[finding] translation-target-unknown 其余集合级仍只读顶层 —— 每包腿上兄弟包声明的 dashboards / flows / globalActions 各报一条假孤儿(已实测),views / pages / apps 结构同形但未探 #19349

Activity

1. os-warren

   added

   pm:dispatched

   and removed

   pm:queue

   on Sep 22, 2026
2. os-warren

   self-assigned this

   on Sep 22, 2026
3. os-warren commented on Sep 22, 2026

   

   os-warren

   on Sep 22, 2026 – with Claude

   Collaborator

   More actions

   Claim: PM loop from seat domain:spec#2
   Session: session_01UDXER3sdqfeVYpEWZs5mZx
   Branch: claude/issue-19349-translation-target-collection-levels
   Worktree: objectstack-issue-19349
   Domain: domain:spec
   Seat: domain:spec#2
   File surface: packages/lint/src/validate-translation-references.ts + its tests; .changeset/. Stop on breach and explain in the report.
   Container & model: S/M, mode:subagent
   Clause-②: no
   Thread-read: none

   ⚠️ Thread-read: none is a reading, not a skip: the thread has 0 fetchable comments on two independent instruments. The counter discrepancy is reconciled below.

   Why this card is claimable right now

   Serial census re-taken in THIS act, ⛔ not inherited. File lists pulled for all 15 open PRs (255 rows) and intersected against this card's landing surface:

   • exact file packages/lint/src/validate-translation-references.ts — free
   • region packages/lint/src/ — free
   • ⭐ no open PR touches any packages/lint/ path at all (lit control: the same census returns 255 rows across 14 other regions, so the instrument sees files)

   ⇒ no file-level and no region-level collision. This was the only card of 69 dispatchable domain:spec / pm:queue candidates whose declared surface is free; 53 collide on file or region, and 15 name no path in their body and were not auto-classified. ⛔ That residue is stated, not silently counted as blocked.

   Prerequisite is landed, verified rather than assumed: the card points the taker at PR #19347 first (the object-level closure). It is merged at 2026-09-20T14:12:06Z and its card #19064 is closed, so its diff reads off main.

   ⚠️ One reconciliation the next reader should have. This issue's comments counter says 1, while both the comments endpoint and the timeline return 0 commented events. Two instruments against one counter, and neither instrument can produce a body — so there is no readable Claim: on this thread and this claim is not a C9 second claim. ⛔ Recorded rather than smoothed over: a stale counter most likely means a deleted comment, and ⛔ this seat does not assert which.

   Scope, as the card itself fences it

   A 层 (3, already reproduced by the filing seat's probe) — dashboards, flows, globalActions/actions.
   B 层 (3, same shape, ⛔ NOT probed) — views, pages, apps; and apps may be partly covered already by #18442's contributed-app level, which the taker must judge first.

   ⛔ The card is explicit that 6 is not "6 defects" and that B must not be skipped merely because it is unprobed. ⛔ Line numbers in the card are from 2277d1fcd1 and the filing seat recorded a ~90-line divergence from the construction seat's numbers without guessing the cause — the taker re-derives line numbers on its own tree and uses neither set.

   ---

   Generated by Claude Code
4. os-warren commented on Sep 22, 2026

   

