diff --git a/test/functional/bip68-sequence.py b/test/functional/bip68-sequence.py index 7b2e6d259..dec3c9475 100755 --- a/test/functional/bip68-sequence.py +++ b/test/functional/bip68-sequence.py @@ -1,464 +1,484 @@ #!/usr/bin/env python3 # Copyright (c) 2014-2016 The Bitcoin Core developers # Distributed under the MIT software license, see the accompanying # file COPYING or http://www.opensource.org/licenses/mit-license.php. # # Test BIP68 implementation # from test_framework.test_framework import BitcoinTestFramework from test_framework.util import * from test_framework.script import * from test_framework.mininode import * from test_framework.blocktools import * +from test_framework.txtools import pad_tx, pad_raw_tx SEQUENCE_LOCKTIME_DISABLE_FLAG = (1 << 31) SEQUENCE_LOCKTIME_TYPE_FLAG = (1 << 22) # this means use time (0 means height) SEQUENCE_LOCKTIME_GRANULARITY = 9 # this is a bit-shift SEQUENCE_LOCKTIME_MASK = 0x0000ffff # RPC error for non-BIP68 final transactions NOT_FINAL_ERROR = "64: non-BIP68-final" class BIP68Test(BitcoinTestFramework): + def add_options(self, parser): + parser.add_argument("--magnetic-anomaly-time", dest="magnetic_anomaly_time", type=long, default=0, + help="Test Magnetic Anomaly activation at a particular time (0 for never, 1 for startup)") + def set_test_params(self): self.num_nodes = 2 self.extra_args = [["-blockprioritypercentage=0"], ["-blockprioritypercentage=0", "-acceptnonstdtxn=0"]] + activation_time = self.options.magnetic_anomaly_time + if activation_time != 0: + for args in self.extra_args: + args.append( + "-magneticanomalyactivationtime={}".format(activation_time)) + def run_test(self): self.relayfee = self.nodes[0].getnetworkinfo()["relayfee"] # Generate some coins self.nodes[0].generate(110) self.log.info("Running test disable flag") self.test_disable_flag() self.log.info("Running test sequence-lock-confirmed-inputs") self.test_sequence_lock_confirmed_inputs() self.log.info("Running test sequence-lock-unconfirmed-inputs") self.test_sequence_lock_unconfirmed_inputs() self.log.info( "Running test BIP68 not consensus before versionbits activation") self.test_bip68_not_consensus() self.log.info("Verifying nVersion=2 transactions aren't standard") self.test_version2_relay(before_activation=True) self.log.info("Activating BIP68 (and 112/113)") self.activateCSV() self.log.info("Verifying nVersion=2 transactions are now standard") self.test_version2_relay(before_activation=False) self.log.info("Passed") # Test that BIP68 is not in effect if tx version is 1, or if # the first sequence bit is set. def test_disable_flag(self): # Create some unconfirmed inputs new_addr = self.nodes[0].getnewaddress() self.nodes[0].sendtoaddress(new_addr, 2) # send 2 BTC utxos = self.nodes[0].listunspent(0, 0) assert(len(utxos) > 0) utxo = utxos[0] tx1 = CTransaction() value = int(satoshi_round(utxo["amount"] - self.relayfee) * COIN) # Check that the disable flag disables relative locktime. # If sequence locks were used, this would require 1 block for the # input to mature. sequence_value = SEQUENCE_LOCKTIME_DISABLE_FLAG | 1 tx1.vin = [ CTxIn(COutPoint(int(utxo["txid"], 16), utxo["vout"]), nSequence=sequence_value)] tx1.vout = [CTxOut(value, CScript([b'a']))] + pad_tx(tx1) tx1_signed = self.nodes[0].signrawtransaction(ToHex(tx1))["hex"] tx1_id = self.nodes[0].sendrawtransaction(tx1_signed) tx1_id = int(tx1_id, 16) # This transaction will enable sequence-locks, so this transaction should # fail tx2 = CTransaction() tx2.nVersion = 2 sequence_value = sequence_value & 0x7fffffff tx2.vin = [CTxIn(COutPoint(tx1_id, 0), nSequence=sequence_value)] tx2.vout = [CTxOut(int(value - self.relayfee * COIN), CScript([b'a']))] + pad_tx(tx2) tx2.rehash() assert_raises_rpc_error(-26, NOT_FINAL_ERROR, self.nodes[0].sendrawtransaction, ToHex(tx2)) # Setting the version back down to 1 should disable the sequence lock, # so this should be accepted. tx2.nVersion = 1 self.nodes[0].sendrawtransaction(ToHex(tx2)) # Calculate the median time past of a prior block ("confirmations" before # the current tip). def get_median_time_past(self, confirmations): block_hash = self.nodes[0].getblockhash( self.nodes[0].getblockcount() - confirmations) return self.nodes[0].getblockheader(block_hash)["mediantime"] # Test that sequence locks are respected for transactions spending # confirmed inputs. def test_sequence_lock_confirmed_inputs(self): # Create lots of confirmed utxos, and use them to generate lots of random # transactions. max_outputs = 50 addresses = [] while len(addresses) < max_outputs: addresses.append(self.nodes[0].getnewaddress()) while len(self.nodes[0].listunspent()) < 200: import random random.shuffle(addresses) num_outputs = random.randint(1, max_outputs) outputs = {} for i in