test: retire 17 family rows strict lint now owns; keep the i18n gate and the props parse sweep (#1582)

[objectstack-fleet]

Fixes #1582
Part of #1583
Fixes #1584
Part of #1585
Fixes #1586
Fixes #1805
Clause-②: no

The one retirement PR for the family re-grade (ruling hotcrm-R74 item 4 B; re-grade 6053840674, fold plan 6053925467). Every row was re-ablated on this PR's base b54380cf (the two-package artifact from #1907) before its local assertion was deleted. 17 of the 19 planned rows retire. 2 rows stay, because the re-ablation found cases that objectstack lint --strict does not report on this base. #1583 and #1585 therefore get Part of, not a closing line: each keeps one row, listed under "Rows kept after re-ablation".

What changed

• F1 (Retire the local tests that re-implement the platform's flow / predicate / readonly-write lint rules, once os lint --strict guards them (epic #1579, step 3, family F1) #1582): removed the inert config.condition sweep. Deleted test/view-tab-label-inert.test.ts, with its walk guard.
• F2 (Retire the local metadata reference-integrity tests the platform's page / view / nav / action / field-group rules already enforce (epic #1579, step 3, family F2) #1583): retired page-field-unknown, field-group-undeclared / -empty / -shadowed, ai-skill-tool-unresolved (plus the transcribed PLATFORM_TOOLS list, which only that row used) and the view-references form-field sweep. Kept: the component-props parse sweep.
• F3 (Retire the local analytics / dataset integrity tests the platform's chart / widget / dataset rules already enforce (epic #1579, step 3, family F3) #1584): retired A1, A3, A4 and A8. Removed the helpers only they used: datasetByName, measuresOf, dimensionsOf and the DATE_RANGE_PRESETS import.
• F4 (Retire the local i18n reference test and the severity-promoting i18n gate, once strict covers the i18n/missing-* and translation-target rules (epic #1579, step 3, family F4) #1585): retired the option-key sweep, the apps.*.navigation orphan sweep and the section-name-missing sweep. Kept: the whole lint:i18n-gate gate. Its script, test, package.json script, verify step, both CI steps and roster entries are byte-identical to main.
• F5 (Retire the local security / sharing / org-axis coverage tests the platform's security-* and sharing-rule-* rules already enforce (epic #1579, step 3, family F5) #1586, Re-run the three-gate test on test/authorization-coverage.test.ts's qualified-FLS-key half once @objectstack/lint ships security-fls-unknown-field (objectstack#16108 → objectstack PR #16998, merged, not yet released) #1805): split the FLS assertion. The unqualified-key and names-a-real-field branches are retired. The unknown-object branch stays, renamed every object-qualified FLS key names a real object. Retired both RLS sweeps, plus USER_VARS / CLAUSE, which only they used. "Every policy targets an object the same set grants" is unchanged.
• Each deletion site has a short ⚰️ RETIRED comment that names the rule id now responsible.

Re-ablation on b54380cf, per case (before deletion)

Method: same as the re-grade. Each case was one anchored mutation through ablation-replace.mjs. It ran pnpm lint --json (objectstack lint --strict), then the owning test file. A trap restored the tree, and the restore was proven three ways: blob hash equals HEAD, git diff HEAD is empty, and git status is clean. Baseline on b54380cf: errors 0, warnings 0, suggestions 18, failing 0, strict true, exit 0. In every row below the local assertion was also red.

