You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
574 lines
23 KiB
574 lines
23 KiB
#!/usr/bin/env python3 |
|
# Copyright (c) 2014-2017 The Bitcoin Core 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 |
|
from collections import deque |
|
import configparser |
|
import datetime |
|
import os |
|
import time |
|
import shutil |
|
import signal |
|
import sys |
|
import subprocess |
|
import tempfile |
|
import re |
|
import logging |
|
|
|
# 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 |
|
|
|
BASE_SCRIPTS= [ |
|
# Scripts that are run by the travis build process. |
|
# Longest test should go first, to favor running tests in parallel |
|
'wallet_hd.py', |
|
'wallet_backup.py', |
|
# vv Tests less than 5m vv |
|
'feature_block.py', |
|
'rpc_fundrawtransaction.py', |
|
'p2p_compactblocks.py', |
|
'feature_segwit.py', |
|
# vv Tests less than 2m vv |
|
'wallet_basic.py', |
|
'wallet_accounts.py', |
|
'p2p_segwit.py', |
|
'wallet_dump.py', |
|
'rpc_listtransactions.py', |
|
# vv Tests less than 60s vv |
|
'p2p_sendheaders.py', |
|
'wallet_zapwallettxes.py', |
|
'wallet_importmulti.py', |
|
'mempool_limit.py', |
|
'rpc_txoutproof.py', |
|
'wallet_listreceivedby.py', |
|
'wallet_abandonconflict.py', |
|
'feature_csv_activation.py', |
|
'rpc_rawtransaction.py', |
|
'wallet_address_types.py', |
|
'feature_reindex.py', |
|
# vv Tests less than 30s vv |
|
'wallet_keypool_topup.py', |
|
'interface_zmq.py', |
|
'interface_bitcoin_cli.py', |
|
'mempool_resurrect.py', |
|
'wallet_txn_doublespend.py --mineblock', |
|
'wallet_txn_clone.py', |
|
'wallet_txn_clone.py --segwit', |
|
'rpc_getchaintips.py', |
|
'interface_rest.py', |
|
'mempool_spend_coinbase.py', |
|
'mempool_reorg.py', |
|
'mempool_persist.py', |
|
'wallet_multiwallet.py', |
|
'wallet_multiwallet.py --usecli', |
|
'interface_http.py', |
|
'rpc_users.py', |
|
'feature_proxy.py', |
|
'rpc_signrawtransaction.py', |
|
'p2p_disconnect_ban.py', |
|
'rpc_decodescript.py', |
|
'rpc_blockchain.py', |
|
'rpc_deprecated.py', |
|
'wallet_disable.py', |
|
'rpc_net.py', |
|
'wallet_keypool.py', |
|
'p2p_mempool.py', |
|
'mining_prioritisetransaction.py', |
|
'p2p_invalid_block.py', |
|
'p2p_invalid_tx.py', |
|
'feature_versionbits_warning.py', |
|
'rpc_preciousblock.py', |
|
'wallet_importprunedfunds.py', |
|
'rpc_signmessage.py', |
|
'feature_nulldummy.py', |
|
'wallet_import_rescan.py', |
|
'mining_basic.py', |
|
'wallet_bumpfee.py', |
|
'rpc_named_arguments.py', |
|
'wallet_listsinceblock.py', |
|
'p2p_leak.py', |
|
'wallet_encryption.py', |
|
'feature_dersig.py', |
|
'feature_cltv.py', |
|
'rpc_uptime.py', |
|
'wallet_resendwallettransactions.py', |
|
'feature_minchainwork.py', |
|
'p2p_fingerprint.py', |
|
'feature_uacomment.py', |
|
'p2p_unrequested_blocks.py', |
|
'feature_logging.py', |
|
'p2p_node_network_limited.py', |
|
'feature_config_args.py', |
|
# Don't append tests at the end to avoid merge conflicts |
|
# Put them in a random line within the section that fits their approximate run-time |
|
] |
|
|
|
EXTENDED_SCRIPTS = [ |
|
# These tests are not run by the travis build process. |
|
# Longest test should go first, to favor running tests in parallel |
|
'feature_pruning.py', |
|
# vv Tests less than 20m vv |
|
'feature_fee_estimation.py', |
|
# vv Tests less than 5m vv |
|
'feature_maxuploadtarget.py', |
|
'mempool_packages.py', |
|
'feature_dbcrash.py', |
|
# vv Tests less than 2m vv |
|
'feature_bip68_sequence.py', |
|
'mining_getblocktemplate_longpoll.py', |
|
'p2p_timeouts.py', |
