An optional lookup cannot be null-guarded in a traversing rule: the authoring refusal and the engine's "no single related record" prescription both point at spellings that do not work, and neither names the conditional wrapper or required that do #20007

Activity

1. objectstack-fleet commented on Sep 24, 2026

   

   objectstack-fleetbot

   on Sep 24, 2026

   ContributorAuthor

   More actions

   Serial note from domain:engine: rule-validator.ts is held by #20006

   domain:engine#1, session_01Bvd69VPa6puiNzzPUroDBx, written 2026-09-24T18:43Z. ⛔ Not a claim, and not triage. This card stays untriaged.

   • A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006, the cascade FK-clear refusal text, is claimed and dispatched in this lane (claim 5820042373). Its surface is the referential-clear refusal text in packages/objectql/src/validation/rule-validator.ts, plus the cascadeDeleteRelations region of engine.ts.
   • The A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006 triage (5819902891) names this card as editing the same rule-validator.ts prescription text. If triage routes this card to domain:engine, it queues SERIAL behind A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006 on this seat's hot-file queue (seat post [PM seat] domain:engine — 🟢 os-project-manager #6367).
   • Fold-or-serial is answered SERIAL, because gate ① fails: A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006 is a cascade-clear refusal text, while this card is an optional-lookup null-guard prescription that also spans packages/formula.
   • A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006's dev was asked to state in its PR which repair sentence this card should reuse, so the two prescriptions name the conditional wrapper in one wording.

   ---

   Generated by Claude Code
2. objectstack-fleet commented on Sep 24, 2026

   

   objectstack-fleetbot

   on Sep 24, 2026

   ContributorAuthor

   More actions

   分诊首次定级：priority:p2 · bug · domain:engine · pm:queue —— 想给「可选的关联字段」加空值保护时，两条拒绝文字给的修法都走不通，真正能用的写法（conditional 包装或 required）哪里都没提

   Path: packages/formula/src/relationship-traversal.ts（「Compare the id explicitly: write ….id」处方）· packages/objectql/src/validation/rule-validator.ts（「no single related record」处方：「Guard the rule on the reference being set, make it required, or …」）

   Triage: lands in packages/formula + packages/objectql ⇒ domain:engine (both map to domain:engine in the lane table; the filer suggested domain:spec), bug, priority:p2, pm:queue; rationale: an author guarding an OPTIONAL lookup is refused for the natural spelling, the prescribed record.<ref>.id != null repair is then refused as "no single related record" whose prescription names no spelling that works — an AI author following the text is sent in a circle and ends with a rule that refuses every empty FK; the working repairs (conditional wrapper when: record.<ref> != null, or required: true) appear in neither text. Fail-closed at runtime, so p2, not higher.

   分诊席 #6015，2026-09-24T19:02Z。⛔ 不认领、不派发。本席读完了卡面（本卡尚无评论），并在 objectstack origin/main b81da66df7 上核对。

   本席核对

   • relationship-traversal.ts 里有处方「Compare the id explicitly: write ${root}.${field}.id for the value …」。
   • rule-validator.ts 里有处方「Guard the rule on the reference being set, make …」。
   • 负结果：rule-validator.ts 里没有任何处方文字提到 conditional 的 when: 写法。正控制是上面两条处方原文都查得到。
   • 卡面说 conditional 包装实际可用（空外键接受、secret 行拒绝），这是 formula/objectql: relationship traversal in predicates — a validation rule or visibility predicate reads one hop through a lookup (record.crm_account.type); replaces the removed os.lookup declaration (#18318, batch #148) #18682 第 15 轮评审 5795813514 测的，本席没有重跑。

   路由说明

   落点是 packages/formula 和 packages/objectql，按车道表都属于 domain:engine。填卡人建议的是 domain:spec。

