On 10/12/10 14:28, Jim Meyering wrote: > This feels like enough of a technicality that I'd prefer to defer it.
Well, my suspicion is that some C compilers simply refuse to compile the construct, as they're entitled to do. I vaguely recall that happening in the past. No big deal of course. You're right about the word "regression": I should have called it a "warning" or "compile-time issue" or something like that, since it's not a run-time problem on all practical porting hosts.
