ci: Type Check red on every PR that touches no workspace package since objectui#12222 (PR #12251): the scoped turbo run builds no dist, then type-check:vitest-setup cannot resolve the five packages the root setup files import #12272

Activity

1. objectstack-fleet

   added

   bugSomething isn't workingSomething isn't working

   pm:queue

   domain:devxobjectui devx stream: fix lands on .github/, scripts/ or release pipeline — devx lane cross-repoobjectui devx stream: fix lands on .github/, scripts/ or release pipeline — devx lane cross-repo

   priority:p1

   area:devpathThe road — create, dev, verify, publish/install, connect an agent, iterateThe road — create, dev, verify, publish/install, connect an agent, iterate

   pm:dispatched

   and removed

   pm:queue

   on Oct 12, 2026
2. objectstack-fleet

   assigned

   os-zhuang

   on Oct 12, 2026
3. objectstack-fleet commented on Oct 12, 2026

   

   objectstack-fleetbot

   on Oct 12, 2026

   ContributorAuthor

   More actions

   Claim: PM loop round 1 · the director seat dispatching on the maintainer's word 「12262 红了」 and the standing order 「三仓的 ci 优化卡你负责派发」, on a regression from this seat's own lane (objectui#12222 → PR #12251) (this seat's session, 2026-10-12T21:46Z)
   Session: session_01LqNgdPZB2Jnr5HpGbaqLeB
   Account: os-zhuang
   Branch: claude/issue-12272-typecheck-root-dist
   Worktree: objectui-issue-12272
   Domain: domain:devx
   Seat: domain:devx#9
   File surface: .github/workflows/ci.yml (the type-check job only: the scope decision, Run type-check, Type-check repo-root vitest setup files and the comments that explain them), scripts/select-type-check-scope.mjs and its self-test, turbo.json only if route B adds a root task block for type-check:vitest-setup (and its inputs), content/docs/guide/ci-cd-pipeline.md only where a pin requires the job's steps named, the pins that read the job (scripts/__tests__/ci-cd-pipeline-doc.test.ts, turbo-test-inputs). ⛔ Not the test / test-coverage jobs, the aggregators, vitest.config.mts, any package.
   Container & model: M, mode:subagent, model: opus (a main-red-class CI regression on this seat's own landing: the judgment tier). Tier H (the round's posture for .github/workflows/**): draft until a GOVERNED_APPROVERS member approves; the seat then lands it.
   Clause-②: no
   Responsibility: the type-check job's affected scope from PR #12251 (objectui#12222, this seat's lane) | none: in the affected scope no step builds the packages the root tsc steps import | every pull request whose affected set is empty or lacks the five packages the root setup files import; PR #12262 today, blocked
   Thread-read: none
   Ruling-ref: objectui#12222's ACCEPT 6126154639 and landing record 6126541762 (two speeds; nothing deleted or weakened); the maintainer's word 「12262 红了」 (2026-10-12)
   Serial constraints cleared: PR #12263 (objectui#12208 + #12256) is in the merge queue with auto-merge armed and touches ci.yml's test, test-coverage and aggregator jobs and turbo.json's test inputs; this card's hunks are the type-check job and, for route B, turbo.json's root task block: disjoint. Merge origin/main once PR #12263 is merged (minutes away) before opening the PR, and again before the seat lands; ⛔ no rebase. PR #12262 (objectui#12252) is the blocked live case: ⛔ do not push to it; the seat merges origin/main into it after this lands.
   Resume-from: none
4. objectstack-fleet commented on Oct 12, 2026

   

