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

.changeset/1582-retire-tests-strict-lint-owns.md

@@ -0,0 +1,4 @@
+---
+---
+
+Tests only, releases nothing. For maintainers: the local test assertions whose every case `objectstack lint --strict` (`pnpm lint`) now reports are retired, row by row — the inert decision `config.condition`, `list.tabs`, page and form field references, field-group integrity, skill tool names, report dataset/dimension/measure bindings, preset filter comparands, the option-key, navigation-orphan and section-name translation checks, FLS key qualification and field names, and RLS predicates (#1582, #1583, #1584, #1585, #1586, #1805). `pnpm verify` still runs `lint:i18n-gate`: on the two-package artifact, `objectstack lint`'s `i18n/missing-*` coverage reports nothing, so the gate was not retired.

content/docs/analytics/cubes.mdx

@@ -84,7 +84,7 @@ So the three questions this section offered — *"ROI by campaign type"*, *"cost
 
 ## How users interact with datasets
 
-What the app guarantees is the binding: a report or a dashboard tile names a dataset and selects its measures and dimensions by name, and `test/analytics-integrity.test.ts` fails the build if any of those names does not resolve. That is what keeps two tiles showing the same number — and it is also why the names in `src/*/datasets/` are worth reading directly rather than through a copy.
+What the app guarantees is the binding: a report or a dashboard tile names a dataset and selects its measures and dimensions by name, and the platform's own checks — `objectstack validate` and `objectstack lint --strict`, which `pnpm verify` and CI both run — fail the build if any of those names does not resolve. That is what keeps two tiles showing the same number — and it is also why the names in `src/*/datasets/` are worth reading directly rather than through a copy.
 
 The drag-and-drop pivot experience this section used to describe — pick a cube, drag dimensions to rows and columns, filter, pivot, chart, save the view as a report, pin it to a dashboard — is a **platform** surface rather than something HotCRM declares. Nothing in this app's metadata configures such a screen, and nothing here tests one, so this page makes no claim about it in either direction. If your workflow depends on self-service pivoting, verify it against your deployment rather than against this page.
 

content/docs/analytics/cubes.zh-Hans.mdx

@@ -84,7 +84,7 @@ description: HotCRM 分析底下的语义层——它是什么、回答哪些业
 
 ## 用户如何与 dataset 打交道
 
-应用能保证的是这层绑定：报表或仪表盘磁贴按名字点出一个 dataset，并按名字选它的度量与维度，而 `test/analytics-integrity.test.ts` 会在其中任何一个名字解析不到时让构建失败。正是这一点让两个磁贴显示同一个数字 —— 这同时也是为什么 `src/*/datasets/` 里的那些名字值得直接去读，而不是隔着一份副本读。
+应用能保证的是这层绑定：报表或仪表盘磁贴按名字点出一个 dataset，并按名字选它的度量与维度，而平台自身的检查（`pnpm verify` 与 CI 都会运行的 `objectstack validate` 和 `objectstack lint --strict`）会在其中任何一个名字解析不到时让构建失败。正是这一点让两个磁贴显示同一个数字 —— 这同时也是为什么 `src/*/datasets/` 里的那些名字值得直接去读，而不是隔着一份副本读。
 
 本节曾经描述的那套拖放透视体验 —— 选一个 cube、把维度拖到行和列、筛选、透视、出图、把视图存成报表、钉到仪表盘 —— 是**平台**侧的界面，而不是 HotCRM 声明出来的东西。本应用的元数据里没有任何东西配置这样一块屏，本仓也没有任何测试覆盖它，所以本页对它不做任何方向上的断言。如果你的工作流依赖自助透视，请对着你自己的部署去验证，而不是对着本页。
 

