Finding: after objectui#8376 four MORE un-trimmed spellings of emptiness remain in plugin-detail — the highlight strip and the H1 summary chips still paint a blank for a whitespace-only value

[os-justin]

Filed by the objectui#8376 dev (branch claude/issue-8376-detailsection-emptiness-authority) while converging DetailSection's three spellings onto one. ⛔ Not claimed. Out of scope there: #8376's fence is DetailSection.tsx, and each site below is a different component with a different consumer and would need its own pins.

What

objectui#8350 fixed the record:details dedupe ladder. objectui#8376 fixed DetailSection's three (isEmptyValue, the em-dash isEmpty branch, canCopy). A census of packages/plugin-detail/src on that branch finds four more raw value === null || value === undefined || value === '' tests, none of which trims, all on the same record-detail page:

site	what it decides	consequence of a whitespace-only value
HeaderHighlight.tsx:171	isEmpty for a chip in the ADR-0085 highlight strip	the strip paints a blank chip instead of the em-dash — the same defect #8376 fixed one band lower on the same screen
DetailView.tsx:979	whether a summaryFields chip renders beside the H1	a blank chip is rendered rather than skipped
RecordMetaFooter.tsx:82	whether created_by / updated_by renders	a blank actor slot
HistoryTimeline.tsx:134	the '—' placeholder for an audit value	a blank cell where the timeline means "nothing"

The first two are the ones that reach a reader: they sit directly under / beside the page H1, whose own definition (@object-ui/core's recordDisplayValueAt) calls that value empty. So on one screen the H1 says the field is empty, the body grid now says empty (#8376), and the strip between them says filled and paints nothing.

Fix shape (a suggestion, not a ruling)

⚠️ #8376 measured that the answer is not "delegate to recordDisplayValueAt" wholesale. That function answers does this resolve to a NAME, so an object value runs through displayNameOfEmbeddedObject and is EMPTY when that yields nothing — right for a title, wrong for a CELL, where { latitude, longitude } / { street, city, … } / ['alpha','beta'] all render real content through a type-aware cell renderer and carry no name-ish key. #8376's hasCellValue is the shape that came out of that measurement: delegate the scalar half to the authority, keep objects as values.

So the likely move here is to lift hasCellValue out of DetailSection.tsx into a small shared module in this package and have these four read it. A taker should decide per site whether the cell question or the title question is the right one — RecordMetaFooter's UserRef in particular is about a reference, which may genuinely want the title predicate.

Whatever the ruling, the pin has to assert the RENDERED outcome (which chip shows the em-dash, which chip is absent), not the predicate in isolation, and needs a non-regression case for object-valued fields — #8376's ablation showed a wholesale delegation is green on every whitespace case and red only on that one.

Measured

Read on branch claude/issue-8376-detailsection-emptiness-authority at b6ed5cdaf:

$ grep -rn "=== null || value === undefined || value === ''" packages/plugin-detail/src --include=*.tsx | grep -v __tests__
packages/plugin-detail/src/HeaderHighlight.tsx:171
packages/plugin-detail/src/RecordMetaFooter.tsx:82
packages/plugin-detail/src/HistoryTimeline.tsx:134

plus DetailView.tsx:979, which spells the same test over a local named val.

Dedup

search_issues once, self-validating (it returned the two cards that ARE about this family): the highlight-strip / whitespace / blank-em-dash wording returned objectui#8376, objectui#8350, objectui#4054 and objectui#8279. #4054 is chip clipping in record:highlights, #8279 is a designer key mismatch — neither names an emptiness predicate. None names these four.

Related

objectui#8376 (the same divergence in DetailSection, fixed) · objectui#8350 (the dedupe ladder, fixed) · ADR-0079 (recordDisplayValueAt, the title definition) · ADR-0085 (highlightFields, the strip's charter)

[os-justin]

Claim: session session_01YBWFb5YgMU5dw8p2VKj16S · branch claude/issue-8394-remaining-emptiness-spellings

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.

Unblocked: objectui#8376 / PR #8396 has landed (verified by content — hasCellValue now appears 5× in DetailSection.tsx), so the shared authority exists and DetailSection's three spellings are already converged.

⛔ Do not repeat the mistake PR #8396's dispatch nearly caused. My brief there ruled "delegate to recordDisplayValueAt wholesale"; measurement proved that wrong at a cell, and PR #8396 shipped a scalar/object split instead. recordDisplayValueAt answers "does this resolve to a NAME" — an object runs through displayNameOfEmbeddedObject and is EMPTY when that yields nothing, while a cell hands objects to type-aware renderers (location, address, select badges, JSON, repeater), none of which carries a name-ish key. Read hasCellValue's docblock in DetailSection.tsx before converging anything.

