Distributed Assertion Checking Using Symbolic Execution

Abstract

Annotating functional correctness properties of code using assertions, in principle, enables systematic checking of code against behavioral properties. In practice however, checking assertions can be costly, especially for complex code annotated with rich behavioral properties. This paper introduces a novel approach for distributing the problem of checking assertions for better scalability. Leveraging that assertions should be side effect free, our approach distributes assertion checking into simpler sub-problems---each focusing on checking one single assertion, so that different assertions are checked in parallel among multiple workers. Furthermore, the sub-problem analysis performed by each worker is guided by the checked assertion to avoid irrelevant path exploration and is prioritized based on the distance towards the checked assertion to provide earlier feedback. A case study shows that our approach can provide a reduction in analysis time required for symbolic execution of Java programs compared to non-distributed approach using the Symbolic PathFinder tool.

Parallel Program Analysis via Range Splitting

“…A common strategy for dividing the search space in sequential or interleaved combinations is to restrict the subsequent verifiers to the yet uncovered search space, e.g., not yet covered test goals [12], open proof obligations [67], or yet unexplored program paths [8,10,19,31,33,41,42,47,53,71]. Some parallel combinations like CoDiDroid [80], distributed assertion checking [93], or the compositional tester sketched in conditional testing [12] decompose the verification statically into separate subtasks. Furthermore, some techniques split the search space to run different instances of the same analysis in parallel on different parts of the program.…”