finding(docs): scroll-area.mdx documents a content member neither ScrollAreaSchema declares nor ui:scroll-area reads — and calls the only working key "alternative" #8234

Activity

1. os-justin

   added

   bugSomething isn't workingSomething isn't working

   documentationImprovements or additions to documentationImprovements or additions to documentation

   pm:dispatched

   finding

   priority:p2

   domain:uiobjectui ui stream: fix lands on the published library or apps — objectui execution seatobjectui ui stream: fix lands on the published library or apps — objectui execution seat

   on Sep 7, 2026
2. os-justin

   self-assigned this

   on Sep 7, 2026
3. os-justin commented on Sep 7, 2026

   

   os-justin

   on Sep 7, 2026 – with Claude

   CollaboratorAuthor

   More actions

   Claim: session_01YBWFb5YgMU5dw8p2VKj16S · branch claude/issue-8234-scroll-area-content-rows

   PM dispatch. Assignee and this claim are set by the PM seat for the dev seat; the dev inherits both and posts no second claim.

   Re-verified on today's main 0fa7a9c83 (card measured f57ca757b)

   Both of the card's claims hold:

   • content/docs/components/complex/scroll-area.mdx:39 publishes content?: SchemaNode | SchemaNode[]; // Scrollable content — and ScrollAreaSchema (packages/types/src/layout.ts:505) declares no such member.
   • :40 labels children? as // Alternative content prop, while packages/components/src/renderers/complex/scroll-area.tsx:34 is renderChildren(schema.children) — the single content read in the renderer.

   ⚠️ Anchor drift: the card cites layout.ts:488-507; the interface actually opens at :505. The mdx line numbers :38-40 happen to be exact. ⛔ Re-derive on your own base and match on text — a line number is a hint about where to look, never an instruction about what to change.

   ⭐ TWO MORE WRONG ROWS on the same page, found while verifying — in scope

   The card did not catch these, and they are the same defect class on the same fence:

   mdx	declaration (packages/types/src/layout.ts)
   :32 height?: string;	:510 height?: string | number;
   :33 width?: string;	:514 width?: string | number;

   The page narrows both to string. An author following it would believe height: 200 is invalid when the declaration accepts it. Fold these in — it is one page, one fence, one re-derivation, and splitting them would mean a second PR over the same four lines.

   ⇒ While you are in there, re-derive the whole fence against the declaration, not just the rows named here, and report anything else that disagrees. Two of four rows checked were wrong; assume the rest need checking too.

   Landing control — established before any edit

   content/docs/components/complex/scroll-area.mdx, measured with grep -cF (fixed-string — a regex grep reads the | as alternation; that has cost this session repeatedly):

   leg	now	required after
   content?: SchemaNode	1	0
   Alternative content prop	1	0
   height?: string; (the narrowed row)	1	0
   width?: string;	1	0
   lit control — SchemaNode on any line	2	must stay non-zero

   The judgement left to you

   The card asks whether the page's "Base properties" section (:45-48) is still the right shape. Decide it and argue it in the PR body from the declaration — either answer is fine if reasoned, neither is fine as an unstated default.

   ⚠️ Note :47 publishes visible?: boolean | string | { dialect?: string; source: string } — check that against BaseSchema too while you are deciding.

   ⭐ Precedent to follow — objectui#8197 / PR #8237, landed hours ago

   Same defect, same page family, same fence class. Its dev ruled no row for the inherited optional children on the collapsible page, on the ground that read-ness, not inheritance, is the discriminator — and used the repo's existing shape for an unread-but-inherited key (prose below the fence, as content/docs/components/overlay/tooltip.mdx:24-27 does). ⚠️ Note this page is the opposite case: here children is read, so it is the content channel and belongs in the fence — do not transplant that ruling mechanically.

   That PR also demonstrated the deliverable end-to-end rather than by grep (zod safeParse refusing the old authoring, a real render probe showing the empty body). Do the same here if it is cheap — a scroll-area authored per the current page versus the corrected one. If it is not cheap, say so rather than substituting a proxy.

   ⚠️ No gate will confirm this for you

   check:doc-types reads only type string literals from doc code blocks, so it never judges a member row. check:doc-snippets compiles only ts/tsx/typescript fences (TS_FENCE_LANGUAGES at scripts/check-doc-snippet-types.mjs:548); this block is plaintext, outside its population by construction. ⇒ your own before/after measurement with a firing control IS the verification. Run the doc gates and report their verdicts as what they are, but ⛔ do not offer a green CI as evidence for these rows.

   Scope

   ⛔ objectui#6805 (the scroll-area demo catalog) is closed and is not this card — that covered demos, this is the reference page. ⛔ Do not touch packages/components/src/renderers/complex/scroll-area.tsx; the renderer is correct, the page is wrong. ⛔ Never git stash. ⛔ Never touch content/docs/releases/. Worktree-first.

   ⛔ Draft PR only — the PM flips and lands. Return the standard structured JSON report, and say bluntly if anything above is wrong; it is a reconstruction and this session has had four corrected by devs already, three of which would have caused damage if followed.