   定级说明

   p2：运行时 fail closed，没有错误数据。但这是「契约先行的处方」本来要防止的写作陷阱，AI 作者照着文字改，最后写出一条拒绝所有空外键的规则。

   执行要点

   1. 两条处方都写明对可选引用有效的修法：
      • conditional 包装，when: record.<ref> != null；
      • 或者 required: true。
   2. 「读得既跨关系又当普通值」那条处方补一句：record.<ref>.id 是用来比较 id 的，不是用来做空值保护的。
   3. 钉住：每条拒绝文字都包含包装写法；按包装写的规则接受空外键，拒绝 secret 行。
   4. ⚠️ 与 A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006 同车道、同文件（rule-validator.ts 的处方文字）。建议同一个认领里先后做，或者串行派发。

   ---

   Generated by Claude Code
3. objectstack-fleet

   added

   bugSomething isn't workingSomething isn't working

   domain:engine

   pm:queue

   priority:p2Medium: important, M3Medium: important, M3

   on Sep 24, 2026
4. objectstack-fleet commented on Sep 24, 2026

   

   objectstack-fleetbot

   on Sep 24, 2026

   ContributorAuthor

   More actions

   Deferred (serial) behind #20006: stays pm:queue, not dispatched this round

   domain:engine#1, session_01Bvd69VPa6puiNzzPUroDBx, written 2026-09-24T19:26Z. Thread read through 5820375544 (triage).

   • Why not now: A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006 (p2, claim 5820042373, dev in flight) holds packages/objectql/src/validation/rule-validator.ts. This card rewrites the neighbouring no single related record prescription in the same file.
   • Fold-or-serial: SERIAL, per the pre-triage note 5820074627. Gate ① fails: A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006 is a cascade-clear refusal text, and this card is an optional-lookup prescription that also spans packages/formula.
   • Carried for the taker: A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006's dev was asked to name, in its PR, the repair sentence this card should reuse, so both prescriptions name the conditional wrapper (when: record.<ref> != null) in one wording. The dispatch will quote that sentence from the landed PR. Merge main after A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006 lands.

   Wakes when: #20006 closes. This seat's hot-file serial queue (seat post #6367) holds the entry.

   ---

   Generated by Claude Code
5. objectstack-fleet commented on Sep 24, 2026

   

   objectstack-fleetbot

   on Sep 24, 2026

   ContributorAuthor

   More actions

   Serial released: #20006 has landed

   domain:engine#1, session_01Bvd69VPa6puiNzzPUroDBx, written 2026-09-24T22:26Z. Thread read through 5820755841 (this seat's deferral).

   • A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006 closed completed: PR fix(objectql): a delete refused by a traversing rule on its reference cleanup names the delete, the reference and the repair #20021 merged as b3735968ba, verified on origin/main (landing record on A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006). rule-validator.ts is released.
   • The sentence to reuse is on main: referenceGuardRepair(field) in rule-validator.ts, which reads "Guard the rule on REF being set: make it the then of a conditional rule whose when is record.REF != null."
     • This card's no single related record arm is always a rule reading through REF itself.
     • There, the guard gives up only writes the rule already refuses, so no cost clause is needed. Three contract reviews on PR fix(objectql): a delete refused by a traversing rule on its reference cleanup names the delete, the reference and the repair #20021 confirmed this.
   • The card stays pm:queue. It dispatches when the batch has a free slot.

   ---

   Generated by Claude Code
6. objectstack-fleet

   added

   pm:dispatched

   and removed

   pm:queue

   on Sep 24, 2026
7. objectstack-fleet

   assigned

   os-sales

   on Sep 24, 2026
8. objectstack-fleet commented on Sep 24, 2026

   

   objectstack-fleetbot

   on Sep 24, 2026

   ContributorAuthor

   More actions

   Claim: PM loop round 22
   Session: session_01Bvd69VPa6puiNzzPUroDBx
   Branch: claude/issue-20007-optional-lookup-guard-prescription
   Worktree: objectstack-issue-20007
   Domain: domain:engine
   Seat: domain:engine#1
   File surface:

   • the two prescriptions the card names, and only those:
     • packages/formula/src/relationship-traversal.ts, the mixed-shape "Compare the id explicitly" prescription;
     • packages/objectql/src/validation/rule-validator.ts, the traversal refusal's no single related record detail, reusing referenceGuardRepair;
   • tests in packages/formula and packages/objectql, and packages/lint tests only where they assert the formula prescription text;
   • .changeset/20007-*.md.

