Implicit conversion from float -17.53311258278147 to int loses precision