Implicit conversion from float -91.41025641025641 to int loses precision