Issue #23839: Various caches now are cleared before running every test file.

8 files changed, 2049 insertions, 0 deletions

diff --git a/Lib/test/libregrtest/__init__.py b/Lib/test/libregrtest/__init__.py
new file mode 100644
index 0000000000..7ba0e6e535
--- /dev/null
+++ b/Lib/test/libregrtest/__init__.py
@@ -0,0 +1,5 @@
+# We import importlib *ASAP* in order to test #15386
+import importlib
+
+from test.libregrtest.cmdline import _parse_args, RESOURCE_NAMES
+from test.libregrtest.main import main
diff --git a/Lib/test/libregrtest/cmdline.py b/Lib/test/libregrtest/cmdline.py
new file mode 100644
index 0000000000..891b00c753
--- /dev/null
+++ b/Lib/test/libregrtest/cmdline.py
@@ -0,0 +1,347 @@
+import argparse
+import os
+import sys
+from test import support
+
+
+USAGE = """\
+python -m test [options] [test_name1 [test_name2 ...]]
+python path/to/Lib/test/regrtest.py [options] [test_name1 [test_name2 ...]]
+"""
+
+DESCRIPTION = """\
+Run Python regression tests.
+
+If no arguments or options are provided, finds all files matching
+the pattern "test_*" in the Lib/test subdirectory and runs
+them in alphabetical order (but see -M and -u, below, for exceptions).
+
+For more rigorous testing, it is useful to use the following
+command line:
+
+python -E -Wd -m test [options] [test_name1 ...]
+"""
+
+EPILOG = """\
+Additional option details:
+
+-r randomizes test execution order. You can use --randseed=int to provide an
+int seed value for the randomizer; this is useful for reproducing troublesome
+test orders.
+
+-s On the first invocation of regrtest using -s, the first test file found
+or the first test file given on the command line is run, and the name of
+the next test is recorded in a file named pynexttest.  If run from the
+Python build directory, pynexttest is located in the 'build' subdirectory,
+otherwise it is located in tempfile.gettempdir().  On subsequent runs,
+the test in pynexttest is run, and the next test is written to pynexttest.
+When the last test has been run, pynexttest is deleted.  In this way it
+is possible to single step through the test files.  This is useful when
+doing memory analysis on the Python interpreter, which process tends to
+consume too many resources to run the full regression test non-stop.
+
+-S is used to continue running tests after an aborted run.  It will
+maintain the order a standard run (ie, this assumes -r is not used).
+This is useful after the tests have prematurely stopped for some external
+reason and you want to start running from where you left off rather
+than starting from the beginning.
+
+-f reads the names of tests from the file given as f's argument, one
+or more test names per line.  Whitespace is ignored.  Blank lines and
+lines beginning with '#' are ignored.  This is especially useful for
+whittling down failures involving interactions among tests.
+
+-L causes the leaks(1) command to be run just before exit if it exists.
+leaks(1) is available on Mac OS X and presumably on some other
+FreeBSD-derived systems.
+
+-R runs each test several times and examines sys.gettotalrefcount() to
+see if the test appears to be leaking references.  The argument should
+be of the form stab:run:fname where 'stab' is the number of times the
+test is run to let gettotalrefcount settle down, 'run' is the number
+of times further it is run and 'fname' is the name of the file the
+reports are written to.  These parameters all have defaults (5, 4 and
+"reflog.txt" respectively), and the minimal invocation is '-R :'.
+
+-M runs tests that require an exorbitant amount of memory. These tests
+typically try to ascertain containers keep working when containing more than
+2 billion objects, which only works on 64-bit systems. There are also some
+tests that try to exhaust the address space of the process, which only makes
+sense on 32-bit systems with at least 2Gb of memory. The passed-in memlimit,
+which is a string in the form of '2.5Gb', determines howmuch memory the
+tests will limit themselves to (but they may go slightly over.) The number
+shouldn't be more memory than the machine has (including swap memory). You
+should also keep in mind that swap memory is generally much, much slower
+than RAM, and setting memlimit to all available RAM or higher will heavily
+tax the machine. On the other hand, it is no use running these tests with a
+limit of less than 2.5Gb, and many require more than 20Gb. Tests that expect
+to use more than memlimit memory will be skipped. The big-memory tests
+generally run very, very long.
+
+-u is used to specify which special resource intensive tests to run,
+such as those requiring large file support or network connectivity.
+The argument is a comma-separated list of words indicating the
+resources to test.  Currently only the following are defined:
+
+    all -       Enable all special resources.
+
+    none -      Disable all special resources (this is the default).
+
+    audio -     Tests that use the audio device.  (There are known
+                cases of broken audio drivers that can crash Python or
+                even the Linux kernel.)
+
+    curses -    Tests that use curses and will modify the terminal's
+                state and output modes.
+
+    largefile - It is okay to run some test that may create huge
+                files.  These tests can take a long time and may
+                consume >2GB of disk space temporarily.
+
+    network -   It is okay to run tests that use external network
+                resource, e.g. testing SSL support for sockets.
+
+    decimal -   Test the decimal module against a large suite that
+                verifies compliance with standards.
+
+    cpu -       Used for certain CPU-heavy tests.
+
+    subprocess  Run all tests for the subprocess module.
