SCP-3305 SCP-3263 fixed Ledger.Constraints.Offchain.updateUtxoIndex

[bwbush]

Prior to this commit, Ledger.Constraints.Offchain.updateUtxoIndex would discard lookups for inputs that aren't script outputs. However, sometimes it is necessary to include particular public-key TxIns in a transaction, via Plutus.Contract.Wallet.ExportTx.lookups: a good example of this is when a Plutus.Contract.Currency.OneShotCurrency needs to consume a specified UTxO in its minting policy. Even though Plutus.Constraints.Tx.mustSpendPubKeyOutput adds a public-key input to the lookups, updateUtxoIndex discards that, with the result that it is only by chance that cardano-wallet would select that necessary input for the balanced transaction and provide it to Plutus validators. (Note that cardano-wallet does not automatically provide the TxIn in the partially constructed transaction to validators: it only provides Cardano.Wallet.PartialTx.inputs and the inputs it has semi-randomly chosen.) The consequence of all of this is that scripts that require a particular input will randomly fail during balancing in wallets with more than one UTxO.

This fix simply adds public-key inputs to the type Ledger.Constraints.OffChain.ScriptOutput so that it can also hold public-key inputs. (There might be a more descriptive name than ScriptOutput now.) It makes minor adjustments to the chain of functions that pass this into an unbalanced transaction before it is sent to cardano-wallet.

Pre-submit checklist:

• Branch
  • Tests are provided (if possible)
  • Commit sequence broadly makes sense
  • Key commits have useful messages
  • Relevant tickets are mentioned in commit messages
  • Formatting, materialized Nix files, PNG optimization, etc. are updated
• PR
  • Self-reviewed the diff
  • Useful pull request description
  • Reviewer requested

[raduom]

This looks good to me, but I have not worked in this area recently, so I asked @sjoerdvisscher for a review also.

[sjoerdvisscher]

The reason we removed public-key inputs is that cardano-wallet doesn't support it. See #57.
So it might be that we were wrongly informed, but did you check that this works with cardano-wallet?

[sjoerdvisscher]

If we do this, the way to go would be to revert #57.

[bwbush]

The reason we removed public-key inputs is that cardano-wallet doesn't support it. See #57. So it might be that we were wrongly informed, but did you check that this works with cardano-wallet?

Yes, I've tested this will wallet and actual Marlowe contract endpoints. The MustSpendPubKeyOutput could have been named MustSpendWalletOutput because it just provides a UTxO that could reside at any of the wallet's addresses.

I stepped through each phase of wallet's processing of the request from the PAB to see how it balances the transaction: the wallet makes provision for spending particular UTxOs in the PartialTx.inputs, which is poplulated from ExportTx. The wallet only provides UTxOs to scripts if they are specified in PartialTx.inputs or if the wallet happens to select the UTxO when it balances; unfortunately, the wallet does not pass the TxIns contained in PartialTx.partialTx :: Tx AlonzoEra to scripts involved in the transaction.

[bwbush]

If we do this, the way to go would be to revert #57.

It looks like reverting #57 would work, but I'd like to test that.

[bwbush]

@sjoerdvisscher, I reverted #57 in d08145b. The Marlowe PAB integration test passes (running the full Marlowe contract on our private testnet) using this revision. Would you please re-review? Thanks.

[sjoerdvisscher]

I like all this red :)