Implicit conversion from float -115.26315789473685 to int loses precision