Currently nsDisplaySubDocument passes mVisibleRect to RecordFrameMetrics. mVisibleRect, like GetBounds, is the final visible rect after composited the scrolled content and then clipping to the scrollport (ie a subset of the scrollport). But we want the visible rect passed to RecordFrameMetrics to be all of the pre-drawn scrolled content in the displayport, like the other RecordFrameMetrics callers. This value is stored on the visible rect of the enclosed display list (mVisibleRect is the visible rect of the item). This doesn't actual seem to cause problems, probably because the important part is that the layers that actually contain the pixels for the scrolled content get the expanded visible rect. But at the very least we should be consistent.