[OE-core] [OE-Core][PATCH v5 0/6] Fix systemd service enables/move towards "stateless" configuration

[Alex Kiernan]

On Tue, Apr 30, 2019 at 1:24 PM Paul Barker <paul at betafive.co.uk> wrote:
>
> On Tue, 30 Apr 2019, at 13:14, Alex Kiernan wrote:
> > On Tue, Apr 30, 2019 at 1:08 PM Paul Barker <paul at betafive.co.uk> wrote:
> > >
> > > On Mon, 29 Apr 2019, at 15:44, Alex Kiernan wrote:
> > > > As a result of the systemd 242 upgrade, which includes 01d2041e41f4
> > > > ("meson: stop creating enablement symlinks in /etc during
> > > > installation"), services such as systemd-networkd are no longer enabled
> > > > in images.
> > > >
> > >
> > > As an aside - is this systemd upgrade why on raspberrypi3 with master branches I recently have no console on tty1 and `/etc/systemd/system/getty.target.wants/getty at tty1.service` is missing?
> > >
> >
> > I suspect so - certainly getty@ is one of the things that's was
> > touched in the upstream commit.
> >
>
> Doing some reading now.
>
> The discussion on https://github.com/systemd/systemd/issues/12345 suggests we should be running `systemctl preset-all` or at least `systemctl enable getty at tty1` on first boot.

That's what this patchset does (`systemctl preset-all`), or rather it
builds images without /etc/machine-id which triggers systemd to do it.

> Maybe this should be handled at the BSP level so we can decide whether to enable getty at tty1 or not on a per-machine basis.
>

Once I'm done you should be able to install preset files which then
determine what gets enabled/disabled, the only real question I've got
is if we should always do this at image build time, or leave it for
first boot and let systemd do it (on read-write images at least).

> However, I can't see an easy way to resolve this at BSP level right now. I know how to add a systemd service file in a package and enable it but in this case we're looking to enable a service which is already installed. Reading through systemd.bbclass and the relevant docs doesn't give any suggestions. Is there any straightforward way to enable an already installed service?
>

Personally, if I needed it fixing right now, I'd probably add a hack
during rootfs build to put the symlinks in the right place.

-- 
Alex Kiernan