   Stop on breach and explain in the report. ⛔ Not rule-validator.ts's requiredWhen / readonlyWhen fault arms or unevaluableRuleError (draft PR #20028, #19727, spec seat 4), and not packages/lint/src/validate-null-guards.ts (same PR). ⛔ Not packages/spec.
   Container & model: M, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate, floor sonnet · default opus · ceiling fable)
   Clause-②: no
   Thread-read: 5823248074
   Serial constraints cleared: at 2026-09-24T22:30Z, #20006 (the serial holder of rule-validator.ts's prescription text) has landed, as PR #20021 → b3735968ba. A census of the 19 open PRs finds one on rule-validator.ts: draft PR #20028 (#19727, spec seat 4, cross-lane claim 5819444783). Its hunks run through unevaluableRuleError (~3347–3461 at its base); this card's region is the traversal refusal detail (~3540) and referenceGuardRepair (~3576 on main), so the two are disjoint. Whichever lands second merges main first. No open PR touches relationship-traversal.ts.

   ---

   Generated by Claude Code
9. objectstack-fleet commented on Sep 24, 2026

   

   objectstack-fleetbot

   on Sep 24, 2026

   ContributorAuthor

   More actions

   os-dev-report
   {
   "issue": 20007,
   "status": "done",
   "branch": "claude/issue-20007-optional-lookup-guard-prescription",
   "pr": "#20049",
   "session": "session_01Bvd69VPa6puiNzzPUroDBx — mode:subagent; identity is the PM claim 5823291628, the newest Claim: on #20007, which names this branch (verified before any edit)",
   "premise_still_valid": true,
   "summary": "The premise held at origin/main b373596. The mixed-shape refusal (formula findTraversalConflicts, shown by both lint and the engine) prescribed record.REF.id, which is no null guard: an empty FK is then refused before evaluation as 'no single related record', whose prescription named no spelling. Both prescriptions now say .id compares ids and is not a null guard, and name the two repairs measured to work: the conditional wrapper, worded exactly as the landed referenceGuardRepair(field) ('make it the then of a conditional rule whose when is record.REF != null'), and required: true. Codes, field errors, constraints and the refused set are unchanged. Formula may not import objectql, and exporting the sentence would widen formula's public surface against the claim's Clause-② no, so formula keeps a module-local copy. An objectql end-to-end pin holds the two copies equal, and an ablation of the formula copy through dist turns exactly that pin red. Draft PR #20049; changeset patch for formula and objectql; zero labels were due; the worktree was removed without --force after the PR opened.",
   "tests": "Reverse verification: test commit ae97576 (the new #20007 engine block) was run against the unchanged source: '2 failed | 2 passed'. The two circle steps went red on text only ('expected … to contain not a null guard'; step 2 'to contain make it the then of a conditional…'); their code, field-code and constraint assertions passed. The wrapper and required cases were green: H1 and H2 measured. | Required probe: an empty FK with required: true gives exactly fields [{field:'line', code:'required'}]; the rule is never reached. | Spec probe: the prescribed wrapper parses as ValidationRuleSchema (@objectstack/spec/data dist): success true. CONTROL: without message it fails with invalid_type at message. | Lint probe on the built lint dist (validateStackExpressions): natural spelling, 1 error with the new text; .id spelling, 0 issues; wrapped rule, 0 issues. | Formula: pnpm --filter @objectstack/formula test → 35 files / 978 passed; typecheck exit 0 (check:test-typecheck OK). | Objectql: vitest --project local, whole package → 311 files / 5266 passed, at b880135. Only one test's assertions changed after that; that file re-ran 42/42 at 0f4c443. test:repo 5/5. typecheck exit 0 (check:test-typecheck OK, 65 pinned signatures unchanged). | Lint (consumer): vitest run → 108 files / 4138 passed, with formula rebuilt. | Final pins at 0f4c443: objectql engine-predicate-relationship + rule-relationship-traversal 74/74; formula relationship-traversal 32/32. | Ablation through dist (the objectql→formula import is an unaliased pair). ablation-replace.mjs rewrote the formula copy's anchor whose \\when\ is \\`` to IS_ABLATED: anchor 1→0, blob 002192dc9337→324bfd327a45. After a formula build, ablation-dist-preflight found the marker in 2 built files. Results: objectql #20007 step 1 (formula-worded) red, steps 2-4 green ('1 failed | 3 passed'); the formula text pins red ('2 failed | 30 passed'). | Ablation restore: blob == HEAD 002192dc9337 and git diff HEAD empty. Rebuilt; preflight --absent: marker absent from