Section: New Results
Software Evolution
Static and Dynamic Checking of Component Interactions
Participants : Laurence Duchien, Anne-Françoise Le Meur, Guillaume Waignier.
Building autonomic applications, i.e., systems that must adapt to their execution context, requires architects to calibrate and validate the adaptation rules by executing their applications in a realistic execution context. Unfortunately, existing works do not allow architects to monitor and visualize the impact of their rules, nor that they let them adjust these rules easily.
In the context of Guillaume Waignier's PhD thesis, we are working on a model-based framework that enables architects to design and debug autonomic systems in an iterative and uniformed process [30] , [55] , [35] . This work has led to the extension of our previous work, called CALICO, a Component AssembLy Interaction Control FramewOrk (cf. sections 5.2 ). At design-time, architects can specify, using models, the application's structure and properties, as well as the desired adaptation rules. At debugging-time, the running application and the models coexist such that the models control the application dynamic adaptation, thanks to a control loop that reified runtime events. Each triggered adaptation is first tested at the model level to check that no application property is broken. Furthermore, architects can modify the models at any time in order to adjust the adaptation rules or even parts of the application. All changes at the model level, if checked correct, are directly propagated to the running application.
Furthermore, our solution is generic regarding the underlying platforms. The current version of CALICO handles three component-based platforms. Moreover, the benchmarks of our implementation show that CALICO is usable to design reliable large systems up-to 10000 components, which is the maximum load of most runtime platforms.
CALICO is available on http://calico.gforge.inria.fr .
Detection of Design Defects
Participants : Laurence Duchien, Anne-Françoise Le Meur.
Following Naouel Moha's Ph.D. thesis [86] , two journal papers have been submitted and accepted.
One paper focuses on the detection of code and design smells in software systems. More precisely, it presents an approach to automate the generation of algorithms from specifications written using a domain-specific language [14] .
The other paper presents the DECOR method, which embodies and defines all the steps necessary for the specification and detection of code and design smells. The paper presents also a concrete implementation and an empirical validation of the method [13] . This paper represents the core of Naouel Moha's Ph.D. thesis [86] .