range(num_outputs): outputs[addresses[i]] = random.randint(1, 20) * 0.01 self.nodes[0].sendmany("", outputs) self.nodes[0].generate(1) utxos = self.nodes[0].listunspent() # Try creating a lot of random transactions. # Each time, choose a random number of inputs, and randomly set # some of those inputs to be sequence locked (and randomly choose # between height/time locking). Small random chance of making the locks # all pass. for i in range(400): # Randomly choose up to 10 inputs num_inputs = random.randint(1, 10) random.shuffle(utxos) # Track whether any sequence locks used should fail should_pass = True # Track whether this transaction was built with sequence locks using_sequence_locks = False tx = CTransaction() tx.nVersion = 2 value = 0 for j in range(num_inputs): sequence_value = 0xfffffffe # this disables sequence locks # 50% chance we enable sequence locks if random.randint(0, 1): using_sequence_locks = True # 10% of the time, make the input sequence value pass input_will_pass = (random.randint(1, 10) == 1) sequence_value = utxos[j]["confirmations"] if not input_will_pass: sequence_value += 1 should_pass = False # Figure out what the median-time-past was for the confirmed input # Note that if an input has N confirmations, we're going back N blocks # from the tip so that we're looking up MTP of the block # PRIOR to the one the input appears in, as per the BIP68 # spec. orig_time = self.get_median_time_past( utxos[j]["confirmations"]) cur_time = self.get_median_time_past(0) # MTP of the tip # can only timelock this input if it's not too old -- # otherwise use height can_time_lock = True if ((cur_time - orig_time) >> SEQUENCE_LOCKTIME_GRANULARITY) >= SEQUENCE_LOCKTIME_MASK: can_time_lock = False # if time-lockable, then 50% chance we make this a time # lock if random.randint(0, 1) and can_time_lock: # Find first time-lock value that fails, or latest one # that succeeds time_delta = sequence_value << SEQUENCE_LOCKTIME_GRANULARITY if input_will_pass and time_delta > cur_time - orig_time: sequence_value = ( (cur_time - orig_time) >> SEQUENCE_LOCKTIME_GRANULARITY) elif (not input_will_pass and time_delta <= cur_time - orig_time): sequence_value = ( (cur_time - orig_time) >> SEQUENCE_LOCKTIME_GRANULARITY) + 1 sequence_value |= SEQUENCE_LOCKTIME_TYPE_FLAG tx.vin.append( CTxIn(COutPoint(int(utxos[j]["txid"], 16), utxos[j]["vout"]), nSequence=sequence_value)) value += utxos[j]["amount"] * COIN # Overestimate the size of the tx - signatures should be less than # 120 bytes, and leave 50 for the output tx_size = len(ToHex(tx)) // 2 + 120 * num_inputs + 50 tx.vout.append( CTxOut(int(value - self.relayfee * tx_size * COIN / 1000), CScript([b'a']))) rawtx = self.nodes[0].signrawtransaction(ToHex(tx))["hex"] if (using_sequence_locks and not should_pass): # This transaction should be rejected assert_raises_rpc_error(-26, NOT_FINAL_ERROR, self.nodes[0].sendrawtransaction, rawtx) else: # This raw transaction should be accepted self.nodes[0].sendrawtransaction(rawtx) utxos = self.nodes[0].listunspent() # Test that sequence locks on unconfirmed inputs must have nSequence # height or time of 0 to be accepted. # Then test that BIP68-invalid transactions are removed from the mempool # after a reorg. def test_sequence_lock_unconfirmed_inputs(self): # Store height so we can easily reset the chain at the end of the test cur_height = self.nodes[0].getblockcount() # Create a mempool tx. txid = self.nodes[0].sendtoaddress(self.nodes[0].getnewaddress(), 2) tx1 = FromHex(CTransaction(), self.nodes[0].getrawtransaction(txid)) tx1.rehash() # As the fees are calculated prior to the transaction being signed, # there is some uncertainty that calculate fee provides the correct # minimal fee. Since regtest coins are free, let's go ahead and # increase the fee by an order of magnitude to ensure this test # passes. fee_multiplier = 10 # Anyone-can-spend mempool tx. # Sequence lock of 0 should pass. tx2 = CTransaction() tx2.nVersion = 2 tx2.vin = [CTxIn(COutPoint(tx1.sha256, 0), nSequence=0)] tx2.vout = [ CTxOut(int(0), CScript([b'a']))] tx2.vout[0].nValue = tx1.vout[0].nValue - \ fee_multiplier * self.nodes[0].calculate_fee(tx2) tx2_raw = self.nodes[0].signrawtransaction(ToHex(tx2))["hex"] tx2 = FromHex(tx2, tx2_raw) tx2.rehash() self.nodes[0].sendrawtransaction(tx2_raw) # Create a spend of the 0th output of orig_tx with a sequence lock # of 1, and test what happens when submitting. # orig_tx.vout[0] must be an anyone-can-spend output def test_nonzero_locks(orig_tx, node, use_height_lock): sequence_value = 1 if not use_height_lock: sequence_value |= SEQUENCE_LOCKTIME_TYPE_FLAG tx = CTransaction() tx.nVersion = 2 tx.vin = [ CTxIn(COutPoint(orig_tx.sha256, 0), nSequence=sequence_value)] tx.vout = [ CTxOut(int(orig_tx.vout[0].nValue - fee_multiplier * node.calculate_fee(tx)), CScript([b'a']))] + pad_tx(tx) tx.rehash() if (orig_tx.hash in node.getrawmempool()): # sendrawtransaction should fail if the tx is in the mempool assert_raises_rpc_error(-26, NOT_FINAL_ERROR, node.sendrawtransaction, ToHex(tx)) else: # sendrawtransaction should succeed if the tx is not in the mempool node.sendrawtransaction(ToHex(tx)) return tx test_nonzero_locks( tx2, self.nodes[0], use_height_lock=True) test_nonzero_locks( tx2, self.nodes[0], use_height_lock=False) # Now mine some blocks, but make sure tx2 doesn't get mined. # Use prioritisetransaction to lower the effective feerate to 0 self.nodes[0].prioritisetransaction( tx2.hash, -1e15, -fee_multiplier * self.nodes[0].calculate_fee(tx2)) cur_time = int(time.time()) for i in range(10): self.nodes[0].setmocktime(cur_time + 600) self.nodes[0].generate(1) cur_time += 600 assert(tx2.hash in self.nodes[0].getrawmempool()) test_nonzero_locks( tx2, self.nodes[0], use_height_lock=True) test_nonzero_locks( tx2, self.nodes[0], use_height_lock=False) # Mine tx2, and then try again self.nodes[0].prioritisetransaction( tx2.hash, 1e15, fee_multiplier * self.nodes[0].calculate_fee(tx2)) # Advance the time on the node so that we can test timelocks self.nodes[0].setmocktime(cur_time + 600) self.nodes[0].generate(1) assert(tx2.hash not in self.nodes[0].getrawmempool()) # Now that tx2 is not in the mempool, a sequence locked spend should # succeed tx3 = test_nonzero_locks( tx2, self.nodes[0], use_height_lock=False) assert(tx3.hash in self.nodes[0].getrawmempool()) self.nodes[0].generate(1) assert(tx3.hash not in self.nodes[0].getrawmempool()) # One more test, this time using height locks tx4 = test_nonzero_locks( tx3, self.nodes[0], use_height_lock=True) assert(tx4.hash in self.nodes[0].getrawmempool()) # Now try combining confirmed and unconfirmed inputs tx5 = test_nonzero_locks( tx4, self.nodes[0], use_height_lock=True) assert(tx5.hash not in self.nodes[0].getrawmempool()) utxos = self.nodes[0].listunspent() tx5.vin.append( CTxIn(COutPoint(int(utxos[0]["txid"], 16), utxos[0]["vout"]), nSequence=1)) tx5.vout[0].nValue += int(utxos[0]["amount"] * COIN) raw_tx5 = self.nodes[0].signrawtransaction(ToHex(tx5))["hex"] assert_raises_rpc_error(-26, NOT_FINAL_ERROR, self.nodes[0].sendrawtransaction, raw_tx5) # Test mempool-BIP68 consistency after reorg # # State of the transactions in the last blocks: # ... -> [ tx2 ] -> [ tx3 ] # tip-1 tip # And currently tx4 is in the mempool. # # If we invalidate the tip, tx3 should get added to the mempool, causing # tx4 to be removed (fails sequence-lock). self.nodes[0].invalidateblock(self.nodes[0].getbestblockhash()) assert(tx4.hash not in self.nodes[0].getrawmempool()) assert(tx3.hash in self.nodes[0].getrawmempool()) # Now mine 2 empty blocks to reorg out the current tip (labeled tip-1 in # diagram above). # This would cause tx2 to be added back to the mempool, which in turn causes # tx3 to be removed. tip = int(self.nodes[0].getblockhash( self.nodes[0].getblockcount() - 1), 16) height = self.nodes[0].getblockcount() for i in range(2): block = create_block(tip, create_coinbase(height), cur_time) block.nVersion = 3 block.rehash() block.solve() tip = block.sha256 height += 1 self.nodes[0].submitblock(ToHex(block)) cur_time += 1 mempool = self.nodes[0].getrawmempool() assert(tx3.hash not in mempool) assert(tx2.hash in mempool) # Reset the chain and get rid of the mocktimed-blocks self.nodes[0].setmocktime(0) self.nodes[0].invalidateblock( self.nodes[0].getblockhash(cur_height + 1)) self.nodes[0].generate(10) def get_csv_status(self): softforks = self.nodes[0].getblockchaininfo()['softforks'] for sf in softforks: if sf['id'] == 'csv' and sf['version'] == 5: return sf['reject']['status'] raise AssertionError('Cannot find CSV fork activation informations') # Make sure that BIP68 isn't being used to validate blocks, prior to # versionbits activation. If more blocks are mined prior to this test # being run, then it's possible the test has activated the soft fork, and # this test should be moved to run earlier, or deleted. def test_bip68_not_consensus(self): assert_equal(self.get_csv_status(), False) txid = self.nodes[0].sendtoaddress(self.nodes[0].getnewaddress(), 2) tx1 = FromHex(CTransaction(), self.nodes[0].getrawtransaction(txid)) tx1.rehash() # Make an anyone-can-spend transaction tx2 = CTransaction() tx2.nVersion = 1 tx2.vin = [CTxIn(COutPoint(tx1.sha256, 0), nSequence=0)] tx2.vout = [ CTxOut(int(tx1.vout[0].nValue - self.relayfee * COIN), CScript([b'a']))] # sign tx2 tx2_raw = self.nodes[0].signrawtransaction(ToHex(tx2))["hex"] tx2 = FromHex(tx2, tx2_raw) + pad_tx(tx2) tx2.rehash() self.nodes[0].sendrawtransaction(ToHex(tx2)) # Now make an invalid spend of tx2 according to BIP68 sequence_value = 100 # 