driver-turso remote: `aggregate` / `find` answer `[]` for a missing table or a missing column, where the local face refuses DATABASE_ERROR / 500 or INVALID_FIELD / INVALID_FILTER / 400 (the #8790 shape on the remote transport)

[objectstack-fleet]

Filing gate: ① a product defect with a named landing site and a reach:. Finding class (a). reach: is a named producer: REST GET /api/v1/data/:object (ObjectQL find → TursoDriver.find, remote face), and service-analytics' ObjectQL aggregate path (AnalyticsService → engine.aggregate → TursoDriver.aggregate, remote face).

Filed by the domain:engine execution seat 1 (session_01N8TPEsoJxPsdSdNKGnNGEN, os-warren) from the #20107 dev's out_of_scope_findings (os-dev-report 5868026272 on #20107, PR #20422). The readings are the dev's. ⛔ Filed bare: routing and grading belong to triage. ⛔ Not a claim.

What happens

Measured by the #20107 dev at 37414c5e8 (PR #20422's head, which predates nothing here), over one libSQL file, with the local TursoDriver over the same file as the control:

read	local face	remote face
aggregate on a mapped table that is really absent	DATABASE_ERROR / 500	[]
aggregate grouped by a declared field whose column is absent	INVALID_FIELD / 400	[]
find whose where names a declared field whose column is absent	INVALID_FILTER / 400	[]

So on a remote Turso database, a schema drift or a missing table reads as "there is no data". The local face and driver-sql refuse it loudly. This is the #8790 shape (closed for driver-sql) on the remote transport. The dev reports it applies to managed objects too, and that it predates PR #20422.

Where

packages/drivers/driver-turso/src/remote-transport.ts:

• the RemoteTransport.aggregate catch, which maps "no such table" / "no such column" to [];
• the RemoteTransport.find $select backstop, which returns [] when no projection is present.

Dedupe

search_issues "turso remote aggregate find returns empty result on missing table or missing column instead of error RemoteTransport" in objectstack-ai/objectstack, open and closed: 14 hits. None is this defect. #20107 is the source card, #20055 / #20094 / #19894 are other remote-face refusals, and #8790 / #11541 (closed) are the driver-sql twins.

Dedupe words: turso remote aggregate no such table empty · turso remote find no such column empty result · RemoteTransport invented empty read

[objectstack-fleet]

Path: an API a customer can call | 缺项 (no item reads a missing table or a drifted column through the remote Turso face) | P2

Triage: first grade — bug · priority:p2 · domain:engine · area:api · pm:queue. Serial after PR #20422, which edits the same file

Triage: lands in packages/drivers/driver-turso/src/remote-transport.ts ⇒ domain:engine. On origin/main 24b70859, the file has:

• the RemoteTransport.aggregate catch that maps no such table / no such column to [] (:1618-1619);
• the find backstop that swallows no such column (:1385).

Rationale:

• Wrong answer, silently. On a remote Turso database, a missing table or a drifted column reads as "there is no data". The local face over the same file refuses it: DATABASE_ERROR / 500, INVALID_FIELD / 400 and INVALID_FILTER / 400. It is the driver-sql: one unresolvable WHERE column, two answers — find() silently returns [] while count() throws a raw dialect error with no ADR-0112 envelope #8790 / driver-sql: an unresolvable WHERE column on aggregate() answers DATABASE_ERROR/500 where find() and count() answer INVALID_FILTER/400 — the #8790 refusal never reached the third read door #11541 shape on the remote transport.
• Reach. REST GET /api/v1/data/:object and the analytics aggregate path.
• p2. Runs but answers wrong, the same grade as the remote-face siblings driver-turso remote: inherited SqlDriver methods the remote face does not override answer from the placeholder :memory: Knex database — introspectSchema() returns no tables, distinct() a 500, findWithWindowFunctions() a raw SQLite error #20055, driver-turso remote: a malformed $between is refused with a bare Error (no code, no status) that names the field and echoes the comparand, where the local face answers INVALID_FILTER / 400, withheld #20094 and driver-turso remote planMediaColumnMove answers empty, so os migrate files-to-references reports nothing to move on a remote Turso database whose media columns are present #19894 (all p2, domain:engine).

Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-09-28T11:06Z. ⛔ Not a claim, ⛔ not a dispatch.

Serial after PR #20422 (#20107), which is open and edits remote-transport.ts (130 lines). The two changes do not depend on each other, so this stays pm:queue, not pm:blocked. Dispatch it on top of PR #20422's merge to avoid a same-file conflict.

Direction. Refuse on the remote face with the local face's codes; ⛔ don't invent new ones.

• The engine lane rule applies: the SQL family shares one filter compiler, so first check whether the local face's refusal is a shared seam the remote face can call. Don't re-derive it.
• Pin each of the three rows as a local-and-remote parity pair.

Duplicate check. A local corpus of 3,418 issues matched RemoteTransport|no such (table|column)|remote … [] / empty 34 times. The only open hits are #20107 (the source card) and the unrelated #20399 / #5930. None is this defect.

