ClipRRectLayer (the round rect version) and ClipPathLayer both use the content bounds for the bounds of the saveLayer when the antiAliasWithSaveLayer flag is used, but ClipRectLayer uses its own clip rect. See:

• https://github.com/flutter/engine/blob/1b10b146cf7c73cf4e543e6762e479eb1b023f1e/flow/layers/clip_rect_layer.cc#L77
• https://github.com/flutter/engine/blob/1b10b146cf7c73cf4e543e6762e479eb1b023f1e/flow/layers/clip_rrect_layer.cc#L78
• https://github.com/flutter/engine/blob/1b10b146cf7c73cf4e543e6762e479eb1b023f1e/flow/layers/clip_path_layer.cc#L78

The clip rect is likely smaller than the bounds of the children, but it doesn't have to be so the saveLayer may be larger than necessary.

All of these layers execute the saveLayer after they set the clip so the resulting saveLayer bounds should end up being the intersection of the clip with the bounds used in the saveLayer (this should be verified), so using the content bounds in the saveLayer should not waste memory.