Computer Supported Software Inspection

Macdonald, F. (1999) Computer Supported Software Inspection. PhD thesis, University Of Strathclyde. (http://www.cis.strath.ac.uk/research/efocs/abstrac...)

Full text not available in this repository.Request a copy

Abstract

For over twenty years, software inspection has been advocated as a simple, cost-effective technique for defect detection in all types of documents. Essentially, a number of participants review a document with the aim of discovering defects. Many positive experience reports have been published demonstrating its benefits, and there are now a number of variations on the basic inspection method. In recent years, there have been a number of attempts to further increase inspection efficiency by the introduction of tool support, resulting in a number of prototype systems. In general, however, existing systems tend to suffer from three major shortcomings. The first concerns their inability to easily support a number of different inspection processes, as well as the inspection of a variety of document types. Existing tools also treat the move to computer-based inspection as a simple change of medium, when there are opportunities to greatly enhance the process. Finally, evaluation of the effectiveness of these tools is sparse. This thesis describes work tackling these deficiencies. Support for multiple inspection processes was achieved by developing a high-level process description language which can be used as input to an inspection tool. A prototype tool was developed which implements this language and also provides a simple mechanism for supporting inspection of multiple document types. This tool was also used to investigate a number of methods for increasing the efficiency of inspection, including cross-referencing within and between documents, active checklists and automatic collation of defect lists. In addition, two controlled experiments comparing paper-based and tool-based inspection were performed, the first of their kind. These experiments could reveal no performance difference between methods.