Support onvoiceschanged for extension-based tts engines.

chrome/browser/speech/tts_controller.cc

Index: chrome/browser/speech/tts_controller.cc
diff --git a/chrome/browser/speech/tts_controller.cc b/chrome/browser/speech/tts_controller.cc
index f4a02423b19247a91bf036307dfb786db6d71bfd..a618234538e7816e3befb6399907be288525ea03 100644
--- a/chrome/browser/speech/tts_controller.cc
+++ b/chrome/browser/speech/tts_controller.cc
@@ -13,6 +13,8 @@
 #include "chrome/browser/profiles/profile.h"
 #include "chrome/browser/speech/extension_api/tts_engine_extension_api.h"
 #include "chrome/browser/speech/extension_api/tts_extension_api.h"
+#include "chrome/browser/speech/tts_engine_extension_observer.h"
+#include "chrome/browser/speech/tts_engine_extension_observer.h"
 #include "chrome/browser/speech/tts_platform.h"
 #include "chrome/common/extensions/api/speech/tts_engine_manifest_handler.h"
 #include "extensions/browser/extension_system.h"
@@ -156,6 +158,9 @@ void TtsController::SpeakNow(Utterance* utterance) {
   bool loaded_built_in =
       GetPlatformImpl()->LoadBuiltInTtsExtension(utterance->profile());
 
+  // Ensure we're observing newly added engines for the given profile.
+  TtsEngineExtensionObserver::GetInstance(utterance->profile());

[dmazzoni, 2014/07/07 15:37:26]

Is there a better place for this?

1) It feels like we should try to do this on more than just "Speak"

2) We're trying to move TtsController into content so I'd like to eliminate
direct dependencies on extension-related code. Can this be abstracted?

[David Tseng, 2014/07/08 23:43:31]

I've moved this call. Should address some of your other concerns I think. Let me
know if that looks like a good change and I think I can then clean up the
observer class a bit more (e.g. no longer a service).

+
   // Get all available voices and try to find a matching voice.
   std::vector<VoiceData> voices;
   GetVoices(utterance->profile(), &voices);