+
+    urlfetch -  It is okay to download files required on testing.
+
+    gui -       Run tests that require a running GUI.
+
+    tzdata -    Run tests that require timezone data.
+
+To enable all resources except one, use '-uall,-<resource>'.  For
+example, to run all the tests except for the gui tests, give the
+option '-uall,-gui'.
+"""
+
+
+RESOURCE_NAMES = ('audio', 'curses', 'largefile', 'network',
+                  'decimal', 'cpu', 'subprocess', 'urlfetch', 'gui', 'tzdata')
+
+class _ArgParser(argparse.ArgumentParser):
+
+    def error(self, message):
+        super().error(message + "\nPass -h or --help for complete help.")
+
+
+def _create_parser():
+    # Set prog to prevent the uninformative "__main__.py" from displaying in
+    # error messages when using "python -m test ...".
+    parser = _ArgParser(prog='regrtest.py',
+                        usage=USAGE,
+                        description=DESCRIPTION,
+                        epilog=EPILOG,
+                        add_help=False,
+                        formatter_class=argparse.RawDescriptionHelpFormatter)
+
+    # Arguments with this clause added to its help are described further in
+    # the epilog's "Additional option details" section.
+    more_details = '  See the section at bottom for more details.'
+
+    group = parser.add_argument_group('General options')
+    # We add help explicitly to control what argument group it renders under.
+    group.add_argument('-h', '--help', action='help',
+                       help='show this help message and exit')
+    group.add_argument('--timeout', metavar='TIMEOUT', type=float,
+                        help='dump the traceback and exit if a test takes '
+                             'more than TIMEOUT seconds; disabled if TIMEOUT '
+                             'is negative or equals to zero')
+    group.add_argument('--wait', action='store_true',
+                       help='wait for user input, e.g., allow a debugger '
+                            'to be attached')
+    group.add_argument('--slaveargs', metavar='ARGS')
+    group.add_argument('-S', '--start', metavar='START',
+                       help='the name of the test at which to start.' +
+                            more_details)
+
+    group = parser.add_argument_group('Verbosity')
+    group.add_argument('-v', '--verbose', action='count',
+                       help='run tests in verbose mode with output to stdout')
+    group.add_argument('-w', '--verbose2', action='store_true',
+                       help='re-run failed tests in verbose mode')
+    group.add_argument('-W', '--verbose3', action='store_true',
+                       help='display test output on failure')
+    group.add_argument('-q', '--quiet', action='store_true',
+                       help='no output unless one or more tests fail')
+    group.add_argument('-o', '--slowest', action='store_true', dest='print_slow',
+                       help='print the slowest 10 tests')
+    group.add_argument('--header', action='store_true',
+                       help='print header with interpreter info')
+
+    group = parser.add_argument_group('Selecting tests')
+    group.add_argument('-r', '--randomize', action='store_true',
+                       help='randomize test execution order.' + more_details)
+    group.add_argument('--randseed', metavar='SEED',
+                       dest='random_seed', type=int,
+                       help='pass a random seed to reproduce a previous '
+                            'random run')
+    group.add_argument('-f', '--fromfile', metavar='FILE',
+                       help='read names of tests to run from a file.' +
+                            more_details)
+    group.add_argument('-x', '--exclude', action='store_true',
+                       help='arguments are tests to *exclude*')
+    group.add_argument('-s', '--single', action='store_true',
+                       help='single step through a set of tests.' +
+                            more_details)
+    group.add_argument('-m', '--match', metavar='PAT',
+                       dest='match_tests',
+                       help='match test cases and methods with glob pattern PAT')
+    group.add_argument('-G', '--failfast', action='store_true',
+                       help='fail as soon as a test fails (only with -v or -W)')
+    group.add_argument('-u', '--use', metavar='RES1,RES2,...',
+                       action='append', type=resources_list,
+                       help='specify which special resource intensive tests '
+                            'to run.' + more_details)
+    group.add_argument('-M', '--memlimit', metavar='LIMIT',
+                       help='run very large memory-consuming tests.' +
+                            more_details)
+    group.add_argument('--testdir', metavar='DIR',
+                       type=relative_filename,
+                       help='execute test files in the specified directory '
+                            '(instead of the Python stdlib test suite)')
+
+    group = parser.add_argument_group('Special runs')
+    group.add_argument('-l', '--findleaks', action='store_true',
+                       help='if GC is available detect tests that leak memory')
+    group.add_argument('-L', '--runleaks', action='store_true',
+                       help='run the leaks(1) command just before exit.' +
+                            more_details)
+    group.add_argument('-R', '--huntrleaks', metavar='RUNCOUNTS',
+                       type=huntrleaks,
+                       help='search for reference leaks (needs debug build, '
+                            'very slow).' + more_details)
+    group.add_argument('-j', '--multiprocess', metavar='PROCESSES',
+                       dest='use_mp', type=int,
+                       help='run PROCESSES processes at once')
+    group.add_argument('-T', '--coverage', action='store_true',
+                       dest='trace',
+                       help='turn on code coverage tracing using the trace '
+                            'module')
+    group.add_argument('-D', '--coverdir', metavar='DIR',
+                       type=relative_filename,
+                       help='directory where coverage files are put')
+    group.add_argument('-N', '--nocoverdir',
+                       action='store_const', const=None, dest='coverdir',
+                       help='put coverage files alongside modules')
+    group.add_argument('-t', '--threshold', metavar='THRESHOLD',
+                       type=int,
+                       help='call gc.set_threshold(THRESHOLD)')
+    group.add_argument('-n', '--nowindows', action='store_true',
+                       help='suppress error message boxes on Windows')
+    group.add_argument('-F', '--forever', action='store_true',
+                       help='run the specified tests in a loop, until an '
+                            'error happens')
+    group.add_argument('--list-tests', action='store_true',
+                       help="only write the name of tests that will be run, "
+                            "don't execute them")
+    group.add_argument('-P', '--pgo', dest='pgo', action='store_true',
+                       help='enable Profile Guided Optimization training')
+
+    return parser
+
+
+def relative_filename(string):
+    # CWD is replaced with a temporary dir before calling main(), so we
+    # join it with the saved CWD so it ends up where the user expects.