|
# vv Tests less than 60s vv |
|
'feature_bip9_softforks.py', |
|
'p2p_feefilter.py', |
|
'rpc_bind.py', |
|
# vv Tests less than 30s vv |
|
'feature_assumevalid.py', |
|
'example_test.py', |
|
'wallet_txn_doublespend.py', |
|
'wallet_txn_clone.py --mineblock', |
|
'feature_notifications.py', |
|
'rpc_invalidateblock.py', |
|
'feature_rbf.py', |
|
] |
|
|
|
# Place EXTENDED_SCRIPTS first since it has the 3 longest running tests |
|
ALL_SCRIPTS = EXTENDED_SCRIPTS + BASE_SCRIPTS |
|
|
|
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", |
|
] |
|
|
|
def main(): |
|
# 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('--combinedlogslen', '-c', type=int, default=0, help='print a combined log (of length n lines) from all test nodes and test framework to the console on failure.') |
|
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-separated-list of scripts to exclude.') |
|
parser.add_argument('--extended', action='store_true', help='run the extended test suite in addition to the basic tests') |
|
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=4, 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") |
|
args, unknown_args = parser.parse_known_args() |
|
|
|
# args to be passed on always start with two dashes; tests are the remaining unknown args |
|
tests = [arg for arg in unknown_args if arg[:2] != "--"] |
|
passon_args = [arg for arg in unknown_args if arg[:2] == "--"] |
|
|
|
# Read config generated by configure. |
|
config = configparser.ConfigParser() |
|
configfile = os.path.abspath(os.path.dirname(__file__)) + "/../config.ini" |
|
config.read_file(open(configfile)) |
|
|
|
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 = "%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 |
|
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. |
|
tests = [re.sub("\.py$", "", t) + ".py" for t in tests] |
|
test_list = [] |
|
for t in tests: |
|
if t in ALL_SCRIPTS: |
|
test_list.append(t) |
|
else: |
|
print("{}WARNING!{} Test '{}' not found in full test list.".format(BOLD[1], BOLD[0], t)) |
|
else: |
|
# No individual tests have been specified. |
|
# Run all base tests, and optionally run extended tests. |
|
test_list = BASE_SCRIPTS |
|
if args.extended: |
|
# place the EXTENDED_SCRIPTS first since the three longest ones |
|
# are there and the list is shorter |
|
test_list = EXTENDED_SCRIPTS + test_list |
|
|
|
# Remove the test cases that the user has explicitly asked to exclude. |
|
if args.exclude: |
|
tests_excl = [re.sub("\.py$", "", t) + ".py" for t in args.exclude.split(',')] |
|
for exclude_test in tests_excl: |
|
if exclude_test in test_list: |
|
test_list.remove(exclude_test) |
|
else: |
|
print("{}WARNING!{} Test '{}' not found in current test list.".format(BOLD[1], BOLD[0], exclude_test)) |
|
|
|
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 (with args removed) and exit. |
|
parser.print_help() |
|
subprocess.check_call([(config["environment"]["SRCDIR"] + '/test/functional/' + test_list[0].split()[0])] + ['-h']) |
|
sys.exit(0) |
|
|
|
check_script_list(config["environment"]["SRCDIR"]) |
|
check_script_prefixes() |
|
|
|
if not args.keepcache: |
|
shutil.rmtree("%s/test/cache" % config["environment"]["BUILDDIR"], ignore_errors=True) |
|
|
|
run_tests(test_list, config["environment"]["SRCDIR"], config["environment"]["BUILDDIR"], config["environment"]["EXEEXT"], tmpdir, args.jobs, args.coverage, passon_args, args.combinedlogslen) |
|
|
|
