Implicit conversion from float -21.411992263056092 to int loses precision