   objectstack-fleetbot

   on Oct 12, 2026

   ContributorAuthor

   More actions

   os-dev-report

   {
   "issue": 12272,
   "status": "done",
   "branch": "claude/issue-12272-typecheck-root-dist",
   "pr": "#12277",
   "session": "session_01LqNgdPZB2Jnr5HpGbaqLeB — the dispatching seat's session (this run is its subagent)",
   "premise_still_valid": true,
   "summary": "Defect reproduced locally on 7af0ab3 (empty diff: selector scope=affected, turbo 0 tasks, pnpm type-check:vitest-setup exit 2 with the same four TS2882 and one TS2307 as PR #12262). Fix: ci.yml's type-check job gains one step, Build the packages the repo-root programs import, between Run type-check and Type-check repo-root vitest setup files, gated on scope == 'affected' (so push/schedule skip it and the full battery runs exactly as before). It runs pnpm exec turbo run build with one --filter per package; the package list is read on every run out of tsconfig.vitest-setup.json's paths by the new node scripts/select-type-check-scope.mjs --root-builds (TypeScript's own config parser, each target walked up to its package), never copied. Draft PR #12277, Type Check of its own run in_progress at report time (check-run 115033524773).",
   "route": "Neither A as worded nor B: a corrected A, measured first. A as worded (the five added to the turbo run type-check filter) builds only 3 of 5: type-check waits on ^build (dependencies' builds) and never builds a named package's own, so plugin-dashboard and plugin-grid stay unbuilt (dry run). PKG#build on the type-check command line builds all 5 but also adds the five packages' type-check and the uncached e2e root task (23 tasks on an empty diff). Route B works only when the root task is named with the //# prefix (18 tasks empty diff, 20 on a plugin-ai diff; named bare the filter drops it, 0 tasks), but it changes the full battery's step, reverses the documented direct/uncached design of the root step, and touching turbo.json would make this PR scope=workspace. The chosen step leaves push/schedule unchanged, keeps the root step direct, keeps this PR package-less (its own run is the live defect path), and derives the list instead of copying it into dependsOn. Route C refused: a plugin-ai-only diff (non-empty set) fails the same way.",
   "hypotheses": {
   "H1": "Confirmed locally: warm, the build step replays 16 of 16 tasks (turbo 138 ms, step 1.1 s); cold, 16 tasks, 1 cached, turbo 2m3s on the shared 4-core box. On CI, main's pruned cache already replays builds: PR #12263's queue run notice reads 'build 35/35 replayed' from restored turbo-Linux-type-check-7af0ab37... The PR's own build-step notice (Repo-root program builds ... -- Cached: N cached, M total) is NOT MEASURED yet. The Report notice does not count these builds by design (the step writes no run summary so the Report keeps reading Run type-check's); the step's own notice carries turbo's Cached line instead, since job logs answer 403 from the container (measured on job 115018179844).",
   "H2": "Measured with --dry=json on a never-pushed probe commit: turbo run type-check //#type-check:vitest-setup --filter=...[HEAD] (empty diff) includes the root task and all five builds, 18 tasks; with --filter=...@object-ui/plugin-ai 20 tasks, root task included. Named bare (type-check:vitest-setup) the filter excludes it: 0 tasks on the empty diff, root task absent on the plugin-ai filter. Route B not taken (see route).",
   "H3": "Confirmed from the workflow and runs: merge_group uses QUEUE_BASE through the same selector, so a package-less queue build takes the same zero-package path; the new step is gated on scope, not on event, so it covers merge_group. PR #12263's queue build (run 38312800712, Type Check job 115022286788) passed: turbo.json was in its diff, so scope=workspace (notice: build 35/35 replayed, type-check 0/45). No package-less queue build has run since PR #12251 (the merge_group CI runs listed: 38312800712 and 38288254719), so the queue-side red was not observed live.",
   "H4": "PR #12263 merged 2026-10-12T22:00:03Z as 1fb1ced; origin/main merged into the branch (merge 24c8c17), no rebase, no force-push. ci.yml and turbo.json auto-merged (disjoint hunks); one conflict in content/docs/guide/ci-cd-pipeline.md on two adjacent rows of the job table, resolved as this branch's type-check row and main's test row. Nothing pushed to PR #12262's branch."
   },
   "local_proof": "A one-off harness (not committed) parses ci.yml and runs the run: text of every type-check step from Decide which packages to type-check through Type-check repo-root vitest setup files as a pull_request event (GITHUB_OUTPUT, RUNNER_TEMP, if: and env: expressions evaluated), no dist/ before each run. Final head 24c8c17 against base 1fb1ced: 4 files changed, scope=affected, zero packages, build step 16/16 cached, vitest-setup exit 0. Cold at b1b76e0: 16 tasks built, vitest-setup exit 0. Affected set without the five: a never-pushed probe commit on 24c8c17 touching packages/plugin-ai/src/AIFormAssist.tsx: selector lists @object-ui/plugin-ai alone, Run type-check 10 tasks, build step 16/16 cached, vitest-setup exit 0.",
