Previous work on tracking source locations has focused on tracking lines through multiple revisions of software. In this paper, we explore a new technique for tracking statements, rather than lines, across multiple revisions of Java source code. We show that our statement-tracking technique achieves comparable accuracy for source code than the most accurate line-tracking techniques, while also safely handling all non-executable formatting changes, such as breaking a single statement across many lines, adding or removing whitespace, moving brackets, or re-ordering methods. Finally, we compare the performance of three of the current state-of-the-art techniques for tracking lines across revisions on a series of benchmarks, and discuss the strengths and weaknesses of each technique.

Two of the most complex operations version control software allows a user to perform are branching and merging. Branching provides the user the ability to create a copy of the source code to allow changes to be stored in version control but outside of the trunk. Merging provides the user the ability to copy changes from a branch to the trunk. Performing a merge can be a tedious operation and one that may be error prone. In this paper, we compare file revisions found on branches with those found on the trunk to determine when a change that is applied to a branch is moved to the trunk. This will allow us to study how developers use merges and to determine if merges are in fact more error prone than other commits.

We describe a method to use the source code change history of a software project to drive and help to refine the search for bugs. Based on the data retrieved from the source code repository, we implement a static source code checker that searches for a commonly fixed bug and uses information automatically mined from the source code repository to refine its results. By applying our tool, we have identified a total of 178 warnings that are likely bugs in the Apache Web server source code and a total of 546 warnings that are likely bugs in Wine, an open-source implementation of the Windows API. We show that our technique is more effective than the same static analysis that does not use historical data from the source code repository.