Re: Fix m4_join

[Ralf Wildenhues]

Hi Eric,

* Eric Blake wrote on Sat, Oct 13, 2007 at 07:35:59PM CEST:
> 
> Another improvement borrowed from libtool.  m4_join was mistakenly
> m4_defun'd instead of m4_define'd.  Additionally, it is nice to avoid
> back-to-back separators for an empty argument.  A quick search didn't turn
> up any clients of m4_join that expected back-to-back separators (since
> most clients provided non-empty arguments in every position, rather than
> taking libtool's approach of m4_join'ing a list of macro expansions, some
> of which might be empty).  And it doesn't hurt that this new
> implementation is slightly faster, or that m4_join was previously
> undocumented (hmm, it looks like I already snuck in the NEWS for this
> change in an earlier patch).

Thanks for your work on this.  I had some year-old notes on m4_join that
I'll try to dig out, but for now I only have this question for your
implementation: is it possible to pass a comma as either argument
(separator, entries) to the current version of m4_join?  Is it possible
to pass a macro that is defined to expand to comma?  This doesn't seem
to work right:
  m4_define([x],[,])
  m4_define([y],[,])
  m4_join([x],[y],[z])

Thanks,
Ralf

> 2007-10-13  Eric Blake  <address@hidden>
> 
>       Change m4_join to match libtool's ltsugar semantics.
>       * lib/m4sugar/m4sugar.m4 (m4_join): Just define this, not defun.
>       Ignore empty arguments, using...
>       (_m4_join): ...this new helper.
>       * tests/m4sugar.at (m4@&address@hidden): New test.
>       * doc/autoconf.texi (Text processing Macros): Document new
>       semantics of m4_join.