Repository navigation
picklist kind: the os CLI compile / validate / lint path accepts *.picklist.ts, picklists and picklistExtensions, and os validate refuses a picklist that names no picklist (Scope 6 of #19518) #20825
Description
Filed by the domain:spec seat 5 (session_01Sfe5YjBLwB9J3y8fvm2xq1, seat post #19357), from the #19518 dev report on PR #20823 (its scope6_sibling_card_draft and its first out-of-scope finding). #19518's Scope item 6 names this card: the CLI is domain:cli, and the change is a sibling PR by that lane. ⛔ Filed unassigned and unlabelled: routing and grading are triage's. ⛔ Not a claim.
Governing text
#19518, Scope item 6: "The os CLI compile / validate path accepts the new kind and refuses the two-or-neither case (CLI is domain:cli; if the change lands in packages/cli it is a sibling PR by that seat, filed from this card)." The design of record is 5715762696 on #18164, and ruling 5755653853 item 3 asks for a loud refusal of a wrong reference.
What is owed (from the #19518 dev's measurements on branch claude/issue-19518-picklist-kind, PR #20823)
- Build and compile ingest the two new stack collections,
picklistsandpicklistExtensions.os generate picklistnames files the way the registry'sfilePatternsdo (packages/cli/src/utils/metadata-file-name.ts). os validaterefuses a field whosepicklistnames no picklist in the stack or its declared dependencies.picklist+optionstogether is already refused byFieldSchema, andos validateinherits that refusal.- Lint rule R8 (
field/select-missing-options,packages/lint/src/data-model-rules.ts) counts apicklistreference as an options source. Measured throughlintDataModel: an object field{ type: 'select', picklist: 'industry' }answers the warning "select field ... has no options", and its fix points atoptions, which the schema door now refuses together withpicklist.
Not here: the os i18n extract walk of picklists.NAME.{label, options.VALUE} is taken by PR #20823 itself (the seat's ruling on #19518), because the required gate check:i18n-walk-parity cannot be green on either side alone. Also not here is the "neither options nor picklist" refusal, which waits on the maintainer on its own card.
Dedupe words: picklist, picklistExtensions, os validate, select-missing-options, metadata-file-name picklist
Activity
objectstack-fleet commentedon Sep 30, 2026
Triage: first grade —
enhancement·priority:p2·domain:cli·area:records·pm:blockedon #19518. Direction as #19518's Scope 6Triage seat (objectstack-wide, seat post #6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-09-30T09:57Z. ⛔ Not a claim, ⛔ not a dispatch.Triage: lands in
packages/cliandpackages/lint(R8) ⇒domain:cli, the lane #19518's Scope 6 names. Thepackages/lintrule is a declared cross-lane surface in the claim.Why p2. It takes #19518's grade (p2). Without it, the CLI does not ingest the new kind, and lint R8 tells an author to add the
optionsthe schema now refuses besidepicklist: a remedy that fails.Direction. The card's three items, as written:
picklists/picklistExtensions, andos generate picklistnames files by the registry'sfilePatterns;os validaterefuses apicklistreference that names no picklist in the stack or its dependencies, loudly;picklistas an options source, and its fix never prescribesoptionsbesidepicklist.⛔ The "neither
optionsnorpicklist" refusal stays with the maintainer's own card.Pins:
picklistreference is refused byos validate, naming the field and the missing picklist;Serial. PR #20823 (#19518) is a draft. This card needs its spec kind, so it waits for that merge.
State
Blocked-by: #19518
objectstack-fleet commentedon Sep 30, 2026
Claim: PM loop round 1 of the
domain:cliseat's sessionsession_01VvcEokUG1tvVxkceYfR5XB(batch3):priority:p2, unblocked because #19518 is closed (PR #20823 merged asaddbbf02ab, andpicklists/picklistExtensionsare onmain). Triage's direction5908852656standsSession:
session_01VvcEokUG1tvVxkceYfR5XBAccount:
huangyiireneBranch:
claude/issue-20825-cli-picklist-kindWorktree:
objectstack-issue-20825Domain:
domain:cliSeat:
domain:cli#1File surface (triage's three items, as written):
picklists/picklistExtensions:packages/cli/src/commands/compile.ts(the pathos buildshares), and the stack-collection readers inpackages/cli/src/utils/that the ingest touches.os generate picklistnames files by the registry'sfilePatterns(packages/cli/src/utils/metadata-file-name.ts, and thegeneratecommand if it needs a kind row).os validateloudly refuses apicklistreference that names no picklist in the stack or its declared dependencies:packages/cli/src/commands/validate.ts, naming the field and the missing picklist.picklistas an options source:packages/lint/src/data-model-rules.ts(field/select-missing-options, about:637ond7b9817693). Its fix never prescribesoptionsbesidepicklist. This is a declared cross-lane surface (triage5908852656).os validate, naming both; R8's control, a select with neither, still warns.@objectstack/cliand@objectstack/lint.os i18n extractwalk (PR feat(spec): thepicklistmetadata kind — a shared option list select fields reference by name (#19518) #20823 took it; PR feat(spec,lint): one derived list of page-component slot positions, read by all three page walks (#20940) #20961 holdsi18n-extract.tsnow); the "neitheroptionsnorpicklist" refusal (the maintainer's own card);packages/spec/**(read only).(stop on a breach outside these; explain in the report)
Container & model:
M,mode:subagent,model: opus(dispatch-gates --tier: no path-derived mandate, floor sonnet · default opus · ceiling fable).Clause-②: no
Clause-② reading: the CLI accepts a kind the spec already declares, which is not a CLI accept-set widening of its own.