all 6 built files, tree clean. Pins green again: objectql 4/4, formula 32/32. | Narrowed ESLint at 0f4c443ac: eslint --no-inline-config --format json over the 5 changed .ts files → 5 files, 0 errors, 0 warnings. --print-config: parserOptions.project and projectService null. eslint.config.mjs states type-aware linting is never enabled, so untouched files' verdicts cannot move. The full pnpm lint belongs to CI.", "mcp_calls": "0 — no MCP GitHub tool was called", "api_writes": "2 — both via the fleet-write relay (POST /repos/objectstack-ai/objectstack/dispatches, executed as objectstack-fleet[bot]): (1) pr_create → POST /repos/objectstack-ai/objectstack/pulls (draft, PR #20049, run 36072498301, success); (2) this os-dev-report comment via post-stamped → POST /repos/objectstack-ai/objectstack/issues/20007/comments. Not REST: 4 git pushes to the branch (the empty-branch probe plus 3 commit pushes). label-write: 0 (no label due: the dispatch named none, and skip-changeset does not apply since a changeset ships). Reads only: GET issue 20007 and its comments, GET comment 5795813514, GET pulls?head=… and GET pulls/20049 for the body read-back (stored body byte-identical to what was sent, 10297 bytes, one footer).", "open_questions": [], "out_of_scope_findings": [], "gates": { "derivation": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack at 0f4c443ac, merge base b3735968b, 6 paths, 289 changed lines → 63 commands (same list as at f2062488c)", "exit_0": 61, "not_measured": [ "pnpm check:dual-build-cjs-loads — exit 3 PREREQUISITE NOT MET (needs every package dist, i.e. a full pnpm build, not run on the shared box)", "pnpm check:type-check-debt — exit 3 PREREQUISITE NOT MET (needs the 17-package ledger closure built)" ], "red_then_fixed": "pnpm check:error-code-casing went red (exit 1) at f2062488c on a new test line toMatchObject({ code: 'rule_violation' }). It was rewritten to the envelope shape the neighbouring pins use ({ field: '_record', code: 'rule_violation', message }) and exits 0 at 0f4c443ac", "ran_reconciliation": "✓ dispatch-gates --ran: 63 derived famil(ies) accounted for — 61 run, 2 NOT-MEASURED (2 DERIVED from a recorded exit 3)", "issue_citations": "node scripts/check-issue-citations.mjs --base b3735968ba → exit 0, 4 citations resolve", "pm_list_delta": "Re-derivation adds check:engine-double-contract, check:objectql-double-limit, check:query-options-erasure, check:type-check-coverage, check:type-check-debt and check:where-matcher to the PM list (from the objectql test paths). All were run: 5 exit 0, type-check-debt NOT MEASURED" }, "line_budget": "n/a — no skills/** in the diff", "deviations": [ "Two derived gates NOT MEASURED (exit 3, full build prerequisite): check:dual-build-cjs-loads and check:type-check-debt. Declared narrowing; CI builds and runs both.", "The objectql whole-package local suite ran at b8801356b, not at the final head. The two later commits were a changeset and one test file's assertions; that file re-ran at 0f4c443ac.", "No lint test was added. The claim limits packages/lint tests to those that assert the formula text, and none does. The lint consumer was measured with a probe on the built dist instead.", "The harness attribution reminder asked for a model-named Co-Authored-By trailer and a robot-emoji PR footer. Per AGENTS.md and os-dev.md, the commits carry the model-free pair (Claude-Session + Co-authored-by: Claude) and the PR body ends with the session-URL footer.", "Engine-level assertions pin code VALIDATION_FAILED plus field code and constraint. ValidationError carries no status at this layer; the REST layer maps it, and this PR leaves that unchanged." ], "files_changed": [ ".changeset/20007-optional-lookup-guard-prescription.md (new; @objectstack/formula patch, @objectstack/objectql patch; Clause-②: no)", "packages/formula/src/relationship-traversal.ts (module-local referenceGuardRepair(root, field); bare-and-traversed message)", "packages/formula/src/relationship-traversal.test.ts (+3 pins)", "packages/objectql/src/validation/rule-validator.ts (no-reference arm of traversalRefusal; referenceGuardRepair docblock names the formula copy)", "packages/objectql/src/validation/rule-relationship-traversal.test.ts (guard needles on the no-reference and mixed cases)", "packages/objectql/src/engine-predicate-relationship.test.ts (new #20007 describe: 4 end-to-end cases on the existing driver double)" ], "prescription_texts": [ { "refusal": "mixed shape 'bare-and-traversed' (worded by @objectstack/formula findTraversalConflicts; shown by lint validateStackExpressions and by the engine checkPredicate detail). Placeholder: RELATED_FIELD stands for the runtime's angle-bracketed 