feat: access list support for signature calculation

[mmyslariane]

Description:
When sending transaction with access list the public key and the address of the transaction sender is not properly calculated and it results in error for example of INVALID_ACCOUNT_ID. We added changes in EthTxData.java and EthTxSigs.java to fix this behaviour by firstly read accessList as object of RLPList instance and put it different variable accessListAsRLP. If this variable is present we transform it to array of Object containing byte array and then we add this to RLPEncoder.
We also put unit tests to check if now transactions on which we found this error are now calculating proper signature.

Related issue(s):
#17040

Notes for reviewer:

Checklist

• Documented (Code comments, README, etc.)
• Tested (unit, integration, etc.)

[david-bakin-sl]

This is awkward. This is a fairly widely used class from a module that is intended to hide certain implementation details. One of those details would be the use of the headlong library - from which the RLPList type comes. And sure enough, the headlong library isn't exported from the module's module-info.java.

Problem is - what to replace it with?

No good suggestions, but what I have is: Since the only other use is in EthTxSigs in this very same module (and I wouldn't anticipate this field being used anywhere else in the near future): You could hold it in the record as an Object - and then EthTxSigs would have to know the intimate details of EthTxData - but it is in the same package ... (And also being an Object would signal to users outside the package that they should stay away from it.)

[mmyslariane]

Ok, I have a question, do you wish to get rid off using of headlong library in the EthTxData class or do not use RLPList as a field in this class? Because if you only wish to not use RLPList as an record than maybe the best solution would be holding this record field accessListAsRLP as an array of Object? Instead of using the new method in EthTxSigs

static Object[] encodeRLPList(RLPList rlpList) {
        return rlpList.elements().stream().map(rlpItem -> {
            if (rlpItem.isList()) {
                return encodeRLPList(rlpItem.asRLPList());
            }
            else {
                return rlpItem.data();
            }
        }).toArray();
    }

I could always move it to EthTxData and do the transformation here and save the result as the new record as an array of Object and then only use the result in calculating the signature. Would that be ok?

[david-bakin-sl]

RIght, not get rid of the use of headlong, just don't expose any headlong type (RLPList in this case) as a record field which means other modules will see it.

Holding it as an Object[] is just fine - just do whatever you need to do in EthTxSigs. Sound OK?

[mmyslariane]

Sounds great! Ok, so I will make this change and push it soon