The construction of oracles for software testing

Abstract

Software testing is important throughout the software life cycle. Testing is the part of the software development process where a computer program is subject to specific conditions to show that the problem meets its intended design. Building a testing oracle is one part of software testing. An oracle is an external mechanism which can be used to check test output for correctness. The characteristics of available oracles have a dominating influence on the cost and quality of software testing. In this thesis, methods of constructing oracles are investigated and classified. There are three kinds of method of constructing oracles: the pseudo-oracle approach, oracles using attributed grammars and oracles based on formal specification. This thesis develops a method for constructing an oracle, based on the Z specification language. A specification language can describe the correct syntax and semantics of software. The contextual part of a specification describes all the legal input to the program and the semantics part describes the meaning of the given input data. Based on this idea, an oracle is constructed and a prototype is implemented according to the method proposed in the thesis.