Let the WebMediaPlayer decide whether to use overlay video.

Source/core/html/HTMLVideoElement.cpp

Index: Source/core/html/HTMLVideoElement.cpp
diff --git a/Source/core/html/HTMLVideoElement.cpp b/Source/core/html/HTMLVideoElement.cpp
index 1fd3884597140f252920d6f5e0924fb1c446b25a..7d9d5aa88d6662af7533dbaf1f51e4856294ce32 100644
--- a/Source/core/html/HTMLVideoElement.cpp
+++ b/Source/core/html/HTMLVideoElement.cpp
@@ -38,6 +38,7 @@
 #include "core/html/parser/HTMLParserIdioms.h"
 #include "core/layout/LayoutImage.h"
 #include "core/layout/LayoutVideo.h"
+#include "platform/RuntimeEnabledFeatures.h"
 #include "platform/UserGestureIndicator.h"
 #include "platform/graphics/GraphicsContext.h"
 #include "platform/graphics/ImageBuffer.h"
@@ -244,6 +245,15 @@ bool HTMLVideoElement::webkitDisplayingFullscreen()
     return isFullscreen();
 }
 
+bool HTMLVideoElement::supportsOverlayFullscreenVideo() const
+{
+    // For WebRTC, the video frame contains all the data and no hardware surface
+    // should be used.
+    return !HTMLMediaElement::isMediaStreamURL(sourceURL().string())
+        && (RuntimeEnabledFeatures::overlayFullscreenVideoEnabled()
+            || (webMediaPlayer() && webMediaPlayer()->supportsOverlayFullscreenVideo()));

[trchen, 2015/07/30 21:29:21]

This condition feels weird to me. I'd prefer it that if
!RuntimeEnabledFeatures::overlayFullscreenVideoEnabled() we never enable the
overlay hack.

[watk, 2015/08/03 21:47:44]

I've added forceCverlayFullscreenVideo which can be used to force enable it for
virtual layout tests. I'll remove overlayFullscreenEnabled in a follow-up CL
once the Android WMP implementation is updated.

+}
+
 void HTMLVideoElement::didMoveToNewDocument(Document& oldDocument)
 {
     if (m_imageLoader)