Implicit conversion from float -17.90513833992094 to int loses precision