Repository navigation
spec: declare an author-settable row ceiling for gallery (and kanban) view configs — the protocol lacks the knob objectui#7390 was ruled to read (principle: 协议不正确的先改协议) #17393
Description
Activity
Claim: PM loop round R45 (wave 4)
Session:session_01JbZnqu8bt6YqfJsr9vaFb3
Branch:claude/issue-17393-view-row-ceiling
Worktree:objectstack-issue-17393
Domain:domain:spec
Seat:domain:spec#2
File surface:packages/spec/src/ui/view.zod.tsand its sibling test(s) first, plus the generated artifacts a regeneration moves and one.changeset/*.md; anything further only where the tree shows the change cannot land without it (stop on breach; explain in the report)
Container & model:M,mode:subagent,model: default judgement tier—node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --tier packages/spec/src/ui/view.zod.tsprinted verbatim: "Model tier — no path-derived mandate: the surface hits none of the 3 declared glob(s)" and "Clause ② SUSPECT surface — a hint, not a verdict", naming that path underpackages/spec/src/**.⚠️ The tool flagged this checkout STALE againstorigin/main; I re-ran it aftergit fetch origin mainand the verdict line was unchanged.
Clause-②: yes
Thread-read: 5690507122
Serial constraints cleared: the named landing filepackages/spec/src/ui/view.zod.tsmeasured FREE across all 19 open PRs (291 distinct held paths), as didpackages/spec/src/ui/view.test.tsandpackages/spec/liveness/view.json. Firing control in the same read:packages/spec/src/data/object.zod.tsreads HELD by #19147. Dark control:packages/spec/src/ui/nonexistent-zzz.tsreads 0.os-verify-lock.sh --statusreadlock is free/queue: empty. Occupancy taken at 2026-09-20T00:01Z; this claim written at 2026-09-20T00:06Z.⚠️ Not cleared, and named rather than hidden: the two GENERATED artifacts this card's regeneration will move —packages/spec/authorable-surface/ui.json(HELD by #19090) andpackages/spec/api-surface-declarations/ui.txt(HELD by #19090 and #19024) — are occupied. That is a landing-order constraint, not a dispatch blocker: generated files are regenerated, never hand-merged, and the relay is the seat's to run (scripts/pm/os-regen-merge.sh, serial hand-off, one PR at a time). Recorded here so the landing window does not rediscover it.⭐ Two line numbers on the card's face are stale and are corrected here rather than left to cost the dev a search: on
origin/main0046a41b43,GalleryConfigSchemais atview.zod.ts:1135(the card says:945) andKanbanConfigSchemaat:1437(the card says:1241). The card's substantive reading — that both arestrictObjects declaring no row ceiling — is re-derived and holds.Clause-②: yesis the card's own declaration (「a new authorable key on a published, strict schema」) and is adopted, not re-derived. The spec lane owes an at-tier contract review every round regardless; a review may overturn the declaration, which is by design.
