Transforming programs and tests in tandem for fault localization

• DOI: 10.1145/3133916
• Publication Date: 2017-10-13T15:15:45Z
• Authors: Xia Li; Lingming Zhang
• Keywords: Benchmark (surveying); Code (set theory); Test suite

Localizing failure-inducing code is essential for software debugging. Manual fault localization can be quite tedious, error-prone, and time-consuming. Therefore, a huge body of research e orts have been dedicated to automated localization. Spectrum-based localization, the most intensively studied approach based on test execution information, may limited effectiveness, since element executed by failed tests not necessarily impact outcome cause failure. To bridge gap, mutation-based has proposed transform programs under check each better However, there are studies effectiveness sufficient number real bugs. In this paper, we perform an extensive study compare techniques with various state-of-the-art spectrum-based 357 bugs from Defects4J benchmark suite. The results firstly demonstrate as well revealing guidelines further improving Based learnt guidelines, outputs/messages obtain mutation information. Then, propose TraPT, Learning-to-Rank technique fully explore obtained information effective experimental show that TraPT localizes 65.12% 94.52% more within Top-1 than spectrum when using default setting LIBSVM.