Implicit conversion from float -6.870554765291615 to int loses precision