Generated by Claude Code
补一个读数时刻 —— 认领评论 5746292223 报了树尖却没报「何时读的」
domain:specseat 2,session_01JbZnqu8bt6YqfJsr9vaFb3,2026-09-20T01:07Z。半态巡检 H44 点名本卡,点得对:H44 #17393 a claim (comment
5746292223) states a tree tip with NO time it was taken — the first such fragment is0046a41b43, and noHH:MM[:SS]Zstamp appears anywhere in its paragraph.⛔ 原评论不改写。补记如下。
补记
origin/main=0046a41b43这个读数,取自与那条认领同一动作里的一次git fetch origin main+git rev-parse --short origin/main。⚠️ 本席能诚实给出的是一个上界,不是一个确切分钟:它被读到的时间不早于 2026-09-20T00:00Z(同一动作里那次占用扫描的时戳),不晚于 2026-09-20T00:06Z(平台记录的那条认领的写入时刻)。⛔ 本席不在这个区间里挑一个分钟写死 —— 从记忆里敲一个精确时戳,正是 H44 这条规矩要防的那个缺陷本身。那条认领里引的两个行号(
view.zod.ts:1135/:1437)是在同一棵树、同一动作上git grep出来的,同此上界。
Generated by Claude Code
os-dev-report
{ "issue": 17393, "status": "done", "branch": "claude/issue-17393-view-row-ceiling", "pr": "https://github.com/objectstack-ai/objectstack/pull/19226", "session": "session_01JbZnqu8bt6YqfJsr9vaFb3", "premise_still_valid": true, "summary": "The row ceiling is declared in @objectstack/spec as a shared `limit` member on the three page-shaped view configs — GalleryConfigSchema, KanbanConfigSchema and TimelineConfigSchema (spec does declare a timeline config, so the card's condition for timeline is met) — spelled `z.number().int().positive().default(100)`, with `DEFAULT_VIEW_ROW_LIMIT = 100` exported so a consumer reads the number instead of re-declaring it. SHAPE, the one call the card delegates: per-view-config, NOT a base view-config member, decided on three properties of the tree: (1) the base shape already carries the row-bounding knob every view type reaches, `pagination.pageSize`, and this file's own `virtualScroll` tombstone prescribes `pagination` for that question — a second base-level row bound would have no declared precedence against it; (2) a base member is reachable from every `type`, the non-grid four included, so it would publish an authorable ceiling on four view kinds no renderer reads (declared-but-unenforced), which the card scopes out by name; (3) the per-kind block is what reaches the renderer — measured in objectui at dda8f3815d, ListView's kanban branch spreads the rest of the merged block flat onto the object-kanban node, so a protocol `kanban.limit` lands exactly where ObjectKanban.tsx:573 reads `schema.limit`, while base-level keys are forwarded into no per-kind node. NAME: `limit`, the name the consumers already read, so the declaration absorbs the consumer-local keys instead of buying a second spelling. DEFAULT: applied by the schema rather than only described, and the describe text and the applied value are pinned to each other in a test; the describe also states the visible truncation signal the renderer owes when the ceiling applies (the schema cannot enforce that half, and the pin asserts the sentence is present). The consumer claim was re-measured, not repeated: kanban's `limit` is declared in @object-ui/types alone (packages/types/src/zod/objectql.zod.ts:1762) and read as `$top: schema.limit ?? DEFAULT_KANBAN_LIMIT` (ObjectKanban.tsx:573, constant = 100 at :84); timeline's is on ObjectTimeline's own props interface (:129), on no published schema, read at :328 against DEFAULT_TIMELINE_LIMIT = 100 (:29); the gallery has neither. objectui was read only — no file in that repo was touched. Changeset is minor with `Clause-②: yes (widening)`; the generated artifacts moved by the repo's own regenerators after a real build, never hand-edited.", "tests": "Final head 5dd391125e, all foreground, exit codes captured before any pipe; heavy runs through scripts/pm/os-verify-lock.sh slot issue-17393. (1) `pnpm --filter @objectstack/spec test` — exit 0, 'Test Files 499 passed (499) / Tests 14606 passed (14606)'. (2) targeted verbose run of the six new pins — exit 0, all six green by name. (3) `pnpm --filter @objectstack/spec typecheck` — exit 0 (tsc --noEmit + scripts project + check:test-typecheck: '54 file(s) / 259 error(s) / 144 pinned signature(s)', ledger unmoved). (4) `pnpm --filter @objectstack/spec check:generated` — exit 0, 'All 16 generated artifacts are up to date', after `pnpm --filter @objectstack/spec build` (a real build; OS_SKIP_DTS was never used) and `check:generated --fix` regenerating exactly the four it proved stale. (5) `pnpm check:adr-anchors` — exit 0, 'OK (53 anchored file(s) ... 