Implicit conversion from float -43.80000000000001 to int loses precision