Implicit conversion from float -17.676348547717836 to int loses precision