Implicit conversion from float -4.30294906166219 to int loses precision