Implicit conversion from float -39.299363057324825 to int loses precision