Implicit conversion from float -46.388888888888886 to int loses precision