Lecture Notes in Computer Science 3472

430 Axel Belinfante, Lars Frantzen, and Christian Schallhart
To generate a test suite a set of test purposes is needed, which represent
sequences of input and output events exchanged between the SUT and its environment
(black box testing). Two modes are offered to generate them. In the interactive
mode the user can define test purposes with the help a SDL-simulator.
Guiding a stepwise simulation of the system one can construct a sequence of
interest.
Based on the SDL specification the tool can automatically complete a set of
test purposes based on a state space exploration to achieve a given percentage
of system-coverage. As in AutoLink the coverage unit is an observational step,
i.e. a sequence of events connecting two states in which the only possible actions
are an input stimuli or a timeout of an internal timer (so called stable states). A
test purpose corresponds to such an observational step which again corresponds
to one or many basic blocks, i.e. blocks of SDL instructions without branching.
It is the same approach as the one from AutoLink and hence there is the same
problem with nondeterminism, see 14.2.15.
In addition to depth-first and supertrace algorithms TestComposer offers a
breadth-first search to traverse the reachability graph. To narrow the search it
is possible to exclude parts of the SDL specification (like transitions, processes
or whole blocks) from the state exploration. To automatically generate postambles
which bring the SUT back to a suitable idle-state, TestComposer allows to
manually define boolean expressions that signalizes such idle states and therefore
allow a search back to them. Test purposes are automatically partitioned
into preamble, test body and postamble. Observer processes can also be used as
abstract test purposes. They do not have to be transformed into MSCs like in
AutoFocus. Such an observer can be used to prune paths of the state space or
generate reports when a given condition holds.
A test purpose does not have to cover a complete sequence of observable
events, it can be incomplete (respectively abstract). TestComposer computes the
missing events needed to bind the specified ones together. There can be many
ways to complete the abstract sequence which allows an abstract test purpose
to describe the set of all possible completions. This is especially useful when the
order of signals does not matter which is a common situation when different
communication channels are involved.
To generate test cases, paths in the SDL specification have to be found which
correspond to the test purposes. Here come the TGV algorithms into operation
which perform also the postamble computation.
Tool Interfaces
SDL specifications serve as inputs. An API (Application Programming Language)
allows the user to construct interfaces with arbitrary test specification
languages. A module for TTCN is already included.
Summary
TestComposer is very similar to AutoLink. Some of the comparative results of
[SEG00] will be addressed in 14.3.