Implicit conversion from float -85.68093385214007 to int loses precision