[Findit] Rerun if the regression range is different.

appengine/findit/crash/crash_pipeline.py

 # 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.
 
 import copy
 import json
 import logging
 
 from google.appengine.api import app_identity
 from google.appengine.ext import ndb
@@ skipping 90 matching lines @@
     with pipeline.After(run_analysis):
       yield PublishResultPipeline(crash_identifiers, client_id)
 
 
 @ndb.transactional
 def _NeedsNewAnalysis(
     crash_identifiers, chrome_version, signature, client_id,
     platform, stack_trace, customized_data):
   analysis = findit_for_client.GetAnalysisForClient(crash_identifiers,
                                                     client_id)
-  if analysis and not analysis.failed:
-    # A new analysis is not needed if last one didn't complete or succeeded.
-    # TODO(http://crbug.com/600535): re-analyze if stack trace or regression
-    # range changed.
+  regression_range = findit_for_client.GetRegressionRange(client_id,
+                                                          customized_data)
+  if (analysis and (regression_range == analysis.regression_range) and

[stgao, 2016/10/07 20:29:35]

Should we also test regression_range is not empty? Sometimes we don't have a
regression range, right?

[Sharu Jiang, 2016/10/07 22:05:03]

For those with empty regressions, we can still find suspected_project and
suspected_components. So we still need to run analyses.

[stgao, 2016/10/07 22:26:04]

What if both regression range are empty? Do we really want to re-analyze?

+      not analysis.failed):

[wrengr, 2016/10/07 17:32:06]

Does it make sense to check analysis.failed after comparing the
regression_ranges? or should analysis.failed be checked first?

[Sharu Jiang, 2016/10/07 22:05:03]

Done.

     logging.info('The analysis of %s has already been done.',
                  repr(crash_identifiers))
     return False
 
   # Create analysis for findit to run if this is not a rerun.
   if not analysis:
     analysis = findit_for_client.CreateAnalysisForClient(crash_identifiers,
                                                          client_id)
 
   findit_for_client.ResetAnalysis(analysis, chrome_version, signature,
-                                  client_id,  platform, stack_trace,
-                                  customized_data)
+                                  client_id, platform, stack_trace,
+                                  customized_data, regression_range)
   return True
 
 
 def ScheduleNewAnalysisForCrash(
     crash_identifiers, chrome_version, signature, client_id,
     platform, stack_trace, customized_data,
     queue_name=constants.DEFAULT_QUEUE):
   """Schedules an analysis."""
   # Check policy and tune arguments if needed.
   pass_policy, updated_analysis_args = findit_for_client.CheckPolicyForClient(
       crash_identifiers, chrome_version, signature,
       client_id, platform, stack_trace,
       customized_data)
 
   if not pass_policy:
     return False
 
   if _NeedsNewAnalysis(*updated_analysis_args):
     analysis_pipeline = CrashWrapperPipeline(crash_identifiers, client_id)
     # Attribute defined outside __init__ - pylint: disable=W0201
     analysis_pipeline.target = appengine_util.GetTargetNameForModule(
         constants.CRASH_BACKEND[client_id])
     analysis_pipeline.start(queue_name=queue_name)
     logging.info('New %s analysis is scheduled for %s', client_id,
                  repr(crash_identifiers))
     return True
 
   return False