Repository navigation
finding(app-shell/plugin-list + spec door): a stored view overlay's top-level options bag is stripped unjudged, persisted verbatim, and rendered on the interface page — options.KIND bypasses its strict schema #10380
Description
Ruled: 5824043998 · letter A (judge options.KIND at the door; persist parsed) · 2026-09-24T23:37Z
Blocked-by: objectstack-ai/objectstack#20051
Filed by the domain:ui execution seat (seat 1, session session_01BA3nKVUwKQJf8DBxrSVtNC) from objectui#10222's dev report (out_of_scope_findings, class b). ⛔ Not graded and not routed: domain:*, priority:*, type, and which repo carries which half are the triage seat's. The seam spans objectstack (the door) and objectui (the consumer).
Fact
A view document written through PUT /api/v1/meta/view as a flattened list overlay can carry a top-level options bag, for example options: { timeline: { metaFields: ['region'] } }. This bag:
- is not judged at the door. The overlay member is
ListViewShapeSchema.extend(flattenedViewOverlayFields()).strip()(objectstackpackages/spec/src/ui/view.zod.ts:5619@44639665).optionsis dropped from the parse output, not refused. The same key written directly astimeline.metaFieldsIS refused (unrecognized_keys); - is persisted anyway.
saveMetaItemkeeps the original request body (「The originalitemis kept verbatim」), so the stripped bag reaches storage. This was read from source and ⛔ not measured at runtime; - is rendered. The console merges the stored overlay verbatim into
objectDef.listViews.InterfaceListPageforwardsview.optionstoListView, which mergesoptions.KINDinto the per-kind node. The object page (app-shellObjectView) does not forwardoptions, so one stored view renders two ways.
The spec's own comment on the overlay member says the .strip() 「exists precisely to carry Studio's auxiliary round-trip keys (isPinned, sortOrder, …)」. options is not an auxiliary round-trip key.
Measured (the objectui#10222 dev, temporary probes, not committed)
- The spec door, with the installed
@objectstack/spec17.4.0: flat overlay +timeline.metaFields⇒success=false. Flat overlay + top-leveloptions.timeline.metaFields⇒success=true, withoptionsabsent from the output. Controls: a ViewItemconfig.optionsand an objectlistViewsentry withoptionsare both refused. - The console:
mergeViewsIntoObjectskeeps"options":{"timeline":{"metaFields":["region"]}}, andInterfaceListPagehandsListViewschema.options = {"timeline":{"metaFields":["region"]}}. Its existing pin 「CONTROL: the legacyoptions.mapbag is still forwarded on its own path」 guards this forward. - The renderer:
ObjectTimelinethen shows theregionchip instead of the defaultstatus/prioritychips.
Seam
spec:VIEW_METADATA_MEMBERS.listOverlay (top-level .strip()) + metadata-protocol:saveMetaItem (verbatim persist) → app-shell:InterfaceListPage (options forward) → plugin-list:ListView (options.KIND merge).
⛔ Not claimed
- ⛔ No fix is chosen: judge or refuse
optionsat the door, persist the parse output, or stop forwardingoptionson the interface page. A door change does not heal rows already stored. - The legacy
options.kanbannesting is a known path (.changeset/8365-stray-kanban-groupby-refused.mdrefusesgroupByin it on thelist-viewroute); how many stored views useoptions.*is not censused. - objectui#10222 (the timeline
metaFieldsretirement) is in the decision box partly because of this path.
Dedupe words
view overlay options bag unjudged · listOverlay strip options persisted verbatim · InterfaceListPage forwards view.options · options.timeline stored view bypass
Activity
objectstack-fleet commentedon Sep 24, 2026
分诊首次定级：
priority:p2·bug·domain:ui·needs-user-decision—— 保存视图时顶层的options包在校验时被悄悄丢掉、却按原样存了下来，接口页再把它渲染出来，于是options.KIND绕过了各视图类型的严格校验；跨两个仓，修在哪里需要维护者定Path: objectstack
packages/spec/src/ui/view.zod.ts（列表覆盖成员ListViewShapeSchema.extend(flattenedViewOverlayFields()).strip()）+saveMetaItem（原样持久化请求体）→ objectuipackages/app-shell/src/views/InterfaceListPage.tsx（把view.options转给ListView）→plugin-listListView（把options.KIND并进对应视图节点）Triage: the objectui consumer half ⇒
domain:ui(the door half is objectstackpackages/spec),bug,priority:p2,needs-user-decision(findingremoved — graded); rationale: a flattened view overlay written throughPUT /api/v1/meta/viewwith a top-leveloptionsbag passes the door because the overlay member.strip()s it (the same key written directly, e.g.timeline.metaFields, is refusedunrecognized_keys),saveMetaItemthen persists the ORIGINAL body verbatim, andInterfaceListPageforwardsview.optionsintoListView, which mergesoptions.KINDinto the per-kind node — so a stored view carries and renders keys the strict schemas refuse, and renders two ways (the object page does not forwardoptions); the repair can sit at the door, the persistence, or the objectui forward, across two repos, and a door fix does not heal stored rows — a contract call for the maintainer.分诊席 #6015，2026-09-24T21:24Z。⛔ 不认领、不派发。本席读完了卡面（本卡尚无评论），并在两个仓的
origin/main上核对。本席核对
view.zod.ts：第 5619 行是ListViewShapeSchema.extend(flattenedViewOverlayFields()).strip()。它上方的注释说这个成员「exists precisely to carry Studio's auxiliary round-trip keys」。InterfaceListPage.tsx：有(view.options as any)?.map ?? …这类读取，确实会把view.options往下转。options；控制台原样合并；ObjectTimeline显示了region标签而不是默认标签。saveMetaItem原样持久化这一条，卡面也注明只是读源码，没有运行时测量。为什么进决策箱
修法有三个落点，分在两个仓：
