On Mon, May 12, 2025 at 02:19:14PM +0200, Alexander Kanavin wrote:
> On Mon, 12 May 2025 at 14:12, Martin Siegumfeldt via
> lists.openembedded.org <mns=gomspace.com@lists.openembedded.org>
> wrote:
> 
> > As mentioned, it appears to be a regression during the styhead -> walnascar development. I notice that the systemctl-native is now built from the systemd sources - is this correct? Should I consequently report this through systemd?
> 
> Yes, we've transitioned to using systemd's implementation rather than our own.
> 

I'm running into this same problem and I think there is a serious flaw
with walnascar / master.  We are using IPKs and what is happening is
that the `postinst` is not running correctly.  The `systemd_postinst` is
where all of the unit files are enabled (based on the SYSTEMD_SERVICE
variable).  If this doesn't run correctly, the image is missing all of
the default enabled services and templates.

The `systemd_postinst` has a guard "if systemctl >/dev/null 2>/dev/null"
and if this fails, none of the unit file enable code is ran.  The thing
is, `systemctl` will always fail because it by default is the same thing
as `systemctl list-units`, which needs to access the systemd bus
connection.

I modified systemd.bbclass as follows:

```
 systemd_postinst() {
+systemctl || echo "It failed for some reason"
 if systemctl >/dev/null 2>/dev/null; then
        OPTS=""
```

After this, my `log.do_rootfs` is full of:
```
Failed to connect to system scope bus via local transport: Operation not permitted (consider using --machine=<user>@.host --user to connect to bus of other user)
It failed for some reason
```

I think there needs to be some change in systemd_postinst to run
something other than the equivalent of `systemctl list-units` to
determine if the executable is valid.


-- 
Patrick Williams