Add a retry to Chromoting tests on Linux builders.

testing/chromoting/browser_tests_launcher.py

Index: testing/chromoting/browser_tests_launcher.py
diff --git a/testing/chromoting/browser_tests_launcher.py b/testing/chromoting/browser_tests_launcher.py
index 0b549f33486d6822b3d6205b277736774324ea87..968e95620b8f0f65ae4c8e8dfc1d322d53a97882 100644
--- a/testing/chromoting/browser_tests_launcher.py
+++ b/testing/chromoting/browser_tests_launcher.py
@@ -11,6 +11,7 @@ from chromoting_test_utilities import CleanupUserProfileDir
 from chromoting_test_utilities import GetJidFromHostLog
 from chromoting_test_utilities import GetJidListFromTestResults
 from chromoting_test_utilities import InitialiseTestMachineForLinux
+from chromoting_test_utilities import MAX_RETRIES
 from chromoting_test_utilities import PrintHostLogContents
 from chromoting_test_utilities import PROD_DIR_ID
 from chromoting_test_utilities import RunCommandInSubProcess
@@ -23,7 +24,6 @@ FAILING_TESTS = ''
 BROWSER_NOT_STARTED_ERROR = (
     'Still waiting for the following processes to finish')
 TIME_OUT_INDICATOR = '(TIMED OUT)'
-MAX_RETRIES = 1
 
 
 def LaunchBTCommand(args, command):
@@ -93,18 +93,7 @@ def LaunchBTCommand(args, command):
     if SUCCESS_INDICATOR in results:
       break
 
-    # Sometimes, during execution of browser-tests, a browser instance is

[joedow, 2016/02/20 17:08:38]

I think this logic is important to keep.  In the long term we will remove the
generic retry logic but this was something you worked on quite a bit to track
down and it is likely to remain a problem (since it was never root caused).  I'd
hate to remove the retry logic and then start hitting his failure again.

It should be too hard to distinguish between this failure and some other
failure, the retry logic is ultimately the same but we would keep this
protection for later.

If you ultimately decide to rip this code out, then please remove the vars at
the top of the file that were referenced here but are no longer used.

[anandc1, 2016/02/22 20:58:08]

Done. Logging a line indicating what type of failure this potentially was and
retrying for any failure, up to MAX_RETRIES.

-    # not started and the test times out. See http://crbug/480025.
-    # To work around it, check if this execution failed owing to that
-    # problem and retry.
-    # There are 2 things to look for in the results:
-    # A line saying "Still waiting for the following processes to finish",
-    # and, because sometimes that line gets logged even if the test
-    # eventually passes, we'll also look for "(TIMED OUT)", before retrying.
-    if not (
-        BROWSER_NOT_STARTED_ERROR in results and TIME_OUT_INDICATOR in results):
-      # Test failed for some other reason. Let's not retry.
-      break
+    # Test failed, retry.
     retries += 1
 
   # Check that the test passed.