fix(platform-objects): the es-ES and ja-JP dashboard gap help text says what its source now says - #16190

Fixes #16165

Two leaves. metadataForms.dashboard.fields.gap.helpText in the es-ES and ja-JP
metadata-form bundles were faithful translations of a source sentence that no longer
exists.

All three strings were re-located by text, not by the line numbers the card recorded.
On this branch's base 4e090ecde the card's quotes and its line numbers both still hold:
es-ES:943, ja-JP:943, English unmoved at dashboard.form.ts:25. The zh-CN round found
its line number eight rows stale, so this was checked rather than assumed; here it was not.

Why the wording

The new English does two things on purpose, and each replacement has to do both: drop the
CSS framework unit an app author never chose and cannot act on, and carry the magnitude
that lets the author size a dashboard. Every register call below was measured against the
bundle it lands in, not picked.

es-ES — Espacio entre widgets, en incrementos de 0.25rem (4 = 1rem)

• The grid framing goes. The English stopped saying "Grid gap", so de cuadrícula
  goes with unidades Tailwind. The author is spacing widgets, not addressing a grid.
• widgets, not componentes. The loanword is what this bundle uses for dashboard
  widgets — 7 occurrences (Widgets del panel, Tarjetas y gráficos colocados en la cuadrícula., Filtros aplicados a todos los widgets, Widgets del panel con posición y tamaño). componente also appears in this file, 4 times, and every one of them is the
  page-layout component concept (Regiones de diseño … con componentes, Componente (por id) que escribe esta variable, the Componente label at :1133). Using it here
  would collide with a different noun in the same bundle.
• 0.25rem keeps the ASCII decimal point. It is a CSS length literal the author types,
  not a quantity in prose. Measured in this file: the only decimal numbers are dot-form
  (Decimales (p. ej., 2 para $10.50) at :409) and there are zero comma-decimals, so
  a Spanish decimal comma would be new to this bundle as well as wrong as CSS.
• en incrementos de over en pasos de. This bundle has no precedent for either — 0
  hits for paso, 0 for incremento — so the choice is general technical Spanish register
  rather than a house style. en incrementos de is the standard rendering for a quantised
  scale; en pasos de reads as a calque of the English words.
• (4 = 1rem) in half-width parens. 140 half-width parens in this file against 0
  full-width. The =-inside-a-parenthetical shape is already the bundle's own way of
  mapping an input value to its meaning: (empty = forever) at :314, (one entry = locked) at :829, (blank = all object fields) at :837. Keeping = also preserves
  the source's equation exactly, which is what a numeric conversion needs.

ja-JP — 「ウィジェット間の間隔、0.25rem 刻み（4 = 1rem）」

• The grid framing goes, same reason: 「グリッド」 is dropped and the head clause becomes
  「ウィジェット間の間隔」. 「ウィジェット」 is this bundle's word throughout
  (「グリッドに配置するカードとチャート。」, 「全ウィジェットに適用するフィルター」,
  「位置とサイズを持つダッシュボードウィジェット」), and 「間隔」 is its word for spacing —
  4 occurrences, including this field's own label 「間隔」. 「余白」 appears 0 times.
• 「刻み」, deliberately not 「単位」. 「単位」 is precisely the noun the stale string used
  (「Tailwind 単位」); reusing it would read as swapping the framework name while keeping the
  "unit" framing the English retired. It also means something else in this bundle — the
  per-X granularity sense at :988 「アプリ単位」, :1471 「オブジェクト単位」, :1514
  「フラットな単位」. 「刻み」 is the ordinary Japanese for a quantised step and carries no
  such collision. It appears 0 times in this file today, so it introduces nothing that
  contradicts an established choice.
• 「、」 as the reading comma, mirroring the English comma and keeping the source's
  three-part shape (head clause, step clause, parenthetical equation). 57 occurrences in
  this file, including inside help text: 「一意識別子（snake_case、作成後は変更不可）」 at
  :345, 「snake_case、environment ごとに一意」 at :647.
• Full-width 「（）」. 82 full-width against 59 half-width across the file, and the
  half-width ones sit in code-ish or untranslated strings. Inside the dashboard subtree
  the split is unanimous: 「（既定 12）」, 「（秒）」, 「（title, subtitle, actions）」 are all
  full-width, so the immediate siblings decide it.
• The space in 0.25rem 刻み follows the bundle's ASCII/Japanese boundary spacing:
  「snake_case の一意識別子」, 「グリッド列（既定 12）」, 「BCP-47 タグ」, 「1 つの」.
• No trailing 「。」: helpText entries in this bundle do not carry one (「グリッド列（既定
  12）」, 「自動更新（秒）」), while description entries do.

