fix(driver-turso): refuse auto_number writes on the remote transport (#6944)

[claude]

Closes #6944.

Ruling carried out

Triage ruled this card at 06:46Z on 2026-08-09 (comment 5230216863): disposition B — explicit refusal, per the #3724 precedent for "a driver face that boots but silently fails to deliver a declared capability". Implementing autonumber on the remote transport (A) stays behind the appetite door. The issue body still poses A-or-B; it is stale and the comment governs.

The defect

TursoDriver picks its transport from url. Local and embedded-replica inherit SqlDriver's write path and issue record numbers from _objectstack_sequences. Remote overrides the write path to RemoteTransport, which builds its own INSERT and never enters fillAutoNumberFields — so auto_number was only a column mapped to TEXT there. Measured on main @ 2f3e79351:

REMOTE create      -> RESOLVED case_number=null
REMOTE bulkCreate  -> RESOLVED [null, null]
REMOTE upsert      -> RESOLVED case_number=null
LOCAL  create      -> RESOLVED case_number="CASE-00001"

Nothing upstream catches it: supports.autonumber is true on this face (inherited via ...super.supports), so the engine defers generation to the driver and never runs its own fallback.

The change

TursoDriver.create / bulkCreate / upsert refuse, in remote mode, a write that would need a number this face cannot issue — NOT_IMPLEMENTED / 501, the same class this package already gives an uncompilable aggregate function (#5907) or an unemittable date bucket (#6212), per ADR-0112. Refused before the statement is built, so it costs no round trip.

Raised on the driver, not the transport — measured, not assumed. RemoteTransport.create(object, data) takes no schema and caches none (arity pinned in the suite), so it cannot tell an auto_number column from any other TEXT one. TursoDriver can: registerRemoteFieldMetadata → registerExternalObject classifies fields at remote schema-sync time and populates autoNumberFields — verified live in remote mode.

Deliberately not refused: a caller-supplied number (seed replay / preserveAudit import — the generate predicate is fillAutoNumberFields' own, reused verbatim); and a merging upsert, which keeps the number already in the column. Only the provably-inserting upsert shape is refused. The residue — an id-bearing upsert that turns out to insert — is pinned as a [known residue, not fixed] assertion rather than left in a comment.

The two absence-assertions, rewritten rather than deleted

#6932 (turso-autonumber-resync.test.ts) and #6943 (turso-autonumber-batch-resync.test.ts) each pinned "RemoteTransport has no autonumber surface at all". Deleting either would remove the only guard against a future silent half-implementation inside the transport; leaving them unchanged would keep a green assertion beside a claim that no longer holds. Both keep the surface probe verbatim and gain the refusal pin (code and status, never a bare toThrow — ADR-0112 / #6144).

Scope

packages/drivers/driver-turso only. All three faces pinned (#6203): local and replica still generate. driver-memory / driver-mongodb are inside the #5499 freeze, declare no supports.autonumber, and are unrelated.

Reverse verification

Predicted before running: with the refusal body emptied, every REMOTE refusal case fails through the harness's "it resolved" branch, not on a code/status comparison; everything else stays green. Measured: 8 failed | 923 passed — the 6 cases in the new suite plus the 2 rewritten pins, all through "it resolved"/"expected a refusal", and each failure message reprints the defect with "case_number":null inside it.

🤖 Generated with Claude Code

https://claude.ai/code/session_01UtmF8wiWUuHFcezsXhYygb

---

Generated by Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
objectstack	Ignored		Aug 9, 2026 2:30pm

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/driver-turso.

6 hand-written doc(s) reference the affected code and may need an implementation-accuracy re-verification:

• content/docs/data-modeling/drivers.mdx (via @objectstack/driver-turso)
• content/docs/deployment/cli.mdx (via @objectstack/driver-turso)
• content/docs/deployment/environment-variables.mdx (via @objectstack/driver-turso)
• content/docs/deployment/self-hosting.mdx (via @objectstack/driver-turso)
• content/docs/getting-started/glossary.mdx (via @objectstack/driver-turso)
• content/docs/plugins/packages.mdx (via @objectstack/driver-turso)

Advisory only. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs origin/main → pass the list as args.docs.