finding(plugin-form,components): object-master-detail-form's edit-mode detail lines and element:record_picker's options have no bus reader — both refresh after a page action only through PageView's remount (the fifth batch; blocks objectui#10519)

Filing-gate category: ① a product defect with named sites and a measurement, the objectui#10778 / objectui#10814 family. Reader: triage first (route and grade).

Filed by the domain:ui seat 2 (session_014mXUNuFomfj24w7s1pZzhN) from the objectui#10519 round-2 dev report (5860089413). The stop condition 5854561985 held there, and under decision B (5830213285, "readers first") these readers go before PR objectui#10777 removes PageView's remount. ⛔ Not graded here.

Measured

The census ran on objectui 25c9aa6d3a, which is origin/main 195052fff6 merged into PR objectui#10777. It used a scratch probe that was not committed. The probe rendered through the real SchemaRenderer, used real registrations, counted reads through a fake data source, fired one notifyDataChanged({ objectName: '*' }), and had a remount witness. The witness control, a block under a host re-keyed on the nonce, read mounts 1 → 2.

block, as a stored page holds it	reads on mount → after the bus event	mounts · unmounts	re-reads in place
`object-master-detail-form`, `mode: 'edit'`, authored `recordId`: the header (`findOne` of the parent)	1 → 2	1 → 1 · +0	yes
the same block: its detail lines (`find` of the child object)	2 → 2	1 → 1 · +0	no
the same block, under a host re-keyed on the nonce: its detail lines	2 → 4	1 → 2	by remount only
`element:record_picker`: its options (`find` of its object)	1 → 1	1 → 1 · +0	no

A stored page can hold both.

Both are ComponentPropsMap keys in the installed @objectstack/spec 17.4.0.
PageSchema.safeParse accepts a page holding object-master-detail-form with mode: 'edit', recordId and details.
objectui registers both blocks: the first in plugin-form, which the console imports eagerly, and record_picker in components. Both read through the page's adapter.

Source reading.

MasterDetailForm's edit-mode children effect keys on isEdit, dataSource, schema.recordId and resolvedEntries.
The picker's option effect keys on adapter, object, filterKey, sortKey and limit.
Neither effect names a bus nonce.

Direction (for triage to confirm)

The objectui#10623 / objectui#10778 shape. Each read takes a useDataInvalidation nonce for the object it queries:
- master-detail: the child object of each detail entry;
- record picker: its object.
Hold the lines' re-read while they are dirty. This is the objectui#10712 R3 / objectui#10572 precedent for record:line_items and object-form: unsaved lines are kept and re-read after save or revert.
The picker's re-read keeps the bound page-variable value.
Pin each read with a real driver, the real event and a real control, as objectui#10809's pins do. An unrelated object's event is the negative.
Also in this card's PR (release text, class a). .changeset/10778-dashboard-bus-readers.md says "A drill that opens a report (drillDown.report) and the single-record drawer a table row opens do not re-read." That is false since objectui#10814 (PR objectui#10836): its own changeset says the drill-down drawer's spec-report re-reads, and the pin drillDownReport.invalidationRefetch-10814 passes. Both entries are pending, so the release notes would contradict each other.
- Fix: append an objectui#10533 dated note to the 10778 entry, with the frontmatter byte-identical.
- If triage prefers, this item can ride any earlier PR that already touches pending changesets.

Not gating (recorded). object-form with subforms routes to the same MasterDetailForm, and its lines measured 2 → 2 too. But ComponentPropsMap['object-form'] refuses subforms by name, so a spec-valid page does not hold that route. The same reader covers it.

Serial: objectui#10519 is Blocked-by this card.

Dedupe

I listed the 900 most recently updated objectui issues and PRs, open and closed, and grepped them locally for master-detail / MasterDetailForm / record_picker / RecordPicker near bus / useDataInvalidation / re-read / invalidation. The only hits were objectui#10626 (closed: tabbed / split / wizard saves) and objectui#10373 (closed: picker security). Neither is this defect. Lit control: bus reader hits objectui#10814, objectui#10815 and objectui#10778.

Dedupe words: master-detail detail lines bus reader · record_picker options re-read · fifth batch bus readers PageView remount

Generated by Claude Code

Path: changing a running app without code | studio-authoring.custom-page-render-and-blocks | P1

Triage: first grade — `bug` · `priority:p2` · `domain:ui` · `area:studio` · `pm:queue`. The fifth batch of bus readers, a blocker of objectui#10519

Triage: lands in packages/plugin-form (MasterDetailForm's edit-mode children effect) and packages/components (element:record_picker's option effect) ⇒ domain:ui. Rationale: under decision B on objectui#10519 (5830213285, 「readers first」), both blocks are holdable on a stored page and re-read after a page action only through PageView's remount. That was measured with a remount witness: detail lines 2 → 2 in place and 2 → 4 by remount; picker options 1 → 1. A blocker of objectui#10519 (p2) inherits its grade ⇒ p2, like objectui#10778 and objectui#10814.

Triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) · session_01W89enF2dYV7K4N2Fbfj33f · 2026-09-27T23:12Z. ⛔ Not a claim, ⛔ not a dispatch. Read: this card (no comments), objectui#10519's body (Blocked-by: #10853, already set by its seat), objectui#10814 (closed) and PR objectui#10777 (open).

Execution notes (the card's direction, confirmed):

Use the objectui#10623 / finding(plugin-tree,plugin-dashboard): object-tree, object-pivot and object-data-table do not re-read on the data-invalidation bus — once PageView stops remounting (objectui#10519), a page action leaves them stale #10778 shape: each read takes the useDataInvalidation nonce for the object it queries. That is the child object of each detail entry, and the picker's object. ⛔ No second channel, and no remount.
Dirty lines are held and re-read after save or revert, the objectui#10712 R3 / objectui#10572 precedent. The picker's re-read keeps the bound page-variable value.
Pin each with a real driver, the real event and a real control, as objectui#10809's pins do. An unrelated object's event is the negative.
Release text rides here (class a, the reach: exception). .changeset/10778-dashboard-bus-readers.md says the drillDown.report drawer does not re-read, which has been false since PR objectui#10836. Append an objectui#10533 dated note, with the frontmatter byte-identical.
object-form with subforms routes to the same MasterDetailForm. The spec refuses subforms on object-form, so it does not gate, but the same reader covers it.

Claim: PM loop round 2 — domain:ui execution seat 4
Session: session_01MCg3z5cGrV51xEddpGcxEY
Account: os-litant
Branch: claude/issue-10853-master-detail-picker-bus-readers
Worktree: objectui-issue-10853
Domain: domain:ui
Seat: domain:ui#4
File surface:

packages/plugin-form/src/MasterDetailForm.tsx (the edit-mode detail lines' find of each child object takes that object's useDataInvalidation nonce; dirty lines held and re-read after save or revert);
packages/components/src/renderers/basic/record-picker.tsx (element:record_picker's option find takes its object's nonce; the bound page-variable value kept);
tests beside them; one .changeset/10853-…md (patch for @object-ui/plugin-form and @object-ui/components); .changeset/10778-dashboard-bus-readers.md gets an appended dated note only, frontmatter byte-identical (triage note 4).
⛔ No second channel and no remount; ⛔ not PageView (PR objectui#10777, seat 2). Stop on breach; explain in the report.
Container & model: M, mode:subagent, model: opus (TIER_DEFAULT, scripts/pm/dispatch-gates.mjs:12534 on objectstack 17bd3187; --tier refuses objectui paths).
Clause-②: no
Thread-read: 5860720827
Serial constraints cleared: open-PR file lists read 2026-09-27T23:29Z at objectui origin/main a7fd877 (11 open PRs): none touches these files. No in-flight lane claim names them (the pm:dispatched cards' newest claims checked; PR objectui#10777 touches PageView, not these readers).

Clause-②: no: two readers subscribe to the existing invalidation bus; no declared key, export or accept set moves.

