Revert "Remove the blocking apis in SpeechMonitor"

This reverts commit 4e3c8ff8.

Reason for revert: ChromeOS bot failures

Bug: chromium:1065746

Original change's description:
> Remove the blocking apis in SpeechMonitor
> 
> - convert all callers to use the async api
> - some misc cleanup in SpokenFeedbackBrowserTest
> 
> Run these against debug and release; use both 5 and 10 parallel jobs which should load the tests enough to catch flakes.
> 
> Test: browser_tests against all relevant suites (SelectToSpeak*, Spoken*, and *.EnableChromeVox).
> Change-Id: I1b425b8640433b7608d84aa67f1cf5900ce26638
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2124701
> Reviewed-by: Nancy Wang <nancylingwang@chromium.org>
> Reviewed-by: Dominic Mazzoni <dmazzoni@chromium.org>
> Commit-Queue: David Tseng <dtseng@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#754319}

TBR=dmazzoni@chromium.org,dtseng@chromium.org,nancylingwang@chromium.org,nancylingwang@google.com

# Not skipping CQ checks because original CL landed > 1 day ago.

Change-Id: I121b1d503547f3fc54597527ecbdb95ee65b9511
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2126328Reviewed-by: Johann Koenig <johannkoenig@google.com>
Commit-Queue: Johann Koenig <johannkoenig@google.com>
Cr-Commit-Position: refs/heads/master@{#754374}

chrome/browser/chromeos/accessibility/select_to_speak_browsertest.cc

@@ -16,6 +16,7 @@
 #include "base/bind.h"
 #include "base/command_line.h"
 #include "base/memory/weak_ptr.h"
+#include "base/strings/pattern.h"
 #include "chrome/browser/chromeos/accessibility/accessibility_manager.h"
 #include "chrome/browser/chromeos/accessibility/speech_monitor.h"
 #include "chrome/browser/profiles/profile.h"
@@ -73,7 +74,7 @@ class SelectToSpeakTest : public InProcessBrowserTest {
     ui_test_utils::NavigateToURL(browser(), GURL(url::kAboutBlankURL));
   }
 
-  SpeechMonitor sm_;
+  SpeechMonitor speech_monitor_;
   std::unique_ptr<ui::test::EventGenerator> generator_;
 
   gfx::Rect GetWebContentsBounds() const {
@@ -139,8 +140,8 @@ class SelectToSpeakTest : public InProcessBrowserTest {
     return browser()->tab_strip_model()->GetActiveWebContents();
   }
 
-  void ExecuteJavaScriptAsync(const std::string& script) {
-    content::ExecuteScriptAsync(GetWebContents(), script);
+  void ExecuteJavaScriptInForeground(const std::string& script) {
+    CHECK(content::ExecuteScript(GetWebContents(), script));
   }
 
  private:
@@ -175,8 +176,8 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SpeakStatusTray) {
   generator_->ReleaseLeftButton();
   generator_->ReleaseKey(ui::VKEY_LWIN, 0 /* flags */);
 
-  sm_.ExpectSpeechPattern("Status tray*");
-  sm_.Replay();
+  EXPECT_TRUE(
+      base::MatchPattern(speech_monitor_.GetNextUtterance(), "Status tray*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, ActivatesWithTapOnSelectToSpeakTray) {
@@ -198,8 +199,8 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, ActivatesWithTapOnSelectToSpeakTray) {
                           bounds.y() + bounds.height());
   generator_->ReleaseLeftButton();
 
