Implicit conversion from float -108.91944990176816 to int loses precision