36673 citation(s) across 4724 file(s) resolve)'; every ADR id in this diff (0122, 0049, 0079, 0087) resolves to a real record, fixtures included. (6) `pnpm lint` — exit 0 over the whole repo (`eslint . --no-inline-config`), so no narrowing claim is needed. (7) Gate battery: the 106 commands `node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack` derives, each run with its exit code recorded before any pipe; `--ran` reconciled to '106 derived famil(ies) accounted for — 103 run, 3 NOT-MEASURED', 0 UNRUN. ABLATION — three legs, each through scripts/ablation-replace.mjs (on-disk anchor counts + blob hash before the run, restore proven as blob == HEAD and `git diff HEAD` empty); view.test.ts imports ./view.zod relatively, so the pins resolve through source and no dist leg applies. LEG A, drop `.int().positive()`: predicted the refusal case reddens alone; observed exactly that — 'refuses a value that could not bound a fetch — and refuses it BY NAME' failed with 'gallery limit=0: expected true to be false', 1 failed / 5 passed, blob 65853f52f3b7 -> 098d5c2e8b2b then restored to 65853f52f3b7. LEG B, delete `limit: rowLimitKey('gallery')`: predicted the acceptance-side cases redden and the scope case holds; observed 5 failed / 1 passed, including 'gallery: expected undefined to be 100' and the parser reporting 'Unrecognized key(s) on this gallery configuration: `limit`'. LEG C, plant the ceiling on GanttConfigSchema: predicted the scope pin reddens naming gantt; observed 1 failed / 5 passed, 'gantt accepts an authorable row ceiling it should not declare: expected [] to include limit'. Leg C exists because A and B never moved the scope pin. TWO HONEST FAILURES ON THE WAY: C's first attempt was VACUOUS — the replacement kept the anchor, ablation-replace refused with 'the anchor count moved 1 -> 1, a drop of 0, not the declared 1', restored and ran nothing; and C's first real run reddened for the wrong reason (no unrecognized_keys issue to stringify, so the assertion complained about argument types instead of about gantt), which is why the pin now asserts on the refused KEY LIST (commit 5dd3911). NOT MEASURED, four families: `pnpm check:dual-build-cjs-loads` exit 3 (reads built output; adapters/apps have no dist here), `pnpm check:lean-entry-closure` exit 3 (loads built entry points; packages/objectql/dist/core.mjs absent), `pnpm check:type-check-debt` exit 3 (--re-measure refuses: 30 workspace dependencies unbuilt), and `pnpm --filter @objectstack/spec check:skill-examples` exit 1 ('packages/client-react/dist holds no .d.ts declarations'), whose exit code is indistinguishable from a finding so `--ran` counted it among the 103 run — the tool's 3 is really 4. All four want a repo-wide build this worktree does not carry; CI builds everything. The three families whose prerequisite WAS bounded were built rather than declared: building @objectstack/lint's 4-package closure turned check:doc-formula-expressions, check:doc-security-posture and check:docs-transcript-drift from exit 3 into exit 0. Six spec gates were re-run after a rebuild when they first refused on a stale dist (the test-file edit moves the build-input hash) and all came back exit 0.", "mcp_calls": "0 — no MCP GitHub tool was called; every GitHub read and write went through the REST proxy with curl.", "api_writes": "2 — POST /repos/objectstack-ai/objectstack/pulls (the draft PR, body written once at creation, read back byte-identical with one footer) and POST /repos/objectstack-ai/objectstack/issues/17393/comments (this report). ZERO label writes on the card or the PR, as the order requires; no PATCH of the PR body; git pushes to the feature branch are not counted here.", "open_questions": [], "out_of_scope_findings": [ "to file (contract violation; dedupe words: prerequisite, exit code, stale dist, api-surface, dispatch-gates): six @objectstack/spec gates refuse a stale `dist` with exit 1 — the code a real finding uses — namely check:api-surface, check:api-surface-declarations, check:exported-any, check:dual-source-exports, check:entry-nameability and check:skill-examples, while this repo declares a distinct exit code for that class and states the reason in scripts/import-prerequisite.mjs: 'Exit 1 from an unmet prerequisite and exit 1 from a real finding are the same reading — which is why the guarded refusal below does NOT keep that number'. Consequence measured on this card: dispatch-gates --ran classifies NOT MEASURED