Use inner HWND for accelerated rendering on windows

chrome/renderer/render_widget.cc

 // Copyright (c) 2010 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.
 
 #include "chrome/renderer/render_widget.h"
 
 #include "app/surface/transport_dib.h"
 #include "base/command_line.h"
 #include "base/logging.h"
 #include "base/message_loop.h"
@@ skipping 505 matching lines @@
     copy_rects.swap(update.paint_rects);
     if (!scroll_damage.IsEmpty())
       copy_rects.push_back(scroll_damage);
 
     for (size_t i = 0; i < copy_rects.size(); ++i)
       PaintRect(copy_rects[i], bounds.origin(), canvas.get());
 
     dib_id = current_paint_buf_->id();
   } else {  // Accelerated compositing path
     // Begin painting.
-    bool finish = next_paint_is_resize_ack();
-    webwidget_->composite(finish);
+    webwidget_->composite(false);
   }
 
   // sending an ack to browser process that the paint is complete...
   ViewHostMsg_UpdateRect_Params params;
   params.bitmap = dib_id;
   params.bitmap_rect = bounds;
   params.dx = update.scroll_delta.x();
   params.dy = update.scroll_delta.y();
   if (is_accelerated_compositing_active_) {
     // If painting is done via the gpu process then we clear out all damage
@@ skipping 83 matching lines @@
   // 1) Ensures that we call WebView::Paint without a bunch of other junk
   //    on the call stack.
   // 2) Allows us to collect more damage rects before painting to help coalesce
   //    the work that we will need to do.
   MessageLoop::current()->PostTask(FROM_HERE, NewRunnableMethod(
       this, &RenderWidget::CallDoDeferredUpdate));
 }
 
 void RenderWidget::didActivateAcceleratedCompositing(bool active) {
   is_accelerated_compositing_active_ = active;
-  Send(new ViewHostMsg_GpuRenderingActivated(
+  Send(new ViewHostMsg_DidActivateAcceleratedCompositing(
     routing_id_, is_accelerated_compositing_active_));
 }
 
 void RenderWidget::scheduleComposite() {
   // TODO(nduca): replace with something a little less hacky.  The reason this
   // hack is still used is because the Invalidate-DoDeferredUpdate loop
   // contains a lot of host-renderer synchronization logic that is still
   // important for the accelerated compositing case. The option of simply
   // duplicating all that code is less desirable than "faking out" the
   // invalidation path using a magical damage rect.
@@ skipping 340 matching lines @@
 
 void RenderWidget::CleanupWindowInPluginMoves(gfx::PluginWindowHandle window) {
   for (WebPluginGeometryVector::iterator i = plugin_window_moves_.begin();
        i != plugin_window_moves_.end(); ++i) {
     if (i->window == window) {
       plugin_window_moves_.erase(i);
       break;
     }
   }
 }