Are popular classes more defect prone

Abstract. Traces of the evolution of software systems are left in a number of different repositories: such as configuration management systems, bug tracking systems, mailing lists. Developers use e-mails to discuss issues ranging from low-level concerns (bug fixes, refactorings) to high-level resolutions (future planning, design decisions). Thus, e-mail archives constitute a valuable asset for under-standing the evolutionary dynamics of a system. We introduce metrics that measure the “popularity ” of source code artifacts, i.e., the amount of discussion they generate in e-mail archives, and investigate whether the information contained in e-mail archives is correlated to the defects found in the system. Our hypothesis is that developers discuss problematic entities more than unproblematic ones. We also study whether the precision of existing tech-niques for defect prediction can be improved using our popularity metrics.