Repository navigation
[finding] install-local: reseed-sample-data over intact sample rows answers 422 RESEED_NO_ROWS "The package declares no seedable records for this runtime" while the package declares 28 records (all already present) #21776
Description
Activity
Triage: first grade —
bug·priority:p3·domain:cli·area:devpath·pm:queue(findingremoved). "Nothing to do" is a successTriage seat (objectstack-wide, seat post #6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-10-04T20:55Z. ⛔ Not a claim, ⛔ not a dispatch.The answer: a reseed whose every declared record is already present answers success, with
inserted: 0andskipped: 28. A reseed is idempotent by its own contract, so an intact baseline is the expected state, not a fault.RESEED_NO_ROWSstays, with its text, only for a package that really declares no seedable records for this runtime (the loader reports 0 declared). Its message then states a true cause.handleReseed(marketplace-install-local-plugin.tsabout:1611–:1660) reads the loader'sskippedcount besideinserted/updated/errors. The response carries all four, so the caller sees why nothing changed.
Why p3. No data is affected. It is a refusal that names a false cause.
Pins:
- over an intact baseline: 2xx with
{ inserted: 0, skipped: 28 }; - after a purge: 2xx with
inserted: 28; - a package declaring no seed records: 422
RESEED_NO_ROWSwith the existing text.
Generated by Claude Code
Claim: PM loop round 1
Session:session_01RWZbGvPFcRKvUqASZtunCU
Account:os-warren(the seat's linked user asget_meanswers it; the card's assignee)
Branch:claude/issue-21776-reseed-intact-baseline
Worktree:objectstack-issue-21776
Domain:domain:cli
Seat:domain:cli#1
File surface, per triage5984295768(read onorigin/main07bf21ff):packages/cloud-connection/src/marketplace-install-local-plugin.ts, thehandleReseedregion only (about:1852; the triage's:1611–:1660is stale after four landings today): a reseed whose every declared record is already present answers success withinserted: 0and theskippedcount;RESEED_NO_ROWSstays only for a package that declares no seedable records.- Tests (unit, and a real-boot door pin);
.changeset/21776-SLUG.md. - One measurement carried from [finding] marketplace install-local installs a manifest whose engines.protocol this runtime cannot satisfy (^16 on 17): 200 success, while POST /api/v1/packages refuses it 422 OS_PROTOCOL_INCOMPATIBLE #21762's review (landed note
5987844456, note (d)): reseed and purge on an entry the rehydrate refused, reported with its measured reach, ⛔ not fixed here. - ⛔ No change to the purge, the listing, the rehydrate or the install route. ⛔ No
packages/specpath. ⛔ No new error code. (stop on breach; explain in the report)
Container & model:M,mode:subagent,model: default (opus);dispatch-gates --tierover the path: no path-derived mandate. TheClause-②: yesbelow owes a same-head review atCONTRACT_REVIEW_TIERbefore landing, reason:landing-operations.md「双肢命中任一 ⇒ 无达档条款②复核 PASS 在案 ⛔ 禁止入队」.
Clause-②: yes (widening) - A reseed over an intact baseline goes from a refusal (
422 RESEED_NO_ROWS) to a success: the door accepts a request it refused. The only published text on the code, its ledger row's comment ("reseed ran but wrote nothing",packages/spec/src/api/error-code-ledger.zod.ts:1038), describes that refusal rather than denying it, so the false-refusal reading ofnodoes not apply.yestakes at leastminorfor@objectstack/cloud-connection(AGENTS.md Post-Task Checklist step 3: 「yestakes at leastminor」). This seat does not build: the dev reports the response shape before and after, and that measurement decides the declaration.
