On Fri, May 30, 2025 at 01:19:25PM +0200, Alexander Kanavin wrote:
> On Wed, 28 May 2025 at 21:47, Patrick Williams <patrick@stwcx.xyz> wrote:
> 
> > Walnascar is currently broken for template services and so is master.
> > My suggestion would be:
> >
> >    - Revert 7a580800db39 in walnascar.
> >    - Improve systemd_create_presets to handle template services in master.
> >    - [Maybe] fix all the services failing per Mathieu on master.
> >    - [Maybe] apply this change so that `systemd_postinst` runs
> >              as it use to.
> 
> Thanks, I appreciate you digging into the issue.

Thank you for the feedback Alex.

> That said, the plan to fix it needs to be reordered. We have a strict
> policy of fixing issues in master first, for good reason. Also,
> backporting the fixes to stable branches is much easier to get
> accepted than major reverts like that.

Agreed.  I tried to do a simple revert locally and it had conflicts
already.  It was easier to just do the needful in
systemd_create_presets.

This is done and ready for feedback:
    https://lore.kernel.org/openembedded-core/20250529174427.2731440-1-patrick@stwcx.xyz/T/#u

> Also, if something currently quietly fails in poky-altcfg (it wasn't
> entirely clear from your analysis), then the fix shouldn't just
> address the issue, it should also make any possible fail a loud and
> hard one. So it doesn't quietly regress again.

poky-altcfg does not currently fail (or fail to perform).
    - There are no template instances in any package in poky-altcfg that
      rely on the use of SYSTEMD_SERVICE.
    - The only template instance in poky-altcfg is created manually with
      symlinks (see `systemd-serialgetty.bb:do_install`).

This problem only affects template instances: 
    SYSTEMD_SERVICE = "foo@bar.service" # Template instance
    SYSTEMD_SERVICE = "foo.service" # Non-template instance

Prior to 7a580800db39, template instances listed in SYSTEMD_SERVICE
would have been enabled by the systemd_postinst hook that called [fake]
systemctl.  After 7a580800db39, the systemd_postinst hook silently
skips.

Prior to 7a580800db39, non-template instances listed in SYSTEMD_SERVICE
were enabled by systemd_create_presets files (and also duplicatively
enabled by systemd_postinst).  After 7a580800db39, non-template
instances are enabled only by the systemd_create_presets.

With my change linked above, the template instances will be enabled by
systemd_create_presets as well.  Thus all instances (template or not) will be
handled by systemd_create_presets, and the systemd_postinst hook only has
effect when installing via an RPM and not when building a rootfs.
entirely.

-- 
Patrick Williams