Simmo Saan

Traditional, or concrete, debuggers allow developers to step through programs and explore the corresponding concrete program states—developers can query current values of program variables. This exploration enables developers to formulate and refine hypotheses about program behaviors. We propose the novel notion of abstract debuggers, which allow developers to explore abstract program states, as computed by sound static analyzers. Giving developers the ability to interactively explore abstract states empowers them to work with hypotheses that are true for all program executions: they can examine and rule out false positives, or better understand a static analysis’s declaration that some code is indeed safe. Abstract debuggers’ interfaces, reminiscent of conventional debuggers, aim to make navigating and interpreting static analysis results more straightforward. We have formalized the concept, applied it by implementing a tool that leverages the static analyzer Goblint, and illustrate its usefulness through case studies.

Goblint is an abstract interpretation framework for C programs with a specialty in concurrency. Using a novel approach, we turn it into a validator of YAML correctness witnesses for all SV-COMP categories. We describe its results at SV-COMP 2024 which includes the first large-scale evaluation of our validator.

Witnesses record automated program analysis results and make them exchangeable. To validate correctness witnesses through abstract interpretation, we introduce a novel abstract operation unassume. This operator incorporates witness invariants into the abstract program state. Given suitable invariants, the unassume operation can accelerate fixpoint convergence and yield more precise results. We demonstrate the feasibility of this approach by augmenting an abstract interpreter with unassume operators and evaluating the impact of incorporating witnesses on performance and precision. Using manually crafted witnesses, we can confirm verification results for multi-threaded programs with a reduction in effort ranging from 7% to 47% in CPU time. More intriguingly, we discover that using witnesses from model checkers can guide our analyzer to verify program properties that it could not verify on its own.