release-please finds merged release PRs by the 'autorelease: pending'
label. Gitea resolves label names with GetLabelIDsInRepoByNames, a SQL IN
filter that silently drops unknown names, so on a repository without the
labels the PR is created unlabeled and every run finds zero merged release
PRs: no tag, no release, and the next release PR instead - with a green
run each time. That is what happened to #20 and #21.
The workflow now ensures both labels exist (idempotent: list, create the
missing ones) and gained workflow_dispatch so a missed release can be
replayed.
release-please finds merged release PRs by the 'autorelease: pending'
label. Gitea resolves label names with GetLabelIDsInRepoByNames, a SQL IN
filter that silently drops unknown names, so on a repository without the
labels the PR is created unlabeled and every run finds zero merged release
PRs: no tag, no release, and the next release PR instead - with a green
run each time. That is what happened to #20 and #21.
The workflow now ensures both labels exist (idempotent: list, create the
missing ones) and gained workflow_dispatch so a missed release can be
replayed.
release-please finds merged release PRs by the 'autorelease: pending'
label. Gitea resolves label names with GetLabelIDsInRepoByNames, a SQL IN
filter that silently drops unknown names, so on a repository without the
labels the PR is created unlabeled and every run finds zero merged release
PRs: no tag, no release, and the next release PR instead - with a green
run each time. That is what happened to #20 and #21.
The workflow now ensures both labels exist (idempotent: list, create the
missing ones) and gained workflow_dispatch so a missed release can be
replayed.
[WARN] .gitea/workflows/release-please.yml:8 - workflow_dispatch lets anyone with write access run this job from any branch, and the job has PR_GITEA_TOKEN and contents: write. Before this change, the workflow only ran on pushes to main, so a modified workflow on a feature branch could not run with these secrets. Fix: remove workflow_dispatch, or gate the job with if: github.ref == 'refs/heads/main' and make sure the secrets are not exposed to other refs. If you keep dispatch, document it as a deliberate trust decision.
[WARN] .gitea/workflows/release-please.yml:19 - There is no concurrency group. Two runs close together (two quick pushes to main, or a push plus a manual dispatch) can both see a missing label and both POST it. The second POST gets a conflict, the step exits 1, and that run never reaches the release action. Fix: add concurrency: { group: release-please, cancel-in-progress: false } at the workflow level.
[WARN] .gitea/workflows/release-please.yml:33 - The existence check reads only ?limit=50 and does not paginate. Once the repo has more than 50 labels, a label that exists on a later page looks missing, the POST returns a conflict, and the step fails on every run. Fix: either paginate with &page=N until an empty page, or drop the pre-check and treat HTTP 409 (or 422 for duplicate name) as "already exists" and continue.
[NIT] .gitea/workflows/release-please.yml:39,46 - /tmp/label.json is a fixed path that can collide if runs share a host. Use "$RUNNER_TEMP/label.json", or pipe the body to stdout.
[NIT] .gitea/workflows/release-please.yml:35 - grep -qF on the raw JSON matches the string anywhere in the response, including a label description. A jq -e --arg n "$name" 'any(.[]; .name == $n)' check is exact and clearer.
The rest of the diff is documentation (OKF_Nuxt/decisions/release-engineering.md, OKF_Nuxt/log.md) and has no reviewable code beyond the workflow above.
VERDICT: APPROVE
[0m
> build · claude-haiku-5-5
[0m→ [0mRead .pr-review-diff.patch
> build · claude-haiku-5-5
[0m→ [0mRead .gitea/workflows/release-please.yml
> build · claude-haiku-5-5
**Findings**
[WARN] .gitea/workflows/release-please.yml:8 - `workflow_dispatch` lets anyone with write access run this job from any branch, and the job has `PR_GITEA_TOKEN` and `contents: write`. Before this change, the workflow only ran on pushes to `main`, so a modified workflow on a feature branch could not run with these secrets. Fix: remove `workflow_dispatch`, or gate the job with `if: github.ref == 'refs/heads/main'` and make sure the secrets are not exposed to other refs. If you keep dispatch, document it as a deliberate trust decision.