The register model these two were matched against

The zh-CN sweep of this same subtree (9f57f1e31, 2026-09-03) states in its own commit
message that its replacements "mirror es-ES / ja-JP, both of which are correct at all five
leaves". So for this subtree es-ES and ja-JP are the reference register, not the followers,
and the job here was to keep them that way rather than to import a Chinese solution.

Verified independently rather than taken on that commit's word: all 13 other source-derived
prose leaves of metadataForms.dashboard — 5 section descriptions and 8 further field help
texts — were read against the current English in both locales and every one tracks it.
gap is the only stale leaf. The English inventory the comparison used is
en.metadata-forms.generated.ts:901..966, 14 prose leaves in total.

No objectui overlay exists for either locale. Read at objectui origin/main 859de84:
metadata-form-i18n.ts resolves bundleLocale() to 'zh-CN' for zh and undefined for
everything else, and its own header says the metadata designer ships en and zh and
"the other eight shipped locales render the producer's English", so "this bundle carries
zh-CN only". The zh-CN round matched that overlay's opening deliberately; here there is
nothing to match, and this bundle is the only translated surface these two locales have for
this field. No objectui card is touched.

Four instruments are blind to this defect, and their green is not evidence

Re-run in this branch's worktree, exit codes captured by redirect-then-read, never through
a pipe. BEFORE is the same tree with both leaves reverted to the base, with the mutation
proven on disk by whole-line anchors and the restore proven afterwards by blob equality
against the HEAD blob plus an empty git diff HEAD, all in one shell:

                             BEFORE (stale leaves)     AFTER (this PR)
pnpm check:i18n                  exit 0                    exit 0
  platform-objects               in sync (11 bundle(s))    in sync (11 bundle(s))
  OK (9 package(s) - all bundles in sync, no undeclared authoring keys)

pnpm check:i18n-stale-fill       exit 0                    exit 0
  scanned 10 bundle set(s), 0 stale-fill leaf/leaves, 0 baselined
  OK (10 bundle set(s) - no new stale fills, 0 baselined)

The numbers are identical either side of the change, to the byte. That is the durable
half of this card: a green gate here is not evidence the change was needed, and not
evidence it worked.

1. check:i18n fills gaps only. It compares committed bundles against a fresh
   extraction, and merge fills gaps — a present-but-stale leaf is not a gap, so --write
   rewrites the en leaf and leaves es-ES and ja-JP alone. Regeneration is structurally
   incapable of fixing this.

2. check:i18n-stale-fill infers a stale fill from cross-locale agreement. It catches a
   leaf that is a copy of the English wearing a translation's name. Separación de cuadrícula (unidades Tailwind) and 「グリッド間隔（Tailwind 単位）」 are real
   translations, written by translators, agreeing with no other locale. Invisible to that
   predicate by construction, not by accident.

3. There is no provenance to fall back on, verified per locale. Neither
   es-ES.source-hashes.generated.ts nor ja-JP.source-hashes.generated.ts carries a
   dashboard entry — case-insensitive, real exit code 1, zero matching lines. The control
   fires in the same files: object matches 463 lines in the es-ES table and 454 in the
   ja-JP one. And the miss is not that the table ignores this bundle: both tables do carry
   metadataForms.* paths, 190 each, across 14 sibling subtrees (action, api, book,
   capability, dataset, doc, field, hook, mapping, object, page, report,
   seed, view). dashboard is simply not one of them, and per that file's own header
   "a path with no entry is LEGACY-TRUSTED and never reported stale".

4. The Docs Drift Check cannot anchor the changed symbol. Re-run locally against the
   base, reading the JSON fields rather than the rendered comment
   (node scripts/docs-audit/affected-docs.mjs --json 4e090ecde):

   docs:                []
   anchors:             []
   anchorlessChanges:   []
   overbroadAnchors:    []
   weakAnchorsDropped:  ["dashboard (symbol)"]
   

   The name it could not anchor is dashboard, not gap — the same reading the zh-CN
   round measured. dashboard was derived as an anchor candidate and dropped as a single
   lowercase word; gap was never derived as a candidate at all, since anchors is empty
   and weakAnchorsDropped has exactly one entry. No page documenting this leaf can surface
   in that work list by either route.

Four independent instruments, none of them broken — each answers a different question
correctly, and this defect falls between all four.

The hand pass the drift check could not do

Searched content/docs by hand, plain directory pathspec, fixed strings:

⚠️ The es and ja zeros are honest but they are not backed by a same-language firing
control, and the difference matters. Measured under the same pathspec: dashboard matches
76 files and refreshInterval 2, so the pathspec works; but no Spanish or Japanese text
exists in content/docs at all — ción, á and ñ each match 0 files, and so do ー,
日本, ダッシュボード, する and 、, while the Chinese control 的 matches 4. So the
right statement is the inventory one: there is no Spanish or Japanese documentation prose
in this tree for either stale string to be sitting in, rather than "a search that could have
found it found nothing".

