Content area
There is little disagreement with the premise that an effective method for testing object-oriented software is needed. An operational object-oriented system may have only been minimally tested for functionality; and during development, only limited testing for correct behavior may have been performed on each individual class. Often no record of testing history is available once development of the object-oriented system is completed. Test cases generated directly from a subset of the C++ source code, namely, from the class declarations, and executed in a specific order as determined from the class declarations can effectively expose a significant number of program errors. This dissertation describes such a testing procedure for object-oriented software. This testing procedure is easy to follow, could be automated, and has successfully unveiled several software faults in publicly available, presumably debugged C++ systems that consisted of a few hundred to several thousand lines of code.