+    return os.path.join(support.SAVEDCWD, string)
+
+
+def huntrleaks(string):
+    args = string.split(':')
+    if len(args) not in (2, 3):
+        raise argparse.ArgumentTypeError(
+            'needs 2 or 3 colon-separated arguments')
+    nwarmup = int(args[0]) if args[0] else 5
+    ntracked = int(args[1]) if args[1] else 4
+    fname = args[2] if len(args) > 2 and args[2] else 'reflog.txt'
+    return nwarmup, ntracked, fname
+
+
+def resources_list(string):
+    u = [x.lower() for x in string.split(',')]
+    for r in u:
+        if r == 'all' or r == 'none':
+            continue
+        if r[0] == '-':
+            r = r[1:]
+        if r not in RESOURCE_NAMES:
+            raise argparse.ArgumentTypeError('invalid resource: ' + r)
+    return u
+
+
+def _parse_args(args, **kwargs):
+    # Defaults
+    ns = argparse.Namespace(testdir=None, verbose=0, quiet=False,
+         exclude=False, single=False, randomize=False, fromfile=None,
+         findleaks=False, use_resources=None, trace=False, coverdir='coverage',
+         runleaks=False, huntrleaks=False, verbose2=False, print_slow=False,
+         random_seed=None, use_mp=None, verbose3=False, forever=False,
+         header=False, failfast=False, match_tests=None, pgo=False)
+    for k, v in kwargs.items():
+        if not hasattr(ns, k):
+            raise TypeError('%r is an invalid keyword argument '
+                            'for this function' % k)
+        setattr(ns, k, v)
+    if ns.use_resources is None:
+        ns.use_resources = []
+
+    parser = _create_parser()
+    # Issue #14191: argparse doesn't support "intermixed" positional and
+    # optional arguments. Use parse_known_args() as workaround.
+    ns.args = parser.parse_known_args(args=args, namespace=ns)[1]
+    for arg in ns.args:
+        if arg.startswith('-'):
+            parser.error("unrecognized arguments: %s" % arg)
+            sys.exit(1)
+
+    if ns.single and ns.fromfile:
+        parser.error("-s and -f don't go together!")
+    if ns.use_mp and ns.trace:
+        parser.error("-T and -j don't go together!")
+    if ns.use_mp and ns.findleaks:
+        parser.error("-l and -j don't go together!")
+    if ns.failfast and not (ns.verbose or ns.verbose3):
+        parser.error("-G/--failfast needs either -v or -W")
+    if ns.pgo and (ns.verbose or ns.verbose2 or ns.verbose3):
+        parser.error("--pgo/-v don't go together!")
+
+    if ns.nowindows:
+        print("Warning: the --nowindows (-n) option is deprecated. "
+              "Use -vv to display assertions in stderr.", file=sys.stderr)
+
+    if ns.quiet:
+        ns.verbose = 0
+    if ns.timeout is not None:
+        if ns.timeout <= 0:
+            ns.timeout = None
+    if ns.use_mp is not None:
+        if ns.use_mp <= 0:
+            # Use all cores + extras for tests that like to sleep
+            ns.use_mp = 2 + (os.cpu_count() or 1)
+    if ns.use:
+        for a in ns.use:
+            for r in a:
+                if r == 'all':
+                    ns.use_resources[:] = RESOURCE_NAMES
+                    continue
+                if r == 'none':
+                    del ns.use_resources[:]
+                    continue
+                remove = False
+                if r[0] == '-':
+                    remove = True
+                    r = r[1:]
+                if remove:
+                    if r in ns.use_resources:
+                        ns.use_resources.remove(r)
+                elif r not in ns.use_resources:
+                    ns.use_resources.append(r)
+    if ns.random_seed is not None:
+        ns.randomize = True
+
+    return ns
diff --git a/Lib/test/libregrtest/main.py b/Lib/test/libregrtest/main.py
new file mode 100644
index 0000000000..f0effc973c
--- /dev/null
+++ b/Lib/test/libregrtest/main.py
@@ -0,0 +1,532 @@
+import datetime
+import faulthandler
+import locale
+import os
+import platform
+import random
+import re
+import sys
+import sysconfig
+import tempfile
+import textwrap
+import time
+from test.libregrtest.cmdline import _parse_args
+from test.libregrtest.runtest import (
+    findtests, runtest,
+    STDTESTS, NOTTESTS, PASSED, FAILED, ENV_CHANGED, SKIPPED, RESOURCE_DENIED,
+    INTERRUPTED, CHILD_ERROR,
+    PROGRESS_MIN_TIME, format_test_result)
+from test.libregrtest.setup import setup_tests
+from test import support
+try:
+    import gc
+except ImportError:
+    gc = None
+
+
+# When tests are run from the Python build directory, it is best practice
+# to keep the test files in a subfolder.  This eases the cleanup of leftover
+# files using the "make distclean" command.
