assistant: disable reading voice query result

Disable screen reader reading voice query result because Assistant
is going to respond with voice anyway.

Bug: b:113734839
Test: locally build and test
Change-Id: Ifa1079a6c551e8682857b8ab8e8afe0dd05218a7
Reviewed-on: https://chromium-review.googlesource.com/c/1232883Reviewed-by: Muyuan Li <muyuanli@chromium.org>
Commit-Queue: Xiaohui Chen <xiaohuic@chromium.org>
Cr-Commit-Position: refs/heads/master@{#597326}

ash/assistant/assistant_interaction_controller.cc

@@ -4,6 +4,9 @@
 
 #include "ash/assistant/assistant_interaction_controller.h"
 
+#include <map>
+#include <utility>
+
 #include "ash/assistant/assistant_controller.h"
 #include "ash/assistant/assistant_screen_context_controller.h"
 #include "ash/assistant/assistant_ui_controller.h"
@@ -85,7 +88,8 @@ void AssistantInteractionController::OnAssistantControllerDestroying() {
 void AssistantInteractionController::OnDeepLinkReceived(
     assistant::util::DeepLinkType type,
     const std::map<std::string, std::string>& params) {
-  using namespace assistant::util;
+  using assistant::util::DeepLinkType;
+  using assistant::util::DeepLinkParam;
 
   if (type == DeepLinkType::kWhatsOnMyScreen) {
     StartScreenContextInteraction();
@@ -426,6 +430,7 @@ void AssistantInteractionController::OnTtsStarted(bool due_to_error) {
             l10n_util::GetStringUTF8(IDS_ASH_ASSISTANT_ERROR_GENERIC)));
   }
 
+  model_.pending_response()->set_has_tts(true);
   // We have an agreement with the server that TTS will always be the last part
   // of an interaction to be processed. To be timely in updating UI, we use
   // this as an opportunity to finalize the Assistant response and update the

ash/assistant/model/assistant_response.h

@@ -43,9 +43,15 @@ class AssistantResponse {
   // Returns all suggestions belongs to the response, mapped to a unique id.
   std::map<int, const AssistantSuggestion*> GetSuggestions() const;
 
+  // Get/Set if the current server response has TTS. This can only be reliably
+  // checked after the response is finalized for obvious reasons.
+  bool has_tts() const { return has_tts_; }
+  void set_has_tts(bool has_tts) { has_tts_ = has_tts; }
+
  private:
   std::vector<std::unique_ptr<AssistantUiElement>> ui_elements_;
   std::vector<AssistantSuggestionPtr> suggestions_;
+  bool has_tts_ = false;
 
   DISALLOW_COPY_AND_ASSIGN(AssistantResponse);
 };

ash/assistant/ui/main_stage/ui_element_container_view.cc

@@ -414,8 +414,12 @@ void UiElementContainerView::OnAllUiElementsAdded() {
 
   // Let screen reader read the query result.
   // NOTE: this won't read webview result, which will be triggered with HTML
-  // ARIA.
-  NotifyAccessibilityEvent(ax::mojom::Event::kAlert, true);
+  // ARIA. Also we don't read when there is a TTS response already to avoid
+  // speaking over the server response.
+  const AssistantResponse* response =
+      assistant_controller_->interaction_controller()->model()->response();
+  if (!response->has_tts())
+    NotifyAccessibilityEvent(ax::mojom::Event::kAlert, true);
 }
 
 bool UiElementContainerView::OnAllUiElementsExitAnimationEnded(