retired row	rule id	cases: lint exit + rule
F1 inert decision config.condition	flow-inert-node-condition	top level 1 · nested in loop/try 1 · service-module flow 1 (two flows) · blank '' 1 expression-invalid · null 0 (same as the re-grade: a null states no predicate, so it is outside the assertion's rationale)
F1 view-tab-label-inert	spec 17.5.0 schema tombstone	list.tabs 1 · listViews.*.tabs 1 (config load refuses view.list.tabs was removed in @objectstack/spec 17.5.0, no rule id)
F2 record:details / highlights / path fields	page-field-unknown	highlights 1 · path statusField 1 · details section field 1 · service-module page 1
F2 field-level group resolves	field-group-undeclared	1
F2 no empty group	field-group-empty	1
F2 no fully hoisted group	field-group-shadowed	1
F2 skill tool names resolve	ai-skill-tool-unresolved	plain 1 · action_ 1 · wildcard 1 · service-module skill 1 · a name only a stack.tools record serves: 0, held by the kept does not lean on AI tool metadata (red)
F2 form section fields	form-field-unknown	default form 1 · formViews object entry 1
F3 A1 dataset	chart-dataset-unknown	chartless 1 · charted 1 · joined block 1 · service-module report 1
F3 A3 xAxis	chart-dimension-unknown	1
F3 A4 rows/columns/values	chart-dimension-unknown · chart-measure-unknown	rows chartless 1 · columns 1 · values 1 · rows charted 1 · rows joined block 1 · service-module rows + values 1 (both ids)
F3 A8 preset comparand	filter-preset-comparand	bare on a widget 1 · $eq on a widget 1 · $in in a report 1
F4 option keys by value	translation-option-key-unknown	extra key en 1 · zh-CN 1 · English-label key 1 · options on a field with none 1 · unknown field 1 translation-target-unknown
F4 apps.*.navigation orphans	translation-target-unknown	unknown nav id 1 · unknown app 1
F4 section label without name	translation-section-name-missing	view form 1 · formViews 1 · page record:details in tabs 1
F5 FLS unqualified / real field	security-fls-unqualified-key · security-fls-unknown-field	bare key 1 · unknown field 1 · empty remainder 1 · unknown object 0 (kept branch, deliberate per objectstack PR #16998)
F5 RLS predicates (both sweeps)	rls-predicate-unknown-field · -unknown-user-variable · -unenforceable	using: unknown field 1 · unknown leading field 1 · current_user.nope 1 · nested path 1 · function call 1 · check (new, not in the re-grade): unknown field 1 · current_user.nope 1 · function call 1 · unknown policy object 0, held by the kept every policy targets an object the same set grants (red)

After deletion, at 68fe1b2c: one case per retired row was re-run with lint only (18 probes). Every probe gave exit 1 with the same rule id, or the same schema refusal for list.tabs.

Rows kept after re-ablation (not retired)

1. F2: component-props parse sweep (every page component's properties parse against its own ComponentPropsMap entry). This case is new; the re-grade did not probe it. A page component nested in another component's properties, with its own properties absent and a contract that requires keys, gets exit 0. Measured on a nested record:related_list, a nested page:accordion and a nested element:text. The local sweep is red on all three. Control: the same edit on a top-level record:path gets exit 1 (component-props-invalid). Mechanism: PageComponentSchema defaults a top-level component's properties to {}, but nothing parses a nested component, and validateComponentProps skips one with if (!props) continue. The three nested types in use today are also held, by chance, by other kept tests. A new nested type with required props would be held by nothing. The test file is unchanged.
2. F4: the lint:i18n-gate gate. This is a regression against the re-grade. On 68fe1b2c (the gate restored) one non-default-locale (zh-CN) entry was deleted per class: object, field, option, section, view, action, navigation, dashboard, widget, dataset, page and flow. All 12 classes gave pnpm lint exit 0 (plain and --json) and gate exit 0. On c529de2b the re-grade measured exit 1 on all 12. Mechanism: the CLI's translation-coverage pass runs computeI18nCoverage on the composed top level. Since Packaging: sales becomes the type: app package and service a type: module — composeStacks(…, { manifest: 'preserve' }), navigation contributions, two-entry packages[] #1907, composeStacks(…, { manifest: 'preserve' }) leaves that level with only manifest, packages and i18n: no translations and no metadata. Run in memory over the union-of-packages stack, the same deletion is reported (warning zh-CN navigation apps.crm_enterprise.navigation.nav_home.label). The kept local completeness tests are red for field, option, section, view, action, widget and page. Nothing in pnpm verify is red for object, navigation, dashboard, dataset or flow. The gate removal could not be shown to lose nothing, so per the dispatch rule the gate stays, and the platform defect is reported for the seat to file upstream.

