Add support for backgrounding processes on the Mac

content/browser/renderer_host/render_process_host_impl.cc

Index: content/browser/renderer_host/render_process_host_impl.cc
diff --git a/content/browser/renderer_host/render_process_host_impl.cc b/content/browser/renderer_host/render_process_host_impl.cc
index 1e20d8e6a899b551bc85a4362a7507ebdfb8bf39..82f1c5a73e85da7b5a9c9ef32b5390cba1a4aeb3 100644
--- a/content/browser/renderer_host/render_process_host_impl.cc
+++ b/content/browser/renderer_host/render_process_host_impl.cc
@@ -1,4 +1,4 @@
-// Copyright (c) 2012 The Chromium Authors. All rights reserved.
+// Copyright 2012 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.
 
@@ -2246,12 +2246,21 @@ void RenderProcessHostImpl::SetBackgrounded(bool backgrounded) {
 
 #if defined(OS_WIN)
   // Same as below, but bound to an experiment (http://crbug.com/458594)
-  // initially on Windows. Enabled by default in the asbence of field trials to
+  // initially on Windows. Enabled by default in the absence of field trials to
   // get coverage on the perf waterfall.
   base::FieldTrial* trial =
       base::FieldTrialList::Find("BackgroundRendererProcesses");
   if (!trial || trial->group_name() != "Disallow")
     child_process_launcher_->SetProcessBackgrounded(backgrounded);
+#elif defined(OS_MACOSX)
+  // Same as below, but bound to an experiment (http://crbug.com/458594) .
+  // Disabled by default (perf waterfall coverage will come through a new
+  // telemetry benchmark).
+  const std::string trial_group_name =
+      base::FieldTrialList::FindFullName("BackgroundRendererProcesses");
+  if (!trial_group_name.empty() && trial_group_name != "Disallow") {

[Alexei Svitkine (slow), 2015/04/23 20:53:28]

This pattern makes it impossible to have a Control group with the same behavior
as the Disallow group. 

Experimentation best practice suggest having a Control group. This used to be
handled by default when the advice was to be "off by default", but now that this
is reversed by the new launch process, you need to be more careful when setting
this up.

Either:
  1) Use experiment params rather than checking for the group name.
  2) Or change the check to not rely on a specific group name - for example you
can use StartsWith()

Please make these changes to both the Windows and Mac experiment branches, since
they suffer from the same problem.

+    child_process_launcher_->SetProcessBackgrounded(backgrounded);
+  }
 #else
   // Control the background state from the browser process, otherwise the task
   // telling the renderer to "unbackground" itself may be preempted by other