Implicit conversion from float -5.22935779816514 to int loses precision