⚠️ Each of the four sites may be asking a different question. HeaderHighlight.tsx (the ADR-0085 strip), DetailView.tsx's summaryFields chips, RecordMetaFooter.tsx and HistoryTimeline.tsx are four surfaces, not one. Converge only where the question really is the same, and say so per site. Where it is not, say what it is instead — a report naming three convergences and one genuine difference is a better result than four forced ones.

The card notes HeaderHighlight.tsx:171 will be, after PR #8396, the only band on the page still calling a whitespace-only value FILLED — it sits between the H1 and the body grid, so the inconsistency is visible in one screenful.

---

Generated by Claude Code

[os-justin]

os-dev-report

{
"issue": 8394,
"status": "done",
"branch": "claude/issue-8394-remaining-emptiness-spellings",
"pr": "#8457",
"premise_still_valid": true,
"summary": "Premise re-derived on origin/main at ca39427: hasCellValue appears 5x in DetailSection.tsx (PR #8396 landed) and all four line numbers the card names are exact. hasCellValue was lifted out of DetailSection.tsx into packages/plugin-detail/src/emptiness.ts and the remaining bands of the record page now read it. Three of the four sites converged outright (HeaderHighlight, HistoryTimeline, DetailView's chip render). RecordMetaFooter is genuinely different and the card's line is NOT where the fix belongs. A FIFTH raw spelling the card's census could not see (DetailView.tsx:438) had to be fixed too, or the fix at the line the card names would have turned a blank chip into a missing chip.",
"per_site_verdict": {
"HeaderHighlight.tsx:171": "CONVERGED — identical question. isEmpty chooses between the em-dash affordance and CellRenderer, the same two branches DetailSection's displayValue chooses between, same detail.noValue label, objects reach the type-aware renderer identically. Now !hasCellValue(value).",
"HistoryTimeline.tsx:134": "CONVERGED — identical question. formatDiffValue chooses between the em-dash and content, and its object branch is JSON.stringify, i.e. real content, so the object half must not delegate for exactly the same reason it must not in a cell.",
"DetailView.tsx:979": "CONVERGED, with a caveat reported below. The chip is drawn by a hand-rolled stringifier (String(val) plus currency/date/percent/select branches), not by a cell renderer. hasCellValue is therefore the NON-REGRESSIVE convergence: it moves whitespace-only strings and nothing else. Delegating to the title predicate would have silently changed which object-valued chips render, which is not what the card asked for and was not measured.",
"DetailView.tsx:438": "NOT IN THE CARD — a fifth spelling, and load-bearing. autoSummaryFields' has() asks the same question one rung earlier, spelled over data?.[n] with the operands reversed, which is why the card's grep pattern missed it. It decides WHICH field becomes a chip. Raw, a whitespace-only status satisfied it and won the single status slot, and a render-only fix then dropped that chip, so the header showed NO status chip where a genuinely filled stage should have taken the slot. Fixed under the bounded in-place exemption: same defect class, same file, mechanical, no new verification surface.",
"RecordMetaFooter.tsx:82": "GENUINELY DIFFERENT — the fix is NOT at that line. Four consumers ask 'is there an actor?' about one value: hasCreated/hasUpdated, the sameUser suppression, the choice between the detail.createdBy and the by-less detail.created label, and MetaEntry's own truthiness gate that mounts UserRef. Only the last reaches UserRef, so line 82 is UNREACHABLE for null/undefined/'' in the first place. Converging it alone removes the blank and leaves 'Created by' plus the middot separator standing over an actor that is not there, which is precisely the dangling phrase that label branch's own comment says it prevents. The footer is normalized at the READ instead (actorOrNone over both audit reads), so all four agree; line 82 stays as a converged defensive floor."
},
"deliberately_not_converged": "ConcurrentUpdateDialog.tsx:65 has the same SHAPE but a different question: it renders the empty string as a pair of quote marks on purpose, because a conflict dialog reports what is STORED, where 'empty string' and 'absent' are facts the reader must be able to tell apart. Converging it would delete information. The reason is written into the shared module's docblock so the next census does not pattern-match its way in.",
"pins_and_whether_a_worse_implementation_would_pass": "packages/plugin-detail/src/tests/detailPage.emptinessAuthority-8394.test.tsx, 11 cases, every one asserting the RENDERED outcome and never the predicate alone; every case carries a control that rendered by value. Presence is read through queryByText into expect(value, message), never getByText. window.innerWidth is pinned to 1280. HeaderHighlight's own affordance is read as [aria-label='No value']:not([data-slot='empty-value']) so it cannot pick up a cell renderer's EmptyValue by accident; the summary chips are read through the per-field aria-label each Badge carries, so WHICH chip is asserted rather than how many. NO — an implementation strictly worse than the bug does not pass: the EMPTY-for-everything ablation reddens 10 of the 11 cases, and the wholesale-delegation ablation reddens 4. The two dedicated non-regression axes (objects still render; 0 and false are still values) are what refuse them.",
