Abstract:
Researchers have hypothesized that if we could estimate the probability that
a fault in a code component will a cause a failure, we could use this estimate to
improve the fault-detection effectiveness of code-coverage-based testing. If this
hypothesis could be supported, it would motivate further research in this area
and could lead to techniques that would help testers distribute testing resources
more effectively and improve the quality of testing. In this research, we developed
a new test adequacy criterion, which incorporates fault exposure potential
estimates into statement-coverage requirements. We conducted empirical studies
to investigate the fault-detection effectiveness of this new criterion. The
results of our studies show that the incorporation of fault exposure potential
information into the statement-coverage test adequacy criterion can indeed improve
its fault-detection effectiveness. However, the effects of incorporating the
estimates vary with the program under test, the nature of faults contained in
the program, and the level of confidence required of the testing. The overall
improvements in fault-detection effectiveness that we observed under our initial
approach were not as large as we might wish, but by improving the method for
estimating fault exposure potential we can obtain better results. The results of
this research provide impetus for future studies in this area.
