Debug Testing and Confidence Testing.
Abstract
The strong point of computer program testing has always been failure. When a test fails, it is clear what to do, and this has led to the maximum that the goal of testing is finding faults. Testing theory, on the other hand, has tried to connect test success to program correctness. Call the kind of testing that seeks failures 'debug testing', and the other 'confidence testing'. A confidence-testing technique might in principle be used for debugging, but debugging tools cannot establish confidence. Debug testing is an activity intertwined with the whole of program development, and its theory must take account of this sociological context; debugging is a human craft. On the other hand, confidence testing theory may take program and test as given, without their human origins. Only by separating the two kinds of testing can reasonable goas be set for testing theory. The difference between debug- and confidence-testing theory is illustrated by detailed analysis of partition testing, and of experiments to validate debugging test tools. Goals for each kind of theory are proposed. Originator-supplied keywords include: Nose-rubbing effect, and Probabilistic correctness.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1984
- Accession Number
- ADA147104
Entities
People
- D. Hamlet
Organizations
- University of Maryland