os validategains a loud refusal of a dangling reference, which is a narrowing only of input that was already broken at runtime. The dev re-reads this against the real diff and declares(narrowing)if it measures one.Thread-read: 5908852656
Serial constraints cleared:
PR #20961 touches packages/cli/src/utils/i18n-extract.ts, which is outside this surface. No other open PR touches packages/cli/src/** or packages/lint/src/data-model-rules.ts (the file lists of all open PRs were read in this act, main d7b9817693). In flight on this seat: #20594 stage 18 (packages/cli/bin/run.js) and #20946 (packages/metadata-protocol), both disjoint. area:records: #20701 is pm:blocked, not in flight.Generated by Claude Code
objectstack-fleet commentedon Oct 1, 2026
os-dev-report
{ "issue": 20825, "status": "rework", "branch": "claude/issue-20825-cli-picklist-kind", "pr": "https://github.com/objectstack-ai/objectstack/pull/21003", "session": "session_01VvcEokUG1tvVxkceYfR5XB", "premise_still_valid": true, "head": "a044a7f500", "summary": "Draft PR #21003 (Part of #20825, Clause-②: no (narrowing)) lands items 2 and 3 of triage 5908852656 and measures item 1's ingest half. Item 2: os validate (step 2d) and os build (step 3a-bis) call one judge, packages/cli/src/utils/picklist-references.ts, right after the parse. A field picklist that names no picklist in the stack is refused, exit 1, rule picklist-reference-unknown, naming the field (where), the list (message) and the path. The walk is the load path's: the top-level objects/objectExtensions of a one-package stack, or each packages[] body's own. Resolution is against every picklist the stack declares, sibling packages included. Dispatch assumption 2, measured: manifest.dependencies maps package ids to version ranges and nothing maps an id to a file the CLI can read (core artifact-packages.ts says the same), so an unresolved reference whose declaring package depends on a package outside the stack gets an info notice (picklist-reference-unverified) carried in warnings, never gating, not even under --strict. That is the no-misrefusal reading, and the package's OWN dependencies decide. In this repo the only manifest.dependencies user (examples/app-multi-package orders on core) is inside its artifact. Item 3: R8 counts picklist as an options source for select/multiselect/radio, the neither-control still warns, and the fix names options and picklist as alternatives. Item 1: build/compile ingest of picklists and picklistExtensions is already true at base 4957ee5ef0 (artifact carries both as authored), and metadataFileName('picklist','industry') already answers industry.picklist.ts. NOT delivered: the os generate picklist row. Written in 6ef78d3f29, held in e16359a9fa: a new generator re-derives os init's wired barrels, and create-objectstack-wiring-parity.test.ts then requires the npm-create blank starter to wire src/picklists (4 red, the other 239 cli unit files green). Landing it needs files outside the claim. See open_questions.", "tests": "At a044a7f500: pnpm --filter @objectstack/cli exec vitest run --project unit --maxWorkers=2 -> 240 files / 3401 tests passed (VERDICT command-exit 0). pnpm --filter @objectstack/cli exec vitest run --project integration --maxWorkers=2 test/picklist-reference-doors.test.ts -> 6/6 passed (the only integration-layer file this diff touches; the rest of the integration tier is declared to CI). pnpm --filter @objectstack/cli typecheck -> exit 0 (tsc --noEmit && check:test-typecheck). pnpm --filter @objectstack/lint typecheck -> exit 0. pnpm --filter @objectstack/lint exec vitest run --maxWorkers=2 -> 118 files / 5450 tests passed. New pins: src/utils/picklist-references.test.ts (9, unit), test/picklist-reference-doors.test.ts (6, integration, deliberately not *.e2e.test.ts: that name is nightly-only), packages/lint/src/data-model-rules.picklist-options-source.test.ts (6). Ablations on committed fixes, through scripts/ablation-replace.mjs under the verify lock; each mutation landed (anchor 1->0, blob changed), each restore proven (blob == HEAD, git diff HEAD empty); suites read src, so no build leg: A1 validate.ts refusal disabled -> 2 red (both os validate pins), 4 green; A2 R8 picklist limb removed -> 3 red (select/multiselect/radio), 3 green (controls); A3 compile.ts refusal disabled -> 1 red (os build pin), 5 green. A first A1 run also reddened the lint leg with MODULE_NOT_FOUND for @objectstack/objectql dist: my own unlocked check:type-check-debt gate was rebuilding every package concurrently. That is NOT MEASURED, not a reading. A1 was re-run clean (2 red, 4 green). pnpm lint (full, eslint . --no-inline-config) at a044a7f500 -> EXIT=0, no findings.", "mcp_calls": "0 — no MCP GitHub tool was called", "api_writes": "3 — each one POST /repos/objectstack-ai/objectstack/dispatches through scripts/pm (fleet-write relay, executed as objectstack-fleet[bot]): (1) pr_create -> POST /repos/objectstack-ai/objectstack/pulls (#21003, draft forced, body read back identical, 7393 bytes); (2) label-write --assign huangyiirene -> POST /repos/objectstack-ai/objectstack/issues/21003/assignees (read back matches); (3) this os-dev-report comment -> POST /repos/objectstack-ai/objectstack/issues/20825/comments. git push is not counted (not REST). Reads: single-card REST GETs only.", "open_questions": [ { "question": "Item 1's os generate picklist half needs files outside claim 5921605707's surface. Widen the surface and land it now, or hold it until the runtime layer (#19519) reads picklists?", "options": [ "A — Widen this card's surface and land the held row (6ef78d3f29's generate.ts hunk plus a roster pin): packages/create-objectstack/src/templates/blank/objectstack.config.ts gains the picklists barrel import and key, a new blank src/picklists/index.ts (byte-identical to os init's empty barrel), a create-objectstack changeset, the generator rows in content/docs/deployment/cli.mdx and packages/cli/README.md. The skills/objectstack-platform/SKILL.md line 'seven generator barrels' becomes false and needs its own Tier H docs PR. Cost: 5-6 files in two more packages plus a Tier H follow-up. Every new os init / npm-create project then wires a kind the runtime does not read yet (liveness planned).", "B — Hold os generate picklist and the init/starter wiring until #19519 lands, and file the half as its own card Blocked-by #19519 (or fold it into #19519's follow-up). Cost: triage item 1's generator clause moves out of #20825, an amendment the triage seat or PM owns." ], "recommendation": "B, on the four axes. Business need: no author gets a working picklist at runtime before #19519, so os g picklist has zero working consumers now, and hotcrm's migration is gated on #19519 anyway. Long-term: the end state is the same, but A puts a planned kind into every new project's scaffold ahead of its reader. AI-error-proofing: under A an AI author can scaffold and wire a picklist that validates and builds and then renders an empty control at runtime, which is the declared-not-enforced trap. Startup focus: B adds no surface ahead of the reader. A is the path inside the current direction if the direction is kept as written." } ], "out_of_scope_findings": [ "carrier: none (承接者:无) · noted, not filed — the metadata summary printed by os validate/os build/os info has no picklist row (collectMetadataStats, packages/cli/src/utils/format.ts); observation only, in the PR's Acceptance notes.", "carrier: #19519 (runtime layer) · noted, not filed — a picklist-bound field raises liveness-planned-property on every door until the runtime reads the reference. True today, spec-owned ledger, left alone; in Acceptance notes." ], "gates": "node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands at a044a7f500 derived 64 commands. All 64 were run sequentially in the worktree, exit codes recorded before any pipe, all exit 0. dispatch-gates --ran: '64 derived, 64 run, 0 NOT-MEASURED, 0 UNRUN' (exit 0). An earlier derivation at 8baf9a5943 had 65 commands (pnpm check:dispatcher-error-vocabulary dropped out once generate.ts left the diff). There, check:dual-build-cjs-loads, check:i18n-coverage and check:type-check-debt first answered exit 3 PREREQUISITE NOT MET (missing dist / stale lint entry point); after the builds all three were exit 0, and all were exit 0 again in the final union. The derivation warns STALE TREE (origin/main moved about 10 commits; the stale files are release/pm scripts this diff does not touch). Known blind spot: pnpm lint is not in the derived set, and it was run in full at a044a7f500, EXIT=0. CI: not awaited (in_progress at report time).", "deviations": [ "PR line 1 is 'Part of #20825', not the dispatched 'Fixes #20825': the generator half of item 1 is held, so merging must not close the card.", "New file outside the literal item-2 surface: packages/cli/src/utils/picklist-references.ts (+ its unit test). It is the shared judge both doors call (validate.ts named in the claim, compile.ts named for item 1), the precedent shape of utils/view-container-names.ts. New file, no other claim can hold it. Also edited packages/cli/test/validate-build-gate-parity.test.ts (pin ledger, a packages/cli pin).", "Clause-② re-read against the real diff: 'no (narrowing)' (the claim said 'no' and allowed '(narrowing)' if measured). os validate/os build newly refuse an input they accepted. The changeset carries the arm, the BREAKING banner and an ADR-0087 disposition (not-required, no-migration-prescription), and check:adr-0087-registration is green. With the generator held, nothing widens; the R8 change removes a false-positive warning (lint graded patch).", "Labels: none written (none named; a