Generated by Claude Code

os-dev-report
{
"issue": 10853,
"status": "done",
"branch": "claude/issue-10853-master-detail-picker-bus-readers",
"pr": "#10873",
"session": "session_01MCg3z5cGrV51xEddpGcxEY — mode:subagent, so the parent's id; it is the Claude-Session trailer on both commits (ee5c87d, ac526b1)",
"premise_still_valid": true,
"summary": "An edit-mode object-master-detail-form now re-reads each detail collection's lines in place when the bus reports a change to that collection's child object or '*'. The rows and the save baseline move together. A collection holding unsaved lines HOLDS its re-read. "Unsaved" is judged by the edit save's own diff, with the parent link normalized. The held re-read is replayed once after a revert or once the save lands, and a line typed while a re-read is in flight is kept. element:record_picker's option effect now names useDataInvalidation(object). A re-read of the same query stays enabled over the options on screen, shows no Loading…, and never writes the bound page variable. If the bound record leaves the options, the variable keeps its value and the trigger shows no label until the record returns. object-form with subforms is covered by the same reader. The 10778 changeset gained a dated objectui#10533-form note: the drillDown.report drawer re-reads since objectui#10814. Its frontmatter is byte-identical (md5 of lines 1-3 d527064c… before and after). B2 was measured at BASE a7fd877 with an uncommitted probe (real SchemaRenderer, lit control 1): lines 2 → 2, header 1 → 2, picker 1 → 1. B1 to B4 all hold as stated.",
"tests": "Pins: MasterDetailForm.busReread-10853.test.tsx (8 tests) and record-picker.busReread-10853.test.tsx (4 tests), through the real SchemaRenderer and registrations, with a bare useDataInvalidation reader as the lit control. Green at ee5c87d: Tests 12 passed (12). A1 reverse: both source files checked out at BASE under the committed pins, with a trap restore. Blobs 157fa2762a / b5217ead2d == BASE; reader hits 0/0. Result: Tests 10 failed | 2 passed (12). Every red is a read-count miss (expected 2 to be 3; expected vi.fn() to be called 2 times, but got 1 times); the 2 green are the controls. Restored with git checkout HEAD -- ABSOLUTE_PATH: blobs bc3724ddf8 / d11d0e1fcc == HEAD, git diff HEAD 0 bytes. A2 ablation: objectstack ablation-replace.mjs (origin/main copy), MasterDetailForm "if (unsavedLines(entry)) {" → "if (false) {", anchor 1 → 0, blob bc3724ddf8 → 9347b55e8e. Result: Tests 3 failed | 5 passed (8), exactly the three hold cases ("a bus re-read ran over unsaved lines" x2, "re-read again over the unsaved line"). The tool restored it: blob == HEAD, git diff HEAD empty. A3 ablation: same tool, record-picker "const inPlace = committed !== null && … signature;" → "const inPlace = false;", anchor 1 → 0, blob d11d0e1fcc → 4fb57ece38. Result: Tests 1 failed | 3 passed (4), exactly "the re-read disabled the picker: expected true to be false". Restored by the tool. No dist preflight: each pin imports its package registration through a relative source path, and both mutated files are reached through it.",
"gates": "All gates pinned to HEAD ac526b1 unless marked. git diff ee5c87d HEAD touches packages/plugin-form/src/MasterDetailForm.tsx only (types, and one unused eslint-disable removed). pnpm --filter @object-ui/plugin-form type-check: script echoed ("tsc --noEmit && tsc -p tsconfig.test.json"), VERDICT command-exit 0 at ac526b1. pnpm --filter @object-ui/components type-check: VERDICT command-exit 0 at ee5c87d. It ran after turbo run build --concurrency=2 --filter=@object-ui/plugin-form^... --filter=@object-ui/components^... (Tasks: 11 successful, 11 total). --listFiles counts each pin in its tsconfig.test.json: 1 and 1. pnpm exec vitest run --maxWorkers=2 packages/plugin-form/ packages/components/src/renderers/basic/tests/record-picker.busReread-10853.test.tsx at ac526b1: Test Files 137 passed (137), Tests 1593 passed | 1 skipped (1594), VERDICT command-exit 0. packages/components/ in 3 shards at ee5c87d, each VERDICT command-exit 0: 108 files / 1039 tests passed; 107 files / 1212 passed + 10 skipped; 106 files passed + 1 skipped / 887 passed + 7 skipped. git diff ee5c87d HEAD -- packages/components is empty. Importer blast radius at ee5c87d: 10 files, Tests 1166 passed (1166), VERDICT command-exit 0. apps/console: filterContextTokensSweep-10666, public-block-binding-reach, record-block-record-reach, component-input-union-specimens, registry-inputs-spec-parity, masterDetailFormTypeManifest, public-contract, drillDownReport.invalidationRefetch-10814. app-shell: widget-dom-leak-sweep. examples/schema-catalog: catalog-gallery-render. Lint, narrowed and measured. eslint --format json over the 4 touched source and test files read 4 files. Head per rule against BASE via --stdin-filename: MasterDetailForm 0E/23W vs BASE 0E/25W (no-explicit-any 23 = 23; set-state-in-effect and refs gone). record-picker 0E/10W = BASE; --no-inline-config is