On Sun, Nov 19, 2017 at 2:48 AM, Sandra Loosemore <san...@codesourcery.com> wrote: > On 11/17/2017 12:12 PM, Wilco Dijkstra wrote: >> >> GCC currently defaults to -fcommon. This is an optional C feature dating >> back to early C implementations. On many targets this means global >> variable >> accesses having an unnecessary codesize and performance penalty in C code >> (the same source generates better code when built as C++). Given there >> isn't >> a lot of software that really requires this (mostly it's accidentally >> forgetting to >> use 'extern' in a header), it is about time to change the default. >> >> What do people think? I presume someone with access to distro source code >> and a fast build machine could try and see how many packages fail to get >> an >> idea how feasible it is. We could keep defaulting to -fcommon with >> -std=c89 >> if necessary. >> >> 2017-11-17 Wilco Dijkstra <wdijk...@arm.com> >> >> * common.opt (fcommon): Change init to 1. >> * doc/invoke.texi (-fcommon): Update documentation. > > > The doc parts of this patch are OK. > > Personally, I think the change is a good idea, but I don't know how much > breakage it would cause. One data point I can contribute is that when we > looked at this several years ago, we found that all competing compilers on > bare-metal targets implemented the -fno-common behavior. So another possible > compromise is keeping -fcommon on Unix-like targets where it is the > traditional behavior, and switching to -fno-common elsewhere where it's not. > I don't see much benefit in tying the default to -std= since it's not > behavior specified in the C standard at all.
A target specific default might be a good idea if we decide to revert. Note I proposed this change a few times already, but the fear was always we'll break too much legacy code. Note you have to make sure GFortran still works! So I think the patch should be changed to make the default behavior be frontend dependent or have a fortran/ adjustment that fixes things up for the fortran dialects that need it. CCing fortran@ Richard. > -Sandra
