Implicit conversion from float -24.125 to int loses precision