Section: New Results

Extraction and Analysis of Knowledge for Automatic Software Repair

Automated software repair aims at assisting developers in order to improve the quality of software systems, for example by recommending some repair actions to fix bugs. Matias Martinez has presented in his PhD thesis [13] that was defended in June 2014, new results in this domain. These results aim at reducing the search space when repairing a software system. The solution relies on two techniques. The first one consists in building change models learnt from repairs performed by other developers. These repairs are mined from existing software repositories of open source projects, and analysed based on their types and frequencies. The second proposed technique is based on the inherent redundancy of code patterns. The assumption is that the probability that the repair code for a particular kind of defect is already present in the software system under study is high. We then take advantage of this inherent redundancy to reduce the search space when looking for repair actions.