Empirical evaluation of the textual differencing regression testing technique

Abstract

Regression testing is a commonly used activity whose purpose is to determine whether the modifications made to a software system have introduced new faults. Textual differencing is a new, safe and fairly precise, selective regression testing technique that works by comparing source files from the old and the new version of the program. We have implemented the textual differencing technique in a tool called Pythia. Pythia has been developed primarily through the integration of standard, well known UNIX programs, and is capable of analyzing large software systems written in C. We present results from a case study involving a software system of approximately 11,000 lines of source code written for the European Space Agency. The results provide empirical evidence that textual differencing is very fast and capable of achieving substantial reductions in the size of the regression test suite.