The single English hit is content/docs/references/ui/dashboard.mdx:36 and needs no
change:

| **gap** | `integer` | optional | Space between widgets, in steps of 0.25rem (4 = 1rem) |

It already carries the current English, because that page is generated from
packages/spec/src/ui/dashboard.zod.ts and regenerates with it. The English documentation
surface was never stale; only the translations were. Nothing matched under
content/docs/releases/ for any of the four English patterns, so nothing here goes near
release notes.

Scope

Two leaves, deliberately. columns is unchanged upstream, so Columnas de cuadrícula (predeterminado 12) and 「グリッド列（既定 12）」 stay accurate. The general sweep of this
subtree is objectstack#14931's and is not started here.

On the state of the zh-CN half: PR #16163 was open and unmerged when this branch was
cut and when this body was written, so on main three leaves of this family are stale, not
two — the correction triage posted on the card. This PR still fixes exactly the two it was
dispatched for; the zh-CN leaf belongs to #15511 and is untouched here. There is no
collision to coordinate: #16163's diff is zh-CN.metadata-forms.generated.ts plus its own
changeset, and this branch touches neither file, so the two merge independently even though
all three leaves sit at :943 of their respective files.

Verification

• Gate family derived from the diff, not judged:
  node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack → 47 families, 55
  runnable commands, derived from the tree at f54d139d3 with the change set taken from the
  merge base 4e090ecde (three-dot, including the changeset file). 54 of 55 green, exit
  codes captured by redirect before any pipe. Re-derived after the changeset landed; the
  family did not grow.
• The one non-green is pnpm --filter @objectstack/spec run check:react-declaration-parity,
  exit 1 with MANIFEST is not set — there is no registry side to compare against and its
  own words "This gate did NOT run. That is a failure, not a skip". That is a prerequisite
  refusal, NOT MEASURED locally, not a red about this diff — the manifest comes from an
  objectui browser build that CI supplies, and this diff touches no file in packages/spec.
• Full pnpm run build first: 72 of 72 turbo tasks successful. It matters for what the
  sweeping gates then measured — check:dts-closure reports 72 built packages swept,
  164/164 declared declaration files present, so its green is a statement about a
  population that includes this package rather than about other packages.
• pnpm --filter @objectstack/platform-objects test — 35 files, 528 tests passed.
• pnpm --filter @objectstack/platform-objects typecheck — clean, and the program was
  proven to include the edited files rather than assumed to: tsc --noEmit --listFiles
  lists 419 files and both es-ES.metadata-forms.generated.ts and
  ja-JP.metadata-forms.generated.ts are among them.
• Contract surface, ablated against the base with the package rebuilt on every leg:
  all 22 published declaration files byte-identical, while 12 of the 66 files dist
  publishes did move in the same run — index, metadata-translations/index and plugin
  in both .js and .mjs, plus their six sourcemaps. That firing control is what makes the
  negative a measurement rather than a blind spot. The rebuild is proven to have re-run by
  mtimes advancing on all three legs, and pnpm run build was invoked per package so no
  turbo cache sat on the path. Mutation proven on disk by whole-line anchors and proven to
  have reached dist (the retired unidades Tailwind appears in 12 dist files while
  mutated, in 0 after restore); restore proven by blob equality against the HEAD blob, an
  empty git diff HEAD, and a byte-for-byte re-comparison of all 66 dist files against the
  pre-ablation snapshot.
• Clause-②: no. The published movement is a string-literal value inside the emitted
  esESMetadataForms / jaJPMetadataForms data structures. The declaration side cannot
  move for it by construction: the published subpath declares
  declare const MetadataFormsTranslations: TranslationBundle;, an opaque bundle type, so
  no literal type carries any leaf text — neither locale constant is even named by symbol in
  any of the 22 declaration files. No exported symbol, signature, or accept/reject behaviour
  moves. That reading is declared in the card's claim comment, which is the carrier the
  enqueue gate reads.
• pnpm check:partof-closing-keyword passed in the sweep on an empty PR_BODY, which is
  a vacuous pass by that gate's design — it reads the body from the environment and never
  from the API. It was re-run against this posted body afterwards; that run is the reading.

Changeset rather than the skip-changeset label, re-derived rather than carried over:
@objectstack/platform-objects declares no private flag and publishes dist, and these
strings reach Spanish- and Japanese-reading Studio authors, so the diff does publish
something. patch, matching the sibling sweeps for the same subtree.

Authored by Claude Code session session_01ARYe3yQTQCUFm5qPYNgKaJ — attribution kept in
prose because a body edit downgrades the session-scoped footer link to the bare form.

Generated by Claude Code