Re: PATCH 2/2: Unconditional use of install-sh for install-strip (Re: cross-install-strip summary)

[Akim Demaille]

Alexandre Duret-Lutz <address@hidden> writes:

> >>> "Akim" == Akim Demaille <address@hidden> writes:
> 
> [...]
> 
>  Akim> I don't understand why Automake would not always use
>  Akim> AC_CHECK_TOOL(STRIP, strip) in AM_INIT_AUTOMAKE, just
>  Akim> like for any other tool it requires.
> 
> This is what the original patch did, however Automake would then
> require config.sub and config.guess for *any* package.
> 
> The start of this discussion is here
>   http://sources.redhat.com/ml/automake/2001-02/msg00317.html
> then in
>   http://sources.redhat.com/ml/automake/2001-02/msg00583.html
> Tom says he doesn't want to change Automake so that
> AC_CHECK_TOOL won't require config.sub & config.guess, and in
>   http://sources.redhat.com/ml/automake/2001-02/msg00652.html
> he says that's not the business of Automake to supply a macro
> that does AC_CHECK_TOOL(STRIP, strip).
> 
> So actually I think that means the user have to
> AC_CHECK_TOOL(STRIP, strip) in his configure.ac, until Autoconf
> enventually supply an AC_PROG_STRIP macro (then Automake can be
> tweaked to require AC_PROG_STRIP when it encounters
> AC_CANONICAL_HOST).

I understand this, but given that Autoconf is soon to be released too,
since the next generation of Autoconf will not require config.guess
when you use AC_CHECK_TOOL, I can't find any actual problem.  What
will happen is that people using 1.5 + 2.13 will have configure want
config.guess.  People using 1.5 + 2.50 will not face the issue.

So, is this really an issue?