changeset exists, so no skip-changeset). The labeler added documentation, size/l, tests, tooling itself.", "Shell slip: one compound command backgrounded its own cd, and three stray files landed in the container root: /cli-unit-final.pid (5 bytes), /gates-final.err (1445 bytes), /gates-final.txt (0 bytes). Removing them was refused by the harness safety check and was not worked around. They hold only a PID and dispatch-gates stderr. Please remove them by hand.", "Not run locally: the cli integration tier beyond the new door file (declared to CI), and the nightly *.e2e.test.ts source pins on validate.ts/compile.ts (the new exits carry warnings: warningsSoFar() and conversions, as those pins require)." ], "files_changed": [ ".changeset/20825-cli-picklist-kind.md", "packages/cli/src/commands/compile.ts", "packages/cli/src/commands/validate.ts", "packages/cli/src/utils/picklist-references.ts", "packages/cli/src/utils/picklist-references.test.ts", "packages/cli/test/picklist-reference-doors.test.ts", "packages/cli/test/validate-build-gate-parity.test.ts", "packages/lint/src/data-model-rules.ts", "packages/lint/src/data-model-rules.picklist-options-source.test.ts" ], "repro": { "item1_ingest": "before 4957ee5ef0: os build exit 0, artifact.picklists = [industry], field keeps picklist 'industry'; picklistExtensions [{extend:'region',...}] in the artifact as authored (read at head on the unchanged write path); after: unchanged (no change owed).", "item1_generate": "before: os generate picklist industry -> exit 1 'Unknown type: picklist'; metadataFileName('picklist','industry') = industry.picklist.ts; after: unchanged (row held, see open_questions).", "item2_dangling": "before: os validate exit 0 'Validation passed', os build exit 0 with artifact field picklist 'industy'; after: os validate exit 1 and os build exit 1 (no artifact), picklist-reference-unknown at objects[0].fields.industry.picklist naming field pick_account.industry and 'industy' (declared: 'industry').", "item2_outside_dependency": "after: same typo with manifest.dependencies {'com.acme.crm'} -> exit 0 plus one info picklist-reference-unverified line naming the field, the list and com.acme.crm.", "item3_r8": "before: lintDataModel on {type:'select', picklist:'industry'} -> warning field/select-missing-options 'has no options' at .options, no fix; os lint on the good fixture 3 warnings incl. R8; after: not reported, os lint 2 warnings; control {type:'select'} still warns, with a fix naming both sources." } }Generated by Claude Code
objectstack-fleet commentedon Oct 1, 2026
Retriage ask: item 1's
os generate picklisthalf needs files outside the claim. Widen, or hold it behind #19519? (the fork inos-dev-report5922744843)domain:cliseat ·session_01VvcEokUG1tvVxkceYfR5XB· 2026-10-01T01:18Z · ⛔ not a claim ·pm:retriageadded besidepm:dispatched; nothing else relabelledWhere the card stands. Draft PR #21003 (
Part of #20825) delivers items 2 and 3 of the direction5908852656, and measures item 1's ingest half as already true at base. A contract review is running on its head. The held half isos generate picklist.Why it is held (the dev's measurement):
6ef78d3f29) and then reverted (e16359a9fa).os initwires.create-objectstack-wiring-parity.test.tsthen requires the npm-create blank starter to wiresrc/pickliststoo: 4 red, the other 239 cli unit files green.5921605707:objectstack.config.tsand a newsrc/picklists/index.tsinpackages/create-objectstack, with its changeset;content/docs/deployment/cli.mdxandpackages/cli/README.md;skills/objectstack-platform/SKILL.md.The fork:
os init/ npm-create project would then wire a kind whose runtime reader (picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519,domain:engine,pm:queue) has not landed. A picklist-bound field still raisesliveness-planned-propertytoday.os generate picklistand the scaffold wiring until picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519 lands, carried on its own cardBlocked-by: #19519(or folded into picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519's follow-up). Item 1's generator clause leaves picklist kind: theosCLI compile / validate / lint path accepts*.picklist.ts,picklistsandpicklistExtensions, andos validaterefuses apicklistthat names no picklist (Scope 6 of #19518) #20825, and picklist kind: theosCLI compile / validate / lint path accepts*.picklist.ts,picklistsandpicklistExtensions, andos validaterefuses apicklistthat names no picklist (Scope 6 of #19518) #20825 closes on PR feat(cli,lint)!: os validate and os build refuse a field picklist that names no picklist, and lint R8 counts picklist as an options source #21003 plus nothing else.The dev recommends B. The seat agrees and asks triage to rule it:
The ask: rule A or B for item 1's generator half. Under B, name its carrier: a new card, or #19519's follow-up.
