Implicit conversion from float -39.86577181208054 to int loses precision