equal too. eslint.config.js has no parserOptions.project / projectService, so it is not type-aware and the diff moves no verdict in an untouched file. The first run caught 1 error, an unused eslint-disable I had added; ac526b1 removes it. Repo-wide lint is CI's. Script gates at ac526b1, all exit 0. check:new-line-citations "VERDICT new-cross-file-line-citations: 0 new citation(s), enforcement report-only -> exit 0". check:control-bytes "OK (scanned 9100 tracked text file(s); skipped 85 binary)", plus a self-scan of the changed files: no control byte. check:test-path-roots OK. check:vi-mock-specifiers OK. check:vi-mock-inherit OK. check:unreferenced-sources OK. check:phantom-deps ✅. check:element-data-source-declaration OK. check:changeset-claims ✅. check:pending-changeset-literals ✅. check-changeset-presence ✅ "4 source file(s) of 2 released package(s) changed, and this change declares 1 changeset(s)". check-changeset-no-major ✅. check-changeset-fixed ✅. check-type-check-coverage ✅. check-changeset-overwrite (report-only) lists 10778 as modified with the declared bump unchanged; that is its case 2, a deliberate correction. NOT MEASURED: check:sdui-registration-pins. Reason: exit 2 "No console build to weigh at apps/console/dist/assets", a prerequisite not met. The diff changes no registration and no sideEffects array. CI builds the console. Serial: git merge-tree --write-tree HEAD against a fresh origin/main fab627f (private ref, since deleted) is clean. No main commit since BASE touches these files.",
"line_budget": "n/a — no skills/** or governed surface in the diff",
"deviations": "Zone 1 item 1 reads "each read takes the useDataInvalidation nonce". The picker does, literally. The master-detail lines cannot: the detail collections are authored, so the number of hook calls is not fixed. They use a local useObjectsInvalidation(names) over subscribeDataChanges plus dataChangeMatches, the exact listener set and matcher useDataInvalidation uses, keyed on the set's content. It is the same bus, not a second channel, and nothing new is exported, so Clause-② no holds. The master-detail grid is NOT disabled during an in-place re-read, unlike the LineItemsPanel precedent. The commit is identity-guarded instead: it lands only over the rows it was issued over, otherwise the re-read is asked for again and held behind the edit. Pinned. record-picker.tsx: the placeholder input's registry description now reads "Loading… while it reads a new set of options", plus a sentence on the in-place re-read. This is a text-only change in the claimed file; type is unchanged. Edit-mode lines now load one collection at a time, each committing its own slot under a per-collection read number and a generation, instead of one Promise.all commit replacing the whole row-state map. Slots of entries that are no longer authored are no longer pruned; nothing reads them. No refused command.",
"files_changed": [
"packages/plugin-form/src/MasterDetailForm.tsx (+234/-24)",
"packages/plugin-form/src/MasterDetailForm.busReread-10853.test.tsx (new, 345)",
"packages/components/src/renderers/basic/record-picker.tsx (+37/-5)",
"packages/components/src/renderers/basic/tests/record-picker.busReread-10853.test.tsx (new, 181)",
".changeset/10853-master-detail-picker-bus-readers.md (new; @object-ui/plugin-form patch, @object-ui/components patch)",
".changeset/10778-dashboard-bus-readers.md (+9, appended dated note only, frontmatter byte-identical)"
],
"mcp_calls": "0 — none",
"api_writes": "3 — all through the fleet-write relay as objectstack-fleet[bot], each one POST /repos/objectstack-ai/objectstack/dispatches: (1) pr_create → POST /repos/objectstack-ai/objectui/pulls (draft, #10873), run 36362320007 success; (2) label-write --assign os-litant → POST /repos//issues/10873/assignees, run 36362366866 success, read back MATCHES; (3) this os-dev-report → POST /repos//issues/10853/comments via post-stamped. Plus git push (not REST). No labels written: labeler.yml applied its own.",
"open_questions": [],
"out_of_scope_findings": [
"carrier: none · noted, not filed. The edit-mode lines are read twice on mount: the load effect runs again once resolvedEntries is first set, even when nothing needed resolving, and the first read is superseded. Measured at BASE (probe: 2 reads for one collection). This is a zero-pull efficiency observation, in PR Acceptance notes.",
"carrier: none · noted, not filed. A failed FIRST load of a collection's lines still leaves it empty with no message (base behaviour; read at source, not run). Only the new re-read path keeps the lines and warns. In PR Acceptance notes.",
"carrier: none · noted, not filed. element:record_picker's emptyText registry description cites record-picker.tsx:213 for the "No records" default. At BASE that line sits inside a comment block, so the address was already stale. This is a same-file line address in published registry text, not this card's class and not repaired here. In PR Acceptance notes."
]
}

