Date: Mon, 6 Jun 2005 20:14:47 -0400 From: "J.R. Oldroyd" <fbsd@opal.com> To: Brooks Davis <brooks@one-eyed-alien.net> Cc: freebsd-rc@freebsd.org Subject: Re: Use of rcorder for local rc.d/*.sh scripts Message-ID: <20050607001447.GG37208@linwhf.opal.com> In-Reply-To: <20050606235426.GA10526@odin.ac.hmc.edu> References: <20050603143803.GP886@linwhf.opal.com> <42A4CA37.1050201@FreeBSD.org> <20050606235426.GA10526@odin.ac.hmc.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
--ibTvN161/egqYuK8
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Jun 06, 16:54, Brooks Davis wrote:
>=20
> This isn't feasable in the general case with current infrastructure.
> The problem is that you need to make it up to mountcritremote before you
> have any assurance that /usr/local/etc exists.
>=20
> Ports contains a work around that allows port to install script in
> /etc/rc.d if they truly need to appear before localpkg. Such ports
> generally still need to have scripts that run after mountcritlocal and
> won't work on systems where /usr/local is remote unless it is on /.
>=20
> From the perspective of someone who works on the diskless scripts, I
> think that sorting scripts in localpkg and using this hack is a decent
> comprimise.
>=20
> Another note, this decision will need to be discussed with ports@ since
> there are a lot of scripts in ${LOCALBASE}/rc.d. There's also at least
> one binary (postfix). I'd like to see this change happen, but there are
> a lot of issues to work out.
>=20
The solution for that would be to:
- rcorder the /etc/rc.d scripts
- execute the scripts up to mountcritlocal and a dummy script
that has a "PROVIDE: MOUNTCRITLOCAL" or similar
- rerun rcorder on the complete script list (including the
locals)
- execute the list again, this time skipping over any in
/etc/rc.d before the MOUNTCRITLOCAL is reached (since they
were done before)
I don't know if that's getting too complicated. I have to say I
prefer to keep things simple.
I also looked at /etc/rc.subr:run_rc_script and discovered that it
handles "foo" and "foo.sh" differently. "foo", if executable, is run
in a subshell, whereas "foo.sh" is sourced into the current shell.
I suspect we don't want all the local rc foo.sh scripts sourcing into
the /etc/rc shell. I.e., to make a change like this, all local rc
scripts would need to be renamed as well as having rcorder tags
added if not already there.
I currently use the hack to localpkg which I originally suggested.
I needed it because several existing ports fail to start without it
and that change does fix the problem.
-jr
--ibTvN161/egqYuK8
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (FreeBSD)
iQDVAwUBQqTm90kkqUax7f6FAQK8oQX7BBy9iw9lbbb9wcHCasjBnPye1PlG+kqG
K5T9FLpjvSDwXdKaTunk+ERNAcm0Y92K9ATGIVSZhXCBzBqFNErSGbbQOgsHNgLG
7/AAHl8H5XcNefqTKopH4r7vXhhACDMIdhMGbyMlrR1OAeSltfZi3x7XbkylXg8d
rkFLQ6SGSYPP1kAdSiEL8h+KwN/CfDVn/Y/l4iaassOJSxxysJSB+b/rNzFk1Oav
yhTHg2gnANhb+LC0FekE9X6C8RQe5GJ+
=KeLv
-----END PGP SIGNATURE-----
--ibTvN161/egqYuK8--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050607001447.GG37208>