   os-warren

   on Sep 22, 2026 – with Claude

   Collaborator

   More actions

   os-dev-report

   {
   "issue": 19349,
   "status": "done",
   "branch": "claude/issue-19349-translation-target-collection-levels",
   "pr": "#19625",
   "session": "session_01UDXER3sdqfeVYpEWZs5mZx",
   "premise_still_valid": true,
   "summary": "All six remaining collection rungs of translation-target-unknown now read the ADR-0130 D4 carrier packages[].manifest.KEY through one generalised reader, artifactProvidedRecords(stack, collection) (#19064's artifactProvidedObjects generalised; 7 call sites). Every level folds RECORDS, for three distinct reasons, not one assumption applied six times. The card's B layer was probed and all three reproduced; apps was partly covered by #18442 and BOTH halves of the declaration question were broken. ⚠️ The Docs Drift Check's empty list is NOT MEASURED, and its declared emitter-only blind spot lands on this diff: a by-shape sweep of content/docs/** and skills/** found THREE published carriers that restate this rule in prose. One of them still teaches the PRE-FIX behaviour in a row that was already stale before this PR. ⛔ None is fixed here — content/docs/ is outside this card's declared file surface and skills/** is a governed Tier H surface; whether they ride this PR is the seat's call. The issue arrived with assignee os-warren already set by the PM dispatch; I did not write it.",
   "tests": "PROBE (throwaway, deleted in the same act, git status --porcelain 0 lines): 6/6 levels reproduced 1 error each on the per-package leg — dashboards.crm_overview, flows.lead_conversion, globalActions.export_all, objects.crm_order._views.board, objects.crm_order._tabs.mine, apps.crm_app; plus a 7th probe proving the second apps half. TEST-FIRST: tests committed red (ffc9273) at 18 failed | 114 passed (132), exit 1. AFTER FIX: 132 passed (132), exit 0. pnpm --filter '@objectstack/lint^...' build exit 0. pnpm --filter @objectstack/lint test — 108 files, 4106 passed / 5 skipped, exit 0. pnpm --filter @objectstack/lint typecheck exit 0. pnpm lint repo-wide eslint . --no-inline-config exit 0 — NOT narrowed, so no narrowing declaration is owed. GATES: dispatch-gates derived 59 families (re-derived after git fetch origin main, byte-identical); --ran with exit codes reconciles 59 derived / 56 run / 3 NOT-MEASURED / 0 UNRUN. check:docs-transcript-drift first exited 3 and was re-run to a real 0 after pnpm --filter '@objectstack/lint...' build. NOT MEASURED (exit 3 = PREREQUISITE NOT MET, neither pass nor finding): check:dual-build-cjs-loads, check:lean-entry-closure, check:type-check-debt — all three refuse without a whole-repo pnpm build; declared to CI. ⊘ ALSO NOT MEASURED: the PR's Docs Drift Check listed nothing and says so is 'not a clean bill of health' — read as exit 3, and compensated by the by-shape sweep recorded under out_of_scope_findings. ABLATION from the COMMITTED fix via scripts/ablation-replace.mjs, one anchor guarding the fold so only the six new rungs are neutralised and #19064's object fold survives: anchor x1 -> x0, blob d6b2706ecbd3 -> 42ee18972671 (mutation LANDED on disk), child 18 failed | 114 passed — the exact pre-fix reading, all 18 reds in this PR's block and zero reds in #19064's block (targeting proof). No build ran between mutation and child and the child went RED, proving the suite resolves this rule through src/ not a stale dist/. RESTORE proven by OBSERVED STATE: disk blob == HEAD blob (d6b2706ecbd3), git diff HEAD 0 lines, git status --porcelain 0 lines. LINE NUMBERS re-derived by SHAPE at merge base 0b4022b: already-reading :733 :828 :886; top-level-only :964 objects, :1034 views, :1040 pages, :1060 actions, :1091 apps, :1146 dashboards, :1177 flows. The card's ~90-line divergence is DETERMINED: every rung differs from the triage seat's set by exactly +96, and #19347's net change to this file is +104/-8 = +96, inserted above every rung — triage measured pre-#19347, construction post-. Neither was wrong. TREE NOTE: my greps ran on PR head 1ef6055, not the check's merge tree bd681fb; all three carrier files are byte-identical between my HEAD and origin/main, so the reading transfers.",
   "mcp_calls": "0 — no MCP GitHub tool was called; all GitHub reads and writes went through the REST proxy with curl",
   "api_writes": "2 — POST /repos/objectstack-ai/objectstack/pulls (draft PR #19625); POST /repos//issues/19349/comments (this report). Zero label writes: the dispatch named no labels and explicitly forbade needs:contract-review, and skip-changeset does not apply (a changeset is included), so the write scope is empty. 4 git push (not REST). PR body written once on create, ⛔ no PATCH; read back and byte-identical except the platform's appended session-URL footer block.",
   "open_questions": [],
   "out_of_scope_findings": [
   "⭐ noted, not filed — PRE-FIX BEHAVIOUR STILL TAUGHT, seat's call whether it rides this PR: content/docs/protocol/kernel/i18n-standard.mdx:1013-1040, the 'Must name' table. The row objects.{object} | an object this stack defines, or a platform object reads, on the per-package leg, as 'a sibling package's object is an orphan' — the pre-fix rule. ⚠️ That row was ALREADY stale before this PR (#19064 widened the object rung; #18441 added the objectExtensions target rung) and #19347's own acceptance notes flagged it. This PR widens the SAME table's understatement across six more rungs: _views, _sections, _actions, apps.{app} / .navigation.{id}, dashboards.* and globalActions.{action} now also resolve against a sibling's declaration. Those rows are vague ('a view's name') rather than actively contradicting. Already absent pre-PR: _tabs, _validations and every flows.* row, and #18442's contributed-app level. ⛔ Not fixed: content/docs/ is outside this card's declared file surface.",
   "noted, not filed — POSITIVE ALIGNMENT, no drift: skills/objectstack-i18n/SKILL.md:196-203 states the reject condition by EXISTENCE ('A key naming an object, field, view, tab, action, param, section, validation rule, app, nav item, dashboard, widget or flow screen that does not exist'). Pre-fix the rule reported keys naming things that DO exist in the artifact, contradicting this published page; post-fix the implementation moves TOWARD it. It is also broader than the docs table (it lists tab, validation rule and flow screen). ⛔ Untouched — skills/** is a governed Tier H surface (Prime Directive #14), so editing it would retier this PR and require a maintainer-authorized approval to land.",
   "noted, not filed — understated, not contradicted: content/docs/deployment/validating-metadata.mdx:625 says the per-package walk is 'handed the artifact's own packages[] as resolution context so a sibling package's OBJECT still resolves'. The mechanism it describes is unchanged and correct; after this PR six more collections resolve through it, so 'object' is now an incomplete instance list rather than a false statement. No stop condition.",
   "SWEEP METHOD, so the above is a reading and not an absence of effort: by SHAPE over content/docs/** (405 pages) and skills/** (44 files), ⛔ not by the identifiers the Docs Drift Check already covers — shapes tried: 'this stack defines' (2), 'does not define' (4), 'sibling package' (4), 'per-package' (14), 'per package' (6), 'orphan' (40), 'packages[]' (13), 'co-ownership' (0), 'does not exist|no such|must name' over skills (9), plus the rule id itself (2). LIT CONTROL: 'translation bundle' returns 27 hits, so the instrument reads this prose and the narrow counts are readings.",
   "noted, not filed: the branch already existed on origin at an ancestor of main with 0 commits ahead — an empty routing probe from the three dead devs. Nothing to build on; started clean from a fast-forward. An observation about this run, not a defect."
   ]
   }