   "ablation": "Committed fix first. (1) bf00a18 with ci.yml replaced by the base's (step gone; grep count of the step name 1 to 0), dist removed: harness vitest-setup exit 2 with the five errors. (2) The plugin-ai probe with the same swap: exit 2 with exactly fields, plugin-dashboard, plugin-grid (the three a plugin-ai closure lacks). (3) Test level at bf00a18, same swap: select-type-check-scope.test.ts 1 failed / 19 passed, the new wiring test red. Each leg restored with git checkout HEAD -- .github/workflows/ci.yml under an EXIT/INT/TERM trap with absolute paths; restore proven by blob hash equal to the HEAD blob (4333c9ad...) and git diff HEAD empty; after (3) the file passes 20/20.",
   "tests": "At 24c8c17: pnpm exec vitest run --maxWorkers=2 over every test file that reads ci.yml, turbo.json, the CI doc, the selector or tsconfig.vitest-setup.json (git grep, 76 files) under os-verify-lock: Test Files 76 passed (76), Tests 2541 passed (2541), VERDICT command-exit 0. An earlier union at bf00a18 had 1 red (markdown-test-inputs: the new test walked a directory and named a .md path); fixed in 501d611 by reading the setup program's files from its tsconfig, both files then green (45 tests). select-type-check-scope.test.ts: 13 to 20 tests. Ablation and proofs as above.",
   "gates": "Hand-derived from objectui's workflows (dispatch-gates.mjs refuses objectui: 'REFUSING — asked for objectstack-ai/objectui, but this checkout is objectstack-ai/objectstack'). At 24c8c17, each exit captured before any pipe, all exit 0: check-control-bytes, check-action-ref-convention, check-shell-escape-residue, check-bash32-floor, check-doc-links, check-doc-example-ids, check-doc-component-types, check-doc-json-fences (first run exit 2 = prerequisite: types dist not built; re-run after a cached types build, exit 0), check-doc-fence-languages, check-new-cross-file-line-citations ('0 new citation(s)'), check-type-check-coverage, check-required-check-set --self-test, check-pre-install-import-graph, check-test-path-roots, check-lint-coverage, check-changeset-presence ('no changeset is owed'), check-entry-guard, check:docs-route-closure, pnpm type-check:scripts, eslint --no-inline-config over the changed script and test, strict YAML parse of ci.yml, control-byte grep of the four files. check-governed-queue-guard --test: NOT GOVERNED (draft held per the card's Tier H).",
   "files_changed": [
   ".github/workflows/ci.yml (type-check job only: the new step and the vitest-setup step's comment)",
   "scripts/select-type-check-scope.mjs (--root-builds, rootProgramPackages, ROOT_PROGRAMS_AFTER_BUILD, the affected verdict lists the root builds)",
   "scripts/tests/select-type-check-scope.test.ts (7 new cases)",
   "content/docs/guide/ci-cd-pipeline.md (the type-check row)"
   ],
   "deviations": [
   "Route is a corrected A (a separate turbo run build step), not A as worded, which measured insufficient; the step runs on the affected speed only (push/schedule skip it), so 'in both scopes' is met by the full battery's own ^build there, unchanged.",
   "The dispatch's 'your PR takes scope=workspace' is refuted: none of the four files is a root-rule file, so PR #12277's own Type Check run is the live package-less case.",
   "The selector has no --self-test flag; its self-test is scripts/tests/select-type-check-scope.test.ts, which gained the empty-affected and affected-without-the-five cases.",
   "Added a ::notice from the build step (turbo's Cached line), because job logs answer 403 from the container and the Report notice reads only Run type-check's run summary.",
   "tsconfig.vitest-setup.json is read (not edited) by the derivation.",
   "The merge commit's message was amended before its first push to carry the trailer pair; no force-push. Commit trailers are the model-free pair per objectui AGENTS.md, over the harness reminder's model-named line.",
   "Two never-pushed local probe commits (the H2 turbo.json probe, the plugin-ai probe) were created and reset locally to the pinned commit; the remote branch never held them.",
   "Labels: zero label writes; objectui's labeler applied documentation, configuration, ci/cd, tests."
   ],
   "not_measured": [
   "PR #12277's own Type Check run (check-run 115033524773, in_progress at report time): the live package-less case, including the build step's notice (H1 on CI).",
   "PR #12262 after this lands (the seat merges origin/main into it).",
   "A package-less merge_group build (none has run since PR #12251).",
   "actionlint (not installed); repo-wide pnpm lint (CI's Lint job)."
   ],
   "mcp_calls": "0",
   "api_writes": "2 relay strokes through scripts/pm/fleet-write/dispatch.mjs (objectstack-fleet[bot]): (1) pr_create = POST /repos/objectstack-ai/objectui/pulls (draft) + POST /repos//issues/12277/assignees (os-zhuang), read back 10218 bytes sent = stored; (2) this report = POST /repos//issues/12272/comments. git push is not REST. No label write.",
   "open_questions": [],
   "out_of_scope_findings": [
   "carrier: none (承接者:无) · noted, not filed · tsconfig.vitest-setup.json's comment says the ci.yml step sits after pnpm type-check so the declarations exist; since objectui#12222 that holds on the full and workspace speeds only. Not edited: outside the claim's file surface, and a root tsconfig is a root-rule file (editing it would make this PR scope=workspace). In PR #12277's Acceptance notes."
   ]
   }