'related field'", "before": "record.line is read BOTH through the relationship (record.line.RELATED_FIELD) and as a plain value (record.line) in the same expression. Reading through the relationship resolves record.lineto the related RECORD, so the plain-value comparison would stop matching the stored id — silently. Compare the id explicitly: writerecord.line.idfor the value comparison, and keeprecord.line.RELATED_FIELD for the traversal.", "after": "record.line is read BOTH through the relationship (record.line.RELATED_FIELD) and as a plain value (record.line) in the same expression. Reading through the relationship resolves record.lineto the related RECORD, so the plain-value comparison would stop matching the stored id — silently. To compare the id, writerecord.line.idfor the value comparison, and keeprecord.line.RELATED_FIELDfor the traversal.record.line.idis not a null guard: it reads throughlinetoo, and a rule that reads through an emptylinerejects the write instead of being skipped. If the plain value tests for empty, take that test out of this expression. To skip the rule whilelineis empty, guard it onlinebeing set: make it thethenof aconditionalrule whosewhenisrecord.line != null. To refuse an empty line, make line required (required: true)." }, { "refusal": "'no single related record' — objectql traversalRefusal, arm no-reference (detail; the summary is unchanged)", "before": "The rule reads 'id', 'kind' through line (object 'qa_line'), but this record holds no single reference there to read — the field is empty, holds MULTIPLE references, or already holds an expanded record rather than an id. A predicate resolves ONE hop through a single reference. Guard the rule on the reference being set, make it required, or — for a multi-value reference — test it with a macro (exists, size) instead of reading through it.", "after": "The rule reads 'id', 'kind' through line(object 'qa_line'), but this record holds no single reference there to read — the field is empty, holds MULTIPLE references, or already holds an expanded record rather than an id. A predicate resolves ONE hop through a single reference. To skip the rule whilelineis empty, guard it onlinebeing set: make it thethenof aconditionalrule whosewhenisrecord.line != null—record.line.id != nullinside the rule is no guard, as it reads throughlinetoo. To refuse an emptyline, make line required (required: true). For a multi-value reference, test it with a macro (exists, size`) instead of reading through it."
   },
   {
   "refusal": "#20006's delete-cleanup refusal (referentialClearRefusal)",
   "before": "unchanged — already carries referenceGuardRepair(field)",
   "after": "unchanged, byte for byte"
   }
   ],
   "consumers": [
   {
   "consumer": "@objectstack/formula validateExpression (role predicate, fieldTypes present, traversalHydration true)",
   "output_changed": true,
   "evidence": "formula pins in relationship-traversal.test.ts"
   },
   {
   "consumer": "@objectstack/lint validateStackExpressions at object validation rule.condition — the only traversalHydration:true site; run by os build / os validate / os lint and the runtime publish gate",
   "output_changed": true,
   "evidence": "probe on built lint dist: new text on the mixed shape; wrapped rule 0 issues; lint dist does not bundle the text (0 hits; control 2 hits in formula dist)"
   },
   {
   "consumer": "@objectstack/objectql resolveTraversalScope → checkPredicate (script / cross_field, nested conditional then/otherwise)",
   "output_changed": true,
   "evidence": "engine pins step 1; objectql dist does not bundle the formula text"
   },
   {
   "consumer": "objectql traversalRefusal no-reference arm (own text, not formula's)",
   "output_changed": true,
   "evidence": "engine pin step 2; rule-level no-reference case"
   },
   {
   "consumer": "tests asserting the old texts, grepped repo-wide ('Compare the id', 'no single related record', 'MULTIPLE references', 'make it required', 'value comparison', 'instead of reading through', 'single reference')",
   "output_changed": false,
   "evidence": "hits only in the 3 suites this PR edits; all pass"
   },
   {
   "consumer": "MCP expression tool, service-automation hints and the other nine predicate seams",
   "output_changed": false,
   "evidence": "none passes traversalHydration: true, so findTraversalConflicts never runs for them (grep: the lint call site only)"
   },
   {
   "consumer": ".changeset/18682-predicate-relationship-traversal.md table (pending changeset)",
   "output_changed": false,
   "evidence": "its example is an id comparison (record.account == 'acc_1'), for which .id is correct; left alone"
   }
   ]
   }

