Coevolution of Software and Tests: an Initial Assessment

Unit testing is the first line of defence against software failure. To make the most of this technique the test code should evolve simultaneously with the product code. First, this dissertation explores the possibilities of using dynamic analysis to extract test dependencies. Then we investigate whether heuristic metrics on these dynamic test dependencies provide a measure for the degree to which the test code evolves with the product code. As a case study we use Apache Ant and look specifically at two different phases in the history of this open source project. We conclude that dynamic test dependencies alone do not suffice to provide such a measure and we propose an alternative solution.