ABSTRACT
A pervasive challenge in designing combinatorial test inputs is managing the state of the system under test (SUT). Dierent states may respond dierently to inputs, so controlling the state is essential to knowing what is being tested and how thoroughly. When a set of test inputs leads to an unplanned result (e.g., initiation of a dierent feature or an exception) instead of the expected result-and if the SUT is not at fault-some of the t-tuples may not be covering the intended test area [63,179,194]. ese test cases mask an expected system behavior. Consequently, gaps in the test coverage may leave some faults undetected. e need then is control of the system state from the set-up of the test case to its expected results. Attention to the state of the system can enable the tester to design a more complete plan for testing more states, to apply the test inputs more accurately with appropriate partitioning, and to be better informed about the inevitable limitations and tradeos of the test design.