Thread-read: 5984295768
Serial constraints cleared: read 2026-10-05T06:38Z: marketplace-install-local-plugin.ts: [finding] install-local: after a restart whose rehydrate refused a protocol-incompatible package, GET /install-local still lists it as installed (200, no "not loaded" marker); with PR #21820 each GET also logs a warn for it #21822 (claim5989168501) is in flight in the SAME file, in other regions (the rehydrate's refusal record andhandleList). This card holds onlyhandleReseed. The maintainer raised this seat's concurrency in this session's chat: 「任务很多，并发加到3」; the file is not a single-claim path, so two disjoint regions are ordinary concurrency and the later lander resolves (execution-duties.md). Fold-or-serial: neither fold (gate ① fails: different defects and fixes) nor serial, on that directive; the earlier shift's "same file hard serial" seat practice yields to it. The dev mergesorigin/mainbefore its PR and again after [finding] install-local: after a restart whose rehydrate refused a protocol-incompatible package, GET /install-local still lists it as installed (200, no "not loaded" marker); with PR #21820 each GET also logs a warn for it #21822 lands.- Open PRs: none touches the file (read at [finding] install-local: after a restart whose rehydrate refused a protocol-incompatible package, GET /install-local still lists it as installed (200, no "not loaded" marker); with PR #21820 each GET also logs a warn for it #21822's claim, 2026-10-05T06:17Z). Verify lock: one holder ([finding] install-local: after a restart whose rehydrate refused a protocol-incompatible package, GET /install-local still lists it as installed (200, no "not loaded" marker); with PR #21820 each GET also logs a warn for it #21822's dev build), queue empty.
domain:cliseat ·session_01RWZbGvPFcRKvUqASZtunCU· 2026-10-05T06:38Zos-dev-report
{
"issue": 21776,
"status": "done",
"branch": "claude/issue-21776-reseed-intact-baseline",
"pr": "#21832",
"session": "session_01RWZbGvPFcRKvUqASZtunCU — subagent run, the parent PM session id",
"premise_still_valid": true,
"summary": "handleReseed (packages/cloud-connection/src/marketplace-install-local-plugin.ts, now :1860) reads skipped beside inserted/updated/errors. A run with nothing written, no errors and skipped above 0 is the intact baseline: it answers 200 with all four counts and sets the ledger as when rows land. Every success now carries skipped. Both refusals are byte-unchanged: errors with nothing written stays 422 RESEED_NO_ROWS with the error count, and a run in which the loader processed no record stays 422 RESEED_NO_ROWS with its existing text, which is now a true cause. Mechanism readings: H1 holds on 07bf21f (:1852, wrote = inserted + updated, no skipped read). H2: seeded.skipped is already carried (runInlineSeed returns result.summary.totalSkipped; the install handler in the same file already counts skipped as rows present), so no new source was needed. H3: 'the loader reports 0 declared' = all four counts 0. The loader reconciles inserted+updated+skipped+errored against the records it processes for this runtime (SeedLoadResult); real cases are every dataset scoped by Seed.env to another environment, or records: []. A package with no seed dataset never reaches the loader and answers 400 RESEED_SKIPPED no-datasets, unchanged and pinned. H4 holds, pinned with details unchanged, also when skipped records sit beside the errors. Response shape, measured on a real boot: before, the intact baseline answered 422 {success:false, error:{code:RESEED_NO_ROWS, message:'Reseed wrote no rows. The package declares no seedable records for this runtime.', details:{inserted:0,updated:0,errors:0}}}; after, 200 {success:true, data:{manifestId:'com.example.crm', inserted:0, updated:0, skipped:28, errors:0, withSampleData:true}}. The success data gains skipped (before: {manifestId, inserted, updated, errors, withSampleData}). Clause-②: yes (widening), minor for @objectstack/cloud-connection, carried in the changeset and as line 3 of the PR body.",
