Re: Running ./config.status

[Ralf Corsepius]

On Fri, 2005-02-04 at 13:01 -0800, Paul Eggert wrote:
> > On Fri, Feb 04, 2005 at 02:49:29PM +0100, Ralf Corsepius wrote:
> >> So adding $(SHELL) might not be wrong.
> 
> But SHELL might disagree with CONFIG_SHELL; the latter is used to run
> config.status.  So adding $(SHELL) might be wrong.
Possible, I didn't check the details.

> Noah Misch <address@hidden> writes:
> 
> > I suspect more systems still in use truncate #! than execute
> > scripts by running them through csh.
> 
> I agree, but the old csh-based systems have been dead for 20 years.
Yep, I have used different *nix-OSes since the late '80s and have never
seen nor heard about this problem, before :-)

> The truncation-based systems have been dead for 10, unless you insist
> on reallly long (>64-byte) file names for standard shells, which is
> obviously lunatic.
Don't forget about folks used to using "that brand-new, fascinating
feature-rich shell" having been installed some very lengthy directory.
So though you are probably right, such situations will rarely be met,
but they probably will still occasionally be tripped.

> So these days, I suspect that SHELL != CONFIG_SHELL mismatch is the
> greatest problem here.
Agreed. I think, it should be CONFIG_SHELL, w/ 
CONFIG_SHELL=${SHELL-/bin/sh}

> However, this discussion is inclining me to say that it's not pressing
> to change Autoconf to behave like Automake (or vice versa) when there
> are reasonable arguments either way.  Let's leave it alone until
> someone actually runs into a problem.
ACK.

Ralf