[objectstack-fleet]

Claim: PM loop round 23
Session: session_01N8TPEsoJxPsdSdNKGnNGEN
Account: os-warren (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-20424-turso-remote-invented-empty
Worktree: objectstack-issue-20424
Domain: domain:engine
Seat: domain:engine#1
File surface:

• packages/drivers/driver-turso/src/remote-transport.ts, only the two places where a missing table or column becomes []: RemoteTransport.aggregate's catch (the no such table / no such column → [] mapping) and RemoteTransport.find's $select backstop. Each is refused with the local face's code, through the local face's own seam where one is shared (the engine lane's SQL-family rule; ⛔ no second copy and no new codes);
• packages/drivers/driver-turso/src/turso-driver.ts, the remote read arms only, and only if the refusal must be raised there rather than in the transport;
• tests in packages/drivers/driver-turso (each of the card's three rows as a local-and-remote parity pair);
• .changeset/20424-*.md.

Stop on breach and explain in the report. ⛔ Not packages/drivers/driver-sql (the local face's refusal is read and called, not changed). ⛔ Not the constructor (#20200, PR #20447 in flight) nor reclaimSpace (landed). ⛔ Not packages/spec.
Container & model: S, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate, floor sonnet · default opus · ceiling fable)
Clause-②: no
Thread-read: 5868632265
Serial constraints cleared: at 2026-09-28T13:28Z, a census of the 15 open PRs' file lists finds one on driver-turso: PR #20447 (#20200, this seat), which touches turso-driver.ts's constructor, its mirror schema and its tests, and not remote-transport.ts. Triage's serial note (5868632265) asked for dispatch on top of PR #20422, which has landed (dbddf02c1). The newest Claim: on the lane's open pm:dispatched cards names no other driver-turso region.

[objectstack-fleet]

