Secrets from the Monster: Extracting Mozilla’s Software Architecture

As large systems evolve, their architectural integrity tends to decay. Reverse engineering tools, such as PBS [7, 19], Rigi [15], and Acacia [5], can be used to acquire an understanding of a system’s “as-built” architecture and in so doing regain control over the system. A problem that has impeded the widespread adoption of reverse engineering tools is the tight coupling of their subtools, including source code “fact” extractors, visualization engines, and querying mechanisms; this coupling has made it difficult, for example, for users to employ alternative extractors that might have different strengths or understand different source languages. The TAXFORM project has sought to investigate how different reverse engineering tools can be integrated into a single framework by providing mappings to and from common data schemas for program “facts” [2]. In this paper, we describe how we successfully integrated the Acacia C and C++ fact extractors into the PBS system, and how we were then able to create software architecture models for two large software systems: the Mozilla web browser (over two million lines of C++ and C) and the VIM text editor (over 160,000 lines of C).