   ---

   Generated by Claude Code
5. os-warren commented on Sep 22, 2026

   

   os-warren

   on Sep 22, 2026 – with Claude

   Collaborator

   More actions

   Report accepted — PR #19625, head 1ef60553c7. The seat's disposition of every open item.

   domain:spec execution seat 2, session session_01UDXER3sdqfeVYpEWZs5mZx, 2026-09-22T02:1xZ.

   ⭐ The card's own unanswered question is now ANSWERED, and the seat verified it first-hand

   This card recorded a ~90-line divergence between the triage seat's line numbers and the construction seat's, and deliberately refused to guess the cause: 「原样记下不猜因」.

   The dev determined it: every rung differs by exactly +96, and #19347's net change to validate-translation-references.ts is +104 / −8 = +96, inserted above every rung ⇒ triage measured pre-#19347, construction post-#19347. ⛔ Neither set was wrong.

   Seat's own check, ⛔ not taken on trust: git show --numstat for that path on #19347's commit returns 104 8. The arithmetic reproduces. ⇒ the card's instruction to re-derive was right, and the mystery is closed rather than inherited.

   What was delivered

   Six rungs now read the ADR-0130 D4 carrier packages[].manifest.KEY through one generalised reader — artifactProvidedRecords(stack, collection), #19064's artifactProvidedObjects generalised, 7 call sites. ⭐ Every level folds records, and the dev gives three distinct reasons rather than one assumption applied six times — which is what the card asked for when it said ⛔ do not assume all six want the same treatment.

   B layer was probed and all three reproduced — so the card's 「同形未探」 is now 「已探」. ⚠️ And apps came back the way the card warned it might: partly covered by #18442, with BOTH halves of the declaration question broken — a 7th probe proved the second half. ⇒ the card was right to say ⛔ judge apps first.

   Ablation, from the committed fix: one anchor guarding the fold so only the six new rungs are neutralised and #19064's object fold survives. Anchor ×1 → ×0, blob d6b2706ecbd3 → 42ee18972671 (mutation landed on disk), child 18 failed / 114 passed — the exact pre-fix reading, with all 18 reds in this PR's block and zero in #19064's (targeting proof). No build ran between mutation and child and the child went red, proving the suite resolves through src/ and not a stale dist/. Restore proven by observed state: disk blob == HEAD blob, git diff HEAD 0 lines, git status --porcelain 0 lines.

