Re: dirlist

[Charles Wilson]

Alexandre Duret-Lutz wrote:

> (aclocal --version) 2>/dev/null | \
>   sed 's/^.* \([0-9][0-9]*\.[0-9][0-9]*[a-z]*\).*$/\1/;1q'


That looks good.  Do you want me to send a new patch, or will you make 
the change yourself?


> | +    libtool_found=no
> | +    for d in $extra_includes $aclocaldir ; do
> | +       if [ "x$d" != "x-I" ] && [ -f "$d/libtool.m4" ] ; then
> | +          libtool_found=yes
> | +       fi
> | +    done
> | +    test "x$libtool_found" = "xyes" || exit 77
> | +    ACLOCAL="$ACLOCAL -I $srcdir/../m4 $extra_includes -I $aclocaldir"
>
> No idea why this appends both `$extra_includes' and `-I $aclocaldir'.
> Shouldn't we add only the directory where the macro was found?
> I guess I'm missing something.


Well, the original code had -I $aclocaldir, so I kept it -- I didn't 
want to run afoul of any unintended side effects, if I only -I included 
the specific directory where libtool.m4 was found.  (e.g. what if 
'libtool.m4' depends on some other .m4 file that is located in one of 
the other directories?  Not true today, of course, but what about 
tomorrow's libtool.m4?)

I figured since what we're really trying to do in the gettext and 
libtool tests is pretend that the new aclocal is fully installed and has 
access to all of the same macros that the currently installed aclocal 
does: we might as well include all of its search paths.  The -I order 
will ensure that our local, new versions will take precedence if there 
are any duplications.

--Chuck