Proof Reuse

Abstract

Experience shows that the prevalent use case of program verification systems is not a single prover run. It is far more likely that a proof attempt fails, and that the program (and/or the specification) has to be revised. Then, after a small change, it is better to adapt and reuse the existing partial proof than to verify the program again from first principles. A particular advantage is that proof reuse can reduce the number of required user interactions.

Here we present such a technique for proof reuse. In fact, towards the end of this chapter ($\Rightarrow$ Sect. 13.9), we will show how our method can improve the user experience for a whole range of verification scenarios. Until then, we limit ourselves to the setting described above, with the further assumption that only the implementation changes and the specification remains unchanged.

After discussing the features of the method, we will introduce a small running example, cover the theoretical and practical details of proof reuse, examine other solutions to the problem, and finally survey the full range of proof reuse applications in deductive verification of Java software.