Re: gnulib-izing sharutils

[Bruce Korb]

Bruno Haible wrote:
> Ah, I see. The version that matters is the one of the GNU gettext macros that
> will be used.
...will be used *at program build time*.
> Therefore `gettext --version` is irrelevant; what matters is
> what 'aclocal' will fetch later. How would you like the notice to be
> formulated?

More like:
> >>>   If your program might get built using GNU gettext version 0.16.1 or 
> >>> older,
> >>>   please add the following options to XGETTEXT_OPTIONS in your 
> >>> po/Makevars:
> >>>     --flag=error:3:c-format --flag=error_at_line:5:c-format

Since most developers won't be able to be certain which flavor of gettext
will get used by Joe Hapless User, most developers would probably want
to put that stuff in their po/Makevars file.

> The autoconf manual has such a section: Node "Autoconf Input Layout".

Oop.  Must have missed it.  Thank you.

>> and I still see the warning: configure.ac:22: warning: The macro 
>> `AC_GNU_SOURCE' is obsolete.
> 
> Needs more analysis.
> 
>> The whole thing ends with this:
>>> configure: creating ./config.status
>>> config.status: creating src/remsync
>>> config.status: creating src/compress-dummy
>>> config.status: creating contrib/Makefile
>>> config.status: creating doc/Makefile
>>> config.status: creating intl/Makefile
>>> config.status: creating lib/Makefile
>>> config.status: creating po/Makefile.in
>>> config.status: creating src/Makefile
>>> config.status: creating src/mail-files
>>> config.status: creating src/mailshar
>>> config.status: creating tests/Makefile
>>> config.status: error: cannot find input file: `Makefile.in
> 
> Uh oh. Looks like some major syntax mistake in the generated configure file.
> You should review your configure.ac and all your own *.m4 macros for proper
> m4 quoting and shell quoting. Usually the way to find this kind of mistake
> is by temporarily removing lots of stuff from configure.ac, until the error
> goes away.

sharutils only has one m4 file now:

> AC_DEFUN([CHECK_COMPRESS],
> [
>   AC_PATH_PROGS(COMPRESS, [compress], [no])
>   if test $COMPRESS != no; then
>     cp $srcdir/COPYING tCOPYING
>     AC_MSG_CHECKING([whether compress works])
>     if compress tCOPYING >/dev/null 2>&1; then
===== Shouldn't this line be:  if $COMPRESS tCOPYING .... ???
>       AC_DEFINE([HAVE_COMPRESS], [1],
>                 [Define if compress is available.])
>       AC_MSG_RESULT([yes])
>     else
>       COMPRESS=no
>       AC_MSG_RESULT([no, installing compress-dummy])
>     fi
>     rm -f tCOPYING tCOPYING.Z
>   fi
>   if test $COMPRESS = no; then
>     ADD_SCRIPT="$ADD_SCRIPT compress-dummy"
>     AC_CONFIG_FILES([src/compress-dummy])
>   fi
> ])
> 
> AC_DEFUN([CHECK_COMPRESS_AND_LINK],
> [
>   CHECK_COMPRESS
>   if test $COMPRESS = no; then
>     AC_ARG_ENABLE([compress-link],
>       [AS_HELP_STRING([--enable-compress-link],
>       [install compress link if the program is missing])])
>     if test "x$enable_compress_link" = xyes; then
>       INSTALL_COMPRESS_LINK="compress-link"
>     else
>       INSTALL_COMPRESS_LINK=
>     fi
>   fi
>   AC_SUBST(INSTALL_COMPRESS_LINK)
> ])

With the bulk of configure.ac trimmed, it's still a mess:

configure: creating ./config.status
config.status: creating contrib/Makefile
config.status: error: cannot find input file: `doc/Makefile.in'

So if I trim it, it dies way early and if I don't trim it then
it doesn't make the top level Makefile.in.

Are there any other tools for debugging this?  THanks -Bruce