Add support for cc_perftests and other non-telemetry gtest based tests.

testing/scripts/common.py

 # Copyright 2014 The Chromium Authors. All rights reserved.
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
 import argparse
 import contextlib
 import json
 import os
 import subprocess
 import sys
 import tempfile
 
 
 SCRIPT_DIR = os.path.abspath(os.path.dirname(__file__))
 SRC_DIR = os.path.abspath(
     os.path.join(SCRIPT_DIR, os.path.pardir, os.path.pardir))
 
 
 # run-webkit-tests returns the number of failures as the return
 # code, but caps the return code at 101 to avoid overflow or colliding
 # with reserved values from the shell.
 MAX_FAILURES_EXIT_STATUS = 101
 
 
 def run_script(argv, funcs):
   def parse_json(path):
     with open(path) as f:
       return json.load(f)
   parser = argparse.ArgumentParser()
   # TODO(phajdan.jr): Make build-config-fs required after passing it in recipe.

[iannucci, 2015/02/23 18:02:49]

TODO(phajdan.jr): help strings for these options so we can tell what they heck
they do :)

   parser.add_argument('--build-config-fs')
   parser.add_argument('--paths', type=parse_json, default={})
   parser.add_argument('--properties', type=parse_json, default={})
+  parser.add_argument('--args', type=parse_json, default=[])

[iannucci, 2015/02/23 18:02:49]

why is the introduction of a new argument necessary? Why can't the relevant info
be shuttled through properties?

[iannucci, 2015/02/23 18:04:33]

Actually... it looks like this isn't even used by the new script? AFAICT
`args.properties` is actually referring to the --properties option above.

[shatch, 2015/02/23 20:14:21]

This was Pawel's preference when we spoke over email.
The new argument is used to get the script args.

[Paweł Hajdan Jr., 2015/02/23 20:46:30]

I consider properties to be the same for all script invocations, and describe
the environment of the build.

Args can be different per invocation, e.g. you could have the same script but
invoked in different steps with different args (e.g. a gtest step with different
test names like base_unittests, net_unittests, etc).

I think the distinction is important and actually results in cleaner code.

 
   subparsers = parser.add_subparsers()
 
   run_parser = subparsers.add_parser('run')
   run_parser.add_argument(
       '--output', type=argparse.FileType('w'), required=True)
   run_parser.add_argument('--filter-file', type=argparse.FileType('r'))
   run_parser.set_defaults(func=funcs['run'])
 
   run_parser = subparsers.add_parser('compile_targets')
@@ skipping 81 matching lines @@
       key += 'passes'
       # TODO(dpranke): crbug.com/357867 ...  Why are we assigning result
       # instead of actual_result here. Do we even need these things to be
       # hashes, or just lists?
       data = result
     else:
       key += 'failures'
     results[key][test] = data
 
   return results