+if sysconfig.is_python_build():
+    TEMPDIR = os.path.join(sysconfig.get_config_var('srcdir'), 'build')
+else:
+    TEMPDIR = tempfile.gettempdir()
+TEMPDIR = os.path.abspath(TEMPDIR)
+
+
+def format_duration(seconds):
+    if seconds < 1.0:
+        return '%.0f ms' % (seconds * 1e3)
+    if seconds < 60.0:
+        return '%.0f sec' % seconds
+
+    minutes, seconds = divmod(seconds, 60.0)
+    return '%.0f min %.0f sec' % (minutes, seconds)
+
+
+class Regrtest:
+    """Execute a test suite.
+
+    This also parses command-line options and modifies its behavior
+    accordingly.
+
+    tests -- a list of strings containing test names (optional)
+    testdir -- the directory in which to look for tests (optional)
+
+    Users other than the Python test suite will certainly want to
+    specify testdir; if it's omitted, the directory containing the
+    Python test suite is searched for.
+
+    If the tests argument is omitted, the tests listed on the
+    command-line will be used.  If that's empty, too, then all *.py
+    files beginning with test_ will be used.
+
+    The other default arguments (verbose, quiet, exclude,
+    single, randomize, findleaks, use_resources, trace, coverdir,
+    print_slow, and random_seed) allow programmers calling main()
+    directly to set the values that would normally be set by flags
+    on the command line.
+    """
+    def __init__(self):
+        # Namespace of command line options
+        self.ns = None
+
+        # tests
+        self.tests = []
+        self.selected = []
+
+        # test results
+        self.good = []
+        self.bad = []
+        self.skipped = []
+        self.resource_denieds = []
+        self.environment_changed = []
+        self.interrupted = False
+
+        # used by --slow
+        self.test_times = []
+
+        # used by --coverage, trace.Trace instance
+        self.tracer = None
+
+        # used by --findleaks, store for gc.garbage
+        self.found_garbage = []
+
+        # used to display the progress bar "[ 3/100]"
+        self.start_time = time.monotonic()
+        self.test_count = ''
+        self.test_count_width = 1
+
+        # used by --single
+        self.next_single_test = None
+        self.next_single_filename = None
+
+    def accumulate_result(self, test, result):
+        ok, test_time = result
+        if ok not in (CHILD_ERROR, INTERRUPTED):
+            self.test_times.append((test_time, test))
+        if ok == PASSED:
+            self.good.append(test)
+        elif ok == FAILED:
+            self.bad.append(test)
+        elif ok == ENV_CHANGED:
+            self.environment_changed.append(test)
+        elif ok == SKIPPED:
+            self.skipped.append(test)
+        elif ok == RESOURCE_DENIED:
+            self.skipped.append(test)
+            self.resource_denieds.append(test)
+
+    def display_progress(self, test_index, test):
+        if self.ns.quiet:
+            return
+        if self.bad and not self.ns.pgo:
+            fmt = "{time} [{test_index:{count_width}}{test_count}/{nbad}] {test_name}"
+        else:
+            fmt = "{time} [{test_index:{count_width}}{test_count}] {test_name}"
+        test_time = time.monotonic() - self.start_time
+        test_time = datetime.timedelta(seconds=int(test_time))
+        line = fmt.format(count_width=self.test_count_width,
+                          test_index=test_index,
+                          test_count=self.test_count,
+                          nbad=len(self.bad),
+                          test_name=test,
+                          time=test_time)
+        print(line, flush=True)
+
+    def parse_args(self, kwargs):
+        ns = _parse_args(sys.argv[1:], **kwargs)
+
+        if ns.timeout and not hasattr(faulthandler, 'dump_traceback_later'):
+            print("Warning: The timeout option requires "
+                  "faulthandler.dump_traceback_later", file=sys.stderr)
+            ns.timeout = None
+
+        if ns.threshold is not None and gc is None:
+            print('No GC available, ignore --threshold.', file=sys.stderr)
+            ns.threshold = None
+
+        if ns.findleaks:
+            if gc is not None:
+                # Uncomment the line below to report garbage that is not
+                # freeable by reference counting alone.  By default only
+                # garbage that is not collectable by the GC is reported.
+                pass
+                #gc.set_debug(gc.DEBUG_SAVEALL)
+            else:
+                print('No GC available, disabling --findleaks',
+                      file=sys.stderr)
+                ns.findleaks = False
+
+        # Strip .py extensions.