-  sm_.ExpectSpeechPattern("This is some text*");
-  sm_.Replay();
+  EXPECT_TRUE(base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                                 "This is some text*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SelectToSpeakTrayNotSpoken) {
@@ -217,9 +218,8 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SelectToSpeakTrayNotSpoken) {
   // The next should be the first thing spoken -- the tray was not spoken.
   ActivateSelectToSpeakInWindowBounds(
       "data:text/html;charset=utf-8,<p>This is some text</p>");
-
-  sm_.ExpectSpeechPattern("This is some text*");
-  sm_.Replay();
+  EXPECT_TRUE(base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                                 "This is some text*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SmoothlyReadsAcrossInlineUrl) {
@@ -230,8 +230,9 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SmoothlyReadsAcrossInlineUrl) {
   // Should combine nodes in a paragraph into one utterance.
   // Includes some wildcards between words because there may be extra
   // spaces. Spaces are not pronounced, so extra spaces do not impact output.
-  sm_.ExpectSpeechPattern("This is some text*with a node*in the middle*");
-  sm_.Replay();
+  EXPECT_TRUE(
+      base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                         "This is some text*with a node*in the middle*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SmoothlyReadsAcrossMultipleLines) {
@@ -244,8 +245,9 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SmoothlyReadsAcrossMultipleLines) {
   // spaces, for example at line wraps. Extra wildcards included to
   // reduce flakyness in case wrapping is not consistent.
   // Spaces are not pronounced, so extra spaces do not impact output.
-  sm_.ExpectSpeechPattern("This is some*text*with*a*node*in*the*middle*");
-  sm_.Replay();
+  EXPECT_TRUE(
+      base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                         "This is some*text*with*a*node*in*the*middle*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SmoothlyReadsAcrossFormattedText) {
@@ -257,8 +259,9 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, SmoothlyReadsAcrossFormattedText) {
   // Should combine nodes in a paragraph into one utterance.
   // Includes some wildcards between words because there may be extra
   // spaces. Spaces are not pronounced, so extra spaces do not impact output.
-  sm_.ExpectSpeechPattern("This is some text*with a node*in the middle*");
-  sm_.Replay();
+  EXPECT_TRUE(
+      base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                         "This is some text*with a node*in the middle*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest,
@@ -266,9 +269,8 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest,
   // Bold or formatted text
   ActivateSelectToSpeakInWindowBounds(
       "data:text/html;charset=utf-8,<canvas>This is some text</canvas>");
-
-  sm_.ExpectSpeechPattern("This is some text*");
-  sm_.Replay();
+  EXPECT_TRUE(base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                                 "This is some text*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, BreaksAtParagraphBounds) {
@@ -277,9 +279,10 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, BreaksAtParagraphBounds) {
       "<p>Second paragraph</p></div>");
 
   // Should keep each paragraph as its own utterance.
-  sm_.ExpectSpeechPattern("First paragraph*");
-  sm_.ExpectSpeechPattern("Second paragraph*");
-  sm_.Replay();
+  EXPECT_TRUE(base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                                 "First paragraph*"));
+  EXPECT_TRUE(base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                                 "Second paragraph*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, LanguageBoundsIgnoredByDefault) {
@@ -290,8 +293,9 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, LanguageBoundsIgnoredByDefault) {
       "<span lang='en-US'>The first paragraph</span>"
       "<span lang='fr-FR'>la deuxième paragraphe</span></div>");
 
-  sm_.ExpectSpeechPattern("The first paragraph* la deuxième paragraphe*");
-  sm_.Replay();
+  EXPECT_TRUE(
+      base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                         "The first paragraph* la deuxième paragraphe*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTestWithLanguageDetection,
@@ -301,9 +305,15 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTestWithLanguageDetection,
       "<span lang='en-US'>The first paragraph</span>"
       "<span lang='fr-FR'>la deuxième paragraphe</span></div>");
 
-  sm_.ExpectSpeechPatternWithLocale("The first paragraph*", "en-US");
-  sm_.ExpectSpeechPatternWithLocale("la deuxième paragraphe*", "fr-FR");
-  sm_.Replay();
+  SpeechMonitorUtterance result1 =
+      speech_monitor_.GetNextUtteranceWithLanguage();
+  EXPECT_TRUE(base::MatchPattern(result1.text, "The first paragraph*"));
+  EXPECT_EQ("en-US", result1.lang);
+
+  SpeechMonitorUtterance result2 =
+      speech_monitor_.GetNextUtteranceWithLanguage();
+  EXPECT_TRUE(base::MatchPattern(result2.text, "la deuxième paragraphe*"));
+  EXPECT_EQ("fr-FR", result2.lang);
 }
 
 // Flaky test. https://crbug.com/950049
@@ -389,16 +399,15 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, ContinuesReadingDuringResize) {
       "<p>Second paragraph is longer than 300 pixels and will wrap when "
       "resized</p></div>");
 
-  sm_.ExpectSpeechPattern("First paragraph*");
+  EXPECT_TRUE(base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                                 "First paragraph*"));
 
