[Buildroot] [PATCH] Binutils: ARC: Fix build failures if makeinfo is missing

[Alexey Brodkin]

Hi Peter, Thomas,

On Tue, 2016-06-07 at 05:42 +0000, Alexey Brodkin wrote:
> Hi Peter, Thomas,
> 
> On Mon, 2016-06-06 at 23:34 +0200, Peter Korsgaard wrote:
> > 
> > > > > > > 
> > > > > > > "Thomas" == Thomas Petazzoni <thomas.petazzoni at free-electrons.com> writes:
> >  > Hello,
> >  > On Mon, 06 Jun 2016 21:36:54 +0200, Peter Korsgaard wrote:
> > 
> >  >> > Signed-off-by: Zakharov Vlad <vzakhar at synopsys.com>  
> >  >> 
> >  >> Committed, thanks.
> > 
> >  > Why?
> > 
> > It was fixing autobuilder issues and seemed like a sensible fix that
> > could be upstreamed. Looking again, I do see that it patched Makefile.in
> > and not Makefile.am, so that's not too nice though.
> Indeed our goal was to fix an issue that causes all autobuilder jobs for ARC
> to fail. The problem is host binutils couldn't be built any longer after we switched
> to arc-2016.03 tools. And now we're unblocked.
> 
> As for patching Makefile.in vs Makefile.am again that's just to do a minimal fix
> in existing sources. Otherwise if we patch Makefile.am we'll need to regenerate
> Makefile.in.
> 
> Even though Vlad has already sent the same patch to Binutils mailing list
> (https://sourceware.org/ml/binutils/2016-06/msg00053.html) that's indeed not
> the right fix - he'll need to fix Makefile.am but at least we're moving
> in upstreaming direction.
> 
> > 
> >  > This problem also exists for other versions of binutils, and also for
> >  > gdb. And we have a patch series from Romain Naour sitting in patchwork
> >  > for several weeks.
> Right so I would think as of now the same patch should be applied to other
> affected instances of binutils and gdb.
> 
> > 
> >  > I don't know if Zlad's version is better or not than Romain's version.
> >  > But at least Romain's version was handling all binutils and gdb
> >  > versions, without patching directly Makefile.in files. So at first
> >  > sight, it looked a lot better than Zlad's version.
> > 
> > Ok, we can always revert if it isn't needed any more once Romain's
> > series is applied.
> 
>
> In opposite Romain's fix only makes sense for BR and I don't really like that
> approach. Why implement hacks on top of
> upstream sources that are not
> upstreamable at all. Why not try "to fix" generic "missing" script if we do think
> it behaves improperly?

So it looks like "makeinfo" is really required when building binutils/gdb from
git sources, see that response in binutils mailing list:
https://sourceware.org/ml/binutils/2016-06/msg00200.html

So shall we just require "makeinfo" installed on build hosts?

-Alexey