A case study in branch testing automation

We present a case study for the empirical validation of the usefulness of two recent research results aimed at improving the branch testing process. The results considered consist of (i) a method for the automatic generation of test paths, and (ii) a bound on the number of tests needed. The validation was conducted within the unit test phase of the software test process of Ericsson Telecomunicazioni on object oriented C++ software, which was developed to control a new generation of telecommunication systems. The outcomes confirmed the validity of the path generation method, in that all the derived paths were found feasible. The case study also provided evidence that this method can reduce the test effort, especially for non-expert testers, though there was not a sufficient amount of results to allow us to quantify this reduction. We provide a statistical validation of the performance of the bound. It proved to be a good estimate of the number of tests needed to achieve branch coverage. © 1997 by Elsevier Science Inc.