[Findit] Use heuristic analysis result for test try jobs(Findit side).

appengine/findit/waterfall/identify_try_job_culprit_pipeline.py

Index: appengine/findit/waterfall/identify_try_job_culprit_pipeline.py
diff --git a/appengine/findit/waterfall/identify_try_job_culprit_pipeline.py b/appengine/findit/waterfall/identify_try_job_culprit_pipeline.py
index eac4f40a5f5d551037250c8bbcc79303c0dcbe07..3f521b737c9273285eb75cf51391174f310ac11a 100644
--- a/appengine/findit/waterfall/identify_try_job_culprit_pipeline.py
+++ b/appengine/findit/waterfall/identify_try_job_culprit_pipeline.py
@@ -168,13 +168,16 @@ class IdentifyTryJobCulpritPipeline(BasePipeline):
         report.get('result', {}))
 
   def _FindCulpritForEachTestFailure(self, blame_list, result):
-    # For test failures, the try job will run against every revision,
-    # so we need to traverse the result dict in chronological order to identify
-    # the culprits for each failed step or test.
+    # For test failures, we need to traverse the result dict in chronological
+    # order to identify the culprits for each failed step or test.
+    # The earliest revision that a test failed is the culprit.
     culprit_map = {}
     failed_revisions = []
+    test_results = result['report'].get('result')
+
     for revision in blame_list:
-      test_results = result['report'].get('result')
+      if not test_results.get(revision):

[stgao, 2016/05/04 21:40:18]

How about let the recipe compute the culprits it finds for each test?

In that case, the recipe could change the way how it tests without impacting the
logic here. WDYT?

[chanli, 2016/05/04 23:41:37]

Done.

+        continue
 
       for step, test_result in test_results[revision].iteritems():
         if (not test_result['valid'] or