On Navigation and Analysis of Software Architecture Evolution

Software systems must evolve during their lifetime in response to changing expectations and environments. As the software evolves, the system becomes harder to understand and maintain without the proper knowledge about how the system had changed in the past and the context of those changes. Studying software evolution has been extraordinarily costly and time consuming, as it lacks a sound theory, effective research approaches, as well as an integrated research environment. In this thesis, we present a research approach for studying software evolution, which incorporates evolution metrics, visualization of system change history, and a method of relating similar program entities between different releases in spite of changed name or location. To validate our approach, we have implemented a prototype research environment called BEAGLE to aid the software maintainer to understand how large software systems have evolved overtime. BEAGLE integrates data from various statistic tools and metric tools, and provides a query engine as well as a web-based visualization and navigation interface. BEAGLE aims to provide help in understanding the long-term evolution of systems that have undergone architectural and structural changes. We performed a case study on the evolution of GNU Compiler Collection (GCC) using a prototype implementation of BEAGLE. We were able to discover several interesting evolution characteristics of GCC, and to answer specific questions related to the evolution history of GCC, such as the relationship between the experimental EGCS release and traditional GCC release, and the different evolution patterns shown by GCC releases at different periods of time.