UML’s support for modeling inter-use-case semantics is very limited.
Briand and Labiche (2002) propose the TOTEM methodology to address the
generation of tests for such dependencies. However, their generation
algorithm, which heavily relies the notion of interleaving, presents
one major problem: even for small examples, the number of sequences of
use cases to test is unmanageable. The question we ask is: is it
feasible to capture use cases and their sequential dependencies in a
specification whose executability can allow system-level testing while
avoiding the explosion of the
number of use case sequences to
generate? To answer positively this question we develop a
specification for TOTEM’s library system in Arnold and Corriveau’s ACL
specification language and explain how this specification can be used
to generate a manageable number of tests. The key idea is that by
modeling use case mutual independence we can considerably reduce the
number of tests.