Re: aclocal problems

[Ralf Wildenhues]

* Eric Blake wrote on Fri, Apr 10, 2009 at 02:12:10PM CEST:
> According to Ralf Wildenhues on 4/10/2009 12:32 AM:
> > <http://thread.gmane.org/gmane.comp.sysutils.automake.general/10429/focus=10431>

> > Good point.  Public third-party macros should be AC_DEFUNed for this
> > reason.  OK to apply?
> 
> Yes, please.

Done, thanks; but see below.

>  But given the recent thread about AC_DEFUN being able to
> hoist contents while m4_define does not, would it ever make sense to add a
> macro in the AC_ namespace that is equivalent to m4_define but still
> traced by aclocal?

Hmm.  That would likely require another sister of AU_DEFUN, too, no?

Ah, there is another reason to use AC_DEFUN: m4_def(ine|un)'d macros are
not upgraded by autoupdate.  AFAICS this bit cannot be worked around, as
the aclocal bit can by having witness macros.

Really I don't think we should advise users away from AC_DEFUN for
public macros.  This is such an enshrined API, everybody knows that.
Changing this usage is likely to take several years.

> Unfortunately, the name AC_DEFINE is already claimed.

Yes.

Cheers,
Ralf