Implicit conversion from float -5.6487341772152035 to int loses precision