 #include "base/memory/ref_counted.h"
 #include "base/memory/scoped_ptr.h"
+#include "base/string16.h"
 #include "base/synchronization/lock.h"
 #include "content/public/browser/notification_observer.h"
 #include "content/public/browser/speech_recognition_event_listener.h"
 
 class Profile;
-class SpeechRecognitionTrayIconController;
 
 namespace content {
 class NotificationRegistrar;
 struct SpeechRecognitionError;
+class SpeechRecognitionManager;
 struct SpeechRecognitionResult;
-class SpeechRecognizer;
 }
 
 namespace net {
@@ -39,10 +39,11 @@ class SpeechInputExtensionInterface {
   virtual void StartRecording(
       content::SpeechRecognitionEventListener* listener,
       net::URLRequestContextGetter* context_getter,
-      int session_id,
+      const string16& extension_name,
       const std::string& language,
       const std::string& grammar,
-      bool filter_profanities) = 0;
+      bool filter_profanities,
+      bool show_notification) = 0;
 
   virtual void StopRecording(bool recognition_failed) = 0;
   virtual bool HasAudioInputDevices() = 0;
@@ -143,19 +144,22 @@ class SpeechInputExtensionManager
   virtual void StartRecording(
       content::SpeechRecognitionEventListener* listener,
       net::URLRequestContextGetter* context_getter,
-      int session_id,
+      const string16& extension_name,
       const std::string& language,
       const std::string& grammar,
-      bool filter_profanities) OVERRIDE;
+      bool filter_profanities,
+      bool show_notification) OVERRIDE;
 
   virtual void StopRecording(bool recognition_failed) OVERRIDE;
 
   // Internal methods.
   void StartOnIOThread(
-      net::URLRequestContextGetter* context_getter,
+      scoped_refptr<net::URLRequestContextGetter> context_getter,
+      const string16& extension_name,
       const std::string& language,
       const std::string& grammar,
-      bool filter_profanities);
+      bool filter_profanities,
+      bool show_notification);
   void ForceStopOnIOThread();
   void IsRecordingOnIOThread(const IsRecordingCallback& callback);
 
@@ -172,7 +176,6 @@ class SpeechInputExtensionManager
                                 const std::string& json_args);
   void ExtensionUnloaded(const std::string& extension_id);
 
-  void SetInputVolumeOnUIThread(float volume);
   void ResetToIdleState();
 
   virtual ~SpeechInputExtensionManager();
@@ -196,10 +199,10 @@ class SpeechInputExtensionManager
   // Used in the UI thread.
   scoped_ptr<content::NotificationRegistrar> registrar_;
   SpeechInputExtensionInterface* speech_interface_;
-  scoped_refptr<SpeechRecognitionTrayIconController> notification_;
 
   // Used in the IO thread.
-  scoped_refptr<content::SpeechRecognizer> recognizer_;
+  bool is_recognition_in_progress_;
+  int speech_recognition_session_id_;
 };
 
 #endif  // CHROME_BROWSER_SPEECH_SPEECH_INPUT_EXTENSION_MANAGER_H_

content/browser/speech/speech_recognition_manager_impl.cc

@@ -327,6 +327,12 @@ void SpeechRecognitionManagerImpl::AbortAllSessionsForListener(
 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);
   FSMState session_state = GetSessionState(session_id);
   DCHECK_LE(session_state, SESSION_STATE_MAX_VALUE);

content/browser/speech/speech_recognizer_impl.cc

@@ -12,7 +12,6 @@
 #include "content/browser/speech/google_one_shot_remote_engine.h"
 #include "content/public/browser/browser_thread.h"
 #include "content/public/browser/speech_recognition_event_listener.h"
-#include "content/public/browser/speech_recognizer.h"
 #include "content/public/common/speech_recognition_error.h"
 #include "content/public/common/speech_recognition_grammar.h"
 #include "content/public/common/speech_recognition_result.h"
@@ -24,7 +23,6 @@ using content::SpeechRecognitionError;
 using content::SpeechRecognitionEventListener;
 using content::SpeechRecognitionGrammar;
 using content::SpeechRecognitionResult;
-using content::SpeechRecognizer;
 using media::AudioInputController;
 using media::AudioManager;
 using media::AudioParameters;
@@ -68,41 +66,6 @@ void KeepAudioControllerRefcountedForDtor(scoped_refptr<AudioInputController>) {
 
