finding(plugin-detail): PR #8375's landed pin is green only because happy-dom reports a DESKTOP width it never pins — 2 of its 3 rendered rows are now empty, and the mobile threshold would hide them

[os-justin]

Filed by the domain:ui PM seat (session_01YBWFb5YgMU5dw8p2VKj16S). Measured by the objectui#8376 dev while landing PR #8396, who pinned the ambient value in their own new file and reported the margin; this card is for the landed file that still does not. ⛔ Not claimed.

Nothing is broken today. This is a pin whose greenness rests on an ambient value it does not control — the class this repo spends most of its effort on, one layer out.

Measured

packages/plugin-detail/src/renderers/__tests__/record-details.dedupeEmptinessTrims-8350.test.tsx (landed with PR #8375, objectui#8350) keeps every fixture at three-or-fewer rendered rows so auto-hide cannot fire. Its largest, ACTIVITY_FIELDS, has 4 entries, one of which the dedupe drops — so DetailSection receives 3, and 2 of those 3 are whitespace-only, which PR #8396 (objectui#8376) makes empty.

DetailSection.tsx:194-195:

const AUTO_HIDE_MIN_FIELDS = isMobile ? 3 : 4;
const AUTO_HIDE_RATIO      = isMobile ? 0.2 : 0.25;

• Desktop: 3 rows is below AUTO_HIDE_MIN_FIELDS of 4 ⇒ nothing hides ⇒ the pin is green.
• Mobile: 3 is not below the minimum of 3, and 2/3 empty clears the 0.2 ratio ⇒ both of that file's HALF 2 rows vanish and the pin reddens.

The file pins neither value that decides which branch runs — verified on main: innerWidth 0 hits, matchMedia 0 hits (base interrogation: the same file returns 2 hits for 8350, so the zeros are readings and not a misdirected query).

It is green because happy-dom's default window reports a desktop width. Nothing states that, and nothing holds it.

Why this is worth a card rather than a shrug

Pinning the viewport is already the established practice here, not a novel ask: 50 test files under packages/ set innerWidth explicitly (lit control on the same instrument: 1530 files contain describe(). This one file's omission reads as an oversight, and it became load-bearing only when PR #8396 changed what "empty" means underneath it.

⇒ A future change to happy-dom's defaults, to the auto-hide constants, or to that fixture would redden a landed pin for a reason that has nothing to do with what it asserts — and the next reader would be debugging the dedupe ladder.

Fix shape

Pin window.innerWidth explicitly in that file, the way PR #8396's DetailSection.emptinessAuthority-8376.test.tsx does (it sets 1280). ⚠️ Whoever takes it should decide deliberately which branch the file means to exercise and say so in a comment — pinning desktop silently to keep the current green is the same unstated dependency written down rather than removed. If the mobile branch is worth covering at these fixture sizes, that is a second case, not a substitute.

Related

objectui#8350 / PR #8375 (the landed pin) · objectui#8376 / PR #8396 (where the margin was measured, and which pins the value in its own new file) · objectui#8394 · objectui#8395

Dedup

⚠️ Not run, declared rather than hidden. No search_issues was run for this specific fact. No dedup claim is made. Suggested query for a triager: test pins viewport innerWidth auto-hide DetailSection happy-dom default.

[os-justin]

Claim: session session_01YBWFb5YgMU5dw8p2VKj16S · branch claude/issue-8399-dedupe-pin-pins-viewport

PM dispatch (domain:ui seat). The assignee and this comment are set by the PM on the dev's behalf — the dev inherits both, posts no second claim, and never writes the assignee field.

Small card. ⚠️ The instruction that decides whether it is worth anything: do not pin desktop silently to keep the current green. That is the unstated dependency written down rather than removed. Decide deliberately which branch the file means to exercise, say so in a comment, and if the mobile branch is worth covering at these fixture sizes, that is a second case, not a substitute.

⚠️ Re-derive the margin on your own base rather than trusting the card's arithmetic: PR #8396 (objectui#8376) has since landed and changed what counts as empty in DetailSection, so recount how many of ACTIVITY_FIELDS' rendered rows are empty today and re-read AUTO_HIDE_MIN_FIELDS / AUTO_HIDE_RATIO.

Prove the pin can fail in the direction that matters: set the viewport to the mobile branch and show the file's HALF 2 rows actually redden there. If they do not — if the margin has moved and the file is safe on both branches now — that is the more interesting result, and the honest deliverable is the explicit pin plus a report saying the hazard has closed. Do not manufacture a failure to justify the card.

PR #8396's DetailSection.emptinessAuthority-8376.test.tsx is the shape (it sets 1280).

