nimbus-eth2/tests/testblockutil.nim

560 lines
20 KiB
Nim
Raw Normal View History

# beacon_chain
# Copyright (c) 2018-2024 Status Research & Development GmbH
# Licensed and distributed under either of
# * MIT license (license terms in the root directory or at https://opensource.org/licenses/MIT).
# * Apache v2 license (license terms in the root directory or at https://www.apache.org/licenses/LICENSE-2.0).
# at your option. This file may not be copied, modified, or distributed except according to those terms.
{.push raises: [].}
import
chronicles,
stew/endians2,
../beacon_chain/consensus_object_pools/sync_committee_msg_pool,
../beacon_chain/spec/datatypes/bellatrix,
../beacon_chain/spec/[
beaconstate, helpers, keystore, signatures, state_transition, validator]
# TODO remove this dependency
from std/random import rand
from eth/common/eth_types_rlp import rlpHash
type
MockPrivKeysT = object
MockPubKeysT = object
const
MockPrivKeys* = MockPrivKeysT()
MockPubKeys* = MockPubKeysT()
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.5/tests/core/pyspec/eth2spec/test/helpers/keys.py
func `[]`*(sk: MockPrivKeysT, index: ValidatorIndex|uint64): ValidatorPrivKey =
var bytes = (index.uint64 + 1'u64).toBytesLE() # Consistent with EF tests
static: doAssert sizeof(bytes) <= sizeof(result)
copyMem(addr result, addr bytes, sizeof(bytes))
proc `[]`*(pk: MockPubKeysT, index: uint64): ValidatorPubKey =
var cache {.threadvar.}: Table[uint64, ValidatorPubKey]
cache.withValue(index, key) do:
return key[]
do:
let key = MockPrivKeys[index].toPubKey().toPubKey()
cache[index] = key
return key
proc `[]`*(pk: MockPubKeysT, index: ValidatorIndex): ValidatorPubKey =
pk[index.uint64]
func makeFakeHash*(i: int): Eth2Digest =
var bytes = uint64(i).toBytesLE()
static: doAssert sizeof(bytes) <= sizeof(result.data)
copyMem(addr result.data[0], addr bytes[0], sizeof(bytes))
proc makeDeposit*(
i: int,
flags: UpdateFlags = {},
cfg = defaultRuntimeConfig): DepositData =
let
privkey = MockPrivKeys[i.ValidatorIndex]
pubkey = MockPubKeys[i.ValidatorIndex]
withdrawal_credentials = makeWithdrawalCredentials(pubkey)
result = DepositData(
pubkey: pubkey,
withdrawal_credentials: withdrawal_credentials,
amount: MAX_EFFECTIVE_BALANCE)
2022-04-14 15:39:37 +00:00
if skipBlsValidation notin flags:
result.signature = get_deposit_signature(cfg, result, privkey).toValidatorSig()
proc makeInitialDeposits*(
n = SLOTS_PER_EPOCH, flags: UpdateFlags = {}, cfg = defaultRuntimeConfig): seq[DepositData] =
for i in 0..<n.int:
result.add makeDeposit(i, flags, cfg = cfg)
func signBlock(
fork: Fork, genesis_validators_root: Eth2Digest, forked: ForkedBeaconBlock,
privKey: ValidatorPrivKey, flags: UpdateFlags = {}): ForkedSignedBeaconBlock =
let
slot = withBlck(forked): forkyBlck.slot
root = hash_tree_root(forked)
signature =
if skipBlsValidation notin flags:
get_block_signature(
fork, genesis_validators_root, slot, root, privKey).toValidatorSig()
else:
ValidatorSig()
ForkedSignedBeaconBlock.init(forked, root, signature)
from eth/eip1559 import EIP1559_INITIAL_BASE_FEE, calcEip1599BaseFee
from eth/common/eth_types import EMPTY_ROOT_HASH, GasInt
proc build_empty_merge_execution_payload(state: bellatrix.BeaconState):
Support for driving multiple EL nodes from a single Nimbus BN (#4465) * Support for driving multiple EL nodes from a single Nimbus BN Full list of changes: * Eth1Monitor has been renamed to ELManager to match its current responsibilities better. * The ELManager is no longer optional in the code (it won't have a nil value under any circumstances). * The support for subscribing for headers was removed as it only worked with WebSockets and contributed significant complexity while bringing only a very minor advantage. * The `--web3-url` parameter has been deprecated in favor of a new `--el` parameter. The new parameter has a reasonable default value and supports specifying a different JWT for each connection. Each connection can also be configured with a different set of responsibilities (e.g. download deposits, validate blocks and/or produce blocks). On the command-line, these properties can be configured through URL properties stored in the #anchor part of the URL. In TOML files, they come with a very natural