Validation Suite Generation

To validate big systems is a very complex task which has not been solved yet. Many researchers have addressed the more restricted problem of language validation Goodenough 81, Whichmann & Sale 80] but still did not succeed. One reason is the lack of a semantical deenition of the language to be validated, the other reason is that such a semantics is often considered as a theoretical object unable to provide more than meanings to programs. To validate a Lisp-like system is still a large problem due to the number of its essential constructs (special forms and other unusual features like eval or multiple values) and to the size of its run-time support of which the garbage collector is the main and most complex part. However | since Scheme has only a few features Rees & Clinger 86] and may be considered as an eecient-calculus Steele & Sussman 75], | since-calculus is the support of Deno-tational Semantics Stoy 77, Schmidt 86] and denotational semantics for Scheme has already been published Muchnick & Pleban 80, Rees & Clinger 86] and eventually | since the equation \program = data" holds in Scheme, it seems therefore possible to make use of the This work has been partially funded by Greco de Programmation. Interpr etation, Compilation et S emantique des Langages Applicatifs denotational semantics of Scheme to generate validation suites. The paper will present this technique, its implementation (via streams) and some experimental results.