Feature-oriented product line evolution

Séminaire PReCISE

Evolving large scale, complex, and highly variable systems is known to be a difficult task. Features, as unit of variability, are defining elements in the design and implementation of such systems. Because of this tight relationship between features and implementation, we can analyze the evolution of a product line through the evolution of its feature model.

During this talk, I will present the ongoing research work I am performing as part of my PhD on product line evolution. First, I will show how, using feature modeling techniques, we helped domain engineers of Philips Healthcare to assess feature change impact on the existing configuration an X-ray machine. With only partial information about the dependencies between the various sub-systems of this machine, we modeled the X-ray machine and its interfaces as several inter-dependent feature models. Using this representation, we devised a simple heuristics to evaluate how a given feature change affects the existing configurations of their systems. Then, I will present FMDiff, a tool we built to gather fine-grained change information from the Linux feature model. Based on the EMF Compare framework, FMDiff allows us to capture changes occurring on features of the Linux kernel down to attribute value changes. The dataset built using feature changes occurring over 15 releases of the kernel provides insights on common change operations and allows us to compare the evolution of the different architectures supported by the kernel.

Contact : Gilles Perrouin - gilles.perrouin@unamur.be
Plus d'info : http://swerl.tudelft.nl/bin/view/NicolasDintzner/WebHome
Télecharger : vCal