Update legacy bisect to post results to perf dashboard.

tools/auto_bisect/bisect_perf_regression_test.py

Index: tools/auto_bisect/bisect_perf_regression_test.py
diff --git a/tools/auto_bisect/bisect_perf_regression_test.py b/tools/auto_bisect/bisect_perf_regression_test.py
index 33f733a7332641521a58b226afa48b69f74f382c..454f27b7cb8927e49352adceba66b3bf0ea52ae1 100644
--- a/tools/auto_bisect/bisect_perf_regression_test.py
+++ b/tools/auto_bisect/bisect_perf_regression_test.py
@@ -131,6 +131,23 @@ def _FakeTestResult(values, bisect_mode_is_return_code):
   return (result_dict, success_code)
 
 
+def _GetMockCallArg(function_mock, call_index):
+  """Gets the list of called arguements for call at "call_index".

[qyearsley, 2016/01/26 18:08:01]

arguements -> arguments

"call_index" -> |call_index|.

[chrisphan, 2016/02/10 02:23:38]

Done.

+
+  Args:
+    function_mock: A Mock object.
+    call_index: The index at which the mocked function was called.

[qyearsley, 2016/01/26 18:08:00]

Is this the index in the sequence of calls? e.g. 0 means first call, 1 means
second call?

[chrisphan, 2016/02/10 02:23:38]

That's correct.

+
+  Returns:
+    The called argument list.
+  """
+  call_args_list = function_mock.call_args_list
+  if not call_args_list or len(call_args_list) <= call_index:
+    return None
+  args, _ = call_args_list[call_index]
+  return args
+
+
 def _GetBisectPerformanceMetricsInstance(options_dict):
   """Returns an instance of the BisectPerformanceMetrics class."""
   opts = bisect_perf_regression.BisectOptions.FromDict(options_dict)
@@ -160,7 +177,7 @@ def _GenericDryRun(options, print_results=False):
   Returns:
     The results dictionary as returned by the bisect Run method.
   """
-  _AbortIfThereAreStagedChanges()
+  #_AbortIfThereAreStagedChanges()

[qyearsley, 2016/01/26 18:08:00]

Uncomment

[chrisphan, 2016/02/10 02:23:38]

Done.

   # Disable rmtree to avoid deleting local trees.
   old_rmtree = shutil.rmtree
   shutil.rmtree = lambda path, on_error: None
@@ -319,6 +336,23 @@ class BisectPerfRegressionTest(unittest.TestCase):
     results = _GenericDryRun(_GetExtendedOptions(1, -100))
     self.assertIsNone(results.error)
 
+  @mock.patch('urllib2.urlopen')
+  def testBisectResultsPosted(self, mock_urlopen):
+    """Bisects with improvement direction matching regression range."""

[qyearsley, 2016/01/26 18:08:00]

Docstring should be deleted or updated.

[chrisphan, 2016/02/10 02:23:38]

Done.

+    # Test result goes from 0 to 100 where lower is better

[qyearsley, 2016/01/26 18:08:01]

This comment appears to be unrelated.

[chrisphan, 2016/02/10 02:23:38]

Done.

+    options_dict = dict(DEFAULT_OPTIONS)
+    options_dict.update({
+        'bisect_mode': bisect_utils.BISECT_MODE_MEAN,
+        'try_job_id': 1234,
+    })
+    opts = bisect_perf_regression.BisectOptions.FromDict(options_dict)
+    results = _GenericDryRun(options_dict, True)
+    bisect_perf_regression._PostBisectResults(results, opts, os.getcwd())

[qyearsley, 2016/01/26 18:08:00]

Optional: Because the purpose of this test method is only to test posting, not
to test a dry run, instead of calling _GenericDryRun above we could just
potentially just construct a results object here directly.

[chrisphan, 2016/02/10 02:23:38]

Done.

+
+    call_args = _GetMockCallArg(mock_urlopen, 0)
+    self.assertIsNotNone(call_args)
+    self.assertIn('"try_job_id": 1234', call_args[1])
+
   def _CheckAbortsEarly(self, results, **extra_opts):
     """Returns True if the bisect job would abort early."""
     global _MockResultsGenerator