eth/channel/test: Replace custom genesis code in SimulatedBackend constructor with core.DeveloperGenesisBlock

[sebastianst]

Our constructor still uses some, now outdated, copied code from core.DeveloperGenesisBlock. We should use this function instead (passing it as faucet address the key that we generate at the top), avoiding code duplication and being automatically up to date with the latest pre-compiles.

I don't know the meaning of parameter period, that's up for research. Looks like it is relevant in our sim backend use case.

Note that the current gas limit seems to be 12.5 Mio (with a jump today to 14 Mio?).

[ggwpez]

The period values seems to set the target block time for the consensus.
Our tests work with all values that i tried, probably since we are not using the PoA consensus algorithm 'clique'.

The gas limit for the genesis is fixed block to 11.5M. The exact value for successive blocks can be calculated with core.CalcGasLimit.

[ggwpez]

As it turned out in hyperledger-labs/go-perun#107, the DeveloperGenesisBlock function is not threadsafe.
It reads+writes a package variable in this line (f), which makes it a race condition since we use it in tests with t.Parallel().

Introducing a package-wide mutex in [backend/eth/channel/test] would be a possibility… that is a bit ugly IMHO.
What do we do now, other suggestions than keeping the hardcoded stuff?
@sebastianst

[sebastianst]

It doesn't write to the package variable but creates a local copy. I'm surprised this causes a race condition.

I agree that a package-wide mutex would be ugly.

[sebastianst]

Ah looks like an unintentional shallow copy 👍🏻

[ggwpez]

Since ethereum/go-ethereum#23068 was merged, this can be properly addressed when the next go-ethereum version is released.
We would still need to wait for miguelmota/go-ethereum-hdwallet#13 and our integration of it, however.