Implicit conversion from float -25.8041958041958 to int loses precision