Implicit conversion from float -36.339969372128635 to int loses precision