options.KIND（objectstack spec），和直接写timeline.metaFields一样按名字拒绝；持久化改为存解析后的结果。objectui 继续转发options，因为过了门的内容已经合法。另外做一次存量视图普查，处理已经存下的行。options。 只改一个仓，但会连带去掉options.map这条有意保留、并有测试钉住的旧路径（「CONTROL: the legacyoptions.mapbag is still forwarded on its own path」）。门口的漏洞也还在。options。 最严格，但会让今天还在用options.*旧写法的视图保存失败。有多少视图在用，卡面没有普查。⇒ 不管选哪个，「一个存下来的视图在两个页面上显示成两个样子」都要消除。门口的修复治不了已经存下的行，所以都需要一次存量处理。
定级说明
p2：严格校验被绕过，存下来的视图带着契约会拒绝的键，而且同一个视图在接口页和对象页显示不同。卡面说 objectui#10222（timeline
metaFields的退役）已经因为这条路径进了决策箱。不给 p1：不涉及权限或数据损坏。裁决后的执行要点
domain:spec），本卡挂Blocked-by:到那张卡；objectui 这边只做必要的配合。options.*，按裁决的方向处理。options.KIND里的越界键按名字拒绝（A / C 的情况）。Generated by Claude Code
objectstack-fleet commentedon Sep 24, 2026
Ruling: batch #223 item 5a · letter A (judge
options.KINDat the door; persist the parsed body; census stored rows) · maintainer 「其他同意」 (chat, director seat summon #29,session_01EcrTi7s5oDYPHS4Pi7h31d) 2026-09-24T23:35ZDirector seat, summon #29. Presented together with objectui#10222, with recommendation A (fallbacks B and C); the maintainer agreed. Thread re-read to its last comment (5822497043) in this act.
Ruled: A.
options.KINDis judged by the kind's strict schema, and an out-of-contract key is refused by name, as the direct spelling is;options.*comes before landing.InterfaceListPagekeeps forwardingoptions, because what passes the door is legal. The pinnedoptions.maplegacy path keeps its control.optionsbag) breaks today's legacy writes without a census.metaFieldsreads) is ruled A in the same batch and does not wait on this card.Execution:
needs-user-decision→pm:blockedin this stroke, withBlocked-by: objectstack-ai/objectstack#20051in the body.priority:p2,buganddomain:uistand.Generated by Claude Code
feat(types,plugin-view,app-shell)!: ObjectViewSchema.listViews mirror…
objectstack-fleet commentedon Sep 30, 2026
Triage: the door half landed, and it is stricter than this card's premise.
pm:blocked→pm:on-holduntil it is installableTriage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-09-30T14:59Z. ⛔ Not a claim, ⛔ not a dispatch.Measured. objectstack-ai/objectstack#20051 closed
completed: PR objectstack-ai/objectstack#20868 merged as9905e61ca2(onmain), the last stage of ruling 甲. Read on objectstackmain:VIEW_ITEM_OPTIONS_REFUSED(packages/spec/src/ui/view.zod.ts): a view item record carries no top-leveloptionsbag. It is refused, with the prescription "move eachoptions.KINDblock toconfig.KIND".projectStorableViewBody).5909085121on that card): no census and no migration of existing rows.What that changes for this card. The ruling-A text above says "
InterfaceListPagekeeps forwardingoptions, because what passes the door is legal". Once objectui runs against this door, no newoptionsbag passes it. The objectui half's aim is unchanged: one stored view renders the same on the interface page and on the object page. The claim re-derives how to align against the landed contract (config.KINDas the declared home) and states what it does with an existing row that still carriesoptions. It keeps ruling A'soptions.maplegacy-path control unless it shows why not.Why on-hold, not queued. A cross-repo unlock means installable. npm
@objectstack/speclatestis still 17.5.0, without PR #20868. Aligning objectui before the door ships would change how existing views render against servers that still acceptoptions.Restart-when: npm
@objectstack/speclatestis above 17.5.0 and its published tarball containsVIEW_ITEM_OPTIONS_REFUSED(objectstack PR #20868,9905e61ca2); the objectui bump that installs it has mergedobjectstack-fleet commentedon Oct 3, 2026
pm:on-hold→pm:queue: the restart condition holds ·domain:uiseat 1 ·session_01FjqrwXPfSMkSfkKYDSRkN2· 2026-10-03T17:33Z · standby unlock scan.5913892290: npm@objectstack/speclatestabove 17.5.0, its tarball containingVIEW_ITEM_OPTIONS_REFUSED(objectstack PR #20868), and the objectui bump that installs it merged.latestis 17.6.0 (published 2026-10-02T03:03Z).VIEW_ITEM_OPTIONS_REFUSED, in 11 files.main'spnpm-lock.yamlresolves@objectstack/spec@17.6.0.5913892290, and no cross-reference since.5824043998, letter A. The hold comment notes the door half "is stricter than this card's premise", so the objectui half is re-derived from the shipped door at claim time.Generated by Claude Code
objectstack-fleet commentedon Oct 3, 2026
Claim: PM loop round 2
Session:
session_01FjqrwXPfSMkSfkKYDSRkN2Account:
os-zhuang(the seat's linked user asGET /useranswers it; the card's assignee)Branch:
claude/issue-10380-interface-options-alignWorktree:
objectui-issue-10380Domain:
domain:uiSeat:
domain:ui#1File surface: ruling
5824043998(A) as re-read by triage5913892290against the landed door, at objectuiorigin/mainab18797, as leads:packages/app-shell/src/views/InterfaceListPage.tsx: today it forwards a stored view's top-leveloptionsbag toListView(about:432and:494).packages/plugin-list/src/ListView.tsx: only if the alignment needs it. Itsschema.options.*reads also serve the twin app-shell's object pages emit, so they are not removed wholesale.5913892290): a view item carries no top-leveloptions. The door refuses it with "move eachoptions.KINDblock toconfig.KIND", stores the parsed body, and the maintainer ruled 「20051 不考虑现有的数据」 (no census, no migration).optionsdoes.options.maplegacy-path control stays unless the PR shows why not.The dev stops on a breach and explains it in the report.
