Implicit conversion from float -45.68329718004338 to int loses precision