Implicit conversion from float -9.92867332382312 to int loses precision