On Measuring Combinatorial Coverage of Manually Created Test Cases for Industrial Software

Combinatorial coverage has been proposed as a way to measure the quality of test cases by using the input interaction characteristics. This paper describes the results of empirically measuring combinatorial coverage of manually created test cases by experienced industrial engineers working with embedded software development. We found that manual test cases achieve on average 78% 2-way combinatorial coverage, 57% 3-way coverage, 40% 4-way coverage, 20% 5-way combinatorial coverage and 13% for 6-way combinatorial coverage. These manual test cases can be augmented to achieve 100% combinatorial coverage for 2-way and 3-way interactions by adding eight and 66 missing test cases on average, respectively. For 4-way interactions, full combinatorial coverage can achieved by adding 658 missing test cases. For 5-way and 6-way interactions, full combinatorial coverage can be achieved by adding 5163 and 6170 missing test cases on average, respectively. The results of this paper suggest that manual test cases created by industrial engineers do not achieve a high combinatorial coverage and can be improved by adding more test cases to cover t-wise interactions at the expense of more test cases to execute.