+        removepy(ns.args)
+
+        return ns
+
+    def find_tests(self, tests):
+        self.tests = tests
+
+        if self.ns.single:
+            self.next_single_filename = os.path.join(TEMPDIR, 'pynexttest')
+            try:
+                with open(self.next_single_filename, 'r') as fp:
+                    next_test = fp.read().strip()
+                    self.tests = [next_test]
+            except OSError:
+                pass
+
+        if self.ns.fromfile:
+            self.tests = []
+            # regex to match 'test_builtin' in line:
+            # '0:00:00 [  4/400] test_builtin -- test_dict took 1 sec'
+            regex = (r'^(?:[0-9]+:[0-9]+:[0-9]+ *)?'
+                     r'(?:\[[0-9/ ]+\] *)?'
+                     r'(test_[a-zA-Z0-9_]+)')
+            regex = re.compile(regex)
+            with open(os.path.join(support.SAVEDCWD, self.ns.fromfile)) as fp:
+                for line in fp:
+                    line = line.strip()
+                    if line.startswith('#'):
+                        continue
+                    match = regex.match(line)
+                    if match is None:
+                        continue
+                    self.tests.append(match.group(1))
+
+        removepy(self.tests)
+
+        stdtests = STDTESTS[:]
+        nottests = NOTTESTS.copy()
+        if self.ns.exclude:
+            for arg in self.ns.args:
+                if arg in stdtests:
+                    stdtests.remove(arg)
+                nottests.add(arg)
+            self.ns.args = []
+
+        # if testdir is set, then we are not running the python tests suite, so
+        # don't add default tests to be executed or skipped (pass empty values)
+        if self.ns.testdir:
+            alltests = findtests(self.ns.testdir, list(), set())
+        else:
+            alltests = findtests(self.ns.testdir, stdtests, nottests)
+
+        if not self.ns.fromfile:
+            self.selected = self.tests or self.ns.args or alltests
+        else:
+            self.selected = self.tests
+        if self.ns.single:
+            self.selected = self.selected[:1]
+            try:
+                pos = alltests.index(self.selected[0])
+                self.next_single_test = alltests[pos + 1]
+            except IndexError:
+                pass
+
+        # Remove all the selected tests that precede start if it's set.
+        if self.ns.start:
+            try:
+                del self.selected[:self.selected.index(self.ns.start)]
+            except ValueError:
+                print("Couldn't find starting test (%s), using all tests"
+                      % self.ns.start, file=sys.stderr)
+
+        if self.ns.randomize:
+            if self.ns.random_seed is None:
+                self.ns.random_seed = random.randrange(10000000)
+            random.seed(self.ns.random_seed)
+            random.shuffle(self.selected)
+
+    def list_tests(self):
+        for name in self.selected:
+            print(name)
+
+    def rerun_failed_tests(self):
+        self.ns.verbose = True
+        self.ns.failfast = False
+        self.ns.verbose3 = False
+        self.ns.match_tests = None
+
+        print("Re-running failed tests in verbose mode")
+        for test in self.bad[:]:
+            print("Re-running test %r in verbose mode" % test, flush=True)
+            try:
+                self.ns.verbose = True
+                ok = runtest(self.ns, test)
+            except KeyboardInterrupt:
+                self.interrupted = True
+                # print a newline separate from the ^C
+                print()
+                break
+            else:
+                if ok[0] in {PASSED, ENV_CHANGED, SKIPPED, RESOURCE_DENIED}:
+                    self.bad.remove(test)
+        else:
+            if self.bad:
+                print(count(len(self.bad), 'test'), "failed again:")
+                printlist(self.bad)
+
+    def display_result(self):
+        if self.interrupted:
+            # print a newline after ^C
+            print()
+            print("Test suite interrupted by signal SIGINT.")
+            executed = set(self.good) | set(self.bad) | set(self.skipped)
+            omitted = set(self.selected) - executed
+            print(count(len(omitted), "test"), "omitted:")
+            printlist(omitted)
+
+        # If running the test suite for PGO then no one cares about
+        # results.
+        if self.ns.pgo:
+            return
+
+        if self.good and not self.ns.quiet:
+            if (not self.bad
+                and not self.skipped
+                and not self.interrupted
+                and len(self.good) > 1):
+                print("All", end=' ')
+            print(count(len(self.good), "test"), "OK.")
+
+        if self.ns.print_slow:
+            self.test_times.sort(reverse=True)
+            print()
+            print("10 slowest tests:")
+            for time, test in self.test_times[:10]:
+                print("- %s: %s" % (test, format_duration(time)))
+
+        if self.bad:
+            print()
+            print(count(len(self.bad), "test"), "failed:")
+            printlist(self.bad)
+
+        if self.environment_changed:
+            print()
+            print("{} altered the execution environment:".format(
+                     count(len(self.environment_changed), "test")))
+            printlist(self.environment_changed)
+
+        if self.skipped and not self.ns.quiet:
+            print()
+            print(count(len(self.skipped), "test"), "skipped:")
+            printlist(self.skipped)
+
+    def run_tests_sequential(self):
+        if self.ns.trace:
+            import trace
+            self.tracer = trace.Trace(trace=False, count=True)
+
+        save_modules = sys.modules.keys()
+
+        print("Run tests sequentially")
+
+        previous_test = None
+        for test_index, test in enumerate(self.tests, 1):
+            start_time = time.monotonic()
+
+            text = test
+            if previous_test:
+                text = '%s -- %s' % (text, previous_test)
+            self.display_progress(test_index, text)
+
+            if self.tracer:
+                # If we're tracing code coverage, then we don't exit with status
+                # if on a false return value from main.
