Re: FYI: function definitions

[Ralf Wildenhues]

* Gary V. Vaughan wrote on Tue, Feb 01, 2005 at 12:49:29PM CET:
> Ralf Wildenhues wrote:
> > 
> > I just realized that, if I first need Autoconf to have a macro that
> > guarantees to provide as_cr_alnum, then we can't use it anyway until
> > the next stable Autoconf is released.
> 
> Not so.

Hmm.

> The old style test scripts don't get an m4sh expansion pass, so _AS_CR_PREPARE
> has no bearing on the contents of test.sh.  Rewriting it on HEAD as a new
> style test might involve adding this somewhere (since we know that either
> Autoconf is new enough and provides a good value for as_cr_alnum, or Autoconf
> is so old that it doesn't provide a value at all):
> 
> AC_PROVIDE_IFELSE([_AR_CR_PREPARE], [], [
> as_cr_alnum='abcdef...XYZ0123456789'
> ])

My understanding was that Autoconf should get a new macro that then
provided this.  Using the internal interface _AS_CR_PREPARE is wrong.

> On branch-2-0, I think the following is fairly future proof:
> 
> test -n "$as_cr_alnum" || as_cr_alnum='abcdef...XYZ0123456789'
> 
> In both cases, I think the slight untidiness is worth the extra safety added
> to sh.test...  what do you think?

In both cases it is plain easier if I write
  lt_cr_alnum=abcdef...89
and be done with it.  I don't even need a variable here.  I really don't
need to abstract out anything here, and IMNSHO this whole discussion is
much longer than what the problem at hand deserves.

Regards,
Ralf