   ⛔ NOT MEASURED, correctly declared rather than banked as green

   • check:dual-build-cjs-loads, check:lean-entry-closure, check:type-check-debt — all three exit 3 (PREREQUISITE NOT MET) without a whole-repo pnpm build. ⇒ neither a pass nor a finding; declared to CI.
   • check:docs-transcript-drift first exited 3 and was re-run to a real 0 after building — the right instinct, ⛔ not a banked skip.
   • ⊘ The PR's Docs Drift Check listed nothing and the dev read that as NOT MEASURED, citing the check's own words 「not a clean bill of health」. ⭐ Correct, and it is the reason the three carriers below were found at all.

   The three published carriers — the seat's call, ⛔ none rides this PR

   The by-shape sweep was this seat's ask, and it paid for itself. Disposition:

   1. ⛔ content/docs/protocol/kernel/i18n-standard.mdx:1013-1040 — filed as its own card, finding(docs): the i18n "Must name" table still states the PRE-FIX orphan rule — row 1 is false on the per-package leg, six more rows are now understated #19626. Row 1 (objects.{object} → "an object this stack defines, or a platform object") is actively false on the per-package leg; the seat read it on origin/main before filing. ⚠️ It was already stale before this PR ([finding] translation-target-unknown's OBJECT rung reports a false orphan on the per-package leg — the rule's universe reads stack.objects only, while its own docblock claims artifactProvidedObjectNames reach #19064 widened the object rung, [finding] translation-target-unknown reports a locale key for an objectExtensions[]-injected field as an orphan at error — indistinguishable from a real typo #18441 added objectExtensions, fix(lint): the translation rule's OBJECT rung reads what the artifact's packages[] provide #19347's acceptance notes flagged it) ⇒ ⛔ not this PR's regression, and ⛔ not a rider: content/docs/ is outside this card's declared file surface, the fix wants the whole table as one unit (six more rows this PR understates, plus _tabs / _validations / flows.* / [finding] translation-target-unknown tells a contributor package that the app it ships labels for is "not defined by this stack" — os build per-package leg, app-name rung #18442's contributed-app level already absent), and PR feat(spec)!: split the translation bundle type — settings is a platform group, not a per-app one (#15178) #19600 currently holds that file — a rider would be a serial collision.
   2. ✅ skills/objectstack-i18n/SKILL.md:196-203 — positive alignment, ⛔ no action. It states the reject condition by existence, so pre-fix the rule contradicted it and post-fix the implementation moves toward it. ⭐ The dev's refusal to touch it was right for a second reason it named itself: skills/** is governed Tier H, so editing it would retier this PR and require a maintainer-authorized approval to land.
   3. ⓘ content/docs/deployment/validating-metadata.mdx:625 — understated, not false. The mechanism it describes is correct; "object" is now an incomplete instance list. ⛔ No stop condition. Recorded on finding(docs): the i18n "Must name" table still states the PRE-FIX orphan rule — row 1 is false on the per-package leg, six more rows are now understated #19626.

   Next

   At-tier contract review dispatched on head 1ef60553c7, with CONTRACT_REVIEW_TIER re-read from origin/main at 2026-09-22T02:08Z immediately beforehand, per the downgrade fuse. ⚠️ Clause-②: no on this card, so ⛔ no needs:contract-review carrier is hung and none is owed — but the landing check still runs the full list, and tonight's incident added a fourth item to it: ④ carrier state verified on both carriers before any enqueue.

   ---

   Generated by Claude Code
6. os-warren

   closed this as completedin #19625on Sep 22, 2026
7. os-warren

   removed

   pm:dispatched

   on Sep 22, 2026
8. os-warren commented on Sep 22, 2026

   

   os-warren

   on Sep 22, 2026 – with Claude

   Collaborator

   More actions

   Release: PR #19625 MERGED 2026-09-22T03:23Z — card closed completed by its own closing keyword. Seat domain:spec#2, session_01UDXER3sdqfeVYpEWZs5mZx.

   What landed

   Squash c3a4c74e36985748a2ef54a3fd10f95b90562928 on main, fix(lint): the translation rule's remaining collection rungs read what the artifact's packages[] provide (#19625). The PR carried 3 commits and main received 1 — verified by git rev-list --parents, which shows exactly one parent, so the merge queue squashed it. ⚠️ Worth recording because the auto_merge response reported merge_method inconsistently across two PRs enqueued minutes apart (squash here, merge on #19623); the tree says squash, and the tree is what counts. ⛔ The field is not the reading.

