Re: 02-m4-warnings.patch

[Akim Demaille]

| | On Oct 27, 2000, Akim Demaille <address@hidden> wrote:
| | >   (_AC_DIAGNOSE, AC_DIAGNOSE): Rename as...
| | >   * m4sugar.m4 (m4_warning_ifelse, _m4_warning_ifelse)
| | 
| | Under the same reasoning, I think we should keep AC_DIAGNOSE as an
| | alias to its m4sugar implementation, and use AC_DIAGNOSE in autoconf
| | code, so that, if we ever decide to make some change in the diagnostic
| | output that is specific to autoconf, we can do it by replacing the
| | alias with a macro, instead of having to modify m4sugar because of
| | some autoconf's need.
| 
| I understand your concern, but I confess I am more attached to keeping
| each stuff in its own name space.
| 
| In the precise case of m4_warn vs AC_DIAGNOSE, since we are talking
| only about M4 itself, I can't see how we could want to depend upon
| something proper to Autoconf.  If we were to change something, then
| chances are high that we'd want to change m4_warn too.  If we need to
| introduce something really new and for Autoconf only, then we can
| introduce a new macro to this end, which will, btw, release us from
| the problems of compatibility (while we would have met some had we
| kept AC_DIAGNOSE).
| 
| So I don't really know.  I'm really attached to spreading the use of
| pure M4sugar in Autoconf, but I'm receptive to your argument.

So what do we do?  Shall I leave AC_DIAGNOSE in Autoconf?

Honestly, if Autoconf were to be developed today, I'd get rid of
AC_WARN and AC_DIAGNOSE in favor of m4_warn, and replace AC_MSG_WARN
with AS_WARN.  My whole point is precisely to keep the prefix
meaningful.

But I understand your preoccupation for de-correlating the layers :(

Hm.  After having talked with myself for a while, I think I'm more
attached to keeping things in its own layer...

Your call Alexandre.  I'd like to hear from you too Pavel.