Implicit conversion from float -99.96247654784241 to int loses precision