Hello Richard,
* Richard Bos wrote on Tue, Feb 19, 2008 at 05:30:49PM CET:
>
> For an open source project, that I work on, it is needed to define a
> webserver
> root directory. This is done using a variable webserver_document_root, that
> is (in my case) defined as /srv/www/htdocs.
> When I now run:
> DISTCHECK_CONFIGURE_FLAGS="--with-dist=suse --without-openpkg" make distcheck
> it fails :(
That is likely because distcheck ensures that with --prefix, all
installed files can be redirected to be inside a subtree.
Do you have a way to override the webserver_document_root? If yes,
assuming that the possibility is --enable-webserver-document-root=DIR,
you can add
DISTCHECK_CONFIGURE_FLAGS='--with-dist=suse --without-openpkg
--enable-webserver-document-root=${sharedstatedir}/htdocs'
Alternatively, you could default the directory in configure.ac
if test -n "$enable_webserver_document_root"; then
...
else
webserver_document_root='${sharedstatedir}/htdocs'
fi
AC_SUBST([webserver_document_root])
> Whatever I do to the webserver_document_root (more details about this below)
> it either fails the 'make distcheck', or 'make install' installs in an
> incorrect location.
>
> I'm now going to use a workaround to (mis)use the configure argument
> --htmldir
> to define the webserver root directory.
--htmldir should be for documentation in HTML format.
> Two questions:
> 1) is a feature request feasible, asking to extend configure so it gets a
> command line option that can be used to define the webserver root directory
> (--webserverdir e.g.)?
No. It is deliberate that all configure scripts understand the same
flags, --enable-* and --with-* are for package-specific extensions (that
are to be ignored by other packages). The GNU Coding Standards mandate
this; the rationale is to allow you to build trees of packages,
connected by AC_CONFIG_SUBDIRS, and not have configure scripts fail due
to arguments it does not know.
> If I search the internet, more people are fighting
> this battle to obtain a webserver root directory with autotools and
> have 'make distcheck' work. Hence, it would serve multiple people.
FWIW, I have seen none of this battle so far.
Cheers,
Ralf
