The quality of test suites is commonly measured using adequacy metrics that focus on error detection, like test coverage. However, the diagnostic performance of spectrum-based fault localization techniques, that can potentially reduce the time spent on debugging, rely on diagnosa
...
The quality of test suites is commonly measured using adequacy metrics that focus on error detection, like test coverage. However, the diagnostic performance of spectrum-based fault localization techniques, that can potentially reduce the time spent on debugging, rely on diagnosability of test suites --- the property of faults to be easily and precisely located. Therefore, in prior work, Perez et al. proposed a new metric, called DDU, that measures the diagnosability of test suites. However, DDU is not yet usable in practice due to its output value between 0 and 1. A developer would not know what test to write next given a certain DDU value. In this study, we explore the performance of DDU in practice by analyzing open source projects. We find no evidence that DDU is correlated to diagnosability and, thus, DDU is currently only useful when combined with test generation techniques.
