Weighted Execution Profiles for Software Testing

Existing execution profiles comprise information about the execution of program elements such as statements, branches, and def-use pairs. They lack any information about what elements are potentially more relevant to failure than others, such information could be leveraged to compute dissimilarity metrics that discriminate better between failing and passing profiles, which is beneficial to software testing. In this work, we explore three heuristics from which we will derive weights to be associated with the covered program elements such that higher weights indicate more potential relevance to failure. We evaluate our heuristics by measuring their impact on an established test suite minimization technique. Our experiments showed somewhat promising results.