os-dev-report

{
  "issue": 20424,
  "status": "done",
  "branch": "claude/issue-20424-turso-remote-invented-empty",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20461",
  "session": "session_01N8TPEsoJxPsdSdNKGnNGEN",
  "premise_still_valid": true,
  "claim_check": "Newest Claim: on #20424 is 5870866980 ('Claim: PM loop round 23'). It names branch claude/issue-20424-turso-remote-invented-empty and reads 'Clause-②: no'. No claim was posted by this run. The run was cut by a usage limit after the empty-branch push and resumed on the coordinator's word, from the same worktree at the same base.",
  "summary": "On the Turso remote face, aggregate, find, findOne and count now answer a missing table or column the way the local face answers it over the same file: DATABASE_ERROR/500, INVALID_FIELD/400 or INVALID_FILTER/400, never [] or null. RemoteTransport.aggregate's catch (no such table / no such column -> []) is removed. RemoteTransport.find's backstop keeps the local ladder's projection rung, gains its ORDER BY rung, and throws the last rung's error instead of answering []. TursoDriver's remote read exits classify through a new private remoteReadFault: anything with a numeric status passes unchanged, and everything else goes to the inherited SqlDriver.aggregateBackendFault (the caller's own query for aggregate, the caller's where alone for find/findOne/count). H1 was confirmed at base 6e3e5462c for all three rows, and H4 holds for managed objects (table or column dropped under a synced object). H2: the aggregate catch came with the method's first version (101d5c345), with no comment, test, card or first-boot case, and the local face refuses both conditions, so nothing was kept. The find terminal [] was the pre-#8790 behaviour that its copy never shed, and its unit pin was replaced. H3 is falsified as worded: the local face has no pre-compile field check for this condition and classifies the backend error after execution. The compositions are protected and reachable, but the class predicate isUnresolvableColumnError is not exported from driver-sql, so the find/count exit reaches it through aggregateBackendFault with a where-only query (arm 1 cannot fire, arm 2 is unresolvableFilterColumnRefusal, the terminal is backendStatementFault). driver-sql and spec are untouched. Two bounded in-place fixes are declared in the PR: count's remote exit (DATABASE_ERROR -> INVALID_FILTER, the #8790 other half, which shares the exit) and the ORDER BY rung (without it the change would turn [] into a misattributed filter refusal). PR #20447 had not landed at either reading.",
  "tests": "Measured head daad09aa1 (branch merged with origin/main 3cf644938; driver-turso dependency closure rebuilt on the merged tree). pnpm --filter @objectstack/driver-turso exec vitest run --maxWorkers=2: 'Test Files 75 passed (75)' / 'Tests 2003 passed | 16 skipped (2019)'. pnpm --filter @objectstack/driver-turso typecheck: exit 0, and tsc --listFiles includes both touched test files (2 hits). New turso-local-remote-missing-table-column-parity.test.ts has 22 cases over a real @libsql/client file: DB with a local driver over the same file. It pins the card's 3 rows as local-and-remote code+status pairs, for a federated and a managed object each (6), plus 8 neighbours: findOne, find with projection and where, count (federated and managed), sum over the field, aggregate with that where, and the ORDER BY and projection recoveries answering the literal rows. It has 6 literal-answer controls, an undeclared-aggregate-function refusal with the same envelope on both faces and no statement sent, and a synthetic already-enveloped refusal passing the remote exit unchanged. remote-transport-unknown-select.test.ts: the '[]' pin is replaced, same input, opposite assertion. H1 probe at 6e3e5462c (throwaway, not committed) gave every row in the PR's H1 table. Reverse verification: the fix was committed first (a10647239). Every leg ran through node scripts/ablation-replace.mjs (anchor 1 -> 0, blob changed, restore proven by blob == HEAD and git diff HEAD empty), with an outer trap re-verifying against the HEAD blob. The subject resolves from src via a relative import with no alias, so there is no dist in the path. Suites: the new parity file, the unit file and the #20107 parity file (79 cases). A (aggregate [] catch restored): 6 failed / 73 passed, exactly the aggregate pairs. B (find terminal back to return []): 5 failed / 74 passed, the find/findOne/find-with-projection pairs and the replaced unit pin. C (remoteReadFault -> backendStatementFault alone): 10 failed / 69 passed, every 400 pair, while both 500 pairs stayed green. D (ORDER BY rung deleted): 1 failed / 78 passed, where the remote face refused with the unnamed INVALID_FILTER wording while the local face answered the rows. E (status gate deleted): 1 failed / 78 passed, the synthetic case re-worded INVALID_FIELD/400. All directions matched the test header's predictions. A first harness dry run was refused by ablation-replace itself (its replacement contained the anchor), so nothing was measured there. Lint, narrowed: eslint --no-inline-config --format json on the 4 changed .ts files gave 4 files, 0 errors, 0 warnings. (1) The population is read from eslint.config.mjs: **/*.{ts,tsx,mts,cts,js,jsx,mjs,cjs} minus ignores. (2) The file count is from the JSON (4). (3) Invariance: the effective parserOptions {ecmaVersion: latest, sourceType: module} have no project, and all 6 active rules are per-file AST rules, so type-aware linting is off and this diff cannot move a verdict on an untouched file. The full pnpm lint is CI's.",
  "gates": {
    "measured_head": "daad09aa1",
    "derivation": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack -> 63 commands (change set: 5 paths vs merge base 3cf644938; tree objectstack-ai/objectstack at daad09aa1)",
    "run": "63 of 63, each exit code written to a file before any pipe",
    "first_pass_exit_3": ["pnpm check:dual-build-cjs-loads", "pnpm check:lean-entry-closure", "pnpm check:type-check-debt"],
    "remedy": "pnpm turbo run build --filter='./packages/*' --filter='./packages/*/*' --concurrency=2 under the verify lock (71 of 71 tasks successful). All three were re-run: exit 0. check:dts-closure (71 packages swept), check:sourcemap-no-sources-content (68) and check:published-files were re-run on the full build: exit 0.",
    "final": "63 exit 0",
    "ran_verdict": "✓ 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).",
    "driver_conformance": "before (6e3e5462c): 'check-driver-conformance: OK — 50 covered cell(s), 0 in the DEBT ledger, 0 exempt.'; after (daad09aa1): 'check-driver-conformance: OK — 50 covered cell(s), 0 in the DEBT ledger, 0 exempt.'; the driver-turso row is ok in every column at both readings",
    "changeset_gates_on_pr_body": "check-changeset-no-major --event (the PR body): exit 0, '✓ LEVEL AXIS: this PR declares clause-② `no`'. check-closing-keyword-parity --body: exit 2 (its 'declarations bind' answer), bound only to 'Fixes #20424' by 2 of 3 parsers.",
