neobytes/test/functional/test_framework/test_framework.py
Alexander Block 2f21e55514 Remove legacy InstantSend code (#3020)
* Remove ppszTypeName from protocol.cpp and reimplement GetCommand

This removes the need to carefully maintain ppszTypeName, which required
correct order and also did not allow to permanently remove old message
types.

To get the command name for an INV type, GetCommandInternal uses a switch
which needs to be maintained from now on.

The way this is implemented also resembles the way it is implemented in
Bitcoin today, but it's not identical. The original PR that introduced the
switch case in Bitcoin was part of the Segwit changes and thus never got
backported. I decided to implement it in a slightly different way that
avoids throwing exceptions when an unknown INV type is encountered.

IsKnownType will now also leverage GetCommandInternal() to figure out if
the INV type is known locally. This has the side effect of old/legacy
message types to return false from now on. We will depend on this side
effect in later commits when we remove legacy InstantSend code.

* Stop handling/relaying legacy IX messages

When we receive an IX message, we simply treat it as a regular TX and relay
it as such.

We'll however still request IX messages when they are announced to us. We
can't simply revert to requesting TX messages in this case as it might
result in the other peer not answering due to the TX not being in mapRelay
yet. We should at some point in the future completely drop handling of IX
messages instead.

* Remove IsNewInstantSendEnabled() and only use IsInstantSendEnabled()

* Remove legacy InstantSend from GUI

* Remove InstantSend from Bitcoin/Dash URIs

* Remove legacy InstantSend from RPC commands

* Remove legacy InstantSend from wallet

* Remove legacy instantsend.h include

* Remove legacy InstantSend from validation code

* Completely remove remaining legacy InstantSend code

* Remove now unused spork

* Fix InstantSend related test failures

* Remove now obsolete auto IS tests

* Make spork2 and spork3 disabled by default

This should have no influence on mainnet as these sporks are actually set
there. This will however affect regtest, which shouldn't have LLMQ based
InstantSend enabled by default.

* Remove instantsend tests from dip3-deterministicmns.py

These were only testing legacy InstantSend

* Fix .QCheckBox#checkUsePrivateSend styling a bit

* s/TXLEGACYLOCKREQUEST/LEGACYTXLOCKREQUEST/

* Revert "verified via InstantSend" back to "verified via LLMQ based InstantSend"

* Use cmd == nullptr instead of !cmd

* Remove last parameter from AvailableCoins call

This was for fUseInstantSend which is not present anymore since rebase
2019-07-09 17:50:08 +03:00