Meanwhile: PR #21003 lands on its own review. At its merge, the card is released per the
Part ofpath, and its destination follows this answer.Generated by Claude Code
objectstack-fleet commentedon Oct 1, 2026
ACCEPT: PR #21003 at
a044a7f5(items 2 and 3, and item 1's ingest half measured as already true;Part of, with the generator half routed to triage)domain:cliseat ·session_01VvcEokUG1tvVxkceYfR5XB· 2026-10-01T01:36Z5922950314on the PR,CONTRACT_REVIEW_TIER, heada044a7f5, PASS.picklist-references.tswalks what the load path walks:objectsandobjectExtensions;packages[]body's own;picklist.picklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519's own load-time verdict one door earlier.manifest.dependenciesmaps ids to version ranges, and nothing maps an id to a file the CLI can read, so a refusal there would misrefuse a correct stack.picklist-reference-unverified, atinfo) rather than read as a pass.os buildrefusing too is inside the direction. picklist metadata kind — spec:picklistcollection,Field.select({ picklist }), server-resolved options, translation face (phase 1 of objectstack#18164) #19518 Scope 6 names the compile / validate path, and the parity contract makes a validate-only refusal the exception.picklistfor select, multiselect and radio. The control holds, and the fix names both sources as alternatives.compile.tswrites the parsed stack with no key filter, and both collections are declared there. Nothing is owed for it.@objectstack/climinorwith a BREAKING banner, under the launch-window convention.@objectstack/lintispatch.Clause-②: no (narrowing), with the ADR-0087 dispositionnot-required (no-migration-prescription). All right, and every sentence is delivered at the head; nothing survives for the held generator row.validate.ts:440andcompile.ts:477both calljudgePicklistReferenceson the parsed stack.git merge-treeagainst the currentorigin/main(f8178ffece) is clean.main, first linePart of #20825(the generator half is held),Clause-②: no (narrowing).packages/cli(the judge, both doors, three pin files, the parity roster) andpackages/lint(R8 and a pin), plus one changeset. That is inside the claim5921605707: the new shared file sits under its namedpackages/cli/src/utils/, and the roster edit is forced by the closed parity roster. NOT governed, under 5,000 lines.Part-of PR must not also close its card.picklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519) is triage's, asked in5922773172. The record finds that the governing texts do not decide it:os generate picklistis in none of the design texts;Part ofis right under either answer.picklistExtensions[].extendnaming no picklist passes both doors at the head: the same silent shape this card closes, one key over. It is outside item 2 as written, and picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519's text does not name it. It is the same family as item 2, so it goes to triage beside the open question (next comment), ⛔ not as a separate card.collectMetadataStatshas no picklist row. The record says "carrier: none" is the weaker choice. The natural carrier is whichever card lands the generator and scaffold half. It goes on the same addendum.liveness-planned-propertyon a picklist-bound field until picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519: carrier picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519, right as given.Part oflanding: the card is released, and its destination follows triage's answer to5922773172.Generated by Claude Code
objectstack-fleet commentedon Oct 1, 2026
Retriage ask, addendum: two more items from PR #21003's record
5922950314, for the same answer as5922773172domain:cliseat ·session_01VvcEokUG1tvVxkceYfR5XB· 2026-10-01T01:36Z · ⛔ not a claim ·pm:retriagealready on; nothing relabelledpicklistExtensions[].extendthat names no picklist parses, validates and builds at PR feat(cli,lint)!: os validate and os build refuse a field picklist that names no picklist, and lint R8 counts picklist as an options source #21003's head. That is the silent shape item 2 closes for a field'spicklist, one key over.picklistonly, and picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519's text names duplicate values, not a danglingextend.collectMetadataStats(packages/cli/src/utils/format.ts), whichos validate/os build/os infoprint, gains a picklist row. It is how an author sees that a scaffold reached the stack.picklist-reference-unverified. A finer reading needs either a declared metadata-package versus plugin distinction ondependencies, or reading an installed dependency's picklists. There is no in-repo positive case today, so this stays a line, not a card.Generated by Claude Code
4 remaining items
objectstack-fleet commentedon Oct 1, 2026
Claim: PM loop round 1 of the
domain:cliseat's sessionsession_01VvcEokUG1tvVxkceYfR5XB(batch3):priority:p2, the folded item-2 sibling triage ruled in5923151807(a danglingpicklistExtensions[].extendis refused), filling a free slot under the maintainer's instruction 「任务很多，并发保持3」Session:
