In theory this should be quite easy to implement.

• Add new argument to koji.daemon.SCM.assert_allowed() → scratch=False
• Add new option allowed_scms_scratch_extra which would be list of SCMs that would be allowed for scratch builds (they will implicitly inherit allowed_scms)
• Pass scratch parameter from the build tasks where SCM.assert_allowed() is called

To keep compatibility with older hubs, wrap in extra try / except calls to assert_allowed()…

@tkopecek does this make sense? if so, I can work on a PR for this.

Yep, I think it is ok - @mikem ?

or, one or more new flag(s) in allowed_scms entry? like *:scratch_only:* or *:regular:scratch:*

@julian I'd like to avoid complicating parsing of those values even more… which is already over-complicated. Plus, it would be even more confusing to work with it.

or, one or more new flag(s) in allowed_scms entry? like :scratch_only: or :regular:scratch:

Agree with Igor here. This is making an already tedious configuration even worse. It might be time to provide a better way to configure the allowed scms.

I am ok with the overall intent here. However, once we start down the road of different scm rules for different scenarios, I wonder how complex we might need to get. Do we need to consider making this a policy?

Do we need to consider making this a policy?

I'm curious if allowed_scms is used per build on some environment. IMO, a policy section supporting channel test is sufficient.

Maybe even simple policy could work here similar to e.g. build_from_scm

build = 
    bool scratch :: allow
    source scratch_only_scm :: deny
    all :: allow

A candidate PR: #2951

This almost does the same as assert_allowed, maybe we could split the setting of use_common and source_cmd out.

Commit 6d4abedb fixes this issue

Commit 47c4b5d7 fixes this issue

Thanks you for implementing this! :wink:

Wow, this is awesome!

This issue has been migrated to Fedora Forge:
https://forge.fedoraproject.org/koji/koji/issues/2757

Please continue any further discussion there.