The Impossibility of Complete Testing
Cem Kaner · 1998
This paper explores three themes: 1. I think that I've figured out how to explain the impossibility of complete testing to managers and lawyers, with examples that they can understand. These are my notes. 2. A peculiar breed of snake-oil sellers reassure listeners that you achieve complete testing by using their coverage monitors. Wrong. Complete line and branch coverage is not complete testing. It will miss significant classes of bugs. 3. If we can't do complete testing, what should we do? It seems to me that at the technical level and at the legal level, we should be thinking about "good enough testing," done with as part of a strategy for achieving "good enough software." Tutorial on Impossibility You probably know this material, but you might find these notes useful for explaining the problem to others. Complete testing is impossible for several reasons: . We can't test all the inputs to the program