Fix for 424682: Draw correct repaint regions instead of a conservative bounding box

cc/layers/picture_layer_impl.cc

Index: cc/layers/picture_layer_impl.cc
diff --git a/cc/layers/picture_layer_impl.cc b/cc/layers/picture_layer_impl.cc
index e819afd860d412784301915fe1fcbd7ac857aa67..47cf03548870d8b01b8d2b9c865a08e5a877be7a 100644
--- a/cc/layers/picture_layer_impl.cc
+++ b/cc/layers/picture_layer_impl.cc
@@ -605,6 +605,20 @@ skia::RefPtr<SkPicture> PictureLayerImpl::GetPicture() {
   return raster_source_->GetFlattenedPicture();
 }
 
+Region PictureLayerImpl::GetInvalidationRegion() {
+  // PictureLayerImpl::invalidation_ is not cleared in the same was as

[danakj, 2015/03/02 20:10:53]

I think I'd reword this a bit.. "truth" is subjective :) Maybe this:

|invalidation_| gives the invalidation contained in the source frame, but is not
cleared after drawing from the layer. However, update_rect() is cleared once the
invalidation is drawn, which is useful for debugging visualizations. This method
intersects the two to give a more exact representation of what was invalidated
that is cleared after drawing.

+  // LayerImpl::update_rect_ is and thus might not give the true state of the
+  // current invalidations at all time.
+  // Ideally, update_rect_ of LayerImpl would be replaced completely by
+  // invalidation_ region as it give thighter bounds but this might be
+  // to much overhead based on where update_rect_ is used today.
+  // That's why we intersect invalidation_ with the update_rect to get the
+  // true invalidations of the present frame.
+  Region current_invalidation(invalidation_);

[danakj, 2015/03/02 20:10:53]

you can do this whole thing as

return IntersectRegions(invalidation_, update_rect());

[daplatz, 2015/03/02 20:28:02]

Acknowledged.

+  current_invalidation.Intersect(update_rect());
+  return current_invalidation;
+}
+
 scoped_refptr<Tile> PictureLayerImpl::CreateTile(
     float contents_scale,
     const gfx::Rect& content_rect) {