Bug#1078600: orphan-sysvinit-scripts: Please consider adding netperf

[Daniel Gröber]

Hi Mark, lorenzo,

On Thu, Aug 15, 2024 at 07:58:42AM +0100, Mark Hindley wrote:
> Remove the runlevels from Default-Start: in the LSB header so that line
> is empty. See insserv(8). I can provide a patch for that in #1078674, if
> you want?

I don't care to spend cycles on figuring out how the coexistence story
works out since in my mind sysvinit is a lost cause, however I'm happy to
collaborate on this as long as it doesn't get in the way and someone
actually takes responsibility for the work supporting it generates (if any,
netperf is pretty tiny).

Mark, if you genuinely want to be responsible for maitaining and supporting
netperf on sysvinit based systems feel free to add yourself to Uploaders,
push straight to git and upload. Remember to update NEWS.

The only technical requirement I have is that user confusion is avoided. So
if /etc/init.d/netperf exists on systemd systems it must control the
systemd unit and /etc/default/netperf should not exist in the default
install since NETSERVER_ENABLE is superceeded by other mechanisms and
keeping it around would be confusing.

> > Frankly, I don't want to support two ways of doing things. Patches and
> > co-maintainers welcome.
> 
> Of course users of non-systemd inits would be happy to provide patches and
> support to keep the LSB initscript within netperf working.
> 
> > If orphan-sysvinit-scripts didn't exist this may have tipped in favor of
> > keeping it, but I decided placing it here (if you'll have it) and putting a
> > note in netperf NEWS is a reasonable compromise, since I'm already adding a
> > NEWS entry for the default disablement change.
> 
> It is Matthew's call as to whether he wants to include netperf LSB
> initscript in orphan-sysvinit-scripts. However, orphan-sysvinit-scripts
> was not a solution of choice and is far from ideal. It was a last resort
> and sub-optimal compromise to deal with situations where maintainers
> refused to keep working LSB initscripts in their packages.

Honestly I don't really understand the desire to keep sysvinit around *in
Debian* if you want to keep developing/running systems using it why not add
initscripts in a derivative distribution like Devuan where (I assume)
you'd actually find a community that cares about this?

> Despite Matthew's considerable care an attention and without any
> criticism of his work, orphan-sysvinit-scripts has a number of unsolvable
> issues[1]. It also still requires coordination: if netperf was to change
> its behaviour in the future 

IMO netperf is done ;)

So while this may be a conern in general do keep in mind what the specifics
of any given package are.


On Fri, Aug 16, 2024 at 09:58:20AM +0200, lorenzo wrote:
> > Remove the runlevels from Default-Start: in the LSB header so that
> > line is empty.
> 
> just use
> 
> override_dh_installinit:
> 	dh_installinit --no-enable --no-start

That's just like the dh_installsystemd then. Neat.

Mark, I'd prefer to go with that way of doing it if you do want to take
this up.

Thanks,
--Daniel
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://www.chiark.greenend.org.uk/pipermail/debian-init-diversity/attachments/20240817/4deea209/attachment.sig>