Implicit conversion from float -26.111111111111114 to int loses precision