783 lines
30 KiB
Python
Executable File

#!/usr/bin/env python3
# Copyright (c) 2014-2016 The Bitcoin Core developers
# Copyright (c) 2014-2019 The Dash Core developers
# Distributed under the MIT software license, see the accompanying
# file COPYING or http://www.opensource.org/licenses/mit-license.php.
"""Base class for RPC testing."""
from collections import deque
from enum import Enum
import logging
import optparse
import os
import shutil
import subprocess
import sys
import tempfile
import time
from concurrent.futures import ThreadPoolExecutor
from .util import (
PortSeed,
GENESISTIME,
MAX_NODES,
assert_equal,
bitcoind_processes,
check_json_precision,
connect_nodes_bi,
connect_nodes,
copy_datadir,
disable_mocktime,
disconnect_nodes,
enable_coverage,
get_mocktime,
get_rpc_proxy,
initialize_datadir,
log_filename,
p2p_port,
rpc_url,
set_cache_mocktime,
set_genesis_mocktime,
set_mocktime,
set_node_times,
satoshi_round,
_start_node,
_start_nodes,
_stop_node,
_stop_nodes,
sync_blocks,
sync_mempools,
sync_masternodes,
wait_for_bitcoind_start,
wait_to_sync)
from .authproxy import JSONRPCException
class TestStatus(Enum):
PASSED = 1
FAILED = 2
SKIPPED = 3
TEST_EXIT_PASSED = 0
TEST_EXIT_FAILED = 1
TEST_EXIT_SKIPPED = 77
class BitcoinTestFramework(object):
"""Base class for a bitcoin test script.
Individual bitcoin test scripts should subclass this class and override the following methods:
- __init__()
- add_options()
- setup_chain()
- setup_network()
- run_test()
The main() method should not be overridden.
This class also contains various public and private helper methods."""
# Methods to override in subclass test scripts.
def __init__(self):
self.num_nodes = 4
self.setup_clean_chain = False
self.nodes = None
def add_options(self, parser):
pass
def setup_chain(self):
self.log.info("Initializing test directory "+self.options.tmpdir)
if self.setup_clean_chain:
self._initialize_chain_clean(self.options.tmpdir, self.num_nodes)
set_genesis_mocktime()
else:
self._initialize_chain(self.options.tmpdir, self.num_nodes, self.options.cachedir)
set_cache_mocktime()
def setup_network(self):
self.setup_nodes()
# Connect the nodes as a "chain". This allows us
# to split the network between nodes 1 and 2 to get
# two halves that can work on competing chains.
for i in range(self.num_nodes - 1):
connect_nodes_bi(self.nodes, i, i + 1)
self.sync_all()
def setup_nodes(self, stderr=None):
extra_args = None
if hasattr(self, "extra_args"):
extra_args = self.extra_args
self.nodes = _start_nodes(self.num_nodes, self.options.tmpdir, extra_args, stderr=stderr)
def run_test(self):
raise NotImplementedError
# Main function. This should not be overridden by the subclass test scripts.
def main(self):
parser = optparse.OptionParser(usage="%prog [options]")
parser.add_option("--nocleanup", dest="nocleanup", default=False, action="store_true",
help="Leave dashds and test.* datadir on exit or error")
parser.add_option("--noshutdown", dest="noshutdown", default=False, action="store_true",
help="Don't stop dashds after the test execution")
parser.add_option("--srcdir", dest="srcdir", default=os.path.normpath(os.path.dirname(os.path.realpath(__file__))+"/../../../src"),
help="Source directory containing dashd/dash-cli (default: %default)")
parser.add_option("--cachedir", dest="cachedir", default=os.path.normpath(os.path.dirname(os.path.realpath(__file__))+"/../../cache"),
help="Directory for caching pregenerated datadirs")
parser.add_option("--tmpdir", dest="tmpdir", help="Root directory for datadirs")
parser.add_option("-l", "--loglevel", dest="loglevel", default="INFO",
help="log events at this level and higher to the console. Can be set to DEBUG, INFO, WARNING, ERROR or CRITICAL. Passing --loglevel DEBUG will output all logs to console. Note that logs at all levels are always written to the test_framework.log file in the temporary test directory.")
parser.add_option("--tracerpc", dest="trace_rpc", default=False, action="store_true",
