Implicit conversion from float -107.6510067114094 to int loses precision