    "not_measured": "The CI-only lanes named by the derivation outside its 63 commands (11 whole-root families, 6 workflow-valued families such as check-shard-attestation --emit and check-issue-citations --census) are declared to CI.",
    "ci": "in_progress (not awaited). At the report: 31 check-runs on daad09aa1, 9 success, 3 skipped, 19 in_progress, 0 failed."
  },
  "line_budget": "552 changed lines (+516 / -36) over 5 files vs the 5000 human-merge threshold: under. No governed surface, no skills/** file, no ledger line budget in play.",
  "files_changed": [
    ".changeset/20424-turso-remote-missing-table-column-refused.md (+27 / -0)",
    "packages/drivers/driver-turso/src/remote-transport.ts (+39 / -17)",
    "packages/drivers/driver-turso/src/turso-driver.ts (+90 / -15)",
    "packages/drivers/driver-turso/src/remote-transport-unknown-select.test.ts (+15 / -4)",
    "packages/drivers/driver-turso/src/turso-local-remote-missing-table-column-parity.test.ts (+345 / -0)"
  ],
  "deviations": [
    "H3 is falsified as worded, and the route changed to the order's intent. There is no pre-compile field check on the local face for this condition (the ingress field checks pass because the field is declared). The local face classifies after execution. The reachable shared seam is the protected SqlDriver.aggregateBackendFault. It is called for aggregate with the caller's query, and for find/findOne/count with { where } alone, because the predicate the local count/findRows exit uses (isUnresolvableColumnError) is not exported from driver-sql and driver-sql may not be edited. There is no second copy of any classification.",
    "remoteReadFault restates one line of backendStatementFault's 'already ours' gate (a numeric status passes unchanged) BEFORE aggregateBackendFault reads a message. A remote door compiles and executes inside one transport call, while the local face guards only execution. No real refusal text distinguishes it today, so it is held by a synthetic pin (ablation E).",
    "Bounded in-place fix 1, declared in the PR body: count's remote exit, DATABASE_ERROR/500 -> INVALID_FILTER/400. It is not one of the card's three rows. It shares remoteReadExit with find, and leaving it would split one list view's two halves (the #8790 rule). All four conditions hold, inside the claimed turso-driver.ts remote read arms.",
    "Bounded in-place fix 2, declared in the PR body: the ORDER BY rung in RemoteTransport.find's ladder. Without it, removing the [] terminal turns a sort on a missing column into a misattributed INVALID_FILTER (measured, leg D), where the local face answers the rows. It is the same catch and the local ladder's second rung.",
    "The unit pin 'still returns empty when even SELECT * fails (e.g. unknown table)' in remote-transport-unknown-select.test.ts is replaced (fixture triage: it pinned the branch removed), not dropped.",
    "One docblock sentence in refuseRemoteColumnMap's header (turso-driver.ts, outside the read arms) moved to the past tense, because the [] it described is now a refusal. Prose only.",
    "origin/main moved to b28550818 after the measured merge daad09aa1: one commit, 6 files, additive in packages/spec only, with no driver-* path. It was not re-merged (AGENTS.md §10 scope), and gates and tests are quoted at daad09aa1.",
    "Commit trailers use AGENTS.md's model-free pair (Claude-Session + Co-authored-by: Claude), not the harness reminder's model-named Co-Authored-By. The PR body ends in AGENTS.md's session-URL footer, not the harness reminder's PR ending. AGENTS.md and os-dev.md take precedence, and the pre-push hook refuses a model identifier.",
    "PR labels documentation, size/l, tests and tooling were written by other actors (the labelers). This run wrote only the PR assignee (os-warren), as the order's budget names. No label was written."
  ],
  "mcp_calls": "0 — no MCP GitHub tool was called",
  "api_writes": "3 — all through the fleet relay (route dispatch, session_01N8TPEsoJxPsdSdNKGnNGEN), each one POST /repos/objectstack-ai/objectstack/dispatches executed by the fleet-write workflow: (1) pr_create, POST /repos/objectstack-ai/objectstack/pulls, draft #20461 (run 36439771558), body read back byte-identical (15407 bytes), so no issue_patch was spent; (2) label-write --assign os-warren, POST /repos/objectstack-ai/objectstack/issues/20461/assignees (run 36439879142), read back MATCHES; (3) post-stamped comment, POST /repos/objectstack-ai/objectstack/issues/20424/comments (this report). Plus git push of the branch 4 times (empty-branch probe, a10647239, b635e43f1, daad09aa1), all fast-forward with no force, which are not REST writes.",
  "open_questions": [
    {
      "question": "Clause-② for this change. The claim reads 'Clause-②: no' (a read that invented [] was a false answer, not a published accept set), and the changeset ships patch. The precedent splits. #11541 (ef52884a8, aggregate 500 -> INVALID_FIELD/400) shipped patch with no narrowing. #8790 (716ac9bf8, find [] -> INVALID_FILTER/400, the same shape as row 3) declared 'BREAKING accept-set narrowing', minor, ADR-0087 registered, with migration entry driver-sql-unresolvable-where-column-refused. That registered entry's surface already names 'driver-sql (and its TursoDriver / SqliteWasmDriver subclasses)' and says find()/findOne() 'now refuse with INVALID_FILTER / 400'. Separately, RemoteTransport.find/aggregate, exported from the package root and used standalone, now raise where they answered [].",
      "options": [
        "A keep 'Clause-②: no' + patch: the narrowing was already declared and registered for TursoDriver by name, so this delivers a published contract rather than narrowing one; the [] was a false answer",
        "B 'Clause-②: yes (narrowing)', BREAKING, following #8790's own declaration: the changeset gains the migration text and an ADR-0087 disposition (the existing entry, or a new one for the remote face)"
      ],
      "recommendation": "A, because the published migration entry already tells TursoDriver users that find/findOne refuse, so no reader of the contract is surprised, and the remote face was the declared-not-enforced gap. The seat owns the line, so it is flagged, not changed."
    }
  ],
  "out_of_scope_findings": [
    "carrier: 承接者:无 · noted, not filed — isUnresolvableColumnError is not exported from @objectstack/driver-sql, so a subclass that runs its own statements (TursoDriver's remote face) reaches the local find/count exit only through aggregateBackendFault with a where-only query. An exported predicate, or a protected where-exit on SqlDriver shared by count and findRows, would be the plainer seam. It needs a driver-sql edit. Acceptance notes.",
    "carrier: 承接者:无 · noted, not filed — RemoteTransport.find keeps its own inline unresolvable-column recognizer (no such column, or column + does not exist) to gate its ladder: a pre-existing copy of the predicate without the MySQL arm, which libSQL never speaks. Acceptance notes.",
    "carrier: 承接者:无 · noted, not filed — if a ladder rung fails with an error that is not a column error, the local face refuses INVALID_FILTER (unnamed wording) and the remote face now answers DATABASE_ERROR/500. Unmeasured, with no known producer. Acceptance notes."
  ]
}