"tests": "At head 4443642 (after merging origin/main 08adfea). pnpm --filter @objectstack/cloud-connection test: 'Test Files 39 passed (39) / Tests 488 passed (488)'. The typecheck (tsc --noEmit && tsc --noEmit -p tsconfig.test.json) exit 0, and --listFiles counts marketplace-install-local-reseed.test.ts 1. That unit file has 7 cases: intact baseline 200 with the whole payload and the ledger flip; rows land with skipped 0; no record processed 422 with code, status and the existing text; errors 422 with details toEqual {inserted:0,updated:0,errors:2}; errors beside skipped 422 unchanged; no dataset 400 RESEED_SKIPPED; partial success. Door pin on a real boot, packages/qa/dogfood/test/install-local-reseed-intact-baseline.dogfood.test.ts, 5 cases on two bootStack(showcase) boots: intact 200 toEqual {manifestId, inserted:0, updated:0, skipped:28, errors:0, withSampleData:true} with every row's updated_at unchanged; after purge 200 with inserted:28, skipped:0; every dataset env ['prod','test'] (bootStack forces NODE_ENV=development, which the loader reads as dev) 422 RESEED_NO_ROWS with the existing text and 0 rows. This file plus the three other reseed-touching dogfood files (purge, listing-sample-data, no-active-organization): 'Test Files 4 passed (4) / Tests 25 passed (25)'. The dogfood tsc --noEmit exit 0 lists the new file. The first door run was red on the env case only: I had assumed NODE_ENV=test, and it measured 'development'; the fix scopes the datasets away from dev. Ablation at dbf5e03, one-off, via scripts/ablation-replace.mjs wrap mode plus a trap restore. The mutation made the intactBaseline condition never true: anchor 'skipped > 0;' x1 -> x0, replacement x0 -> x1, blob 53f51a73 -> 27c721a6. Then pnpm --filter @objectstack/cloud-connection build exit 0, and ablation-dist-preflight '@objectstack/cloud-connection' with the marker exit 0 ('marker present in 2 built files'). Unit: 'Tests 1 failed | 6 passed (7)' (intact case: expected 422 to be 200). Dogfood: 'Tests 1 failed | 4 passed (5)' (intact case; it printed the before body quoted in summary). Restore: blob == HEAD 53f51a73 and git diff HEAD empty. Rebuild, preflight --absent exit 0 ('marker absent from all 6 built files', 'working tree clean against HEAD'). Unit 7/7 and dogfood 5/5 passed again. Direction: red, as predicted.",
"mcp_calls": "0",
"api_writes": "3 — each one repository_dispatch to POST /repos/objectstack-ai/objectstack/dispatches (HTTP 204), executed by the fleet-write relay as objectstack-fleet[bot]: (1) pr_create -> POST /repos/objectstack-ai/objectstack/pulls (draft #21832; read-back: 8934 bytes sent, 8934 stored, identical); (2) assign via scripts/pm/label-write.mjs -> POST /repos//issues/21832/assignees [os-warren] (read-back matches); (3) this os-dev-report via scripts/pm/post-stamped.mjs -> POST /repos//issues/21776/comments. git push (3 pushes to the branch) is not a REST write. 0 label writes: the dispatch names no label, and skip-changeset does not apply because the package publishes.",
"open_questions": [
{
"question": "The ruling's third pin reads 'a package declaring no seed records: 422 RESEED_NO_ROWS with the existing text', and its body defines that case as 'the loader reports 0 declared'. Measured on main, a package with NO seed dataset at all never reaches the loader: applySideEffects leaves mode 'skipped' reason 'no-datasets', and the reseed answers 400 RESEED_SKIPPED before the RESEED_NO_ROWS branch. Which answer belongs to that case?",
"options": [
"A — keep it as built: 422 RESEED_NO_ROWS only where the loader ran and processed 0 records (datasets scoped to another environment, or records: []), pinned on a real boot; no-datasets stays 400 RESEED_SKIPPED, unchanged and pinned at unit level.",
"B — also map no-datasets to 422 RESEED_NO_ROWS: a code and status change for a case the ruling did not name in its parenthetical, which contradicts the published changeset text of #21774 (it lists 'a package with no seed datasets' among the declines that stay 400 RESEED_SKIPPED)."
],
"recommendation": "A, because the ruling's own parenthetical names the loader's count. It is not blocking, and the PR implements A."
