Repeat Failed Tests in Serial

Before considering a test failed in run_test_cases.py try running the failed tests serial, since they may have only failed before because they were conflicting with other tests running at the same time.


BUG=


Review URL: https://chromiumcodereview.appspot.com/10831330

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@151893 0039d316-1c4b-4281-b951-d872f2087c98

2 files changed, 99 insertions, 21 deletions

diff --git a/tools/isolate/run_test_cases.py b/tools/isolate/run_test_cases.py
index 9beccc8..3200c02 100755
--- a/tools/isolate/run_test_cases.py
+++ b/tools/isolate/run_test_cases.py
@@ -454,13 +454,13 @@ def list_test_cases(executable, index, shards, disabled, fails, flaky):
 
 
 class Runner(object):
-  def __init__(self, executable, cwd_dir, timeout, progress):
+  def __init__(self, executable, cwd_dir, timeout, progress, retry_count=3):
     # Constants
     self.executable = executable
     self.cwd_dir = cwd_dir
     self.timeout = timeout
     self.progress = progress
-    self.retry_count = 3
+    self.retry_count = retry_count
     # It is important to remove the shard environment variables since it could
     # conflict with --gtest_filter.
     self.env = os.environ.copy()
@@ -546,6 +546,14 @@ def get_test_cases(executable, whitelist, blacklist, index, shards):
   return tests
 
 
+def LogResults(result_file, results):
+  """Write the results out to a file if one is given."""
+  if not result_file:
+    return
+  with open(result_file, 'wb') as f:
+    json.dump(results, f, sort_keys=True, indent=2)
+
+
 def run_test_cases(executable, test_cases, jobs, timeout, result_file):
   """Traces test cases one by one."""
   progress = Progress(len(test_cases))
@@ -556,9 +564,7 @@ def run_test_cases(executable, test_cases, jobs, timeout, result_file):
     results = pool.join(progress, 0.1)
     duration = time.time() - progress.start
   results = dict((item[0]['test_case'], item) for item in results)
-  if result_file:
-    with open(result_file, 'wb') as f:
-      json.dump(results, f, sort_keys=True, indent=2)
+  LogResults(result_file, results)
   sys.stdout.write('\n')
   total = len(results)
   if not total:
@@ -581,10 +587,34 @@ def run_test_cases(executable, test_cases, jobs, timeout, result_file):
     else:
       assert False, items
 
+  # Retry all the failures serially to see if they are just flaky when
+  # run at the same time.
+  if fail:
+    print 'Retrying failed tests serially.'
+    progress = Progress(len(fail))
+    function = Runner(
+        executable, os.getcwd(), timeout, progress, retry_count=1).map
+    test_cases_retry = fail[:]
+
+    for test_case in test_cases_retry:
+      output = function(test_case)
+      progress.print_update()
+      results[output[0]['test_case']].append(output)
+      if not output[0]['returncode']:
+        fail.remove(test_case)
+        flaky.append(test_case)
+
+    LogResults(result_file, results)
+    sys.stdout.write('\n')
+
+  print 'Summary:'
   for test_case in sorted(flaky):
     items = results[test_case]
     print '%s is flaky (tried %d times)' % (test_case, len(items))
 
+  for test_case in sorted(fail):
+    print '%s failed' % (test_case)
+
   print 'Success: %4d %5.2f%%' % (len(success), len(success) * 100. / total)
   print 'Flaky:   %4d %5.2f%%' % (len(flaky), len(flaky) * 100. / total)
   print 'Fail:    %4d %5.2f%%' % (len(fail), len(fail) * 100. / total)
diff --git a/tools/isolate/run_test_cases_smoke_test.py b/tools/isolate/run_test_cases_smoke_test.py
index 62ba92ef..2174468 100755
--- a/tools/isolate/run_test_cases_smoke_test.py
+++ b/tools/isolate/run_test_cases_smoke_test.py
@@ -3,6 +3,7 @@
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
+import json
 import logging
 import os
 import re
@@ -16,14 +17,19 @@ sys.path.append(os.path.join(ROOT_DIR, 'data', 'gtest_fake'))
 import gtest_fake_base
 
 
-def RunTest(test_file):
+def RunTest(test_file, dump_file=None):
   target = os.path.join(ROOT_DIR, 'data', 'gtest_fake', test_file)
   cmd = [
       sys.executable,
       os.path.join(ROOT_DIR, 'run_test_cases.py'),
-      '--no-dump',
-      target,
   ]
+
+  if dump_file:
+    cmd.extend(['--result', dump_file])
+  else:
+    cmd.append('--no-dump')
+
+  cmd.append(target)
   logging.debug(' '.join(cmd))
   proc = subprocess.Popen(
       cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
@@ -39,6 +45,12 @@ class TraceTestCases(unittest.TestCase):
     os.environ.pop('GTEST_SHARD_INDEX', '')
     os.environ.pop('GTEST_TOTAL_SHARDS', '')
 
