snap upload rejected with "refresh-timer" interface attribute

[Kyle Nitzsche]

With the transition from snapd's refresh.schedule to refresh.timer config key [1], a customer is declaring an interface attribute accordingly in snapcraft.yaml, but the store rejects the upload.

I confirm in my test. That is, if my snapcraft.yaml has the following:
plugs:
  snapd:
    interface: snapd-control
    refresh-timer: managed

apps:
  daemon:
    daemon: simple
    command: daemon
    plugs: [snapd]

And then I push to the store I get this:
Error while processing...
The store was unable to accept this snap.
  - unknown attribute 'refresh-timer' for interface 'snapd-control' (plugs)

But, if I use the interface attribute refresh-schedule (which corresponds to the deprecated snapd config refresh.schedule key), the upload succeeds.

Please also see this snapd bug: https://bugs.launchpad.net/snapd/+bug/1771835

[1] https://snapdocs.labix.org/configuration-options/87

[Daniel Manrique (roadmr)]

That error is actually given by the click reviewer tools, which the store runs to scan every incoming snap upload.

I'll add a task for the click reviewer tools; once a version is available which is happy with the new attribute name, I can update what we have in the store so this works as intended.

[Jamie Strandboge (jdstrand)]

I think https://snapdocs.labix.org/configuration-options/87 is out of date. There is no code in snapd supporting 'refresh-timer' in snapd master, but refresh-schedule is still supported in snapd-control.

Perhaps you are thinking of the new 'refresh-mode'? Eg:

name: test-stop-and-refresh-mode
...
apps:
 endure:
   ...
   daemon: simple
   refresh-mode: endure

See https://forum.snapcraft.io/t/process-lifecycle-on-snap-refresh/140/36 for details.

[Kyle Nitzsche (knitzsche)]

The key to control refresh scheduling is now named refresh.timer whereas it used to be named refresh.schedule. However, snapd still apparently uses refresh.schedule for the case of "managed", which is a bug I filed and listed in the description.

The question is: should reviewer tools also transition to refresh-timer as the interface attribute? I suppose it should, as that other bug is fixed.

JdStrand: Are you sure this bug is invalid?