Implicit conversion from float -40.3475935828877 to int loses precision