Research On Software Auto-Testing Technology Based On Message Mechanism

Software Testing, as an important branch of Software Engineering, is a new and vital subject of research. Statistics show that, in typical software development projects, the work of software testing accounts for 40% or more of all. Due to the monotony, complexity, and time-consumingness of most work in software testing, it is hard to accomplish by traditional manual work, therefore, automated testing approaches should be employed.How to make an automated Black-box testing on human-machine interactive software is a challenging subject. The existing theories and methods of automated testing, like modeling technology of interface objects, Z-specification-based testing sample auto-generating technology, and SOFL-based testing sample generating model, improved, to a certain extent, the formalization level function testing. However, in regard to complicatedly human-machine interactive software systems of various systematic states, such as CAD system, these methods are hard to function.Our research is supported by a national 863 research project which is named as Research on Automatic Test Technology for 3-Dimensions CAD System (Project No. is 2002AA411510). This thesis, basing on Windows Message Mechanism, proposes a structural model of automated black box testing aimed at graphic user interface software. Through picking up various message at the bottom of Windows and meanwhile reorganizing the message order according to the relevant flow specification and generative calculation, this model derives abundant testing samples and then drives the system being tested to work automatically by the reorganized derivative message order, thus achieving the goal of automated testing.According to this, a message-mechanism based automated testing system is constructed in this thesis. Experiments show that the testing system constructed following this idea, by assuaging the work intensity of testing staff, lower the testingcost, and improve the efficiency of testing, achieves good testing results. This system can be widely applied in regression testing.