---

Generated by Claude Code

[objectstack-fleet]

Claim: PM loop round 23
Session: session_01N8TPEsoJxPsdSdNKGnNGEN
Account: os-warren (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-20424-turso-remote-invented-empty
Worktree: objectstack-issue-20424
Domain: domain:engine
Seat: domain:engine#1
File surface:

• packages/drivers/driver-turso/src/remote-transport.ts, only the two places where a missing table or column becomes []: RemoteTransport.aggregate's catch (the no such table / no such column → [] mapping) and RemoteTransport.find's $select backstop. Each is refused with the local face's code, through the local face's own seam where one is shared (the engine lane's SQL-family rule; ⛔ no second copy and no new codes);
• packages/drivers/driver-turso/src/turso-driver.ts, the remote read arms: the exits of aggregate, find, findOne and count through one private classifier over SqlDriver.aggregateBackendFault. That includes count's remote exit (DATABASE_ERROR → INVALID_FILTER, the driver-sql: one unresolvable WHERE column, two answers — find() silently returns [] while count() throws a raw dialect error with no ADR-0112 envelope #8790 other half, which shares the exit), and one past-tense docblock sentence in refuseRemoteColumnMap's header;
• RemoteTransport.find's recovery ladder gains the local ladder's ORDER BY rung, so a sort on a missing column answers the rows as the local face does;
• tests in packages/drivers/driver-turso (each of the card's three rows as a local-and-remote parity pair, for a federated and a managed object), and the one remote-transport-unknown-select.test.ts pin that asserted the removed [], replaced;
• .changeset/20424-*.md.

Stop on breach and explain in the report. ⛔ Not packages/drivers/driver-sql (the local face's refusal is read and called, not changed). ⛔ Not the constructor (#20200, PR #20447 in flight) nor reclaimSpace (landed). ⛔ Not packages/spec.
Container & model: S, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate, floor sonnet · default opus · ceiling fable)
Clause-②: no (narrowing)
Thread-read: 5872594845
Serial constraints cleared: at 2026-09-28T13:28Z, a census of the 15 open PRs' file lists finds one on driver-turso: PR #20447 (#20200, this seat), which touches turso-driver.ts's constructor, its mirror schema and its tests, and not remote-transport.ts. Triage's serial note (5868632265) asked for dispatch on top of PR #20422, which has landed (dbddf02c1). The newest Claim: on the lane's open pm:dispatched cards names no other driver-turso region.

Amended (this re-posted claim supersedes 5870866980's file surface and its Clause-② line). Seat answers to the os-dev report on #20424 (PR #20461):

1. Open question 1, Clause-②: B. Before this change, a remote-face read that named a missing table or column answered [] (or null); now it refuses. That is a narrowing of what the remote face accepted. driver-sql: one unresolvable WHERE column, two answers — find() silently returns [] while count() throws a raw dialect error with no ADR-0112 envelope #8790 declared the same shape (find [] → INVALID_FILTER / 400) a BREAKING accept-set narrowing under a maintainer ruling. Its registered entry, driver-sql-unresolvable-where-column-refused, already names TursoDriver as a subclass that "now refuses", and prescribes the remedy this change needs (name a real column, or run schema sync). So:
   • the line reads no (narrowing);
   • the changeset ships minor with the BREAKING banner;
   • the ADR-0087 marker is not-required (already-registered driver-sql-unresolvable-where-column-refused). The aggregate legs are the same drifted-schema family with the same remedy, and the marker's why says so.
     The dev's A reading (the contract was already published) is true of the entry, but remote-face users meet the refusal for the first time here, so the banner is owed.
2. The two bounded in-place fixes are accepted: count's remote exit, and the ORDER BY rung. Both are in the surface above.
3. The H3 route is accepted: the protected aggregateBackendFault with a where-only query for find / findOne / count, no driver-sql edit, no second copy of a classification. The unexported isUnresolvableColumnError stays an Acceptance note.

[objectstack-fleet]

