Re: HEADS UP: plan to switch many ports over to GCC 12 soon

["David H. Gutteridge" <david%gutteridge.ca@localhost>]

On Fri, 21 Jun 2024 at 15:07:41 -0400, Greg Troxel wrote:
> Christoph Badura <bad%bsd.de@localhost> writes:
>
> > For me gcc/usr.bin/host-libcpp failed to build on macOS 12.7.
> >
> > Stuff like:
> > src/external/gpl3/gcc/dist/libcpp/include/line-map.h:1677:44: warning: 
> > deleted function definitions are a C++11 extension 
> > [-Wc++11-extensions]
> > src/external/gpl3/gcc/dist/libcpp/include/cpplib.h:319:13: warning: 
> > commas at the end of enumerator lists are a C++11 extension 
> > [-Wc++11-extensions]
> > src/external/gpl3/gcc/dist/libcpp/include/cpplib.h:1144:6: warning: 
> > scoped enumerations are a C++11 extension [-Wc++11-extensions]
> > etc.
> >
> > A patch similar what David Gutteridge did on r1.113 of 
> > src/tools/gcc/Makefile
> > fixed that.
> >
> > diff --git a/external/gpl3/gcc/usr.bin/host-libcpp/Makefile 
> > b/external/gpl3/gcc/usr.bin/host-libcpp/Makefile
> > index 77ea9e0b2ba..e7968c85f55 100644
> > --- a/external/gpl3/gcc/usr.bin/host-libcpp/Makefile
> > +++ b/external/gpl3/gcc/usr.bin/host-libcpp/Makefile
> > @@ -1,5 +1,7 @@
> >  #	$NetBSD: Makefile,v 1.6 2023/05/19 10:42:34 lukem Exp $
> >
> > +.include <bsd.init.mk>
> > +
> >  BUILD_ENV= \
> >  		AR=${HOST_AR:Q} \
> >  		AWK=${TOOL_AWK:Q} \
> > @@ -19,6 +21,12 @@ BUILD_ENV= \
> >  		RANLIB=${HOST_RANLIB:Q} \
> >  		YACC=${YACC:Q}
> >
> > +# Force C++11 so builds succeed on hosts where this isn't the 
> > compiler
> > +# default, e.g., NetBSD 8.x with GCC 5.5.
> > +.if ${HAVE_GCC} > 10
> > +HOST_CXXFLAGS+=		-std=c++11
> > +.endif
> > +
> >  GMAKE_J_ARGS?=	${MAKEFLAGS:[*]:M*-j*:C/(^|.* )(-j 
> > ?[0-9][0-9]*).*/\2/W}
> >  BUILD_COMMAND=	/usr/bin/env -i ${BUILD_ENV} ${TOOL_GMAKE} 
> > ${GMAKE_J_ARGS} -e
> >
> > I'll commit that if there are no objections.
>
> I'm not really comfortable with this being conditional.   If the 
> program
> is in c++11, it makes sense to just always pass that -std.   If the
> default is c++11, it's a noop.  If the default is c++17, it asks that
> the sources be treated as c++11 and thus would catch if anyone added a
> c++17ism via an edit.

If you're referring here to ".if ${HAVE_GCC} > 10", that definitely
needs to be conditional. HAVE_GCC refers to the version of GCC being
built as a tool, not the version of the host compiler in use. Building
GCC < 11 does not require C++11 support from the host compiler, and
wouldn't be correct to pass (it would knock out older but still viable
host compilers in that context, at minimum).

Regards,

Dave