Cardano Updates

The cleanup that follows `cabal v2-build` walks the slice's `$out` and
deletes the lndir-composed dep-slice content — those symlinks are
pure overhead (NAR scan, fixupPhase walk, reference scan all stat
them) and downstream consumers compose dep slices directly from
`nix-support/transitive-deps`.

GHC bakes `<this-slice>/store/.../<dep-uid>/lib` into the slice's own
shared libraries' DT_RUNPATH / LC_RPATH so iserv-dyn / dyld can find
each dep's transitive `.so` at TH-eval time.  Plain "wipe dep dirs"
breaks this: stdenv's `shrink-rpath` (Linux fixupPhase) sees the
entry's dir is empty / missing and drops it, and on Darwin
install_name_tool wasn't touching them either — at runtime dyld
fails with `No such file or directory`.

Before wiping, walk every ELF / Mach-O file in the slice's own
unit dirs and rewrite its rpath: for each entry pointing into
`$out/store/`, follow a symlink under it via `readlink -f` to find
the dep slice's actual `/nix/store/<dep-slice>/store/.../<dep-uid>/lib`
and substitute that.  Linux uses `patchelf --set-rpath`; Darwin uses
`install_name_tool -rpath`.  After the rewrite, `shrink-rpath` sees
populated dirs and keeps the entries.

Confirmed:
  * x86_64-linux native with-packages.run (paired with the awk fix
    in the v2-shell env-file generator)
  * x86_64-linux musl64 js-template-haskell.build — previously
    failed with `Error loading shared library libHSth-abstraction-…
    .so: No such file or directory` because shrink-rpath had
    stripped the dep-slice path
  * x86_64-linux native cabal-sublib-shell.run
  * armv7a-android-prebuilt c-ffi.run (cross)

One bisect-safe implementation slice folds T010 (RED: six failing
parity scenarios in TransactionLedgerSpec), T011 (GREEN:
ScriptWitnesses extension + plumbing in Transaction/Ledger.hs +
default updates at all five call sites), and T012 (regression
re-run of the broader TransactionLedger pattern) into a single
subagent commit. The brief in tasks.md is self-contained — the
subagent loads no process skill.

T000-T002 (bootstrap + spec + plan stops) are already done.
T020-T023 are orchestrator-owned finalisation chores (gate
re-run, PR body refresh, finalization audit, drop gate.sh +
mark ready) and do not require a subagent.

The previous SQL paginated by epoch_stake.id (an auto-increment assigned
at insertion time by db-sync). Two replicas with identical chain data
can have different epoch_stake.id ranges for the same epoch — db-sync
recomputes the snapshot per epoch, and rollback/replay can leave gaps
or reorder insertions. Result: page 1 returned a different subset of
stake_address rows on each backend.

Verified on two backends with bit-identical data (md5 of the full set
matched) but different id ranges (91359..92937 vs 93823..95434 for the
same epoch 121, 1579 rows each).

Fix: order by the slot_no of the `delegation` row that established the
(addr, pool) pairing effective in this epoch. slot_no is pure chain
data, stable across any replica. Tiebreak on addr_id.

Applied to:
- /epochs/:number/stakes              (paged + unpaged)
- /epochs/:number/stakes/:pool_id     (paged + unpaged)

Same response shape, same set of rows — just deterministic ordering.

Plan cost: a correlated subquery on `delegation` fires once per
epoch_stake row, served by `idx_delegation_addr_id`. ~7 ms total on
preview's epoch 121 (1579 rows), unchanged from the previous 6.5 ms.

Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

Signed-off-by: Eric Torreborre <[email protected]>

Cabal pretty-prints long `id:` hashes onto a separate indented line:

  id:
      lens-5.3.6-9653fc7a...

The awk pattern `/^id:[[:space:]]/ { print $2 }` only matched the
boot-db single-line form (`id:    base-4.22.0.0-inplace`), so every
composed-store conf was silently skipped — the `package-id` lines for
lens, transformers, ... never made it into the env file even though
their `.conf`s were in the package-db, so ghc reported them as
hidden:

  >>> Could not load module `Control.Lens'.
  >>>     It is a member of the hidden package `lens-5.3.6'.

Read the next line when `id:` has no value on the same line.

- Introduced `applyDRepCert` function to streamline DRep certificate operations (register, update, retire) in `buildDRepCertTx.ts`.
- Updated `RegisterDRep`, `Retire`, and `UpdateDRep` components to utilize the new `applyDRepCert` function for better code reuse and clarity.
- Added Jest tests for DRep certificate functionality in `drepCert.test.ts`, covering various scenarios including registration, update, and retirement.
- Created utility functions and fixtures for testing in `cborUtils.ts`, `fixtures.ts`, and `mockProvider.ts`.
- Established a GitHub Actions workflow for running unit tests and uploading coverage reports.
- Set coverage thresholds for critical files to ensure code quality.

Signed-off-by: Akhil Repala <[email protected]>

utxoFromMempool initialized with UTxOInBabbageEra, causing all
evaluations to hit the Babbage rejection branch even for Conway
transactions. Changed to UTxOInConwayEra.

Reorder deserialiseCBOR to try Conway and Dijkstra before Babbage.
Conway transactions can successfully deserialize as Babbage (the
Babbage decoder accepts Conway-era CBOR), wrapping them as
GenTxBabbage and rejecting them as "unsupported era". Older eras
kept as fallbacks for specific diagnostics.

