Re: Starting services in runlevels 0 and 6
[Christian Perrier] From what I see, having this will not happen for etch. Do you think it could be a release goal for etch+1? I believe it should be a release goal for etch+1 to switch sysv-rc to use a dependency based sequencing, yes. With that already decided and widely accepted, it wouldn't probably be very hard to begin a campaign to make packages include dependency information just after we release etch It is already happening, so I am confident we will get even more scripts to include those headers after Etch. The nice thing about the change is that it only add a comment to the script, so the change is very low-risk. Lintian already give a warning for init.d scripts without these headers, and this help a lot of maintainers to notice the problem. Friendly, -- Petter Reinholdtsen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Starting services in runlevels 0 and 6
[Jörg Sommer] It was pointed out to me, that even the scripts starting with S are called with argument 'stop' for runlevels 0 and 6 by /etc/init.d/rc. However, the reason why it is implemented that way is still not clear. Because the S scripts are run after the K scripts. This way it is possible to run special scripts at very last. Well, it is not really a good explanation, as the ordering of K01 S01 could also be done with K01 K02. No need to use start-symlinks as stop scripts to get the proper ordering. I've been told that the sysv-rc system behave the way it does because Solaris did it when the Debian boot system was written, and it was used as the example. I guess that is as good explanation as any on the historical reasons for the strange setup in the Debian boot system. Friendly, -- Petter Reinholdtsen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Starting services in runlevels 0 and 6
Le jeudi 12 octobre 2006 à 20:33 -0700, Jurij Smakov a écrit : It was pointed out to me, that even the scripts starting with S are called with argument 'stop' for runlevels 0 and 6 by /etc/init.d/rc. However, the reason why it is implemented that way is still not clear. IIRC, it is so that they are executed after the scripts starting with K. -- .''`. Josselin Mouette/\./\ : :' : [EMAIL PROTECTED] `. `'[EMAIL PROTECTED] `- Debian GNU/Linux -- The power of freedom
Re: Starting services in runlevels 0 and 6
[Henrique de Moraes Holschuh] On Thu, 12 Oct 2006, Jurij Smakov wrote: It was pointed out to me, that even the scripts starting with S are called with argument 'stop' for runlevels 0 and 6 by /etc/init.d/rc. However, the reason why it is implemented that way is still not clear. Braindamage inherited from SysV. Yes, historical reasons. And it isn't even required. SuSe for example only store K* type symlinks in those directories, instead of handling them specially. We can do the same, if we switch to a dependency based init.d system, but the current symlinks need to be renamed when that happen. The insserv package provide a script to do this while it reorders the boot sequence based on the LSB header dependency info. It is not enough to just rename them, as all postinst script call update-rc.d with other assumptions, and the shutdown order will be wrong unless the sequence numbers are changed as well. The brave can try to install insserv and run BAD_INSSERV_HACKER=true dpkg-reconfigure insserv to enable the dependency based boot. But remember, if it break you get to keep both the pieces. Running it again and disabling the dependency based boot might be possible, but there is no guarantee. The reason this might break is that several init.d scripts are missing dependency information, so the boot order generated by insserv might be incorrect. insserv include override files for some of these scripts, for the base system and a normal desktop install, but there are probably hundred of init.d scripts without known dependency info. I use it myself on one of my test machines, and it works for me. But I have added override files for the packages I use. :) I would like us to switch to a dependency based boot sequence system after Etch. First all init.d scripts need to include dependency info. Next, we need to locate and fix any dependency loops. Last we can automate the switch to dependency based sequencing, possible only on new installations and when the system administrator decide to convert. The nice thing about documenting dependencies is that we can automatically detect bugs in the current boot sequence. We already found and fixed a few of those with the info currently in the packages. :) Friendly, -- Petter Reinholdtsen -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Starting services in runlevels 0 and 6
Hello Jurij, Jurij Smakov [EMAIL PROTECTED] wrote: On Thu, Oct 12, 2006 at 07:34:19PM -0700, Jurij Smakov wrote: /etc/rc0.d/S30urandom /etc/rc0.d/S35networking /etc/rc0.d/S36ifupdown /etc/rc0.d/S37sendsigs (start action for this one is a no-op) /etc/rc0.d/S48cryptdisks /etc/rc0.d/S59cryptdisks-early and similar stuff in /etc/rc6.d. I cannot find a rational explanation for starting services right before shutdown. Is it intentional, or are those just bugs? It was pointed out to me, that even the scripts starting with S are called with argument 'stop' for runlevels 0 and 6 by /etc/init.d/rc. However, the reason why it is implemented that way is still not clear. Because the S scripts are run after the K scripts. This way it is possible to run special scripts at very last. Bye, Jörg. -- Life can only be understood backwards, but it must be lived forwards. (Soren Kierkegaard) -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Starting services in runlevels 0 and 6
The reason this might break is that several init.d scripts are missing dependency information, so the boot order generated by insserv might be incorrect. insserv include override files for some of these From what I see, having this will not happen for etch. Do you think it could be a release goal for etch+1? With that already decided and widely accepted, it wouldn't probably be very hard to begin a campaign to make packages include dependency information just after we release etch signature.asc Description: Digital signature
Re: Starting services in runlevels 0 and 6
On Thu, Oct 12, 2006 at 07:34:19PM -0700, Jurij Smakov wrote: Greetings, I've accidentally noticed that there are a few services which are *started* when entering runlevels 0 and 6 (halt and reboot, respectively). For example, on my mostly up-to-date sid system I have: /etc/rc0.d/S30urandom /etc/rc0.d/S35networking /etc/rc0.d/S36ifupdown /etc/rc0.d/S37sendsigs (start action for this one is a no-op) /etc/rc0.d/S48cryptdisks /etc/rc0.d/S59cryptdisks-early and similar stuff in /etc/rc6.d. I cannot find a rational explanation for starting services right before shutdown. Is it intentional, or are those just bugs? It was pointed out to me, that even the scripts starting with S are called with argument 'stop' for runlevels 0 and 6 by /etc/init.d/rc. However, the reason why it is implemented that way is still not clear. Best regards, -- Jurij Smakov [EMAIL PROTECTED] Key: http://www.wooyd.org/pgpkey/ KeyID: C99E03CC -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Starting services in runlevels 0 and 6
On Thu, 12 Oct 2006, Jurij Smakov wrote: It was pointed out to me, that even the scripts starting with S are called with argument 'stop' for runlevels 0 and 6 by /etc/init.d/rc. However, the reason why it is implemented that way is still not clear. Braindamage inherited from SysV. -- One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie. -- The Silicon Valley Tarot Henrique Holschuh -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