   // Resize before second is spoken. If resizing caused errors finding the
   // inlineTextBoxes in the node, speech would be stopped early.
-  sm_.Call([this]() {
-    ExecuteJavaScriptAsync(
-        "document.getElementById('resize').style.width='100px'");
-  });
-  sm_.ExpectSpeechPattern("*when*resized*");
-  sm_.Replay();
+  ExecuteJavaScriptInForeground(
+      "document.getElementById('resize').style.width='100px'");
+  EXPECT_TRUE(
+      base::MatchPattern(speech_monitor_.GetNextUtterance(), "*when*resized*"));
 }
 
 IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, WorksWithStickyKeys) {
@@ -420,12 +429,11 @@ IN_PROC_BROWSER_TEST_F(SelectToSpeakTest, WorksWithStickyKeys) {
                           bounds.y() + bounds.height());
   generator_->ReleaseLeftButton();
 
-  sm_.ExpectSpeechPattern("This is some text*");
+  EXPECT_TRUE(base::MatchPattern(speech_monitor_.GetNextUtterance(),
+                                 "This is some text*"));
 
   // Reset state.
-  sm_.Call([]() { AccessibilityManager::Get()->EnableStickyKeys(false); });
-
-  sm_.Replay();
+  AccessibilityManager::Get()->EnableStickyKeys(false);
 }
 
 }  // namespace chromeos

chrome/browser/chromeos/accessibility/speech_monitor.cc

@@ -12,9 +12,12 @@
 namespace chromeos {
 
 namespace {
+const char kChromeVoxEnabledMessage[] = "ChromeVox spoken feedback is ready";
+const char kChromeVoxAlertMessage[] = "Alert";
+const char kChromeVoxUpdate1[] = "chrome vox Updated Press chrome vox o,";
+const char kChromeVoxUpdate2[] = "n to learn more about chrome vox Next.";
 
 constexpr int kPrintExpectationDelayMs = 3000;
-
 }  // namespace
 
 SpeechMonitor::SpeechMonitor() {
@@ -28,6 +31,52 @@ SpeechMonitor::~SpeechMonitor() {
     CHECK(replay_called_) << "Expectation was made, but Replay() not called.";
 }
 
+std::string SpeechMonitor::GetNextUtterance() {
+  return GetNextUtteranceWithLanguage().text;
+}
+
+SpeechMonitorUtterance SpeechMonitor::GetNextUtteranceWithLanguage() {
+  if (utterance_queue_.empty()) {
+    loop_runner_ = new content::MessageLoopRunner();
+    loop_runner_->Run();
+    loop_runner_.reset();
+  }
+  SpeechMonitorUtterance result = utterance_queue_.front();
+  utterance_queue_.pop_front();
+  return result;
+}
+
+bool SpeechMonitor::SkipChromeVoxEnabledMessage() {
+  return SkipChromeVoxMessage(kChromeVoxEnabledMessage);
+}
+
+bool SpeechMonitor::DidStop() {
+  return did_stop_;
+}
+
+void SpeechMonitor::BlockUntilStop() {
+  if (!did_stop_) {
+    loop_runner_ = new content::MessageLoopRunner();
+    loop_runner_->Run();
+    loop_runner_.reset();
+  }
+}
+
+bool SpeechMonitor::SkipChromeVoxMessage(const std::string& message) {
+  while (true) {
+    if (utterance_queue_.empty()) {
+      loop_runner_ = new content::MessageLoopRunner();
+      loop_runner_->Run();
+      loop_runner_.reset();
+    }
+    SpeechMonitorUtterance result = utterance_queue_.front();
+    utterance_queue_.pop_front();
+    if (result.text == message)
+      return true;
+  }
+  return false;
+}
+
 bool SpeechMonitor::PlatformImplAvailable() {
   return true;
 }
@@ -46,6 +95,7 @@ void SpeechMonitor::Speak(int utterance_id,
 }
 
