Plutus predicate failure to use Plutus error instead of string

[JaredCorduan]

It would be helpful to downstream libraries if the plutus script predicate returned the actual plutus error instead of a string:

https://github.com/input-output-hk/cardano-ledger-specs/blob/8ef01f28e31789503879f1391b230aa1e50c6ab2/eras/alonzo/impl/src/Cardano/Ledger/Alonzo/TxInfo.hs#L632

[MaximilianAlgehed]

I agree that this is a problem. We've bumped into issues / warts / frustrations related to this when implementing testing tools for Plutus contracts. For example over at IntersectMBO/plutus#4023.

As for solutions I don't know what's best but the information that's needed by downstream code is at least:

1. What term caused the error (in the case of e.g. builtins failing)
2. What error message did the contract developer intend to throw. Right now this information is propagated via the log but in a perfect world the interpreter would put this error in the datastructure it returns when a computation fails.

[JaredCorduan]

We unfortunately cannot run Plutus scripts in verbose mode inside the node, but there will soon be two ways to see the verbose logs.

• The CLI will soon be using the newer version of evaluateTransactionExecutionUnits when building transactions, so that when a Plutus scripts fails it can return the logs.
• When the node returns the predicate failure for a failed script, in addition to the show e that is inside the Text, there is debug information in the ByteString. I believe our plan is to make using the debug information to retrieve the verbose logs available from the CLI. This is just a matter of running debugPutus with the information from the predicate failure.