Basic display management for mus

components/mus/ws/platform_screen.h

Index: components/mus/ws/platform_screen.h
diff --git a/components/mus/ws/platform_screen.h b/components/mus/ws/platform_screen.h
new file mode 100644
index 0000000000000000000000000000000000000000..c6f672d5551574a7e5e493c7e242cf4c0a28ac2a
--- /dev/null
+++ b/components/mus/ws/platform_screen.h
@@ -0,0 +1,43 @@
+// Copyright 2016 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.
+
+#ifndef COMPONENTS_MUS_WS_PLATFORM_SCREEN_H_
+#define COMPONENTS_MUS_WS_PLATFORM_SCREEN_H_
+
+#include <stdint.h>
+
+#include "base/callback.h"
+
+namespace gfx {
+class Rect;
+}
+
+namespace mus {
+namespace ws {
+
+// PlatformScreen provides the necessary functionality to configure all
+// attached physical displays.
+class PlatformScreen {
+ public:
+  PlatformScreen();

[sky, 2016/05/04 17:05:44]

You shouldn't need to have this.

[rjkroege, 2016/05/04 22:50:13]

Done.

+  virtual ~PlatformScreen();

[sky, 2016/05/04 17:05:44]

And inline this (we generally inline destructor for pure virtual classes like
this).

[rjkroege, 2016/05/04 22:50:13]

Done. Aside: wasn't there a time when destructors needed to be in a .cc file to
link correctly?

+
+  // Creates a PlatformScreen instance.
+  static std::unique_ptr<PlatformScreen> Create();
+
+  // Initializes platform specific screen resources.
+  virtual void Init() = 0;
+
+  using ConfiguredDisplayCallback = base::Callback<void(int64_t, gfx::Rect)>;

[sky, 2016/05/04 17:05:44]

const gfx::Rect&

[rjkroege, 2016/05/04 22:50:13]

Done.

+
+  // ConfigurePhysicalDisplay() configures a single physical display and returns
+  // its id and bounds for it via |callback|.
+  virtual void ConfigurePhysicalDisplay(
+      const ConfiguredDisplayCallback& callback) = 0;

[sky, 2016/05/04 17:05:44]

Do you expect this to be called more than once? It seems like this is a one time
call. Can this function be nuked and the callback supplied to Init()?

[rjkroege, 2016/05/04 22:50:13]

Eventually yes: split-mus restarting the mus-gpu or mash reconfiguring the
screen. At present, no.
I think this is not a good idea as it combines code with different ordering
intents:

PlatformScreen::Init  - sets up the PlatformScreen class -- a process that might
be time consuming and should be started early as possible (e.g. before entering
a sandbox.)

PlatformScreen::ConfigurePhysicalDisplay - the ws thread is saying that it is
waiting for a configured physical display as soon as possible.

+};
+
+}  // namespace ws
+}  // namespace mus
+
+#endif  // COMPONENTS_MUS_WS_PLATFORM_SCREEN_H_