def run_tests(test_list, src_dir, build_dir, exeext, tmpdir, jobs=1, enable_coverage=False, args=[], combined_logs_len=0): |
|
# Warn if bitcoind is already running (unix only) |
|
try: |
|
if subprocess.check_output(["pidof", "bitcoind"]) is not None: |
|
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 = "%s/test/cache" % build_dir |
|
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"] = build_dir + '/src/bitcoind' + exeext |
|
os.environ["BITCOINCLI"] = build_dir + '/src/bitcoin-cli' + exeext |
|
|
|
tests_dir = src_dir + '/test/functional/' |
|
|
|
flags = ["--srcdir={}/src".format(build_dir)] + 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 jobs > 1: |
|
# Populate cache |
|
try: |
|
subprocess.check_output([tests_dir + 'create_cache.py'] + flags + ["--tmpdir=%s/cache" % tmpdir]) |
|
except subprocess.CalledProcessError as e: |
|
sys.stdout.buffer.write(e.output) |
|
raise |
|
|
|
#Run Tests |
|
job_queue = TestHandler(jobs, tests_dir, tmpdir, test_list, flags) |
|
time0 = time.time() |
|
test_results = [] |
|
|
|
max_len_name = len(max(test_list, key=len)) |
|
|
|
for _ in range(len(test_list)): |
|
test_result, testdir, stdout, stderr = job_queue.get_next() |
|
test_results.append(test_result) |
|
|
|
if test_result.status == "Passed": |
|
logging.debug("\n%s%s%s passed, Duration: %s s" % (BOLD[1], test_result.name, BOLD[0], test_result.time)) |
|
elif test_result.status == "Skipped": |
|
logging.debug("\n%s%s%s skipped" % (BOLD[1], test_result.name, BOLD[0])) |
|
else: |
|
print("\n%s%s%s failed, Duration: %s s\n" % (BOLD[1], test_result.name, BOLD[0], test_result.time)) |
|
print(BOLD[1] + 'stdout:\n' + BOLD[0] + stdout + '\n') |
|
print(BOLD[1] + 'stderr:\n' + BOLD[0] + stderr + '\n') |
|
if combined_logs_len and os.path.isdir(testdir): |
|
# Print the final `combinedlogslen` lines of the combined logs |
|
print('{}Combine the logs and print the last {} lines ...{}'.format(BOLD[1], combined_logs_len, BOLD[0])) |
|
print('\n============') |
|
print('{}Combined log for {}:{}'.format(BOLD[1], testdir, BOLD[0])) |
|
print('============\n') |
|
combined_logs, _ = subprocess.Popen([os.path.join(tests_dir, 'combine_logs.py'), '-c', testdir], universal_newlines=True, stdout=subprocess.PIPE).communicate() |
|
print("\n".join(deque(combined_logs.splitlines(), combined_logs_len))) |
|
|
|
print_results(test_results, max_len_name, (int(time.time() - time0))) |
|
|
|
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.was_successful, test_results)) |
|
|
|
sys.exit(not all_passed) |
|
|
|
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.was_successful |
|
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 TestHandler: |
|
""" |
|
Trigger the test scripts passed in via the list. |
|
""" |
|
|
|
def __init__(self, num_tests_parallel, tests_dir, tmpdir, test_list=None, flags=None): |
|
assert(num_tests_parallel >= 1) |
|
self.num_jobs = num_tests_parallel |
|
self.tests_dir = tests_dir |
|
self.tmpdir = tmpdir |
|
self.test_list = test_list |
|
self.flags = flags |
|
self.num_running = 0 |
|
# 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) |
|
self.portseed_offset = int(time.time() * 1000) % 625 |
|
self.jobs = [] |
|
|
|
def get_next(self): |
|
while self.num_running < self.num_jobs and self.test_list: |
|
# Add tests |
|
self.num_running += 1 |
|
t = self.test_list.pop(0) |
|
portseed = len(self.test_list) + self.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() |
|
testdir = "{}/{}_{}".format(self.tmpdir, re.sub(".py$", "", test_argv[0]), portseed) |
|
