Implicit conversion from float -7.4342928660826 to int loses precision