Reduce First, Debug Later

Abstract

The delta debugging minimization algorithm ddmin provides an efficient procedure for the simplification of failing test-cases. Despite its contribution towards the automation of debugging, ddmin still requires a significant number of iterations to complete. The delta debugging (DD) search space can be narrowed down by providing the test-case circumstances that are most likely relevant to the occurred failure. This paper proposes a novel approach to the problem of failure simplification consisting of two consecutive phases: 1) failure reduction by rewriting (performed offline), and 2) DD invocation (performed online). In the best case scenario, the reduction phase may already deliver a simplified failure, otherwise, it potentially supplies DD with extra information about where to look for the failure. The proposed solution has been prototyped as a web application debugging tool, which was evaluated on a shopping cart web application - Flex Store. The evaluation shows an improvement of the DD execution time if the offline reduction over-approximates the failure.