What were you trying to do that didn't work?

RHEL 10 recently removed tracer, so I am looking for how to replace its functionality in Cockpit. We used tracer to determine what needs to happen after applying OS updates:

1. Nothing, just continue
2. Restart outdated systemd services
3. Show outdated processes which are not managed by systemd, and offer to reboot
4. Reboot due to a kernel (or low-level package like dbus) update
5. Not reboot if a kernel update came via a kpatch

1. and 2. are covered by dnf needs-restarting This is about 3., which the current version cannot do.

Please provide the package NVR for which bug is seen:

dnf-plugins-core-4.7.0-5.el10.noarch

How reproducible: Always

Steps to reproduce

podman.service is an example where a package update does not restart the service in the RPM scripts (because reasons, that's ok). It times out on idle, so start it and keep it busy:

systemctl start podman.socket
while curl --unix-socket /run/podman/podman.sock http://nothing/v1.12/libpod/containers/json; do sleep 1; done

Now you can follow

journalctl -f -u podman.service

to get the requests and see that it stays alive (but this isn't necessary for the reproducer)

Now update podman:

dnf update podman

If the package is already up to date, you can also do

dnf reinstall -y podman

Now check

dnf needs-restarting
dnf needs-restarting --services
dnf needs-restarting --useronly

These are all the modes that exist, aside from --reboothint which works correctly.

‐‐services is correct, it outputs "podman.service". But the default mode and ‐‐useronly both show

2489 : /usr/bin/podman --log-level=info system service

But that is not useful – this is already covered by services. So we need a mode that filters the systemd-controlled processes from that list. I take it there was a reason to show all processes in that output, and by now it's established behaviour, so it may take a new mode/filter like --no-services?