Implicit conversion from float -17.539118065433854 to int loses precision