Re: gcc8 and "fatal error: had to relocate PCH" with boost-libs

[maya%netbsd.org@localhost]

On Mon, Nov 05, 2018 at 11:20:32AM +0000, David Brownlee wrote:
> On Mon, 29 Oct 2018 at 09:43, David Brownlee <abs%absd.org@localhost> wrote:
> >
> > On Mon, 29 Oct 2018 at 08:32, Martin Husemann <martin%duskware.de@localhost> wrote:
> > >
> > > On Mon, Oct 29, 2018 at 04:18:21AM +0000, maya%netbsd.org@localhost wrote:
> > > > Nice. I lacked a way to reproduce the PCH problems, which makes
> > > > it hard to update the patch and push it upstream.
> > > >
> > > > for boost-libs we have:
> > > > .if ${OPSYS} == "NetBSD" && ((${MACHINE_ARCH} == "i386") || (!empty(MACHINE_ARCH:Mearm*) && !empty(PKGSRC_COMPILER:Mgcc)))
> > > > BJAM_ARGS+=           pch=off
> > > > .endif
> > > >
> > > > But maybe it needs to be all netbsd.
> > >
> > > Can we just use the existing + working gcc fix instead?
> >
> > Apologies, I think I may be missing something.
> >
> > Do we have a gcc fix for "error: had to relocate PCH" which can be
> > applied to the pkgsrc gcc7 and gcc8? That would be great.
> >
> > Otherwise the boost-libs workaround seems a good idea - either
> > extending to all NetBSD, or all NetBSD on gcc7 & gcc8 if we want to
> > not change the gcc5/6 behaviour (it could well be needed on more
> > os/arch/gcc combinations, maybe noone has tried building recent
> > boost-libs on vax -:p)
> 
> Hi - just pinging to see if there are any objections for me disabling
> PCH in boost-libs for all NetBSD gcc versions, or do people prefer me
> to keep it limited (all >= gcc7)?
> 
> David

Feel free to disable PCH whenever it causes you trouble, it's an
optimization to make builds faster, it should have no effect on the
result. We don't have rules that the builds must be super duper fast :-)

Go for it!