help="Print out all RPC calls as they are made")
parser.add_option("--portseed", dest="port_seed", default=os.getpid(), type='int',
help="The seed to use for assigning port numbers (default: current process id)")
parser.add_option("--coveragedir", dest="coveragedir",
help="Write tested RPC commands into this directory")
parser.add_option("--configfile", dest="configfile",
help="Location of the test framework config file")
self.add_options(parser)
(self.options, self.args) = parser.parse_args()
if self.options.coveragedir:
enable_coverage(self.options.coveragedir)
PortSeed.n = self.options.port_seed
os.environ['PATH'] = self.options.srcdir+":"+self.options.srcdir+"/qt:"+os.environ['PATH']
check_json_precision()
# Set up temp directory and start logging
if self.options.tmpdir:
os.makedirs(self.options.tmpdir, exist_ok=False)
else:
self.options.tmpdir = tempfile.mkdtemp(prefix="test")
self._start_logging()
success = TestStatus.FAILED
try:
self.setup_chain()
self.setup_network()
self.run_test()
success = TestStatus.PASSED
except JSONRPCException as e:
self.log.exception("JSONRPC error")
except SkipTest as e:
self.log.warning("Test Skipped: %s" % e.message)
success = TestStatus.SKIPPED
except AssertionError as e:
self.log.exception("Assertion failed")
except KeyError as e:
self.log.exception("Key error")
except Exception as e:
self.log.exception("Unexpected exception caught during testing")
except KeyboardInterrupt as e:
self.log.warning("Exiting after keyboard interrupt")
if not self.options.noshutdown:
self.log.info("Stopping nodes")
try:
if self.nodes:
self.stop_nodes()
except BaseException as e:
success = False
self.log.exception("Unexpected exception caught during shutdown")
else:
self.log.info("Note: dashds were not stopped and may still be running")
if not self.options.nocleanup and not self.options.noshutdown and success != TestStatus.FAILED:
self.log.info("Cleaning up")
shutil.rmtree(self.options.tmpdir)
else:
self.log.warning("Not cleaning up dir %s" % self.options.tmpdir)
if os.getenv("PYTHON_DEBUG", ""):
# Dump the end of the debug logs, to aid in debugging rare
# travis failures.
import glob
filenames = [self.options.tmpdir + "/test_framework.log"]
filenames += glob.glob(self.options.tmpdir + "/node*/regtest/debug.log")
MAX_LINES_TO_PRINT = 1000
for fn in filenames:
try:
with open(fn, 'r') as f:
print("From" , fn, ":")
print("".join(deque(f, MAX_LINES_TO_PRINT)))
except OSError:
print("Opening file %s failed." % fn)
traceback.print_exc()
if success == TestStatus.PASSED:
self.log.info("Tests successful")
sys.exit(TEST_EXIT_PASSED)
elif success == TestStatus.SKIPPED:
self.log.info("Test skipped")
sys.exit(TEST_EXIT_SKIPPED)
else:
self.log.error("Test failed. Test logging available at %s/test_framework.log", self.options.tmpdir)
logging.shutdown()
sys.exit(TEST_EXIT_FAILED)
# Public helper methods. These can be accessed by the subclass test scripts.
def start_node(self, i, dirname, extra_args=None, rpchost=None, timewait=None, binary=None, stderr=None):
return _start_node(i, dirname, extra_args, rpchost, timewait, binary, stderr)
def start_nodes(self, num_nodes, dirname, extra_args=None, rpchost=None, timewait=None, binary=None, stderr=None):
return _start_nodes(num_nodes, dirname, extra_args, rpchost, timewait, binary, stderr)
def stop_node(self, num_node):
_stop_node(self.nodes[num_node], num_node)
def stop_nodes(self):
_stop_nodes(self.nodes)
def split_network(self):
"""
Split the network of four nodes into nodes 0/1 and 2/3.
"""
disconnect_nodes(self.nodes[1], 2)
disconnect_nodes(self.nodes[2], 1)
self.sync_all([self.nodes[:2], self.nodes[2:]])
def join_network(self):
"""
Join the (previously split) network halves together.
"""
connect_nodes_bi(self.nodes, 1, 2)
self.sync_all()
def sync_all(self, node_groups=None):
if not node_groups:
node_groups = [self.nodes]
for group in node_groups:
sync_blocks(group)
sync_mempools(group)
