Implicit conversion from float -5.867507886435334 to int loses precision