Use eraProtVerHigh in transaction deserialization to accept
transactions from any protocol version within an era (e.g.
Conway PV9 through PV11).

Co-Authored-By: Claude <[email protected]>

Prefer oneOf vs anyOf for the interval representations.
Use mutual exclusion (not/required) to make oneOf entries
non-overlapping when both bounds are present.

Co-Authored-By: Claude <[email protected]>

Schema fixes for Dijkstra-era validation:
- Add SubmitTransactionFailure entries: 3169 (PointerAddressInCollateralReturn),
  3170 (SpendingOutputFromSubTransaction)
- Add ScriptPurpose<Guard> and Clause<Guard> to cardano.json
- Fix AccountBalanceInterval: use anyOf instead of oneOf
- Make ProposedProtocolParameters minFeeReferenceScripts sub-fields optional

Enable Dijkstra ApplyTxError generator and Dijkstra GenTx generator.
GenTx uses genDijkstraShelleyTx which strips sub-transactions to
avoid MemoBytes CBOR round-trip failures in mock protocol peers.

Dijkstra block generation remains excluded: randomly generated blocks
fail CBOR round-tripping in the ChainSync mock peer, likely due to
block header/body hash inconsistency. The consensus codec itself
supports Dijkstra's 5-segment block body.

Add test dependencies: cardano-data, cardano-ledger-dijkstra:testlib.
Import Dijkstra Arbitrary instances.

Co-Authored-By: Claude <[email protected]>

- Add GenTxIdDijkstra to genGenTxId
- Add Dijkstra context error to genContextError
- Add Dijkstra-era MissingScript to genScriptFailure
- Add "guard" to script purpose enum in cardano.json
- Remove stale FIXME comment in Generators.hs
- Regenerate EvaluateTransactionResponse test vectors

Co-Authored-By: Claude <[email protected]>

Add schema definitions for new Dijkstra era fields:
- Transaction: requiredExtraGuards, directDeposits,
  accountBalanceIntervals, subTransactions, requiredTopLevelGuards
- ProtocolParameters: maxReferenceScriptsSizePerBlock
- New types: AccountBalanceInterval (with oneOf for bound variants),
  TopLevelGuard, SubTransaction

New JSON field names introduced across all Dijkstra commits:

  Transaction body (TopTx):
  - requiredExtraGuards: script hash guards (new Dijkstra field)
  - directDeposits: account address to coin map
  - accountBalanceIntervals: list of interval constraints per account
  - subTransactions: nested sub-transactions
  - fee: moved to TopTx-only (SubTx has no fee)

  Transaction body (SubTx):
  - requiredTopLevelGuards: guard credentials with optional datums

  AccountBalanceInterval fields:
  - greaterThanOrEqualTo: inclusive lower bound (coin)
  - strictlyLessThan: exclusive upper bound (coin)

  TopLevelGuard fields:
  - credential: guard credential (key hash or script hash)
  - datum: optional inline datum

  ProtocolParameters:
  - maxReferenceScriptsSizePerBlock: new Dijkstra param

  Script purpose enum:
  - "guard": new Dijkstra script purpose

  Error codes:
  - 3169: PointerAddressInCollateralReturn
  - 3170: SpendingOutputFromSubTransaction

Co-Authored-By: Claude <[email protected]>

Co-Authored-By: Claude <[email protected]>

Both full PParams and PParamsUpdate nest range and multiplier inside
minFeeReferenceScripts to preserve Conway's API shape. For full
PParams, all three sub-fields are always present. For updates, only
the sub-fields being updated are included, and the key is omitted
entirely when none of the three are updated.

Co-Authored-By: Claude <[email protected]>

Gate Dijkstra behind nodeToClientV > V_23 since V_23 is the latest
NodeToClientVersion in ouroboros-consensus 3.0.1.0 and is negotiated
by all networks with node 11.0.1.

On Dijkstra networks (future V_24+), Dijkstra is tried first. On
Conway networks, Conway is tried first. Each list is ordered from
most probable era to least probable.

Verified on preview and preprod that GenTxIdDijkstra does not crash
the node (returns false, loop continues), so this is a performance
optimization to avoid unnecessary round-trips on non-Dijkstra
networks.

Co-Authored-By: Claude <[email protected]>

Signed-off-by: Akhil Repala <[email protected]>

Signed-off-by: dependabot[bot] <[email protected]>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

`haskell-nix-cabal-project-local-sync` was previously included in
the shell's `nativeBuildInputs` and shellHook only when the
project's effective `cabalProjectLocal` had non-empty content.
For targets where none of the platform mkIfs fire and the user
hasn't set anything (e.g. ghcjs), the script was missing from
PATH — so any test that called it directly errored with
`command not found`, and the test's `import:
cabal.project.<targetPrefix>local` then failed because the file
didn't exist either.

Always include the script and always run the shellHook.  When
the project's `cabalProjectLocal` is empty, the script writes an
empty `cabal.project.<targetPrefix>local`, which cabal imports
harmlessly (no-op).  Removes the empty-content gate, so the
shape is the same across all targets.

Verified the ghcjs cabal-sublib-shell test now passes
(`aarch64-darwin.unstable.ghc9141.ghcjs.tests.cabal-sublib-shell.run`)
and the existing native/musl64/wasi32 paths are unchanged.