os-dev-report

{
  "issue": 20424,
  "round": "patch round 1 (seat ruling in amended claim 5872640432: Clause-② B)",
  "status": "done",
  "branch": "claude/issue-20424-turso-remote-invented-empty",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20461",
  "session": "session_01N8TPEsoJxPsdSdNKGnNGEN",
  "premise_still_valid": true,
  "claim_check": "Newest Claim: on #20424 is the amended 5872640432 (created after 5870866980 and after os-dev-report 5872594845). It names branch claude/issue-20424-turso-remote-invented-empty and reads 'Clause-②: no (narrowing)'. It accepts the two bounded in-place fixes and the H3 route. No claim was posted by this run.",
  "head": "3b56eb769",
  "summary": "The changeset and the PR body changed; no code or test changed. `git diff daad09aa1 3b56eb769 -- packages/drivers` is empty. Changeset: '@objectstack/driver-turso' patch -> minor. 'Clause-②: no' -> 'Clause-②: no (narrowing)'. The ADR-0087 marker is 'not-required (already-registered driver-sql-unresolvable-where-column-refused)', and its reason says the entry names TursoDriver and prescribes this remedy, and that the aggregate legs are the same drifted-schema family with the same remedy. It gains the **BREAKING** banner in the launch-window form (copied from the repo's narrowing changesets, e.g. 20418's), and a FROM -> TO line: remote aggregate/find/findOne/count reads that answered []/null now refuse with DATABASE_ERROR/500, INVALID_FIELD/400 or INVALID_FILTER/400 as the local face does. The line names the fix (run schema sync, or name a column the table has) and says RemoteTransport.find/aggregate, exported from the package root, now raise where they answered []. PR body, via the relay issue_patch, read back byte-identical (18144 bytes, footer kept): the Clause-② line reads 'no (narrowing)', with a seat-ruling line naming claim 5872640432 under it. The 'Clause-② and the changeset' section now says minor, the banner and the disposition. The Files line reads +33 / -0 for the changeset (+522 / -36 total), a '## Patch round 1' section records this round and its gate exits, and one Acceptance note on main's movement is updated. origin/main was merged at 2304b1608 (a true merge commit, 5 incoming commits, no driver-* path), giving head 3b56eb769.",
  "tests": "No code or test changed. Because the merge moved packages/spec (a dependency), the driver-turso dependency closure was rebuilt on the merged tree at 3b56eb769 under the verify lock. pnpm --filter @objectstack/driver-turso exec vitest run --maxWorkers=2 gave 'Test Files 75 passed (75)' / 'Tests 2003 passed | 16 skipped (2019)', and pnpm --filter @objectstack/driver-turso typecheck exited 0 (os-verify-lock VERDICT command-exit 0). Round-0 ablations and the full 63-gate run at daad09aa1 stand: packages/drivers is byte-identical.",
  "gates": {
    "measured_head": "3b56eb769",
    "pnpm check:adr-0087-registration": "exit 0 — '✓ check-adr-0087-registration: 1 declared-breaking changeset(s), each carrying an ADR-0087 disposition.' / '.changeset/20424-turso-remote-missing-table-column-refused.md [BREAKING+clause-②-narrowing] not-required (already-registered)'; self-test 441 assertions",
    "check-changeset-no-major --base origin/main --event (the new PR body)": "exit 0 — '✓ This diff introduces no `major` bump.' / '✓ LEVEL AXIS: this PR declares clause-② `no (narrowing)`, and no package whose `packages/**/src/**` it moves is graded `patch`.' (direction arm narrowing, ships minor in the launch window)",
    "check-empty-changeset --base origin/main": "exit 0 — '✓ No empty-frontmatter changeset introduced by this diff (1 declaring changeset(s) added).' / '✓ No changeset from the merge base modified or deleted by this diff (#17712).'",
    "check-closing-keyword-parity --body (the new PR body)": "exit 2 (its 'declarations bind' answer), bound only to 'Fixes #20424'",
    "full_gate_run": "not re-run this round: the round-0 run (63 of 63 exit 0, --ran 0 NOT-MEASURED) was at daad09aa1, and packages/drivers is byte-identical at 3b56eb769. The change set this round adds the changeset edit plus main's merged commits, which are CI's",
    "ci": "in_progress (not awaited). At the report: 38 check-runs on 3b56eb769, 19 success, 5 skipped, 14 in_progress, 0 failed. The round-0 head daad09aa1 runs were cancelled at 15:04-15:05Z, superseded by the push of 3b56eb769 (Lint & Repo Gates and Test Core shards 1/3/4 cancelled; the 'Test Core' rollup reads failure over the cancelled shards — its log was not read)."
  },
  "line_budget": "This round: the changeset +6 lines net (27 -> 33). Total vs origin/main: +522 / -36 over 5 files, under the 5000 human-merge threshold. No governed surface.",
  "files_changed": [
    ".changeset/20424-turso-remote-missing-table-column-refused.md (+33 / -0 vs origin/main; this round +8 / -2)"
  ],
  "deviations": [
    "origin/main moved once more, to fbeb56e4b, after the round's merge at 2304b1608: one commit (a packages/spec test pin, packages/spec/vitest.repo-tests.json, scripts/cross-package-test-inputs.mjs, turbo.json), with no driver-* path. It was not re-merged, and the gates are quoted at 3b56eb769.",
    "The driver-turso tests and typecheck were re-run on the merged tree although the round asked for no code or test change (nothing in them changed). The merge moved packages/spec, which driver-turso depends on (AGENTS.md §10: rebuild what the merge touched).",
    "Commit trailers use AGENTS.md's model-free pair (Claude-Session + Co-authored-by: Claude), not the harness reminder's model-named Co-Authored-By, and the PR body keeps AGENTS.md's session-URL footer."
  ],
  "mcp_calls": "0 — no MCP GitHub tool was called",
  "api_writes": "2 — both through the fleet relay (route dispatch, session_01N8TPEsoJxPsdSdNKGnNGEN), each one POST /repos/objectstack-ai/objectstack/dispatches executed by the fleet-write workflow: (1) issue_patch on 20461 with body, PATCH /repos/objectstack-ai/objectstack/issues/20461 (run 36441821070), read back byte-identical (18144 bytes); (2) post-stamped comment, POST /repos/objectstack-ai/objectstack/issues/20424/comments (this report). Plus git push of c9c3293c8 and the merge 3b56eb769 (one fast-forward push, no rebase, no force), which is not a REST write.",
  "open_questions": [],
  "out_of_scope_findings": []
}

---

Generated by Claude Code

[objectstack-fleet]

Claim: PM loop round 23
Session: session_01N8TPEsoJxPsdSdNKGnNGEN
Account: os-warren (the seat's linked user as GET /user answers it; always the card's assignee)
Branch: claude/issue-20424-turso-remote-invented-empty
Worktree: objectstack-issue-20424
Domain: domain:engine
Seat: domain:engine#1
File surface:

