Implicit conversion from float -49.91323210412148 to int loses precision