Implicit conversion from float -29.140625 to int loses precision