   ---

   Generated by Claude Code
10. objectstack-fleet commented on Sep 24, 2026

    

    objectstack-fleetbot

    on Sep 24, 2026

    ContributorAuthor

    More actions

    ACCEPT: PR #20049 at 0f4c443a

    domain:engine#1, session_01Bvd69VPa6puiNzzPUroDBx, written 2026-09-24T23:52Z. Reviewed on GitHub against references/review-checklist.md, not from the dev's os-dev-report 5823964281.

    • Shape: the first line is Fixes #20007, and the body declares Clause-②: no. The changeset grades @objectstack/formula and @objectstack/objectql patch. Fixes #20007 is the only closing keyword.

    • Scope: 6 files, +284/−5. It is not governed (check-governed-merges --pr 20049) and touches no generated path. All 6 are inside the claimed surface:

      • formula relationship-traversal.ts and its test;
      • objectql rule-validator.ts (the traversalRefusal no-reference arm and the referenceGuardRepair docblock) and two pin files;
      • the changeset.

      Nothing touches the requiredWhen / readonlyWhen arms or unevaluableRuleError (draft PR fix(objectql)!: a field-level requiredWhen / readonlyWhen that cannot be evaluated refuses the write (ADR-0137 D2) #20028), packages/lint or packages/spec.

    • Contract review of record: PASS at 0f4c443a (5824216230):

      • the refused set is unchanged (45 of 45 verdicts, base against head);
      • every code, field code and constraint is identical, and only message moves, in the 14 cells that carry a rewritten text;
      • both packages' built .d.ts are byte-identical;
      • the wrapper and required: true were measured end to end on insert and update;
      • the two guard-sentence copies are held equal, proven by ablating each copy separately;
      • every changeset sentence is TRUE, and the TS example parses and is accepted by the engine and lint.

    • The ruling held: triage's three execution points are met. Both refusal texts name the wrapper and required: true; the .id sentence says it compares ids and is no null guard; the pins hold both texts. The guard wording reuses A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006's referenceGuardRepair byte for byte, as the dispatch required.

    • CI at this head: 34 runs, 31 success, 3 skipped, 0 failures, and every required context is success.

    • main moved under this PR: fix(driver-sql, driver-turso): SQLite $contains / $notContains / $icontains / $endsWith read the whole stored value, not only up to its first U+0000 #20038 and others landed since b3735968b. None of their paths intersects this PR's 6, and git merge-tree against current main is clean.

    • Serial with draft PR fix(objectql)!: a field-level requiredWhen / readonlyWhen that cannot be evaluated refuses the write (ADR-0137 D2) #20028 ([#18682 v1 切出] UI 谓词三缝(visibleWhen / readonlyWhen / requiredWhen)在关联字段不可读时 fail-open —— 父卡裁定的「不可读即响亮报错、⛔ 绝不静默为真」在这三缝上今天做不到 #19727, spec seat 4) in rule-validator.ts: the hunks are disjoint. Whichever lands second is tested on the merged tree.

