ci: call the central auto-merge sweep — its coverage moved first - #719
Merged
Conversation
This repo was the last of the two deliberately excluded from the fleet rollout. Its copy of auto-merge-sweep.sh shipped merges by re-arming CI so that cd.yml (workflow_run on CI) would fire — a chain with a known hole: a CI run started by GITHUB_TOKEN emits no workflow_run event, so CD never fires for an automated merge. The canonical sweep closes that with a RECONCILER, enabled here via `deploy_workflow: cd.yml`: every sweep compares main's tip against the last successful cd.yml run and dispatches it directly when they differ. The `post-main` hand-off in ci.yml stays as a second path; the reconciler is the net under both. THE TESTS MOVED BEFORE THE SCRIPT WAS DELETED. This repo was one of only two in the fleet that ever wrote tests for the sweep, and its suite covered four things the canonical suite did not — the in-progress defer, the stale-run wait, and the full red-base carve-out set (green-on-failing-jobs merges; no coverage refused; PARTIAL coverage refused). All of it is now in dotfiles/scripts/ci/test-auto-merge-sweep.sh (17 cases, mutation-proven), ported in bitbaum/dotfiles#34 so no assertion is lost between the two commits. One block of auto-merge-base-guard.test.ts was NOT about the sweep: the source assertions on refresh-e2e-reset-tokens.mjs (fixture keyed on GITHUB_RUN_ATTEMPT, leftover fixture cleared instead of failed on). That script stays in this repo, so its guard stays too — moved to e2e-reset-fixture.test.ts, its own file, because the thing it tests no longer shares context with a script this repo no longer carries. Verified before deleting the local copy: diffed it against the canonical and accounted for every differing line — all of them are main_* vs base_* naming for identical logic, the cancelled-rerun the canonical generalises, and the CI re-arm the canonical does via REARM_WORKFLOWS. The update-branch behaviour this repo's branch-currency tests pinned is in the canonical verbatim. CLAUDE.md repointed: the policy section and the two-hand-offs section both named scripts/ci/auto-merge-sweep.sh, which would have become a lie the moment this merged. Fleet duplication: 8 copies -> 7. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
catomean
added a commit
that referenced
this pull request
Aug 20, 2026
This repo was the last of the two deliberately excluded from the fleet rollout. Its copy of auto-merge-sweep.sh shipped merges by re-arming CI so that cd.yml (workflow_run on CI) would fire — a chain with a known hole: a CI run started by GITHUB_TOKEN emits no workflow_run event, so CD never fires for an automated merge. The canonical sweep closes that with a RECONCILER, enabled here via `deploy_workflow: cd.yml`: every sweep compares main's tip against the last successful cd.yml run and dispatches it directly when they differ. The `post-main` hand-off in ci.yml stays as a second path; the reconciler is the net under both. THE TESTS MOVED BEFORE THE SCRIPT WAS DELETED. This repo was one of only two in the fleet that ever wrote tests for the sweep, and its suite covered four things the canonical suite did not — the in-progress defer, the stale-run wait, and the full red-base carve-out set (green-on-failing-jobs merges; no coverage refused; PARTIAL coverage refused). All of it is now in dotfiles/scripts/ci/test-auto-merge-sweep.sh (17 cases, mutation-proven), ported in bitbaum/dotfiles#34 so no assertion is lost between the two commits. One block of auto-merge-base-guard.test.ts was NOT about the sweep: the source assertions on refresh-e2e-reset-tokens.mjs (fixture keyed on GITHUB_RUN_ATTEMPT, leftover fixture cleared instead of failed on). That script stays in this repo, so its guard stays too — moved to e2e-reset-fixture.test.ts, its own file, because the thing it tests no longer shares context with a script this repo no longer carries. Verified before deleting the local copy: diffed it against the canonical and accounted for every differing line — all of them are main_* vs base_* naming for identical logic, the cancelled-rerun the canonical generalises, and the CI re-arm the canonical does via REARM_WORKFLOWS. The update-branch behaviour this repo's branch-currency tests pinned is in the canonical verbatim. CLAUDE.md repointed: the policy section and the two-hand-offs section both named scripts/ci/auto-merge-sweep.sh, which would have become a lie the moment this merged. Fleet duplication: 8 copies -> 7. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
