Cardano Updates

yaci-store stores cursor_.era as a numeric ordinal (7 = Conway), not the
string "Conway". Map the ordinal to an era name and gate the hard-fork
readiness note on era 7, so crossing into the Conway / Van Rossem window
is detected correctly during the mainnet full-sync regression.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

cardano-crypto-wallet links against libsodium for Argon2id KDF and
XChaCha20-Poly1305 AEAD. Without this override, the Windows
cross-compiled DLL links against the default system libsodium, which
is not available in the Wine iserv environment (only libsodium-vrf is
present there, used by all other crypto packages).

This caused the iserv to crash (exit 1) at TH-module boundaries when
compiling cardano-wallet for Windows/ucrt64.

- wallet-key-export.hs: pattern-match HashedCredentials instead of
  tuple; V2 keys exit with a clear error (no XPrv export possible)
- wallet-key-export-test.hs: wrap serializeXPrv args in HashedCredentialsV1
- PersistPrivateKeySpec.hs: bind IO-returning encryptedCreate with >>=
- LayerSpec.hs: remove now-redundant XPrv import (Werror)
- Cardano.Wallet: use encryptedChainCode in decryptV2 to supply the
  missing 32-byte chain code so CC.xprv receives the full 96 bytes

The merged cardano-base PR changed encryptedCreateDirectWithTweak,
encryptedValidatePassphrase, and encryptedKeyMaterial to return IO
instead of pure Either, since they operate on locked memory.

- Bump SRP tag to cb068e0 (merged HEAD of cardano-base master)
- mkV2Credentials: lift to IO, bind encryptedCreateDirectWithTweak
- attachPrivateKeyFromPwd: bind mkV2Credentials with <-
- withRootKey: separate STM read from IO validation so
  encryptedValidatePassphrase can run outside atomically
- migrateV1toV2: bind encryptedCreateDirectWithTweak with >>=
- decryptV2: lift to IO; use mlsbToByteString + mlsbFinalize instead
  of BA.convert (MLockedSizedBytes has no ByteArrayAccess instance)

  The accumulator-based headIsFinalizedWith used txInfoOutputs directly,
  which includes the wallet's change output appended by coverFee. Since
  that output is not in the accumulator, the KZG membership proof failed
  for every real on-chain fanout.

  Restore numberOfFanoutOutputs :: Integer to the Fanout redeemer (as in
  the original hash-based validator) so the validator slices txInfoOutputs
  to only the distributed UTxOs before running the pairing check and value
  conservation. Also update the FanOut.hs head output to carry the UTxO
  value so mustConserveValue passes, and add a unit test that appends a
  trailing wallet change output and asserts the transaction still
  evaluates — this would have caught the regression at the contract test
  level rather than requiring a devnet E2E run.

Signed-off-by: Sasha Bogicevic <[email protected]>

  The new accumulator-based headIsFinalizedWith validator checks that the
  head input value covers all distributed outputs. The test head output was
  missing UTxO.totalValue, causing H4 (HeadValueIsNotPreserved) failures.

Signed-off-by: Sasha Bogicevic <[email protected]>

Signed-off-by: Sasha Bogicevic <[email protected]>

Signed-off-by: Sasha Bogicevic <[email protected]>

  Remove utxoHash/alphaUTxOHash/omegaUTxOHash from ClosedDatum, all
  Close/Contest redeemers, and the snapshot signing tuple. The BLS
  accumulator already commits to the full UTxO set (utxo ∪ alpha ∪
  omega), making the three separate SHA256 hashes redundant.

  Full fanout now verifies outputs via a KZG membership proof (same as
  partial fanout) rather than hash comparison. The Fanout redeemer gains
  proof and crsRef fields; the three output-count fields are dropped.

  Snapshot signing shrinks from a 7-tuple to a 4-tuple
  (headId, version, snapshotNumber, accumulatorHash).

Signed-off-by: Sasha Bogicevic <[email protected]>

Also reduce a diff in the test code

Signed-off-by: Sasha Bogicevic <[email protected]>

Signed-off-by: Sasha Bogicevic <[email protected]>

  fanoutChunkSize and fanoutOutputThreshold are now defined once in
  Test.Hydra.Tx.Fixture and imported wherever needed, removing the
  per-file duplicates.

  prepareTxToPost gains explicit FanoutTx/FinalPartialFanoutTx branches
  (error) so GHC's exhaustiveness checker catches any future unhandled
  constructor, instead of silently falling through a wildcard.

  The two deadline-slot conversions in mkChain.postTx now throw
  FailedToConstructFanoutTx (a PostTxError Tx) instead of userError,
  so the exception is caught by Node.hs's PostTxError handler rather
  than propagating uncaught and crashing the node.

  Replace the hardcoded numToDistribute = totalUTxO - 1 in the partial
  fanout benchmarks with a binary search that finds the largest chunk
  actually fitting within both the tx size and execution budget, mirroring
  findFittingFanoutTx in the real node. The Remaining column in the output
  table now carries meaningful information instead of always showing 1.

  Also add ContestationDeadlineOutsideTimeHorizon to the PostTxError
  oneOf in api.yaml, which was defined in Chain.hs but missing from the
  schema, causing ServerOutputSpec to fail.

Signed-off-by: Sasha Bogicevic <[email protected]>

  computeFinalPartialFanOutCost was capped at 7 (fanoutChunkSize), hiding
  the real tx-size limit for the terminal fanout step. Now it sweeps a wide
  range and searches for the actual maximum, using a single preceding output
  as minimal setup to reach FanoutProgress.

  computePartialFanOutMixedCost also dropped its fanoutChunkSize cap and
  now distributes all-but-one outputs, consistent with the nominal benchmark.
  The stale description mentioning fanoutChunkSize is corrected accordingly.

Signed-off-by: Sasha Bogicevic <[email protected]>

  HeadLogic no longer uses fanoutChunkSize or fanoutOutputThreshold to decide
  which tx type to emit. For a fresh fanout it always emits FanoutTx; for an
  in-progress fanout it always emits FinalPartialFanoutTx. Handlers now owns the
  sizing decision: it tries the preferred tx first, then falls back to
  PartialFanoutTx with a decreasing chunk size (starting at N-1) until one fits
  within the execution budget, using a single shared findFittingFanoutTx helper.

  PartialFanoutTx is removed from PostChainTx — it is now an internal Handlers
  detail, never emitted by HeadLogic. fanoutChunkSize and fanoutOutputThreshold
  are removed from KZGTrustedSetup and all call sites.

Adds a remote-driven QA harness to validate hard-fork (Van Rossem)
readiness end-to-end against mainnet, modelled on cf-token-metadata-registry:

- docker-compose.yml + .env: Postgres + API (built from source), syncing
  mainnet from the first ADA Handle mint (slot 47931310) to tip. DNS-TTL
  relay-wedge mitigation applied via JAVA_TOOL_OPTIONS (the image entrypoint
  is a fixed `java -jar`, so JAVA_OPTS would be ignored).
- .claude/skills/cf-adahandle-resolver-mainnet-integration-test-full-sync:
  SSH-driven skill (preflight -> prepare -> compose up -> tunnels -> monitor
  sync to tip -> parse-error scan -> REST round-trip spot-checks). Pass =
  full sync through Conway with zero genuine block-parse errors + DB-vs-API
  consistency on sampled handles.
- .gitignore: track .claude/, ignore skill runtime config (config.env,
  tunnels.pid) and settings.local.json.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>