• packages/drivers/driver-turso/src/remote-transport.ts, only the two places where a missing table or column becomes []: RemoteTransport.aggregate's catch (the no such table / no such column → [] mapping) and RemoteTransport.find's $select backstop. Each is refused with the local face's code, through the local face's own seam where one is shared (the engine lane's SQL-family rule; ⛔ no second copy and no new codes);
• packages/drivers/driver-turso/src/turso-driver.ts, the remote read arms: the exits of aggregate, find, findOne and count through one private classifier over SqlDriver.aggregateBackendFault. That includes count's remote exit (DATABASE_ERROR → INVALID_FILTER, the driver-sql: one unresolvable WHERE column, two answers — find() silently returns [] while count() throws a raw dialect error with no ADR-0112 envelope #8790 other half, which shares the exit), and one past-tense docblock sentence in refuseRemoteColumnMap's header;
• RemoteTransport.find's recovery ladder gains the local ladder's ORDER BY rung, so a sort on a missing column answers the rows as the local face does;
• amended, text only: the one comment in remote-transport.ts (about :2909–:2911) that still says find()'s backstop swallows a no such column into [], and the one header prediction line in turso-local-remote-external-object-parity.test.ts (about :48). This PR made both false;
• tests in packages/drivers/driver-turso (each of the card's three rows as a local-and-remote parity pair, for a federated and a managed object), and the one remote-transport-unknown-select.test.ts pin that asserted the removed [], replaced;
• .changeset/20424-*.md.

Stop on breach and explain in the report. ⛔ Not packages/drivers/driver-sql (the local face's refusal is read and called, not changed). ⛔ Not the constructor (#20200, PR #20447 in flight) nor reclaimSpace (landed). ⛔ Not packages/spec.
Container & model: S, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate, floor sonnet · default opus · ceiling fable)
Clause-②: no (narrowing)
Thread-read: 5872895784
Serial constraints cleared: at 2026-09-28T13:28Z, a census of the 15 open PRs' file lists finds one on driver-turso: PR #20447 (#20200, this seat), which touches turso-driver.ts's constructor, its mirror schema and its tests, and not remote-transport.ts. Triage's serial note (5868632265) asked for dispatch on top of PR #20422, which has landed (dbddf02c1). The newest Claim: on the lane's open pm:dispatched cards names no other driver-turso region.

Amended again (this re-posted claim supersedes 5872640432's file surface; Clause-②: no (narrowing) is unchanged). The at-tier review 5873239611 on PR #20461 PASSes 3b56eb769 and flags two things this round fixes:

• a comment and a test-header line this PR made false;
• PR fix(driver-turso)!: new TursoDriver refuses syncUrl under a forced remote mode, and sync with no syncUrl (#20200) #20447 (driver-turso: new TursoDriver accepts syncUrl / sync under mode: 'remote' and ignores them — isSyncEnabled() answers true, no sync runs, and sync() rejects SYNC_NOT_SUPPORTED #20200) landed after that head's CI started, so the combined tree is unmeasured. origin/main is merged in the same round and the driver-turso suite re-runs on it.

The review's migration-entry item (the aggregate legs are not named by driver-sql-unresolvable-where-column-refused's text) is a packages/spec edit outside this claim; the seat files it as a spec-lane card.

First amendment, kept for the record (5872640432 superseded 5870866980's file surface and its Clause-② line). Seat answers to the os-dev report on #20424 (PR #20461):

1. Open question 1, Clause-②: B. Before this change, a remote-face read that named a missing table or column answered [] (or null); now it refuses. That is a narrowing of what the remote face accepted. driver-sql: one unresolvable WHERE column, two answers — find() silently returns [] while count() throws a raw dialect error with no ADR-0112 envelope #8790 declared the same shape (find [] → INVALID_FILTER / 400) a BREAKING accept-set narrowing under a maintainer ruling. Its registered entry, driver-sql-unresolvable-where-column-refused, already names TursoDriver as a subclass that "now refuses", and prescribes the remedy this change needs (name a real column, or run schema sync). So:
   • the line reads no (narrowing);
   • the changeset ships minor with the BREAKING banner;
   • the ADR-0087 marker is not-required (already-registered driver-sql-unresolvable-where-column-refused). The aggregate legs are the same drifted-schema family with the same remedy, and the marker's why says so.
     The dev's A reading (the contract was already published) is true of the entry, but remote-face users meet the refusal for the first time here, so the banner is owed.
2. The two bounded in-place fixes are accepted: count's remote exit, and the ORDER BY rung. Both are in the surface above.
3. The H3 route is accepted: the protected aggregateBackendFault with a where-only query for find / findOne / count, no driver-sql edit, no second copy of a classification. The unexported isUnresolvableColumnError stays an Acceptance note.

[objectstack-fleet]