Container & model:
M,mode:subagent,model: opus(default judgment tier;dispatch-gates --tier --repo objectstack-ai/objectuiprinted "no path-derived mandate" forInterfaceListPage.tsxandListView.tsx)Clause-②: no
Why no: the door that judges a view already ships in
@objectstack/spec17.6.0, and this card aligns a consumer's rendering to it. No published schema, type, input or accept set moves. If the change turns out to need one, it becomesyesand the PR says so.Thread-read: 5971703185
Serial constraints cleared: read 2026-10-03T17:41Z.
InterfaceListPage.tsxorListView.tsx.packages/types,plugin-kanban) are claimed in this act, with disjoint files.Generated by Claude Code
objectstack-fleet commentedon Oct 3, 2026
os-dev-report
{ "issue": 10380, "status": "needs_decision", "branch": "claude/issue-10380-interface-options-align", "pr": null, "session": "session_01FjqrwXPfSMkSfkKYDSRkN2", "premise_still_valid": true, "summary": "Stopped at the brief's fork rule, with no code and no PR. P1 is FALSE, measured with the spec's own parse on the installed @objectstack/spec 17.6.0. Only the ViewItem record (VIEW_ITEM_OPTIONS_REFUSED) and the form overlay (FORM_OVERLAY_OPTIONS_REFUSED) refuse a top-level `options` bag. The flattened list overlay, which is the card's shape, DECLARES `options` as ListViewOverlayOptionsSchema. It judges each `options.KIND` key by key against that kind's block, keeps it in the parse, and so stores it. The landed changelogs say the same: 6a4aec7 says the bag 'is judged key by key, and objectui pins it', and 9905e61 says 'The flattened list overlay's legacy options bag is unchanged'. So the triage re-read 5913892290's line 'no new options bag passes it' does not hold for this member. The brief quotes that line in its not-re-openable block. I flag the conflict here and do not resolve it. P2 HOLDS for app-shell ObjectView: every per-kind rung reads `viewDef.KIND` only, and a stored row's `options` never reaches ListView from that page. ObjectDataPage is unanchored and renders no stored view, so it is not a party to the divergence. The card's premise still holds on main, re-measured with door-LEGAL rows (see tests). One stored overlay hands the renderer different per-kind config on the two pages: timeline groupByField, timeline titleField, kanban titleField, and the map locationField. The card's own fixture (`options.timeline.metaFields`) is now refused at the door. Its region chip no longer renders, because the chip list was made a constant by the objectui#10222 retirement. That last point was read from source and the render was not measured. Aligning needs either (i) a spec change that retires the list overlay's bag, or (ii) the consumer-side read of `options.KIND` under `KIND` on the object page that the fork rule forbids, or (iii) a consumer narrowing that leaves a declared bag that no renderer honours. That makes it a maintainer call. The four-axis analysis is in open_questions. The branch was pushed empty, equal to origin/main ab18797, as the claim's landing marker. It has no commits.", "tests": "P1, the spec's own parse: a node probe imports the INSTALLED @objectstack/spec 17.6.0 and runs ViewMetadataSchema.safeParse plus diagnoseViewMetadata. Flat list overlay `{object, viewKind:'list', options:{...}}`: options.timeline.startDateField, options.map.locationField, options.kanban.groupByField, options.calendar.startDateField and options.gallery.coverField each give success=true, branch=listOverlay, and the parsed output KEEPS options. A full overlay with type+columns+options.timeline gives success=true. options.timeline.metaFields (the card fixture) gives success=false, unrecognized_keys at options.timeline. options.kanban.groupBy gives unrecognized_keys at options.kanban. options.grid gives unrecognized_keys at options. ViewItem record `{name:'task.board', object, viewKind:'list', config:{...}}` plus top-level options.kanban or options.map gives success=false: invalid_type at options, 'A view item record carries no top-level `options` bag...'. CONTROL: the same record without options gives success=true. Form overlay plus options gives refused at options. The repro rows R2, R3, R4 and R5 below, judged as flat overlays, each give success=true, branch=listOverlay, with options kept. R1 gives refused. The storable projection (projectStorableViewBody, objectstack metadata-protocol on origin/main) stores the parsed value of every authored key the parse kept, so an accepted bag is stored. That was read from source and not measured at runtime. REPRO on main ab18797: a temporary vitest probe in packages/app-shell/src/views, run from the repo root as `pnpm exec vitest run FILE` through os-verify-lock with slot issue-10380-dev. Result: 'Tests 6 passed (6)' and 'os-verify-lock: VERDICT command-exit 0'. The probe was deleted and never committed, and `git status --porcelain` is empty. For each stored flat overlay row in objectDef.listViews, it captures the schema each page hands ListView. It then feeds that schema to the REAL ListView with spy object-timeline and object-kanban renderers, which shows the config the renderer receives. R6 CONTROL (no options, top-level timeline {startDateField, titleField:'name'}): both pages hand the renderer an identical timeline {startDateField:'start_date', titleField:'name'}. R4 (door-legal, same top-level timeline, plus options.timeline.groupByField 'region'): object page has no groupByField, interface page has groupByField 'region'. R2 (options.timeline.titleField 'region'): titleField is 'name' on the object page and 'region' on the interface page. R5 (kanban {groupByField, columns} plus options.kanban.titleField 'region'): titleField is 'name' on the object page and 'region' on the interface page. R3 (options.map {locationField:'location', titleField:'region'}, the path the CONTROL pin guards): the interface page forwards both keys, and the object page's options.map carries only titleField 'name' with no locationField. R1 (the card fixture): metaFields reaches only the interface page's timeline props. No gate was run, because there is no diff.", "mcp_calls": "0", "api_writes": "1: POST /repos/objectstack-ai/objectui/issues/10380/comments (this report, via scripts/pm/post-stamped.mjs). REST reads (not writes): GET issues/10380 and GET issues/10380/comments. git push of the empty branch (not REST). No pr_create and no label-write, because the fork rule stopped the build.", "gates": [], "files_changed": [], "line_budget": "n/a", "deviations": [ "Fork rule taken: P1 was measured false, so no code, no draft PR and no PR-assignee label-write were made. The budgeted git push was spent on the empty claim branch, as rule 1 requires.", "A conflict is flagged here and not resolved. Brief section 1 quotes triage 5913892290 ('Once objectui runs against this door, no new options bag passes it') as not re-openable. The measured 17.6.0 door and the landed changelogs (6a4aec7, 9905e61) both say the flattened list overlay's bag is accepted and judged. I treated that line as a reading of the door, matching the brief's own framing of P1 as a premise to measure, and not as a ruling.", "The card fixture's rendered region chip was not re-measured at render level. The probe used spy renderers. ObjectTimeline's chip list is now the constant status/priority, per its source and the pin ObjectTimeline.metaFieldsRetired-10222.test.tsx." ], "open_questions": [ { "question": "The landed door accepts a judged top-level `options.KIND` bag on the flattened list overlay and stores it. The interface page renders that bag and the object page does not. Which way should one stored view render the same on both pages? Ruling A's text, 'InterfaceListPage keeps forwarding options, because what passes the door is legal', is still literally true of this member. Followed literally, though, it would make the object page read the bag, which is the remap the brief's fork rule forbids building without a ruling.", "options": [ "A: retire the bag at the spec door, then align objectui. On the objectstack spec, the list overlay refuses `options` by name, as the ViewItem record and the form overlay already do, with the prescription 'move each options.KIND block to the top-level KIND block'. There is no migration, per the maintainer's 「20051 不考虑现有的数据」. objectui's half then becomes the brief's lead. InterfaceListPage stops forwarding a stored row's `options` and keeps its own synthesized `options.map` default. The ruling-A `options.map` CONTROL pin is retired together with its premise. Business need: there is no measured first-party producer. No objectui source originates a stored `options` bag: the only `options: {` per-kind literals are the two ListView props the pages build. The console's updateView only carries an existing bag forward. The objectstack census in 6a4aec7 found zero authored bodies. Production sys_metadata rows are NOT MEASURED. Long term: one home per kind block, contract-first, and all three view members agree. AI error: the loudest form, a 422 at save that names the prescription, with no second spelling and no silent no-op. Startup scope: immediate retirement of a legacy spelling is the stated default, and no new gate is added. Cost: two repos in sequence (spec first, a breaking accept-set narrowing, Clause-② yes on that card). It re-opens ruling A's '⛔ C' line, whose census condition the maintainer has since waived. A stored legacy row then renders the way the object page renders, and the next console save of that row (a rename or pin re-saves it) is refused.", "B: the object page honours the judged bag, in objectui only. ObjectView reads `viewDef.options.KIND` as a per-key underlay of `viewDef.KIND`, which is the semantics ListViewOverlayOptionsSchema declares and ListView already applies. InterfaceListPage and the `options.map` control stay unchanged. Business need: it serves no measured producer. Long term: two spellings stay live on a second page, and every per-kind rung must read both. AI error: the door judges the bag, so nothing invalid passes. Still, two legal homes with a per-key precedence is the dialect the spec itself discourages ('prefer the top-level spelling'). Startup scope: it widens a legacy spelling instead of retiring it. This is the consumer-side remap of options.KIND that the fork rule forbids building without a ruling.", "C: objectui alone stops forwarding (the brief's lead), and the spec keeps accepting the bag. Business need: the same as A. Long term: the door keeps a declared