comments on insserv

[Michael Biebl]

Petter Reinholdtsen wrote:
> [Michael Biebl]
>> after (re)enabling insserv this week, I thought I'd share my
>> experiences. insserv successfully installed and reordered the boot sequence.
> 
> Thank you for the feedback. :)
> 
>> This probably means, the LSB header information is not in sync with
>> the information that is passed to update-rc.d
> 
> Yes, this is correct.  And for all but the runlevel 0 and 6 cases, I
> expect the update-rc.d calls to be correct, and the headers to be
> wrong.
> 
>> Example: console-tools.postinst has
>>   update-rc.d console-screen.sh start 48 S .
>> But the LSB header has
>>   # Default-Start:     S 2 3 4 5
>>   # Default-Stop:      0 1 6
> 
> Definitely wrong.  Should have
> 
>   # Default-Start:     S
>   # Default-Stop:
> 
>> I'd assume, those LSB headers are not entirely correct. I'll leave
>> it up to Petter to decide and file bug reports.
> 
> Will try to find time for it.

Ok, thanks. Will you also look at the other init scripts (sudo, 
util-linux [hwclock.sh], netbase [networking])?
> 
>> 2.) The start/stop priorities have strange "holes".
>> Take ls /etc/rc2.d
>> The first start priorities are
>>   S15,S16 [*no* 17,18],S19,S20,S21,S22, [*no* 23], S24
>> Why does it start at 15 at all, and not S01? Why are some start
>> priorities missing?
>> Similar issue in /etc/rc[06].d/
>> There
>> K02,03,04,05,06,07,08,09,10,11,12,13, [*no* 14,15,16],17,18, [*no* 19],
>> 20,[*no* 21],22,23,24
>> Some questions here: Why is there no K01, why are those "holes" in the
>> sequence?
> 
> This is because of the way insserv works internally.  It build one
> dependency graph for all runlevels, and assign sequence numbers using
> this dependency graph, independend of what runlevels are enabled for a
> given service.

Hm, that doesn't explain though, why there are no priorities like 17,18 
or 23 in rc2.d, or why the sequence numbers start at 15, or does it?

>> 3.) console-setup is not started after console-screen.sh anymore.
>> I installed console-setup to configure my console font and keyboard
>> layout. In the old init sequence, console-screen.sh was run at S48,
>> console-setup S49.
>> Now console-setup is S19, console-screen.sh S20, so my changes get reset
>> by console-screen.sh again.
>> => I'd say console-setup should be fixed to run after console-screen.sh
>> Again, I'll leave it up to you to file a bug report and request the
>> correct action (as I don't really know what's the right thing to do).
> 
> Sound like a dependency bug.  Not sure how it is best solve.  Two
> options are to let console-setup depend on console-screen.sh, or to
> let console-screen.sh reverse depend on console-setup.  I do not know
> the packages, so I am not sure which of these are "best". :)

You mean a
   Should-Start: console-screen.sh
in console-setup or alternatively a
   X-Start-Before: console-setup
in console-screen.sh?

Does insserv support X-Start-Before?

>> 4.) Imho you should not only divert update-rc.d, but also it's man page.
>> It's weird when update-rc.d and update-rc.d.8 are out of sync.
>> The man page of update-rc.d-innserv is insufficient at its current state
>> though. A lot of vital information is missing (e.g. how to disable
>> services with insserv).
> 
> Good point.  Should be filed as a wishlist bug against insserv to make
> sure it is not forgotten.

Will you do that, should I do it?

Cheers,
Michael


-- 
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 252 bytes
Desc: OpenPGP digital signature
Url : http://lists.alioth.debian.org/pipermail/initscripts-ng-devel/attachments/20080412/da06d995/attachment.pgp