Abstract
Many researchers have been focusing on building combinatorial test generators having the best possible perfor- mances, in
terms of smaller test suites and shorter generation times. The majority of tools generates test suites from scratch. This
means that when the test suite must be regenerated, the old tests are discarded and a new test suite is built. However, there
are many cases in which old test cases, possibly written by hand, need to be (or could be) included in the fi nal test suite,
and the test suite completed with new tests in order to reach the desired level of combinatorial coverage. These existing
tests that are reused are generally called seed tests. Seed tests could be important for testing domain-specifi c critical
parts of the system, or they could represent old test suites that must be enriched to reach the desired (possibly higher)
strength of coverage. In this paper, we propose a new architecture for incremental test generation that starts from existing
test seeds. This new architecture is supported by the pMEDICI+ tool which extends our previous effort done for pMEDICI. We
evaluate the proposed approach on the benchmarks given in the context of the second edition of the CT-Competition and w.r.t.
two application scenarios. For each scenario, we automatically generate seed tests and then we apply pMEDICI+ to obtain the
desired test suite. The experiments highlight that using incremental test generation can contribute signifi cantly in the
reduction of test generation time and, in many cases, in the reduction of the test suite size.
[download the pdf file] [DOI]