Towards scalable compositional test generation

One difficulty of automated test case generation is to deal with compositional units that brings in compositional space explosion of program states. We present a new dynamic execution framework which analyzes program behaviors dynamically for automatic test inputs generation. We utilize forward slicing to explore those functions affecting conditional predicates in program under test. The functions that do not affect the conditional predicates are not in need of being analyzed symbolically. Pointer alias analysis is adopted to make slicing in the presence of pointers more precise. A dynamic partial execution technique is proposed to accelerate the speed of searching the unit space. The proposed approach can be applied to real programs and the experiments are also very encouraging.