Implicit conversion from float -289.746835443038 to int loses precision