Implicit conversion from float -8.618784530386733 to int loses precision