Much effort has been expended in attempting to discover how errors are introduced into software, with the hope of being able to reduce the potentially serious faults occurring in applications. One area of interest is clones--fragments of code that are repeated within a piece of software. It has long been recognized that sections of repeated code can cause problems when one section is changed to correct an error while a repeated section is not updated; as a result, the repeated section still contains the uncorrected problem. It is also common for clones to change in separate ways (inconsistently) during code development.

Previous work has studied how individual clones change during an application’s development and how to identify unwanted inconsistencies. The work reported here tries to go one step further by investigating consecutive changes to clones over a substantial period (five years) of the lifetime of a number of applications. The authors provide a systematic description of consecutive changes to clones and go on to analyze three applications with a full Subversion history. They then use one application (a tool they had written to identify and track clones in software) to see whether they can identify any changes that led to undiscovered errors.

As the case study failed to provide any strong conclusions, this paper should be regarded as a first step; therefore, it’s primarily only of interest to researchers in this area. A less specialized reader would have benefited greatly if the example given near the end of the paper had been brought forward and described in more detail.