Reland of Android: support multiple displays on C++ side

ui/android/screen_android.h

Index: ui/android/screen_android.h
diff --git a/ui/android/screen_android.h b/ui/android/screen_android.h
new file mode 100644
index 0000000000000000000000000000000000000000..ddd6ba0eebe5bab2b436c7af48eaba1e06f6597e
--- /dev/null
+++ b/ui/android/screen_android.h
@@ -0,0 +1,23 @@
+// 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 UI_ANDROID_SCREEN_ANDROID_H_
+#define UI_ANDROID_SCREEN_ANDROID_H_
+
+#include <jni.h>
+#include "ui/android/ui_android_export.h"
+
+namespace display {
+class Screen;
+}
+
+namespace ui {
+
+bool RegisterScreenAndroid(JNIEnv* env);
+
+UI_ANDROID_EXPORT display::Screen* CreateScreenAndroid();

[boliu, 2016/11/07 17:05:33]

what if we also create the screen inside RegisterScreenAndroid? a bit
unidiomatic, but seems much easier?

[Tima Vaisburd, 2016/11/07 18:26:04]

I thought we wanted to be similar to other platforms... I will try this idea.

[Tima Vaisburd, 2016/11/09 01:17:01]

Couldn't do it: in Chrome the call to
Java_DisplayAndroidManager_onNativeSideCreated() failed because it missed the
command line object. In WebView, the JNI registration is skipped (btw, how does
it work?), so RegisterScreenAndroid is not called at all.

[boliu, 2016/11/09 03:34:25]

meh, init is so complicated

how about lazily when the primary display is registered?

as you can see, if you want to add an explicit init call, then you need to add
that call in *a lot* of places..
Oh yeah.. forgot, jni registration is lazy in webview

[Tima Vaisburd, 2016/11/09 09:01:43]

Sigh.. maybe in some common part in the content layer initialization?

+
+}  // namespace display
+
+#endif  // UI_ANDROID_SCREEN_ANDROID_H_