 bool SpeechMonitor::StopSpeaking() {
+  did_stop_ = true;
   return true;
 }
 
@@ -64,8 +114,21 @@ void SpeechMonitor::GetVoices(std::vector<content::VoiceData>* out_voices) {
 void SpeechMonitor::WillSpeakUtteranceWithVoice(
     content::TtsUtterance* utterance,
     const content::VoiceData& voice_data) {
+  // Blacklist some phrases.
+  // Filter out empty utterances which can be used to trigger a start event from
+  // tts as an earcon sync.
+  if (utterance->GetText() == "" ||
+      utterance->GetText() == kChromeVoxAlertMessage ||
+      utterance->GetText() == kChromeVoxUpdate1 ||
+      utterance->GetText() == kChromeVoxUpdate2)
+    return;
+
+  VLOG(0) << "Speaking " << utterance->GetText();
   utterance_queue_.emplace_back(utterance->GetText(), utterance->GetLang());
   delay_for_last_utterance_ms_ = CalculateUtteranceDelayMS();
+  if (loop_runner_.get())
+    loop_runner_->Quit();
+
   MaybeContinueReplay();
 }
 
@@ -120,19 +183,11 @@ void SpeechMonitor::ExpectSpeech(const std::string& text,
 
 void SpeechMonitor::ExpectSpeechPattern(const std::string& pattern,
                                         const base::Location& location) {
-  ExpectSpeechPatternWithLocale(pattern, "", location);
-}
-
-void SpeechMonitor::ExpectSpeechPatternWithLocale(
-    const std::string& pattern,
-    const std::string& locale,
-    const base::Location& location) {
   CHECK(!replay_loop_runner_.get());
-  replay_queue_.push_back({[this, pattern, locale]() {
+  replay_queue_.push_back({[this, pattern]() {
                              for (auto it = utterance_queue_.begin();
                                   it != utterance_queue_.end(); it++) {
-                               if (base::MatchPattern(it->text, pattern) &&
-                                   (locale.empty() || it->lang == locale)) {
+                               if (base::MatchPattern(it->text, pattern)) {
                                  // Erase all utterances that came before the
                                  // match as well as the match itself.
                                  utterance_queue_.erase(

chrome/browser/chromeos/accessibility/speech_monitor.h

@@ -25,13 +25,35 @@ struct SpeechMonitorUtterance {
 };
 
 // For testing purpose installs itself as the platform speech synthesis engine,
-// allowing it to intercept all speech calls. Provides an api to make
-// asynchronous function calls and expectations about resulting speech.
+// allowing it to intercept all speech calls, and then provides a method to
+// block until the next utterance is spoken.
 class SpeechMonitor : public content::TtsPlatform {
  public:
   SpeechMonitor();
   virtual ~SpeechMonitor();
 
+  // Blocing api.
+  // Use the following apis to write a synchronous test e.g.
+  // DoSomething();
+  // EXPECT_EQ("foo", speech_monitor_.GetNextUtterance());
+
+  // Blocks until the next utterance is spoken, and returns its text.
+  std::string GetNextUtterance();
+  // Blocks until the next utterance is spoken, and returns its text.
+  SpeechMonitorUtterance GetNextUtteranceWithLanguage();
+
+  // Wait for next utterance and return true if next utterance is ChromeVox
+  // enabled message.
+  bool SkipChromeVoxEnabledMessage();
+  bool SkipChromeVoxMessage(const std::string& message);
+
+  // Returns true if StopSpeaking() was called on TtsController.
+  bool DidStop();
+
+  // Blocks until StopSpeaking() is called on TtsController.
+  void BlockUntilStop();
+
+  // Non-blocking api.
   // Use these apis if you want to write an async test e.g.
   // sm_.ExpectSpeech("foo");
   // sm_.Call([this]() { DoSomething(); })
@@ -45,10 +67,6 @@ class SpeechMonitor : public content::TtsPlatform {
                     const base::Location& location = FROM_HERE);
   void ExpectSpeechPattern(const std::string& pattern,
                            const base::Location& location = FROM_HERE);
-  void ExpectSpeechPatternWithLocale(
-      const std::string& pattern,
-      const std::string& locale,
-      const base::Location& location = FROM_HERE);
   void ExpectNextSpeechIsNot(const std::string& text,
                              const base::Location& location = FROM_HERE);
   void ExpectNextSpeechIsNotPattern(const std::string& pattern,
@@ -93,17 +111,17 @@ class SpeechMonitor : public content::TtsPlatform {
   void MaybeContinueReplay();
   void MaybePrintExpectations();
 
