Other Related Topics - Automatic Generation of Combinatorial Test Data - page 81

Chapter 8
Other Related Topics
Abstract This chapter briefly discusses some other topics that are related to test
data generation. They include, among others, how to construct a combinatorial testing
model for the software under test, how to select a subset of test cases from an existing
test suite, and how to generate special-purpose test cases in the process of debugging.
8.1 Problems Related to Test Input Generation
In the previous chapters, we described many techniques and tools for generating the
input data for combinatorial testing (CT). We have taken a simplistic view of the
problem: Given a model (consisting of parameters, values, constraints, etc.), find a
small (or perhaps the smallest) test suite satisfying the constraints and achieving the
required coverage.
In practice, test input generation is a part of the software development process.
There are several other related activities in the process, like test case execution and
debugging. See Fig. 8.1 .
To apply combinatorial testing to the system under test (SUT), we usually need a
model (unless we already have some test cases). Given that model, we can use various
techniques and tools (as described in the previous chapters) to obtain a suitable test
suite. Then, we test the system using the test suite. Ideally, all the test cases pass. But
it may happen that a few test cases fail, which means the system has some defect(s).
Then, we should identify a small set of parameters, which leads to the failure. Such
a combination of parameters may help us find the bugs in the system. After the bugs
are fixed or the developers change the system due to new requirements, we should
test the system again. This is an iterative process.
Given the above global view of the software development process, we may ask
some questions about testing. For example, can we reuse some existing test cases?
If not, we need to generate a test suite. But where does the model come from? Is it
always good to generate a very small set of test cases? Do we need to put the test
cases in certain order?