   ---

   Generated by Claude Code
4. os-justin commented on Sep 7, 2026

   

   os-justin

   on Sep 7, 2026 – with Claude

   CollaboratorAuthor

   More actions

   os-dev-report

   {
     "issue": 8234,
     "status": "done",
     "branch": "claude/issue-8234-scroll-area-content-rows",
     "pr": "https://github.com/objectstack-ai/objectui/pull/8267",
     "premise_still_valid": true,
     "summary": "The whole `plaintext` Schema fence on content/docs/components/complex/scroll-area.mdx was re-derived member by member against ScrollAreaSchema (packages/types/src/layout.ts:505) and BaseSchema (packages/types/src/base.ts:70-468). FIVE things were wrong, not four: the phantom `content?` row (deleted), `children?` mislabelled 'Alternative content prop' (relabelled as the only content channel), `height?: string` and `width?: string` (both widened to `string | number`), and a FIFTH nobody flagged — `width`'s `(default: '100%')` annotation names a default that does not exist at render time; '100%' lives only in the registration's designer defaultProps (scroll-area.tsx:50) and the renderer never applies it. Measured: omit `width` and no width reaches the style attribute at all. Every remaining row was checked and is correct, so untouched: `type` (layout.ts:506), `orientation` and its `(default: 'vertical')` which IS a real render-time default (scroll-area.tsx:24), `className` (base.ts:171), and the three Base-properties rows. A short prose paragraph was added below the fence in the repo's established shape for this (tooltip.mdx:24-27, collapsible.mdx after PR #8237), stating the content channel and — because it is measured, see `tests` — that nothing rejects the old authoring. One file, +15/-4, no changeset (gate verdict plus landed precedent).",
     "base_properties_ruling": "KEEP IT, UNCHANGED — argued in the PR body from the declaration, on four grounds. (1) Every row is true against the declaration: `id?: string` = base.ts:81; `testId?: string` = base.ts:424; and `visible?: boolean | string | { dialect?: string; source: string }` is the EXACT structural expansion of base.ts:292's `visible?: boolean | ExpressionWire`, since ExpressionWire = `string | { dialect?: string; source: string }` (expression.ts:67). All three visible arms were measured working end to end (boolean false, predicate string, CEL envelope — each renders nothing, innerHTML length 0). The claim comment flagged :47 for checking; it checks out exactly. (2) A curated subset is the right shape: BaseSchema has 21 members, so an exhaustive section would add 18 rows here and on every component page, several inert for a scroll box (bind, data, placeholder, label); deleting the section would remove the only thing explaining why id/visible/testId are absent from ScrollAreaSchema itself. (3) It is not a local invention — the identical three rows with the identical `visible` expansion appear verbatim on 16 component pages, so rewriting it here forks a convention across 16 files. (4) Read-ness (the #8197 discriminator) agrees for two of three: `id` reaches the DOM as id= plus data-obj-id, `visible` gates rendering. `testId` is the one row where read-ness and the declaration disagree — and correcting it HERE would mean making this page disagree with the declaration to route around a defect in another file, the lenient-fallback shape AGENTS.md #0.1 forbids. Filed at the producer instead as #8268.",
     "tests": "DEMONSTRATED END TO END, not by grep. Three throwaway probe files, run from the repo root (`pnpm exec vitest run packages/components/src/renderers/complex/__tests__/zz-probe*.test.tsx`, the canonical invocation the guard demands), all run under the shared verify lock, all VERDICT command-exit 0, all deleted before the commit — nothing from them is in the diff. Readings: (A) zod on the page-as-published authoring — `ScrollAreaSchema.safeParse({type:'scroll-area', content:[...]})` success=TRUE, issues=null, and `content` is PRESERVED in the parsed output. (B) `safeParse({height:200,width:300,children:[...]})` success=true, issues=null. (C) render authored per the OLD page: body text present=FALSE, container.textContent is only Radix's own injected style tag. (D) render authored per the NEW page: body text present=TRUE. (E) height:200 numeric renders `height: 200px; width: 300px` in the style attribute and the body renders; (F) control leg, height:'200px', renders the identical `height: 200px`. (G) through SchemaRenderer the phantom key is not stripped as metadata, so it is spread as a React prop and lands on the element as `content=\"[object Object]\"`. (H) omit width: style carries no width. LANDING CONTROL (grep -cF, fixed-string), before -> after: `content?: SchemaNode` 1 -> 0; `Alternative content prop` 1 -> 0; `height?: string;` 1 -> 0; `width?: string;` 1 -> 0; lit control `SchemaNode` on any line 2 -> 1 (stays lit — the corrected `children?: SchemaNode | SchemaNode[];` row is still there, so the four zeroes are a deletion of the wrong rows, not of the fence). GATES, verdict lines, exit codes captured by redirect-then-capture, never through a pipe: check:doc-fences 0; check:doc-types 0 ('Every documented component type is registered'); check:doc-snippets 0 after the scoped 34-package build ('618 of 618 block(s) judged, 0 failed', its own sentinel control firing) — the BARE run is exit 2 PRECONDITION NOT MET, i.e. NOT MEASURED, not a failure; docs:check-links 0; check:control-bytes 0; check:doc-example-readers 0; check:docs-route-closure 0; check-changeset-presence 0 ('no changeset is owed'); check-governed-queue-guard --test says NOT GOVERNED. Also self-scanned the file for control bytes with `grep -naP` — no hits. NO GATE CONFIRMS THE ROWS and none is offered as evidence for them: check:doc-types reads only `type` string literals; check:doc-snippets compiles only ts/tsx/typescript fences (TS_FENCE_LANGUAGES at check-doc-snippet-types.mjs:548, verified) and this block is plaintext. The A-H measurements above are the verification. No affected-package tests are owed: the diff is one file under content/docs.",
