Implicit conversion from float -49.18558077436583 to int loses precision