UW - CSE - 11 - 05 - 02 2 nd May 2011 ( Revised 5 th October 2011 )

Developers often copy, or clone, code in order toreuse or modify functionality. When they do so, they also clone any bugs in the original code. Or, different developers mayindependently make the same mistake. As one example of abug, multiple products in a product line may use a componentin a similar wrong way. This paper makes two contributions. First, it presents an empirical study of cloned buggy code. In alarge industrial product line, about 4% of the bugs areduplicated across more than one product or file. In three opensource projects (the Linux kernel, the Git version control system, and the PostgreSQL database) we found 282, 33, and33 duplicated bugs, respectively. Second, this paper presents atool, CBCD, that searches for code that is semanticallyidentical to given buggy code. CBCD tests graph isomorphism over the Program Dependency Graph (PDG) representationand uses four optimizations. We evaluated CBCD by searchingfor known clones of buggy code segments in the three projectsand compared the results with text-based, token-based, and AST-based code clone detectors, namely Simian, CCFinder,and CloneDr. The results of the evaluation show that CBCD isapplicable for its principal use: it is fast when searching forpossible clones of the buggy code in a large system and it is more precise than the other code clone detectors. KeywordsValidation, Debugging aids