Research On The Technique Of Automated Software Repair Based On The Similar Code

With the growing of software size,the costs of software maintenance have raised a lot.The techniques of software testing and debugging can reduce the effort of maintenance,but the bugs hidden in the software are still the burdens for developers.With little or none assistance from developers,the technique of program repair can fix the bugs by reusing existing code,program synthesis and other methods.Recently,much attention has been paid on reusing code and reusing fix operations from other software.Compared to the technique of reusing the fix operations,reusing code can fix the bug with no need of fix operations from historical versions of software.The similar code can be used to fix the bug.However,this technique is limited by the huge searching space of fix ingredients from the similar code.Finding out those similar code fragments which contain proper fix ingredients efficiently will improve a lot the efficiency of repair.This paper proposes reusability metrics of similar codes for program repair,which combine the similarity and difference from program syntax tree to pick out the most suitable repair candidate.After analysing the automated repair tools of reusing code,we find that they are not guided by the efficent technique of locating the fix ingrediants to reducing the search space.We have implemented a prototype tool based on RSRepair,which can guided by different techniques of fix localization.With efficent techniques of fix localization,RSRepair can improve the efficiency of repair.