On Mon, 10 Sep 2012, Ralf Baechle wrote: > > A dummy "break" is the usual solution though. I don't think GCC ever > > complains if it sees it unreachable after a "return" -- in a sense it is > > just as unreachable as this null instruction is. > > I wasn't overly picky. Whatever gets the stuff to build correctly. I'm > doing one final round of test builds over all -stable branches before > dropping most of them like radioctive rocks. But more on that later. Yeah, sure -- I just noted this is not breaking a new ground really. There was a time GCC used to support genuinely empty switch cases and that was removed at one point for better ISO C compliance. There was a rush fixing code all over the place at that point, including some "proper" GNU software such as I reckon bison, and the common approach taken was that I referred to. Maciej