100 block relative locktime tx3 = CTransaction() tx3.nVersion = 2 tx3.vin = [CTxIn(COutPoint(tx2.sha256, 0), nSequence=sequence_value)] tx3.vout = [ CTxOut(int(tx2.vout[0].nValue - self.relayfee * COIN), CScript([b'a']))] + pad_tx(tx3) tx3.rehash() assert_raises_rpc_error(-26, NOT_FINAL_ERROR, self.nodes[0].sendrawtransaction, ToHex(tx3)) # make a block that violates bip68; ensure that the tip updates tip = int(self.nodes[0].getbestblockhash(), 16) block = create_block( tip, create_coinbase(self.nodes[0].getblockcount() + 1)) block.nVersion = 3 - block.vtx.extend([tx1, tx2, tx3]) + if self.options.magnetic_anomaly_time == 0: + block.vtx.extend([tx1, tx2, tx3]) + else: + block.vtx.extend( + sorted([tx1, tx2, tx3], key=lambda tx: tx.get_id())) block.hashMerkleRoot = block.calc_merkle_root() block.rehash() block.solve() self.nodes[0].submitblock(ToHex(block)) assert_equal(self.nodes[0].getbestblockhash(), block.hash) def activateCSV(self): # activation should happen at block height 576 csv_activation_height = 576 height = self.nodes[0].getblockcount() assert_greater_than(csv_activation_height - height, 1) self.nodes[0].generate(csv_activation_height - height - 1) assert_equal(self.get_csv_status(), False) disconnect_nodes(self.nodes[0], 1) self.nodes[0].generate(1) assert_equal(self.get_csv_status(), True) # We have a block that has CSV activated, but we want to be at # the activation point, so we invalidate the tip. self.nodes[0].invalidateblock(self.nodes[0].getbestblockhash()) connect_nodes(self.nodes[0], 1) sync_blocks(self.nodes) # Use self.nodes[1] to test standardness relay policy def test_version2_relay(self, before_activation): inputs = [] outputs = {self.nodes[1].getnewaddress(): 1.0} rawtx = self.nodes[1].createrawtransaction(inputs, outputs) rawtxfund = self.nodes[1].fundrawtransaction(rawtx)['hex'] tx = FromHex(CTransaction(), rawtxfund) tx.nVersion = 2 tx_signed = self.nodes[1].signrawtransaction(ToHex(tx))["hex"] try: tx_id = self.nodes[1].sendrawtransaction(tx_signed) assert(before_activation == False) except: assert(before_activation) if __name__ == '__main__': BIP68Test().main() diff --git a/test/functional/test_runner.py b/test/functional/test_runner.py index b82e44c0a..76b8db2eb 100755 --- a/test/functional/test_runner.py +++ b/test/functional/test_runner.py @@ -1,705 +1,706 @@ #!/usr/bin/env python3 # Copyright (c) 2014-2016 The Bitcoin Core developers # Copyright (c) 2017 The Bitcoin developers # Distributed under the MIT software license, see the accompanying # file COPYING or http://www.opensource.org/licenses/mit-license.php. """Run regression test suite. This module calls down into individual test cases via subprocess. It will forward all unrecognized arguments onto the individual test scripts. Functional tests are disabled on Windows by default. Use --force to run them anyway. For a description of arguments recognized by test scripts, see `test/functional/test_framework/test_framework.py:BitcoinTestFramework.main`. """ import argparse import configparser import datetime import os import time import shutil import signal import sys import subprocess import tempfile import re import logging import xml.etree.ElementTree as ET import json import threading import multiprocessing from queue import Queue, Empty # Formatting. Default colors to empty strings. BOLD, BLUE, RED, GREY = ("", ""), ("", ""), ("", ""), ("", "") try: # Make sure python thinks it can write unicode to its stdout "\u2713".encode("utf_8").decode(sys.stdout.encoding) TICK = "✓ " CROSS = "✖ " CIRCLE = "○ " except UnicodeDecodeError: TICK = "P " CROSS = "x " CIRCLE = "o " if os.name == 'posix': # primitive formatting on supported # terminal via ANSI escape sequences: BOLD = ('\033[0m', '\033[1m') BLUE = ('\033[0m', '\033[0;34m') RED = ('\033[0m', '\033[0;31m') GREY = ('\033[0m', '\033[1;30m') TEST_EXIT_PASSED = 0 TEST_EXIT_SKIPPED = 77 NON_SCRIPTS = [ # These are python files that live in the functional tests directory, but are not test scripts. "combine_logs.py", "create_cache.py", "test_runner.py", ] TEST_PARAMS = { # Some test can be run with additional parameters. # When a test is listed here, the it will be run without parameters # as well as with additional parameters listed here. # This: # example "testName" : [["--param1", "--param2"] , ["--param3"]] # will run the test 3 times: # testName # testName --param1 --param2 # testname --param3 + "bip68-sequence.py": [["--magnetic-anomaly-time=1"]], "txn_doublespend.py": [["--mineblock"]], "txn_clone.py": [["--mineblock"]] } # Used to limit the number of tests, when list of tests is not provided on command line # When --extended is specified, we run all tests, otherwise # we only run a test if its execution time in seconds does not exceed EXTENDED_CUTOFF