Implicit conversion from float -1.8302387267904408 to int loses precision