Re: M4

[Gary V . Vaughan]

On Tuesday 07 August 2001 9:17 am, Akim Demaille wrote:
> [Cc'ed to some lists.]
>
> >>>>> "Gary" == Gary V Vaughan <address@hidden> writes:
>
> Gary> Salut!
>
> Gettextized Gary :)

I don't have a dk.po file though ;-)

> Gary> I haven't looked at that stuff much... I didn't arrive until
> Gary> about 1.4n or so.  But changing the semantics when 1.4 is so
> Gary> well established is definitely a bug.
>
> Well, it is not very clearly written somewhere, but when you read the
> meaning of --error-output,
>
>   -o, --error-output=FILE      redirect debug and trace output
>
> it seems quite clear dumpdef etc. have nothing to do here.  AFAICS,
> this is a side effect of nice factorization: dumpdef uses
> M4_DEPBUG_PRINTF and the like, which are bound to -o.  So it does seem
> like a nice things from the code point of view, but the dirty side
> effect is pollution of -o.

I see.  Perhaps there is a compromise to be found, where we can retain the 
1.4 semantics without spoiling the factorisation too much?

> Gary> I have a not-yet-building working directory which holds a fair
> Gary> few enhancements from several months ago which I need to clean
> Gary> up and check in.
>
> That would be great!  There are several problems with GNU M4
> currently.  The most painful AFAIC is obstack.h.  You sure don't run a
> system with a native obstack.h :)  Because then you'd have seen that,
> yes, when you have it, no m4/obstacks.h is installed, so the
> compilation fails (we want m4/obstack.h, but only
> /usr/include/obstack.h is there).
>
> I first thought creating m4/obstack.h as a mere #include <obstack.h>
> was fine, but it is not: then you need #include_next, which I don't
> think is portable.
>
> Currently, I see no simple idea to have it work for both the being
> built m4, and the installed headers.

Ah yes.  I did fix this by installing m4_obstack.h if the system file was
missing, but it met with some resistance so I reverted it.  I don't recall 
the particular issues that were debated... is there an archive of m4-forum?  
If not I have a near complete archive of my old mail on CD that I could hunt 
through to refresh my memory, and save retreading the same ground again...

> >> I have also found benign issues, ANSIfication to perform etc.
>
> Gary> Okay.  I guess you are aware of my views on ANSI style after the
> Gary> Goat Book debates? ;-) I won't labour the point except to say
> Gary> that bootstrap tools need to be portable to non-ANSI
> Gary> environments.
>
> Well, my problem is about aesthetics: some files have mixed style.

Eeew.  I expect that is fixed with my unchecked-in working copy.  Even so, 
the files that are maintained outside of m4 (obstack.c for example) might be 
ANSI...

> Gary> And GCC requires Autoconf requires GNU M4.
>
> Huh?  GCC does come with its configure, no?  Sounds weird to me for
> GCC to require Autoconf.

I can't argue with that.  I would ask that you let me clean up my working 
copy and check it in before the source-quake though ;-)

> I would like to clean up the package a little bit:
>
> add config/Makefile.am  to make it a regular package

Why do you think this is a clean-up?  Since nothing is built from the config 
directory, I actually prefer to distribute the config files from the 
top-level directory, and save on file clutter.  There is also one less 
useless Make recursion for every build/check/dist etc...

> require 2.52

Agreed.

Cheers,
        Gary.
-- 
  ())_. Gary V. Vaughan     gary@(oranda.demon.co.uk|gnu.org)
  ( '/  Research Scientist  http://www.oranda.demon.co.uk       ,_())____
  / )=  GNU Hacker          http://www.gnu.org/software/libtool  \'      `&
`(_~)_  Tech' Author        http://sources.redhat.com/autobook   =`---d__/