Update gtest-parallel and introduce gtest-parallel-wrapper.

third_party/gtest-parallel/gtest-parallel-wrapper.py

Index: third_party/gtest-parallel/gtest-parallel-wrapper.py
diff --git a/third_party/gtest-parallel/gtest-parallel-wrapper.py b/third_party/gtest-parallel/gtest-parallel-wrapper.py
new file mode 100755
index 0000000000000000000000000000000000000000..1d8bcac71323fcf2572861cc5e24e5d8b9709d3d
--- /dev/null
+++ b/third_party/gtest-parallel/gtest-parallel-wrapper.py
@@ -0,0 +1,37 @@
+#!/usr/bin/env python
+
+# Copyright (c) 2016 The WebRTC project authors. All Rights Reserved.
+#
+# Use of this source code is governed by a BSD-style license
+# that can be found in the LICENSE file in the root of the source
+# tree. An additional intellectual property rights grant can be found
+# in the file PATENTS.  All contributing project authors may
+# be found in the AUTHORS file in the root of the source tree.
+
+import os
+import subprocess
+import sys
+
+# GTEST_SHARD_INDEX and GTEST_TOTAL_SHARDS must be removed from the environment

[kjellander_webrtc, 2016/11/21 07:14:31]

Ah, I didn't know having these set caused bugs. Then I'm much more supportive to
adding this wrapper!

+# otherwise it will be picked up by the binary, causing a bug where only tests
+# in the firsh shard are executed.
+test_env = os.environ.copy()
+gtest_shard_index = test_env.pop('GTEST_SHARD_INDEX', '0')
+gtest_total_shards = test_env.pop('GTEST_TOTAL_SHARDS', '1')
+
+gtest_parallel_path = os.path.dirname(os.path.abspath(__file__))
+gtest_parallel_path = os.path.join(gtest_parallel_path, 'gtest-parallel')
+
+command = [
+    sys.executable,
+    gtest_parallel_path,
+    '--shard_count',
+    gtest_total_shards,
+    '--shard_index',
+    gtest_shard_index,
+] + sys.argv[1:]
+
+print 'gtest-parallel-wrapper: Executing command %s' % ' '.join(command)
+sys.stdout.flush()
+
+sys.exit(subprocess.call(command, env=test_env, cwd=os.getcwd()))

[kjellander_webrtc, 2016/11/21 07:14:31]

What if one runs this script and presses ctrl+c to abort execution. Does that
termination signal propagate properly? I remember gtest-parallel had issues with
this in the early days so I want to make sure adding this new layer on top of it
makes it as easy to abort as before. Even if it's unlikely to be run by
developers aborting a buildbot could be a similar scenario.

[ehmaldonado_webrtc, 2016/11/21 13:23:57]

It seems like I can abort it just fine. Maybe you can patch and confirm?