Delegate root layer scroll offset to android_webview.

android_webview/browser/in_process_view_renderer.h

Index: android_webview/browser/in_process_view_renderer.h
diff --git a/android_webview/browser/in_process_view_renderer.h b/android_webview/browser/in_process_view_renderer.h
index 4175dd1cdeeb6239a505eeb6a5b98bdc7ccbf774..be876649c5c10803c7987b9ebb26b8b8b0c510c1 100644
--- a/android_webview/browser/in_process_view_renderer.h
+++ b/android_webview/browser/in_process_view_renderer.h
@@ -53,6 +53,8 @@ class InProcessViewRenderer : public BrowserViewRenderer,
   virtual void DidDestroyCompositor(
       content::SynchronousCompositor* compositor) OVERRIDE;
   virtual void SetContinuousInvalidate(bool invalidate) OVERRIDE;
+  virtual void SetTotalRootLayerScrollOffset(gfx::Vector2dF new_value) OVERRIDE;
+  virtual gfx::Vector2dF GetTotalRootLayerScrollOffset() OVERRIDE;
 
   void WebContentsGone();
 
@@ -91,6 +93,9 @@ class InProcessViewRenderer : public BrowserViewRenderer,
   // Last View scroll before hardware rendering is triggered.
   gfx::Point hw_rendering_scroll_;
 
+  // TODO(mkosiba): Plumb through to Java
+  gfx::Vector2dF root_layer_scroll_offset_;

[joth, 2013/06/06 20:19:32]

I'm tempted to name this next_frame_scroll_offet_

c.f. ContentViewCore always carries info about the last frame. We'd like this
guy to be copied down java->native at the start of the frame, and remain our
authoritative source until the frame is complete (when we copy it back to java
if it changed).
the dance is avoid doing JNI in the middle of the draw functor, which is
apparently Not Cool.

If you want to mention this is root-layer scroll offset you can do that in a
comment, but at this level in the stack it's basically the only scroll we care
about so I don't think the differentiation is needed in the name

(The TODO is fine in just the .cc, don't think it needs repeating here.)

[mkosiba (inactive), 2013/06/07 16:18:27]

Calling it next_frame_scroll_offset_ would be a bit misleading (at least to me)
because it might imply that once the value is set it doesn't change till the
next frame.

AS far as not making any JNI in the draw functor - I think we should 1) add
DCHECKS in the JNI layers to enforce this and, 2) see if it actually happens.
Also, in SW mode, what are you going to do about the guy who thought it's a good
idea to change the scroll offset from his onDraw implementation?
Ok.
Ok.

+
   base::WeakPtrFactory<InProcessViewRenderer> weak_factory_;
 
   DISALLOW_COPY_AND_ASSIGN(InProcessViewRenderer);