Test-to-code traceability links model the relationships between test artefacts and code artefacts. When utilised during the development process, these links help developers to keep test code in sync with tested code, reducing the rate of test failures and missed faults. Test-to-code traceability links can also help developers to maintain an accurate mental model of the system, reducing the risk of architectural degradation when making changes. However, establishing and maintaining these links manually places an extra burden on developers and is error-prone. This paper presents TCtracer, an approach and implementation for the automatic establishment of test-to-code traceability links. Unlike existing work, TCtracer operates at both the method level and the class level, allowing us to establish links between tests and functions, as well as between test classes and tested classes. We improve over existing techniques by combining an ensemble of new and existing techniques and exploiting a synergistic flow of information between the method and class levels. An evaluation of TCtracer using four large, well-studied open source systems demonstrates that, on average, we can establish test-to-function links with a mean average precision (MAP) of 78% and test-class-to-class links with an MAP of 93%.

Commercial entities have started using peer-to-peer (P2P) algorithms for file distribution. For example, Sub Pop Records and Blizzard Entertainment use BitTorrent [2] to distribute large files to their clients. Given that last-mile bandwidths to home users are increasing rapidly, a P2P approach is perhaps the only viable solution to avoid a bottleneck at the servers. Furthermore, the fact that P2P algorithms exploit the bandwidth available to the clients and reduces the provisioning costs at the server, makes P2P even more attractive.