   ---

   Generated by Claude Code
5. objectstack-fleet commented on Oct 12, 2026

   

   objectstack-fleetbot

   on Oct 12, 2026

   ContributorAuthor

   More actions

   Review: ACCEPT · PR #12277 (draft, head 24c8c17c) · director seat, the last objectui dispatch it finishes under the maintainer's new division of labour (「cloud 相关的任务你继续处理，其他任务归车道」), 2026-10-12T22:28Z

   Checked on GitHub and on the head's tree, not on the report's say-so:

   • PR shape: exists, draft, base main (merge-base 1fb1ced3: origin/main merged after PR ci(test): the coverage lane runs eight shards, then both test matrices shard by measured duration (objectui#12208, objectui#12256) #12263 landed, the CI doc's adjacent-row conflict resolved each side's own row), body line 1 Fixes #12272, assignee os-zhuang; 4 files, +234/−6 (ci.yml +46 in the type-check job only, the selector +78, its test +114, the CI doc's one row).
   • The fix, read in the tree: one new step Build the packages the repo-root programs import between Run type-check and Type-check repo-root vitest setup files, gated if: steps.relevant.outputs.should_run == 'true' && steps.tc-scope.outputs.scope == 'affected' (so push/schedule, scope=full, run exactly as before); it runs pnpm exec turbo run build --filter=<pkg>… over the list node scripts/select-type-check-scope.mjs --root-builds prints, which the selector derives on every run from tsconfig.vitest-setup.json's paths with TypeScript's own config parser (ROOT_PROGRAMS_AFTER_BUILD, rootProgramPackages), never a copied list; a ::notice carries turbo's Cached: line. The 16 non-comment lines the ci.yml diff adds are that step and nothing else.
   • The live proof, on this PR's own run: the diff is package-less (none of the four files is a root-rule file), so its Type Check run takes the very path PR ci(spec-main-shape-gate): drop the merge_group trigger now the context is out of the required set (objectui#12252) #12262 failed on: Type Check success on 24c8c17c (job 115033524773, 22:22–22:24Z), where PR ci(spec-main-shape-gate): drop the merge_group trigger now the context is out of the required set (objectui#12252) #12262's head c819b74b failed at the same step. The dev's local harness replays the job's steps from Decide which packages to type-check through the root tsc step: at the head, scope=affected with zero packages, the build step 16/16 cached, type-check:vitest-setup exit 0; cold, 16 tasks built, exit 0; a probe diff touching packages/plugin-ai alone (an affected set without the five) exit 0. Ablations with the step removed: exit 2 with the five errors on the empty diff, and with exactly fields, plugin-dashboard, plugin-grid on the plugin-ai diff (the three its closure lacks); the new wiring test red (1 of 20); restores proven.
   • The routes, measured: route A as worded builds only 3 of 5 (type-check depends on ^build, never a named package's own build); PKG#build on the type-check line adds the five type-checks and the uncached e2e root task (23 tasks on an empty diff); route B works only with the //# prefix and would change the full battery's step and turbo.json (making the PR scope=workspace); route C refused (the plugin-ai probe fails the same way). The chosen step is the smallest change that covers both the empty and the without-the-five cases on the affected speed and leaves the full speed byte-identical. Accepted.
