[Findit] Modify tryjob pipelines to trigger try jobs for test failure.

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 568b5a8cb0060d04f9970ba4bcdc90e94e9322d3..83d8fbb4282df3a6bea19390581c6e30df2c7a39 100644
--- a/appengine/findit/waterfall/identify_try_job_culprit_pipeline.py
+++ b/appengine/findit/waterfall/identify_try_job_culprit_pipeline.py
@@ -7,11 +7,31 @@ from common.http_client_appengine import HttpClientAppengine as HttpClient
 from model import wf_analysis_status
 from model.wf_try_job import WfTryJob
 from pipeline_wrapper import BasePipeline
+from waterfall.try_job_type import TryJobType
+
+
+GIT_REPO = GitRepository(
+    'https://chromium.googlesource.com/chromium/src.git', HttpClient())
 
 
 class IdentifyTryJobCulpritPipeline(BasePipeline):
   """A pipeline to identify culprit CL info based on try job compile results."""
 
+  def _GetCulpritInfo(self, failed_revisions):
+    """Gets commit_positions and review_urls for revisions."""
+    culprits = {}
+    for failed_revision in failed_revisions:
+      culprits[failed_revision] = {
+          'revision': failed_revision
+      }
+      change_log = GIT_REPO.GetChangeLog(failed_revision)
+      if change_log:
+        culprits[failed_revision]['commit_position'] = (
+            change_log.commit_position)
+        culprits[failed_revision]['review_url'] = change_log.code_review_url
+
+    return culprits
+
   @staticmethod
   def _GetFailedRevisionFromResultsDict(results_dict):
     """Finds the failed revision from the given dict of revisions.
@@ -101,43 +121,89 @@ class IdentifyTryJobCulpritPipeline(BasePipeline):
 
     return failed_revision
 
-  @staticmethod
-  def _GetCulpritFromFailedRevision(failed_revision):
-    """Returns a culprit (dict) using failed_revision, or None."""
-    if not failed_revision:
-      return None
-
-    git_repo = GitRepository(
-        'https://chromium.googlesource.com/chromium/src.git', HttpClient())
-    change_log = git_repo.GetChangeLog(failed_revision)
-
-    if not change_log:
-      return None
-
-    return {
-        'revision': failed_revision,
-        'commit_position': change_log.commit_position,
-        'review_url': change_log.code_review_url
-    }
+  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.
+    culprit_map = {}
+    failed_revisions = []
+    for revision in blame_list:
+      for step, step_result in result['report'][revision].iteritems():
+        if step_result['valid'] and step_result['status'] == 'failed':

[qyearsley, 2016/02/03 18:55:14]

Optional change if you want to use one less level of indentation:

if not step_result['valid'] or step_result['status'] != 'failed':
  continue

Either way is probably equally good.

[chanli, 2016/02/03 23:44:13]

Done.

+          if revision not in failed_revisions:
+            failed_revisions.append(revision)
+
+          if step not in culprit_map:
+            culprit_map[step] = {
+                'tests': {}
+            }
+
+          if (not step_result['failures'] and
+              not culprit_map[step].get('revision')):
+            # Non swarming test failures, only have step level failure info.
+            culprit_map[step]['revision'] = revision
+
+          for failed_test in step_result['failures']:
+            # Swarming tests, gets first failed revision for each test.
+            if failed_test not in culprit_map[step]['tests']:
+              culprit_map[step]['tests'][failed_test] = {
+                  'revision': revision
+              }
+    return culprit_map, failed_revisions
+
+  def _UpdateCulpritMapWithCulpritInfo(self, culprit_map, culprits):
+    """Fills in commit_position and review_url for each failed rev in map."""
+    for step_culprit in culprit_map.values():
+      if step_culprit.get('revision'):
+        culprit = culprits[step_culprit['revision']]
+        step_culprit['commit_position'] = culprit['commit_position']
+        step_culprit['review_url'] = culprit['review_url']
+      for test_culprit in step_culprit.get('tests', {}).values():
+        test_revision = test_culprit['revision']
+        test_culprit.update(culprits[test_revision])
 
   # Arguments number differs from overridden method - pylint: disable=W0221
-  def run(self, master_name, builder_name, build_number, try_job_id,
-          compile_result):
-    culprit = None
-    failed_revision = self._GetFailedRevisionFromCompileResult(compile_result)
-    culprit = self._GetCulpritFromFailedRevision(failed_revision)
+  def run(
+      self, master_name, builder_name, build_number, blame_list, try_job_type,
+      try_job_id, result):
+    """Identifies the information for failed revisions.
+
+    Please refer to try_job_result_format.md for format check.
+    """
+    culprits = None
+    if result and result.get('report'):
+      if try_job_type == TryJobType.COMPILE:
+        # For compile failures, the try job will stop if one revision fails, so
+        # the culprit will be the last revision in the result.
+        failed_revision = self._GetFailedRevisionFromCompileResult(
+            result)
+        failed_revisions = [failed_revision] if failed_revision else []
+        culprits = self._GetCulpritInfo(failed_revisions)
+        if culprits:
+          result['culprit'] = culprits[failed_revision]
+      else:  # try_job_type is 'test'.
+        culprit_map, failed_revisions = self._FindCulpritForEachTestFailure(
+            blame_list, result)
+        culprits = self._GetCulpritInfo(failed_revisions)
+        if culprits:
+          self._UpdateCulpritMapWithCulpritInfo(culprit_map, culprits)
+          result['culprit'] = culprit_map
 
     # Store try job results.
     try_job_result = WfTryJob.Get(master_name, builder_name, build_number)
-    if culprit:
-      compile_result['culprit'] = culprit
-      if (try_job_result.compile_results and
-          try_job_result.compile_results[-1]['try_job_id'] == try_job_id):
-        try_job_result.compile_results[-1].update(compile_result)
+    if culprits:
+      result_to_update = (
+          try_job_result.compile_results if
+          try_job_type == TryJobType.COMPILE else
+          try_job_result.test_results)
+      if (result_to_update and
+          result_to_update[-1]['try_job_id'] == try_job_id):
+        result_to_update[-1].update(result)
+
       else:  # pragma: no cover
-        try_job_result.compile_results.append(compile_result)
+        result_to_update.append(result)
 
     try_job_result.status = wf_analysis_status.ANALYZED
     try_job_result.put()
 
-    return culprit
+    return result.get('culprit', None) if result else None

[qyearsley, 2016/02/03 18:55:14]

Equivalent to `result.get('culprit') if result else None`, since the default
default for dict.get() is None. But this is good as is too, since it is
explicit.

[chanli, 2016/02/03 23:44:13]

Done.