# Private helper methods. These should not be accessed by the subclass test scripts.
def _start_logging(self):
# Add logger and logging handlers
self.log = logging.getLogger('TestFramework')
self.log.setLevel(logging.DEBUG)
# Create file handler to log all messages
fh = logging.FileHandler(self.options.tmpdir + '/test_framework.log')
fh.setLevel(logging.DEBUG)
# Create console handler to log messages to stderr. By default this logs only error messages, but can be configured with --loglevel.
ch = logging.StreamHandler(sys.stdout)
# User can provide log level as a number or string (eg DEBUG). loglevel was caught as a string, so try to convert it to an int
ll = int(self.options.loglevel) if self.options.loglevel.isdigit() else self.options.loglevel.upper()
ch.setLevel(ll)
# Format logs the same as bitcoind's debug.log with microprecision (so log files can be concatenated and sorted)
formatter = logging.Formatter(fmt = '%(asctime)s.%(msecs)03d000 %(name)s (%(levelname)s): %(message)s', datefmt='%Y-%m-%d %H:%M:%S')
formatter.converter = time.gmtime
fh.setFormatter(formatter)
ch.setFormatter(formatter)
# add the handlers to the logger
self.log.addHandler(fh)
self.log.addHandler(ch)
if self.options.trace_rpc:
rpc_logger = logging.getLogger("BitcoinRPC")
rpc_logger.setLevel(logging.DEBUG)
rpc_handler = logging.StreamHandler(sys.stdout)
rpc_handler.setLevel(logging.DEBUG)
rpc_logger.addHandler(rpc_handler)
def _initialize_chain(self, test_dir, num_nodes, cachedir, extra_args=None, stderr=None):
"""Initialize a pre-mined blockchain for use by the test.
Create a cache of a 200-block-long chain (with wallet) for MAX_NODES
Afterward, create num_nodes copies from the cache."""
assert num_nodes <= MAX_NODES
create_cache = False
for i in range(MAX_NODES):
if not os.path.isdir(os.path.join(cachedir, 'node' + str(i))):
create_cache = True
break
if create_cache:
self.log.debug("Creating data directories from cached datadir")
# find and delete old cache directories if any exist
for i in range(MAX_NODES):
if os.path.isdir(os.path.join(cachedir, "node" + str(i))):
shutil.rmtree(os.path.join(cachedir, "node" + str(i)))
# Create cache directories, run dashds:
set_genesis_mocktime()
for i in range(MAX_NODES):
datadir = initialize_datadir(cachedir, i)
args = [os.getenv("DASHD", "dashd"), "-server", "-keypool=1", "-datadir=" + datadir, "-discover=0", "-mocktime="+str(GENESISTIME)]
if i > 0:
args.append("-connect=127.0.0.1:" + str(p2p_port(0)))
if extra_args is not None:
args.extend(extra_args)
bitcoind_processes[i] = subprocess.Popen(args, stderr=stderr)
self.log.debug("initialize_chain: dashd started, waiting for RPC to come up")
wait_for_bitcoind_start(bitcoind_processes[i], rpc_url(i), i)
self.log.debug("initialize_chain: RPC successfully started")
self.nodes = []
for i in range(MAX_NODES):
try:
self.nodes.append(get_rpc_proxy(rpc_url(i), i))
except:
self.log.exception("Error connecting to node %d" % i)
sys.exit(1)
# Create a 200-block-long chain; each of the 4 first nodes
# gets 25 mature blocks and 25 immature.
# Note: To preserve compatibility with older versions of
# initialize_chain, only 4 nodes will generate coins.
#
# blocks are created with timestamps 10 minutes apart
# starting from 2010 minutes in the past
block_time = GENESISTIME
for i in range(2):
for peer in range(4):
for j in range(25):
set_node_times(self.nodes, block_time)
self.nodes[peer].generate(1)
block_time += 156
# Must sync before next peer starts generating blocks
sync_blocks(self.nodes)
# Shut them down, and clean up cache directories:
self.stop_nodes()
self.nodes = []
disable_mocktime()
for i in range(MAX_NODES):
os.remove(log_filename(cachedir, i, "debug.log"))
os.remove(log_filename(cachedir, i, "db.log"))
os.remove(log_filename(cachedir, i, "peers.dat"))
os.remove(log_filename(cachedir, i, "fee_estimates.dat"))
for i in range(num_nodes):
from_dir = os.path.join(cachedir, "node" + str(i))
to_dir = os.path.join(test_dir, "node" + str(i))