[WARN] .gitea/workflows/release-please.yml:19 - There is no `concurrency` group. Two runs close together (two quick pushes to `main`, or a push plus a manual dispatch) can both see a missing label and both POST it. The second POST gets a conflict, the step exits 1, and that run never reaches the release action. Fix: add `concurrency: { group: release-please, cancel-in-progress: false }` at the workflow level.
[WARN] .gitea/workflows/release-please.yml:33 - The existence check reads only `?limit=50` and does not paginate. Once the repo has more than 50 labels, a label that exists on a later page looks missing, the POST returns a conflict, and the step fails on every run. Fix: either paginate with `&page=N` until an empty page, or drop the pre-check and treat HTTP 409 (or 422 for duplicate name) as "already exists" and continue.
[NIT] .gitea/workflows/release-please.yml:39,46 - `/tmp/label.json` is a fixed path that can collide if runs share a host. Use `"$RUNNER_TEMP/label.json"`, or pipe the body to stdout.
[NIT] .gitea/workflows/release-please.yml:35 - `grep -qF` on the raw JSON matches the string anywhere in the response, including a label description. A `jq -e --arg n "$name" 'any(.[]; .name == $n)'` check is exact and clearer.
The rest of the diff is documentation (`OKF_Nuxt/decisions/release-engineering.md`, `OKF_Nuxt/log.md`) and has no reviewable code beyond the workflow above.
VERDICT: APPROVE
Noel
merged commit f9ddbb7e1a into main2026-10-27 11:07:46 +00:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
release-please finds merged release PRs by the 'autorelease: pending'
label. Gitea resolves label names with GetLabelIDsInRepoByNames, a SQL IN
filter that silently drops unknown names, so on a repository without the
labels the PR is created unlabeled and every run finds zero merged release
PRs: no tag, no release, and the next release PR instead - with a green
run each time. That is what happened to #20 and #21.
The workflow now ensures both labels exist (idempotent: list, create the
missing ones) and gained workflow_dispatch so a missed release can be
replayed.
[0m
[WARN] .gitea/workflows/release-please.yml:8 -
workflow_dispatchlets anyone with write access run this job from any branch, and the job hasPR_GITEA_TOKENandcontents: write. Before this change, the workflow only ran on pushes tomain, so a modified workflow on a feature branch could not run with these secrets. Fix: removeworkflow_dispatch, or gate the job withif: github.ref == 'refs/heads/main'and make sure the secrets are not exposed to other refs. If you keep dispatch, document it as a deliberate trust decision.[WARN] .gitea/workflows/release-please.yml:19 - There is no
concurrencygroup. Two runs close together (two quick pushes tomain, or a push plus a manual dispatch) can both see a missing label and both POST it. The second POST gets a conflict, the step exits 1, and that run never reaches the release action. Fix: addconcurrency: { group: release-please, cancel-in-progress: false }at the workflow level.[WARN] .gitea/workflows/release-please.yml:33 - The existence check reads only
?limit=50and does not paginate. Once the repo has more than 50 labels, a label that exists on a later page looks missing, the POST returns a conflict, and the step fails on every run. Fix: either paginate with&page=Nuntil an empty page, or drop the pre-check and treat HTTP 409 (or 422 for duplicate name) as "already exists" and continue.[NIT] .gitea/workflows/release-please.yml:39,46 -
/tmp/label.jsonis a fixed path that can collide if runs share a host. Use"$RUNNER_TEMP/label.json", or pipe the body to stdout.[NIT] .gitea/workflows/release-please.yml:35 -
grep -qFon the raw JSON matches the string anywhere in the response, including a label description. Ajq -e --arg n "$name" 'any(.[]; .name == $n)'check is exact and clearer.The rest of the diff is documentation (
OKF_Nuxt/decisions/release-engineering.md,OKF_Nuxt/log.md) and has no reviewable code beyond the workflow above.VERDICT: APPROVE