github-actions Bot
pushed a commit
that referenced
this pull request
Aug 20, 2026
* deps: bump tailwindcss from 3.4.18 to 4.3.3 Bumps [tailwindcss](https://github.com/tailwindlabs/tailwindcss/tree/HEAD/packages/tailwindcss) from 3.4.18 to 4.3.3. - [Release notes](https://github.com/tailwindlabs/tailwindcss/releases) - [Changelog](https://github.com/tailwindlabs/tailwindcss/blob/main/CHANGELOG.md) - [Commits](https://github.com/tailwindlabs/tailwindcss/commits/v4.3.3/packages/tailwindcss) --- updated-dependencies: - dependency-name: tailwindcss dependency-version: 4.3.3 dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <support@github.com> * deps: bump eslint from 9.39.5 to 10.8.1 (#708) * deps: bump eslint from 9.39.5 to 10.8.1 Bumps [eslint](https://github.com/eslint/eslint) from 9.39.5 to 10.8.1. - [Release notes](https://github.com/eslint/eslint/releases) - [Commits](eslint/eslint@v9.39.5...v10.8.1) --- updated-dependencies: - dependency-name: eslint dependency-version: 10.8.1 dependency-type: direct:development update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <support@github.com> * fix(lint): make eslint 10 flat config work with eslint-config-next ESLint 10 removed context.getFilename() (used by eslint-plugin-react's version-detect codepath) and dropped the implicit scopeManager.addGlobals Next's bundled babel-eslint-parser relied on for plain JS. Pin an explicit React version so detect never runs, and use plain espree for non-TS files. Also bumps eslint-plugin-unused-imports to 4.4.1 for ESLint 10 peer support. --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Mao Nakamoto <41178744+maonakamoto@users.noreply.github.com> * ci: call the central auto-merge sweep — its coverage moved first (#719) This repo was the last of the two deliberately excluded from the fleet rollout. Its copy of auto-merge-sweep.sh shipped merges by re-arming CI so that cd.yml (workflow_run on CI) would fire — a chain with a known hole: a CI run started by GITHUB_TOKEN emits no workflow_run event, so CD never fires for an automated merge. The canonical sweep closes that with a RECONCILER, enabled here via `deploy_workflow: cd.yml`: every sweep compares main's tip against the last successful cd.yml run and dispatches it directly when they differ. The `post-main` hand-off in ci.yml stays as a second path; the reconciler is the net under both. THE TESTS MOVED BEFORE THE SCRIPT WAS DELETED. This repo was one of only two in the fleet that ever wrote tests for the sweep, and its suite covered four things the canonical suite did not — the in-progress defer, the stale-run wait, and the full red-base carve-out set (green-on-failing-jobs merges; no coverage refused; PARTIAL coverage refused). All of it is now in dotfiles/scripts/ci/test-auto-merge-sweep.sh (17 cases, mutation-proven), ported in bitbaum/dotfiles#34 so no assertion is lost between the two commits. One block of auto-merge-base-guard.test.ts was NOT about the sweep: the source assertions on refresh-e2e-reset-tokens.mjs (fixture keyed on GITHUB_RUN_ATTEMPT, leftover fixture cleared instead of failed on). That script stays in this repo, so its guard stays too — moved to e2e-reset-fixture.test.ts, its own file, because the thing it tests no longer shares context with a script this repo no longer carries. Verified before deleting the local copy: diffed it against the canonical and accounted for every differing line — all of them are main_* vs base_* naming for identical logic, the cancelled-rerun the canonical generalises, and the CI re-arm the canonical does via REARM_WORKFLOWS. The update-branch behaviour this repo's branch-currency tests pinned is in the canonical verbatim. CLAUDE.md repointed: the policy section and the two-hand-offs section both named scripts/ci/auto-merge-sweep.sh, which would have become a lie the moment this merged. Fleet duplication: 8 copies -> 7. Co-authored-by: Claude Opus 5 <noreply@anthropic.com> * fix(a11y): close the white-on-accent class with a gate, not a third sweep (#720) PR #703 swept 13 files to text-on-accent, but three labels survived: the Button 'accent' cva variant in config/design-system.ts (which is exactly what renders the header's Get Started and the landing page's Meet your Cat — measured live at 3.10:1), plus two inline chips. White on --public-accent is 3.10:1 against a 4.5:1 floor; text-on-accent is 6.10:1. Second fix of the same class, so the class is now closed: a new check:accent-ink gate (mutation-tested — reverting the variant makes it exit 1) runs at the head of verify. The broader design:check stays unwired because legacy violations would turn the base red; this gate is narrow and green today. Co-authored-by: Georgy Butaev <41178744+g-but@users.noreply.github.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> * fix(build): migrate to Tailwind v4 Keeps tailwind.config.ts as the single source of truth via v4's @config bridge directive (same pattern already validated in solon) rather than a full CSS-first rewrite — that migration (OKLCH, dropping the JS config entirely) is tracked separately in this repo's own docs as future work, not something to bundle into a dependency-version bump. - postcss.config.js: tailwindcss + autoprefixer -> @tailwindcss/postcss (v4 handles vendor prefixing internally) - globals.css: @tailwind directives -> @import "tailwindcss" + @config - tailwind.config.ts: darkMode now needs a 2-element tuple; safelist no longer exists on Config (v4 moved pattern-based safelisting to @source inline() in CSS) — ported the dynamic initiative-color and gradient-opacity patterns there verbatim - globals.css: three dead custom-class-into-@apply chains (input, card, viewport-fix+safe-area-padding) that v3 silently no-op'd are now hard errors in v4 — inlined the actual declarations they meant Verified: full production build succeeds, lint clean, and visually spot-checked 3 pages (home, pricing, ecosystem) in a real browser against the built output — borders, colors, cards, badges all intact. --------- Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Mao Nakamoto <41178744+maonakamoto@users.noreply.github.com> Co-authored-by: Claude Opus 5 <noreply@anthropic.com> Co-authored-by: Georgy Butaev <41178744+g-but@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The last of the two repos deliberately excluded from the fleet rollout. Its copy of
auto-merge-sweep.shshipped merges by re-arming CI so thatcd.yml(workflow_runon CI) would fire — a chain with a known hole: a CI run started byGITHUB_TOKENemits noworkflow_runevent, so CD never fires for an automated merge.The canonical sweep closes that with a reconciler, enabled here via
deploy_workflow: cd.yml: every sweep compares main's tip against the last successfulcd.ymlrun and dispatches it directly when they differ. Thepost-mainhand-off inci.ymlstays as a second path; the reconciler is the net under both.The tests moved before the script was deleted
This repo was one of only two in the fleet that ever wrote tests for the sweep, and its suite covered four things the canonical suite did not — the in-progress defer, the stale-run wait, and the full red-base carve-out set (green-on-failing-jobs merges; no coverage refused; partial coverage refused). All of it is now central — bitbaum/dotfiles#34, 17 cases, mutation-proven — so no assertion is lost between the two commits.
One block was not about the sweep: the source assertions on
refresh-e2e-reset-tokens.mjs(fixture keyed onGITHUB_RUN_ATTEMPT"; leftover fixture cleared, not failed on). That script stays in this repo, so its guard stays too — moved toe2e-reset-fixture.test.ts`.Verified before deleting
Diffed the local copy against the canonical and accounted for every differing line:
main_*vsbase_*naming for identical logic, the cancelled-rerun the canonical generalises, and the CI re-arm the canonical does viaREARM_WORKFLOWS. The update-branch behaviour the branch-currency tests pinned is in the canonical verbatim.CLAUDE.mdrepointed — two sections named the local script and would have become lies the moment this merged.Fleet duplication: 8 copies → 7.
🤖 Generated with Claude Code