Re: unbreak devel/gmake on armv7

On Fri, Nov 01, 2019 at 11:04:47AM +0100, Jeremie Courreges-Anglas wrote:
>
> +cc phessler who hit the same problem.
>
> On Fri, Nov 01 2019, Matthieu Herrb <matthieu@openbsd.org> wrote:
> > Ok, so here is a correct patch.
> > The problem is that makeint.h considers that __arm is not really
> > POSIX.1 compliant. I think it was meant to only apply to SVR3 ?
>
> I'm not sure about that:
>
> http://git.savannah.gnu.org/cgit/make.git/commit/?id=660a23d6493910c3cf5fa0699811d1c1c3f3e655

Thanks for digging the gmake history.
>
> But I guess your patch would have almost the same result as removing
> __arm from those tests.
>
> > But any variant that fixes the condition on OpenBSD is ok.
>
> Yes, but IMHO it would be better if the fix conveyed the right message.
> Could you please consider the diff below instead? (Untested)
>
> > I've left the patch for the missing semi-column in place, although
> > this code should never be used.
>

That works too. ok matthieu@
>
>
> Index: patches/patch-job_c
> ===================================================================
> RCS file: /cvs/ports/devel/gmake/patches/patch-job_c,v
> retrieving revision 1.2
> diff -u -p -r1.2 patch-job_c
> --- patches/patch-job_c 13 Sep 2019 16:59:34 -0000 1.2
> +++ patches/patch-job_c 1 Nov 2019 10:01:40 -0000
> @@ -46,7 +46,7 @@ Index: job.c
> +static void
> +unblock_sigs ()
> +{
> -+ sigsetmask (siggetmask (0) & ~fatal_signal_mask)
> ++ sigsetmask (siggetmask (0) & ~fatal_signal_mask);
> +}
> +
> +void
> Index: patches/patch-makeint_h
> ===================================================================
> RCS file: patches/patch-makeint_h
> diff -N patches/patch-makeint_h
> --- /dev/null 1 Jan 1970 00:00:00 -0000
> +++ patches/patch-makeint_h 1 Nov 2019 10:01:40 -0000
> @@ -0,0 +1,16 @@
> +$OpenBSD$
> +
> +Remove too broad #ifdef so that we also use modern code on arm.
> +
> +Index: makeint.h
> +--- makeint.h.orig
> ++++ makeint.h
> +@@ -114,7 +114,7 @@ extern int errno;
> +