Implicit conversion from float -68.61111111111111 to int loses precision