Add helper script for running CTS APKs stored in the GS buckets.

android_webview/tools/run_cts.py

Index: android_webview/tools/run_cts.py
diff --git a/android_webview/tools/run_cts.py b/android_webview/tools/run_cts.py
new file mode 100755
index 0000000000000000000000000000000000000000..07fc8cf2f592958b5cf6c28baac15ecd41e9aba3
--- /dev/null
+++ b/android_webview/tools/run_cts.py
@@ -0,0 +1,124 @@
+#!/usr/bin/env python
+#
+# Copyright 2016 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.
+
+"""Runs the CTS test APKs stored in GS."""
+
+import argparse
+import json
+import os
+import shutil
+import sys
+import tempfile
+
+
+sys.path.append(os.path.join(
+    os.path.dirname(__file__), os.pardir, os.pardir, 'build', 'android'))
+import devil_chromium  # pylint: disable=import-error
+from devil.utils import cmd_helper  # pylint: disable=import-error
+
+sys.path.append(os.path.join(
+    os.path.dirname(__file__), os.pardir, os.pardir, 'build'))
+import find_depot_tools  # pylint: disable=import-error
+
+_CTS_BUCKET = 'gs://chromium-cts'
+
+_GSUTIL_PATH = os.path.join(find_depot_tools.DEPOT_TOOLS_PATH, 'gsutil.py')
+_TEST_RUNNER_PATH = os.path.join(
+    os.path.dirname(__file__), os.pardir, os.pardir,
+    'build', 'android', 'test_runner.py')
+
+_EXPECTED_FAILURES_FILE = os.path.join(
+    os.path.dirname(__file__), 'cts_config', 'expected_failure_on_bot.json')
+_WEBVIEW_CTS_GCS_PATH_FILE = os.path.join(
+    os.path.dirname(__file__), 'cts_config', 'webview_cts_gcs_path.json')
+
+
+def GetCtsPath(arch, platform):
+  """Gets relative path the CTS APK is stored at."""
+  with open(_WEBVIEW_CTS_GCS_PATH_FILE) as f:
+    cts_gcs_path_info = json.load(f)
+  try:
+    return cts_gcs_path_info[arch][platform]['apk']
+  except KeyError:
+    raise Exception('No CTS test available for arch:%s, android:%s' %
+                    (arch, platform))
+
+
+def GetExpectedFailures():
+  """Gets list of tests expected to fail."""
+  with open(_EXPECTED_FAILURES_FILE) as f:
+    expected_failures_info = json.load(f)
+  expected_failures = []
+  for class_name, methods in expected_failures_info.iteritems():
+    expected_failures.extend(['%s#%s' % (class_name, m['name'])
+                              for m in methods])
+  return expected_failures
+
+
+def DownloadAndRunCTS(args, test_runner_args):
+  base_cts_dir = None
+  delete_cts_dir = False
+  try:
+    relative_cts_path = GetCtsPath(args.arch, args.platform)
+
+    if args.apk_dir:
+      base_cts_dir = args.apk_dir
+    else:
+      base_cts_dir = tempfile.mkdtemp()
+      delete_cts_dir = True
+
+    local_cts_path = os.path.join(base_cts_dir, relative_cts_path)
+    google_storage_cts_path = '%s/%s' % (_CTS_BUCKET, relative_cts_path)
+
+    # Download CTS APK if needed.
+    if not os.path.exists(local_cts_path):
+      if cmd_helper.RunCmd(
+          [_GSUTIL_PATH, 'cp', google_storage_cts_path, local_cts_path]):
+        raise Exception('Error downloading CTS from Google Storage.')
+
+    test_runner_args += ['--test-apk', local_cts_path]
+    # TODO(mikecase): This doesn't work at all with the

[jbudorick, 2016/12/14 16:22:19]

Oh, I think I understand what you mean now -- if someone calling this script
specifies their own -f/--gtest-filter/etc flag and --skip-expected-failures,
they conflict and only the one that appears last on the command-line is used.

Should this script check that a user only specifies one or the other but not
both?

+    # --gtest-filter test runner option currently. The
+    # filter options will just override eachother.
+    if args.skip_expected_failures:
+      test_runner_args += ['-f=-%s' % ':'.join(GetExpectedFailures())]
+    return cmd_helper.RunCmd(
+        [_TEST_RUNNER_PATH, 'instrumentation'] + test_runner_args)
+  finally:
+    if delete_cts_dir and base_cts_dir:
+      shutil.rmtree(base_cts_dir)
+
+
+def main():
+  parser = argparse.ArgumentParser()
+  parser.add_argument(
+      '--arch',
+      choices=['arm_64'],

[sgurun-gerrit only, 2016/12/14 01:20:58]

does this only work for 64 bits? how about L platform, do we have a L device
there?

[the real yoland, 2016/12/14 01:29:29]

Yes, one can run CTS L with this script. This arch is used for CTS test package,
not meant for device architecture

I think the "arm_64" is based on what we got from Android Build. From android
build, even for L platform, the CTS artifacts are separated into cts_arm_64,
cts_misp, cts_x86_64. And I believe tests apks themselves have no different for
32bits and 64bits

[sgurun-gerrit only, 2016/12/14 18:00:29]

yeah the tests should be in Java all.

+      default='arm_64',
+      help='Arch for CTS tests.')
+  parser.add_argument(
+      '--platform',
+      choices=['L', 'M', 'N'],
+      required=True,
+      help='Android platform version for CTS tests.')
+  parser.add_argument(
+      '--skip-expected-failures',
+      action='store_true',
+      help='Option to skip all tests that are expected to fail.')
+  parser.add_argument(
+      '--apk-dir',
+      help='Directory to load/save CTS APKs. Will try to load CTS APK '
+           'from this directory before downloading from Google Storage '
+           'and will then cache APK here.')
+
+  args, test_runner_args = parser.parse_known_args()
+  devil_chromium.Initialize()
+
+  return DownloadAndRunCTS(args, test_runner_args)
+
+
+if __name__ == '__main__':
+  sys.exit(main())