accept-set that no renderer honours, which is declared-but-unenforced. AI error: the worst of the three. An AI-written `options.timeline.groupByField` gets a 200 and renders nowhere, a silent no-op trap. Startup scope: the cheapest change now, but it is the consumer-side narrowing that the basic principle forbids ('声明而未兑现是实现缺口,补实现或退役,⛔ 不在消费端收窄'). It also contradicts ruling A's text 'InterfaceListPage keeps forwarding options'." ], "recommendation": "A. It is the only option that leaves neither a declared-but-unhonoured bag nor a two-spelling dialect, and it wins on all four axes. Business need: no producer measured. Long term: one home per block. AI error: a loud refusal beats consumer tolerance. Startup scope: immediate retirement. Its costs are the cross-repo sequencing and an amendment to ruling A, which is why this is the maintainer's call. If the maintainer wants an objectui-only change, choose B over C: B honours what the door declares, and C drops it silently." } ], "out_of_scope_findings": [ "carrier: the ruling on this card's open question (option A's cost line) · noted, not filed. Under 17.6.0, a stored legacy flat overlay whose bag carries a key the door now refuses (e.g. options.timeline.metaFields) is refused on its next console save, because data-objectstack updateView merges the current row and re-saves it on rename or pin. The landed 9905e61 changelog records this behaviour ('refused on its next save'). Not measured at runtime. Dedupe words: view overlay options bag refused next save · updateView round-trip legacy options · stored row options metaFields 422" ] }Generated by Claude Code
objectstack-fleet commentedon Oct 3, 2026
Seat answer to the dev's
needs_decision(report5971873117): ruling A already decides it. The object page honours the judged bag; not escalateddomain:uiseat 1 ·session_01FjqrwXPfSMkSfkKYDSRkN2· 2026-10-03T17:57Z. Checked against the published source, not the report.The dev's P1 finding holds, so triage's re-read
5913892290is corrected for the list overlay. In npm@objectstack/spec17.6.0'ssrc/ui/view.zod.ts:VIEW_ITEM_OPTIONS_REFUSEDguards the view item RECORD only (options: z.never(…), about:5305).FORM_OVERLAY_OPTIONS_REFUSEDguards the form overlay (about:6225).options: ListViewOverlayOptionsSchema.optional()(about:6193). Its docblock (about:5934–:5960) reads:options.KINDis judged by the kind's own strict schema";options.mappath objectui pins)";describe: "the same schema as the top-level block of that kind, which wins per key where both set one".optionsbag passes it" is true of the record and the form overlay, ⛔ not of the list overlay.Why this is not a decision-box item. Ruling A (
5824043998) is the governing text, and it already answers the open question:InterfaceListPagekeeps forwardingoptions, because what passes the door is legal";options.maplegacy path keeps its control".What passes the door is now judged and legal. The interface page keeps forwarding, so the only alignment that keeps every sentence of the ruling is that the object page renders the same bag. That is the dev's option B.
The two other options are outside this seat's power:
optionsbag)". The spec's own docblock records that exclusion. Only the maintainer can re-open it. The dev's four-axis case for it is noted in this seat's round report for the maintainer, ⛔ not acted on.Direction for the build (same claim, same branch):
ObjectView) handsListViewthe stored row's judgedoptionsbag, asInterfaceListPagedoes.ListView's existing per-key merge, which already applies the declared precedence: the top-levelKINDblock wins per key. ⛔ No second merge or remap is written in app-shell.options.mapdefaults) layer the way the interface page'smapCfgdoes. Theoptions.mapcontrol pin stays green.optionscontrol. A per-key collision pin shows the top-level block winning on both pages.5971770388amended here:packages/app-shell/src/views/ObjectView.tsxjoins the surface.InterfaceListPage.tsxmay need no change.Clause-②: nostands: no published schema, type or input moves, and the page starts honouring what the spec already declares and judges.The dev's out-of-scope finding (a stored legacy row whose bag holds a key the door now refuses is refused on its next console save) is landed objectstack behaviour. The changeset of objectstack PR #20868 (
9905e61c) says "Stored rows are not migrated … A row keeps its bytes until its next save", under the maintainer's 「20051 不考虑现有的数据」. The same changeset says "The flattened list overlay's legacyoptionsbag is unchanged: it is judged key by key, and objectui pins it". ⇒ Acceptance notes, not filed.Generated by Claude Code
objectstack-fleet commentedon Oct 3, 2026
os-dev-report
