[Aurelien Jarno] > Let's give an example. Our script, among other things does: > - Ensure that /etc/mtab is a symlink to /proc/mounts, and that has to be > done before mtab.sh so that it don't remove /dev. It also needs to start > after checkroot to be able to write on the file system. > - Setup the permissions in /dev. It roughly corresponds to udev, and > needs to provide udev, otherwise insserv is not able to solve > the dependencies.
Right. Then I understand the problem. But I believe making it provide udev is a bad idea, as the provides really should be unique across the archive. Which scripts fail to work when udev is not provided? As far as I know, udev should be an optional dependency, and I would claim that any script having it as a required are buggy. A quick search in all the init.d scripts in the archive indicate that these scripts are buggy: alsa-utils, brltty, eeepc-acpi-scripts, keyboard-setup, lvm2, mouseemu, multipath-tools and pcmciautils. I believe all these should be changed. > So freebsd-utils needs checkroot, which needs mountdevsubfs, which > should start udev which is provided by freebsd-utils => insserv > fails. Right. > The patch among other things make sure that mtab.sh don't destroy > /dev and ensure that /etc/mtab is a symlink to /proc/mounts, so the > dependency on checkroot can be dropped. Thank you. Happy hacking, -- Petter Reinholdtsen -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