   @objectstack/lint patch, Clause-②: no. Six collection rungs of translation-target-unknown now read packages[].manifest.* through artifactProvidedRecords, so a package translating what a SIBLING package of the same artifact declares is no longer reported as an orphan on the per-package leg.

   Landing preconditions, as satisfied

   1. At-tier PASS on record — comment 5770448504, Served-tier: 225/225 CONTRACT_REVIEW_TIER, naming head 1ef60553c7. ⚠️ That record SUPERSEDED an earlier FAIL (5770384505) whose sole basis was CI. ⛔ The seat did not convert that FAIL into a PASS by assertion — the reviewing session was resumed and re-read CI itself at the same unchanged head.
   2. check-clause2-carriers --pair 19625 — EXIT 0.
   3. CI green by job conclusion, latest run per check NAME — 34 names, 31 success, 3 skipped (all rostered), 0 failure, 0 cancelled. ⛔ Never an aggregate roll-up in place of member lanes: the Test Core aggregate was green AND all six shards independently green.
   4. needs:contract-review absent on both carriers (this card and the PR), verified before the enqueue.

   Governed-surface reading: governedPathsIn over the final file list returns matched.length = 0. ⚠️ Read as matched, ⛔ never as the tier string — governedTierFor answers 'H' as a fail-closed default on an empty match and would have misread this ungoverned PR as Tier H. Control lit: the same call over skills/objectstack-i18n/SKILL.md + AGENTS.md returns 2, tier H.

   The CI incident this round, and what it left behind

   Test Core (6/6) failed once at this head on packages/plugins/plugin-dev/src/dev-plugin-tenancy-posture.test.ts:154, Test timed out in 5000ms. It is a member lane, so a real failure rather than an aggregator refusal, and it was diagnosed rather than re-queued: plugin-dev declares no dependency on @objectstack/lint (15 deps + 4 devDeps, none), main was green in the same window, and every other shard was green. The one sanctioned re-run passed ⇒ intermittent, not deterministic.

   ⚠️ Carded rather than routed around, as #19631 — and ⛔ NOT as a plain flake: @objectstack/cli was scheduled on this same shard 6 and packages/cli does declare @objectstack/lint, so a packages/lint diff invalidates the workspace's heaviest test package on exactly the shard that timed out. The failing test's own header (#3060) declares it vulnerable to precisely that. ⛔ The card states that a testTimeout raise is the LAST rung of the ordered remedy, never the opening move.

   ⚠️ One figure on #19631 was corrected after filing: @objectstack/cli's weight is 458.15s in scripts/test-shard-timings.json, the dataset the partitioner reads. The 548.6 / 474.4 first cited there live only in the HEADER of scripts/partition-test-shards.mjs:29, inside the historical table arguing why test-file COUNT was the wrong weight. The mechanism is unaffected; the citation was off the wrong instrument. Caught by the at-tier review and confirmed first-hand before the edit.

   Residue — ⛔ nothing here blocks, all of it is named so it is not lost

   • finding(docs): the i18n "Must name" table still states the PRE-FIX orphan rule — row 1 is false on the per-package leg, six more rows are now understated #19626 — the i18n Must-name table's row 1 is actively false on the per-package leg. Pre-dates this PR.
   • finding(spec): three more VIEW_BINDING_BLOCKS rows assert renderer fallbacks objectui deleted — gantt, timeline (half) and map, and the map row cannot move without its message and test pin #19630 — three more VIEW_BINDING_BLOCKS rows whose premises moved. ⚠️ Its body carried a stale reading of spec(ui): the list-view calendar guard gates on appearance.allowedVisualizations ONLY — type: 'calendar' with no calendar: block parses CLEAN at all three doors #16577 as a pending ruling; corrected in place, since spec(ui): the list-view calendar guard gates on appearance.allowedVisualizations ONLY — type: 'calendar' with no calendar: block parses CLEAN at all three doors #16577 closed completed on 2026-09-11.
   • Two prose nits the review raised and this PR did not take, both non-blocking: the :871 claim that no manifest form exists beside this one is loose for objects (which ManifestSchema declares as glob patterns — the operative claim about RECORDS is true and recordsOf returns [] on a glob array), and the apps-unions vs dashboards/flows-first-wins precedence asymmetry would read better with one cross-reference.

   ---

   Generated by Claude Code
9. akarma-synetal

   added a commit that references this issue on Sep 28, 2026

   fix(lint): the translation rule's remaining collection rungs read wha…

   ...

   This issue will close once commit c3a4c74 is merged into the 'main' branch.

   c3a4c74