Implicit conversion from float -12.52387448840382 to int loses precision