Re: RFC: cool hack for simplifying do_subst generation of scripts

[Stefano Lattarini]

Hi Eric.

On 01/12/2012 08:16 PM, Eric Blake wrote:
> I was chatting with Federico on IRC about a way to avoid duplication
> between Makefile.am and configure.ac.  His particular case had a script
> file that wanted a few full substitutions (which implies a make-time
> substitution, per [1]), but he didn't want to have to copy the whole
> list of AC_SUBST from configure.ac into Makefile.am in his do_subst
> script, and didn't want to build a file.in.in that goes through
> configure AC_CONFIG_FILES only to have file.in then go through a
> make-time expansion.  We came up with the autoconf-documented use of
> $(top_builddir)/config.status --file=- (from [2]).
>
Thanks for the link, I wasn't aware of this feature.

> And to demonstrate the ease of use, I've done this preliminary patch
> against automake itself.
>
I like the idea.  Thanks for working on this!

> Is this something we are interested in doing (since it _does_ provide a
> reduced maintenance burden, where adding a new AC_SUBST that does not
> need make-time expansion can now be done in just one file instead of
> two)?  And if so, should I expand this RFC patch into something that
> passes the testsuite as well as mentioning it in the automake manual?
> [For the purposes of my quick test, I started a 'make check',
>
About this: note that a mere "make check" might not catch all the possible
blunders, since the testsuite by default uses wrapper scripts that redefine
various internal directories affected by AC_SUBST (e.g., the directories
where automake looks for *.am fragments and for its extra perl libraries,
and the directories where aclocal looks for *.m4 files by default).  You'd
be probably better off trying with "make installcheck" (unfortunately
available only in master), which is more faithful in this regard.

> still underway as I type, and noticed that at least aclocal5.test failed,
> but everything else through f90only.test is happy so far.  But I won't
> bother to investigate failures unless we decide to go forward with this
> plan.]
>

> [1] https://www.gnu.org/software/automake/manual/html_node/Scripts.html
> [2] 
> https://www.gnu.org/software/autoconf/manual/autoconf.html#config_002estatus-Invocation
> 
>
> diff --git i/Makefile.am w/Makefile.am
> index 8fe9c0f..0ef39c1 100644
> --- i/Makefile.am
> +++ w/Makefile.am
>
> [SNIP CODE]
>
Have you considered doing something similar also for the recipe for the
creation of `tests/defs-static' in master?  That would be useful as well.

Thanks,
  Stefano