Motivated by Structural Statistical Software Testing (SSST), this paper is interested in sampling the feasible execution paths in the control flow graph of the program being tested. For some complex programs, the fraction of feasible paths becomes tiny, ranging in $[10^{-10}, 10^{-5}]$. When relying on the uniform sampling of the program paths, SSST is thus hindered by the non-Markovian nature of the ``feasible path'' concept, due to the long-range dependencies between the program nodes. A divide and generate approach relying on an extended Parikh Map representation is proposed to address this limitation; experimental validation on real-world and artificial problems demonstrates gains of orders of magnitude compared to the state of the art.
@InProceedings{baskiotis_et_al:DagSemProc.08351.7, author = {Baskiotis, Nicolas and Sebag, Mich\`{e}le and Gaudel, Marie-Claude}, title = {{SoftwareTesting with Active Learning in a Graph}}, booktitle = {Evolutionary Test Generation}, pages = {1--12}, series = {Dagstuhl Seminar Proceedings (DagSemProc)}, ISSN = {1862-4405}, year = {2009}, volume = {8351}, editor = {Holger Schlingloff and Tanja E. J. Vos and Joachim Wegener}, publisher = {Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik}, address = {Dagstuhl, Germany}, URL = {https://drops.dagstuhl.de/entities/document/10.4230/DagSemProc.08351.7}, URN = {urn:nbn:de:0030-drops-20149}, doi = {10.4230/DagSemProc.08351.7}, annote = {Keywords: Structural Statistical Software Testing, Active Learning, Control Flow Graph, Feaisble Paths, Parikh maps.} }
Feedback for Dagstuhl Publishing