Français Anglais
Accueil Annuaire Plan du site
Home > Research results > Dissertations & habilitations
Research results
Ph.D de

Group : Formal Testing and System Exploration

Using Combinatorial Structures for Statistical Testing

Starts on 01/09/2000
Advisor : GAUDEL, Marie-Claude
[Alain Denise et Marie-Claude Gaudel]

Funding : A
Affiliation : Université Paris-Saclay
Laboratory : LRI

Defended on 24/06/2004, committee :
A. Denise (LRI, Université Paris-sud 11),
P. Flajolet (rapporteur, INRIA Rocquencourt),
M.-C. Gaudel (LRI, Université Paris-sud 11),
B. Marre (CEA Saclay),
F. Ouabdesselam (rapporteur, LSR-IMAG, Grenoble),
P. Thévenod-Fosse (présidente, LAAS Toulouse).

Research activities :
   - Verification
   - Software Testing
   - Formal Methods for Software Engineering

Abstract :
In this thesis, we describe a new generic method for statistical testing
of software procedures, according to any given graphical description
of the behavior of the system under test (control flow graph,
statecharts, etc.). Its main originality is that it combines results
and tools from combinatorics (random generation of combinatorial
structures) with symbolic constraint solving, yielding a fully automatic
test generation method.
Instead of drawing input values as with classical testing methods,
uniform random generation routines are used for drawing paths from
the set of possible execution paths or traces of the system under test.
Then a constraint resolution step is used for finding actual values for
activating the generated paths.
Moreover, we show how linear programming techniques may help to improve the quality of test set.

A first application has been performed for structural statistical testing,
first defined by Thevenod-Fosse and Waeselynck (LAAS) and a tool has
been developed. Some experiments (more than 10000 on four programs
of an industrial software) has been made in order to evaluate our
approach and its stability.

These experiments show that our approach is comparable to the one of
the LAAS, is stable and has the additional advantage to be completely
automated. Moreover, these first experiments show also that the
method scales up well.
More generally, this approach could provide a basis for a new class of
tools in the domain of software testing, combining random generation of
combinatorial structures, linear programming techniques, and
constraint solvers.

More information: