Consensus
Mon 29 Dec 2025 2:17AM

Switching from GitLab to Codeberg

AB Alyaza Birze Public Seen by 14

CE

Chaia Eran
<span class="translation_missing" title="translation missing: en.poll_proposal_options.minor objection">Minor Objection</span>
Sun 28 Dec 2025 7:40PM

URL churn seems like a pain in the ass, and if we do switch, it'd probably be best to go straight to self-hosted Forgejo, but these are minor quibbles at best.

K

Katja
<span class="translation_missing" title="translation missing: en.poll_proposal_options.no objection">No Objection</span>
Sun 28 Dec 2025 7:40PM

I don't have the technical knowledge needed to have strong opinions on this, but given the rationale expressed here, it sounds like a net positive to take this step!

K

kouhai Wed 31 Dec 2025 8:22AM

as the person who stood up some of the original infra in a hurry, and has used burnout recovery time to set up gitea, forgejo, woodpecker (multiple times!) and also used github/gitea/forgejo actions i don't dislike the idea. but there's a couple missing details.

questions:
will the ci be self hosted or rely on codeberg's limited resources?
has the forgejo permissions system been tested? are the specific permissions expressible?

W

WholeWheatBagels Tue 13 Jan 2026 5:04AM

@kouhai The current idea is to use their CI as we have very minimal CI requirements, and our upstream projects are also using their CI and we can reuse their config with that.

I have not personally tested their perm system but it is similar to our current on GL and restricts admin actions on repos, so we can keep those permissions to infra keyholders and such. It has better permissions for branch protection that will let us protect e.g. the allowlist automation more completely.