Generating Software Test Data by Evolution：（软件测试数据生成的进化）.pdfVIP

Technical Rep ort RSTR Generating Software Test Data by Evolution Gary McGraw Christoph Michael and Michael Schatz RST Corp oration Suite Ridgetop Circle Sterling VA February Abstract This pap er discusses the use of genetic algorithms GAs for automatic software test data generation This research extends previous work on dynamic test data generation where the problem of test data generation is reduced to one of minimizing a function Miller and Sp o oner Korel In our work the function is minimized by using one of two genetic algorithms in place of the lo cal minimization techniques used in earlier research We describ e the implementation of our GAbased system and examine the eectiveness of this approach on a numb er of programs one of which is signicantly larger than those for which results have previously b een rep orted in the literature We also examine the eect of program complexity on the test data generation problem by executing our system on a numb er of synthetic programs that have varying complexities Intro duction An imp ortant asp ect of software testing involves judging how well a series of test inputs tests a piece of co de Usually the goal is to uncover as many faults as p ossible with a p otent set of tests since a test series that has the p otential to uncover many faults is obviously b etter than one that can only uncover a few Unfortunately it is almost imp ossible to predict how many faults will b e uncovered by a given test set This is not only b ecause of the diversity of the faults themselves but b ecause the very concept of a fault is only vaguely dened Test adequacy cri