Release pipeline: fork can't cut releases on forgejo.philflow.me (release.yml targets codeberg, no runner) #11
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
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?
Context
The fork has no working release path on its own Forgejo:
.forgejo/workflows/release.ymlis the upstream goern workflow —runs-on: codeberg-smalland hardcodedcodeberg.org/api/v1/...URLs.origin(forgejo.philflow.me) — releases were never cut here..tekton/).Current interim build path (used for the v2.23.0 fixes, #6/#7/#8)
Built reproducibly in a container with no local Go and no runner:
then atomic-installed to
~/.local/bin/forgejo-mcp(rename, not in-place — running binary is ETXTBSY) and pinned~/.local/bin/forgejo-mcp.head=git rev-parse HEAD. This works but isn't a real release.What's needed for a proper Fork-Release-CI on philflow.me
release.yml:runs-on→ an existing Philflow runner label; replacecodeberg.org/api/v1/...in the.mcpbstep with the fork host; set.goreleaser.ymlgitea_urlsto forgejo.philflow.me.RELEASE_TOKENsecret.v2.23.0→ goreleaser buildsforgejo-mcp_2.23.0_linux_amd64.tar.gzas a release asset; updateflow.raven-deployment/scripts/setup-raven-dev-mcp-toolchain.sh(and a workstation equivalent) to pull from the fork release instead of codeberg.forgejo-mcp.headfromgit rev-parse HEAD(the pin was stale since 2026-05-22 and nearly misled a provenance analysis).Also
ADR-KV-018 says "~95 tools" — after the slim (commit
633927a, 102→73) the real surface is 73. Update the ADR.Found by the forgejo-mcp tool-call forensic audit, 2026-06-17.