Publishing Forgejo v1.20 was the highlight of these past few weeks and also required more time than the previous major releases. The new features are the most attractive part of the announcement but most of the work went into listing and explaining the breaking changes in the release notes. They have to be as clear as possible for Forgejo admins asking themselves: does it matter to me? What should I do then? Unfortunately a few issues were missed and one of them caused a downtime when upgrading Codeberg. A robust and durable solution was developed as a followup.
Meanwhile the development quietly went on, a project to produce Debian packages entered the Forgejo contrib organization, the website was reorganized internally, new documentation chapters were created, the hardware infrastructure did not cause any trouble and the Codeberg moderation dealt with spam bots efficiently. Are these signs that Forgejo is a healthy project? Definitely. Does this mean someone willing to help would be turned down because there is nothing else to do? Absolutely not! If you are serious about creating a world where Free Software can be developed with Free Software, step in, your help is needed.
Forgejo v1.20 was published 24 July 2023 and there are reasons to be happy about the new features described in the blog post. And also by the fact that the entire release process is now based on Forgejo Actions. It is a challenging release for Forgejo admins because there are many breaking changes. A lot of attention went into ensuring a seamless experience when upgrading from any Gitea version, v1.20 included, and there has been no report of failed upgrades so far.
Codeberg was successfully upgraded to Forgejo v1.20.2-0 on 10 August 2023 after a few tests of database upgrades. In the days prior to the upgrade, Forgejo contributors and Codeberg volunteers worked together to figure out the potential issues and prepare the ground. A mail was sent to all users a few days before, warning them to re-create their personal access tokens in order to ensure their scope were not unexpectedly modified. This was the most prominent breaking change since it required all users to be aware of its impact.
Shortly after the upgrade all packages became unexpectedly unavailable and the feature was disabled while investigating. It turned out to be a regression caused by conflicting storage settings in the
app.ini file and the problem was fixed a few hours later. The lesson was learned and the Forgejo v1.20.3-0 release will include a safeguard against that particular issue. Other bugs (with less impact) were discovered and some of them were fixed right away (profile rendering, auth icons ratio, dropzone filename hidden, …).
In addition to the development areas discussed below, it is worth mentioning there has been an unprecedented amount of pull requests proposed and merged to fix bugs or implement features such as OpenStreetMap links in profiles, banning dots in usernames or unifying project views. A number of these changes were contributed back to Forgejo dependencies to keep the Forgejo codebase size to a minimum.
A pull request started a few months ago to allow for setting the update times of issues and comments via the API is now two tests away from being merged. It matters to federation because mirroring an issue from one forge to another so it can be displayed requires the dates are preserved which is not currently possible.
The F3 driver for Forgejo now relies on the gof3 CLI instead of implementing its own. It was made possible by a refactor of gof3 that uses urfave/v2 to be compatible with Forgejo. This kind of factorization reduces the redundant work required to maintain independent projects and keeps their combined codebases DRY.
forgejo-cli f3 mirror subcommand will do what it suggests: mirrors projects, issues, pull requests, etc. from one forge to the other. The implementation is not complete but made progress together with tests.
Around 6 August work started to deal with a difficult problem regarding storage settings (see the final draft blog post if you are reading this before it is published). There are multiple cascading issues:
- The bug is found in all Forgejo releases
- Upgrading to a release fixing the bug requires manual intervention
- Implementing a safeguard has to be done before the database upgrade and after loading the settings but there is no logic to do that
- The safeguard has to prevent downgrading from a point release but the information about the previously running Forgejo instance is not available
- Testing the safeguard can only be done reliably with automated upgrade tests which do not exist
A solution for those issues was implemented and backported to v1.20. It was almost done when the Codeberg upgrade to v1.20 happened on 10 August and discovered one manifestation of the bug that was overlooked (conflicting
[storage.packages] sections). It was an unfortunate oversight but had the benefit of improving the fix to be released in v1.20.3-0.
Codeberg hit the issue shortly after extensive manual upgrade testing was conducted. All of which had to be repeated after fixing the newly found issue. It was a great example of why manual testing is generally a bad idea and eventually more time consuming than writing automated tests. Instead of running the upgrade tests manually again, automated upgrade tests were implemented. Although they launch multiple versions of Forgejo a dozen times, they run under five minutes.
