Implicit conversion from float -28.974358974358978 to int loses precision