shutil.copytree(from_dir, to_dir)
initialize_datadir(test_dir, i) # Overwrite port/rpcport in dsah.conf
def _initialize_chain_clean(self, test_dir, num_nodes):
"""Initialize empty blockchain for use by the test.
Create an empty blockchain and num_nodes wallets.
Useful if a test case wants complete control over initialization."""
for i in range(num_nodes):
initialize_datadir(test_dir, i)
MASTERNODE_COLLATERAL = 1000
class MasternodeInfo:
def __init__(self, proTxHash, ownerAddr, votingAddr, pubKeyOperator, keyOperator, collateral_address, collateral_txid, collateral_vout):
self.proTxHash = proTxHash
self.ownerAddr = ownerAddr
self.votingAddr = votingAddr
self.pubKeyOperator = pubKeyOperator
self.keyOperator = keyOperator
self.collateral_address = collateral_address
self.collateral_txid = collateral_txid
self.collateral_vout = collateral_vout
class DashTestFramework(BitcoinTestFramework):
def __init__(self, num_nodes, masterodes_count, extra_args, fast_dip3_enforcement=False):
super().__init__()
self.mn_count = masterodes_count
self.num_nodes = num_nodes
self.mninfo = []
self.setup_clean_chain = True
self.is_network_split = False
# additional args
self.extra_args = extra_args
self.extra_args += ["-sporkkey=cP4EKFyJsHT39LDqgdcB43Y3YXjNyjb5Fuas1GQSeAtjnZWmZEQK"]
self.fast_dip3_enforcement = fast_dip3_enforcement
if fast_dip3_enforcement:
self.extra_args += ["-dip3params=30:50"]
def create_simple_node(self):
idx = len(self.nodes)
args = self.extra_args
self.nodes.append(self.start_node(idx, self.options.tmpdir, args))
for i in range(0, idx):
connect_nodes(self.nodes[i], idx)
def prepare_masternodes(self):
for idx in range(0, self.mn_count):
self.prepare_masternode(idx)
def prepare_masternode(self, idx):
bls = self.nodes[0].bls('generate')
address = self.nodes[0].getnewaddress()
txid = self.nodes[0].sendtoaddress(address, MASTERNODE_COLLATERAL)
txraw = self.nodes[0].getrawtransaction(txid, True)
collateral_vout = 0
for vout_idx in range(0, len(txraw["vout"])):
vout = txraw["vout"][vout_idx]
if vout["value"] == MASTERNODE_COLLATERAL:
collateral_vout = vout_idx
self.nodes[0].lockunspent(False, [{'txid': txid, 'vout': collateral_vout}])
# send to same address to reserve some funds for fees
self.nodes[0].sendtoaddress(address, 0.001)
ownerAddr = self.nodes[0].getnewaddress()
votingAddr = self.nodes[0].getnewaddress()
rewardsAddr = self.nodes[0].getnewaddress()
port = p2p_port(len(self.nodes) + idx)
if (idx % 2) == 0:
self.nodes[0].lockunspent(True, [{'txid': txid, 'vout': collateral_vout}])
proTxHash = self.nodes[0].protx('register_fund', address, '127.0.0.1:%d' % port, ownerAddr, bls['public'], votingAddr, 0, rewardsAddr, address)
else:
self.nodes[0].generate(1)
proTxHash = self.nodes[0].protx('register', txid, collateral_vout, '127.0.0.1:%d' % port, ownerAddr, bls['public'], votingAddr, 0, rewardsAddr, address)
self.nodes[0].generate(1)
self.mninfo.append(MasternodeInfo(proTxHash, ownerAddr, votingAddr, bls['public'], bls['secret'], address, txid, collateral_vout))
self.sync_all()
def remove_mastermode(self, idx):
mn = self.mninfo[idx]
rawtx = self.nodes[0].createrawtransaction([{"txid": mn.collateral_txid, "vout": mn.collateral_vout}], {self.nodes[0].getnewaddress(): 999.9999})
rawtx = self.nodes[0].signrawtransaction(rawtx)
self.nodes[0].sendrawtransaction(rawtx["hex"])
self.nodes[0].generate(1)
self.sync_all()
self.mninfo.remove(mn)
def prepare_datadirs(self):
# stop faucet node so that we can copy the datadir
self.stop_node(0)
start_idx = len(self.nodes)
for idx in range(0, self.mn_count):
copy_datadir(0, idx + start_idx, self.options.tmpdir)
# restart faucet node
self.nodes[0] = self.start_node(0, self.options.tmpdir, self.extra_args)
def start_masternodes(self):
start_idx = len(self.nodes)
for idx in range(0, self.mn_count):
self.nodes.append(None)
executor = ThreadPoolExecutor(max_workers=20)
def do_start(idx):
args = ['-masternode=1',
'-masternodeblsprivkey=%s' % self.mninfo[idx].keyOperator] + self.extra_args