+                cmd = ('result = runtest(self.ns, test); '
+                       'self.accumulate_result(test, result)')
+                ns = dict(locals())
+                self.tracer.runctx(cmd, globals=globals(), locals=ns)
+                result = ns['result']
+            else:
+                try:
+                    result = runtest(self.ns, test)
+                except KeyboardInterrupt:
+                    self.interrupted = True
+                    self.accumulate_result(test, (INTERRUPTED, None))
+                    break
+                else:
+                    self.accumulate_result(test, result)
+
+            previous_test = format_test_result(test, result[0])
+            test_time = time.monotonic() - start_time
+            if test_time >= PROGRESS_MIN_TIME:
+                previous_test = "%s in %s" % (previous_test, format_duration(test_time))
+            elif result[0] == PASSED:
+                # be quiet: say nothing if the test passed shortly
+                previous_test = None
+
+            if self.ns.findleaks:
+                gc.collect()
+                if gc.garbage:
+                    print("Warning: test created", len(gc.garbage), end=' ')
+                    print("uncollectable object(s).")
+                    # move the uncollectable objects somewhere so we don't see
+                    # them again
+                    self.found_garbage.extend(gc.garbage)
+                    del gc.garbage[:]
+
+            # Unload the newly imported modules (best effort finalization)
+            for module in sys.modules.keys():
+                if module not in save_modules and module.startswith("test."):
+                    support.unload(module)
+
+        if previous_test:
+            print(previous_test)
+
+    def _test_forever(self, tests):
+        while True:
+            for test in tests:
+                yield test
+                if self.bad:
+                    return
+
+    def run_tests(self):
+        # For a partial run, we do not need to clutter the output.
+        if (self.ns.verbose
+            or self.ns.header
+            or not (self.ns.pgo or self.ns.quiet or self.ns.single
+                    or self.tests or self.ns.args)):
+            # Print basic platform information
+            print("==", platform.python_implementation(), *sys.version.split())
+            print("==  ", platform.platform(aliased=True),
+                          "%s-endian" % sys.byteorder)
+            print("==  ", "hash algorithm:", sys.hash_info.algorithm,
+                  "64bit" if sys.maxsize > 2**32 else "32bit")
+            print("==  cwd:", os.getcwd())
+            print("==  encodings: locale=%s, FS=%s"
+                  % (locale.getpreferredencoding(False),
+                     sys.getfilesystemencoding()))
+            print("Testing with flags:", sys.flags)
+
+        if self.ns.randomize:
+            print("Using random seed", self.ns.random_seed)
+
+        if self.ns.forever:
+            self.tests = self._test_forever(list(self.selected))
+            self.test_count = ''
+            self.test_count_width = 3
+        else:
+            self.tests = iter(self.selected)
+            self.test_count = '/{}'.format(len(self.selected))
+            self.test_count_width = len(self.test_count) - 1
+
+        if self.ns.use_mp:
+            from test.libregrtest.runtest_mp import run_tests_multiprocess
+            run_tests_multiprocess(self)
+        else:
+            self.run_tests_sequential()
+
+    def finalize(self):
+        if self.next_single_filename:
+            if self.next_single_test:
+                with open(self.next_single_filename, 'w') as fp:
+                    fp.write(self.next_single_test + '\n')
+            else:
+                os.unlink(self.next_single_filename)
+
+        if self.tracer:
+            r = self.tracer.results()
+            r.write_results(show_missing=True, summary=True,
+                            coverdir=self.ns.coverdir)
+
+        print()
+        duration = time.monotonic() - self.start_time
+        print("Total duration: %s" % format_duration(duration))
+
+        if self.bad:
+            result = "FAILURE"
+        elif self.interrupted:
+            result = "INTERRUPTED"
+        else:
+            result = "SUCCESS"
+        print("Tests result: %s" % result)
+
+        if self.ns.runleaks:
+            os.system("leaks %d" % os.getpid())
+
+    def main(self, tests=None, **kwargs):
+        global TEMPDIR
+
+        if sysconfig.is_python_build():
+            try:
+                os.mkdir(TEMPDIR)
+            except FileExistsError:
+                pass
+
+        # Define a writable temp dir that will be used as cwd while running
+        # the tests. The name of the dir includes the pid to allow parallel
+        # testing (see the -j option).
+        test_cwd = 'test_python_{}'.format(os.getpid())
+        test_cwd = os.path.join(TEMPDIR, test_cwd)
+
+        # Run the tests in a context manager that temporarily changes the CWD to a
+        # temporary and writable directory.  If it's not possible to create or
+        # change the CWD, the original CWD will be used.  The original CWD is
+        # available from support.SAVEDCWD.
+        with support.temp_cwd(test_cwd, quiet=True):
+            self._main(tests, kwargs)
+
+    def _main(self, tests, kwargs):
+        self.ns = self.parse_args(kwargs)
+
+        if self.ns.slaveargs is not None:
+            from test.libregrtest.runtest_mp import run_tests_slave
+            run_tests_slave(self.ns.slaveargs)
+
+        if self.ns.wait:
+            input("Press any key to continue...")
