Add a script to test Chrome with the NVDA screen reader.

tools/accessibility/nvda/nvda_chrome_tests.py

Index: tools/accessibility/nvda/nvda_chrome_tests.py
diff --git a/tools/accessibility/nvda/nvda_chrome_tests.py b/tools/accessibility/nvda/nvda_chrome_tests.py
new file mode 100644
index 0000000000000000000000000000000000000000..61c0820c7916442972f2b83211657f33e8a941d8
--- /dev/null
+++ b/tools/accessibility/nvda/nvda_chrome_tests.py
@@ -0,0 +1,229 @@
+#!/usr/bin/env python
+# Copyright (c) 2014 The Chromium Authors. All rights reserved.

[David Tseng, 2014/07/15 21:34:45]

nit: I think the (c) isn't needed.

[dmazzoni, 2014/07/17 06:52:43]

Done.

+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+"""Semi-automated tests of Chrome with NVDA.
+
+This file performs (semi) automated tests of Chrome with NVDA
+(NonVisual Desktop Access), a popular open-source screen reader for
+visually impaired users on Windows. It works by launching Chrome in a
+subprocess, then launching NVDA in a special environment that simulates
+speech rather than actually speaking, and ignores all events coming from
+processes other than a specific Chrome process ID. Each test automates
+Chrome with a series of actions and asserts that NVDA gives the expected
+feedback in response.
+
+The tests are "semi" automated in the sense that they are not intended to be
+run from any developer machine, or on a buildbot - it requires setting up the
+environment according to the instructions in README.txt, then running the
+test script, then filing bugs for any potential failures. If the environment
+is set up correctly, the actual tests should run automatically and unattended.
+"""
+
+import os
+import pywinauto
+import re
+import shutil
+import signal
+import subprocess
+import sys
+import tempfile
+import time
+import unittest
+
+CHROME_PROFILES_PATH = os.path.join(os.getcwd(), 'chrome_profiles')
+CHROME_PATH = os.path.join(os.environ['USERPROFILE'],
+                           'AppData',
+                           'Local',
+                           'Google',
+                           'Chrome SxS',
+                           'Application',
+                           'chrome.exe')
+NVDA_PATH = os.path.join(os.getcwd(),
+                         'nvdaPortable',
+                         'nvda_noUIAccess.exe')
+NVDA_PROCTEST_PATH = os.path.join(os.getcwd(),
+                                  'nvda-proctest')
+NVDA_LOGPATH = os.path.join(os.getcwd(),
+                            'nvda_log.txt')
+WAIT_FOR_SPEECH_TIMEOUT_SECS = 3.0
+
+class NvdaChromeTest(unittest.TestCase):
+  @classmethod
+  def setUpClass(cls):
+    print 'user data: %s' % CHROME_PROFILES_PATH
+    print 'chrome: %s' % CHROME_PATH
+    print 'nvda: %s' % NVDA_PATH
+    print 'nvda_proctest: %s' % NVDA_PROCTEST_PATH
+
+    print
+    print 'Clearing user data directory and log file from previous runs'
+    if os.access(NVDA_LOGPATH, os.F_OK):
+      os.remove(NVDA_LOGPATH)
+    if os.access(CHROME_PROFILES_PATH, os.F_OK):
+      shutil.rmtree(CHROME_PROFILES_PATH)
+    os.mkdir(CHROME_PROFILES_PATH, 0777)
+
+    def handler(signum, frame):
+      print 'Test interrupted, attempting to kill subprocesses.'
+      self.tearDown()
+      sys.exit()
+    signal.signal(signal.SIGINT, handler)
+
+  def setUp(self):
+    user_data_dir = tempfile.mkdtemp(dir = CHROME_PROFILES_PATH)
+    args = [CHROME_PATH,
+            '--user-data-dir=%s' % user_data_dir,
+            '--no-first-run',
+            'about:blank']
+    print
+    print ' '.join(args)
+    self._chrome_proc = subprocess.Popen(args)
+    self._chrome_proc.poll()
+    if self._chrome_proc.returncode is None:
+      print 'Chrome is running'
+    else:
+      print 'Chrome exited with code', self._chrome_proc.returncode
+      sys.exit()
+    print 'Chrome pid: %d' % self._chrome_proc.pid
+
+    os.environ['NVDA_SPECIFIC_PROCESS'] = str(self._chrome_proc.pid)
+
+    args = [NVDA_PATH,
+            '-m',
+            '-c',
+            NVDA_PROCTEST_PATH,
+            '-f',
+            NVDA_LOGPATH]
+    self._nvda_proc = subprocess.Popen(args)
+    self._nvda_proc.poll()
+    if self._nvda_proc.returncode is None:
+      print 'NVDA is running'
+    else:
+      print 'NVDA exited with code', self._nvda_proc.returncode
+      sys.exit()
+    print 'NVDA pid: %d' % self._nvda_proc.pid
+
+    app = pywinauto.application.Application()
+    app.connect_(process = self._chrome_proc.pid)
+    self._pywinauto_window = app.top_window_()
+
+    try:
+      self._WaitForSpeech(['Address and search bar edit', 'about:blank'])
+    except:
+      self.tearDown()
+
+  def tearDown(self):
+    print
+    print 'Shutting down'
+
+    self._chrome_proc.poll()
+    if self._chrome_proc.returncode is None:
+      print 'Killing Chrome subprocess'
+      self._chrome_proc.kill()
+    else:
+      print 'Chrome already died.'
+
+    self._nvda_proc.poll()
+    if self._nvda_proc.returncode is None:
+      print 'Killing NVDA subprocess'
+      self._nvda_proc.kill()
+    else:
+      print 'NVDA already died.'
+
+  def _GetSpeechFromNvdaLogFile(self):
+    """Return everything NVDA would have spoken as a list of strings.
+
+    Parses lines like this from NVDA's log file:
+      Speaking [LangChangeCommand ('en'), u'Google Chrome', u'window']
+      Speaking character u'slash'
+
+    Returns a single list of strings like this:
+      [u'Google Chrome', u'window', u'slash']
+    """
+    if not os.access(NVDA_LOGPATH, os.F_OK):
+      return []
+    lines = open(NVDA_LOGPATH).readlines()
+    regex = re.compile(r"u'((?:[^\'\\]|\\.)*)\'")
+    result = []
+    for line in lines:
+      for m in regex.finditer(line):
+        speech_with_whitespace = m.group(1)
+        speech_stripped = re.sub(r'\s+', ' ', speech_with_whitespace).strip()
+        result.append(speech_stripped)
+    return result

