Implicit conversion from float -8.794416243654823 to int loses precision