Bug#826215: Bug#810018: procps pidof (was: Re: Processed: forcibly merging 851747 826215)

[Craig Small]

On Mon, 12 Jan 2026 at 01:01, Luca Boccassi <bluca at debian.org> wrote:

> On Thu, 30 Oct 2025 19:58:37 +1100 Craig Small <csmall at debian.org> wrote:
> > Path B: We decide to use the procps pidof. Then there are two questions.
> > 1) Should the procps pidof package be Essential?
> > 2) Should the procps pidof package be separate to procps and libproc2?
> >
> > My preference is for 1) the answer is no. This package would only be
> needed
> > because sysvinit-utils needs it, so a dependency should cover it.
> > The "main" procps package would probably need a dependency/recommends on
> it
> > just so pidof is there for users.
>
> Sounds reasonable
>
I had a look into this and it seems the only valid required user of pidof
is /usr/lib/lsb/init-functions
Some init scripts call this, but they should be sourcing init-functions
(most do) and using pidofproc() (some do).
My understanding is the init-functions file is required for sysv init
scripts only.

So:
 * a sysv init system uses init scripts, which use init-functions, which
needs pidof
 * systemd init system uses unit files and doesn't need init-functions or
pidof

I'm not 100% sure of that second bullet point. I'd really like that
confirmed.

Does sysvinit-utils need to be Essential at all? Is it just merely making
sysvinit-core depend on sysvinit-utils (it does already)
and then sysvinit-utils requires whatever package has pidof?

Looking into it more and adjusting Helmut's suggested migration path[1]

Is it a matter of:
1) sysvinit-core depends on sysvinit-utils (already does with a version)
2) sysvinit-utils has its Essential tag removed, its being pulled in by
sysvinit-core
3) sysvinit-utils depends on and provides virtual package pidof
4) If there is anything that needs pidof but doesn't need sysvinit-utils it
also depends on virtual package pidof
5) At some time sysvinit-utils drops the virtual package, doesn't install
pidof, procps picks those up

Start without adding a new package which is operationally easier (no
> new queue to clear) and see how it goes? It can always be added later,
> if it turns out it's needed
>
Looking again at that small intersection. Having procps pidof by itself
would mean the (non systemd) sysctl isn't installed.
I'm not sure there are that many systems like that.

 - Craig


1: https://lists.debian.org/debian-devel/2023/11/msg00105.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.chiark.greenend.org.uk/pipermail/debian-init-diversity/attachments/20260113/cbec29c2/attachment.htm>