| Recent work in the area of software engineering has focused on the development of sophisticated interactive environments to support the development and maintenance of software systems. In such environments, software development usually proceeds in cycles of program modification followed by testing and debugging. The testing and debugging of modified programs is a major factor contributing to the high cost of maintaining evolving software systems. To reduce this cost, program changes must be made and tested in an efficient manner. This dissertation describes a systematic approach for reducing the cost of regression testing and fault localization through the use of incremental data flow analysis. Incremental data flow analysis is used to identify the portion of a program affected by a change so that testing effort may be focused accordingly. The analysis allows the partitioning of existing test cases into relevant, nonrelevant, and invalid classes. This greatly reduces the effort associated with validating a program following modification. New algorithms for definition-use dependency analysis and data flow anomaly detection which is capable of handling recursive procedures are also presented. |
