Derivation Strategies for Experienced-Based Test Oracles
José Augusto Evangelho Hernández · Calhoun: The Naval Postgraduate School Institutional Archive (Naval Postgraduate School) · 1989
Traditionally, large software systems are tested to demonstrate that the system satisfies the set of functional requirements and specifications from which it was derived. Various methodologies exist for conducting this type of testing. However, when the requirements document, or specification, has become outdated or incomplete to the point that they are irrelevant, then testing must take a different approach in order to verify and validate. There can be many reasons why a large software system gets developed without a clear specification; notwithstanding testing must proceed even when confronted with a non-existent specification. Testing in such situations is difficult since there is no separation of specified function from implemented function, and thus no objective standard for judging the correctness of test results. This research proposes a strategy for verification and validation of large software systems when no effective requirements specification exists. To derive an objective correctness standard, the strategy employs requirements information gained from a variety of sources: user conferences, analyst conferences, new user manuals, inverse transformation of code to specification, a validated "kernel" system, and previous test strategies.