KEEP rows in the files this PR edits: byte-identity

Method: each it(…) block was extracted from b54380cf and from HEAD, and the text was hashed (sha1, first 12 hex). Across the 8 edited test files, 118 blocks are byte-identical. Seventeen were removed (the retired rows) and one was added (the FLS split). One block is comment-only: sees a real navigation tree and an apps.*.navigation table in every locale, whose inline comment was rewritten because it pointed at the retired orphan sweep. Its non-comment lines hash equal (44031195) at both revisions. The KEEP rows of the table, and the holders named above:

file › assertion	base	HEAD
flow-decision-authority › every decision node actually decides something	ff12fefc4618	ff12fefc4618
metadata-references › every page component's properties parse against its own ComponentPropsMap entry (kept, row 1 above)	7a3fefb3d34d	7a3fefb3d34d
metadata-references › every page component type has a props contract to parse against	985131628dce	985131628dce
field-groups-coverage › %s declares fieldGroups	3c63094cf981	3c63094cf981
view-references › kanban groupByField is a select field with options	64c37d829e9b	64c37d829e9b
skills-integrity › does not lean on AI tool metadata, which never executes (holder)	8af703ba057d	8af703ba057d
i18n-references › translated object and field keys name real objects and fields	4c90ff38134e	4c90ff38134e
i18n-references › no locale carries a _views entry for a view the stack does not ship	8222ee4af2da	8222ee4af2da
i18n-references › every authored field label and help string is translated	4bcc95a5d1d6	4bcc95a5d1d6
i18n-references › every option value has a translated label in every locale	341603826787	341603826787
i18n-references › every form and detail-page section heading is translated	c59cb84472f6	c59cb84472f6
i18n-references › every canonical view label is translated in every locale	983e0ac81a02	983e0ac81a02
i18n-references › every action has a label entry in every locale pack	11584fdb8f9e	11584fdb8f9e
i18n-references › every dashboard widget has a translated title and description	764a555ec7bf	764a555ec7bf
i18n-references › every page has translated nav copy and header copy	89afa7bb9a85	89afa7bb9a85
i18n-references › every reference-rail object has a translated label in every locale	0775ced90511	0775ced90511
i18n-references › every view empty state is translated	8ff1ee70b6e9	8ff1ee70b6e9
i18n-references › every action parameter label is translated	00d69619006f	00d69619006f
authorization-coverage › every business object is granted in at least one permission set	c878f28a53d3	c878f28a53d3
authorization-coverage › every object nav item is readable by at least one permission set	bb3a81aea126	bb3a81aea126
authorization-coverage › every policy targets an object the same set grants (holder)	140aa8fe8c8a	140aa8fe8c8a
authorization-coverage › a sharing rule only targets an object whose OWD lets shares apply	a317ec6e9049	a317ec6e9049
authorization-coverage › the guest set carries no export bit (ADR-0090 D9 anchor rule)	2ca385723718	2ca385723718

The gate's files have an empty git diff b54380cf HEAD: scripts/check-lint-i18n-gate.mjs, test/lint-i18n-gate.test.ts, package.json, both workflows, test/script-main-guard.test.ts, scripts/lib/main-module.mjs and test/verify-log-decoy-pin.test.ts.

