[systemd-devel] Why does sysv generator translate Required-Start keyword into an After= ordering dep only ?

[Andrei Borzenkov]

10.03.2016 18:05, Francis Moreau пишет:
> On Tue, Mar 8, 2016 at 10:19 AM, Andrei Borzenkov <arvidjaar at gmail.com> wrote:
>> 08.03.2016 11:33, Francis Moreau пишет:
>>> On Tue, Mar 8, 2016 at 9:23 AM, Andrei Borzenkov <arvidjaar at gmail.com> wrote:
>>>> 08.03.2016 11:07, Francis Moreau пишет:
>>>>> On Tue, Mar 8, 2016 at 7:51 AM, Andrei Borzenkov <arvidjaar at gmail.com> wrote:
>>>>>> 07.03.2016 10:04, Francis Moreau пишет:
>>>>>>> Hello,
>>>>>>>
>>>>>>> Sorry for the long delay.
>>>>>>>
>>>>>>> On Fri, Feb 26, 2016 at 5:05 AM, Andrei Borzenkov <arvidjaar at gmail.com> wrote:
>>>>>>>> 26.02.2016 00:55, Francis Moreau пишет:
>>>>>>>>>
>>>>>>>>> But now I'm wondering how the following case is handled: a sysinit
>>>>>>>>> script "a" has "Required-Start: b". But "b" is a native systemd
>>>>>>>>> service. I don't think the tool that enable/disable sysv services can
>>>>>>>>> enable and order correctly the native service.
>>>>>>>>>
>>>>>>>>
>>>>>>>> What difference does it make?
>>>>>>>
>>>>>>> The difference is that in my current understanding nothing will pull "b" in.
>>>>>>
>>>>>> That was answered in part you trimmed off. sysvinit never actively
>>>>>> pulled "b" in either so nothing really changed here.
>>>>>>
>>>>>
>>>>> In my understanding insserv is part of the sysvinit implementation.
>>>>>
>>>>> Therefore to enable a service with sysvinit, we do:
>>>>>
>>>>>  - insserv a (this will create S<xx>a *and* S<yy>b" with yy < xx)
>>>>
>>>> That would be new to me. insserv creates links ("enables") exactly those
>>>> services that you specify. So if you say "insserv a" you will get only
>>>> "a" enabled; this /may/ rearrange other services including "b" if they
>>>> are already enabled but this will not enable "b".
>>>>
>>>
>>> That's how I understood Lennart's excerpt I was referring to previously.
>>>
>>
>> Hmm ... I tested on SLES11 and indeed, while "insserv a" will not enable
>> "b" it will refuse to enable "a" if "b" is not enabled. And conversely
>> it will not disable "b" if "a" is enabled.
>>
>> So at least it tries to ensure that set of enabled services is consistent.
> 
> Is this system uses systemd ?
> 

No; it is too old for this. It is classical sysvinit.