{ "issue": 10380, "status": "done", "branch": "claude/issue-10380-interface-options-align", "pr": "https://github.com/objectstack-ai/objectui/pull/11571", "session": "session_01FjqrwXPfSMkSfkKYDSRkN2", "premise_still_valid": true, "summary": "Built option B under ruling A, as the seat's answer 5971906253 directed. app-shell ObjectView now forwards the stored row's judged legacy `options` bag, read by a new helper, storedLegacyOptions. For each kind the bag carries, the bag's block replaces the page's synthesized block (the mapCfg rule), and the view's own top-level block goes out at the top level. ListView's EXISTING per-key merge (`{ ...options.KIND, ...KIND }`) then lays the declared block over the bag, and no merge is written in app-shell. A row with no bag is relayed exactly as before. InterfaceListPage keeps forwarding `options` and the `options.map` path. Its derived defaults (kanban lane, calendar and timeline date, gallery cover, gantt dates) no longer fill a kind the bag carries, extending mapCfg's `bag.map ?? derived` rule; without this, R2 does not render the same (see deviations). The relay census lost its eight `relayed-nested` rows: the per-kind blocks now have conditional top-level rungs that read viewDef, and the census requires a relayed member to carry no absence row. Changeset: @object-ui/app-shell patch. PR #11571 is a draft on head 44e55cf, assigned to os-zhuang. One named residual: a chart view whose chart block lives only in the bag still differs, because the object page's dedicated ObjectChart route bypasses ListView (see open_questions).", "tests": "New pin `ObjectView.storedOptionsBag-10380.test.tsx`. It renders each stored flat overlay row through both pages, captures the schema each page hands ListView, mounts the REAL ListView on it with spy object-timeline / object-kanban / object-map renderers, and compares what the renderer receives. Rows, mapped to the probe labels: R2 options.timeline.titleField; R3 options.map; R4 options.timeline.groupByField under a declared timeline; R5 options.kanban.titleField under a declared kanban; R6 CONTROL with no bag; two collision rows (timeline titleField, kanban lane) where the top-level block must win on both pages; plus a unit case for storedLegacyOptions. The 17.6.0 door accepts every row: the installed spec's ViewMetadataSchema.safeParse gives success=true, branch=listOverlay. RED ON MAIN FIRST: with the fix committed (4d0668d), ObjectView.tsx and InterfaceListPage.tsx were checked out at ab18797 ('grep -c legacyOptions' read 0 in both files). Result: 'Tests 6 failed | 2 passed (8)'. The four bag rows were RED as predicted, the control and the timeline collision GREEN as predicted, and storedLegacyOptions RED because the export is absent. The kanban collision was RED, which was NOT predicted: the lanes agreed, and it failed on the object page's `titleField: 'name'` floor. Restored with `git checkout HEAD --`; blob hashes matched HEAD and `git diff HEAD` was empty. ABLATION on the fixed tree: the declared block was merged UNDER the bag in app-shell (`{ ...viewDef.KIND, ...legacyOptions.KIND }`, top level, for timeline and kanban). Anchors on disk went 1 to 0 and the injection 0 to 1. Predicted: 2 RED. Observed: 'Tests 3 failed | 5 passed (8)', which is both collisions plus R4, because ListView reads timeline groupByField as a flat key off the top-level block only. Restored and verified the same way. FINAL at 44e55cf: 'pnpm --filter @object-ui/app-shell type-check' exit 0, and its tsconfig.test.json covers the new test (it flagged a TS18046 there on an earlier commit, since fixed). 'pnpm exec vitest run packages/app-shell/src/views/ObjectView packages/app-shell/src/views/InterfaceListPage packages/app-shell/src/views/__tests__/PageView packages/app-shell/src/no-refresh-key-remount.ratchet.test.ts packages/app-shell/src/views/view-filter-fold.ratchet.test.ts' gave 'Test Files 55 passed (55)' and 'Tests 504 passed (504)', with 'os-verify-lock: VERDICT command-exit 0'. Build closure: 'turbo run build --filter=@object-ui/app-shell^... --concurrency=2' gave 'Tasks: 28 successful, 28 total'. NOT MEASURED: the full app-shell suite (979 test files), reason: narrowed to the suites that render or read the two edited modules, because the whole suite exceeds the foreground cap; CI runs it. Lint, narrowed: 'eslint --no-inline-config' on the 4 touched ts/tsx files gave files=4 and 0 errors. The population is eslint.config.js's '**/*.{ts,tsx}' block with tseslint recommended, which has no parserOptions.project and no projectService, so it is not type-aware and untouched files cannot move. Warnings base to head: ObjectView.tsx 171 to 173, InterfaceListPage.tsx 41 to 41, census 0 to 0.", "mcp_calls": "0", "api_writes": "4 across this run, all through the scripts/pm fleet-write relay as objectstack-fleet[bot]: (1) POST /repos/objectstack-ai/objectui/issues/10380/comments, the needs_decision report 5971873117; (2) POST /repos/objectstack-ai/objectui/pulls, pr_create draft #11571, relay run 37144530796, read back 12438 bytes sent and 12438 stored, identical; (3) POST /repos/objectstack-ai/objectui/issues/11571/assignees adding os-zhuang via label-write.mjs, relay run 37144572894, read back MATCHES; (4) POST /repos/objectstack-ai/objectui/issues/10380/comments, this report. git pushes of the branch are not REST. REST reads: issues/10380 and its comments, comments 5971906253 and 5971873117, pulls/11571.", "gates": [ { "command": "pnpm --filter @object-ui/app-shell type-check", "exit": 0, "verdict": "tsc --noEmit && tsc -p tsconfig.test.json, no diagnostics (os-verify-lock VERDICT command-exit 0)" }, { "command": "pnpm exec vitest run (55 ObjectView / InterfaceListPage / PageView / ratchet files)", "exit": 0, "verdict": "Test Files 55 passed (55) · Tests 504 passed (504)" }, { "command": "pnpm exec turbo run build --filter='@object-ui/app-shell^...' --concurrency=2", "exit": 0, "verdict": "Tasks: 28 successful, 28 total" }, { "command": "pnpm check:control-bytes", "exit": 0, "verdict": "check-control-bytes: OK (scanned 10184 tracked text file(s); skipped 85 binary)" }, { "command": "pnpm check:new-line-citations", "exit": 0, "verdict": "VERDICT new-cross-file-line-citations: 0 new citation(s), enforcement report-only, exit 0" }, { "command": "pnpm check:changeset-claims", "exit": 0, "verdict": "report-only: names 7070-no-invented-gantt-date-fields.md and 7218-rowcolor-host-relay.md; both paragraphs re-read, still true" }, { "command": "pnpm check:pending-changeset-literals", "exit": 0, "verdict": "No test source names a pending changeset." }, { "command": "node scripts/check-changeset-presence.mjs", "exit": 0, "verdict": "4 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s)" }, { "command": "node scripts/check-changeset-no-major.mjs", "exit": 0, "verdict": "No changeset declares a major bump." }, { "command": "node scripts/check-test-path-roots.mjs", "exit": 0, "verdict": "check-test-path-roots: OK" }, { "command": "node scripts/check-vi-mock-specifiers.mjs", "exit": 0, "verdict": "check-vi-mock-specifiers: OK" }, { "command": "node scripts/check-vi-mock-inherit.mjs", "exit": 0, "verdict": "check-vi-mock-inherit: OK" }, { "command": "node scripts/check-vi-mock-override-shape.mjs", "exit": 0, "verdict": "check-vi-mock-override-shape: OK" }, { "command": "pnpm check:spec-symbols", "exit": 0, "verdict": "spec symbol derivation: 0 untriaged collisions; spec member citations: nothing cites a key its spec symbol does not declare" }, { "command": "pnpm check:self-import", "exit": 0, "verdict": "No package names itself inside its own src/." }, { "command": "pnpm check:phantom-deps", "exit": 0, "verdict": "Every in-scope import is declared by the package that publishes it." } ], "files_changed": [ "packages/app-shell/src/views/ObjectView.tsx", "packages/app-shell/src/views/InterfaceListPage.tsx", "packages/app-shell/src/views/ObjectView.relayRungCensus-7559.test.ts", "packages/app-shell/src/views/ObjectView.storedOptionsBag-10380.test.tsx", ".changeset/10380-object-page-stored-options-bag.md" ], "line_budget": "n/a", "deviations": [ "InterfaceListPage.tsx changed, though the seat said it 'may need no change'. On main, R2 (a bag timeline with no declared timeline, timeline whitelisted) got a derived top-level startDateField on the interface page and none on the object page. The derived block outranked the bag on every key it set, so R2 could not render the same without it. Its derived defaults now yield to the bag per kind, the rule mapCfg already applies to map. The `options` forward and the `options.map` path are untouched.", "The census ledger (ObjectView.relayRungCensus-7559.test.ts) lost its eight `relayed-nested` rows, with a comment in their place. The new conditional top-level rungs read viewDef, and the census's own check 'declares nothing that IS relayed' requires the rows to go. The nested paths are still asserted by name.", "Observed directions differed from predictions twice, and both are recorded in the test header and the PR: the kanban collision was RED on main (title floor), and the ablation turned 3 cases RED, not 2.", "Residual not aligned: the object page's dedicated chart route (`viewDef.type === 'chart'`) reads `viewDef.chart` only. Aligning it needs a second chart resolution in app-shell, which the direction forbids, or a new plugin-list export. Raised as an open question.", "Commit trailers use AGENTS.md's model-free pair (`Co-authored-by: Claude`), not the harness's model-named line. The PR footer is the os-dev session-URL form." ], "open_questions": [ { "question": "Residual of this card: a `type: 'chart'` row whose chart block lives only in the bag passes the 17.6.0 door (measured: success=true, branch=listOverlay). The object page renders it through its dedicated ObjectChart route, which reads `viewDef.chart` alone, so it still differs from the interface page. Keep the PR's closing `Fixes` line with this as a named residual, or change that line to `Part of` and open a follow-up?", "options": [ "A: keep `Fixes`, and record the residual in Acceptance notes (already done). Business need: no producer of bag-only chart rows is measured; the console's CreateViewDialog writes a top-level `chart`. Long term: there is no second resolution in app-shell. AI error: the gap is visible only to a hand- or AI-written bag-only chart row. Startup scope: nothing is added.", "B: a follow-up exports ListView's `resolveListChartBinding` from @object-ui/plugin-list, and the chart route calls it, so the precedence has one owner. It adds a published name, and its Clause-② needs a look.", "C: a follow-up writes `viewDef.chart || legacyOptions.chart` in the chart route. ⛔ This is the second resolution the direction forbids; listed only for completeness." ], "recommendation": "A. The residual has no measured producer, and the startup-scope axis says not to add a published name for it. B is the right shape if a producer ever appears." } ], "out_of_scope_findings": [ "class: a · reach: named producer: the console's CreateViewDialog creates views with a top-level `tree: { parentField }` / `chart: {...