Implicit conversion from float -40.81967213114754 to int loses precision