You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I've tried to tackle some of the proposed challenges with a deductive verifier (Verus), following the VeriFast approach. While feasible, the problem is that in practice it's impossible to meet the requirements of challenges that impose a rather small TCB compared to what tools like Verus has. For example challenge 21 allows us to assume correctness of the slice module, but doesn't mention the rest of standard library, which is trusted for tools that implement compositional verification like Verus and Verifast; we would need facts about Strings, integer comparisons, iterations, strings.
I would like some guidance/thoughts from the challenge creators:
Accept the extra assumptions around other std-lib components for tools that perform compositional verification (e.g., if you are verifying Pattern you can assume String).
Re-think whether we need different challenges for these tools, such as verifying the simpler stdlib modules.
Activity
[-]Tool TCB's and challenges[/-][+]Trusted assumptions and challenges[/+]