+  scoped_refptr<content::MessageLoopRunner> loop_runner_;
   // Our list of utterances and specified language.
   base::circular_deque<SpeechMonitorUtterance> utterance_queue_;
-
+  bool did_stop_ = false;
   std::string error_;
 
+  // Delayed utterances.
   // Calculates the milliseconds elapsed since the last call to Speak().
   double CalculateUtteranceDelayMS();
-
   // Stores the milliseconds elapsed since the last call to Speak().
   double delay_for_last_utterance_ms_;
-
   // Stores the last time Speak() was called.
   std::chrono::steady_clock::time_point time_of_last_utterance_;
 

chrome/browser/chromeos/accessibility/spoken_feedback_app_list_browsertest.cc

@@ -8,12 +8,15 @@
 #include "ash/app_list/test/test_search_result.h"
 #include "ash/app_list/views/app_list_view.h"
 #include "ash/shell.h"
+#include "base/strings/pattern.h"
 #include "base/strings/utf_string_conversions.h"
 #include "chrome/browser/chromeos/accessibility/spoken_feedback_browsertest.h"
+#include "chrome/browser/profiles/profile_manager.h"
 #include "chrome/browser/ui/app_list/app_list_client_impl.h"
 #include "chrome/common/chrome_switches.h"
 #include "chrome/grit/generated_resources.h"
 #include "chromeos/constants/chromeos_switches.h"
+#include "components/account_id/account_id.h"
 #include "components/user_manager/user_names.h"
 #include "testing/gtest/include/gtest/gtest.h"
 #include "ui/base/l10n/l10n_util.h"
@@ -101,8 +104,8 @@ INSTANTIATE_TEST_SUITE_P(TestAsNormalAndGuestUser,
 IN_PROC_BROWSER_TEST_P(SpokenFeedbackAppListTest, LauncherStateTransition) {
   EnableChromeVox();
 
-  sm_.Call(
-      [this]() { EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF)); });
+  EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF));
+
   sm_.ExpectSpeechPattern("Launcher");
   sm_.ExpectSpeech("Button");
   sm_.ExpectSpeech("Shelf");
@@ -135,8 +138,8 @@ IN_PROC_BROWSER_TEST_P(SpokenFeedbackAppListTest,
                        DisabledFullscreenExpandButton) {
   EnableChromeVox();
 
-  sm_.Call(
-      [this]() { EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF)); });
+  EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF));
+
   sm_.ExpectSpeech("Shelf");
 
   // Press space on the launcher button in shelf, this opens peeking launcher.
@@ -172,8 +175,8 @@ IN_PROC_BROWSER_TEST_P(SpokenFeedbackAppListTest,
 
   EnableChromeVox();
 
-  sm_.Call(
-      [this]() { EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF)); });
+  EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF));
+
   sm_.ExpectSpeech("Press Search plus Space to activate");
   // Press space on the launcher button in shelf, this opens peeking
   // launcher.
@@ -209,8 +212,8 @@ IN_PROC_BROWSER_TEST_P(SpokenFeedbackAppListTest,
 
   EnableChromeVox();
 
-  sm_.Call(
-      [this]() { EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF)); });
+  EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF));
+
   sm_.ExpectSpeech("Press Search plus Space to activate");
   // Press space on the launcher button in shelf, this opens peeking
   // launcher.
@@ -254,12 +257,10 @@ IN_PROC_BROWSER_TEST_P(SpokenFeedbackAppListTest,
 IN_PROC_BROWSER_TEST_P(SpokenFeedbackAppListTest, NavigateAppLauncher) {
   EnableChromeVox();
 
-  sm_.Call([this]() {
-    // Add one app to the applist.
-    PopulateApps(1);
+  // Add one app to the applist.
+  PopulateApps(1);
 
-    EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF));
-  });
+  EXPECT_TRUE(PerformAcceleratorAction(ash::FOCUS_SHELF));
 
   // Wait for it to say "Launcher", "Button", "Shelf", "Tool bar".
   sm_.ExpectSpeechPattern("Launcher");

chrome/browser/ui/ash/launcher/chrome_launcher_controller_browsertest.cc

@@ -2333,46 +2333,48 @@ IN_PROC_BROWSER_TEST_F(HotseatShelfAppBrowserTest, EnableChromeVox) {
   chromeos::SpeechMonitor speech_monitor;
 