content/docs/analytics/cubes.zh-Hant.mdx

@@ -86,7 +86,7 @@ description: HotCRM 分析底下的語義層——它是什麼、回答哪些業
 
 ## 使用者如何與 dataset 打交道
 
-應用能保證的是這層繫結：報表或儀表板磁貼按名字點出一個 dataset，並按名字選它的度量與維度，而 `test/analytics-integrity.test.ts` 會在其中任何一個名字解析不到時讓建構失敗。正是這一點讓兩個磁貼顯示同一個數字 —— 這同時也是為什麼 `src/*/datasets/` 裡的那些名字值得直接去讀，而不是隔著一份副本讀。
+應用能保證的是這層繫結：報表或儀表板磁貼按名字點出一個 dataset，並按名字選它的度量與維度，而平台自身的檢查（`pnpm verify` 與 CI 都會執行的 `objectstack validate` 和 `objectstack lint --strict`）會在其中任何一個名字解析不到時讓建構失敗。正是這一點讓兩個磁貼顯示同一個數字 —— 這同時也是為什麼 `src/*/datasets/` 裡的那些名字值得直接去讀，而不是隔著一份副本讀。
 
 本節曾經描述的那套拖放樞紐體驗 —— 選一個 cube、把維度拖到列和欄、篩選、樞紐、出圖、把檢視存成報表、釘到儀表板 —— 是**平台**側的介面，而不是 HotCRM 宣告出來的東西。本應用的中繼資料裡沒有任何東西組態這樣一塊螢幕，本倉也沒有任何測試涵蓋它，所以本頁對它不做任何方向上的斷言。如果你的工作流依賴自助樞紐分析，請對著你自己的部署去驗證，而不是對著本頁。
 

docs/ARCHITECTURE.md

@@ -225,7 +225,8 @@ the platform assistant by surface affinity.
 Skills declare no bespoke tools (ADR-0109). They compose the platform's data
 tools with the `action_<name>` tools the runtime materialises from Actions that
 opt in via `ai.exposed` (ADR-0011) — every name a skill declares must resolve to
-one of those, which `test/skills-integrity.test.ts` enforces.
+one of those, which `pnpm lint` (`objectstack lint --strict`, rule
+`ai-skill-tool-unresolved`) enforces.
 
 The `live_data` skill explicitly requires live schema inspection before
 answering record questions, because admins can change metadata over time.

docs/developers/code_examples.md

@@ -298,8 +298,7 @@ sources resolve, and only two:
   guess at. The ones a CRM skill wants are `describe_object`, `list_objects`,
   `query_records`, `query_data`, `get_record`, `aggregate_data`,
   `search_knowledge` and `visualize_data`. From 17.0 the authoritative set is
-  exported as `PLATFORM_PROVIDED_TOOL_NAMES` from `@objectstack/spec/system`;
-  before 17.0 it is transcribed into `test/skills-integrity.test.ts`. Note
+  exported as `PLATFORM_PROVIDED_TOOL_NAMES` from `@objectstack/spec/system`. Note
   `search_knowledge` retrieves over a *declared knowledge source*, and there is
   currently nowhere in a skills-only app to declare one.
 - **`action_<name>`** — materialised from an Action that opts in with
@@ -309,8 +308,10 @@ sources resolve, and only two:
 Authoring a `defineTool` record does **not** create a third source: `ToolSchema`
 is a read-only projection for Studio discovery with no `implementation` field
 and no executor. Reasoning — scoring, drafting, forecasting — belongs in
-`instructions`, not in a tool (ADR-0109). `test/skills-integrity.test.ts`
-enforces all of this at PR time.
+`instructions`, not in a tool (ADR-0109). At PR time `pnpm lint` resolves every
+tool name (`ai-skill-tool-unresolved`), and `test/skills-integrity.test.ts`
+refuses `stack.tools` records and checks that a referenced Action is AI-exposed
+with a headless path.
 
