Capturing the diversity of analyses on the Linux kernel variability

Description

As its variability management architecture is complex, the Linux kernel is a constant subject of study for analyzing different aspects of its variability. It relies on a configuration-aware build system, preprocessor directives in the code, and a configuration tool. While many studies have focused on detecting anomalies within these parts or between them, all concepts and denominations are different among contributions, with similar properties devised with varied formalisms, or with no easy relationship between them. This actually hampers the understanding of all variability issues and proposed analyses, as well as their application to other highly configurable systems. In this paper, we analyse the different properties that have been studied on the variability of the kernel and propose a formalism based on the generic concepts of configurator and derivator. We instantiate them to represent the Kconfig, the Kbuild, and CPP in a unified model that enables to represent all the consistency properties. With this model, we manage to categorize the main related studies, establishing their coverage on the defined properties, showing also overlapping and divergences between studies.

Abstract

International audience

Additional details