The trick generally is to break programs into pieces and...

The trick generally is to break programs into pieces and have those pieces be individually testable and so then when you move on to the other pieces you treat it as a black box knowing that it either works or doesn't work.