cardano-testnet: separate in types the options encoded in genesis files, from other options

[smelc]

Description

This PR implements a decision made on Slack: we wanted to change that parameters of cardanoTestnet are containing some values both in the CardanoTestnetOptions type and in the genesis files, causing confusion.

So this PR changes cardanoTestnet input types so that this duplication is removed: CardanoTestnetOptions now has less fields.

Note that Haskell callers to cardanoTestnet should now obtain the default value for CardanoTestnetDefault by calling def (from the Default class), instead of using cardanoDefaultTestnetOptions (which was removed).

How to review this PR

The commits are iterative, because I was testing out the design while writing this PR, and so every commit compiles and makes sense on its own. But it's nevertheless better to just look at the global diff in one shot.

Checklist

• Commit sequence broadly makes sense and commits have useful messages
• Self-reviewed the diff

[smelc]

I don't know sure what's causing these changes. What I did in this PR should have no runtime effect, so I'm wondering.

@carbolymer> would you have a clue?

[carbolymer]

I haven't read your code thoroughly but you're using data-default-class, maybe it pulled a sane plutus v2 cost model default from ledger instead of 2^63-1?

[smelc]

OK found from where the data comes. It's those costs that are showing up now: https://github.com/IntersectMBO/cardano-api/blob/4b3e3305e8bcb5b7eef33cdd1baf1b93d330a7b2/cardano-api/internal/Cardano/Api/Genesis.hs#L726

Still don't know what causes this though.

[smelc]

OK found it: https://github.com/IntersectMBO/cardano-node/pull/5976/files#r1749955020

The change is correct: before, we were not honoring the user-supplied era in cardanoTestnetDefault.

[smelc]

This explains the change in the protocol parameters golden file: before we were not honoring the era passed within the CardanoTestnetOptions parameters. We were wrongly using cardanoDefaultTestnetOptions.

Now we take the era from the input parameter.

[erikd]

This needs a rebase but otherwise LGTM.

[Jimbo4350]

We now have two testnet options types here. This is not ideal.

[smelc]

Maybe the naming is bad. Would ShelleyTestnetOptions better be named GenesisOptions maybe?

[Jimbo4350]

GenesisOptions is better.

[smelc]

PR doing the renaming: #5987

[Jimbo4350]

What is the difference between these two types? We should not have two types that describe the testnet options.

[smelc]

@Jimbo4350> the difference is explained in the haddocks of the definitions of those types:

-- | Options which, contrary to 'ShelleyTestnetOptions' are not implemented
-- by tuning the genesis files.
data CardanoTestnetOptions = CardanoTestnetOptions

and

-- | Options that are implemented by writing fields in the Shelley genesis file.
data ShelleyTestnetOptions = ShelleyTestnetOptions

[Jimbo4350]

This split seems arbitrary. Why are we introducing another configuration type?

[smelc]

I think the split is clear: ShelleyTestnetOptions contains values that are encoded in genesis files whereas CardanoTestnetOptions contains the rest of the values.

This is visible e.g. in the new prototype of cardanoTestnet that contains CardanoTestnetOptions and the genesis files, as opposed to cardanoTestnetDefault. cardanoTestnetDefault has both ShelleyTestnetOptions and CardanoTestnetOptions, because the genesis files are not yet around at this stage.

[Jimbo4350]

I think the source of confusion for me came from cardanoMaxSupply existing in both types. The naming is also confusing. Cardano vs Shelley Tesnetoptions. The difference should be immediately obvious in the names (you suggested GenesisOptions which is better).

We have:

data CardanoTestnetOptions = CardanoTestnetOptions
  { -- | List of node options. Each option will result in a single node being
    -- created.
    cardanoNodes :: [TestnetNodeOptions]
  , cardanoNodeEra :: AnyShelleyBasedEra -- ^ The era to start at
  , cardanoMaxSupply :: Word64 -- ^ The amount of Lovelace you are starting your testnet with (forwarded to shelley genesis)
                               -- TODO move me to ShelleyTestnetOptions when https://github.com/IntersectMBO/cardano-cli/pull/874 makes it to cardano-node
  , cardanoEnableP2P :: Bool
  , cardanoNodeLoggingFormat :: NodeLoggingFormat
  , cardanoNumDReps :: Int -- ^ The number of DReps to generate at creation
  , cardanoEnableNewEpochStateLogging :: Bool -- ^ if epoch state logging is enabled
  } deriving (Eq, Show)

I think we should aim for:

data CardanoTestnetOptions = CardanoTestnetOptions
  { cardanoNodes :: [TestnetNodeOptions]
  }

There are other random configuration values here e.g cardanoEnableP2P and cardanoNumDReps which do not belong here and then we have to reconcile configuration overrides which just leads to more confusion. We are trying to do too much wrt the configuration and I think leaning more on cardano-cli to generate the configuration is the right direction. We will eliminate a lot of potential bug sources doing this.

[Jimbo4350]

These overrides can live in create-testnet-data but we also have the issue of relisting all of the potential configuration values in the create-testnet-data which obviously isn't sensible. Maybe we can move the existing overrides to create-testnet-data.