Implicit conversion from float -32.69938650306749 to int loses precision