[David Tseng, 2014/07/15 21:34:45]

Do you want/care to clear the speech log? (so that you don't end up parsing the
entire contents every time you wait).

[dmazzoni, 2014/07/17 06:52:43]

I'm not sure how to do this easily without potentially breaking NVDA, since it
has the file open for writing. I could save the byte number, I guess.

In practice, each test is pretty small and this takes negligible time, so I'm
not going to worry about it for now. We can optimize it later when we have a
longer test.

+
+  def _WaitForSpeech(self, expected):
+    """Block until the last speech in NVDA's log file is the given string(s).
+
+    Repeatedly parses the log file until the last speech line(s) in the
+    log file match the given strings, or it times out.
+
+    Args:
+      expected: string or a list of string - only succeeds if these are the last
+        strings spoken, in order.
+
+    Returns when those strings are spoken, or throws an error if it times out
+      waiting for those strings.
+    """
+    if type(expected) is type(''):
+      expected = [expected]
+    start_time = time.time()
+    while True:
+      lines = self._GetSpeechFromNvdaLogFile()
+      if (lines[-len(expected):] == expected):

[David Tseng, 2014/07/15 21:34:45]

What if some other speech slips in right after your expected text?

[dmazzoni, 2014/07/17 06:52:43]

Currently these tests really want to assert that it was the last thing spoken,
otherwise they could get confused when the same utterance repeats more than
once. However, if we encounter flakiness due to unrelated speech, we can address
that later. So far that hasn't happened.

+        break
+
+      if time.time() - start_time >= WAIT_FOR_SPEECH_TIMEOUT_SECS:
+        print '** Speech from NVDA so far:'
+        for line in lines:
+          print '"%s"' % line
+        print '** Was waiting for:'
+        for line in expected:
+          print '"%s"' % line
+        raise Exception('Timed out')
+      time.sleep(0.1)
+
+  #
+  # Tests
+  #
+
+  def testTypingInOmnibox(self):
+    # Ctrl+A: Select all.
+    self._pywinauto_window.TypeKeys('^A')
+    self._WaitForSpeech('selecting about:blank')
+
+    # Type three characters.
+    self._pywinauto_window.TypeKeys('xyz')
+    self._WaitForSpeech(['x', 'y', 'z'])
+
+    # Arrow back over two characters.
+    self._pywinauto_window.TypeKeys('{LEFT}')
+    self._WaitForSpeech(['z', 'z', 'unselecting'])
+
+    self._pywinauto_window.TypeKeys('{LEFT}')
+    self._WaitForSpeech('y')
+
+  def testFocusToolbarButton(self):
+    # Alt+Shift+T.
+    self._pywinauto_window.TypeKeys('%+T')
+    self._WaitForSpeech('Reload button Reload this page')
+
+  def testReadAllOnPageLoad(self):
+    # Ctrl+A: Select all
+    self._pywinauto_window.TypeKeys('^A')
+    self._WaitForSpeech('selecting about:blank')
+
+    # Load data url.
+    self._pywinauto_window.TypeKeys('data:text/html,Hello<p>World.')
+    self._WaitForSpeech('dot')
+    self._pywinauto_window.TypeKeys('{ENTER}')
+    self._WaitForSpeech(
+        ['document',
+         'Hello',
+         'World.'])
+
+if __name__ == '__main__':
+  unittest.main()
+