> The only completely accurate semantic check is to run > the program and see if it produces the right result. If that were possible, we could solve the halting problem :-) > In other words, don't LBYL, but Just Do It, and use > unit tests. A fine idea when it's possible. Unfortunately, it's not always possible.