Implicit conversion from float -2.8406466512702053 to int loses precision