syntax (althrough the URL scheme is also supported). * The previously scattered EL-related state and logic is now moved to `eth1_monitor.nim` (this module will be renamed to `el_manager.nim` in a follow-up commit). State is assigned properly either to the `ELManager` or the to individual `ELConnection` objects where appropriate. The ELManager executes all Engine API requests against all attached EL nodes, in parallel. It compares their results and if there is a disagreement regarding the validity of a certain payload, this is detected and the beacon node is protected from publishing a block with a potential execution layer consensus bug in it. The BN provides metrics per EL node for the number of successful or failed requests for each type Engine API requests. If an EL node goes offline and connectivity is resoted later, we report the problem and the remedy in edge-triggered fashion. * More progress towards implementing Deneb block production in the VC and comparing the value of blocks produced by the EL and the builder API. * Adds a Makefile target for the zhejiang testnet
2023-03-05 01:40:21 +00:00
bellatrix.ExecutionPayloadForSigning =
## Assuming a pre-state of the same slot, build a valid ExecutionPayload
## without any transactions from a non-merged block.
doAssert not is_merge_transition_complete(state)
let
latest = state.latest_execution_payload_header
timestamp = compute_timestamp_at_slot(state, state.slot)
randao_mix = get_randao_mix(state, get_current_epoch(state))
var payload = bellatrix.ExecutionPayload(
parent_hash: latest.block_hash,
state_root: latest.state_root, # no changes to the state
receipts_root: EMPTY_ROOT_HASH,
block_number: latest.block_number + 1,
prev_randao: randao_mix,
gas_limit: 30000000, # retain same limit
gas_used: 0, # empty block, 0 gas
timestamp: timestamp,
base_fee_per_gas: EIP1559_INITIAL_BASE_FEE)
payload.block_hash = rlpHash blockToBlockHeader(bellatrix.BeaconBlock(body:
bellatrix.BeaconBlockBody(execution_payload: payload)))
Support for driving multiple EL nodes from a single Nimbus BN (#4465) * Support for driving multiple EL nodes from a single Nimbus BN Full list of changes: * Eth1Monitor has been renamed to ELManager to match its current responsibilities better. * The ELManager is no longer optional in the code (it won't have a nil value under any circumstances). * The support for subscribing for headers was removed as it only worked with WebSockets and contributed significant complexity while bringing only a very minor advantage. * The `--web3-url` parameter has been deprecated in favor of a new `--el` parameter. The new parameter has a reasonable default value and supports specifying a different JWT for each connection. Each connection can also be configured with a different set of responsibilities (e.g. download deposits, validate blocks and/or produce blocks). On the command-line, these properties can be configured through URL properties stored in the #anchor part of the URL. In TOML files, they come with a very natural syntax (althrough the URL scheme is also supported). * The previously scattered EL-related state and logic is now moved to `eth1_monitor.nim` (this module will be renamed to `el_manager.nim` in a follow-up commit). State is assigned properly either to the `ELManager` or the to individual `ELConnection` objects where appropriate. The ELManager executes all Engine API requests against all attached EL nodes, in parallel. It compares their results and if there is a disagreement regarding the validity of a certain payload, this is detected and the beacon node is protected from publishing a block with a potential execution layer consensus bug in it. The BN provides metrics per EL node for the number of successful or failed requests for each type Engine API requests. If an EL node goes offline and connectivity is resoted later, we report the problem and the remedy in edge-triggered fashion. * More progress towards implementing Deneb block production in the VC and comparing the value of blocks produced by the EL and the builder API. * Adds a Makefile target for the zhejiang testnet
2023-03-05 01:40:21 +00:00
bellatrix.ExecutionPayloadForSigning(executionPayload: payload,
blockValue: Wei.zero)
from stew/saturating_arith import saturate
proc build_empty_execution_payload(
state: bellatrix.BeaconState,
feeRecipient: Eth1Address): bellatrix.ExecutionPayloadForSigning =
## Assuming a pre-state of the same slot, build a valid ExecutionPayload
## without any transactions.
let
latest = state.latest_execution_payload_header
timestamp = compute_timestamp_at_slot(state, state.slot)
randao_mix = get_randao_mix(state, get_current_epoch(state))
base_fee = calcEip1599BaseFee(GasInt.saturate latest.gas_limit,
GasInt.saturate latest.gas_used,
latest.base_fee_per_gas)
var payload = bellatrix.ExecutionPayloadForSigning(
executionPayload: bellatrix.ExecutionPayload(
parent_hash: latest.block_hash,
fee_recipient: bellatrix.ExecutionAddress(data: distinctBase(feeRecipient)),
state_root: latest.state_root, # no changes to the state
receipts_root: EMPTY_ROOT_HASH,
block_number: latest.block_number + 1,
prev_randao: randao_mix,
gas_limit: latest.gas_limit, # retain same limit
gas_used: 0, # empty block, 0 gas
timestamp: timestamp,
base_fee_per_gas: base_fee),
blockValue: Wei.zero)
payload.executionPayload.block_hash =
bellatrix.BeaconBlock(body: bellatrix.BeaconBlockBody(execution_payload:
payload.executionPayload)).compute_execution_block_hash()
payload
proc addTestBlock*(
state: var ForkedHashedBeaconState,
cache: var StateCache,
eth1_data: Eth1Data = Eth1Data(),
attestations: seq[Attestation] = newSeq[Attestation](),
deposits: seq[Deposit] = newSeq[Deposit](),
sync_aggregate: SyncAggregate = SyncAggregate.init(),
graffiti: GraffitiBytes = default(GraffitiBytes),
flags: set[UpdateFlag] = {},
nextSlot: bool = true,
cfg: RuntimeConfig = defaultRuntimeConfig): ForkedSignedBeaconBlock =
# Create and add a block to state - state will advance by one slot!
Revamp attestation pool This is a revamp of the attestation pool that cleans up several aspects of attestation processing as the network grows larger and block space becomes more precious. The aim is to better exploit the divide between attestation subnets and aggregations by keeping the two kinds separate until it's time to either produce a block or aggregate. This means we're no longer eagerly combining single-vote attestations, but rather wait until the last moment, and then try to add singles to all aggregates, including those coming from the network. Importantly, the branch improves on poor aggregate quality and poor attestation packing in cases where block space is running out. A basic greed scoring mechanism is used to select attestations for blocks - attestations are added based on how much many new votes they bring to the table. * Collect single-vote attestations separately and store these until it's time to make aggregates * Create aggregates based on single-vote attestations * Select _best_ aggregate rather than _first_ aggregate when on aggregation duty * Top up all aggregates with singles when it's time make the attestation cut, thus improving the chances of grabbing the best aggregates out there * Improve aggregation test coverage * Improve bitseq operations * Simplify aggregate signature creation * Make attestation cache temporary instead of storing it in attestation pool - most of the time, blocks are not being produced, no need to keep the data around * Remove redundant aggregate storage that was used only for RPC * Use tables to avoid some linear seeks when looking up attestation data * Fix long cleanup on large slot jumps * Avoid some pointers * Speed up iterating all attestations for a slot (fixes #2490)
2021-04-12 20:25:09 +00:00
if nextSlot:
var info = ForkedEpochInfo()
process_slots(
cfg, state, getStateField(state, slot) + 1, cache, info, flags).expect(
"can advance 1")
let
proposer_index = get_beacon_proposer_index(
2023-01-11 12:29:21 +00:00
state, cache, getStateField(state, slot)).expect("valid proposer index")
privKey = MockPrivKeys[proposer_index]
randao_reveal =
if skipBlsValidation notin flags:
get_epoch_signature(
getStateField(state, fork),
getStateField(state, genesis_validators_root),
getStateField(state, slot).epoch, privKey).toValidatorSig()
else:
ValidatorSig()
let message = withState(state):
let execution_payload =
when consensusFork > ConsensusFork.Bellatrix:
default(consensusFork.ExecutionPayloadForSigning)
elif consensusFork == ConsensusFork.Bellatrix:
if cfg.CAPELLA_FORK_EPOCH != FAR_FUTURE_EPOCH:
# Can't keep correctly doing this once Capella happens, but LVH search
# test relies on merging. So, merge only if no Capella transition.
default(bellatrix.ExecutionPayloadForSigning)
else:
# Merge shortly after Bellatrix
if forkyState.data.slot >
cfg.BELLATRIX_FORK_EPOCH * SLOTS_PER_EPOCH + 10:
if is_merge_transition_complete(forkyState.data):
const feeRecipient = default(Eth1Address)
build_empty_execution_payload(forkyState.data, feeRecipient)
else:
build_empty_merge_execution_payload(forkyState.data)
else:
Support for driving multiple EL nodes from a single Nimbus BN (#4465) * Support for driving multiple EL nodes from a single Nimbus BN Full list of changes: * Eth1Monitor has been renamed to ELManager to match its current responsibilities better. * The ELManager is no longer optional in the code (it won't have a nil value under any circumstances). * The support for subscribing for headers was removed as it only worked with WebSockets and contributed significant complexity while bringing only a very minor advantage. * The `--web3-url` parameter has been deprecated in favor of a new `--el` parameter. The new parameter has a reasonable default value and supports specifying a different JWT for each connection. Each connection can also be configured with a different set of responsibilities (e.g. download deposits, validate blocks and/or produce blocks). On the command-line, these properties can be configured through URL properties stored in the #anchor part of the URL. In TOML files, they come with a very natural syntax (althrough the URL scheme is also supported). * The previously scattered EL-related state and logic is now moved to `eth1_monitor.nim` (this module will be renamed to `el_manager.nim` in a follow-up commit). State is assigned properly either to the `ELManager` or the to individual `ELConnection` objects where appropriate. The ELManager executes all Engine API requests against all attached EL nodes, in parallel. It compares their results and if there is a disagreement regarding the validity of a certain payload, this is detected and the beacon node is protected from publishing a block with a potential execution layer consensus bug in it. The BN provides metrics per EL node for the number of successful or failed requests for each type Engine API requests. If an EL node goes offline and connectivity is resoted later, we report the problem and the remedy in edge-triggered fashion. * More progress towards implementing Deneb block production in the VC and comparing the value of blocks produced by the EL and the builder API. * Adds a Makefile target for the zhejiang testnet
2023-03-05 01:40:21 +00:00
default(bellatrix.ExecutionPayloadForSigning)
else:
default(bellatrix.ExecutionPayloadForSigning)
makeBeaconBlock(
cfg,
state,
2023-01-11 12:29:21 +00:00
proposer_index,
randao_reveal,
# Keep deposit counts internally consistent.
Eth1Data(
deposit_root: eth1_data.deposit_root,
deposit_count: forkyState.data.eth1_deposit_index + deposits.lenu64,
block_hash: eth1_data.block_hash),
graffiti,
attestations,
deposits,
BeaconBlockValidatorChanges(),
sync_aggregate,
execution_payload,
noRollback,
cache,
verificationFlags = {skipBlsValidation})
if message.isErr:
raiseAssert "Failed to create a block: " & $message.error
let
new_block = signBlock(
getStateField(state, fork),
getStateField(state, genesis_validators_root), message.get(), privKey,
flags)
new_block
proc makeTestBlock*(
state: ForkedHashedBeaconState,
cache: var StateCache,
eth1_data = Eth1Data(),
attestations = newSeq[Attestation](),
deposits = newSeq[Deposit](),
sync_aggregate = SyncAggregate.init(),
graffiti = default(GraffitiBytes),
cfg = defaultRuntimeConfig): ForkedSignedBeaconBlock =
# Create a block for `state.slot + 1` - like a block proposer would do!
# It's a bit awkward - in order to produce a block for N+1, we need to
# calculate what the state will look like after that block has been applied,
# because the block includes the state root.
let tmpState = assignClone(state)
addTestBlock(
tmpState[], cache, eth1_data,
attestations, deposits, sync_aggregate, graffiti, cfg = cfg)
func makeAttestationData*(
state: ForkyBeaconState, slot: Slot, committee_index: CommitteeIndex,
beacon_block_root: Eth2Digest): AttestationData =
let
current_epoch = get_current_epoch(state)
start_slot = start_slot(current_epoch)
epoch_boundary_block_root =
if start_slot == state.slot: beacon_block_root
else: get_block_root_at_slot(state, start_slot)
doAssert slot.epoch == current_epoch,
"Computed epoch was " & $slot.epoch &
" while the state current_epoch was " & $current_epoch
# https://github.com/ethereum/consensus-specs/blob/v1.4.0-beta.7/specs/phase0/validator.md#attestation-data
AttestationData(
slot: slot,
index: committee_index.uint64,
beacon_block_root: beacon_block_root,
source: state.current_justified_checkpoint,
target: Checkpoint(
epoch: current_epoch,
root: epoch_boundary_block_root
)
)
2023-08-23 16:40:48 +00:00
func makeAttestationSig(
fork: Fork, genesis_validators_root: Eth2Digest, data: AttestationData,
committee: openArray[ValidatorIndex],
bits: CommitteeValidatorsBits): ValidatorSig =
let signing_root = compute_attestation_signing_root(
fork, genesis_validators_root, data)
var
agg {.noinit.}: AggregateSignature
first = true
for i in 0..<bits.len():
if not bits[i]: continue
let sig = blsSign(MockPrivKeys[committee[i]], signing_root.data)
if first:
agg.init(sig)
first = false
else:
agg.aggregate(sig)
if first:
ValidatorSig.infinity()
else:
agg.finish().toValidatorSig()
func makeAttestationData*(
state: ForkedHashedBeaconState, slot: Slot, committee_index: CommitteeIndex,
beacon_block_root: Eth2Digest): AttestationData =
## Create an attestation / vote for the block `beacon_block_root` using the
## data in `state` to fill in the rest of the fields.
## `state` is the state corresponding to the `beacon_block_root` advanced to
## the slot we're attesting to.
withState(state):
makeAttestationData(
forkyState.data, slot, committee_index, beacon_block_root)
2023-08-23 16:40:48 +00:00
func makeAttestation(
state: ForkedHashedBeaconState, beacon_block_root: Eth2Digest,
committee: seq[ValidatorIndex], slot: Slot, committee_index: CommitteeIndex,
validator_index: ValidatorIndex, cache: var StateCache,
flags: UpdateFlags = {}): Attestation =
let
index_in_committee = committee.find(validator_index)
data = makeAttestationData(state, slot, committee_index, beacon_block_root)
doAssert index_in_committee != -1, "find_beacon_committee should guarantee this"
var aggregation_bits = CommitteeValidatorsBits.init(committee.len)
aggregation_bits.setBit index_in_committee
2022-04-14 15:39:37 +00:00
let sig = if skipBlsValidation in flags:
ValidatorSig()
else:
makeAttestationSig(
getStateField(state, fork),
getStateField(state, genesis_validators_root),
data, committee, aggregation_bits)
Attestation(
data: data,
aggregation_bits: aggregation_bits,
signature: sig
)
func find_beacon_committee(
state: ForkedHashedBeaconState, validator_index: ValidatorIndex,
cache: var StateCache): auto =
let epoch = epoch(getStateField(state, slot))
for epoch_committee_index in 0'u64 ..< get_committee_count_per_slot(
state, epoch, cache) * SLOTS_PER_EPOCH:
let
slot = ((epoch_committee_index mod SLOTS_PER_EPOCH) +
epoch.start_slot.uint64).Slot
index = CommitteeIndex(epoch_committee_index div SLOTS_PER_EPOCH)
committee = get_beacon_committee(state, slot, index, cache)
if validator_index in committee:
return (committee, slot, index)
doAssert false
func makeAttestation*(
state: ForkedHashedBeaconState, beacon_block_root: Eth2Digest,
validator_index: ValidatorIndex, cache: var StateCache): Attestation =
let (committee, slot, index) =
find_beacon_committee(state, validator_index, cache)
makeAttestation(state, beacon_block_root, committee, slot, index,
validator_index, cache)
func makeFullAttestations*(
state: ForkedHashedBeaconState, beacon_block_root: Eth2Digest, slot: Slot,
cache: var StateCache,
flags: UpdateFlags = {}): seq[Attestation] =
# Create attestations in which the full committee participates for each shard
# that should be attested to during a particular slot
let committees_per_slot = get_committee_count_per_slot(
state, slot.epoch, cache)
for committee_index in get_committee_indices(committees_per_slot):
let
committee = get_beacon_committee(state, slot, committee_index, cache)
data = makeAttestationData(state, slot, committee_index, beacon_block_root)
2020-03-04 21:27:11 +00:00
doAssert committee.len() >= 1
var attestation = Attestation(
aggregation_bits: CommitteeValidatorsBits.init(committee.len),
data: data)
for i in 0..<committee.len:
attestation.aggregation_bits.setBit(i)
attestation.signature = makeAttestationSig(
getStateField(state, fork),
getStateField(state, genesis_validators_root), data, committee,
attestation.aggregation_bits)
result.add attestation
proc makeSyncAggregate(
state: ForkedHashedBeaconState,
syncCommitteeRatio: float,
cfg: RuntimeConfig): SyncAggregate =
if syncCommitteeRatio <= 0.0:
return SyncAggregate.init()
let
syncCommittee =
withState(state):
when consensusFork >= ConsensusFork.Altair:
if (forkyState.data.slot + 1).is_sync_committee_period():
forkyState.data.next_sync_committee
else:
forkyState.data.current_sync_committee
else:
return SyncAggregate.init()
fork =
getStateField(state, fork)
genesis_validators_root =
getStateField(state, genesis_validators_root)
slot =
getStateField(state, slot)
latest_block_id =
withState(state): forkyState.latest_block_id
