Re: Proposed release goal for lenny: Switch to dependency based boot system
- To: debian-release@lists.debian.org
- Subject: Re: Proposed release goal for lenny: Switch to dependency based boot system
- From: Petter Reinholdtsen <pere@hungry.com>
- Date: Wed, 04 Jul 2007 19:27:52 +0200
- Message-id: <[🔎] 2flved0nko7.fsf@saruman.uio.no>
- References: <E1I1OBO-00004i-Ny@saruman.uio.no> <20070622140653.GB13358@dario.dodds.net> <20070626223755.GC5195@kitenet.net> <20070629074215.GA23521@dario.dodds.net>
[Steve Langasek]
> Well, that seems... suboptimal. :) And something of a barrier to adoption.
Yeah, but it make it easier to detect the problematic ones. :)
> What would be a better heuristic? Even running these scripts last
> instead of first would seem better. Would it make sense to keep
> track of the sysvinit number for each script, and for each script
> without dependency information, treat /all/ lower-numbered scripts
> as dependencies?
I plan to assume start and stop dependency '$remote_fs $syslog', start
runlevel 2345 and stop runlevel 016, as this arguably is the current
meaning of the 'default' setting in update-rc.d. This is not
implemented in insserv yet, thought.
For those init.d scripts where this is not the correct setting, I plan
to provide override files in the insserv package until the scripts get
correct headers themselves.
But first and foremost, I want the scripts themselves to have correct
dependency information, as it is the only way to keep the boot
sequence correct.
Happy hacking,
--
Petter Reinholdtsen
Reply to: