Integrated Module Testing and Module Verification

Abstract

In this dissertation an integrated approach to formal module verification by model checking and module testing is described. The main focus lays on the verification of C functions. Specification-based testing and functional verification require a formalized module specification. For this purpose an annotation language as an extension of a pre-/post-condition syntax is developed and discussed. This annotation language allows the definition of logical conditions relating the program s pre-state to its post-state after executing the module. For requirements tracking a test case specification is developed. The correctness conditions can be refined by the introduction of auxiliary variables. Besides the specification of the module under test, the presented annotation language allows to model the behavior of external functions called by the module under test. By the specification of pre- and post-conditions as well as test cases, test data generation for both structural and functional testing is reduced to a reachability problem (as known from bounded model checking) within the module s control flow graph. These reachability problems are investigated using symbolic execution. The strength of symbolic execution is in its precision and its ability to reason about multiple program inputs simultaneously, but it also has limitations like aliasing or external function calls. These in turn are analyzed and new algorithms are developed which overtake most of the detected limitations. The expansion and selection strategies for test case selection are developed and described. They allow to minimize the size of investigated states and the number of generated test cases, while achieving maximal branch coverage. The developed algorithms and strategies are implemented in the test generator CTGEN, which generates test data for C1 structural coverage and for functional coverage. It also supports automated stub generation where the data returned by a stub during test execution depends on the specification provided by the user. CTGEN is evaluated and compared with competing tools and produces competitive results.