[patch][arm] gcc-7-branch: Fix bootstrap on FreeBSD

[Kyrill Tkachov]

On 17/10/17 13:42, Andreas Tobler wrote:
> Hi Kyrill,
>
> On 17.10.17 12:02, Kyrill Tkachov wrote:
>
> > On 16/10/17 20:00, Andreas Tobler wrote:
> >> Hi all,
> >>
> >> I struggled over a bootstrap issue while building gcc-7 for
> >> armv7-*-freebsd*
> >>
> >> I got a 'permission denied' while creating the arm-tables.opt file.
> >>
> >> The source tree is located on a nfs server.
> >>
> >> The below patch fixed it for me.
> >>
> >> Ok to apply?
> >>
> >> TIA,
> >> Andreas
> >>
> >> 2017-10-16  Andreas Tobler <andreast@gcc.gnu.org>
> >>
> >>          * config/arm/t-arm (MD_INCLUDES): Create arm-tables.opt via
> >>          intermediate arm-tables.new like the other awk generated 
> files.
> >>
> >> Index: config/arm/t-arm
> >> ===================================================================
> >> --- config/arm/t-arm    (revision 253792)
> >> +++ config/arm/t-arm    (working copy)
> >> @@ -75,8 +75,8 @@
> >>    $(srcdir)/config/arm/arm-tables.opt: 
> $(srcdir)/config/arm/parsecpu.awk \
> >>      $(srcdir)/config/arm/arm-cpus.in
> >>           $(AWK) -f $(srcdir)/config/arm/parsecpu.awk -v cmd=opt \
> >> -               $(srcdir)/config/arm/arm-cpus.in > \
> >> -               $(srcdir)/config/arm/arm-tables.opt
> >> +               $(srcdir)/config/arm/arm-cpus.in > arm-tables.new
> >> +       mv arm-tables.new $(srcdir)/config/arm/arm-tables.opt
> >>
> >
> > This looks ok to me as it makes the rule consistent with the other
> > awk-generated files.
> >
> > Out of interest, this looks like a small subset of Richard's patch [1]
> > at r249971.
>
> Hehe, now as you say, yes. But I wasn't aware about it. I just tried to
> fix my bootstrap issue and compared the snippet with main. And tried if
> it helps to use an intermediate file.
>
> > Have you tried that patch on the branch?
>
> No, is this patch going to appear on the gcc-7 branch?
> If it is, then I'll not apply my patchlet above.
>

AFAIK that patch was part of a series to further improve the 
architecture features selection mechanism for GCC 8
and wasn't thus considered for the GCC 7 branch, but it looks like a 
fairly standalone improvement, so unless Richard
has any objections to it, I think it would be good to take that patch 
for the branch (assuming it passes validation there).

Kyrill

> > [1] https://gcc.gnu.org/ml/gcc-patches/2017-07/msg00223.html
>
> Thanks,
> Andreas
>