On Don, 2005-03-24 at 19:53 +0100, Edwin van Vliet wrote: > While installing LFS (development version) I have been reading the > mailing lists, and repeatedly bumped into remarks which imply that LFS > should be guided toward being an LSB compliant installation. > > Since I have recently installed a second LFS server (for testing and > development), I just thought I might want to convert the existing > bootscripts into LSB compatible ones. In fact, I have created a minimal > set of scripts to boot the system, and for now it seems to work.
I think it's great to get LSB compatible bootscripts into LFS. BTW: system init scripts don't need to follow LSB conventions, but I also like to be consistent throughout the whole system. > > But I need a little feedback here. The LSB init functions for providing > feedback to the end-user seem to be functions that provide this > information *after* a process has started. It might not always be clear > what the boot script is waiting for. Maybe you all can help find a > solution for this. Let me describe what I have written so far: > > The /lib/lsb/init-functions provides some basic functionality as > described in the LSB specification. For example: > > log_failure_msg "Could not start process" > [...] > Well, if the command takes 10 seconds to complete, the bootscript is > waiting, but the end user doesn't know what he's waiting for. Also, the > command might create some output. I worked around that to provide debian-style output as it's very bad not to see what's going on. For example see [1] > More questions: the LSB specifications mention "facilities" like > $local_fs and $network, but is not very clear on the implementation of > it all. I have therefore just created /etc/init.d/local_fs and > /etc/init.d/network as I think that the install_initd scripts should > provide any implementation logic (and should therefore be included with > the LSB compliant package). Also: if the /etc/init.d scripts themselves > include their own requirements, the install_initd script should probably > automatically create the symbolic links and should make sure the > scripts are called in the correct order. This problem can be resolved > later on. I've written a rc script [2] to handle init script dependencies on the fly without any symlinks at all. I don't say that's the right way but it's shown to be very convenient. My package init scripts are all linked from the respective paldo package page, feel free to have a look and comment on them. The system scripts can be found in the boot-scripts package [3]. Regards, JÃrg [1] http://www.paldo.org/paldo/sources/cron/init-cron-20041228 [2] http://www.paldo.org/paldo/sources/boot-scripts/rc-20041014 [3] http://www.paldo.org/index.php?section=packages&page=main&query=boot-scripts -- JÃrg Billeter <[EMAIL PROTECTED]> -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