}` block, and `tree` / `chart` are allowed interface-page visualizations · evidence: temporary probe (deleted, not committed) on 44e55cf. A stored row `{ type: 'tree', tree: { parentField: 'parent_id' } }` reaches ListView on the object page as options.tree {parentField:'parent_id', labelField:'name'}. On the interface page (allowedVisualizations ['tree']), viewType is 'tree' and the row has no `tree` and no `options.tree`, so ListView's tree branch gets no parentField. InterfaceListPage's schema literal has no `tree` or `chart` rung (chart read from source). It is independent of the bag, so it was not fixed here · dedupe words: InterfaceListPage tree chart not forwarded · interface page drops declared tree block · interface page chart binding missing · allowedVisualizations tree chart interface page", "carrier: PR #11571 Acceptance notes (per the seat's answer 5971906253) · noted, not filed: a stored legacy flat overlay whose bag holds a key the 17.6.0 door refuses is served as stored until its next console save (rename or pin via data-objectstack updateView), which the door then refuses by name; landed objectstack behaviour (9905e61c)" ] }Generated by Claude Code
objectstack-fleet commentedon Oct 3, 2026
✅ ACCEPT: PR objectui#11571 at head
44e55cf78(dev report5972231377); it lands through the merge queue when every check is greendomain:uiseat 1 ·session_01FjqrwXPfSMkSfkKYDSRkN2· 2026-10-03T18:40Z. Reviewed against GitHub, not the report.main. The first line isFixes #10380, with no other closing keyword.Clause-②: nois at line start, with its reason. 5 files, +510 / −25, no governed path (check-governed-merges: NOT governed). The PR assignee isos-zhuang.Clause-②: nore-checked: the one new export,storedLegacyOptions, is a module export ofviews/ObjectView.tsxbesidekanbanViewOptions. Neither is onpackages/app-shell/src/index.ts. No published schema, type or input moves.patchfor@object-ui/app-shell✓.5971906253, carried out:ObjectViewhandsListViewthe stored row's judgedoptionsbag. For each kind the bag carries, the bag's block replaces this page's synthesized block, and the view's own top-level block goes out at the top level.ListView's EXISTING per-key merge ({ ...options.KIND, ...KIND }) then applies the spec's precedence. No merge is written in app-shell.InterfaceListPagekeeps forwardingoptionsand theoptions.mappath (ruling A).5971770388amended here:InterfaceListPage.tsxchanges, which the seat had said it "may need no change": its derived defaults (kanban lane, calendar and timeline date, gallery cover, gantt dates) no longer fill a kind the bag carries. That extends the page's ownmapCfgrule (view.options.map ?? derived). A derived block placed at the top level would otherwise outrank the row's declaration.ObjectView.relayRungCensus-7559.test.tsloses its eightrelayed-nestedrows, because its own "declares nothing that IS relayed" check requires it.mainthe object page's synthesized block used to fill one. Layering derived < bag < declared would need the second merge the direction forbids. No producer of bag rows is measured: the objectstack census of objectstack PR #20868 found zero authored bodies, and no objectui source originates a stored bag. So the cost is theoretical, and both pages now render the declared metadata the same way.ObjectView.storedOptionsBag-10380.test.tsxmounts the REALListViewfrom both pages with spy renderers, over door-legal rows R2–R5, the no-bag control R6 and two collision rows.titleFieldfloor.type: 'chart'row whose chart block lives only in the bag) ⇒ A, answered by this seat.Fixesstays, and the residual is in Acceptance notes. The object page's dedicated chart route readsviewDef.chartonly. No producer of bag-only chart rows is measured:CreateViewDialogwrites a top-levelchart. A shared export (B) is the right shape if one appears. C (a second resolution) is ⛔.tree/chartrung, so aCreateViewDialogrow with a top-leveltree: { parentField }reachesListView's tree branch with noparentFieldthere. This seat files it as a card.5971906253.Generated by Claude Code
objectstack-fleet commentedon Oct 3, 2026
Landed and verified by content: closed
completedby PR objectui#11571 ·domain:uiseat 1 ·session_01FjqrwXPfSMkSfkKYDSRkN2· 2026-10-03T19:06Z.The merge: PR objectui#11571 went through the merge queue to
mainas0685646at 2026-10-03T19:05Z (seat ACCEPT5972279043).Content check on
main0685646against its first parenta7557a7:mainObjectView.tsx:export function storedLegacyOptionsObjectView.tsx: the...legacyOptions,forward intooptionsInterfaceListPage.tsx:derives('kanban')(derived defaults yield to the bag)InterfaceListPage.tsx: theoptionsforward, kept (ruling A)The pin
ObjectView.storedOptionsBag-10380.test.tsxand the changeset10380-object-page-stored-options-bag.mdare in the tree.No other card closed in that window.
Carried, not on this card:
chartresidual is recorded in the PR's Acceptance notes.tree/chartblocks are filed as objectui#11572.pm:dispatchedis removed in the same act.domain:ui, the priority and the type stay.Generated by Claude Code
Merge main (0a53c67) into claude/issue-11466-node-slot-union-v1 (obje…
fix(app-shell): the object page renders a stored view's judged legacy…