On Monday 26 March 2007, Marco d'Itri wrote: > On Mar 26, Brian Cox <[EMAIL PROTECTED]> wrote: > > The subject about says it all. Rather than configuring the > > networking script in /etc/init.d to stop in runlevels 0 and 6, the > > postinst hook for netbase sets it up to start, as you can see from > > this snippet (starting at line 100): > > I am almost sure that there was a reason for this...
Marco, I just discovered what that reason is, courtesy of Kel Modderman on the wpasupplicant package (which exhibits the same behavior).[1] It turns out that the rc scripts of sysv-rc and file-rc treat both start and stop sets assigned to runlevels 0 and 6 as a stop action. This was apparently done because some init scripts were being stopped too soon during shutdown, by virtue of the fact that stop actions are run before start actions. Certain scripts devoted purely to shutdown-related tasks seem to be coded and/or ordered to run during the 'start' phase, and in cases where some service (like /etc/init.d/networking, for example) needs to be stopped *after* those other scripts, the only solution is to assign a start set to the networking script and have /etc/init.d/rc run that script using the stop action. I'm sorry about bothering you with this report over a non-bug. I had naively assumed that the start and stop sets act the same way in all runlevels when, had I scanned through Debian Policy, I would have found this behavior documented in section 9.3.1. /user puts on brown paper bag Thanks, Brian [1] See bug #416234 for a discussion related to another non-bug I rather unfortunately filed. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
