Analysing the Linux kernel feature model changes using FMDiff

Abstract

Evolving a large scale, highly variable system is a challenging task. For such a system, evolution operations often require to update consistently both their implementation and its feature model. In this context, the evolution of the feature model closely follows the evolution of the system. The purpose of this work is to show that fine-grained feature changes can be used to guide the evolution of the highly variable system. In this paper, we present an approach to obtain fine-grained feature model changes with its supporting tool "FMDiff". Our approach is tailored for Kconfig-based variability models and proposes a feature change classification detailing changes in features, their attributes and attribute values. We apply our approach to the Linux kernel feature model, extracting feature changes occurring in sixteen official releases. In contrast to previous studies, we found that feature modifications are responsible for most of the changes. Then, by taking advantage of the multi-platform aspect of the Linux kernel, we observe the effects of a feature change across the different architecture-specific feature models of the kernel. We found that between 10 and 50 % of feature changes impact all the architecture-specific feature

Introduction to the theme issue on variability modeling of software-intensive systems

“…Thus evolution of variability models closely follows evolution of systems. Dintzner et al [3] study fine grained evolution of the feature model and feature implementations in the Linux kernel project. The model is specified in Kconfig, a domain specific language developed by the kernel project.…”

Feature Models in Linux

“…Analyzing the impact of a change helps to know exactly which architectures need to be (build) tested, and can thereby improve performance of current testing frameworks of the Linux community since only affected and not all architectures will be checked. By reading the related papers [12,15] about FMDiff, we stumbled across a seemingly contradictory observation with previous work from our group. On one hand, Dietrich et al [7] state that more than 60 percent of all features are shared among the FMs.…”

“…On one hand, Dietrich et al [7] state that more than 60 percent of all features are shared among the FMs. On the other hand, Dintzner et al [15] discovered when diffing the FMs of Linux over a range of Linux versions, that changes to shared features affect the FMs in the same way but only 37 percent of changes affect all FMs. Since changes to code are proportional to its size [18], the results actually seem to be inverted.…”

Safe Evolution Patterns for Software Product Lines

“…Through this work, we obtained a detailed understanding of feature changes and their impact in two different contexts, and means to capture and analyse them. We extended our work on the Linux kernel, obtaining further knowledge on the inner workings of its FM and its evolution [17]. More recently, I cooperated with Leonardo Passos on the extension of his work on coevolution pattern identification [15].…”

“…More recently, I cooperated with Leonardo Passos on the extension of his work on coevolution pattern identification [15]. My work so far resulted in the following three publications [7], [16], [17], and my participation in a fourth [15].…”