     "brief_corrections": [
       "WRONG, and it would have cost a leg: the brief asks for '#8197's zod safeParse refusing the old authoring'. That does NOT reproduce here, structurally. CollapsibleSchema declares `content` as REQUIRED, so omitting it fails — but ScrollAreaSchema.children is OPTIONAL and the zod BaseSchema is `.passthrough()` (packages/types/src/zod/base.zod.ts:241), so the phantom `content` key VALIDATES and is preserved in the parsed output. Measured, not reasoned. Nothing anywhere rejects the old authoring, which makes this page's row MORE dangerous than #8197's, not less. I reported the true reading rather than hunting for a refusal that cannot exist, and the page now says so in prose.",
       "WRONG: the brief tells me to run `check:internal-doc-links`. No such script exists in this repo. The link gate is `docs:check-links` (scripts/check-doc-links.mjs), which is what I ran (exit 0).",
       "INCOMPLETE: the brief's table has four wrong rows. There is a fifth, on the same `width` line the table already flags — `(default: '100%')` names a render-time default that does not exist. The claim comment's instruction to re-derive the WHOLE fence is what caught it; the table alone would not have.",
       "CORRECT and worth confirming, since both were flagged as uncertain: :47's `visible?: boolean | string | { dialect?: string; source: string }` checks out as an exact expansion of `boolean | ExpressionWire`, and the anchor drift warning is right — ScrollAreaSchema opens at layout.ts:505, height at :510, width at :514, and the mdx line numbers :32/:33/:39/:40 were exact. The warning about the guard on `pnpm --filter PKG exec vitest run FILE` is also correct; the canonical form is `pnpm exec vitest run PATH` from the repo root."
     ],
     "mcp_calls": "8 — issue_read get, issue_read get_comments, search_issues (the one targeted dedup hit), create_pull_request, issue_write create (#8268), pull_request_read (body read-back), add_issue_comment (this), issue_read get_comments (read-back of this). CHANNEL SWITCH DECLARED: the repo-scoped REST read probe returned 403 for this seat (`GET /repos/objectstack-ai/objectui/issues?state=open&per_page=1`), so dedup went through one targeted MCP search_issues rather than a REST list plus local grep. That search returned 11 semantically on-target results including #7088 and #7455, so it is a live reading and needed no control word.",
     "open_questions": [],
     "out_of_scope_findings": [
       "filed as #8268: `BaseSchema.testId`'s JSDoc (base.ts:419-423) promises 'Rendered as data-testid attribute'. Nothing in the repo reads `schema.testId` — the only two hits are schema-builder.ts:84, which WRITES it, and its own test. It is absent from SchemaRenderer's metadata destructure (SchemaRenderer.tsx:1451-1478), so it is spread as a raw React prop and React writes it to the DOM as a lowercase `testid=\"...\"`, which no getByTestId or [data-testid=...] query finds. Measured, with `id` as a contrast leg in the same run (that one reaches the DOM correctly). Exact class-mirror of #7088 on a second key, and the row is published on 16 component doc pages. Two directions written up in the issue; not chosen, it is a contract decision.",
       "NOT FILED, handing to the PM: `BaseSchema` declares TWO content keys — `body?` (base.ts:254, 'Child components or content') and `children?` (base.ts:260, 'Alternative name for children (React-style). Some components use children instead of body'). Each renderer reads exactly one and the declaration does not say which, which is precisely the shape that produced #8234, #8197 and #6939. Measured here: authoring `body` on a `ui:scroll-area` renders NOTHING, identically to the phantom `content` — a second silent empty-box channel on a key BaseSchema does declare. Not filed because my one permitted targeted dedup search was spent on the testId finding and the REST list channel is 403 for this seat, so I could not dedup this one; per the rule I am reporting it rather than filing blind or dropping it.",
       "NOT FILED, observation only, likely already owned: this page's `plaintext` fence is entry `['content/docs/components/complex/scroll-area.mdx', 1]` at scripts/check-doc-fence-languages.mjs:380 — one of the 80 declared files carrying 89 blocks in objectui#5867's SHRINK-ONLY population. Converting it to a `ts` fence would put these member rows under check:doc-snippets and close the gate blindness the card names, but that is #5867's campaign and its ledger, not this card."
     ]
   }

   ---

   Generated by Claude Code
5. os-justin

   closed this as completedin #8267on Sep 7, 2026