}
],
"out_of_scope_findings": [
"class: b · reach: measured on a bootStack(showcase) boot whose ledger holds com.example.crm with engines.protocol ^16 on protocol 17 (the rehydrate logged OS_PROTOCOL_INCOMPATIBLE '…is NOT loaded into this runtime…', and 0 of the 5 CRM objects are registered). POST /api/v1/marketplace/install-local/com.example.crm/reseed-sample-data answered 400 RESEED_SKIPPED 'Reseed did not run: seed-error: Object 'crm_account' not found'. Before that, it loaded the refused package's translation bundle into the i18n service (2 loadTranslations calls) and merged its 5 seed datasets into the kernel's shared seed-datasets list (0 -> 5, with the seed-replayer present). There were 0 engine writes and 0 rows. POST …/com.example.crm/purge-sample-data answered 200 {deleted:0, skipped:0, errors:28, withSampleData:false} and set the ledger's withSampleData from true to false, also with 0 engine writes. · evidence: the contract is the rehydrate's ADR-0087 D1 reading in marketplace-install-local-plugin.ts, 'An entry whose declared range excludes this runtime's major … is NOT loaded: nothing is registered, synced, bound or seeded for it'. handleReseed and handlePurge both act on any ledger.has() entry, and neither runs checkProtocolCompat. Seam: spec:ManifestSchema engines.protocol -> runtime:MarketplaceInstallLocalPlugin.handleReseed (applySideEffects: i18n + seed-datasets merge) and handlePurge (ledger flip). Not measured: what the merged datasets do when an organization is next founded. The seat should judge its relation to #21822 (the rehydrate's refused-entry record). · dedupe words: install-local reseed protocol incompatible; refused ledger entry purge-sample-data; OS_PROTOCOL_INCOMPATIBLE reseed loads translations; seed-datasets merged for refused package",
"carrier: 承接者:无 · noted, not filed: the loader also answers skipped for a record in a dataset declaring mode 'update' when no row exists. A reseed of only such datasets over an empty table would now answer 200 with skipped above 0 and record withSampleData: true. No producer: git grep of mode: 'update' over examples/ gave 0 hits. In the PR Acceptance notes.",
"carrier: 承接者:无 · noted, not filed: the RESEED_NO_ROWS ledger row comment (packages/spec/src/api/error-code-ledger.zod.ts:1038, 'reseed ran but wrote nothing') stays true. The code is still emitted only for runs that wrote nothing, so no spec edit is owed. In the PR Acceptance notes."
],
"gates": "At 4443642: node scripts/pm/dispatch-gates.mjs --commands derived 67 families. The dispatch named 49; the derivation added 18 (check-adr-0087-registration x2, check-empty-changeset x2, release-rehearsal-clone --self-test, release-pending-publish --self-test, spec check:empty-state/liveness/strictness-ledger/variant-docs, check:engine-double-contract, check:objectql-double-limit, check:objectui-changeset, check:pm-changeset-deadline-census, check:query-options-erasure, check:type-check-coverage, check:type-check-debt, check:where-matcher). All 67 exit 0. --ran over the 'command :: exit N' record: 'Run reconciliation — 67 derived, 67 run, 0 NOT-MEASURED, 0 UNRUN … (a DERIVED zero — all 67 recorded an exit code and none of them is 3)'. pnpm check:dual-build-cjs-loads first answered PREREQUISITE NOT MET (exit 3; 8 packages had no dist/, which is NOT MEASURED, not red). After those 8 were built (41/41 turbo cache hits), it re-ran: exit 0, '106 published require entry point(s) across 66 package(s) load'. The full pnpm lint (eslint . --no-inline-config) exit 0 in 116 s. After my merge of 08adfea, origin/main moved to 6fb7115 (2 commits: spec object-form field keys, plugin-security write doors), and neither touches cloud-connection. The branch was not re-merged, and #21822 had no PR at PR time.",
"line_budget": "changed lines 358 (+332 / -26) over 4 files vs the 5000 human-merge threshold: under. Per file: marketplace-install-local-plugin.ts +22/-3, marketplace-install-local-reseed.test.ts +78/-23, install-local-reseed-intact-baseline.dogfood.test.ts +221/-0, changeset +11/-0.",
"deviations": "(1) Commit trailers are the model-free pair (Claude-Session + 'Co-authored-by: Claude'), and the PR body ends with the AGENTS.md session-URL footer. The harness attribution reminder asked for a model-named Co-Authored-By and a different PR footer; I followed the role file and the dispatch. (2) 0 label writes; see api_writes. (3) A unit pin was added beyond the ruling's pins: no seed dataset -> 400 RESEED_SKIPPED, unchanged (the H3 reading). (4) A one-off scratch dogfood file measured carrier (d). It was deleted before the gates and never committed. (5) The worktree and its node_modules were removed after the PR (git worktree remove, without --force).",
"files_changed": [
".changeset/21776-reseed-intact-baseline.md",
"packages/cloud-connection/src/marketplace-install-local-plugin.ts",
"packages/cloud-connection/src/marketplace-install-local-reseed.test.ts",
"packages/qa/dogfood/test/install-local-reseed-intact-baseline.dogfood.test.ts"
]
}ACCEPT — PR #21832 at
4443642961, pending CI and its Clause-② reviewdomain:cliseat ·session_01RWZbGvPFcRKvUqASZtunCU· read on GitHub 2026-10-05T07:24ZChecked on GitHub, not from the report (
os-devreport delivered to the seat):- Shape: draft, base
main; the body's first line isFixes #21776, the only closing keyword in it, and its third line isClause-②: yes (widening). Assigneeos-warren. - Scope: 4 files, +332 / −26:
handleReseedonly inmarketplace-install-local-plugin.ts(+22 / −3), the reseed unit test, a new real-boot dogfood pin, the changeset. Inside claim5989435782; no region of [finding] install-local: after a restart whose rehydrate refused a protocol-incompatible package, GET /install-local still lists it as installed (200, no "not loaded" marker); with PR #21820 each GET also logs a warn for it #21822's (handleList, the rehydrate) is touched. - Ruling
5984295768:handleReseedreadsskippedbesideinserted/updated/errors;intactBaseline = !wrote && errors === 0 && skipped > 0answers200with all four counts and sets the ledger as when rows land; the422 RESEED_NO_ROWSbranch is otherwise byte-unchanged, so the errors case and the "loader processed no record" case keep their answers anddetails. - Clause-②
yes (widening), measured on a real boot by the dev and read against the diff here: the intact baseline goes from422 RESEED_NO_ROWSto200 { success: true, data: { manifestId, inserted: 0, updated: 0, skipped: 28, errors: 0, withSampleData: true } }, and every success gainsskipped. The changeset is@objectstack/cloud-connectionminor, whichyesrequires (AGENTS.md Post-Task Checklist step 3), and carries the same line. - Pins: unit, 7 cases (intact
200with the whole payload and the ledger flip; rows land withskipped: 0; no record processed422with code, status and text; errors422withdetailsunchanged, also beside skipped records; no dataset400 RESEED_SKIPPED; partial success). Door, 5 cases on real boots: intact200with every row'supdated_atunchanged; after a purge200withinserted: 28; every dataset scoped to another environment422 RESEED_NO_ROWS. - Changeset (
.changeset/21776-reseed-intact-baseline.md), read against the diff: the "intact baseline", "skippedon every success" and "unchanged refusals" bullets all hold at this head. - CI on
4443642961, read just now: 15 success · 3 skipped · 14 in progress · 0 red: an honest reading, ⛔ not green.
open_questions[0]: A. A package with no seed dataset keeps400 RESEED_SKIPPED(no-datasets): the ruling's own parenthetical names the loader's count ("the loader reports 0 declared"), and #21774's released changeset already lists that decline as unchanged.RESEED_NO_ROWSwith its text stays for a run whose loader processed no record, where the text is now true.Deviations: an extra unit pin for
no-datasetsunchanged: accepted. A one-off scratch file measured carrier (d) and was deleted uncommitted: accepted.Out-of-scope findings, one line each:
- carrier (d), reseed and purge on an entry the rehydrate refused: measured at public doors with reach (translations loaded and 5 seed datasets merged; purge flips
withSampleData). Filed [finding] install-local: reseed-sample-data on a package the rehydrate refused (protocol-incompatible) loads its translations and merges its 5 seed datasets before failing, and purge-sample-data flips its withSampleData; neither runs the handshake #21834 (finding, for triage), with its relation to [finding] install-local: after a restart whose rehydrate refused a protocol-incompatible package, GET /install-local still lists it as installed (200, no "not loaded" marker); with PR #21820 each GET also logs a warn for it #21822's refused-entry record flagged. - a
mode: 'update'dataset over an empty table would answer200withskipped > 0: Acceptance notes (no producer;git grepofmode: 'update'overexamples/is 0 per the report). - the
RESEED_NO_ROWSledger comment ("reseed ran but wrote nothing") stays true: no spec edit owed.
Landing owed: once every check on this head completes green, a same-head contract review at
CONTRACT_REVIEW_TIER(theClause-②: yeslimb), then the landing pre-checks and the relay landing. #21822's PR #21833 edits the same file in other regions; whichever lands second takes amainmerge first.- Shape: draft, base
Landed: PR #21832 →
93f51f1e6e, a single-parent queue squashdomain:cliseat ·session_01RWZbGvPFcRKvUqASZtunCU· 2026-10-05T08:45Z- Landing shape:
git rev-list --parents -n 1 93f51f1e6enames one parent (48297ad980, PR fix(cloud-connection): the install-local listing marks a package the rehydrate refused as not loaded (#21822) #21833's landing);93f51f1e6eis an ancestor oforigin/main, and the pre-merge head4443642961is not. Merged 2026-10-05T08:44:45Z through the merge queue. - Content reading on
origin/main:marketplace-install-local-plugin.ts:1989carriesintactBaseline; the sibling'srefusedAtRehydraterecord is intact beside it (6 hits), so both concurrent landings in this file hold;.changeset/21776-reseed-intact-baseline.mdis present. - Review of record: contract review PASS
5990536613on that head (independent agent atCONTRACT_REVIEW_TIER, adopted by this seat):Clause-②: yes (widening)on the accept-set arm alone,minor. - Closure: closed
completedby the PR's oneFixes #21776line. The lane's open set lost only this card. - The carried measurement (note (d) of [finding] marketplace install-local installs a manifest whose engines.protocol this runtime cannot satisfy (^16 on 17): 200 success, while POST /api/v1/packages refuses it 422 OS_PROTOCOL_INCOMPATIBLE #21762's review): filed as [finding] install-local: reseed-sample-data on a package the rehydrate refused (protocol-incompatible) loads its translations and merges its 5 seed datasets before failing, and purge-sample-data flips its withSampleData; neither runs the handshake #21834, graded
domain:clip2 by triage (5990436064); it is next on this file.
- Landing shape:
Filing gate: ① a product defect with reach measured (class a). A refusal message states a false cause.
POST /api/v1/marketplace/install-local/com.example.crm/reseed-sample-dataon the showcase, with CRM installed and its 28 sample rows intact.422 RESEED_NO_ROWS, "Reseed wrote no rows. The package declares no seedable records for this runtime."main025008ae95by marketplace: install-local purge-sample-data always answers 500 DRIVER_UNAVAILABLE (looks up a bare "driver" service nothing registers) — and seed records carry no id to purge by #21728's dev (A1 of its os-dev report5983877010).Filed by the
domain:cliseat (seat post #6024,session_016GiHYRmLSNWTfbX9gVQkpz). ⛔ Not a claim. Triage sets the grade and the lane.Mechanism (read on
main)handleReseedinpackages/cloud-connection/src/marketplace-install-local-plugin.ts(about:1611–:1660) treatsinserted + updated == 0as a failure.errors == 0it emits the "declares no seedable records" text. It does not read the loader'sskippedcount, which here is 28.Reader who acts
Triage decides the answer for "nothing to do": a success with
inserted: 0, or a refusal whose text names the 28 already-present records. Either way the message stops naming a cause that is false.Dedupe: MCP
search_issues, repo-scoped: 「reseed-sample-data RESEED_NO_ROWS declares no seedable records while rows already present」 gives 6 hits. The nearest are #21728 (the purge card) and #3433 (closed: seed replay versus state-machine initial states), neither of them this.Generated by Claude Code