by exit code, so it reported '103 run, 3 NOT-MEASURED' over a record in which four families measured nothing. Successor: the owner of scripts/import-prerequisite.mjs's exit-code vocabulary.", "noted, not filed: the card's two declaration-site line numbers are stale (`:945` / `:1241`); on 0046a41b43 the sites are view.zod.ts:1135 and :1437, as the seat's claim comment already corrected. The card's substantive claim — both are strictObjects declaring no row ceiling — was re-derived and holds. Successor: this PR itself, which lands on the real sites.", "noted, not filed: objectui's timeline row cap is declared on a component props interface only (ObjectTimeline.tsx:129) and on no published schema, unlike kanban's, which at least reaches @object-ui/types. That asymmetry is exactly what objectui#7390 is blocked on and is that card's to close once this releases. Successor: objectui#7390." ], "deviations": [ "ONE FILE OUTSIDE THE DECLARED SURFACE, taken because the tree showed the change cannot land without it: packages/spec/src/type-alias-convention.pin.test.ts. An applied default gives KanbanConfigSchema a second shape, so its ADR-0122 isomorphic pin (Iso829) stops holding — tsc reported it as 'Type false does not satisfy the constraint true' through check:test-typecheck. The prescribed follow-through landed with it: KanbanConfigParsed declared beside the bare alias in view.zod.ts, the pin line removed with its own receipt in the file's own style, and the pinned count plus both prose statements of it moved 785 -> 784. check:spec-parsed-alias is green on the result. GalleryConfig and TimelineConfig needed nothing — both already carried defaults, hence both halves of the alias pair.", "content/docs/references/** (three .mdx files) is outside packages/spec but inside the order's 'the GENERATED artifacts a regeneration moves': it is regenerated by `pnpm --filter @objectstack/spec gen:docs`, never hand-edited, and check:docs is green. Same for the four non-ui api-surface-declarations files, which move only where the view schema is embedded in them.", "NOT MEASURED families are four, not the three dispatch-gates --ran reports — check:skill-examples refuses with exit 1, which the reconciliation cannot classify. Each is named with its exit code and its reason in `tests` and in the PR body.", "The ablation's leg C was vacuous on its first attempt (anchor kept, ablation-replace refused, nothing ran) and reddened for the wrong reason on its second; both are reported rather than quietly re-run, and the pin was sharpened so its red names the view type.", "Commit trailers carry the model-free pair AGENTS.md mandates (Claude-Session + Co-authored-by: Claude), not the harness reminder's model-bearing Co-Authored-By line: the pre-push hook refuses a model identifier in that pair, and AGENTS.md is this repo's source of truth. The pre-push gate confirmed it: 'check:commit-card-trailers: 1 commit message(s) on this push carry no card relation and no model identifier in the trailer pair.'", "CI was not awaited: the report is delivered at the end of local verification, as the standing contract requires. The PR is a draft and stays one — it owes the spec lane's at-tier contract review, which is the seat's to run." ] }
Generated by Claude Code
os-dev-report
{ "issue": 17393, "status": "done", "branch": "claude/issue-17393-view-row-ceiling", "head": "e1ae025756fe1b1d13c2947d9845414ef7ffafc8", "pr": "https://github.com/objectstack-ai/objectstack/pull/19226", "session": "session_01JbZnqu8bt6YqfJsr9vaFb3", "premise_still_valid": true, "summary": "Round 2 — the generated-artifact sync the queue ejection called for, nothing else. `bash scripts/pm/os-regen-merge.sh` ran in a fresh worktree on the branch; its order was followed exactly and no generated file was hand-edited. Step 1 merged origin/main (e233db9dbb) and git auto-committed it (5c753f997a). Step 2 chose a side PER FILE against the pre-merge base 0046a41b43: it TOOK main's side of the four os-regen paths both sides had moved — api-surface/ui.json, api-surface-declarations/ui.txt, authorable-surface/ui.json, export-origins/ui.json — and KEPT the branch's bytes on the seven it alone had moved (the three docs references pages, the four non-ui api-surface-declarations files, authorable-defaults/ui.json). Step 3's commit was refused by the pre-commit hook naming those four as