Implicit conversion from float -6.76781002638522 to int loses precision