Support skopeo auth when inspecting prev-ref

[cthtrifork]

This solves #19

[antheas]

Mmmm. Can you try to pull the skopeo command outside? Then people can do podman login in their action runner and it will work. I am afraid this might cause regressions and it only applies to the github registry

[cthtrifork]

Mmmm. Can you try to pull the skopeo command outside? Then people can do podman login in their action runner and it will work. I am afraid this might cause regressions and it only applies to the github registry

Would it not be a big refactor as you are calling sudo podman run multiple times in your action.yaml ?
This is the reason i added it inside the 3_chunk.sh script but also added environment variable checks to not break existing calls.

I agreed it is limited to the github registry and I could add custom inputs to solve this.

The easiest way to solve your requirement would be probably inject the auth file instead - would that be preferable or am I missing out on something

[antheas]

The skopeo command just downloads the manifest file. If you pull it outside the podman script it will be able to access the creds of your action and will work. You can also opt to add an action parameter that allows you to specify a path for the manifest and if it exists skip the skopeo call

[cthtrifork]

I tried implementing the registry_auth_file, but I can change it again to volume mount a previous manifest file which has been pulled outside your action. The volume mount path would then be a input action

[antheas]

Actually the way you did it might be good enough. Does it work?

[cthtrifork]

Just tested it. It works!

[antheas]

Does it work if the auth file does not exist?

[cthtrifork]

I think you are right. I will test more scenarios. I forgot that volume mounts only create nonexisting directories but not nonexisting files

[cthtrifork]

@antheas bash was being unfriendly but I have tested multiple scenaries including happy path, non existing auth file etc.

[cthtrifork]

Let me know if you prefer a rebase or you are just going to squash

[cthtrifork]

@antheas What do you prefer? or do you want a different approach altogether. Right now I am using my fork but I would like to use upstream.