+
+        support.PGO = self.ns.pgo
+
+        setup_tests(self.ns)
+
+        self.find_tests(tests)
+
+        if self.ns.list_tests:
+            self.list_tests()
+            sys.exit(0)
+
+        self.run_tests()
+        self.display_result()
+
+        if self.ns.verbose2 and self.bad:
+            self.rerun_failed_tests()
+
+        self.finalize()
+        sys.exit(len(self.bad) > 0 or self.interrupted)
+
+
+def removepy(names):
+    if not names:
+        return
+    for idx, name in enumerate(names):
+        basename, ext = os.path.splitext(name)
+        if ext == '.py':
+            names[idx] = basename
+
+
+def count(n, word):
+    if n == 1:
+        return "%d %s" % (n, word)
+    else:
+        return "%d %ss" % (n, word)
+
+
+def printlist(x, width=70, indent=4):
+    """Print the elements of iterable x to stdout.
+
+    Optional arg width (default 70) is the maximum line length.
+    Optional arg indent (default 4) is the number of blanks with which to
+    begin each line.
+    """
+
+    blanks = ' ' * indent
+    # Print the sorted list: 'x' may be a '--random' list or a set()
+    print(textwrap.fill(' '.join(str(elt) for elt in sorted(x)), width,
+                        initial_indent=blanks, subsequent_indent=blanks))
+
+
+def main(tests=None, **kwargs):
+    """Run the Python suite."""
+    Regrtest().main(tests=tests, **kwargs)
diff --git a/Lib/test/libregrtest/refleak.py b/Lib/test/libregrtest/refleak.py
new file mode 100644
index 0000000000..3b3d2458b1
--- /dev/null
+++ b/Lib/test/libregrtest/refleak.py
@@ -0,0 +1,269 @@
+import errno
+import os
+import re
+import sys
+import warnings
+from inspect import isabstract
+from test import support
+
+
+try:
+    MAXFD = os.sysconf("SC_OPEN_MAX")
+except Exception:
+    MAXFD = 256
+
+
+def fd_count():
+    """Count the number of open file descriptors"""
+    if sys.platform.startswith(('linux', 'freebsd')):
+        try:
+            names = os.listdir("/proc/self/fd")
+            return len(names)
+        except FileNotFoundError:
+            pass
+
+    count = 0
+    for fd in range(MAXFD):
+        try:
+            # Prefer dup() over fstat(). fstat() can require input/output
+            # whereas dup() doesn't.
+            fd2 = os.dup(fd)
+        except OSError as e:
+            if e.errno != errno.EBADF:
+                raise
+        else:
+            os.close(fd2)
+            count += 1
+    return count
+
+
+def dash_R(the_module, test, indirect_test, huntrleaks):
+    """Run a test multiple times, looking for reference leaks.
+
+    Returns:
+        False if the test didn't leak references; True if we detected refleaks.
+    """
+    # This code is hackish and inelegant, but it seems to do the job.
+    import copyreg
+    import collections.abc
+
+    if not hasattr(sys, 'gettotalrefcount'):
+        raise Exception("Tracking reference leaks requires a debug build "
+                        "of Python")
+
+    # Save current values for dash_R_cleanup() to restore.
+    fs = warnings.filters[:]
+    ps = copyreg.dispatch_table.copy()
+    pic = sys.path_importer_cache.copy()
+    try:
+        import zipimport
+    except ImportError:
+        zdc = None # Run unmodified on platforms without zipimport support
+    else:
+        zdc = zipimport._zip_directory_cache.copy()
+    abcs = {}
+    for abc in [getattr(collections.abc, a) for a in collections.abc.__all__]:
+        if not isabstract(abc):
+            continue
+        for obj in abc.__subclasses__() + [abc]:
+            abcs[obj] = obj._abc_registry.copy()
+
+    nwarmup, ntracked, fname = huntrleaks
+    fname = os.path.join(support.SAVEDCWD, fname)
+    repcount = nwarmup + ntracked
+    rc_deltas = [0] * repcount
+    alloc_deltas = [0] * repcount
+    fd_deltas = [0] * repcount
+
+    print("beginning", repcount, "repetitions", file=sys.stderr)
+    print(("1234567890"*(repcount//10 + 1))[:repcount], file=sys.stderr,
+          flush=True)
+    # initialize variables to make pyflakes quiet
+    rc_before = alloc_before = fd_before = 0
+    for i in range(repcount):
+        indirect_test()
+        alloc_after, rc_after, fd_after = dash_R_cleanup(fs, ps, pic, zdc,
+                                                         abcs)
+        print('.', end='', flush=True)
+        if i >= nwarmup:
+            rc_deltas[i] = rc_after - rc_before
+            alloc_deltas[i] = alloc_after - alloc_before
+            fd_deltas[i] = fd_after - fd_before
+        alloc_before = alloc_after
+        rc_before = rc_after
+        fd_before = fd_after
+    print(file=sys.stderr)
+    # These checkers return False on success, True on failure
+    def check_rc_deltas(deltas):
+        return any(deltas)
+    def check_alloc_deltas(deltas):
+        # At least 1/3rd of 0s
+        if 3 * deltas.count(0) < len(deltas):
+            return True
+        # Nothing else than 1s, 0s and -1s
+        if not set(deltas) <= {1,0,-1}:
+            return True
+        return False
+    failed = False
+    for deltas, item_name, checker in [
+        (rc_deltas, 'references', check_rc_deltas),
+        (alloc_deltas, 'memory blocks', check_alloc_deltas),
+        (fd_deltas, 'file descriptors', check_rc_deltas)]:
+        if checker(deltas):
+            msg = '%s leaked %s %s, sum=%s' % (
+                test, deltas[nwarmup:], item_name, sum(deltas))
+            print(msg, file=sys.stderr, flush=True)
+            with open(fname, "a") as refrep:
+                print(msg, file=refrep)
+                refrep.flush()
+            failed = True
+    return failed
+
+
+def dash_R_cleanup(fs, ps, pic, zdc, abcs):
+    import gc, copyreg
+    import collections.abc
+    from weakref import WeakSet
+
+    # Restore some original values.