"ablations": "Seven legs, each run from the COMMITTED implementation, each mutation proven to reach disk (git hash-object differs from git rev-parse HEAD:PATH, plus a removed-text / injected-text grep count in both directions), each restored BY STATE (hash back to the HEAD blob AND git diff HEAD empty), never by an exit code; trap on EXIT INT TERM with absolute paths; final tree verified clean. No build was needed and none was faked: the pin imports every mutated module by RELATIVE source path (../HeaderHighlight, ../DetailView, ../HistoryTimeline, ../RecordMetaFooter, ../emptiness), so nothing resolves through dist. Legs and the named rows they redden: (1) HeaderHighlight back to raw, 8273834c -> 832fde55, reddens 'AFFORDANCE — a whitespace-only highlight draws the No value em-dash, not a blank chip'. (2) DetailView render back to raw, d838646f -> da6752a2, reddens 'CHIP — a whitespace-only summary field renders no chip, and its neighbour still does'. (3) DetailView PICKER back to raw (this is exactly the render-only fix the card would have produced), d838646f -> 535b9266, reddens 'PICKER AND RENDER AGREE — a whitespace-only status must not consume the auto-detected status slot'. (4) HistoryTimeline back to raw, 6cef6a42 -> 8e4bf786, reddens 'PLACEHOLDER — a whitespace-only audit value reads as the em-dash, not as a blank cell'. (5) RecordMetaFooter renderer-only fix (reads left un-normalized), 4e80a845 -> 4b664024, reddens 'THE BY-LESS LABEL — a whitespace-only created_by is no actor at all'. (6) hasCellValue returns EMPTY for everything, d765171f -> c05896fe, reddens 10 of 11 including every CONTROL and both 0/false cases. (7) hasCellValue delegates the object half wholesale, d765171f -> d118792c, reddens 4: the geolocation/multiselect chips, the object audit value, the bare {id} actor, and the MEASUREMENT case.",
"tests": "pnpm --filter @object-ui/plugin-detail test => 'Test Files 138 passed (138) / Tests 1242 passed (1242)', exit 0 (captured before any pipe). pnpm --filter @object-ui/plugin-detail type-check => exit 0 (tsc --noEmit && tsc -p tsconfig.test.json; note objectui spells it type-check with a hyphen). Package-scoped eslint . => exit 0, 905 pre-existing warnings and 0 errors. Affected packages from TURBO_SCM_BASE=ca3942729 turbo ls --affected (11), every one with a test script run from the REPO ROOT with explicit positional paths: app-shell 649 files / 6261 passed + 1 skipped (1417s wall, three other agents' vitest runs were sharing the box), console 89/1069, plugin-kanban 30/204, plugin-calendar 30/216, plugin-gantt 63/496, runner 4/17, schema-catalog 29/2088 — all green. Gates: check-changeset-presence verdict line '5 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s)'; check-changeset-no-major, check-changeset-fixed, check-changeset-overwrite, check-control-bytes ('scanned 6668 tracked text file(s)'), check-unreferenced-sources ('Every shipped source file in every covered package is reachable') all exit 0. check-governed-queue-guard --test on all eight changed paths => 'NOT GOVERNED'. Control-byte self-scan of the eight changed files: zero hits, with a lit control (a file containing U+000B) proving the grep fires — my first control used a TAB, which that character class deliberately excludes, so it did not fire and the instrument looked broken when it was not.",
"changeset": "minor on @object-ui/plugin-detail (.changeset/8394-record-page-emptiness-authority.md), matching the landed precedent PR #8396 for a record:details render change. No major (the repo forbids it). skip-changeset NOT applied — it is a phantom label here and a real changeset is owed.",
"mcp_calls": "7 — issue_read(get), issue_read(get_comments), search_issues (rate-limited, returned no results), create_pull_request, pull_request_read (full body read-back), add_issue_comment (this one), and one further read-back of this comment. REST was probed first and returns 403 for reads from this container, while git push works.",
"contradicts_this_brief": [
"THE CENSUS SAYS FOUR; THERE ARE FIVE. DetailView.tsx:438 spells the same test with the operands reversed (data?.[n] !== undefined && data?.[n] !== null && data?.[n] !== ''), so neither the card's grep nor its 'over a local named val' footnote catches it. It is not cosmetic: it governs which fields ever reach line 979, and the picker-only ablation reddens a dedicated case. Fixing only the four named lines would have replaced a blank status chip with NO status chip.",
"THE CARD'S GUESS ABOUT UserRef IS WRONG BY MEASUREMENT. The card says RecordMetaFooter's UserRef 'is about a reference, which may genuinely want the title predicate'. It does not: UserRef hands objects to LookupCellRenderer, which resolves them through its own display chain. Measured on this branch — a bare {id:'u1'} actor renders 'u1' on screen, and the title authority calls that payload empty, so delegating would have dropped the actor and fallen the label back to the by-less 'Created'. That is now a pinned non-regression case, and it is red under the wholesale-delegation ablation.",