DEFAULT_EXTENDED_CUTOFF = 40 DEFAULT_JOBS = (multiprocessing.cpu_count() // 3) + 1 class TestCase(): """ Data structure to hold and run information necessary to launch a test case. """ def __init__(self, test_num, test_case, tests_dir, tmpdir, flags=None): self.tests_dir = tests_dir self.tmpdir = tmpdir self.test_case = test_case self.test_num = test_num self.flags = flags def run(self, portseed_offset): t = self.test_case portseed = self.test_num + portseed_offset portseed_arg = ["--portseed={}".format(portseed)] log_stdout = tempfile.SpooledTemporaryFile(max_size=2**16) log_stderr = tempfile.SpooledTemporaryFile(max_size=2**16) test_argv = t.split() tmpdir = [os.path.join("--tmpdir=%s", "%s_%s") % (self.tmpdir, re.sub(".py$", "", t), portseed)] name = t time0 = time.time() process = subprocess.Popen([os.path.join(self.tests_dir, test_argv[0])] + test_argv[1:] + self.flags + portseed_arg + tmpdir, universal_newlines=True, stdout=log_stdout, stderr=log_stderr) process.wait() log_stdout.seek(0), log_stderr.seek(0) [stdout, stderr] = [l.read().decode('utf-8') for l in (log_stdout, log_stderr)] log_stdout.close(), log_stderr.close() if process.returncode == TEST_EXIT_PASSED and stderr == "": status = "Passed" elif process.returncode == TEST_EXIT_SKIPPED: status = "Skipped" else: status = "Failed" return TestResult(name, status, int(time.time() - time0), stdout, stderr) def on_ci(): return os.getenv('TRAVIS') == 'true' or os.getenv('TEAMCITY_VERSION') != None def main(): # Read config generated by configure. config = configparser.ConfigParser() configfile = os.path.join(os.path.abspath( os.path.dirname(__file__)), "..", "config.ini") config.read_file(open(configfile)) src_dir = config["environment"]["SRCDIR"] build_dir = config["environment"]["BUILDDIR"] tests_dir = os.path.join(src_dir, 'test', 'functional') # Parse arguments and pass through unrecognised args parser = argparse.ArgumentParser(add_help=False, usage='%(prog)s [test_runner.py options] [script options] [scripts]', description=__doc__, epilog=''' Help text and arguments for individual test script:''', formatter_class=argparse.RawTextHelpFormatter) parser.add_argument('--coverage', action='store_true', help='generate a basic coverage report for the RPC interface') parser.add_argument( '--exclude', '-x', help='specify a comma-seperated-list of scripts to exclude. Do not include the .py extension in the name.') parser.add_argument('--extended', action='store_true', help='run the extended test suite in addition to the basic tests') parser.add_argument('--cutoff', type=int, default=DEFAULT_EXTENDED_CUTOFF, help='set the cutoff runtime for what tests get run') parser.add_argument('--force', '-f', action='store_true', help='run tests even on platforms where they are disabled by default (e.g. windows).') parser.add_argument('--help', '-h', '-?', action='store_true', help='print help text and exit') parser.add_argument('--jobs', '-j', type=int, default=DEFAULT_JOBS, help='how many test scripts to run in parallel. Default=4.') parser.add_argument('--keepcache', '-k', action='store_true', help='the default behavior is to flush the cache directory on startup. --keepcache retains the cache from the previous testrun.') parser.add_argument('--quiet', '-q', action='store_true', help='only print results summary and failure logs') parser.add_argument('--tmpdirprefix', '-t', default=tempfile.gettempdir(), help="Root directory for datadirs") parser.add_argument('--junitouput', '-ju', default=os.path.join(build_dir, 'junit_results.xml'), help="file that will store JUnit formated test results.") args, unknown_args = parser.parse_known_args() # Create a set to store arguments and create the passon string tests = set(arg for arg in unknown_args if arg[:2] != "--") passon_args = [arg for arg in unknown_args if arg[:2] == "--"] passon_args.append("--configfile=%s" % configfile) # Set up logging logging_level = logging.INFO if args.quiet else logging.DEBUG logging.basicConfig(format='%(message)s', level=logging_level) # Create base test directory tmpdir = os.path.join("%s", "bitcoin_test_runner_%s") % ( args.tmpdirprefix, datetime.datetime.now().strftime("%Y%m%d_%H%M%S")) os.makedirs(tmpdir) logging.debug("Temporary test directory at %s" % tmpdir) enable_wallet = config["components"].getboolean("ENABLE_WALLET") enable_utils = config["components"].getboolean("ENABLE_UTILS") enable_bitcoind = config["components"].getboolean("ENABLE_BITCOIND") if config["environment"]["EXEEXT"] == ".exe" and not args.force: # https://github.com/bitcoin/bitcoin/commit/d52802551752140cf41f0d9a225a43e84404d3e9 # https://github.com/bitcoin/bitcoin/pull/5677#issuecomment-136646964 print( "Tests currently disabled on Windows by default. Use --force option to enable") sys.exit(0) if not (enable_wallet