Test planning for software reuse

This thesis describes a methodology to support the decision to reuse a software asset from a repository, using the information of previous testing to best advantage. The context of this work is statistical testing with usage models represented as Markov chains.

The methodology provides techniques to compare a new usage chain with other usage chains in the repository to determine the similarity between the prospective new use and the testing history. To determine if enough testing was performed, the method compares the prospective new use with the testing chains in the repository, and also analyzes the reliability and confidence measures when no failures are found.

The methodology also describes techniques to combine previous uses of the software in such a way as to have more information about the behavior of the software in general as well as for a particular environment. A technique to disregard the irrelevant (redundant) testing experience is also provided so that the stopping criteria in testing are met sooner than by testing exclusively with the new use, or by including all the testing history.