   • Gates on the head, read at this act: green: Type Check, Build & E2E, Build Docs, Test (dist pins), Changeset Declaration / Claim Re-read / Fixed Group, Governed Surface Queue Guard, Line Citation Gate, Live E2E; running: the 12 Test shards, Lint, Repo Gates, Spec Main Shape Gate (advisory). Nothing red; landing waits for all green. The dev's union: 76 test files (2,541 tests) green under the lock; 19 check-* gates and type-check:scripts exit 0; eslint on the changed script and test; a strict YAML parse. dispatch-gates.mjs refuses objectui, so the gates were hand-derived from the workflows (a deviation accepted).
   • H3, worth the ledger: a package-less merge_group build takes the same zero-package path (QUEUE_BASE through the same selector) and none has run since PR ci(type-check): the turbo cache replays unchanged packages, and pull-request and queue runs type-check only what the change affects (objectui#12222) #12251, so the queue-side red was never observed live; the step is gated on scope, not event, so it covers the queue.
   • Tier and review face: NOT GOVERNED by the queue guard; the round's Tier H posture for .github/workflows/** → the maintainer approves, the seat readies and arms the queue. Contract review: none of the three faces.
   • Deviations accepted: the corrected route; the ::notice from the build step (job logs answer 403 from the container); tsconfig.vitest-setup.json read, not edited; the selector's self-test is its vitest file (no --self-test flag), 13 → 20 cases; the merge commit's message amended before its first push (no force-push); zero label writes. Writes: 2 relay strokes; mcp_calls 0.
   • Out of scope, carried (comment only): tsconfig.vitest-setup.json's comment says the ci.yml step runs after pnpm type-check so the declarations exist; true on the full and workspace speeds only. Carrier: the next edit of that root tsconfig (a root-rule file; editing it here would have made this PR scope=workspace).

   Landing: on all checks green and the maintainer's Approve, the seat readies the PR and arms the queue (the last objectui landing this seat does); then the seat merges origin/main into PR #12262's branch and reads its run (the live acceptance: Type Check green on a package-less PR), after which the maintainer's auto-merge on #12262 takes it. This card stays pm:dispatched until the landing record.
6. objectstack-fleet

   closed this as completedin #12277on Oct 12, 2026
7. objectstack-fleet commented on Oct 12, 2026

   

   objectstack-fleetbot

   on Oct 12, 2026

   ContributorAuthor

   More actions

   LANDED — PR #12277 is MERGED (2026-10-12T23:11Z, approved by the maintainer, review 5496847263 on head 24c8c17c; readied and auto-merge armed by the seat on that approval; merge commit 48fb558d) and read in place on origin/main 48fb558d. Director seat, summon #37, session_01LqNgdPZB2Jnr5HpGbaqLeB, 2026-10-12T23:11Z: the last objectui landing this seat does under the maintainer's division of labour (「cloud 相关的任务你继续处理，其他任务归车道」).

   probe on origin/main	want	got
   ci.yml: the step Build the packages the repo-root programs import, gated on scope == 'affected'	1	1
   scripts/select-type-check-scope.mjs: rootProgramPackages and --root-builds	present	present
   the selector's test carries the new cases	present	present
   Run type-check step byte-identical to 1fb1ced3 (the full speed untouched)	identical	identical