and enable_utils and enable_bitcoind): print( "No functional tests to run. Wallet, utils, and bitcoind must all be enabled") print( "Rerun `configure` with -enable-wallet, -with-utils and -with-daemon and rerun make") sys.exit(0) # Build list of tests all_scripts = get_all_scripts_from_disk(tests_dir, NON_SCRIPTS) if tests: # Individual tests have been specified. Run specified tests that exist # in the all_scripts list. Accept the name with or without .py # extension. test_list = [t for t in all_scripts if (t in tests or re.sub(".py$", "", t) in tests)] cutoff = sys.maxsize # do not cut off explicitly specified tests else: # No individual tests have been specified. # Run all tests that do not exceed test_list = all_scripts cutoff = args.cutoff if args.extended: cutoff = sys.maxsize # Remove the test cases that the user has explicitly asked to exclude. if args.exclude: for exclude_test in args.exclude.split(','): if exclude_test + ".py" in test_list: test_list.remove(exclude_test + ".py") # Use and update timings from build_dir only if separate # build directory is used. We do not want to pollute source directory. build_timings = None if (src_dir != build_dir): build_timings = Timings(os.path.join(build_dir, 'timing.json')) # Always use timings from scr_dir if present src_timings = Timings(os.path.join( src_dir, "test", "functional", 'timing.json')) # Add test parameters and remove long running tests if needed test_list = get_tests_to_run( test_list, TEST_PARAMS, cutoff, src_timings, build_timings) if not test_list: print("No valid test scripts specified. Check that your test is in one " "of the test lists in test_runner.py, or run test_runner.py with no arguments to run all tests") sys.exit(0) if args.help: # Print help for test_runner.py, then print help of the first script # and exit. parser.print_help() subprocess.check_call( [os.path.join(tests_dir, test_list[0]), '-h']) sys.exit(0) if not args.keepcache: shutil.rmtree(os.path.join(build_dir, "test", "cache"), ignore_errors=True) run_tests(test_list, build_dir, tests_dir, args.junitouput, config["environment"]["EXEEXT"], tmpdir, args.jobs, args.coverage, passon_args, build_timings) def run_tests(test_list, build_dir, tests_dir, junitouput, exeext, tmpdir, num_jobs, enable_coverage=False, args=[], build_timings=None): # Warn if bitcoind is already running (unix only) try: pidofOutput = subprocess.check_output(["pidof", "bitcoind"]) if pidofOutput is not None and pidofOutput != b'': print("%sWARNING!%s There is already a bitcoind process running on this system. Tests may fail unexpectedly due to resource contention!" % ( BOLD[1], BOLD[0])) except (OSError, subprocess.SubprocessError): pass # Warn if there is a cache directory cache_dir = os.path.join(build_dir, "test", "cache") if os.path.isdir(cache_dir): print("%sWARNING!%s There is a cache directory here: %s. If tests fail unexpectedly, try deleting the cache directory." % ( BOLD[1], BOLD[0], cache_dir)) # Set env vars if "BITCOIND" not in os.environ: os.environ["BITCOIND"] = os.path.join( build_dir, 'src', 'bitcoind' + exeext) os.environ["BITCOINCLI"] = os.path.join( build_dir, 'src', 'bitcoin-cli' + exeext) flags = [os.path.join("--srcdir={}".format(build_dir), "src")] + args flags.append("--cachedir=%s" % cache_dir) if enable_coverage: coverage = RPCCoverage() flags.append(coverage.flag) logging.debug("Initializing coverage directory at %s" % coverage.dir) else: coverage = None if len(test_list) > 1 and num_jobs > 1: # Populate cache subprocess.check_output( [os.path.join(tests_dir, 'create_cache.py')] + flags + [os.path.join("--tmpdir=%s", "cache") % tmpdir]) # Run Tests time0 = time.time() test_results = execute_test_processes( num_jobs, test_list, tests_dir, tmpdir, flags) runtime = int(time.time() - time0) max_len_name = len(max(test_list, key=len)) print_results(test_results, max_len_name, runtime) save_results_as_junit(test_results, junitouput, runtime) if (build_timings is not None): build_timings.save_timings(test_results) if coverage: coverage.report_rpc_coverage() logging.debug("Cleaning up coverage data") coverage.cleanup() # Clear up the temp directory if all subdirectories are gone if not os.listdir(tmpdir): os.rmdir(tmpdir) all_passed = all( map(lambda test_result: test_result.status == "Passed", test_results)) sys.exit(not all_passed) def execute_test_processes(num_jobs, test_list, tests_dir, tmpdir, flags): update_queue = Queue() job_queue = Queue() test_results = [] poll_timeout = 10 # seconds # In case there is a graveyard of zombie bitcoinds, we can apply a # pseudorandom offset to hopefully jump over them. # (625 is PORT_RANGE/MAX_NODES) portseed_offset = int(time.time() * 1000) % 625 ## # Define some helper functions we will need for threading. ## def