Forgejo Actions: run the weekly update here, same behavior as GitHub #1
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "claude/forgejo-actions"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Runs the weekly GeoNames update on Forgejo itself, so this repo no longer depends on GitHub. Behavior is the same as
.github/workflows/update_allCountries.yml: every Sunday at 00:00 UTC it downloadsallCountries.zipand publishes release<branch>-<n>, with the zip and the same notes.Merge method: Fast-forward only, one commit on current
main.Changes
.forgejo/workflows/update_allCountries.yml(new): a step-for-step port of the GitHub workflow, on the Unraid runner asruns-on: ubuntu-latest. While.forgejo/workflows/exists, Forgejo ignores.github/, and GitHub ignores.forgejo/. Each difference from the GitHub workflow is forced by Forgejo:n=run_number + 114. This repo's run counter starts at 1, but the releases copied from GitHub end atmain-114, so the first scheduled release here ismain-115or later.forgejo-release@v2.13.4creates the release, instead ofsoftprops/action-gh-release, which only talks to GitHub's API. The notes go through a file, not an expression, so their backticks never reach a shell.$FORGEJO_OUTPUT(runner v13). Artifacts useforgejo/upload-artifact@v4.dry_runinput that does everything except publish.update_allCountries.py: the releases URL comes fromRELEASES_API_URL, with the GitHub URL as the default, so the GitHub workflow is unaffected. I also removed the duplicate URL constant insideget_previous_checksum. The file keeps its CRLF line endings.Kept as-is on purpose
get_previous_checksumnever matches, because the notes wrap the MD5 in backticks and the regex does not allow for them. So every successful run publishes a release, as on GitHub. Aareon chose to keep that behavior (2026-10-09).After merge
main. The first scheduled run is Sunday 2026-10-11, 00:00 UTC.Tested
update_allCountries.pyfrom this branch, run locally on Python 3.10 withRELEASES_API_URLpointed at this repo's Forgejo API:Bearerauth was accepted) and logged the same "MD5 checksum not found" warning the GitHub runs do;update_status.txt=updateand notes in the same format.ubuntu-latestimage (catthehacker/ubuntu:act-24.04) has Python 3.12, pip, venv and node 24.actions/setup-python,actions/checkoutandforgejo/upload-artifactall resolve from data.forgejo.org.Not tested
main; the first real one is Sunday.🤖 Generated with Aki acting as Claude (Agent: Aki)
Runner tests on
claude/forgejo-actions(3c51288):dry_run=true, success.setup-pythoninstalled CPython 3.10.22, and the deps installed.update.claude/forgejo-actions-115, so the+114offset works.release-notesartifact uploaded.dry_run=false, success.forgejo-releasepublishedclaude/forgejo-actions-116with the automatic token, so no secret is needed. It was not a draft or a prerelease.allCountries.zipis 19,700,298 bytes, and the MD5 of the downloaded file (2e9a381a…) matches the one in the notes.I deleted the test release and its tag afterwards. The repo is back to 102 releases and 105 tags.
Still untested: the
scheduletrigger, which registers only frommain. The first real run is Sunday 2026-10-11 00:00 UTC, run #3, so releasemain-117.main-115andmain-116are skipped, the same way failed runs left gaps on GitHub.