   • The card was closed by the PR's Fixes line; pm:dispatched stripped and the assignee released in this act.
   • The live acceptance, in this act: the seat updates PR ci(spec-main-shape-gate): drop the merge_group trigger now the context is out of the required set (objectui#12252) #12262's branch from main through the relay (pr_update_branch), so its next run takes the affected speed with the new step; Type Check green there is the card's acceptance on the very PR that found the defect. Read and recorded here when that run completes; the maintainer's auto-merge on ci(spec-main-shape-gate): drop the merge_group trigger now the context is out of the required set (objectui#12252) #12262 then takes it.
   • The merge group's package-less path (H3 of the ACCEPT) is read on the first package-less queue build that comes; none has run since PR ci(type-check): the turbo cache replays unchanged packages, and pull-request and queue runs type-check only what the change affects (objectui#12222) #12251.
8. objectstack-fleet

   removed

   pm:dispatched

   on Oct 12, 2026
9. objectstack-fleet

   unassigned

   os-zhuang

   on Oct 12, 2026
10. objectstack-fleet commented on Oct 12, 2026

    

    objectstack-fleetbot

    on Oct 12, 2026

    Last edited by objectstack-fleet

    ContributorAuthor

    More actions

    Live acceptance read: PR #12262's Type Check is green on the affected speed with the new step. Director seat, summon #37, session_01LqNgdPZB2Jnr5HpGbaqLeB, 2026-10-12T23:20Z.

    PR #12262's branch was updated from main (48fb558d, the merge of PR #12277) through the relay at 2026-10-12T23:13Z; its new head 654f8bbb is the same package-less diff that failed Type Check at c819b74b (run 38311534599, job 115018179844, 21:35–21:36Z: the defect this card was filed on).

    on head 654f8bbb, run 38320983456, job 115048751482 (23:13–23:15Z)	want	got
    Type Check conclusion	success	success
    the step Build the packages the repo-root programs import ran (scope affected)	ran	ran: @object-ui/components @object-ui/fields @object-ui/i18n @object-ui/plugin-dashboard @object-ui/plugin-grid -- Cached: 16 cached, 16 total, FULL TURBO in 118 ms
    pnpm type-check:vitest-setup (tsc -p tsconfig.vitest-setup.json)	exit 0	exit 0, no diagnostics
    the type-check turbo run itself	zero tasks, as before	"No task ran in this turbo run" (the zero-task path PR #12251 opened; now harmless)

    • The card's acceptance is met on the PR that found the defect: the same diff, the same speed, red before 48fb558d and green after it, with the five packages' builds restored from the turbo cache. Job 115048751482 is the citation.
    • Remaining on PR ci(spec-main-shape-gate): drop the merge_group trigger now the context is out of the required set (objectui#12252) #12262 at this reading: the 12 Test shards, Lint, Repo Gates and the advisory Spec Main Shape Gate still running; Build & E2E, Build Docs, Test (dist pins), the three Changeset checks, Governed Surface Queue Guard, Line Citation Gate, Live E2E green. The maintainer's auto-merge takes it when the required set is green; the three-merge-group reading for objectui#12252 follows the merge (the seat's promised reading, left on ci(spec-main-shape-gate): drop the merge_group trigger once the ruleset reads INTACT on seven — an advisory verdict on a queue build is read by nobody (objectui#12240 follow-up, letter B) #12252 when done).
    • H3 (the merge group's package-less path) stays unread: no package-less queue build has run since PR ci(type-check): the turbo cache replays unchanged packages, and pull-request and queue runs type-check only what the change affects (objectui#12222) #12251. PR ci(spec-main-shape-gate): drop the merge_group trigger now the context is out of the required set (objectui#12252) #12262 itself will be the first, when it enters the queue.

    Corrected in place by the seat at 2026-10-12T23:24Z: the failing job cited for c819b74b is 115018179844; the id first written was wrong.
11. objectstack-fleet commented on Oct 12, 2026

    

    objectstack-fleetbot

    on Oct 12, 2026

    ContributorAuthor

    More actions

    H3 read: the merge group's package-less path is covered. Director seat, 2026-10-12T23:43Z. The first package-less queue build since PR #12251 was PR #12262's merge group (run 38322151265, head a1a917a0, 23:26–23:38Z): Type Check success (job 115052481389, 23:26–23:28Z); its step Build the packages the repo-root programs import ran (success, 23:28:19–23:28:20Z, a cache replay) and Type-check repo-root vitest setup files exited 0 right after. With the live acceptance above (the pull_request speed, 6134498869) the card's acceptance is complete on both speeds the fix covers. Nothing further is owed here.