Multiply-deployed residual testing at the object level

Software testing is a proven technique widely used to increase confidence that a software product behaves as expected and to gather system usage information to support evolution. Many structural coverage criteria techniques are used to perform and measure testing activities, but their complete satisfaction is rarely achieved resulting in code release with neglected test obligations. This paper suggests a remedy for this situation – monitor the software after release to measure coverage criteria (either by beta testers or by end users). Our approach explores residual testing in multiple deployments of an application to determine incrementally but continually how well selected coverage criteria are satisfied. We introduce a prototype and demonstrate the value of the approach to real-world applications.