[Mac] Maximise timer slack for background tabs

chrome/browser/resource_throttle.h

Index: chrome/browser/resource_throttle.h
diff --git a/chrome/browser/resource_throttle.h b/chrome/browser/resource_throttle.h
new file mode 100644
index 0000000000000000000000000000000000000000..bb52b2be40aa63fb7751672de32a1d46a09fb9da
--- /dev/null
+++ b/chrome/browser/resource_throttle.h
@@ -0,0 +1,39 @@
+// Copyright (c) 2014 The Chromium Authors. All rights reserved.

[Avi (use Gerrit), 2014/05/20 14:41:39]

No (c)

+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef CHROME_BROWSER_RESOURCE_THROTTLE_H_
+#define CHROME_BROWSER_RESOURCE_THROTTLE_H_
+
+#include "base/basictypes.h"
+#include "base/compiler_specific.h"
+#include "chrome/browser/media/media_capture_devices_dispatcher.h"
+#include "content/public/browser/web_contents_observer.h"
+#include "content/public/browser/web_contents_user_data.h"
+
+class ResourceThrottle

[tonyg, 2014/05/20 14:14:52]

Kari and I have been talking a lot about how to solve the general problem of
resource management. She's collected a handful of signals like this which we
think'd be useful for coalescing and prioritizing the network, cpu and other
system resources. We've been wondering whether we should have a single manager
or multiple.

Currently content/browser/loader/resource_scheduler.h does all of the network
scheduling and she's working on adding these signals to it as well. We could
either generalize one system_resource_manager which listens to all of the
signals we care about and sets all resource priorities or else we could do a
manager-per-resource: rename resource_scheduler.h to network_resource_manager.h
and name this cpu_resource_manager.h or something like that. Thoughts?

In any case, this functionality probably belongs in content/ not chrome/.

[aiolos (Not reviewing), 2014/05/20 17:45:23]

I agree with Tony that you'll want to move this to content instead of chrome if
possible. Either way I'd use content::MediaObserver here instead of
MediaCaptureDevicesDispatcher::Observer.

The debate over whether we should have one scheduler to rule them all is an
interesting one and I definitely think it's worth more discussion. There are
many signals that are important to every scheduler, like whether a tab/app/etc
is frontmost or playing audio, but there are also a few more specialized ones,
like whether a tab is pinned, that only seem important in specific cases.
Perhaps it would be cleaner to have a resource_manager with all the generally
important signals that can then be extended for each kind of resource. I don't
have a good enough feel for what would be the best option yet, so it might be
better to just rename this for the moment and start a broader conversation on
what to do moving forward. Thoughts?

On 2014/05/20 14:14:52, tonyg wrote:

+    : public content::WebContentsUserData<ResourceThrottle>,
+      content::WebContentsObserver,
+      MediaCaptureDevicesDispatcher::Observer {
+ public:
+  virtual ~ResourceThrottle();
+
+  // From WebContentsObserver .

[Avi (use Gerrit), 2014/05/20 14:41:39]

Why the space before the .? (here and below too)

+  virtual void WasShown() OVERRIDE;
+  virtual void WasHidden() OVERRIDE;
+
+  // From MediaCaptureDevicesDispatcher::Observer .
+  virtual void OnCreatingAudioStream(int render_process_id,
+                                       int render_frame_id) OVERRIDE;
+
+ private:
+  void SetTimerCoalescingEnabled(bool enabled);
+
+  explicit ResourceThrottle(content::WebContents* contents);
+  friend class content::WebContentsUserData<ResourceThrottle>;
+
+  content::WebContents* contents_;
+  bool timer_coalescing_enabled_;
+};
+
+#endif  // CHROME_BROWSER_RESOURCE_THROTTLE_H_