cardano-testnet: Dijkstra hard fork and nested transaction swap tests

[Jimbo4350]

Description

Two cardano-testnet integration tests for Dijkstra, plus the config change they need.

Hard fork to Dijkstra starts the default three-node cluster in Conway at protocol version 10 and hard forks it through governance, the way mainnet will: a HardForkInitiation action to PV 11.0 (voted for by the three default DReps and the default SPO, ratified, enacted), then a chained HardForkInitiation action to PV 12.0. It then asserts that the node reports the Dijkstra era and keeps producing blocks.

Nested transaction swap additionally mints 100 TokenA to Alice in Conway and funds Bob, then, once at PV12, settles a multi-asset swap with a Dijkstra nested transaction built with cardano-cli dijkstra transaction sub-transaction (IntersectMBO/cardano-cli#1453, via CARDANO_CLI): Alice's sub-transaction gives 100 TokenA for 53 ADA, Bob's gives 60 ADA for the 100 TokenA plus 7 ADA change, the batcher's top-level transaction embeds both and pays the fee. Each party signs only its own sub-transaction. Proof is read back from query utxo: Bob holds the tokens, Alice gained exactly 50 ADA, Bob lost exactly 50 ADA, the batcher paid the fee.

Testnet/Defaults.hs now sets ExperimentalHardForksEnabled: true for every era. Without it the node declares PV11 as its maximum and, once PV12 is enacted, rejects its own blocks with ObsoleteNode and the chain stalls. Golden config updated. Testnet.Components.Query.stopEpochStateView lets a test stop the client-side ledger fold before the chain enters an era that fold cannot process.

Status

cabal.project temporarily points ouroboros-consensus at IntersectMBO/ouroboros-consensus@0ebed78d (branch jordan/4.2.1.0-allow-hardfork-into-dijkstra): the 4.2.1.0 release plus the cherry-picked commit from IntersectMBO/ouroboros-consensus#2167 ("Allow hardfork into Dijkstra"). No 4.x release contains it; without it the Conway ledger config is built with TriggerHardForkNotDuringThisExecution and the node never enters Dijkstra. To be replaced by a released version.

• Hard fork to Dijkstra passes with the SRP: PV 10 -> 11 -> 12 is enacted through governance, the node enters Dijkstra, query tip reports Dijkstra, blocks keep coming. On plain 4.2.1.0 it fails at the era assertion with The node is not in the Dijkstra era after protocol version 12 was enacted: Just (String "Conway").
• Nested transaction swap reaches the Dijkstra ledger and is rejected at submission with SubBadInputsUTxO for every sub-transaction input, plus ValueNotConservedUTxO at the top level. Cause: UTxO-HD key fetching (getTransactionKeySets / getBlockKeySets via allInputsTxBodyF) does not load sub-transaction inputs, so the ledger sees them as missing. Present in consensus 4.2.1.0 and main, and in ledger's Dijkstra allInputsTxBodyF (0.3.0.0 and master). With a consensus-side fix that includes sub-transaction inputs, the swap passes end to end (verified locally, not part of this PR).

Also found: cardano-api's ConvertLedgerEvent instance for the hard fork block has no DijkstraLedgerEvent case (11.7.0.0 and master), so foldBlocks / foldEpochState clients crash on the first Dijkstra block. The tests therefore stop the epoch state view before PV12 and keep the epoch state logger off.

Run with:

DISABLE_RETRIES=1 cabal test cardano-testnet-test --test-options '-p "/Hard fork to Dijkstra/"'
CARDANO_CLI=/path/to/pr-1453/cardano-cli DISABLE_RETRIES=1 cabal test cardano-testnet-test --test-options '-p "/Nested transaction swap/"'

Checklist

• Commit sequence broadly makes sense and commits have useful messages
• New tests are added if needed and existing tests are updated.
• Any changes are noted in the CHANGELOG.md for affected package (changelog fragment in cardano-testnet/.changes/)
• The version bounds in .cabal files are updated
• CI passes.
• Self-reviewed the diff

🤖 Generated with Claude Code