Dave Roberts (Finding Lisp) recently posted about the gcl ansi-tests' Common Lisp random tester and random testing in general.

I use random testing not only for testing CL compilers, but also for more focused testing of specific language features. For example, I randomly check SUBTYPEP (which caused crhodes some grief), various sequence functions, settings of control variables for printing/reading, and various numeric functions.

The idea of using randomized inputs to test compilers has been tried before, aside from McKeeman. Don Slutz at Microsoft wrote a random test generator for SQL database systems. B. A. Wichmann did random stress testing of compilers for various languages. Ariel Faigon talks about random testing of compiler backends targeting National Semiconductor 3200 series microprocessors.