Repository navigation
Publish accepts an activityMilestones[].field or displayNameField that names no field; both are served and do nothing #33101
Description
qa#3 · os-tesla · session_01L9qibw6yXcLi9aRRBpPT4W
QA-source: #33096 · studio-authoring.no-entry-property-matrix · observed beside the clauses
Found: QA run of touched:ee9ad834 at objectstack main 8e3d4a54 (18.4.0 round 2, part B, browser slice), during the off-clause probe. Minor.
Repro (writable package PKG, admin)
PUT /api/v1/meta/object/x_probe?mode=draft&package=PKGwith{name:'x_probe', sharingModel:'public_read_write', fields:{name:{type:'text'}}, activityMilestones:[{field:'nope', value:'won', summary:'x'}]}→ 200.- The same with
displayNameField:'nope'→ 200. POST /api/v1/packages/PKG/publish-drafts→outcome:'published'.GET /api/v1/meta/object/x_probeserves both, withnameField:'nope'.- Result: the milestone can never fire, and a created row's
sys_activity.record_labelis the record id.
Expected: refused at publish, as the membershipKeys pointers (membership-key-*-invalid) and searchableFields (searchable-field-unknown) already are.
Mechanism
validateObjectFieldRefsleaves these scalars out on purpose: "left to a separate decision" (packages/lint/src/validate-object-field-refs.ts:172–195).- The publish probe (
packages/metadata-protocol/src/build-probes.ts:396–414) runs only that rule.
This is a documented deferral, so it may fold into the decision that comment names.
Activity
objectstack-fleet commentedon Nov 13, 2026
triage#1 · hotlong · session_01Q4EjjMnyLqJdzuktEmRPXH
Path: none
Triage: first touch →
domain:spec· p3 ·pm:queue· 2026-11-13T12:26Z⛔ Not a claim.
activityMilestones[].fieldanddisplayNameFieldnaming no field. Both are inert (the milestone never fires, and the record label falls back to the id). These are minor author errors with no data loss, p3.packages/lintvalidate-object-field-refs.ts(the lint is anchored to the spec lane, per SKILL.md).objectstack-fleet commentedon Nov 14, 2026
domain:spec#1 · os-project-manager · session_01FEkQ5mswnWm181XgYcCUag
Claim: PM loop round 1 · 2026-11-14T16:34Z
Session:
session_01FEkQ5mswnWm181XgYcCUag(seat); card threadsession_01FEkQ5oaMukqihPa4PYwzvHis the devAccount:
os-project-managerBranch:
claude/project-thread-9r10fs(the card thread's designated branch; the platform lets a thread push only that branch, so notclaude/issue-33101-*)Worktree:
objectstack-issue-33101Domain:
domain:specSeat:
domain:spec#1File surface:
packages/lint/src/validate-object-field-refs.ts(or the owning rule if measured, e.g.validate-record-title.ts), their tests,rule-explanations.tsif the rule id set changes;packages/metadata-protocol/src/build-probes.tsonly if the probe must call another rule; theactivityMilestones[].field/displayNameFielddescribe inpackages/spec/src/data/object.zod.tsonly to state the refusal;examples/**fixtures only if measured to name a missing field (reported first); root.changeset/(stop on breach; explain in the report)Container & model:
M,mode:thread,model: default tier(narrowing built at the judgment tier; contract review at CONTRACT_REVIEW_TIER before enqueue, MANDATORY path: narrowing face),effort: highClause-②: no (narrowing)
Metadata-protocol: narrowed
Responsibility: the lint rule owner (
packages/lint, objectstack) leaves these pointers unjudged by a code-comment deferral | the platform path that covers the family isvalidateObjectFieldRefsunder the #19332 ruling ("a misspelling is refused loudly at authoring") | any author of an object at the publish door oros build; QA run #33096 reached itThread-read: 6852874653
Premise: holds at
origin/main24f2f401:validate-object-field-refs.ts~:172–195 listsdisplayNameFieldandactivityMilestones[].fieldas NOT owned. Taken under the maintainer's standing order 「p1 p2 处理完了，就可以处理 p3」 (2026-10-26; Ruling-ref), no spec p1/p2 dispatchable. Order:/mnt/project-files/pm-orders/spec/order-33101.md(project share).Serial constraints cleared: OPENPRS
objectstack-fleet commentedon Nov 14, 2026
domain:spec#1 · os-project-manager · session_01FEkQ5mswnWm181XgYcCUag
Claim correction (claim 6880767436) · 2026-11-14T16:37Z
The claim's last line went out with a placeholder. The reading it stands for:
Serial constraints cleared: no open PR touches
validate-object-field-refs.ts,validate-record-title.tsorbuild-probes.ts. PR #33205 (seat 2, #33154) editspackages/lint/src/authoring-rules.tsandvalidate-expressions.ts; if this card must register a rule inauthoring-rules.ts, it merges main once after #33205 lands.objectstack-fleet commentedon Nov 14, 2026
domain:spec#1 · os-project-manager · session_01FEkQ5mswnWm181XgYcCUag
File surface amended (claim 6880767436) · 2026-11-14T16:45Z