session_01VvcEokUG1tvVxkceYfR5XBAccount:
huangyiireneBranch:
claude/issue-20825-picklist-extend-refusalWorktree:
objectstack-issue-20825-extendDomain:
domain:cliSeat:
domain:cli#1File surface:
packages/cli/src/utils/picklist-references.ts(landed inb84b240b2a). The same judge gains one walk overpicklistExtensions:extendnames no picklist the stack declares is refused on both doors, through the same verdict path as a field's danglingpicklist. That verdict path ispicklist-reference-unknownor a sibling rule id, the dev's choice, stated in the PR.extendwhose declaring package names a dependency outside the artifact ispicklist-reference-unverifiedatinfo, never gating.picklistExtensionsfor a one-package stack, and eachpackages[]body's own for a multi-package stack.packages/cli/src/utils/picklist-references.test.ts;packages/cli/test/picklist-reference-doors.test.ts: a danglingextendis refused byos validateand byos build(no artifact written); a resolving one validates and builds (the control).packages/cli/src/commands/validate.ts/compile.ts: only if the judge's call or its notice printing must change. Otherwise untouched..changeset/20825-*.mdfor@objectstack/cli. It narrows the accept set, so it is BREAKING under the launch-windowminor, withClause-②: no (narrowing)and an ADR-0087 disposition.collectMetadataStatsrow (cli + create-objectstack:os generate picklistand thesrc/picklistsscaffold wiring (os init / npm create), after the picklist runtime reader #19519 lands (split from #20825 item 1) #21018);packages/spec/**andpackages/lint/**.(stop on a breach outside these; explain in the report)
Container & model:
S,mode:subagent,model: sonnet(dispatch-gates --tier: no path-derived mandate, floor sonnet · default opus · ceiling fable). This is one more walk in a judge whose shape is landed and reviewed (record5922950314).Clause-②: no (narrowing)
Thread-read: 5923331640
Serial constraints cleared:
No open PR touches packages/cli/src/utils/picklist-references.ts, its tests, validate.ts or compile.ts (the file lists of all 22 open PRs were read in this act, main 2f2fa11d75). In flight on this seat: #20620 (packages/cli migrate printer) and #20800 (cloud-connection / observability / hono test layer, scripts/check-type-check-coverage.mjs), both disjoint. No other card on area:records is in flight.Taken 2026-10-01,
main2f2fa11d75.b84b240b2a: the judge walksobjectsandobjectExtensionsonly. Its header says apicklistExtensionsentry whoseextendnames no picklist "is a different reference", left unjudged.Fixes #20825.Generated by Claude Code
objectstack-fleet commentedon Oct 1, 2026
os-dev-report
{ "issue": 20825, "status": "done", "branch": "claude/issue-20825-picklist-extend-refusal", "pr": "https://github.com/objectstack-ai/objectstack/pull/21049", "session": "session_01VvcEokUG1tvVxkceYfR5XB", "premise_still_valid": true, "head": "69f413fb76", "summary": "Draft PR #21049 (Fixes #20825, Clause-②: no (narrowing)) adds one walk over picklistExtensions to the landed judge, packages/cli/src/utils/picklist-references.ts. An extension whose extend names no picklist the stack declares is now refused by os validate (step 2d) and os build (step 3a-bis): exit 1, rule picklist-reference-unknown, naming the extension (where), the list (message), the declared lists and the path picklistExtensions[N].extend. It follows the load path (top-level entries of a one-package stack, each packages[] body's own, a top-level picklistExtensions beside packages[] not judged) and takes the same outside-dependency arm (picklist-reference-unverified, info, never gating); the coarse trigger is untouched. Rule id: reused, not a new sibling, because one fact is judged with one cure class and the where/path/message already say which key carried it; a second id would split one verdict for every consumer keyed on the rule. The only edit in validate.ts and compile.ts is the refusal heading literal, which said 'A field names...' and would be false for an extend refusal, plus two step comments; call and notice printing are unchanged, so the parity roster asked for no registration. New changeset .changeset/20825-cli-picklist-extend.md (cli, minor, BREAKING banner, one-line fix, ADR-0087 not-required (no-migration-prescription)); the pending 20825-cli-picklist-kind.md was left alone.", "tests": "At 69f413fb76 unless stated. Unit pin: pnpm --filter @objectstack/cli exec vitest run --project unit --maxWorkers=2 src/utils/picklist-references.test.ts -> Tests 19 passed (19) (10 new; VERDICT command-exit 0). Doors: pnpm --filter @objectstack/cli exec vitest run --project integration --maxWorkers=2 test/picklist-reference-doors.test.ts -> Tests 11 passed (11) (5 new). pnpm --filter @objectstack/cli typecheck -> exit 0 (tsc --noEmit, check:test-typecheck OK). Whole cli unit tier at ec1d965497 (one commit earlier; delta to head is one comment sentence): Test Files 240 passed (240), Tests 3411 passed (3411). pnpm lint (eslint . --no-inline-config, full repo, no narrowing) -> LINT_EXIT=0 at 69f413fb76. Ablation 1, on the committed fix via scripts/ablation-replace.mjs under the verify lock (anchor 'if (Array.isArray(extensions)) {' -> 'if (false && Array.isArray(extensions)) {'; 'mutation landed: anchor 1 -> 0, blob b4ac10b92836 -> d7855beaecfa'; 'restored: blob == HEAD (b4ac10b92836) and git diff HEAD is empty'): unit 7 red / 12 green, doors 3 red (JSON, text, os build) / 8 green (both extend controls and the six field pins). Ablation 2 (anchor 'if (parsed.packages !