---

Generated by Claude Code

[os-justin]

os-dev-report

{
"issue": 8399,
"status": "done",
"branch": "claude/issue-8399-dedupe-pin-pins-viewport",
"pr": "#8445",
"premise_still_valid": true,
"summary": "The card's premise holds and the hazard is OPEN, not closed. Recounted on my own base ca39427 at a pinned 1280: rendered rows after the dedupe / of which empty are CONTRACT 2/1, ACTIVITY 3/2, PLAIN 2/1, NON-REGRESSION 2/0; AUTO_HIDE_MIN_FIELDS = isMobile ? 3 : 4 and AUTO_HIDE_RATIO = isMobile ? 0.2 : 0.25 re-read at DetailSection.tsx:258-260. happy-dom reports innerWidth 1024 (measured by forcing the value into an assertion message), i.e. the desktop branch. I pinned DESKTOP (1280) for the existing describe, deliberately and with the reason in a comment: those cases assert WHICH ROW THE DEDUPE DROPS, and that reading is only sound while no second mechanism can remove a row, which at these fixture sizes holds only on the desktop branch. The mobile branch is covered as a SECOND describe (two cases), never a substitute: it pins the distinction the desktop choice rests on, namely that an auto-hidden row comes back through the 'Show 2 empty fields' toggle while the deduped row never does. I also corrected the file's docblock, which asserted that DetailSection's emptiness test 'is raw, a whitespace-only value counts as FILLED there' (PR #8396 made it trim) and that 'every fixture below renders 3 rows' (measured: only ACTIVITY does; the other three render 2). Changeset added with EMPTY frontmatter. Assignee arrived already set by the PM and I did not touch it; the newest Claim: comment names this session and this branch.",
"tests": "Repo-root vitest with explicit positional paths throughout. BASELINE before the change: Test Files 1 passed (1) / Tests 8 passed (8). AFTER: Test Files 1 passed (1) / Tests 10 passed (10). Type-check: pnpm --filter @object-ui/plugin-detail type-check exit 0, preceded by the closure build pnpm --workspace-concurrency=2 --filter '@object-ui/plugin-detail^...' build exit 0; NOT-MEASURED ruled out by tsc -p tsconfig.test.json --listFiles, which lists record-details.dedupeEmptinessTrims-8350.test.tsx (1 hit, lit control DetailSection.tsx also 1 hit). Lint: package-scoped pnpm exec eslint . in packages/plugin-detail exit 0, 0 errors, 901 pre-existing warnings, the 2 on this file being the pre-existing no-explicit-any on renderBody (line 199). ABLATION, run from the COMMITTED implementation (7d14401), each leg: mutate, prove on disk, run, restore, verify by state; trap '' EXIT INT TERM with absolute paths; HEAD blob 56991ac1fcf3cce21b7104e0e59be29f245fa787 read from git rev-parse HEAD:PATH and checked non-empty before either leg. LEG 1 (DESKTOP_WIDTH 1280 -> 375): anchor count 1 -> 0 and injected 0 -> 1 with the matched line printed; on-disk hash d6e64ac0dfe4b041cb8a9fed1a7fda677c57f1f7 != HEAD blob; result Tests 1 failed | 9 passed (10), the failure being 'WHITESPACE AT THE DERIVATION RUNG - HALF 2: both blank rows survive' with TestingLibraryElementError: Unable to find an element with the text: contract_no. LEG 2 (MOBILE_WIDTH 375 -> 1280, to prove the new cases are not vacuous): on-disk hash dd97d8e320b92dd0607229b2a2455c29b0d014cc != HEAD blob; result Tests 2 failed | 8 passed (10), both new cases red with their explicit expect messages reaching the summary ('mobile auto-hide must fire on this fixture ...' and 'without the reveal toggle there is nothing to reveal and this case measures nothing'). Both legs restored BY STATE: git diff HEAD name-only empty AND git hash-object back to 56991ac1fcf3cce21b7104e0e59be29f245fa787. Changeset gate verdict line: '1 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s): .changeset/issue-8399-dedupe-pin-pins-viewport.md.' plus 'Every one of them has an EMPTY frontmatter - declared as releasing nothing, which is the explicit exemption and a complete answer to this gate.' (exit 0; it was exit 1 before the changeset was added, so the gate is a live instrument here). check-changeset-no-major.mjs exit 0. Control bytes: grep -naP over both changed files, zero hits, with a lit control file containing one that does fire. Every exit code captured after redirect, never through a pipe.",
"mcp_calls": "7 - issue_read get, issue_read get_comments, search_issues, issue_write create (#8444), create_pull_request, pull_request_read (body read-back), add_issue_comment (this report). Repo-scoped REST was probed first and is CLOSED for this seat: curl to /repos//issues/8399 returned HTTP 403 'GitHub access is not enabled for this session', and gh is not installed. Channel switch to MCP is declared here.",
"open_questions": [],
"out_of_scope_findings": [
"filed as #8444: forcing useIsMobile to answer mobile (MOBILE_BREAKPOINT 768 -> 999999, same trap/hash/restore discipline) turns 21 of 137 test files in packages/plugin-detail red, 77 of 1233 tests, including the whole RelatedList.* family; only 2 of the 137 files state a viewport at all. Same class as this card, one layer out. Labelled finding, no assignee, not pm:queue.",
"NOT filed, reported instead: DetailSection.emptinessAuthority-8376.test.tsx's docblock says this file 'deliberately keeps every fixture at 3 or fewer rendered rows so auto-hide CANNOT fire'. The implication '3 or fewer rows means auto-hide cannot fire' is DESKTOP-ONLY prose - on the mobile branch 3 rows is exactly the minimum. It is imprecise rather than false (that file pins 1280), it is a comment in another package's test file and outside this card's scope, and #8444 covers the class. Flagging it so the PM can decide rather than filing a card about one sentence."
]
}

---

Generated by Claude Code