The Effectiveness of Software Testing Techniques

已经提出了许多用于测试软件的形式化技术，但是相对较少的证据被收集来表明这些技术是多么有效。本研究计画的目标是：1）比较各种先前开发的测试技术的错误侦测能力，包括基于控制流、基于数据流和变异测试技术; 2）评估路径表达式的部分符号评估（PSE）的有效性，PSE是一种启发式技术，可解决测试中出现的某些语义分析问题。故障检测能力的各种测试数据的充分性标准将通过实验进行比较，产生大量的测试集的几个主题程序，并测量适当的测试集的比例（根据各种标准），暴露故障。故障检测能力也将通过调查条件下，一个标准是保证比另一个根据一定的概率措施分析检查。PSE的有效性将通过测量启发式成功案例的比例来研究。

Many formal techniques for testing software have been proposed, but relatively little evidence has been gathered to indicate how effective these techniques are. The goals of this research project are 1) to compare the fault-detecting ability of various previously developed testing techniques, including control flow based, data flow based, and mutation testing techniques, and 2) to evaluate the effectiveness of partial symbolic evaluation of path expressions (PSE), a heuristic technique that addresses certain semantic analysis issues arising in testing. Fault-detecting ability of various test data adequacy criteria will be compared experimentally by generating large numbers of test sets for each of several subject programs, and measuring the proportion of adequate test sets (according to the various criteria) that expose faults. Fault-detecting ability will also be examined analytically by investigating conditions under which one criterion is guaranteed to be better than another according to certain probabilistic measures. Effectiveness of PSE will be investigated by measuring the proportion of cases in which the heuristic succeeds.