SHF: Small: Reverse Engineering Variability Implementations

点击翻译按钮获取中文摘要

Variability in software systems is very common and important to tailorthe system to user needs and to use the system flexibly in differentcontexts without developing a new system from scratch for each context.For example, embedded systems with resource constraints,high-performance optimizations, and end-user customizable systems allexploit variability. However variability in software systems creates afundamental tension between (a) reuse and (b) development andmaintenance costs. On the one hand, planning variability upfront enablessystematic reuse and promises to significantly reduce developmenteffort, costs, and time to market. On the other hand, variabilityrequires an upfront investment and causes long-term maintenance costs,because variations give rise to an exponential number of configurations with potentially intricate interactions. Especially if introduced inan ad-hoc fashion and not managed appropriately, variability canincrease maintenance costs to a level that outweighs the expectedbenefits. This research develops and integrates reverse-engineeringtechniques for variability to lower upfront investments for variabilityand to reduce the costs of long-term maintenance caused by accidentalcomplexity of existing ad-hoc variability implementations.This project revisits how variable and reusable software is implementedby encouraging lightweight and incremental implementation mechanisms andproviding migration mechanisms for all ad-hoc legacy implementationscommon in practice, including conditional compilation (#ifdef),branches, clones, textual patches, command-line parameters, andplug-ins. It integrates research from different communities, includingproduct- line analysis, static analysis, refactoring, concern location,and architecture recovery toward a common goal of reverse engineeringvariability implementations . This integration will allow scalingresearch and practice of analyzing and migrating variability from corecalculi to real-world systems of the size and complexity of the Linuxkernel.