Re: RFC: 77-gary-diagnose-version-mismatch.patch

[Alexandre Duret-Lutz]

On Mon, Feb 02, 2004 at 11:43:41AM +0000, Scott James Remnant wrote:
> In the interests of forward compatibility, it probably shouldn't *error*
> out...
>
> If in 1.6.2 we add a new option, we necessarily want 1.6 to suddenly
> stop working because people have included the new one.
>
> We know that Autoconf doesn't error when unknown configure arguments are
> given, what does Automake do if you give AM_INIT_AUTOMAKE an unknown
> option?  (afaict it doesn't)

Automake diagnoses unknown options as errors.  It checks this at
automake-time, not autoconf-time.

I don't think they could be warnings.  The idea is that when a
maintainer selected an option, he has a reason for doing so.  So we
should not ignore it.  Just see options as features: new features
require new versions.  It's not unlike `ls --new-option' either.