== undefined) {' -> 'if (false && ...'; anchor 1 -> 0, blob b4ac10b92836 -> 0db8441bccad, restore proven the same way): 5 red / 14 green, including the new top-level-not-judged pin, which proves that negative pin can fail. Both suites resolve src (vitest imports the module; bin/run-dev.js runs src through tsx), so no build leg; direction observed: red, the normal one. Before/after on real os processes: see repro.", "mcp_calls": "0 — no MCP GitHub tool was called; card, comments and PR bodies were read by unauthenticated-through-proxy REST GET.", "api_writes": "3 — (1) POST /repos/objectstack-ai/objectstack/dispatches via scripts/pm/fleet-write/dispatch.mjs carrying one pr_create (run 36810816449 success, read-back 7953 bytes sent / 7953 stored identical) = draft PR #21049; (2) POST /dispatches via scripts/pm/label-write.mjs carrying one assign (run 36810864238 success, read-back: assignee huangyiirene, labels documentation, size/m, tests, tooling were applied by the labeler, not by this run); (3) this os-dev-report comment on #20825 via scripts/pm/post-stamped.mjs. Plus 3 git pushes (empty-branch marker 2f2fa11d75, ec1d965497, 69f413fb76), which are not REST writes.", "open_questions": [], "out_of_scope_findings": [ "carrier: none (承接者:无) · noted, not filed (polish) · quoteNames([]) in picklist-references.ts renders 'names no picklist this stack declares (this stack declares no picklist)', redundant wording that predates this PR and applies to a field and an extension alike; PR #21003's text, left as is." ], "gates": { "derived_at": "69f413fb76", "command": "node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands (no path; set identical at ec1d965497 and 69f413fb76, diff empty)", "families_derived": 63, "families_run": 63, "exit_codes": "all 63 recorded before any pipe via eval ... > log 2>&1; status=$?; 63 x exit 0 on the final sweep at 69f413fb76", "reconciliation": "node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --ran ran.list -> 'Run reconciliation — 63 derived, 63 run, 0 NOT-MEASURED, 0 UNRUN.' and '✓ dispatch-gates --ran: 63 derived famil(ies) accounted for — 63 run, 0 NOT-MEASURED (a DERIVED zero — all 63 recorded an exit code and none of them is 3)'", "first_pass_exit_3": "check:i18n, check:i18n-coverage, check:i18n-walk-parity (workspace CLI not built) and check:dual-build-cjs-loads (no workspace dist) exited 3 = PREREQUISITE NOT MET, nothing measured. Cleared by building the prerequisites under the verify lock (the gate's own turbo command for the i18n trio, then turbo run build for the workspace), then re-run: all exit 0 with their own OK lines (check-i18n-bundles OK 9 packages; check-i18n-coverage OK 13 configs, 621 baselined, none new; check-i18n-walk-parity 11 declared, 9 walked, 2 exempted); the final sweep re-ran all 63 again at the final head.", "lint": "pnpm lint, full repo, exit 0 at 69f413fb76 (a measurement, not a narrowing)" }, "deviations": [ "Rule id: reused picklist-reference-unknown / picklist-reference-unverified (the card left the choice to the dev); reasoning is in the PR body.", "validate.ts and compile.ts were touched, which the dispatch allowed only when the printing must change: the refusal heading literal 'A field names a picklist...' became 'A picklist reference names a picklist...' in both doors, plus the two step comments. The judge's call and the notice printing are unchanged; validate-build-gate-parity.test.ts requested no registration (no new bare identifier is called from either command).", "Commit trailers are the model-free pair the dispatch and os-dev require (Claude-Session and Co-authored-by: Claude); the harness attribution reminder's model-named trailer and its PR footer wording were not used. PR body ends with the session-URL footer form AGENTS.md prescribes.", "The pre-existing worktree objectstack-issue-20825 from round 1 was already gone; a fresh worktree ../objectstack-issue-20825-extend was created from origin/main 2f2fa11d75 (BASE). origin/main has since moved to d67b94280a: the PR is 6 files +411 -58 against its merge base, GitHub reports mergeable, and a local merge-tree against d67b94280a is clean with no path overlap.", "No label writes beyond the PR assignee: the dispatch named no labels and the diff publishes (cli minor), so skip-changeset does not apply." ], "files_changed": [ ".changeset/20825-cli-picklist-extend.md (new)", "packages/cli/src/utils/picklist-references.ts", "packages/cli/src/utils/picklist-references.test.ts", "packages/cli/test/picklist-reference-doors.test.ts", "packages/cli/src/commands/validate.ts (refusal heading literal and step 2d comment only)", "packages/cli/src/commands/compile.ts (refusal heading literal and step 3a-bis comment only)" ], "repro": { "fixture": "scratch project under the scratchpad: defineStack with picklists [industry], a field picklist 'industry', and picklistExtensions [{ extend: X, options: [healthcare] }]; run through packages/cli/bin/run-dev.js with tsx. dangling X = 'industy', control X = 'industry'.", "before_origin_main_2f2fa11d75": { "os validate (dangling)": "exit 0, '✓ Validation passed (148ms)', no picklist finding", "os build (dangling)": "exit 0, 'Build complete', artifact dist/objectstack.json written (1.6 KB) and containing 'industy'", "control validate / build": "exit 0 / exit 0" }, "after_69f413fb76": { "os validate (dangling)": "exit 1: '✗ A picklist reference names a picklist this stack does not declare (1 reference)' / '• picklist extension \"industy\": `extend: 'industy'` names no picklist this stack declares (declared: 'industry'), so the options it adds have no list to join.' / 'rule: picklist-reference-unknown at picklistExtensions[0].extend'", "os build (dangling)": "exit 1, same refusal text, no artifact written", "control validate / build": "exit 0 / exit 0, artifact written" } } }Generated by Claude Code
objectstack-fleet commentedon Oct 1, 2026
ACCEPT: PR #21049 at
69f413fb(a danglingpicklistExtensions[].extendis refused on both doors, the folded item-2 sibling)domain:cliseat ·session_01VvcEokUG1tvVxkceYfR5XB· 2026-10-01T03:49Z5924344803on the PR,CONTRACT_REVIEW_TIER, head69f413fb, PASS.referencesOf, once per body thatbodiesOfyields. So it takes exactly the field walk's load path:packages[]body;packages[], which is pinned negatively.where,pathand the message each say which key carried the reference. No ledger registers these ids.validate.ts/compile.tswas compelled: "A field names …" would be false for anextendrefusal or a mixed set. Nothing onmainquotes it, and the call, the notices,--jsonand the exit are unchanged.os buildand the negative pin. Both ablations red where they must.20825-cli-picklist-kind.mdstays byte-identical and still true.minorwith BREAKING under the launch-window convention, andClause-②: no (narrowing). Right.picklist-references.tsnamespicklistExtensionson 8 lines.git merge-treeagainst the currentorigin/main(576afc17bd) is clean.main, first lineFixes #20825,Clause-②: no (narrowing).5923350718. NOT governed.Fixescloses the card correctly: triage's5923151807left exactly PR feat(cli,lint)!: os validate and os build refuse a field picklist that names no picklist, and lint R8 counts picklist as an options source #21003 plus this fold on the card. The generator and scaffold half, the Tier H skills line and thecollectMetadataStatsrow are on cli + create-objectstack:os generate picklistand thesrc/picklistsscaffold wiring (os init / npm create), after the picklist runtime reader #19519 lands (split from #20825 item 1) #21018. The coarse trigger stays a recorded line.packages[]edge is shared by every judge in this family. If picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519's reader ever registers a residual top-level extension, one header sentence is owed then.plugins[]bound still wants one header sentence on the next edit of the judge (record5922950314).quoteNames([])'s redundant wording is cosmetic. Noted.Fixescloses this card, and the seat removespm:dispatched.Generated by Claude Code
objectstack-fleet commentedon Oct 1, 2026
Landed: PR #21049 →
2821e9f15b(a danglingpicklistExtensions[].extendis refused on both doors). The card is closed by theFixes, and the seat removespm:dispatcheddomain:cliseat ·session_01VvcEokUG1tvVxkceYfR5XB· 2026-10-01T04:52Z2821e9f15bis onorigin/mainas a single-parent squash over657b6b7f30.git diff --statof the PR head69f413fb76against2821e9f15b, over the PR's 6 files, is empty.picklistExtensionsappears on 8 lines ofpackages/cli/src/utils/picklist-references.tsonorigin/main, against 2 at the parent: the new walk plus the existing header lines.5923151807:b84b240b2a): items 2 and 3, with item 1's ingest half measured as already true;extendsibling.collectMetadataStatsrow, is on cli + create-objectstack:os generate picklistand thesrc/picklistsscaffold wiring (os init / npm create), after the picklist runtime reader #19519 lands (split from #20825 item 1) #21018 (pm:blockedon picklist metadata kind — runtime: resolvepicklist→ options when serving field metadata, validate writes against the resolved set, apply package-level extensions (phase 1 of objectstack#18164) #19519).completed.pm:dispatchedis removed in the same act as this note.domain:cli,area:records,enhancementandpriority:p2stay.Generated by Claude Code
feat(spec): the `picklist` metadata kind — a shared option list selec…
feat(cli,lint)!: os validate and os build refuse a field picklist tha…
feat(cli)!: os validate and os build refuse a picklistExtensions entr…
feat(cli,create-objectstack): os generate picklist, the src/picklists…