Implicit conversion from float -107.80219780219778 to int loses precision