Skip to content

Run CI on the release branch, because the release pull requests had no checks at all - #50

Merged
omercelikdev merged 1 commit into
mainfrom
release-branch-ci
Sep 8, 2026
Merged

Run CI on the release branch, because the release pull requests had no checks at all#50
omercelikdev merged 1 commit into
mainfrom
release-branch-ci

Conversation

@omercelikdev

Copy link
Copy Markdown
Contributor

changeset-release/main has failed CI twice, and chasing it found something worse than a broken build.

Run 33958092041 has zero jobs and concludes failure. And pull request #48 — the one that shipped 0.6.0 — merged on 2026-09-05 with no checks reported whatsoever. So did #46 and #44 before it.

The cause is deliberate GitHub behaviour

The changesets bot opens the version pull request with GITHUB_TOKEN, and GitHub does not trigger pull_request workflows for a token-authored pull request — the guard that stops workflows triggering themselves recursively.

The usual fix is a personal access token or a GitHub App, which means a stored long-lived secret on a public repository. Running CI on the branch costs nothing and restores the coverage: the version commit is pushed to changeset-release/main, so a push trigger on that branch pattern sees it.

G1 stays pull-request-only, and that is correct

The release branch consumes the changeset files. A gate demanding one would fail on the very branch whose job is to remove them — so leaving G1 scoped to pull requests is the right call, not an oversight.

Why it is worth a pull request

Low stakes in content: the version pull request carries a version bump and a changelog, nothing else. High stakes in habit — a merge with no checks reported is exactly the state that teaches a team to stop reading them.

🤖 Generated with Claude Code

…o checks at all

The changeset-release/main branch has failed CI twice, and chasing it found something worse
than a broken build: run 33958092041 has ZERO jobs and concludes failure, and pull request
#48 — the one that shipped 0.6.0 — merged on 2026-09-05 with no checks reported whatsoever.
So did #46 and #44 before it.

The cause is deliberate GitHub behaviour, not a misconfiguration here: the changesets bot
opens the version pull request with GITHUB_TOKEN, and GitHub does not trigger pull_request
workflows for a token-authored pull request, to stop workflows from recursively triggering
themselves.

The usual fix is a personal access token or a GitHub App, which means a stored long-lived
secret on a public repository. Running CI on the BRANCH costs nothing and restores the
coverage: the version commit is pushed to changeset-release/main, so a push trigger on that
branch pattern sees it.

G1 stays pull-request-only and that is correct rather than an oversight — the release branch
CONSUMES the changeset files, so a gate demanding one would fail on the very branch whose job
is to remove them.

Low stakes in content, since the version pull request only carries a version bump and a
changelog. High stakes in habit: a merge with no checks reported is exactly the state that
teaches a team to stop reading them.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@omercelikdev
omercelikdev merged commit 3fea7c5 into main Sep 8, 2026
2 checks passed
@omercelikdev
omercelikdev deleted the release-branch-ci branch September 8, 2026 22:55
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant