feat: added special address with block number and block hash

[Mohiiit]

Small fix for SNOS to run, it requires special address in the state diff

Pull Request type

Please add the labels corresponding to the type of changes your PR introduces:

• Feature

What is the current behavior?

Currently we were not adding the block hash and block number of previous block with special address which was breaking the SNOS

Resolves: #NA

What is the new behavior?

• The above mention issue is resolved

Does this introduce a breaking change?

No

[apoorvsadana]

let's panic ig?

[Mohiiit]

resolved

[apoorvsadana]

let's add a comment

[Mohiiit]

added

[apoorvsadana]

Suggested change

	block_time: Duration::from_secs(4),
	pending_block_update_time: Duration::from_secs(1),

[Mohiiit]

resolved

[jbcaron]

This logic for block hashes is already implemented for blockifier with the StateAdapter, it should be changed accordingly while keeping a compatibility of re-execution of the mainnet.

madara/crates/client/exec/src/blockifier_state_adapter.rs

Lines 33 to 55 in fe09932

	if *contract_address.key() == Felt::ONE {
	let requested_block_number = felt_to_u64(key.0.key()).map_err(|_| StateError::OldBlockHashNotProvided)?;
	// Not found if in the last 10 blocks.
	if !block_hash_storage_check_range(
	&self.backend.chain_config().chain_id,
	self.block_number,
	requested_block_number,
	) {
	return Ok(Felt::ZERO);
	}
	return self
	.backend
	.get_block_hash(&BlockId::Number(requested_block_number))
	.map_err(|err| {
	log::warn!("Failed to retrieve block hash for block number {requested_block_number}: {err:#}");
	StateError::StateReadError(format!(
	"Failed to retrieve block hash for block number {requested_block_number}",
	))
	})?
	.ok_or(StateError::OldBlockHashNotProvided);
	}

[apoorvsadana]

@jbcaron, I am mostly missing something but do we need this custom logic in the reader? The feeder gateway does return state diffs where it says 0x1 has the following storage keys and values. In full node mode, if we are applying these state diffs in the DB, won't normal fetching from DB work same as this custom logic we're using here?

[jbcaron]

we should just delete this code in the same PR but be sure that we can re-execute the mainnet blocks (I don't know from when this feature was introduced)
I checked on sepolia, there is no issue

[apoorvsadana]

we should just delete this code in the same PR but be sure that we can re-execute the mainnet blocks

makes sense, we can delete and see if the test cases pass + that we can resync a few hundred blocks on mainnet + get_storage_at on 0x1 address works as expected

EDIT: @Mohiiit has also already informed the visoft team for e2e cases for this specific case

Cannot merge now

[cchudant]

this match can be simplified to

let prev_block_hash = self.backend.get_block_hash(&BlockId::Number(prev_block_number))
  .map_err(|err| Error::Unexpected(
      format!("Error fetching block hash for block {prev_block_number}: {err:#}").into()
  ))?
  .ok_or_else(|| Error::Unexpected(
      format!("No block hash found for block number {prev_block_number}").into()
  ))?;
let address = Felt::ONE;
new_state_diff.storage_diffs.push(ContractStorageDiffItem {
    address,
    storage_entries: vec![StorageEntry { key: Felt::from(block_n), value: prev_block_hash }],
});

[Mohiiit]

resolved

[jbcaron]

I think we should only return 0x0 in this case
we can therefore remove this check
this is already the implementation in the state adapter but maybe this logic should be in the db
@cchudant what do you think?

[Mohiiit]

I checked with the juno rpc call and it has this behaviour hence implemented this

[cchudant]

apparently you have to return ContractNotFound
https://github.com/starkware-libs/starknet-specs/blob/master/api/starknet_api_openrpc.json#L216

but we need to bring this up into full node collab channel - getNonce also returns ContractNotFound but I think it's possible to increment the nonce of an undeclared contract