os-dev-report

{
  "issue": 20424,
  "round": "patch round 2 (at-tier review 5873239611 PASS at 3b56eb769; seat's amended claim 5873267997)",
  "status": "done",
  "branch": "claude/issue-20424-turso-remote-invented-empty",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20461",
  "session": "session_01N8TPEsoJxPsdSdNKGnNGEN",
  "premise_still_valid": true,
  "claim_check": "Newest Claim: on #20424 is 5873267997 (after 5872640432 and os-dev-report 5872895784). It names branch claude/issue-20424-turso-remote-invented-empty, reads 'Clause-②: no (narrowing)', and amends the file surface to add the two text-only fixes. No claim was posted by this run.",
  "head": "0c47b7bbe",
  "summary": "Text only plus a merge, with no logic or test-assertion change. (1) remote-transport.ts, the $-prefixed-key comment (was :2909-2911): the sentence saying find()'s no-such-column backstop 'swallows the error into [] anyway' under SQLITE_DQS=0 now says the backstop refuses, as the local face does (INVALID_FILTER / 400, see #20424), and that it used to answer [] as a second road to the same answer. Nothing else in the block changed (+6 / -5). (2) turso-local-remote-external-object-parity.test.ts:47-49, the #20107 header's first ablation prediction: aggregate now refuses the missing table as DATABASE_ERROR / 500 instead of answering [] in place of the sum (+3 / -2). (3) origin/main merged at acd009521 with a true merge commit, carrying PR #20447 (bea6d2ea3). The merge was clean, with no regen pending. Head 0c47b7bbe. (4) PR body via issue_patch, read back byte-identical (20506 bytes, session-URL footer kept). The #20447 Acceptance note now says it landed as bea6d2ea3, is merged here, and driver-turso was re-tested on the combined tree. The refuseRemoteColumnMap line now also names the added parenthetical sentence. The Files list and totals read 6 files, +531 / -43. A '## Patch round 2' section records this round.",
  "tests": "Combined tree 0c47b7bbe, after rebuilding the driver-turso dependency closure under the verify lock: pnpm --filter @objectstack/driver-turso test gave 'Test Files 76 passed (76)' / 'Tests 2036 passed | 18 skipped (2054)' (the added file and cases are #20447's), and pnpm --filter @objectstack/driver-turso typecheck exited 0 (os-verify-lock VERDICT command-exit 0). Narrowed eslint --no-inline-config --format json over the 5 changed .ts files gave 5 files, 0 errors, 0 warnings (same config evidence as round 0: no parserOptions.project, 6 per-file AST rules, so no untouched file's verdict can move).",
  "gates": {
    "measured_head": "0c47b7bbe",
    "derivation": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack -> 63 commands (change set: 6 paths vs merge base acd009521; +531 / -43), the same 63 as round 0",
    "build_first": "pnpm turbo run build --filter='./packages/*' --filter='./packages/*/*' --concurrency=2 under the verify lock: the first call was cut by its own 590s timeout after queueing (32 of 42 tasks done, exit 124, no orphaned process left), and the resumed call ran 71 of 71 successful (32 cached)",
    "run": "63 of 63, each exit code written to a file before any pipe",
    "final": "63 exit 0 on the first pass (no exit 3: the full build preceded the run)",
    "ran_verdict": "✓ 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).",
    "driver_conformance": "at 0c47b7bbe: 'check-driver-conformance: OK — 50 covered cell(s), 0 in the DEBT ledger, 0 exempt.' (unchanged from 6e3e5462c and daad09aa1)",
    "check-changeset-no-major --event (the round-2 PR body)": "exit 0 — '✓ LEVEL AXIS: this PR declares clause-② `no (narrowing)`, and no package whose `packages/**/src/**` it moves is graded `patch`.'",
    "check-closing-keyword-parity --body (the round-2 PR body)": "exit 2 (its 'declarations bind' answer), bound only to 'Fixes #20424'",
    "ci": "in_progress (not awaited). At the report: 41 check-runs on 0c47b7bbe, 32 success, 5 skipped, 4 in_progress, 0 failed, 0 cancelled."
  },
  "line_budget": "This round: +9 / -7 over 2 files (both comments). Total vs origin/main: +531 / -43 over 6 files, under the 5000 human-merge threshold. No governed surface.",
  "files_changed": [
    "packages/drivers/driver-turso/src/remote-transport.ts (this round +6 / -5, comment only)",
    "packages/drivers/driver-turso/src/turso-local-remote-external-object-parity.test.ts (this round +3 / -2, header comment only)"
  ],
  "deviations": [
    "Found and left unchanged, because it is outside the amended surface ('the one header prediction line ... about :48'): the same #20107 header's third prediction, 'A filter on the renamed field answers []' (turso-local-remote-external-object-parity.test.ts, around :54), is also out of date after this PR. With the column-map refusal deleted, that filter would now be refused INVALID_FILTER / 400. It is prediction prose, and no assertion depends on it. It is named in the PR body's Patch round 2 section.",
    "Commit trailers use AGENTS.md's model-free pair (Claude-Session + Co-authored-by: Claude), not the harness reminder's model-named Co-Authored-By, and the PR body keeps AGENTS.md's session-URL footer."
  ],
  "mcp_calls": "0 — no MCP GitHub tool was called",
  "api_writes": "2 — both through the fleet relay (route dispatch, session_01N8TPEsoJxPsdSdNKGnNGEN), each one POST /repos/objectstack-ai/objectstack/dispatches executed by the fleet-write workflow: (1) issue_patch on 20461 with body, PATCH /repos/objectstack-ai/objectstack/issues/20461 (run 36449236174), read back byte-identical (20506 bytes); (2) post-stamped comment, POST /repos/objectstack-ai/objectstack/issues/20424/comments (this report). Plus one git push of ecca6921c and the merge 0c47b7bbe (fast-forward, no rebase, no force), which is not a REST write.",
  "open_questions": [],
  "out_of_scope_findings": [
    "carrier: the seat (this card's surface) · noted, not filed — turso-local-remote-external-object-parity.test.ts header, third reverse-verification prediction ('A filter on the renamed field answers []') is out of date after this PR (would now be INVALID_FILTER / 400). Prediction prose only, outside the amended surface."
  ]
}

---

Generated by Claude Code