Automatic Generation of Test Cases

muchine, " a progrum f o r uutomuticully generating syntacticully correct progrums (test cusrs> f o r checking compiler front ends. The notion of " clynumic grammur " is introduced und is used in a syntax-defining notution thut procides f o r context-sensitiuity. Exurnples demonstrute use of the syntax machine. The " syntax machine " discussed here automatically generates random test cases for any suitably defined programming language.' The test cases it produces are syntactically valid programs. But they are not " meaningful, " and if an attempt is made to execute them, the results are unpredictable and uncheckable. For this reason, they are less valuable than handwritten test cases. However, as an inexhaustible source of new test material, the syntax machine has shown itself to be a valuable tool. In the following sections, we characterize the use of this tool in testing different types of language processors, introduce the concept of " dynamic grammar " of a programming language, outline the structure of the system, and show what the syntax machine does by means of some examples. Test cases Test cases for a language processor are programs written following the rules of the language, as documented. The test cases, when processed, should give known results. If this does not happen, then either the processor or its documentation is in error. We can distinguish three categories of language processors and assess the usefulness of the syntax machine for testing them. For an interpreter, the syntax machine test cases are virtually useless,