rng = HmacDrbgContext.new()
syncCommitteePool = newClone(SyncCommitteeMsgPool.init(rng, cfg))
type
Aggregator = object
subcommitteeIdx: SyncSubcommitteeIndex
validatorIdx: ValidatorIndex
selectionProof: ValidatorSig
let
minActiveParticipants =
if syncCommitteeRatio >= 2.0 / 3: # Ensure supermajority is hit
(SYNC_COMMITTEE_SIZE * 2 + 2) div 3
else:
0
maxActiveParticipants = (syncCommitteeRatio * SYNC_COMMITTEE_SIZE).int
var
aggregators: seq[Aggregator]
numActiveParticipants = 0
for subcommitteeIdx in SyncSubcommitteeIndex:
let
firstKeyIdx = subcommitteeIdx.int * SYNC_SUBCOMMITTEE_SIZE
lastKeyIdx = firstKeyIdx + SYNC_SUBCOMMITTEE_SIZE - 1
var processedKeys = initHashSet[ValidatorPubKey]()
for idx, validatorKey in syncCommittee.pubkeys[firstKeyIdx .. lastKeyIdx]:
if validatorKey in processedKeys:
continue
processedKeys.incl validatorKey
let
validatorIdx =
block:
var res = 0
for i, validator in getStateField(state, validators):
if validator.pubkey == validatorKey:
res = i
break
res.ValidatorIndex
selectionProofSig = get_sync_committee_selection_proof(
fork, genesis_validators_root,
slot, subcommitteeIdx,
MockPrivKeys[validatorIdx])
if is_sync_committee_aggregator(selectionProofSig.toValidatorSig):
aggregators.add Aggregator(
subcommitteeIdx: subcommitteeIdx,
validatorIdx: validatorIdx,
selectionProof: selectionProofSig.toValidatorSig)
if numActiveParticipants >= minActiveParticipants and
rand(1.0) > syncCommitteeRatio:
continue
var positions: seq[uint64]
for pos, key in syncCommittee.pubkeys[firstKeyIdx + idx .. lastKeyIdx]:
if numActiveParticipants >= maxActiveParticipants:
break
if key == validatorKey:
positions.add (idx + pos).uint64
inc numActiveParticipants
if positions.len == 0:
continue
let signature = get_sync_committee_message_signature(
fork, genesis_validators_root,
slot, latest_block_id.root,
MockPrivKeys[validatorIdx])
syncCommitteePool[].addSyncCommitteeMessage(
slot,
latest_block_id,
uint64 validatorIdx,
signature,
subcommitteeIdx,
positions)
for aggregator in aggregators:
var contribution: SyncCommitteeContribution
if syncCommitteePool[].produceContribution(
slot, latest_block_id, aggregator.subcommitteeIdx, contribution):
let
contributionAndProof = ContributionAndProof(
aggregator_index: uint64 aggregator.validatorIdx,
contribution: contribution,
selection_proof: aggregator.selectionProof)
contributionSig = get_contribution_and_proof_signature(
fork, genesis_validators_root,
contributionAndProof,
MockPrivKeys[aggregator.validatorIdx])
signedContributionAndProof = SignedContributionAndProof(
message: contributionAndProof,
signature: contributionSig.toValidatorSig)
syncCommitteePool[].addContribution(
signedContributionAndProof,
latest_block_id, contribution.signature.load.get)
syncCommitteePool[].produceSyncAggregate(latest_block_id, slot + 1)
iterator makeTestBlocks*(
state: ForkedHashedBeaconState,
cache: var StateCache,
blocks: int,
eth1_data = Eth1Data(),
attested = false,
allDeposits = newSeq[Deposit](),
syncCommitteeRatio = 0.0,
graffiti = default(GraffitiBytes),
cfg = defaultRuntimeConfig): ForkedSignedBeaconBlock =
var state = assignClone(state)
for _ in 0..<blocks:
let
parent_root = withState(state[]): forkyState.latest_block_root
attestations =
if attested:
makeFullAttestations(
state[], parent_root, getStateField(state[], slot), cache)
else:
@[]
stateEth1 = getStateField(state[], eth1_data)
stateDepositIndex = getStateField(state[], eth1_deposit_index)
deposits =
if stateDepositIndex < stateEth1.deposit_count:
let
lowIndex = stateDepositIndex
numDeposits = min(MAX_DEPOSITS, stateEth1.deposit_count - lowIndex)
highIndex = lowIndex + numDeposits - 1
allDeposits[lowIndex .. highIndex]
else:
newSeq[Deposit]()
sync_aggregate = makeSyncAggregate(state[], syncCommitteeRatio, cfg)
yield addTestBlock(
state[], cache,
eth1_data = eth1_data,
attestations = attestations,
deposits = deposits,
sync_aggregate = sync_aggregate,
graffiti = graffiti,
cfg = cfg)