SMOTL—A System to Construct Samples for Data Processing Program Debugging

Constraint-Based Testing

“…Using a dedicated homemade constraint solver, it was then possible to generate a test input which would reach the mutated statement and reveal the different behavior of the mutated program. Taking inspiration of ideas originally proposed by Bicevskis et al in [45], the constraint solver propagated inequality constraints extracted from program statements to reduce the variation domain of input variables. Following a distinct approach based on program executions, metaheuristics were also explored at the same time for automatic test data generation with the research work of Korel [46], Gupta et al [47], and Tracey et al [48].…”

“…Historically, linear programming techniques in combination with ad hoc techniques were the first automatic constraint solving the be used in order to generate test inputs for path conditions [3,33,45,67,68]. By switching path conditions solving into a constrained optimization problem, Boyer et al and Clarke [33,67] have explored several distinct linear programming techniques, including the cut generation algorithm of Gomory for integer computations and the Benders decomposition technique.…”

“…Note that, in the context of test input generation, Boyer et al [67] also proposed to use the conjugate gradient algorithm to deal with quadratic and linear inequalities. Interestingly, Bicevskis et al [45] proposed a tool called SMOTL, which contained an ad hoc procedure for solving inequalities over integer variables based on a domain pruning procedure. The constraint system is processed iteratively: the inequalities are used to prune the interval domain of each variable until no more pruning is possible.…”

Euclide: A Constraint-Based Testing Framework for Critical C Programs

“…But, constraint propagation alone does not guarantee satisfiability, as it just prunes the variation domains without looking at potential solutions. And it must be coupled with other mechanisms in order to find solutions or to show inconsistency 3 Dynamic Linear Relaxations (DLRs). In [14], we introduced DLRs to relax dynamically all the constraints of an Euclide program, including the non-linear ones, within a Linear Programming framework.…”

Constructing an automated testing oracle: an effort to produce reliable software

“…For some specialised classes of programs, there exist methods for constructing a finite set of test cases whose successful execution can establish correctness of the program for all possible inputs [5,9). This is not possible in the general case: testing can show the presence of software errors but it cannot certify their absence for unconstrained programs.…”