An Experimental Evaluation of Software Testing.
Abstract
The techniques evaluated are path (branch) coverage testing and static analysis. The basic approach was to prepare programs for testing by seeding them with errors whose type and frequency are typical of new software at the integration- or system-level of testing. The experiments were conducted in three phases. The first used eight small programs from a popular programming manual, the second and third used a 5000-line FORTRAN program used to simulate ballistic-missile defense engagements. For the most part, both the path testing and static analysis used the SQLAB tool, with the techniques used singly and in combination. In Phase 1, the DAVE system's static analysis capabilities were also used. In Phase 3, the techniques were compared with the techniques of intermediate-value printout and control-flow tracing. The inescapable conclusion remains, however, that fully automated computer-aided testing is not possible at present. Further, the errors that are not detected are generally considered difficult to locate by conventional techniques.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1979
- Accession Number
- ADA071050
Entities
People
- C. Gannon
- N. B. Brooks
- R. N. Meeson