Chain state in head state not updated on replayed observation

[ch1bo]

Context & versions

Any current version, master

Steps to reproduce

1. Create and open a Head at block x+1, check the persisted state, it should have an Open chain state
2. Configure hydra-node to --start-chain-from block x
3. Restart the hydra-node and check state again as the node resynchronizes with the L1

Actual behavior

The chainState is reset to Idle and weirdly has previous chain states attached.

Expected behavior

The chainState is Idle at first with no previous, and update accordingly as the chain component synchronizes with the L1. Expect to see an Open chain state next to an Open head state in the end.

Solution idea

In the HeadLogic.update function, we could be writing the chainState also on any observation (if not matched and resulting in a headState transition already)

  (currentState, OnChainEvent Observation{newChainState}) ->
      NewState (setChainState newChainState currentState) []

[ffakenz]

From my pov, it is not clear what is needed for this story.

[ffakenz]

We started a draft branch (fix-start-chain-from) on Monday to explore this, but we could not get the problem.

[ffakenz]

We tried to reproduced the problem using this scenario but the persisted state seems to be fine.

startChainFromReplayed :: FilePath -> Tracer IO EndToEndLog -> RunningNode -> TxId -> IO ()
startChainFromReplayed tmpDir tracer node@RunningNode{networkId, nodeSocket} hydraScriptsTxId = do
  (aliceCardanoVk, _) <- keysFor Alice
  let aliceSk = generateSigningKey "alice-timed"
  let alice = deriveParty aliceSk
  let contestationPeriod = UnsafeContestationPeriod 2
  aliceChainConfig <- chainConfigFor Alice tmpDir nodeSocket [] contestationPeriod
  (chainPointBeforeComit, committedUTxOByAlice, headId) <- withHydraNode tracer aliceChainConfig tmpDir 1 aliceSk [] [1] hydraScriptsTxId $ \n1 -> do
    waitForNodesConnected tracer [n1]
    let lovelaceBalanceValue = 100_000_000

    -- Funds to be used as fuel by Hydra protocol transactions
    seedFromFaucet_ node aliceCardanoVk lovelaceBalanceValue (contramap FromFaucet tracer)
    send n1 $ input "Init" []
    headId <- waitForAllMatch 10 [n1] $ headIsInitializingWith (Set.fromList [alice])

    -- Get some UTXOs to commit to a head
    (aliceExternalVk, aliceExternalSk) <- generate genKeyPair
    committedUTxOByAlice <- seedFromFaucet node aliceExternalVk aliceCommittedToHead (contramap FromFaucet tracer)
    chainPointBeforeComit <- queryTip networkId nodeSocket
    requestCommitTx n1 committedUTxOByAlice <&> signTx aliceExternalSk >>= submitTx node

    waitFor tracer 3 [n1] $ output "HeadIsOpen" ["utxo" .= committedUTxOByAlice, "headId" .= headId]
    pure (chainPointBeforeComit, committedUTxOByAlice, headId)

  let aliceChainConfigBeforeOpen = aliceChainConfig{startChainFrom = Just chainPointBeforeComit}
  withHydraNode tracer aliceChainConfigBeforeOpen tmpDir 1 aliceSk [] [1] hydraScriptsTxId $ \n1 -> do
    threadDelay 5
    send n1 $ input "Close" []
    waitMatch 3 n1 $ \v -> do
      guard $ v ^? key "tag" == Just "HeadIsClosed"
    stateContents <- readFileLBS $ tmpDir </> "state-1" </> "state"
    putLBSLn stateContents

This likely got fixed when implementing #913