handle_message(message, running_jobs): """ handle_message handles a single message from handle_test_cases """ if isinstance(message, TestCase): running_jobs.add(message.test_case) print("{}{}{} started".format(BOLD[1], message.test_case, BOLD[0])) return if isinstance(message, TestResult): test_result = message running_jobs.remove(test_result.name) test_results.append(test_result) if test_result.status == "Passed": print("%s%s%s passed, Duration: %s s" % ( BOLD[1], test_result.name, BOLD[0], test_result.time)) elif test_result.status == "Skipped": print("%s%s%s skipped" % (BOLD[1], test_result.name, BOLD[0])) else: print("%s%s%s failed, Duration: %s s\n" % (BOLD[1], test_result.name, BOLD[0], test_result.time)) print(BOLD[1] + 'stdout:' + BOLD[0]) print(test_result.stdout) print(BOLD[1] + 'stderr:' + BOLD[0]) print(test_result.stderr) return assert False, "we should not be here" def handle_update_messages(): """ handle_update_messages waits for messages to be sent from handle_test_cases via the update_queue. It serializes the results so we can print nice status update messages. """ printed_status = False running_jobs = set() while True: message = None try: message = update_queue.get(True, poll_timeout) if message is None: break # We printed a status message, need to kick to the next line # before printing more. if printed_status: print() printed_status = False handle_message(message, running_jobs) update_queue.task_done() except Empty as e: if not on_ci(): print("Running jobs: {}".format(", ".join(running_jobs)), end="\r") sys.stdout.flush() printed_status = True def handle_test_cases(): """ job_runner represents a single thread that is part of a worker pool. It waits for a test, then executes that test. It also reports start and result messages to handle_update_messages """ while True: test = job_queue.get() if test is None: break # Signal that the test is starting to inform the poor waiting # programmer update_queue.put(test) result = test.run(portseed_offset) update_queue.put(result) job_queue.task_done() ## # Setup our threads, and start sending tasks ## # Start our result collection thread. t = threading.Thread(target=handle_update_messages) t.setDaemon(True) t.start() # Start some worker threads for j in range(num_jobs): t = threading.Thread(target=handle_test_cases) t.setDaemon(True) t.start() # Push all our test cases into the job queue. for i, t in enumerate(test_list): job_queue.put(TestCase(i, t, tests_dir, tmpdir, flags)) # Wait for all the jobs to be completed job_queue.join() # Wait for all the results to be compiled update_queue.join() # Flush our queues so the threads exit update_queue.put(None) for j in range(num_jobs): job_queue.put(None) return test_results def print_results(test_results, max_len_name, runtime): results = "\n" + BOLD[1] + "%s | %s | %s\n\n" % ( "TEST".ljust(max_len_name), "STATUS ", "DURATION") + BOLD[0] test_results.sort(key=lambda result: result.name.lower()) all_passed = True time_sum = 0 for test_result in test_results: all_passed = all_passed and test_result.status != "Failed" time_sum += test_result.time test_result.padding = max_len_name results += str(test_result) status = TICK + "Passed" if all_passed else CROSS + "Failed" results += BOLD[1] + "\n%s | %s | %s s (accumulated) \n" % ( "ALL".ljust(max_len_name), status.ljust(9), time_sum) + BOLD[0] results += "Runtime: %s s\n" % (runtime) print(results) class TestResult(): """ Simple data structure to store test result values and print them properly """ def __init__(self, name, status, time, stdout, stderr): self.name = name self.status = status self.time = time self.padding = 0 self.stdout = stdout self.stderr = stderr def __repr__(self): if self.status == "Passed": color = BLUE glyph = TICK elif self.status == "Failed": color = RED glyph = CROSS elif self.status == "Skipped": color = GREY glyph = CIRCLE return color[1] + "%s | %s%s | %s s\n" % (self.name.ljust(self.padding), glyph, self.status.ljust(7), self.time) + color[0] def get_all_scripts_from_disk(test_dir, non_scripts): """ Return all available test script from script directory (excluding NON_SCRIPTS) """ python_files = set([t for t in os.listdir(test_dir) if t[-3:] == ".py"]) return list(python_files - set(non_scripts)) def get_tests_to_run(test_list, test_params, cutoff, src_timings, build_timings=None): """ Returns only test that will not run longer that cutoff. Long running tests are returned first to favor running tests in parallel Timings from build directory override those from src directory """ def get_test_time(test): if build_timings is not None: timing = next( (x['time'] for x in build_timings.existing_timings if x['name'] == test), None) if timing is not None: return timing # try source directory. Return 0 if test is unknown to always run it