[systemd-devel] [ANNOUNCE] systemd v221

[Lennart Poettering]

On Fri, 19.06.15 16:40, Michael Biebl (mbiebl at gmail.com) wrote:

> 2015-06-19 16:38 GMT+02:00 Lennart Poettering <lennart at poettering.net>:
> > On Fri, 19.06.15 16:29, Michael Biebl (mbiebl at gmail.com) wrote:
> >
> >> > If something is not in shape we'll revert it. Regardless of the
> >> > general merits of the patch set: this one actually broke stuff, it
> >> > was incomplete. Either you make the man pages dynamic, or you ship
> >> > them pre-built. The patch set did both. That's broken, and hence has
> >> > no place in a release. And I'd much rather see that stuff removed
> >> > again than having to delay the release further.
> >> >
> >> >> Not amused, not amused at all.
> >> >
> >> > I am sorry you feel that way.
> >> >
> >> > I am not sure though what you suggest: delay releases until zero
> >> > bugfixes have been applied for a week? Well, that would mean we'd
> >> > never do releases again, sorry. We have to release some time. On
> >>
> >> Bullshit. That's been in master for a while and the justification for
> >> reverting appear to totally made up.
> >
> > I wasn't the one who reverted it or even involved in the
> > discussions. But what I saw is that the patch was borked, since it one
> > one hand tried to ship the man pages pre-built but also wanted to make
> > them different depending on ./configure runs. And that's just
> > *broken*.
> 
> The justification for the revert basically boil down to:
> let's make it as hard as possible and use systemd as a stick to force
> them to not use split-usr.

Well, to make that clear, we support split-usr for compatibility
reasons, to be nice to debian and ubuntu. But we don't like the
concept, and we'll not compromise for it. For example, ProtectSystem=
isn't really effective with split-usr either. Neither does nspawn's
--volatile= switch or anything else related to the stateless/factory
reset stuff we are working on work with it.

Also, when it comes to file search paths I am strongly of the opinion
that we never should document the paths in the root dir, since they
generally are a form of API: it's an extension point for 3rd party
packages, and we should communicate clearly where that point is and
that it is always the same. If we instead say "it can be anything,
depending on your distro", then we aren't better than sysvinit.

So yes, the split-usr thing is unloved by many of the systemd upstream
devs. We'll continue to support it though, but only the minimal level
necessary to make things work and not regress. And: the quirks it
causes we'll not necessarily document, simply because we want the man
pages clear and clean and not contain a list of compat quirks that
apply to some systems only.

----

The above is my personal opinion, and again in the specific decision
and discussion about the reverted man patches I was not involved. And
as far as I can see the communication around it wasn't very good. I am
very sorry for that, we should do this better,

Lennart

-- 
Lennart Poettering, Red Hat