@odatano/nightgate v0.10.0

@odatano/nightgate 0.10.0 (2026-07-24)

• AttestationVault passport hardening in ONE contract change (one redeploy):
  bindPassport rebind-takeover guard + registrar-gated pre-registration
  (registerPassport) against first-bind squatting.
• Deterministic batch apply order: submitContractCallBatch rewrites segment
  ids into call order before proving, so DEPENDENT calls run as one
  transaction. Lifts the 0.9.3 order-independence restriction.

Highlights

bindPassport rebind-takeover guard

• passport_bindings.insert overwrites and only payload_hash ownership was
  checked, so any attester could re-bind a bound passportId onto their own
  attestation and hijack the QR resolution (same takeover class as the 0.6.1
  attest() fix). An already-bound, unregistered id now re-binds only for
  the owner of its currently bound attestation; same-owner rebinding stays
  allowed.

Registrar-gated pre-registration (registerPassport)

• A new constructor locks the DEPLOYER's attester identity as registrar;
  the registrar-only circuit registerPassport(passportId, owner_id) fills
  passport_owners. Registered ids bind/re-bind ONLY for their registered
  attester: blocks first-bind squatting and recovers a squatted id (registrar
  re-points it, the owner rebinds over the foreign binding). Unregistered ids
  stay first-come-first-served (squattable by design).
• New OData action registerPassport(...): async job like
  grant/revokeDisclosure (30/h per session), result
  { passportId, ownerId, contractAddress, txHash }. Browser helper
  prepareRegisterPassport.

Deterministic batch apply order (submitContractCallBatch)

• The SDK randomizes each call's segment id while the ledger applies merged
  intents in ascending segment order, so a dependent batch only landed by
  luck (~1/3). NIGHTGATE now rewrites the batch's existing segment ids into
  call order before proving (ledger-v8 recomputes binding);
  attest -> bindPassport -> anchorContentRoot runs as ONE tx.
• Fail-closed: if the ordering cannot be established for a multi-call batch,
  the submission aborts BEFORE proving (nothing submitted) instead of
  silently proving in randomized order. Duplicate circuit names keep a random
  relative order among themselves. New srv/midnight/batch-segment-order.ts.

Notes

• Recompiled managed/ artifact (compactc 0.31.0), now 9 circuits
  (anchorContentRoot/proveFieldPredicate keys shifted too). No CDS schema
  change. The deploying session's secret defines the registrar: deploy from
  the session that should manage passport ownership.
• Consumer action (NIGHTPASS): redeploy the vault. A redeployed vault
  starts with an EMPTY ledger: switch the contract address AND re-create
  needed state (attest, bindPassport, anchorContentRoot, grants), ideally
  registering known passport ids first, otherwise QR resolution appears
  lost.
• Job-pool fix: grantDisclosure/revokeDisclosure/registerPassport now
  run in the HEAVY pool (full ZK proof generation; light since 0.9.0 was an
  oversight).
• Verified: 68 suites / 1231 tests + 34-check contract regression script,
  and TWO full live rounds on preprod (fresh vaults 81576468…/da9b0bcf…):
  deploy, first-try dependent 3-call batch with on-chain SUCCESS
  (txs 86bc9a0b…/4db4dea6…), registerPassport, bind on the registered
  id. Two first-try batches had ~11% combined odds under the old randomized
  order; the statistical 6/6-vs-0/6 stress proof runs consumer-side.