 }  // namespace
 
-// TODO(primiano) Create(...) is transitional (until we fix speech input
-// extensions) and should be removed soon. The manager should be the only one
-// knowing the existence of SpeechRecognizer(Impl), thus the only one in charge
-// of instantiating it.
-SpeechRecognizer* SpeechRecognizer::Create(
-    SpeechRecognitionEventListener* listener,
-    int session_id,
-    const std::string& language,
-    const std::string& grammar,
-    net::URLRequestContextGetter* context_getter,
-    bool filter_profanities,
-    const std::string& hardware_info,
-    const std::string& origin_url) {
-  speech::SpeechRecognitionEngineConfig remote_engine_config;
-  remote_engine_config.language = language;
-  if (!grammar.empty())
-    remote_engine_config.grammars.push_back(SpeechRecognitionGrammar(grammar));
-  remote_engine_config.audio_sample_rate =
-      speech::SpeechRecognizerImpl::kAudioSampleRate;
-  remote_engine_config.audio_num_bits_per_sample =
-      speech::SpeechRecognizerImpl::kNumBitsPerAudioSample;
-  remote_engine_config.filter_profanities = filter_profanities;
-  remote_engine_config.hardware_info = hardware_info;
-  remote_engine_config.origin_url = origin_url;
-
-  // SpeechRecognizerImpl takes ownership of google_remote_engine.
-  speech::GoogleOneShotRemoteEngine* google_remote_engine =
-      new speech::GoogleOneShotRemoteEngine(context_getter);
-  google_remote_engine->SetConfig(remote_engine_config);
-
-  return new speech::SpeechRecognizerImpl(listener,
-                                          session_id,
-                                          google_remote_engine);
-}
-
 namespace speech {
 
 const int SpeechRecognizerImpl::kAudioSampleRate = 16000;

content/browser/speech/speech_recognizer_impl.h

@@ -11,7 +11,6 @@
 #include "base/memory/scoped_ptr.h"
 #include "content/browser/speech/endpointer/endpointer.h"
 #include "content/browser/speech/speech_recognition_engine.h"
-#include "content/public/browser/speech_recognizer.h"
 #include "content/public/common/speech_recognition_error.h"
 #include "content/public/common/speech_recognition_result.h"
 #include "media/audio/audio_input_controller.h"
@@ -27,15 +26,13 @@ class AudioManager;
 
 namespace speech {
 
-// TODO(primiano) Next CL: Remove the Impl suffix and the exported
-// /content/public/browser/speech_recognizer.h interface since this class should
-// not be visible outside (currently we need it for speech input extension API).
+// TODO(primiano) Next CL: Remove the Impl suffix.
 
 // Handles speech recognition for a session (identified by |session_id|), taking
 // care of audio capture, silence detection/endpointer and interaction with the
 // SpeechRecognitionEngine.
 class CONTENT_EXPORT SpeechRecognizerImpl
-    : public NON_EXPORTED_BASE(content::SpeechRecognizer),
+    : public base::RefCountedThreadSafe<SpeechRecognizerImpl>,
       public media::AudioInputController::EventHandler,
       public NON_EXPORTED_BASE(SpeechRecognitionEngineDelegate) {
  public:
@@ -50,18 +47,15 @@ class CONTENT_EXPORT SpeechRecognizerImpl
       int session_id,
       SpeechRecognitionEngine* engine);
 
-  // content::SpeechRecognizer methods.
-  virtual void StartRecognition() OVERRIDE;
-  virtual void AbortRecognition() OVERRIDE;
-  virtual void StopAudioCapture() OVERRIDE;
-  virtual bool IsActive() const OVERRIDE;
-  virtual bool IsCapturingAudio() const OVERRIDE;
+  void StartRecognition();
+  void AbortRecognition();
+  void StopAudioCapture();
+  bool IsActive() const;
+  bool IsCapturingAudio() const;
   const SpeechRecognitionEngine& recognition_engine() const;
 
- protected:
-  virtual ~SpeechRecognizerImpl();
-
  private:
+  friend class base::RefCountedThreadSafe<SpeechRecognizerImpl>;
   friend class SpeechRecognizerImplTest;
 
   enum FSMState {
@@ -96,6 +90,8 @@ class CONTENT_EXPORT SpeechRecognizerImpl
     content::SpeechRecognitionError engine_error;
   };
 
