> Well, there is an infinite amount of "real-life correct" codes that fail > type checking; do you propose a clear guideline that the compiler > should follow to accept them? Do not periodically let the kernel folks think you introduce random changes to make their life harder. -- Eric Botcazou