Implicit conversion from float -1.1214953271027923 to int loses precision