Implicit conversion from float -49.40677966101697 to int loses precision