Compilers and [lack of] Termination Revisted

The author originally posted an article showing how compiler optimizations disproved Fermat's Last Theorem (http://en.wikipedia.org/wiki/Fermat%27s_last_theorem): http://blog.regehr.org/archives/140

The submission is a follow-up to that article discussing the topic in greater depth. It includes the relevant specifications for the behavior of non-terminating, non-side-effect-causing loops in a few different programming languages. It piqued my interest since I hadn't even considered that such behavior would be specified.