Implicit conversion from float -49.837398373983746 to int loses precision