Only label milestones for ChromiumPerf and ChromiumPerfFyi and corresponding test.

dashboard/dashboard/file_bug.py

Index: dashboard/dashboard/file_bug.py
diff --git a/dashboard/dashboard/file_bug.py b/dashboard/dashboard/file_bug.py
index 3bca816800b837309b6bc4d0bd216bf4212b2ae6..cc30ad9641c1c5ba24b94cff0df55cbe091d29df 100644
--- a/dashboard/dashboard/file_bug.py
+++ b/dashboard/dashboard/file_bug.py
@@ -232,24 +232,26 @@ def _MilestoneLabel(alerts):
   Chromium milestones, do not label them (see
   https://github.com/catapult-project/catapult/issues/2906).
   """
-  revisions = [a.start_revision for a in alerts if hasattr(a, 'start_revision')]
+  revisions = [a.end_revision for a in alerts if hasattr(a, 'end_revision')]
   if not revisions:
     return None
-  start_revision = min(revisions)
-
+  end_revision = min(revisions)
   for a in alerts:
-    if a.start_revision == start_revision:
-      master = utils.TestPath(a.GetTestMetadataKey()).split('/')[0]
-      if master != 'ChromiumPerf' and master != 'ChromiumPerfFyi':
+    if a.end_revision == end_revision:
+      row_key = utils.GetRowKey(a.test, a.end_revision)
+      row = row_key.get()
+      if hasattr(row, 'r_commit_pos'):
+        end_revision = row.r_commit_pos
+      else:
         return None
       break
   try:
-    milestone = _GetMilestoneForRevision(start_revision)
+    milestone = _GetMilestoneForRevision(end_revision)
   except KeyError:
     logging.error('List of versions not in the expected format')
   if not milestone:
     return None
-  logging.info('Matched rev %s to milestone %s.', start_revision, milestone)
+  logging.info('Matched rev %s to milestone %s.', end_revision, milestone)
   return 'M-%d' % milestone
 
 
@@ -260,7 +262,7 @@ def _GetMilestoneForRevision(revision):
   by a suspected regression. We do this by locating in the list of current
   versions, regardless of platform and channel, all the version strings (e.g.
   36.0.1234.56) that match revisions (commit positions) later than the earliest
-  possible start_revision of the suspected regression; we then parse out the
+  possible end_revision of the suspected regression; we then parse out the
   first numeric part of such strings, assume it to be the corresponding
   milestone, and return the lowest one in the set.