 Note there is no `permissions` key on a skill — `SkillSchema` has no such field,
 so one is silently stripped ([#511](https://github.com/objectstack-ai/hotcrm/pull/511)).

src/sales/profiles/marketing-user.profile.ts

@@ -55,8 +55,9 @@ export const MarketingUserProfile = {
       name: 'opportunity_private_owner_only_marketing',
       // A deal flagged Private is visible only to its owner, even to holders of
       // org-wide opportunity read. (`rowLevelSecurity[].label` / `description`
-      // have no runtime effect — liveness: dead in 17.6.0 — so the intent lives
-      // in this comment.)
+      // are display copy — the 17.7.0 liveness ledger marks both live: the
+      // permission editor draws them on the policy card. This policy authors
+      // neither, so its intent lives in this comment.)
       object: 'crm_opportunity',
       operation: 'select' as const,
       using: 'is_private == false || owner_id == current_user.id',

src/sales/profiles/sales-manager.profile.ts

@@ -110,8 +110,9 @@ export const SalesManagerProfile = {
       name: 'opportunity_private_owner_only',
       // A deal flagged Private is visible only to its owner, even to holders of
       // org-wide opportunity read. (`rowLevelSecurity[].label` / `description`
-      // have no runtime effect — liveness: dead in 17.6.0 — so the intent lives
-      // in this comment.)
+      // are display copy — the 17.7.0 liveness ledger marks both live: the
+      // permission editor draws them on the policy card. This policy authors
+      // neither, so its intent lives in this comment.)
       object: 'crm_opportunity',
       operation: 'select' as const,
       using: 'is_private == false || owner_id == current_user.id',

test/analytics-integrity.test.ts

@@ -1,19 +1,18 @@
 // Copyright (c) 2025 ObjectStack. Licensed under the Apache-2.0 license.
 
 import { describe, it, expect } from 'vitest';
-import { DATE_RANGE_PRESETS } from '@objectstack/spec/ui';
 import stack from './helpers/composed-stack';
 
 /**
  * Analytics metadata guards (#492).
  *
- * `os validate` / `build` check that a report HAS a chart and a dataset name,
- * but never that the chart's axes resolve to measures/dimensions the dataset
- * actually defines, nor that filter time-windows stay relative at runtime.
- * Each rule below was a real defect:
+ * `os validate` / `build` check that a report HAS a chart and a dataset name.
+ * Whether the dataset and its dimensions and measures resolve is checked by
+ * `objectstack lint --strict` (`pnpm lint`; see the retired block below), which
+ * catches the defect this file was first written for: case/opportunity report
+ * charts named measures (`case_number`, `amount`) that exist in no dataset, so
+ * the yAxis rendered empty. What stays here, each rule a real defect:
  *
- *  - case/opportunity report charts named measures (`case_number`, `amount`)
- *    that exist in no dataset — the yAxis rendered empty;
  *  - churn/opportunity reports computed ISO dates at module load, freezing
  *    "last 30 days" into dist/objectstack.json at whatever day the artifact
  *    was built;
@@ -22,16 +21,9 @@ import stack from './helpers/composed-stack';
  */
 
 type AnyRec = Record<string, any>;
-const datasets: AnyRec[] = (stack as any).datasets ?? [];
 const reports: AnyRec[] = (stack as any).reports ?? [];
 const dashboards: AnyRec[] = (stack as any).dashboards ?? [];
 
-const datasetByName = new Map(datasets.map((d) => [d.name, d]));
-const measuresOf = (dataset: string): Set<string> =>
-  new Set((datasetByName.get(dataset)?.measures ?? []).map((m: AnyRec) => m.name));
-const dimensionsOf = (dataset: string): Set<string> =>
-  new Set((datasetByName.get(dataset)?.dimensions ?? []).map((d: AnyRec) => d.name));
-
 /** A report or a joined-report block — anything carrying a dataset binding. */
 const reportBlocks = (r: AnyRec): AnyRec[] =>
   r.type === 'joined' && Array.isArray(r.blocks)
@@ -40,41 +32,12 @@ const reportBlocks = (r: AnyRec): AnyRec[] =>
 const allBlocks = reports.flatMap(reportBlocks);
 const labelOf = (b: AnyRec) => (b.__parent ? `${b.__parent}/${b.name}` : b.name);
 