node = self.start_node(idx + start_idx, self.options.tmpdir, args)
self.mninfo[idx].nodeIdx = idx + start_idx
self.mninfo[idx].node = node
self.nodes[idx + start_idx] = node
wait_to_sync(node, True)
def do_connect(idx):
for i in range(0, idx + 1):
connect_nodes(self.nodes[idx + start_idx], i)
jobs = []
# start up nodes in parallel
for idx in range(0, self.mn_count):
jobs.append(executor.submit(do_start, idx))
# wait for all nodes to start up
for job in jobs:
job.result()
jobs.clear()
# connect nodes in parallel
for idx in range(0, self.mn_count):
jobs.append(executor.submit(do_connect, idx))
# wait for all nodes to connect
for job in jobs:
job.result()
jobs.clear()
sync_masternodes(self.nodes, True)
executor.shutdown()
def setup_network(self):
self.nodes = []
# create faucet node for collateral and transactions
self.nodes.append(self.start_node(0, self.options.tmpdir, self.extra_args))
required_balance = MASTERNODE_COLLATERAL * self.mn_count + 1
while self.nodes[0].getbalance() < required_balance:
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(1)
# create connected simple nodes
for i in range(0, self.num_nodes - self.mn_count - 1):
self.create_simple_node()
sync_masternodes(self.nodes, True)
# activate DIP3
if not self.fast_dip3_enforcement:
while self.nodes[0].getblockcount() < 500:
self.nodes[0].generate(10)
self.sync_all()
# create masternodes
self.prepare_masternodes()
self.prepare_datadirs()
self.start_masternodes()
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(1)
# sync nodes
self.sync_all()
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
mn_info = self.nodes[0].masternodelist("status")
assert (len(mn_info) == self.mn_count)
for status in mn_info.values():
assert (status == 'ENABLED')
def create_raw_tx(self, node_from, node_to, amount, min_inputs, max_inputs):
assert (min_inputs <= max_inputs)
# fill inputs
inputs = []
balances = node_from.listunspent()
in_amount = 0.0
last_amount = 0.0
for tx in balances:
if len(inputs) < min_inputs:
input = {}
input["txid"] = tx['txid']
input['vout'] = tx['vout']
in_amount += float(tx['amount'])
inputs.append(input)
elif in_amount > amount:
break
elif len(inputs) < max_inputs:
input = {}
input["txid"] = tx['txid']
input['vout'] = tx['vout']
in_amount += float(tx['amount'])
inputs.append(input)
else:
input = {}
input["txid"] = tx['txid']
input['vout'] = tx['vout']
in_amount -= last_amount
in_amount += float(tx['amount'])
inputs[-1] = input
last_amount = float(tx['amount'])
assert (len(inputs) >= min_inputs)
assert (len(inputs) <= max_inputs)
assert (in_amount >= amount)
# fill outputs
receiver_address = node_to.getnewaddress()
change_address = node_from.getnewaddress()
fee = 0.001
outputs = {}
outputs[receiver_address] = satoshi_round(amount)
outputs[change_address] = satoshi_round(in_amount - amount - fee)
rawtx = node_from.createrawtransaction(inputs, outputs)
ret = node_from.signrawtransaction(rawtx)
decoded = node_from.decoderawtransaction(ret['hex'])
ret = {**decoded, **ret}
return ret
def wait_for_instantlock(self, txid, node):
# wait for instantsend locks
start = time.time()
locked = False
while True:
try:
is_tx = node.getrawtransaction(txid, True)
if is_tx['instantlock']:
locked = True
break
except:
# TX not received yet?
pass
if time.time() > start + 10:
break
time.sleep(0.5)
return locked
def wait_for_sporks_same(self, timeout=30):
st = time.time()
while time.time() < st + timeout:
if self.check_sporks_same():
return
time.sleep(0.5)
raise AssertionError("wait_for_sporks_same timed out")
def check_sporks_same(self):
sporks = self.nodes[0].spork('show')
for node in self.nodes[1:]:
sporks2 = node.spork('show')
if sporks != sporks2:
return False
return True
def wait_for_quorum_phase(self, phase, check_received_messages, check_received_messages_count, timeout=30):
t = time.time()
while time.time() - t < timeout:
all_ok = True
for mn in self.mninfo:
s = mn.node.quorum("dkgstatus")["session"]
