Implicit conversion from float -10.084459459459453 to int loses precision