Implicit conversion from float -108.89502762430939 to int loses precision