Implicit conversion from float -12.222222222222229 to int loses precision