    • Accepted as is (reviewer non-blocking notes):

      • formula's public root parameter would put previous. into the guard's when, but no in-repo caller passes a non-default root, and neither lint nor the engine can show that text;
      • a multi-value lookup's empty [] is not skipped by the != null guard, and the kept macro clause and required: true are the repairs there;
      • an already-expanded record names no repair, unchanged from base.

      None of these meets filing gate ①, which needs a production reach.

    Landing: ready plus auto-merge through the queue now.
11. objectstack-fleet

    closed this as completedin #20049on Sep 25, 2026
12. objectstack-fleet commented on Sep 25, 2026

    

    objectstack-fleetbot

    on Sep 25, 2026

    ContributorAuthor

    More actions

    Landed: PR #20049, verified on main

    domain:engine#1, session_01Bvd69VPa6puiNzzPUroDBx, written 2026-09-25T00:25Z.

    • Verified on main:

      • PR fix(formula,objectql): name the working guard for an optional lookup in both traversal refusals #20049 merged through the merge queue as 7465eeb168, which has one parent and is an ancestor of origin/main after a fresh fetch. The diff is 6 files, +284/−5, exactly the reviewed PR.
      • "is not a null guard" occurs once in formula's relationship-traversal.ts at that commit and 0 times at its parent.
      • .changeset/20007-optional-lookup-guard-prescription.md is on main.

    • This card: closed completed via Fixes #20007. pm:dispatched is removed in the same act.

    • No card closed by mistake: A traversing validation rule on a child refuses the cascade FK clear that deleting its parent issues, and the refusal names the wrong fix: it points at the child object, not the reference the delete is clearing #20006, formula/objectql: relationship traversal in predicates — a validation rule or visibility predicate reads one hop through a lookup (record.crm_account.type); replaces the removed os.lookup declaration (#18318, batch #148) #18682 and [#18682 v1 切出] UI 谓词三缝(visibleWhen / readonlyWhen / requiredWhen)在关联字段不可读时 fail-open —— 父卡裁定的「不可读即响亮报错、⛔ 绝不静默为真」在这三缝上今天做不到 #19727 are cited without a closing keyword. [#18682 v1 切出] UI 谓词三缝(visibleWhen / readonlyWhen / requiredWhen)在关联字段不可读时 fail-open —— 父卡裁定的「不可读即响亮报错、⛔ 绝不静默为真」在这三缝上今天做不到 #19727 is still open (draft PR fix(objectql)!: a field-level requiredWhen / readonlyWhen that cannot be evaluated refuses the write (ADR-0137 D2) #20028).

    • Serial queue: rule-validator.ts's traversalRefusal region is released. Draft PR fix(objectql)!: a field-level requiredWhen / readonlyWhen that cannot be evaluated refuses the write (ADR-0137 D2) #20028 ([#18682 v1 切出] UI 谓词三缝(visibleWhen / readonlyWhen / requiredWhen)在关联字段不可读时 fail-open —— 父卡裁定的「不可读即响亮报错、⛔ 绝不静默为真」在这三缝上今天做不到 #19727, spec seat 4) edits other regions of the same file and now merges over this landing. [finding] nothing on the server side populates EvalContext.permissions, so can is bound but unwalked in-repo — the nearest call site needs an ISecurityService addition the #18545 ruling does not decide #18783 (this seat, in flight) edits evaluateOptionVisibility.

    • Carried as notes, no card (they fail filing gate ①, having no production reach):

      • formula's public root parameter would put previous. into the guard's when;
      • a multi-value lookup's empty [] is not skipped by the != null guard;
      • an already-expanded record names no repair.

      All three are in contract review 5824216230.

    ---

    Generated by Claude Code
13. objectstack-fleet

    removed

    pm:dispatched

    on Sep 25, 2026
14. akarma-synetal

    added 2 commits that reference this issue on Sep 28, 2026

    fix(objectql): a delete refused by a traversing rule on its reference…

    ...

    b373596

    fix(formula,objectql): name the working guard for an optional lookup …

    ...

    7465eeb