The Forgejo semantic version was used when implementing the storage setting sanity checks. It is not yet used for releases or public facing Forgejo version numbering. But it can be used internally.
When a user is blocked and is also a collaborator on a repository that the blocker owns, that collaboration is removed.
It was proposed to publish Forgejo development versions on a weekly basis. Although this was done as a byproduct of the weekly Forgejo rebase in August, the discussion is still ongoing. They are only for experimentation and could be used to run https://next.forgejo.org which is meant to help figure out if a bug is happening in the most recent development branch.
The forgejo-contrib organization is where projects related to Forgejo can find a home while they are being developed or when the long term maintenance is uncertain.
A pull request was proposed to the Helm chart for forgejo to deploy the Forgejo runner alongside Forgejo. The offline registration implemented in Forgejo and the runner were designed to help that particular use case.
The website was restructured to use Astro content collections to better separate content from the source and improve maintainability over the ad-hoc system previously in use. There also were smaller changes such as custom 404 pages.
The Forgejo development branch is rebased weekly and that worked out nicely for almost a year. The same was done until last month for stable branches. It also worked well but had one confusing side effect: the tags of the stable versions could not be found in the stable version branch. For instance the v1.93.3-0 tag is not on a commit found in the forgejo/v1.19 branch. Although it did not seem to create any actual problems, it is unusual and cause for confusion.
Early July a proposal was made for a different workflow to address this issue. It has been adopted when the first v1.20 release was published and the v1.20.1-0 and v1.20.2-0 are now both found in the forgejo/v1.20 branch.
There has been no issue with any of the hardware in the Forgejo infrastructure. It is worth mentioning resources used daily for the CI and the releases did not cause trouble.
This kind of stability is taken for granted when relying on cloud providers but it does not happen magically. Their staff is hard at work to maintain a stable environment over time. Forgejo chose to rely on Free Software only and runs its own hardware. Part of the infrastructure is using an OpenStack provider, similar to AWS but based on a Free Software stack and API. The rest is running on dedicated LXC hypervisors.
Being independent from cloud providers running proprietary software means more work for Forgejo community members. But since Free Software is one of the core values of Forgejo, it is also a requirement. The promise is to deliver a Free Software codebase but also to ensure it can be developed and released in a Free Software environment. The big question is: can the Forgejo devops team durably maintain a reliable and secure infrastructure? It is too early to tell but over four weeks of uninterrupted and stable service is a very positive sign.
Wikidata acts as central storage for the structured data of its Wikimedia sister projects including Wikipedia, Wikivoyage, Wiktionary, Wikisource, and others. Forgejo has an entry which was thoroughly updated. The Forgejo releases are also listed and updated by Wikidata editors when they are published.
The NLnet grant is progressing and some tasks related to the release process were completed. A “Requests for Payment” was approved and another is pending. The relationship between NLnet and Forgejo is not a contract in the traditional sense and a “Request for Payment” is the equivalent of an invoice, only it is payed to the beneficiary as a charitable donation. In this case the payment will go to Codeberg e.V.
Time is as valuable as funding and Forgejo durability heavily depends on how much time volunteers are willing to devote to moderation, sorting bugs, fixing them etc. It is, in fact, arguably the most precious resource any Free Software project has. The time spent by one volunteer was added to the sustainability repository to account for that. It is not a requirement and each volunteer decides whether or not they want their time to be accounted for in this way.
Two spam bots posted a dozen messages that have been sent via email to people watching the Forgejo repositories. The bots were removed by the Codeberg moderation team before the incident was even noticed by the Forgejo moderation team.
A moderation report was published and reminds Forgejo community members that “nobody can, under any circumstances, unilaterally decide to reveal private information in Forgejo spaces. It does not only go against one of the core values of Forgejo, it also goes against the most basic expectation for privacy of every person entering Forgejo spaces”.
Forgejo is a community of people who contribute in an inclusive environment. We forge on an equal footing, by reporting a bug, voicing an idea in the chatroom or implementing a new feature. The following list of contributors is meant to reflect this diversity and acknowledge all contributions since the last monthly report was published. If you are missing, please ask for an update.
A minority of Forgejo contributors earn a living by implementing the roadmap co-created by the Forgejo community, see the sustainability repository for the details.