F4 gate-removal proof: not obtained, gate kept

deleted zh-CN entry (class)	pnpm lint	pnpm lint --json failing	gate	local test red
object, navigation, dashboard, dataset, flow	0	0	0	none
field, option, section, view, action, widget, page	0	0	0	the matching completeness assertion in i18n-references

Stale prose corrected in this PR

• test/i18n-references.test.ts. The completeness block comment said pnpm lint runs with --skip-i18n and that lint exits 0 on warnings. It now gives what was measured: the run is strict, but its i18n coverage reports nothing on this artifact. The View-labels comment, the nav block comment and the "no _sections category" sentence (17.7.0 has i18n/missing-section) were corrected the same way.
• src/sales/profiles/{sales-manager,marketing-user}.profile.ts, comments only. They said "liveness: dead in 17.6.0". The installed 17.7.0 liveness/permission.json marks props.rowLevelSecurity.children.label and .description "status": "live" (verifiedAt 2026-10-03, "RE-GRADED dead → live … the permission editor's Row-Level Security section heads each policy card with pol.label").
• Comments in the edited test files that named a retired assertion. They now name the rule id instead: flow-decision-authority, field-groups-coverage, skills-integrity, analytics-integrity, view-references (also its view-tab-label-inert pointer) and the metadata-references split map.

Outside the claimed surface, edited because this PR's retirements made the text false:

• content/docs/analytics/cubes.mdx, .zh-Hans.mdx and .zh-Hant.mdx. These said test/analytics-integrity.test.ts fails the build on an unresolved dataset, measure or dimension name; they now credit objectstack validate and objectstack lint --strict.
• docs/ARCHITECTURE.md and docs/developers/code_examples.md. These named test/skills-integrity.test.ts as the tool-name resolver and said PLATFORM_TOOLS is transcribed there.

Not edited here, now stale. Both are src comments, and the dispatch allows only the two profile comments under src/. They are listed for #1919, the src-comment card that follows this PR:

• src/revenue/objects/opportunity_line_item.object.ts:69: "test/field-groups-coverage.test.ts pins it" (field-group-shadowed).
• src/service/flows/case-escalation.flow.ts:123: "test/flow-decision-authority.test.ts enforces that".

Verification

• pnpm verify at 68fe1b2c exited 0 through the verify lock (VERDICT command-exit 0). Validate: "✓ Validation passed". Typecheck: clean. Lint: "18 suggestion(s)", exit 0. lint:i18n-gate: "✓ i18n lint gate: 0 i18n/missing-* issues". Hygiene: "✓ source hygiene clean". Token ratchet: "✓ source token ratchet clean". Build: "✓ Build complete". Test: "Test Files 178 passed (178) · Tests 3814 passed | 1 skipped (3815)".
• Probe totals: 104 probe runs in 15 batches. Every probe log records RESTORED-PROVEN, the batch runner reported no unproven restore, and every batch ended with "TREE CLEAN after batch". Four probes were refused by ablation-replace because the anchor was contained in its replacement (no-op, restored), and each was redone with a unique anchor.

Acceptance notes (for the seat; not filed by this PR)

• Platform defect (F4 kept row): objectstack lint translation coverage is blind on a composeStacks artifact. Evidence is above. Since Packaging: sales becomes the type: app package and service a type: module — composeStacks(…, { manifest: 'preserve' }), navigation contributions, two-entry packages[] #1907, pnpm verify catches no missing non-default-locale label for object, navigation, dashboard, dataset or flow.
• Platform gap (F2 kept row): component-props-* skips a nested page component whose properties is absent.
• ai-skill-tool-unresolved counts a stack.tools record as resolving, while ToolSchema says a metadata-authored tool is never executed in the open edition. The kept skills test holds this case today.
• test/i18n-shared-widget-parity.test.ts:23 still says "pnpm lint --skip-i18n skips the bundles". This was stale before this PR, and the file is outside this surface.
• The restored gate's own comments (the gate script header, and the CI step comment "Under --strict (Turn on os lint --strict in this repo's verify chain and prove the gate reds (epic #1579, step 2b — the half that needs a release) #1581) it now fails on both, so this step overlaps it") are kept byte-identical with the row, and are not true on this base.

---

Generated by Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
hotcrm	Ignored		Oct 8, 2026 8:02am UTC