Integrated module testing and module verification

In this dissertation an integrated approach to formal module verification by model checking andmodule 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 purposean 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-stateto its post-state after executing the module. For requirements tracking a test case specification isdeveloped. 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 modelthe behavior of external functions called by the module under test.By the specification of preand post-conditions as well as test cases, test data generation for bothstructural and functional testing is reduced to a reachability problem (as known from bounded modelchecking) within the module’s control flow graph. These reachability problems are investigated usingsymbolic execution. The strength of symbolic execution is in its precision and its ability to reasonabout multiple program inputs simultaneously, but it also has limitations like aliasing or externalfunction calls. These in turn are analyzed and new algorithms are developed which overtake most ofthe detected limitations. The expansion and selection strategies for test case selection are developedand described. They allow to minimize the size of investigated states and the number of generatedtest cases, while achieving maximal branch coverage.The developed algorithms and strategies are implemented in the test generator CTGEN, whichgenerates test data for C1 structural coverage and for functional coverage. It also supports automatedstub generation where the data returned by a stub during test execution depends on the specifica-tion provided by the user. CTGEN is evaluated and compared with competing tools and producescompetitive results.