Section: New Results

Automated Software Repair

Participant : Martin Monperrus [correspondant] .

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. In [15] , we present some major results in this direction by mining fix transactions of existing software repositories. From the empirical study of 14 software repositories containing 89,993 versioning transactions, we show that we can learn a probability distribution of repair actions. We show that certain distributions over repair actions can result in an infinite time (in average) to find a repair shape while other fine-tuned distributions enable to find a repair shape in hundreds of repair attempts. We now aim at going beyond this empirical study and theoretical analysis by exploring how to use this learned knowledge for new software systems.