[Findit] Flake Checker: Get swarming task triggered by flake try jobs.

appengine/findit/waterfall/flake/process_flake_try_job_result_pipeline.py

Index: appengine/findit/waterfall/flake/process_flake_try_job_result_pipeline.py
diff --git a/appengine/findit/waterfall/flake/process_flake_try_job_result_pipeline.py b/appengine/findit/waterfall/flake/process_flake_try_job_result_pipeline.py
index 2a5fb799d056b705a018a87b9a376221e355ba68..0c34193ccdfebbeca1b1d5ad51704d3f8c16d93b 100644
--- a/appengine/findit/waterfall/flake/process_flake_try_job_result_pipeline.py
+++ b/appengine/findit/waterfall/flake/process_flake_try_job_result_pipeline.py
@@ -5,8 +5,31 @@
 from google.appengine.ext import ndb
 
 from common.pipeline_wrapper import BasePipeline
-from model.flake.flake_try_job import FlakeTryJob
+from gae_libs.http.http_client_appengine import HttpClientAppengine
 from model.flake.master_flake_analysis import DataPoint
+from waterfall import swarming_util
+
+
+def _GetSwarmingTaskIdForTryJob(report, revision, step_name):
+  """Check json output for each task and return id of the one with test result.
+  """
+  if not report:
+    return None
+
+  http_client = HttpClientAppengine()
+  task_ids = report.get('result', {}).get(revision, {}).get(

[stgao, 2017/01/12 07:43:27]

If there is only one task id, do we still have to query Swarming or Isolated
server?

If the test doesn't exist in the tested revision, should we just pick any task
id or return None?

[chanli, 2017/01/12 20:28:29]

Good point. Just updated.
If test doesn't exist, would there be swarming tasks?

[stgao, 2017/01/13 02:18:18]

Example: a test is newly added in the suspected build number. Bisecting at
commits before the test being added will still run Swarming tasks, but all
Swarming tasks won't have any iteration data for the test.

+      step_name, {}).get('step_metadata', {}).get('swarm_task_ids', [])
+
+  for task_id in task_ids:
+    output_json = swarming_util.GetIsolatedOutputForTask(task_id, http_client)
+    if output_json:
+      for data in output_json.get('per_iteration_data', []):
+        # If this task doesn't have result, per_iteration_data will look like
+        # [{}, {}, ...]
+        if data:
+          return task_id
+
+  return None
 
 
 class ProcessFlakeTryJobResultPipeline(BasePipeline):
@@ -65,5 +88,6 @@ class ProcessFlakeTryJobResultPipeline(BasePipeline):
     data_point.pass_rate = pass_rate
     data_point.try_job_id = try_job.try_job_ids[-1]
     data_point.try_job_url = try_job.flake_results[-1].get('url')
-    # TODO: Add swarming task data.
+    data_point.task_id = _GetSwarmingTaskIdForTryJob(
+        try_job.flake_results[-1].get('report'), revision, step_name)
     flake_analysis.data_points.append(data_point)