Using automated source code analysis for software evolution

Software maintenance is one of the most expensive and time-consuming phases in the software life-cycle. The size and complexity of commercial applications probably present the greatest difficulty that maintainers face when making changes to their applications. As a result of the corresponding loss of understanding, business knowledge encapsulated within the system becomes fragmented, and any changes made as a result of new business initiatives become difficult to implement and hence may mean a loss of business opportunities. The paper outlines an approach to regaining understanding of software which has been used in the Release project at Durham University. This approach involves determining the calling structure of a program in terms of a call-graph, and from this call-graph extracting a dominance tree. Various problems which have been encountered during the construction of tools to perform this task are described.