packages/spec/src/migrations/entries/semantic/18.object-field-pointer-unknown-refused.ts(the ADR-0087 semantic ledger entry aMetadata-protocol: narrowedchangeset owes, ascheck-adr-0087-registrationrefusesnot-required (no-migration-prescription)on it; same shape as cli: os validate / os build resolve a dataset reader through the reader's declared closure, as the runtime does (authoring half of #32929, 18.4.1) #32967 / lint: the object_extension publish door runs no field-formula judgment, so a computed field os build refuses still publishes through PUT /meta/object_extension #33072 / An object extension's validations are judged at neither the publish door nor os build, yet the registry registers them on the target object #33154), and@objectstack/specpatch in the changeset frontmatter. Requested by the card thread.validate-record-title.tsandbuild-probes.tsstay untouched (the title rule never reports a dangling pointer, sovalidateObjectFieldRefsis the one owner).objectstack-fleet commentedon Nov 14, 2026
domain:spec#1 · os-project-manager · session_01FEkQ5oaMukqihPa4PYwzvH
os-dev-report
{
"issue": 33101,
"status": "done",
"branch": "claude/project-thread-9r10fs",
"pr": "#33370",
"session": "session_01FEkQ5oaMukqihPa4PYwzvH",
"premise_still_valid": true,
"summary": "validateObjectFieldRefs (object-field-ref-unknown, error; os build/validate/lint and the runtime object door via runtimeTypes ['flow','object']) now judges nameField, displayNameField and activityMilestones[N].field verbatim against the authored field map plus the registry-injected columns; a name the title pair shares is reported once, at displayNameField. Mechanism hypothesis measured: validateRecordTitle never reaches the publish door (CLI_ONLY advisory) and never reports a dangling pointer even on the CLI (objectTitleCompleteness grades it 'synthesized', the rule reports only 'none'), and validateSemanticRoles skips an unresolved title pointer, so no rule owned the position: one owner, validate-record-title.ts and build-probes.ts untouched (the probe already runs this rule). Header owned/NOT-owned lists and os explain updated. ADR-0087: new semantic entry object-field-pointer-unknown-refused (the gate refuses no-migration-prescription on narrowed; surface amended in 6880909508); changesets lint minor (fix(lint)!, BREAKING) + spec patch.",
"tests": "HEAD 8e1531c. validate-object-field-refs.test.ts 103/103 (3 new describes: title pair, milestones, runtime door with the card's repro body; it.each over 7 injected columns per position; coverage controls + os explain facts widened). pnpm --filter @objectstack/lint test 208 files/8914 passed + typecheck exit 0; @objectstack/spec test 840 files/26832 passed (1 todo); spec src/migrations 198/198; @objectstack/metadata-protocol test 339 files passed (9 skipped)/22844 passed. Ablation via scripts/ablation-replace.mjs on committed head, each restore proven blob==HEAD and git diff HEAD empty: (1) title judge off -> 9 failed/94 passed; (2) milestone read -> [] -> 5 failed/98; (3) object-graph injected acceptance removed -> 25 failed/78 incl. all 14 new injected-column pins. Lint tests import src (no dist leg). Corpus sweep (scratch, not committed): 132 *.object.ts / 137 objects under packages/ + examples/ (92 title pointers, 1 milestone) through runRuntimeAuthoringRules type object + the probe's single-object call: 0 findings at the new positions; positive control on sys_cloud_binding_credential fires both paths. HotCRM ac4dd3d measured on source text (not installed): 20 title pointers, 5 milestones, each a declared field.",
"gates": "90 derived by dispatch-gates --commands on 8e1531c (base 24f2f40); all exit 0; --ran: 90 run, 0 NOT MEASURED, 0 unrun. Reruns named: check:dts-closure exit 1 first (dogfood closure build rewriting metadata-core/dist concurrently) -> exit 0 rerun; check:type-check-debt exit 3 first (lock budget behind dogfood) -> exit 0 under the lock (0 ledger entries above their number); check:dual-build-cjs-loads exit 3 first (PREREQUISITE NOT MET) -> exit 0 after the full build. check-changeset-no-major + check-empty-changeset replayed as a pull_request event with the PR body: exit 0; check-adr-0087-registration exit 0 (registered, new here). PR check-runs: in_progress at report time.",
"dogfood_battery": "8e1531c2 · VERDICT command-exit 0 (held 2017s) · Test Files 464 passed | 3 skipped · Tests 3927 passed | 42 skipped · no failures",
"mcp_calls": "0",
"api_writes": "2 — fleet-write pr_create (POST pulls via relay, #33370, body read back identical) and this os-dev-report comment (post-stamped.mjs, POST issues/33101/comments); plus 1 git push (not REST)",
"open_questions": [],
"security_findings": "none",
"out_of_scope_findings": [
"carrier: none · noted, not filed — tenancy.tenantField / tenancy.organizationField / lifecycle.ttl.field stay unjudged (the header's deferral, each needs its own injected-column measurement); not a defect class (a)-(c) by itself, recorded in the PR's Acceptance notes",
"carrier: none · noted — main moved 20 commits past base 24f2f40 during the run; none touches the diff's files, so no merge was taken"
]
}
Generated by Claude Code
objectstack-fleet commentedon Nov 14, 2026
domain:spec#1 · os-project-manager · session_01FEkQ5mswnWm181XgYcCUag
ACCEPT → PR #33370 comment 6882167902 · 2026-11-14T18:23Z · ready + auto-merge armed; card closes on merge.