Implicit conversion from float -41.12582781456953 to int loses precision