   // Enable ChromeVox.
+  {
     ASSERT_FALSE(
         chromeos::AccessibilityManager::Get()->IsSpokenFeedbackEnabled());
     chromeos::AccessibilityManager::Get()->EnableSpokenFeedback(true);
+    EXPECT_TRUE(speech_monitor.SkipChromeVoxEnabledMessage());
+
+    // Disable earcons (https://crbug.com/396507).
+    const std::string script("ChromeVox.earcons.playEarcon = function() {};");
+    extensions::ExtensionHost* host =
+        extensions::ProcessManager::Get(browser()->profile())
+            ->GetBackgroundHostForExtension(
+                extension_misc::kChromeVoxExtensionId);
+    ASSERT_TRUE(content::ExecuteScript(host->host_contents(), script));
+  }
 
   ash::RootWindowController* controller =
       ash::Shell::GetRootWindowControllerWithDisplayId(
           display::Screen::GetScreen()->GetPrimaryDisplay().id());
+
+  // Gesture tap at the home button.
   views::View* home_button = ash::ShelfTestApi().GetHomeButton();
   ui::test::EventGenerator event_generator(controller->GetRootWindow());
-  auto* generator_ptr = &event_generator;
-
-  // Wait for ChromeVox to start reading anything.
-  speech_monitor.ExpectSpeechPattern("*");
-  speech_monitor.Call([generator_ptr, home_button]() {
-    // Gesture tap at the home button.
-    generator_ptr->GestureTapAt(home_button->GetBoundsInScreen().CenterPoint());
-  });
-  speech_monitor.ExpectSpeech("Launcher");
-  speech_monitor.ExpectSpeech("Button");
-  speech_monitor.ExpectSpeech("Shelf");
-  speech_monitor.ExpectSpeech("Tool bar");
-  speech_monitor.ExpectSpeech(", window");
-
-  speech_monitor.Call([controller]() {
-    // Hotseat is expected to be extended if spoken feedback is enabled.
-    ASSERT_EQ(ash::HotseatState::kExtended,
-              controller->shelf()->shelf_layout_manager()->hotseat_state());
-  });
-
-  speech_monitor.Call([generator_ptr]() {
-    // Press the search + right. Expects that the browser icon receives the
-    // accessibility focus and the hotseat remains in kExtended state.
-    generator_ptr->PressKey(ui::VKEY_RIGHT, ui::EF_COMMAND_DOWN);
-  });
+  event_generator.GestureTapAt(home_button->GetBoundsInScreen().CenterPoint());
+
+  ASSERT_EQ("Launcher", speech_monitor.GetNextUtterance());
+  ASSERT_EQ("Button", speech_monitor.GetNextUtterance());
+  ASSERT_EQ("Shelf", speech_monitor.GetNextUtterance());
+  ASSERT_EQ("Tool bar", speech_monitor.GetNextUtterance());
+  ASSERT_EQ(", window", speech_monitor.GetNextUtterance());
+
+  // Hotseat is expected to be extended if spoken feedback is enabled.
+  ASSERT_EQ(ash::HotseatState::kExtended,
+            controller->shelf()->shelf_layout_manager()->hotseat_state());
+
+  // Press the search + right. Expects that the browser icon receives the
+  // accessibility focus and the hotseat remains in kExtended state.
+  event_generator.PressKey(ui::VKEY_RIGHT, ui::EF_COMMAND_DOWN);
   const int browser_index =
       ash::ShelfModel::Get()->GetItemIndexForType(ash::TYPE_BROWSER_SHORTCUT);
-  speech_monitor.ExpectSpeech(
-      base::UTF16ToASCII(ash::ShelfModel::Get()->items()[browser_index].title));
-  speech_monitor.Replay();
-
+  EXPECT_EQ(
+      base::UTF16ToASCII(ash::ShelfModel::Get()->items()[browser_index].title),
+      speech_monitor.GetNextUtterance());
   EXPECT_EQ(ash::HotseatState::kExtended,
             controller->shelf()->shelf_layout_manager()->hotseat_state());