Implicit conversion from float -129.8198198198198 to int loses precision