stale, which the script's own header calls that sequence's designed collection point, so the regeneration followed on a committed merge — never in MERGE state, so the authorable-surface anchor could not roll back. Step 4: `pnpm --filter @objectstack/spec build` (real build, OS_SKIP_DTS never used), then gen:export-origins, gen:schema, gen:api-surface, gen:api-surface-declarations in the order the script printed, then `check:generated --fix` which proved exactly one more artifact stale (content/docs/references/ui/component.mdx) and regenerated it. The repair was then committed as e1ae025756 after reading the STAGED diff, not the worktree diff, and the hook cleared its own marker ('os-regen: all deferred artifacts are current — marker cleared'). The silent drop was real and is visible in that staged diff: it restored #19219's ui/ObjectKanbanProps:navigation, ui/ObjectCalendarProps:navigation and the whole ObjectTimelineProps block, which the driver had dropped when it kept this branch's side. Pushed, no rebase, no force. The PR remains a draft: no ready flip, no queue, no auto-merge, no label write, no PR-body edit.", "tests": "All at head e1ae025756, foreground, exit codes captured before any pipe; heavy runs through scripts/pm/os-verify-lock.sh slot issue-17393 (one hold waited 417s behind issue-18639's phase2 — queued, not stuck, and waited out in-turn with a blocking `tail --pid`). BOTH-SIDES ASSERTION, quoted-exact-name greps on the regenerated artifacts: this branch's rows present — 'ui/GalleryConfig:limit' 1, 'ui/KanbanConfig:limit' 1, 'ui/TimelineConfig:limit' 1, 'DEFAULT_VIEW_ROW_LIMIT (const)' 1, 'KanbanConfigParsed (type)' 1, export-origins 'DEFAULT_VIEW_ROW_LIMIT' 1; sibling #19219's entries present — 'ui/ObjectKanbanProps:navigation' 1, 'ui/ObjectCalendarProps:navigation' 1, 'ObjectTimelinePropsSchema (const)' 1; dark control 'ui/GalleryConfig:zzNoSuchKey__' 0. GATES: `pnpm --filter @objectstack/spec check:generated` exit 0 ('All 16 generated artifacts are up to date'); `check:api-surface` exit 0; `check:api-surface-declarations` exit 0; `check:authorable-surface` exit 0; `check:export-origins` exit 0; `check:docs` exit 0; `check:spec-parsed-alias` exit 0; `pnpm check:adr-anchors` exit 0 ('OK (53 anchored file(s) ... 36681 citation(s) across 4724 file(s) resolve)'); `pnpm --filter @objectstack/spec test` exit 0 — 'Test Files 500 passed (500) / Tests 14632 passed (14632)', up from 499/14606 because the merge brought #19219's own suite in, the six row-ceiling pins still green; `pnpm --filter @objectstack/spec typecheck` exit 0 (test-typecheck ledger unmoved at '54 file(s) / 259 error(s) / 144 pinned signature(s)'); `pnpm lint` exit 0 over the whole repo. BATTERY: the derived list is byte-identical to round 1's (106 commands, diff empty) over a change set of 17 paths against the new merge base e233db9db — the merge smuggled nothing into the diff; all 106 re-run at this head, `--ran` reconciled to '106 derived famil(ies) accounted for — 103 run, 3 NOT-MEASURED', 0 UNRUN. Same four NOT MEASURED as round 1, same reasons: check:dual-build-cjs-loads (exit 3), check:lean-entry-closure (exit 3), check:type-check-debt (exit 3), and check:skill-examples (exit 1, 'packages/client-react/dist holds no .d.ts declarations' — the exit-code class now filed as #19227). The three bounded-prerequisite families were built and re-run rather than declared: rebuilding @objectstack/lint's 4-package closure in this worktree turned check:doc-formula-expressions, check:doc-security-posture and check:docs-transcript-drift from exit 3 into exit 0. MERGEABILITY, measured two ways: `git merge-base --is-ancestor e233db9dbb HEAD` exit 0 (self-proving, no control leg owed), and the driver-free probe the script's header prescribes — `git clone --bare --shared . PROBE.git` then `git --git-dir=PROBE.git merge-tree --write-tree --name-only e233db9dbb e1ae025756` — exit 0 printing a tree sha and no conflicted path. GitHub's own `mergeable` read back `null` / `mergeable_state: unknown` at the moment of the read, i.e. still computing; that is reported as what it was rather than waited out. No ablation was re-run: the implementation bytes did not move this round (view.zod.ts and view.test.ts are untouched by the merge), and round 1's three legs stand at their recorded readings.", "mcp_calls": "0 — no MCP GitHub tool was called; every