Re: Broken makefile given Autoconf version mismatch

[Stepan Kasal]

Hello Ralf,

On Thu, May 25, 2006 at 03:43:11PM +0200, Ralf Wildenhues wrote:
> > On Wed, Apr 12, 2006 at 09:47:36PM +0200, Stepan Kasal wrote:
> > > BTW: if someone does
> > >   AC_SUBST([mydatadir], [$datadir/my])
> > > then our hack won't help him.  But:

thank you for writing the patch.  A few details:

>    AC_MSG_WARN([$ac_file_inputs seems to ignore the --datarootdir setting])
...
> +  AC_MSG_WARN([ $ac_file seems to ignore the --datarootdir setting])

I'd delete the space before $ac_file.  (The value of ac_file_inputs
starts by a space, but that is not intentional, it just is not worth
the hassle to fix it.)

I don't like that the two messages are similar: the former one
indicates a problem, which is most probably worked around, while the
latter one indicates a bug which will most probably hit.

I think the latter one should be much longer and should speak about
``referenece to undefined variable'' or some such.  Perhaps:

AC_MSG_WARN([$ac_file contains reference to variable `datarootdir'
which seems to be undefined there. Please make sure it is defined.])

And the check

> @@ -562,6 +562,10 @@
>  dnl  else
>  dnl     rm -f $ac_file; mv "$tmp/out" $ac_file
>  dnl  fi
> +
> +test -z "$ac_datarootdir_hack$ac_datarootdir_seen" &&
> +  grep '\${datarootdir}' $ac_file &&
> +  AC_MSG_WARN([ $ac_file seems to ignore the --datarootdir setting])
>  ])# _AC_OUTPUT_FILE

comes too late, $ac_file can be `-'.  You should check $tmp/out as
soon as it is created, a few lines above.

> The Debian switch to 2.59c has already uncovered another (wrong) usage
> case: in an AC_DEFINE*.

I have no idea what workaround or warning could be added to fix this.
But I think we can let it be, this is less common than the AC_SUBST
problem.

> (It's on purpose that I'm putting warning signs everywhere near the
> wrong AC_SUBST ... too many people picking up stuff from anywhere.)

I wasn't able to understand this, sorry.

Have a nice day,
        Stepan