From: Andi Kleen Date: Mon, 20 Jun 2011 21:30:01 GMT Subject: Re: [PATCH 2/3] Add a lot of dummy returns to avoid warnings with NO_NORETURN Message-ID: <20110620213001.GB32765@one.firstfloor.org> In-Reply-To: <7vsjr4b3tf.fsf@alter.siamese.dyndns.org> On Mon, Jun 20, 2011 at 02:17:32PM -0700, Junio C Hamano wrote: > Andi Kleen writes: > > > From: Andi Kleen > > > > Add a lot of dummy returns to silence "control flow reaches > > end of non void function" warnings with disabled noreturn. > > > > If NO_NORETURN is not disabled they will be all optimized away. > > I think this is probably a bad move, given that the previous patch is a This is basically the patch you suggested. Do you have some other suggestion now? FWIW I preferred my original minimal patch and I can go back to that one. > temporary workaround until gcc 4.6 is fixed. With -Wunreachable-code on, gcc mainline (and possibly 4.6.2) has it already fixed, but it's reasonable to assume 4.6.0 will be in use for a long time. There's nothing "temporary" about compiler workarounds, unless you wait 10 years or so. > these will introduce noise for build without NO_NORETURN (either when > profile feedback is not used, or when profile feedback build is in use and > it no longer requires the NO_NORETURN workaround). I fixed the noise in a followon patch. -Andi -- ak@linux.intel.com -- Speaking for myself only.