On stability in the saddle-point sense