Revert of Enable more benchmark_smoke_unittest coverage (patchset #4 id:100001...

Revert of Enable more benchmark_smoke_unittest coverage (patchset #4 id:100001 of https://codereview.chromium.org/1144193002/)

Reason for revert:
Reverting due to flaky perf tests (see crbug.com/490130 for details).

Original issue's description:
> Enable more benchmark_smoke_unittest coverage
>
> Previously, benchmark_smoke_unittest has a heuristic to pick
> a benchmark for smoke testing if the benchmark uses a test
> in measurement directory.
> This patch removes that check, and instead provide a black
> list of benchmarks module that we exclude the
> benchmark_smoke_unittest to avoid high CQ time.
>
> This helped increases benchmark smoke coverage from 14 benchmarks -> 33 benchmarks (one in each benchmark module).
>
> On my local linux machine, this increases the cycle time of
> benchmark_smoke_unittest from 20s -> 1m7s.
>
> BUG=490130
> CQ_EXTRA_TRYBOTS=tryserver.chromium.perf:linux_perf_bisect;tryserver.chromium.perf:mac_perf_bisect;tryserver.chromium.perf:android_nexus5_perf_bisect
>
> Committed: https://crrev.com/65014162e48a7f5881e8b282dd9e8dc17c5aa7c4
> Cr-Commit-Position: refs/heads/master@{#330909}

TBR=sullivan@chromium.org,dtu@chromium.org,nednguyen@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=490130

Review URL: https://codereview.chromium.org/1143363006

Cr-Commit-Position: refs/heads/master@{#330923}

tools/perf/benchmarks/benchmark_smoke_unittest.py

@@ -10,21 +10,14 @@ of every benchmark would run impractically long.
 """
 
 import os
-import sys
-import time
 import unittest
 
 from telemetry import benchmark as benchmark_module
 from telemetry.core import discover
+from telemetry.page import page_test
 from telemetry.unittest_util import options_for_unittests
 from telemetry.unittest_util import progress_reporter
 
-from benchmarks import dom_perf
-from benchmarks import rasterize_and_record_micro
-from benchmarks import spaceport
-from benchmarks import speedometer
-from benchmarks import jetstream
-
 
 def SmokeTestGenerator(benchmark):
   # NOTE TO SHERIFFS: DO NOT DISABLE THIS TEST.
@@ -63,42 +56,34 @@ def SmokeTestGenerator(benchmark):
     benchmark.ProcessCommandLineArgs(None, options)
     benchmark_module.ProcessCommandLineArgs(None, options)
 
-    current = time.time()
-    try:
-      self.assertEqual(0, SinglePageBenchmark().Run(options),
-                       msg='Failed: %s' % benchmark)
-    finally:
-      print 'Benchmark %s run takes %i seconds' % (
-          benchmark.Name(), time.time() - current)
+    self.assertEqual(0, SinglePageBenchmark().Run(options),
+                     msg='Failed: %s' % benchmark)
 
   return BenchmarkSmokeTest
 
 
-# The list of benchmark modules to be excluded from our smoke tests.
-_BLACK_LIST_TEST_MODULES = {
-    dom_perf,   # Always fails on cq bot.
-    rasterize_and_record_micro,  # Always fails on cq bot.
-    spaceport,  # Takes 451 seconds.
-    speedometer,  # Takes 101 seconds.
-    jetstream,  # Take 206 seconds.
-}
-
-
 def load_tests(loader, standard_tests, pattern):
   del loader, standard_tests, pattern  # unused
   suite = progress_reporter.TestSuite()
 
   benchmarks_dir = os.path.dirname(__file__)
   top_level_dir = os.path.dirname(benchmarks_dir)
+  measurements_dir = os.path.join(top_level_dir, 'measurements')
 
+  all_measurements = discover.DiscoverClasses(
+      measurements_dir, top_level_dir, page_test.PageTest).values()
   # Using the default of |index_by_class_name=False| means that if a module
   # has multiple benchmarks, only the last one is returned.
   all_benchmarks = discover.DiscoverClasses(
       benchmarks_dir, top_level_dir, benchmark_module.Benchmark,
       index_by_class_name=False).values()
   for benchmark in all_benchmarks:
-    if sys.modules[benchmark.__module__] in _BLACK_LIST_TEST_MODULES:
+    if hasattr(benchmark, 'test') and benchmark.test not in all_measurements:
+      # If the benchmark does not have a measurement, then it is not composable.
+      # Ideally we'd like to test these as well, but the non-composable
+      # benchmarks are usually long-running benchmarks.
       continue
+
     # TODO(tonyg): Smoke doesn't work with session_restore yet.
     if (benchmark.Name().startswith('session_restore') or
         benchmark.Name().startswith('skpicture_printer')):

tools/telemetry/telemetry/user_story/user_story_set.py

@@ -43,7 +43,7 @@ class UserStorySet(object):
     self._cloud_storage_bucket = cloud_storage_bucket
     if base_dir:
       if not os.path.isdir(base_dir):
-        raise ValueError('Invalid directory path of base_dir: %s' % base_dir)
+        raise ValueError('Must provide valid directory path for base_dir.')
       self._base_dir = base_dir
     else:
       self._base_dir = os.path.dirname(inspect.getfile(self.__class__))