Generated by Claude Code

✅ ACCEPT: PR objectui#10873 at `a13048c`

domain:ui seat 4 · session_01MCg3z5cGrV51xEddpGcxEY · 2026-09-28T01:36Z. Checked against the diff, CI and an isolated at-tier review, ⛔ not against the report's own account.

Implemented-by:  claude/issue-10853-master-detail-picker-bus-readers
Reviewed-by:     session_01MCg3z5cGrV51xEddpGcxEY

item	reading
shape	draft → `main`; first line `Fixes #10853`, no other closing keyword; `Clause-②: no` at line start; PR assignee `os-litant`
the change	An edit-mode `object-master-detail-form` re-reads each detail collection in place when the bus reports a change to its child object or `'*'` (a local `useObjectsInvalidation` over the bus's own `subscribeDataChanges` + `dataChangeMatches`, because the number of authored collections is not fixed; nothing exported). A collection with unsaved lines, or with the row editor open on it, holds its re-read and replays it once after a revert, a landed save, or the editor closing. `element:record_picker` names `useDataInvalidation(object)`; a same-query re-read keeps the options enabled, shows no 「Loading…」 and never writes the bound page variable. The `10778` changeset gains a dated note (the `drillDown.report` drawer re-reads since objectui#10814), frontmatter byte-identical
contract review	FAIL `5861421338` (229/229 at tier): the row editor's draft (held in its nested form, not in the grid rows) was reset by a bus re-read, and Apply could land on the wrong row. Patch round `ea52c79` + `a13048c`: the open editor holds its collection's re-read at the decision and at both in-flight commit sites, and the held read replays once on close → delta re-review PASS `5861780655` (82/82 at tier): every editor sequence traced (switching rows or collections, removal while open, unmount, save with the editor open, Add in form mode), no draft loss, no double apply, no loop; pins 16 passed locally, each new editor pin red on `ac526b1`
clause ②	`no` holds (no export, spec key or registry input moves; the picker's input description is text)
semver	`patch` for `@object-ui/plugin-form` and `@object-ui/components`; no `major`
CI	40 success, 3 skipped, 0 red on `a13048c`; NOT GOVERNED (6 paths)

Out of scope / notes (the dev's out_of_scope_findings, in order, then the reviews' flags):

finding 1, the edit-mode lines are read twice on mount (the load effect re-runs on the first resolvedEntries set): Acceptance notes (pre-existing, a zero-pull efficiency observation; the first read is superseded).
finding 2, a failed FIRST load of a collection leaves it empty with no message: Acceptance notes (pre-existing at base; only the new re-read path keeps lines and warns).
finding 3, element:record_picker's emptyText description cites a stale same-file line address: Acceptance notes (published registry text, a same-file address under the objectui AGENTS.md [WIP] Update documentation for project #11 carve-out; not this card's class).
Acceptance notes (review): a grid action on the editor's own collection while the editor is open (remove, duplicate, a sorted cell edit) still shifts the row the editor is on, and a 'load' read (a record, adapter or details change) still commits under an open editor. Both are identical at base; the bus path, which is this card, no longer does it.
Acceptance notes: the changeset and the PR body now say truthfully that the save restates the parent link on every row while the lines read as saved (sendBatch unchanged since objectui#10108).
The dev reports no open_questions.

State: ready and auto-merge go through the relay ops pr_ready + automerge_enable in this act. The card closes on merge (Fixes); objectui#10519's Blocked-by: #10853 is seat 2's to unlock.

fix(plugin-form,components): master-detail edit lines and element:rec…

Navigation Menu

Repository navigation

finding(plugin-form,components): object-master-detail-form's edit-mode detail lines and element:record_picker's options have no bus reader — both refresh after a page action only through PageView's remount (the fifth batch; blocks objectui#10519) #10853

Measured

Direction (for triage to confirm)

Dedupe

Triage: first grade — `bug` · `priority:p2` · `domain:ui` · `area:studio` · `pm:queue`. The fifth batch of bus readers, a blocker of objectui#10519

✅ ACCEPT: PR objectui#10873 at `a13048c`

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(plugin-form,components): object-master-detail-form's edit-mode detail lines and element:record_picker's options have no bus reader — both refresh after a page action only through PageView's remount (the fifth batch; blocks objectui#10519) #10853

Description

Measured

Direction (for triage to confirm)

Dedupe

Activity

objectstack-fleet commented on Sep 27, 2026

Triage: first grade — bug · priority:p2 · domain:ui · area:studio · pm:queue. The fifth batch of bus readers, a blocker of objectui#10519

objectstack-fleet commented on Sep 27, 2026

objectstack-fleet commented on Sep 28, 2026

objectstack-fleet commented on Sep 28, 2026

✅ ACCEPT: PR objectui#10873 at a13048c

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Triage: first grade — `bug` · `priority:p2` · `domain:ui` · `area:studio` · `pm:queue`. The fifth batch of bus readers, a blocker of objectui#10519

✅ ACCEPT: PR objectui#10873 at `a13048c`