Skip to content

docs(release): correct two steps the v0.3.0 cut proved wrong - #145

Merged
ArangoGutierrez merged 1 commit into
mainfrom
docs/release-template-corrections
Aug 19, 2026
Merged

docs(release): correct two steps the v0.3.0 cut proved wrong#145
ArangoGutierrez merged 1 commit into
mainfrom
docs/release-template-corrections

Conversation

@ArangoGutierrez

Copy link
Copy Markdown
Owner

Two defects in the release checklist, both found by following it during the v0.3.0 cut rather than by reading it.

The release-count check counted drafts. A bare --jq 'length' includes drafts for anyone with write access, so it reads "one more than before" as soon as phase 3 creates the draft and before anything is public. Measured: 4 published, draft created, bare length returned 5 while an anonymous /releases/latest still returned v0.2.2. Both occurrences now filter on .draft==false.

Phase 5 told you to push to main. That branch is protected and requires a PR plus build-test, so the push is rejected with GH013. Phase 5 now says so, and records the technique that removed the red window on this cut: create the signed tag locally, drive the site commit green on a branch while nothing is public, push the tag only then.

Testing done

  • swift test rc=0, 1235 tests in 12 suites, one pre-existing known issue.
  • The corrected jq was executed against the live API and returns 5, matching the five published releases.
  • YAML re-parsed after editing: 12 body blocks, form intact.

Breaking changes

None. Issue-template text only.

Both were found by following the checklist rather than by reading it.

The release-count check used a bare `--jq 'length'`. That counts DRAFTS for
anyone with write access, so it reports "one more than before" the moment phase 3
creates the draft, before anything is public. Measured during the v0.3.0 cut: 4
published, draft created, bare `length` returned 5 while an anonymous
`/releases/latest` still returned v0.2.2. Both occurrences now filter on
`.draft==false`.

Phase 5 said to push the tag and merge the site commit back to back. `main` is
protected: `refs/heads/main` requires a pull request and the `build-test` check,
so a direct push is rejected with GH013 and the window is as long as CI takes.

Phase 5 now also records the technique that removes the window entirely. Create
the signed tag LOCALLY and do not push it. `newestReleaseTag()` reads `git tag`,
so the local tag alone turns the site guards red and the phase 5 commit can be
driven green on a branch while nothing is public. Push the tag only once that
branch is green: `main` never sees a red run, because CI runs on push and the
next push to `main` is the one carrying the fix. The v0.3.0 cut did this and
`main` went green to green.

Signed-off-by: Carlos Eduardo Arango Gutierrez <eduardoa@nvidia.com>
@ArangoGutierrez
ArangoGutierrez marked this pull request as ready for review August 19, 2026 14:38
@ArangoGutierrez
ArangoGutierrez merged commit b333f41 into main Aug 19, 2026
2 checks passed
@ArangoGutierrez
ArangoGutierrez deleted the docs/release-template-corrections branch August 19, 2026 14:43
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