GitHub read and write went through the REST proxy with curl.", "api_writes": "1 this round — POST /repos/objectstack-ai/objectstack/issues/17393/comments (this report). ZERO label writes (needs:contract-review was re-hung by the seat, not by me), no PR-body PATCH, no ready flip, no queue, no auto-merge arm. Round 1's two writes stand as reported.", "open_questions": [], "out_of_scope_findings": [ "noted, not filed: the os-regen routing count on this PR, re-derived per file with `git check-attr merge` rather than estimated by category, reads 13 of the 17 changed paths at this head — the seat's 12 was correct for the pre-merge 16-path diff, and the delta is exactly content/docs/references/ui/component.mdx, which this round's regeneration added. The four unrouted paths are the changeset, view.zod.ts, view.test.ts and type-alias-convention.pin.test.ts. Successor: this PR's own landing window.", "noted, not filed: the round-1 finding about prerequisite refusals exiting 1 is now card #19227 (filed by the seat, deduped onto the #13983 / #14008 migration). It fired again unchanged this round on check:skill-examples, which is why this report's NOT MEASURED count is four where dispatch-gates --ran can only see three. Successor: #19227." ], "deviations": [ "The worktree was recreated for this round (round 1 removed it after opening the PR) and `pnpm install` re-run; the lint closure had to be rebuilt in it before the three bounded-prerequisite gates could be measured.", "Step 3's commit was refused by the pre-commit hook and the regeneration was folded into the single commit e1ae025756, exactly as the refusal text prescribes ('Regenerate the 4 stale artifact(s) above, git add them, and commit again'). So this round carries one commit on top of the merge commit rather than the two the script's prose sketches; the staged diff was read before committing, per the runbook sentence.", "GitHub's mergeable_state was still 'unknown' when read, so no post-sync verdict from the platform is claimed here — only the two local measurements above.", "The contract-review record 5746847791 was pinned to head 5dd391125e and this round moves the head to e1ae025756, so that record no longer points at what would land. Re-hanging needs:contract-review and running a fresh at-tier review is the seat's act, and it has already been done on the card — I wrote no label and took no other action on it." ] }
Generated by Claude Code
- added 3 commits that reference this issue
on Sep 28, 2026
Filed by the director seat (summon #21,
session_01QVMnxyWBx8cAQMsV6akDV9) executing the maintainer's principle, verbatim and untranslated: 「我们的项目以objectstack 协议为准，文档应该以实际实现为准。协议不正确的应该先修改协议。」 (director chat, 2026-09-10T11:0xZ). Routeddomain:specby the anchoring rule;pm:queue. ⛔ Not claimed.Why this is a protocol change first
objectui#7390 is ruled C: an author-settable row ceiling for the gallery view (
$top: limit ?? default, plus a visible truncation footnote). Measured onorigin/mainpackages/spec/src/ui/view.zod.ts:GalleryConfigSchema(:945) is a strictObject of exactlycoverField·coverFit·cardSize·titleField·visibleFields;KanbanConfigSchema(:1241) is exactlygroupByField·summarizeField·columns. Neither declares a row ceiling. objectui already reads a ceiling on kanban (ObjectKanbanSchema.limit, declared in@object-ui/typesonly) and on timeline (its own props interface) — consumer-local author-settable keys the protocol never declared, which is the divergence the third clause forbids. ⇒ The knob enters the protocol first; objectui#7390 blocks on this card, and the kanban/timeline consumer-only keys are brought under the same declaration.Deliverable
@objectstack/specfor the view kinds that page-shaped consumers cap by author choice: gallery and kanban at minimum (timeline if spec declares a timeline config), with a describe stating the default and that the renderer must show a visible truncation signal when it applies. Name and placement are the spec seat's (a sharedlimiton those config schemas is the smallest shape; a base view-config member is the alternative — say which and why).Clause-②: yes(a new authorable key on a published, strict schema); changesetminor; the authorable-surface ledger (packages/spec/authorable-surface/ui.json) gains the rows.Refs: objectui#7390 (ruling 5615809797 and its amendment), objectui#7210, ADR-0021.
Generated by Claude Code