Implicit conversion from float -105.45454545454544 to int loses precision