if "llmq_5_60" not in s:
all_ok = False
break
s = s["llmq_5_60"]
if "phase" not in s:
all_ok = False
break
if s["phase"] != phase:
all_ok = False
break
if check_received_messages is not None:
if s[check_received_messages] < check_received_messages_count:
all_ok = False
break
if all_ok:
return
time.sleep(0.1)
raise AssertionError("wait_for_quorum_phase timed out")
def wait_for_quorum_commitment(self, timeout = 15):
t = time.time()
while time.time() - t < timeout:
all_ok = True
for node in self.nodes:
s = node.quorum("dkgstatus")
if "minableCommitments" not in s:
all_ok = False
break
s = s["minableCommitments"]
if "llmq_5_60" not in s:
all_ok = False
break
if all_ok:
return
time.sleep(0.1)
raise AssertionError("wait_for_quorum_commitment timed out")
def mine_quorum(self, expected_contributions=5, expected_complaints=0, expected_justifications=0, expected_commitments=5):
quorums = self.nodes[0].quorum("list")
# move forward to next DKG
skip_count = 24 - (self.nodes[0].getblockcount() % 24)
if skip_count != 0:
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(skip_count)
sync_blocks(self.nodes)
# Make sure all reached phase 1 (init)
self.wait_for_quorum_phase(1, None, 0)
# Give nodes some time to connect to neighbors
time.sleep(2)
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(2)
sync_blocks(self.nodes)
# Make sure all reached phase 2 (contribute) and received all contributions
self.wait_for_quorum_phase(2, "receivedContributions", expected_contributions)
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(2)
sync_blocks(self.nodes)
# Make sure all reached phase 3 (complain) and received all complaints
self.wait_for_quorum_phase(3, "receivedComplaints", expected_complaints)
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(2)
sync_blocks(self.nodes)
# Make sure all reached phase 4 (justify)
self.wait_for_quorum_phase(4, "receivedJustifications", expected_justifications)
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(2)
sync_blocks(self.nodes)
# Make sure all reached phase 5 (commit)
self.wait_for_quorum_phase(5, "receivedPrematureCommitments", expected_commitments)
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(2)
sync_blocks(self.nodes)
# Make sure all reached phase 6 (mining)
self.wait_for_quorum_phase(6, None, 0)
# Wait for final commitment
self.wait_for_quorum_commitment()
# mine the final commitment
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(1)
while quorums == self.nodes[0].quorum("list"):
time.sleep(2)
set_mocktime(get_mocktime() + 1)
set_node_times(self.nodes, get_mocktime())
self.nodes[0].generate(1)
sync_blocks(self.nodes)
new_quorum = self.nodes[0].quorum("list", 1)["llmq_5_60"][0]
# Mine 8 (SIGN_HEIGHT_OFFSET) more blocks to make sure that the new quorum gets eligable for signing sessions
self.nodes[0].generate(8)
sync_blocks(self.nodes)
return new_quorum
# Test framework for doing p2p comparison testing, which sets up some bitcoind
# binaries:
# 1 binary: test binary
# 2 binaries: 1 test binary, 1 ref binary
# n>2 binaries: 1 test binary, n-1 ref binaries
class SkipTest(Exception):
"""This exception is raised to skip a test"""
def __init__(self, message):
self.message = message
class ComparisonTestFramework(BitcoinTestFramework):
def __init__(self):
super().__init__()
self.num_nodes = 2
self.setup_clean_chain = True
def add_options(self, parser):
parser.add_option("--testbinary", dest="testbinary",
default=os.getenv("BITCOIND", "dashd"),
help="dashd binary to test")
parser.add_option("--refbinary", dest="refbinary",
default=os.getenv("BITCOIND", "dashd"),
help="dashd binary to use for reference nodes (if any)")
def setup_network(self):
extra_args = [['-whitelist=127.0.0.1']]*self.num_nodes
if hasattr(self, "extra_args"):
extra_args = self.extra_args
self.nodes = self.start_nodes(
self.num_nodes, self.options.tmpdir, extra_args,
binary=[self.options.testbinary] +
[self.options.refbinary]*(self.num_nodes-1))