Implicit conversion from float -9.840000000000003 to int loses precision