Implicit conversion from float -9.865771812080538 to int loses precision