"THE CARD MIS-ATTRIBUTES WHAT LINE 82 DECIDES. Its table says RecordMetaFooter.tsx:82 decides 'whether created_by / updated_by renders'. MetaEntry already gates the renderer on the raw truthiness of the same value, so line 82 is unreachable for null / undefined / '' and only ever sees truthy values. The blank-actor consequence the card describes cannot be fixed at that line alone.",
"THE BRIEF'S INSTRUMENT RULE IS CONFIRMED BY A REAL FALSE RED, and it is worth recording as a measured instance rather than advice. pnpm --filter @object-ui/example-schema-catalog run test reported 'Test Files 2 failed | 27 passed / Tests 93 failed | 1995 passed', every failure an ENOENT for examples/schema-catalog/apps/site/... — that suite reads process.cwd(), which pnpm's filtered run sets to the package directory. The identical selection run from the repo root is 29/29 files and 2088/2088 tests green. A seat that trusted the filtered run would have reported 93 red tests on another team's package."
],
"open_questions": [],
"out_of_scope_findings": [
"NOT FILED — search_issues is rate-limited ('API rate limit already exceeded for user ID 323634890') and REST reads 403 from this container, so per the dispatch I am reporting it in full for the PM to file rather than filing it unsearched. FINDING 1: the record page's summary chips stringify an object-valued field to the literal text [object Object]. Site: packages/plugin-detail/src/DetailView.tsx, the effectiveSummaryFields.map render beside the page H1 (the chip whose guard this PR converged). Unlike every other band of the page, that chip has NO type-aware renderer: display is String(val) with special cases only for currency, date/datetime, percent and select/status/multiselect. So an object-valued summary field — an expanded lookup payload, a location, an address — renders a Badge reading [object Object] next to the H1, and its aria-label carries the same text. Reachable both ways: schema.summaryFields is author-declared and never filtered by type, and the auto-detection can pick a select field whose stored value is an array. Deliberately NOT fixed in PR #8457: hasCellValue is the non-regressive convergence (it moves whitespace-only strings and nothing else), whereas changing what an object does at this site is a display decision needing its own ruling — either route the chip through getCellRenderer the way HeaderHighlight does, or skip object values. Suggested labels: bug, plugin, domain:ui.",
"NOT FILED — same rate-limit reason, reported in full for the PM. FINDING 2: RelatedList contains TWO emptiness spellings that disagree with each other, in the same component. packages/plugin-detail/src/RelatedList.tsx:913-917 isValueEmpty — used by pruneEmpty to drop all-empty columns — TRIMS: it tests (typeof v === 'string' && v.trim() === ''). packages/plugin-detail/src/RelatedList.tsx:963, the placeholder branch inside makeCell, tests only value === null || value === undefined: it does not test the empty string and does not trim. So one rule says a whitespace-only column is empty and prunes it away entirely, while the other says an individual whitespace-only or empty-string cell in a surviving column is filled and renders it visually blank instead of the muted em-dash that the very same function draws for null. Same family as #8394, but a different component with a different consumer (the related-list grid, not the record body), so it is outside this card's fence and needs its own pins. Suggested labels: bug, plugin, domain:ui, finding."
],
"notes": [
"The claim was verified before any edit: the newest (and only) Claim comment names session_01YBWFb5YgMU5dw8p2VKj16S and branch claude/issue-8394-remaining-emptiness-spellings. No second claim was posted and the assignee field was never written.",
"PR #8457 is a DRAFT for PM contract review. Its body was read back in full after creation: it survived the sanitizer intact, and the platform appended its own bare attribution block below the session-URL footer, as expected. mergeable_state is 'behind' because origin/main advanced from ca39427 to 1706d8b while this ran; the merge queue rebuilds on current main, so no rebase was done and nothing was force-pushed.",
"In-flight overlap declared in the PR body: claude/issue-8395-copyfield-object-values sits on the same base and is expected to touch DetailSection.tsx. The overlap is textual only and far apart — this PR removes hasCellValue plus its docblock from the top of that file and adds one import, while #8395's subject is handleCopyField's String(value) further down.",
"CI is not waited on, per the dispatch contract: the report lands at draft-PR time. Nothing is left running — no dev server, no background monitor, no watcher. The one monitor armed during this run (an until-condition on the app-shell suite) fired once and ended with the process it was watching."
]
}

---

Generated by Claude Code