-describe('report dataset bindings resolve', () => {
-  it('every report block names a defined dataset', () => {
-    const bad = allBlocks
-      .filter((b) => b.dataset && !datasetByName.has(b.dataset))
-      .map((b) => `${labelOf(b)}: dataset "${b.dataset}" is not defined`);
-    expect(bad, bad.join('\n  ')).toEqual([]);
-  });
-
-  it('every chart xAxis is a dimension of the report dataset', () => {
-    const bad: string[] = [];
-    for (const b of allBlocks) {
-      if (!b.chart?.xAxis || !b.dataset) continue;
-      if (!dimensionsOf(b.dataset).has(b.chart.xAxis)) {
-        bad.push(`${labelOf(b)}: xAxis "${b.chart.xAxis}" is not a dimension of "${b.dataset}"`);
-      }
-    }
-    expect(bad, `chart xAxis names a missing dimension:\n  ${bad.join('\n  ')}`).toEqual([]);
-  });
-
-  it('every rows / columns / values entry resolves against the dataset', () => {
-    const bad: string[] = [];
-    for (const b of allBlocks) {
-      if (!b.dataset || !datasetByName.has(b.dataset)) continue;
-      const dims = dimensionsOf(b.dataset);
-      const meas = measuresOf(b.dataset);
-      for (const d of [...(b.rows ?? []), ...(b.columns ?? [])]) {
-        if (!dims.has(d)) bad.push(`${labelOf(b)}: dimension "${d}" not in "${b.dataset}"`);
-      }
-      for (const v of b.values ?? []) {
-        if (!meas.has(v)) bad.push(`${labelOf(b)}: measure "${v}" not in "${b.dataset}"`);
-      }
-    }
-    expect(bad, `report bindings name missing dataset members:\n  ${bad.join('\n  ')}`).toEqual([]);
-  });
-});
+// ⚰️ RETIRED (#1584): "every report block names a defined dataset" (A1),
+// "every chart xAxis is a dimension of the report dataset" (A3) and "every
+// rows / columns / values entry resolves against the dataset" (A4).
+// `objectstack lint --strict` reports them as `chart-dataset-unknown`,
+// `chart-dimension-unknown` and `chart-measure-unknown` — on a charted, a
+// chartless and a joined-report block alike — and `pnpm lint` fails on each.
 
 describe('metric tiles carry no fabricated trend deltas', () => {
   /**
@@ -170,28 +133,10 @@ describe('time windows stay relative at runtime', () => {
     expect(bad, `absolute dates frozen into metadata:\n  ${bad.join('\n  ')}`).toEqual([]);
   });
 
-  it('no filter comparand is a bare date-range PRESET name (objectstack#8690)', () => {
-    // A preset name is the picker's vocabulary, not the query layer's. The
-    // console lowers `last_30_days` into `{ from: '{30_days_ago}', to:
-    // '{today}' }` before any filter is sent; a preset name written straight
-    // into metadata never gets that lowering. It is not a `{macro}` either, so
-    // the resolver does not reject it — it reaches the driver as a literal
-    // string, compares false against every row, and the query answers HTTP 200
-    // with ZERO rows and no diagnostic. The symptom is an all-zero dashboard,
-    // indistinguishable at a glance from the #460 defect that cost this
-    // dashboard its date picker for a release. Filed upstream as
-    // objectstack#8690; guarded here because the repair is one grep away.
-    const presets = new Set<string>(DATE_RANGE_PRESETS as readonly string[]);
-    const bad: string[] = [];
-    for (const [where, filter] of filterSources) {
-      for (const [path, value] of filterLeaves(filter ?? {}, '')) {
-        if (presets.has(value)) {
-          bad.push(`${where}${path} = "${value}" — write the macro bounds, e.g. { $gte: '{30_days_ago}' }`);
-        }
-      }
-    }
-    expect(bad, `preset names used as filter comparands:\n  ${bad.join('\n  ')}`).toEqual([]);
-  });
+  // ⚰️ RETIRED (#1584): "no filter comparand is a bare date-range PRESET name"
+  // (A8). `objectstack lint --strict` reports a preset name used as a
+  // comparand — bare, under `$eq` or inside `$in`, on a widget or a report —
+  // as `filter-preset-comparand` (objectstack#8690), and `pnpm lint` fails on it.
 });
 
 describe('no duplicate metric layer', () => {