+  virtual ~SpeechRecognizerImpl();
+
   // Entry point for pushing any new external event into the recognizer FSM.
   void DispatchEvent(const FSMEventArgs& event_args);
 

content/content_browser.gypi

@@ -139,8 +139,8 @@
     'public/browser/speech_recognition_preferences.h',
     'public/browser/speech_recognition_session_config.cc',
     'public/browser/speech_recognition_session_config.h',
+    'public/browser/speech_recognition_session_context.cc',
     'public/browser/speech_recognition_session_context.h',
-    'public/browser/speech_recognizer.h',
     'public/browser/trace_controller.h',
     'public/browser/trace_subscriber.h',
     'public/browser/user_metrics.h',

content/public/browser/speech_recognition_session_context.cc

+// Copyright (c) 2012 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/public/browser/speech_recognition_session_context.h"
+
+namespace content {
+
+SpeechRecognitionSessionContext::SpeechRecognitionSessionContext()
+    : render_process_id(0),
+      render_view_id(0),
+      render_request_id(0),
+      requested_by_page_element(true),
+      is_first_request_for_context(false) {
+}
+
+SpeechRecognitionSessionContext::~SpeechRecognitionSessionContext() {
+}
+
+}  // namespace content

content/public/browser/speech_recognition_session_context.h

@@ -6,6 +6,7 @@
 #define CONTENT_PUBLIC_BROWSER_SPEECH_RECOGNITION_SESSION_CONTEXT_H_
 #pragma once
 
+#include "base/string16.h"
 #include "content/common/content_export.h"
 #include "ui/gfx/rect.h"
 
@@ -21,18 +22,32 @@ namespace content {
 // SpeechRecognitionManager::GetSessionContext and
 // SpeechRecognitionManager::LookupSessionByContext methods).
 struct CONTENT_EXPORT SpeechRecognitionSessionContext {
-  SpeechRecognitionSessionContext()
-      : render_process_id(0),
-        render_view_id(0),
-        render_request_id(0),
-        js_handle_id(0) {}
-  ~SpeechRecognitionSessionContext() {}
+  SpeechRecognitionSessionContext();
+  ~SpeechRecognitionSessionContext();
 
   int render_process_id;
   int render_view_id;
   int render_request_id;
   int js_handle_id;
+
+  // Determines whether recognition was requested by a page element (in which
+  // case its coordinates are passed in |element_rect|).
+  bool requested_by_page_element;
+
+  // The coordinates of the page element for placing the bubble (valid only when
+  // |requested_by_page_element| = true).
   gfx::Rect element_rect;
+
+  // Determines whether this is the first time that this context (identified by
+  // |context_name|) is requesting a recognition.
+  // TODO(primiano) This is really temporary, remove after CL1.12 which will
+  // refactor SpeechRecognitionPreferences and move this check entirely whithin
+  // chrome, without involving content.
+  bool is_first_request_for_context;
+
+  // A texual description of the context (website, extension name) that is
+  // requesting recognition, for prompting security notifications to the user.
+  string16 context_name;
 };
 
 }  // namespace content

content/public/browser/speech_recognizer.h

-// Copyright (c) 2012 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.
-
-#ifndef CONTENT_PUBLIC_BROWSER_SPEECH_RECOGNIZER_H_
-#define CONTENT_PUBLIC_BROWSER_SPEECH_RECOGNIZER_H_
-#pragma once
-
-#include <string>
-
-#include "base/memory/ref_counted.h"
-#include "content/common/content_export.h"
-
-namespace net {
-class URLRequestContextGetter;
-}
-
-namespace content {
-
-class SpeechRecognitionEventListener;
-
-// Records audio, sends recorded audio to server and translates server response
-// to recognition result.
-// TODO(primiano) remove this public interface as soon as we fix speech input
-// extensions.
-class SpeechRecognizer : public base::RefCountedThreadSafe<SpeechRecognizer> {
- public:
-  CONTENT_EXPORT static SpeechRecognizer* Create(
-      SpeechRecognitionEventListener* event_listener,
-      int session_id,
-      const std::string& language,
-      const std::string& grammar,
-      net::URLRequestContextGetter* context_getter,
-      bool filter_profanities,
-      const std::string& hardware_info,
-      const std::string& origin_url);
-
-  // Starts audio recording and the recognition process. The same
-  // SpeechRecognizer instance can be used multiple times for speech recognition
-  // though each recognition request can be made only after the previous one
-  // completes (i.e. after receiving
-  // SpeechRecognitionEventListener::OnRecognitionEnd).
-  virtual void StartRecognition() = 0;
-
-  // Stops recording audio and cancels recognition. Any audio recorded so far
-  // gets discarded.
-  virtual void AbortRecognition() = 0;
-
-  // Stops recording audio and finalizes recognition, possibly getting results.
-  virtual void StopAudioCapture() = 0;
-
-  // Checks wether the recognizer is active, that is, either capturing audio
-  // or waiting for a result.
-  virtual bool IsActive() const = 0;
-
-  // Checks whether the recognizer is capturing audio.
-  virtual bool IsCapturingAudio() const = 0;
-
- protected:
-  friend class base::RefCountedThreadSafe<SpeechRecognizer>;
-  virtual ~SpeechRecognizer() {}
-};
-
-}  // namespace content
-
-#endif  // CONTENT_PUBLIC_BROWSER_SPEECH_RECOGNIZER_H_