Abstract Path Testing with PathCrawler

摘要

PathCrawler is a tool developed by CEA List for the automatic generation of test inputs to ensure the coverage of all feasible execution paths of a C function. Due to its concolic approach and depth-first exhaustive search strategy implemented in Prolog, PathCrawler is particularly efficient in the generation of tests to cover the fully expanded tree of feasible paths. However, for many tested functions this coverage criterion demands too many tests and a weaker criterion must be used. In order to efficiently generate tests for a new criterion whilst still using a concolic approach, we must modify the search strategy. To facilitate the definition and comparison of different coverage criteria, we propose a new type of tree, trees of abstract paths, and define the different types of abstract node in these trees. We demonstrate how several criteria can be conveniently defined in terms of coverage of these new trees. Moreover, efficient generation of tests to satisfy these criteria using the concolic approach can be designed as different strategies to explore these trees.