Implicit conversion from float -49.45827232796486 to int loses precision