Re: [PATCH] more portable bootstrap procedure

[Ralf Wildenhues]

* Jim Meyering wrote on Sat, Dec 13, 2008 at 01:52:46PM CET:
> Ralf Wildenhues <address@hidden> wrote:
> > * Jim Meyering wrote on Sat, Dec 13, 2008 at 01:09:10PM CET:
> >> Ralf Wildenhues <address@hidden> wrote:
> >> >
> >> > There is one question I have on this: coreutils requires 'tar', my patch
> >> > will allow $TAR as GNU tar.  However, Automake's macros will honor
> >> > $AMTAR only, not $TAR.  Is this where tar is used, or are there other
> >> > instances I am not aware of?  If yes, should I fix check_versions to
> >> > check $AMTAR instead of $TAR?
> >
> >> Whichever you prefer is fine with me.  I use neither.
> >
> > Well, why is tar then listed as a build prerequisite in bootstrap.conf?
> > Have you had a bad experience with non-GNU tar in 'make dist'?
> 
> Personally, I'd like tar to honor the TAR_OPTIONS envvar,
> as set in GNUmakefile.  But that's useful only when creating
> a distribution tarball, and even then, it's not critical.

Hmm, ok, I understand the point.  Well, up to you, that s/TAR/AMTAR/
change would be fine with me.

> >> The rest of those changes look fine and passed a smoke test,
> >> so I'll push it (but without the Signed-off-by line) unless
> >
> > You don't use Signed-off-by for coreutils?  Why not?
> 
> Adding a "Signed-off-by: NAME" line is redundant when
> Author: already lists that same NAME.

Not if the semantics of Signed-off-by: are those described in
git/Documentation/SubmittingPatches.  Of course GNU software
is different anyway in that you have at some point signed an
actual piece of paper.  Then, as long as the changes are only
by the person in Author:, then yes, I guess it is redundant.

Maybe I should revisit my commit habits for autotools.
OTOH, using 'git commit -s' shouldn't hurt, and fingers are
easily trained.

Cheers,
Ralf