tmpdir_arg = ["--tmpdir={}".format(testdir)] |
|
self.jobs.append((t, |
|
time.time(), |
|
subprocess.Popen([self.tests_dir + test_argv[0]] + test_argv[1:] + self.flags + portseed_arg + tmpdir_arg, |
|
universal_newlines=True, |
|
stdout=log_stdout, |
|
stderr=log_stderr), |
|
testdir, |
|
log_stdout, |
|
log_stderr)) |
|
if not self.jobs: |
|
raise IndexError('pop from empty list') |
|
while True: |
|
# Return first proc that finishes |
|
time.sleep(.5) |
|
for j in self.jobs: |
|
(name, time0, proc, testdir, log_out, log_err) = j |
|
if os.getenv('TRAVIS') == 'true' and int(time.time() - time0) > 20 * 60: |
|
# In travis, timeout individual tests after 20 minutes (to stop tests hanging and not |
|
# providing useful output. |
|
proc.send_signal(signal.SIGINT) |
|
if proc.poll() is not None: |
|
log_out.seek(0), log_err.seek(0) |
|
[stdout, stderr] = [l.read().decode('utf-8') for l in (log_out, log_err)] |
|
log_out.close(), log_err.close() |
|
if proc.returncode == TEST_EXIT_PASSED and stderr == "": |
|
status = "Passed" |
|
elif proc.returncode == TEST_EXIT_SKIPPED: |
|
status = "Skipped" |
|
else: |
|
status = "Failed" |
|
self.num_running -= 1 |
|
self.jobs.remove(j) |
|
|
|
return TestResult(name, status, int(time.time() - time0)), testdir, stdout, stderr |
|
print('.', end='', flush=True) |
|
|
|
class TestResult(): |
|
def __init__(self, name, status, time): |
|
self.name = name |
|
self.status = status |
|
self.time = time |
|
self.padding = 0 |
|
|
|
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] |
|
|
|
@property |
|
def was_successful(self): |
|
return self.status != "Failed" |
|
|
|
|
|
def check_script_prefixes(): |
|
"""Check that at most a handful of the |
|
test scripts don't start with one of the allowed name prefixes.""" |
|
|
|
# LEEWAY is provided as a transition measure, so that pull-requests |
|
# that introduce new tests that don't conform with the naming |
|
# convention don't immediately cause the tests to fail. |
|
LEEWAY = 10 |
|
|
|
good_prefixes_re = re.compile("(example|feature|interface|mempool|mining|p2p|rpc|wallet)_") |
|
bad_script_names = [script for script in ALL_SCRIPTS if good_prefixes_re.match(script) is None] |
|
|
|
if len(bad_script_names) > 0: |
|
print("INFO: %d tests not meeting naming conventions:" % (len(bad_script_names))) |
|
print(" %s" % ("\n ".join(sorted(bad_script_names)))) |
|
assert len(bad_script_names) <= LEEWAY, "Too many tests not following naming convention! (%d found, maximum: %d)" % (len(bad_script_names), LEEWAY) |
|
|
|
|
|
def check_script_list(src_dir): |
|
"""Check scripts directory. |
|
|
|
Check that there are no scripts in the functional tests directory which are |
|
not being run by pull-tester.py.""" |
|
script_dir = src_dir + '/test/functional/' |
|
python_files = set([t for t in os.listdir(script_dir) if t[-3:] == ".py"]) |
|
missed_tests = list(python_files - set(map(lambda x: x.split()[0], ALL_SCRIPTS + NON_SCRIPTS))) |
|
if len(missed_tests) != 0: |
|
print("%sWARNING!%s The following scripts are not being run: %s. Check the test lists in test_runner.py." % (BOLD[1], BOLD[0], str(missed_tests))) |
|
if os.getenv('TRAVIS') == 'true': |
|
# On travis this warning is an error to prevent merging incomplete commits into master |
|
sys.exit(1) |
|
|
|
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=%s' % 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((" - %s\n" % 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 |
|
|
|
|
|
if __name__ == '__main__': |
|
main()
|
|
|