Add flaky test classification feature to test suites.

tools/run-tests.py

Index: tools/run-tests.py
diff --git a/tools/run-tests.py b/tools/run-tests.py
index 761d03fe3351f4a15a46c3745d3561bc5abac86e..03722a5ca3caa29dbf222d31757e692760dcdf35 100755
--- a/tools/run-tests.py
+++ b/tools/run-tests.py
@@ -94,6 +94,9 @@ def BuildOptions():
                     default=False, action="store_true")
   result.add_option("--cat", help="Print the source of the tests",
                     default=False, action="store_true")
+  result.add_option("--classification",

[Jakob Kummerow, 2013/08/06 11:11:01]

"classification" is too generic. How about "--flaky-tests" or
"--run-flaky-tests" instead? The options' names could stay the same.

+                    help="Class of tests to run (all|flaky|non-flaky)",
+                    default="all")
   result.add_option("--command-prefix",
                     help="Prepended to each shell command used to run a test",
                     default="")
@@ -204,6 +207,9 @@ def ProcessOptions(options):
     # This is OK for distributed running, so we don't need to set no_network.
     options.command_prefix = (["python", "-u", run_valgrind] +
                               options.command_prefix)
+  if not options.classification in ["all", "flaky", "non-flaky"]:
+    print "Unknown classification %s" % options.classification
+    return False
   return True
 
 
@@ -315,7 +321,7 @@ def Execute(arch, mode, args, options, suites, workspace):
     if len(args) > 0:
       s.FilterTestCasesByArgs(args)
     all_tests += s.tests
-    s.FilterTestCasesByStatus(options.warn_unused)
+    s.FilterTestCasesByStatus(options.warn_unused, options.classification)
     if options.cat:
       verbose.PrintTestSource(s.tests)
       continue