A C++ data model supporting reachability analysis and dead codedetection

摘要

A software repository provides a central information source fornunderstanding and reengineering code in a software project. Complexnreverse engineering tools can be built by analyzing information storednin the repository without reparsing the original source code. The mostncritical design aspect of a repository is its data model, which directlynaffects how effectively the repository supports various analysis tasks.nThis paper focuses on the design rationales behind a data model for anC++ software repository that supports reachability analysis and deadncode detection at the declaration level. These two tasks are frequentlynneeded in large software projects to help remove excess softwarenbaggage, select regression tests and support software reuse studies. Thenlanguage complexity introduced by class inheritance, friendship, andntemplate instantiation in C++ requires a carefully designed model toncatch all necessary dependencies for correct reachability analysis. Wenexamine the major design decisions and their consequences in our modelnand illustrate how future software repositories can be evaluated forncompleteness at a selected abstraction level. Examples are given tonillustrate how our model also supports variants of reachabilitynanalysis: impact analysis, class visibility analysis, and dead codendetection. Finally, we discuss the implementation and experience of ournanalysis tools on a few C++ software projects