return next( (x['time'] for x in src_timings.existing_timings if x['name'] == test), 0) # Some tests must also be run with additional parameters. Add them to the list. tests_with_params = [] for test_name in test_list: # always execute a test without parameters tests_with_params.append(test_name) params = test_params.get(test_name) if params is not None: tests_with_params.extend( [test_name + " " + " ".join(p) for p in params]) result = [t for t in tests_with_params if get_test_time(t) <= cutoff] result.sort(key=lambda x: (-get_test_time(x), x)) return result class RPCCoverage(): """ Coverage reporting utilities for test_runner. Coverage calculation works by having each test script subprocess write coverage files into a particular directory. These files contain the RPC commands invoked during testing, as well as a complete listing of RPC commands per `bitcoin-cli help` (`rpc_interface.txt`). After all tests complete, the commands run are combined and diff'd against the complete list to calculate uncovered RPC commands. See also: test/functional/test_framework/coverage.py """ def __init__(self): self.dir = tempfile.mkdtemp(prefix="coverage") self.flag = '--coveragedir={}'.format(self.dir) def report_rpc_coverage(self): """ Print out RPC commands that were unexercised by tests. """ uncovered = self._get_uncovered_rpc_commands() if uncovered: print("Uncovered RPC commands:") print("".join((" - {}\n".format(i)) for i in sorted(uncovered))) else: print("All RPC commands covered.") def cleanup(self): return shutil.rmtree(self.dir) def _get_uncovered_rpc_commands(self): """ Return a set of currently untested RPC commands. """ # This is shared from `test/functional/test-framework/coverage.py` reference_filename = 'rpc_interface.txt' coverage_file_prefix = 'coverage.' coverage_ref_filename = os.path.join(self.dir, reference_filename) coverage_filenames = set() all_cmds = set() covered_cmds = set() if not os.path.isfile(coverage_ref_filename): raise RuntimeError("No coverage reference found") with open(coverage_ref_filename, 'r') as f: all_cmds.update([i.strip() for i in f.readlines()]) for root, dirs, files in os.walk(self.dir): for filename in files: if filename.startswith(coverage_file_prefix): coverage_filenames.add(os.path.join(root, filename)) for filename in coverage_filenames: with open(filename, 'r') as f: covered_cmds.update([i.strip() for i in f.readlines()]) return all_cmds - covered_cmds def save_results_as_junit(test_results, file_name, time): """ Save tests results to file in JUnit format See http://llg.cubic.org/docs/junit/ for specification of format """ e_test_suite = ET.Element("testsuite", {"name": "bitcoin_abc_tests", "tests": str(len(test_results)), # "errors": "failures": str(len([t for t in test_results if t.status == "Failed"])), "id": "0", "skipped": str(len([t for t in test_results if t.status == "Skipped"])), "time": str(time), "timestamp": datetime.datetime.now().isoformat('T') }) for test_result in test_results: e_test_case = ET.SubElement(e_test_suite, "testcase", {"name": test_result.name, "classname": test_result.name, "time": str(test_result.time) } ) if test_result.status == "Skipped": ET.SubElement(e_test_case, "skipped") elif test_result.status == "Failed": ET.SubElement(e_test_case, "failure") # no special element for passed tests ET.SubElement(e_test_case, "system-out").text = test_result.stdout ET.SubElement(e_test_case, "system-err").text = test_result.stderr ET.ElementTree(e_test_suite).write( file_name, "UTF-8", xml_declaration=True) class Timings(): """ Takes care of loading, merging and saving tests execution times. """ def __init__(self, timing_file): self.timing_file = timing_file self.existing_timings = self.load_timings() def load_timings(self): if os.path.isfile(self.timing_file): with open(self.timing_file) as f: return json.load(f) else: return [] def get_merged_timings(self, new_timings): """ Return new list containing existing timings updated with new timings Tests that do not exists are not removed """ key = 'name' merged = {} for item in self.existing_timings + new_timings: if item[key] in merged: merged[item[key]].update(item) else: merged[item[key]] = item # Sort the result to preserve test ordering in file merged = list(merged.values()) merged.sort(key=lambda t, key=key: t[key]) return merged def save_timings(self, test_results): # we only save test that have passed - timings for failed test might be # wrong (timeouts or early fails) passed_results = [t for t in test_results if t.status == 'Passed'] new_timings = list(map(lambda t: {'name': t.name, 'time': t.time}, passed_results)) merged_timings = self.get_merged_timings(new_timings) with open(self.timing_file, 'w') as f: json.dump(merged_timings, f, indent=True) if __name__ == '__main__': main()