+    self.filename = 'test.results'
+
+  def tearDown(self):
+    if os.path.exists(self.filename):
+      os.remove(self.filename)
+
   def _check_results(self, expected_out_re, out, err):
     if sys.platform == 'win32':
       out = out.replace('\r\n', '\n')
@@ -52,8 +64,20 @@ class TraceTestCases(unittest.TestCase):
     self.assertEquals([], lines)
     self.assertEquals('', err)
 
+  def _check_results_file(self, expected_file_contents_entries):
+    self.assertTrue(os.path.exists(self.filename))
+
+    with open(self.filename) as f:
+      file_contents = json.load(f)
+
+    self.assertEqual(len(expected_file_contents_entries), len(file_contents))
+    for (entry_name, entry_count) in expected_file_contents_entries:
+      self.assertTrue(entry_name in file_contents)
+      self.assertEqual(entry_count, len(file_contents[entry_name]))
+
   def test_simple_pass(self):
-    out, err, return_code = RunTest('gtest_fake_pass.py')
+    out, err, return_code = RunTest('gtest_fake_pass.py',
+                                    dump_file=self.filename)
 
     self.assertEquals(0, return_code)
 
@@ -61,37 +85,53 @@ class TraceTestCases(unittest.TestCase):
       r'\[\d/\d\]   \d\.\d\ds .+',
       r'\[\d/\d\]   \d\.\d\ds .+',
       r'\[\d/\d\]   \d\.\d\ds .+',
+      re.escape('Summary:'),
       re.escape('Success:    3 100.00%'),
       re.escape('Flaky:      0  0.00%'),
       re.escape('Fail:       0  0.00%'),
       r'\d+\.\ds Done running 3 tests with 3 executions. \d+\.\d test/s',
     ]
-
     self._check_results(expected_out_re, out, err)
 
+    expected_result_file_entries = [
+        ('Foo.Bar1', 1),
+        ('Foo.Bar2', 1),
+        ('Foo.Bar3', 1)
+    ]
+    self._check_results_file(expected_result_file_entries)
+
   def test_simple_fail(self):
-    out, err, return_code = RunTest('gtest_fake_fail.py')
+    out, err, return_code = RunTest('gtest_fake_fail.py', self.filename)
 
     self.assertEquals(1, return_code)
 
-    expected_out_re = [
-      r'\[\d/\d\]   \d\.\d\ds .+',
-      r'\[\d/\d\]   \d\.\d\ds .+',
-      r'\[\d/\d\]   \d\.\d\ds .+',
-      r'\[\d/\d\]   \d\.\d\ds .+',
-      r'\[\d/\d\]   \d\.\d\ds .+',
-      r'\[\d/\d\]   \d\.\d\ds .+',
+    test_fail_output = [
       re.escape('Note: Google Test filter = Baz.Fail'),
       r'',
     ] + [
-        re.escape(l) for l in
-        gtest_fake_base.get_test_output('Baz.Fail').splitlines()
+      re.escape(l) for l in
+      gtest_fake_base.get_test_output('Baz.Fail').splitlines()
     ] + [
       '',
     ] + [
       re.escape(l) for l in gtest_fake_base.get_footer(1, 1).splitlines()
     ] + [
-      '',
+      ''
+    ]
+
+    expected_out_re = [
+      r'\[\d/\d\]   \d\.\d\ds .+',
+      r'\[\d/\d\]   \d\.\d\ds .+',
+      r'\[\d/\d\]   \d\.\d\ds .+',
+      r'\[\d/\d\]   \d\.\d\ds .+',
+      r'\[\d/\d\]   \d\.\d\ds .+',
+      r'\[\d/\d\]   \d\.\d\ds .+',
+    ] + test_fail_output + [
+      re.escape('Retrying failed tests serially.'),
+      r'\[\d/\d\]   \d\.\d\ds .+',
+    ] + test_fail_output + [
+      re.escape('Summary:'),
+      re.escape('Baz.Fail failed'),
       re.escape('Success:    3 75.00%'),
       re.escape('Flaky:      0  0.00%'),
       re.escape('Fail:       1 25.00%'),
@@ -99,6 +139,14 @@ class TraceTestCases(unittest.TestCase):
     ]
     self._check_results(expected_out_re, out, err)
 
+    expected_result_file_entries = [
+        ('Foo.Bar1', 1),
+        ('Foo.Bar2', 1),
+        ('Foo.Bar3', 1),
+        ('Baz.Fail', 4)
+    ]
+    self._check_results_file(expected_result_file_entries)
+
   def test_simple_gtest_list_error(self):
     out, err, return_code = RunTest('gtest_fake_error.py')