+    warnings.filters[:] = fs
+    copyreg.dispatch_table.clear()
+    copyreg.dispatch_table.update(ps)
+    sys.path_importer_cache.clear()
+    sys.path_importer_cache.update(pic)
+    try:
+        import zipimport
+    except ImportError:
+        pass # Run unmodified on platforms without zipimport support
+    else:
+        zipimport._zip_directory_cache.clear()
+        zipimport._zip_directory_cache.update(zdc)
+
+    # clear type cache
+    sys._clear_type_cache()
+
+    # Clear ABC registries, restoring previously saved ABC registries.
+    for abc in [getattr(collections.abc, a) for a in collections.abc.__all__]:
+        if not isabstract(abc):
+            continue
+        for obj in abc.__subclasses__() + [abc]:
+            obj._abc_registry = abcs.get(obj, WeakSet()).copy()
+            obj._abc_cache.clear()
+            obj._abc_negative_cache.clear()
+
+    clear_caches()
+
+    # Collect cyclic trash and read memory statistics immediately after.
+    func1 = sys.getallocatedblocks
+    func2 = sys.gettotalrefcount
+    gc.collect()
+    return func1(), func2(), fd_count()
+
+
+def clear_caches():
+    import gc
+
+    # Clear the warnings registry, so they can be displayed again
+    for mod in sys.modules.values():
+        if hasattr(mod, '__warningregistry__'):
+            del mod.__warningregistry__
+
+    # Flush standard output, so that buffered data is sent to the OS and
+    # associated Python objects are reclaimed.
+    for stream in (sys.stdout, sys.stderr, sys.__stdout__, sys.__stderr__):
+        if stream is not None:
+            stream.flush()
+
+    # Clear assorted module caches.
+    # Don't worry about resetting the cache if the module is not loaded
+    try:
+        distutils_dir_util = sys.modules['distutils.dir_util']
+    except KeyError:
+        pass
+    else:
+        distutils_dir_util._path_created.clear()
+    re.purge()
+
+    try:
+        _strptime = sys.modules['_strptime']
+    except KeyError:
+        pass
+    else:
+        _strptime._regex_cache.clear()
+
+    try:
+        urllib_parse = sys.modules['urllib.parse']
+    except KeyError:
+        pass
+    else:
+        urllib_parse.clear_cache()
+
+    try:
+        urllib_request = sys.modules['urllib.request']
+    except KeyError:
+        pass
+    else:
+        urllib_request.urlcleanup()
+
+    try:
+        linecache = sys.modules['linecache']
+    except KeyError:
+        pass
+    else:
+        linecache.clearcache()
+
+    try:
+        mimetypes = sys.modules['mimetypes']
+    except KeyError:
+        pass
+    else:
+        mimetypes._default_mime_types()
+
+    try:
+        filecmp = sys.modules['filecmp']
+    except KeyError:
+        pass
+    else:
+        filecmp._cache.clear()
+
+    try:
+        struct = sys.modules['struct']
+    except KeyError:
+        pass
+    else:
+        struct._clearcache()
+
+    try:
+        doctest = sys.modules['doctest']
+    except KeyError:
+        pass
+    else:
+        doctest.master = None
+
+    try:
+        ctypes = sys.modules['ctypes']
+    except KeyError:
+        pass
+    else:
+        ctypes._reset_cache()
+
+    try:
+        typing = sys.modules['typing']
+    except KeyError:
+        pass
+    else:
+        for f in typing._cleanups:
+            f()
+
+    gc.collect()
+
+
+def warm_caches():
+    # char cache
+    s = bytes(range(256))
+    for i in range(256):
+        s[i:i+1]
+    # unicode cache
+    [chr(i) for i in range(256)]
+    # int cache
+    list(range(-5, 257))
diff --git a/Lib/test/libregrtest/runtest.py b/Lib/test/libregrtest/runtest.py
new file mode 100644
index 0000000000..ba0df0a317
--- /dev/null
+++ b/Lib/test/libregrtest/runtest.py
@@ -0,0 +1,245 @@
+import faulthandler
+import importlib
+import io
+import os
+import sys
+import time
+import traceback
+import unittest
+from test import support
+from test.libregrtest.refleak import dash_R, clear_caches
+from test.libregrtest.save_env import saved_test_environment
+
+
+# Test result constants.
+PASSED = 1
+FAILED = 0
+ENV_CHANGED = -1
+SKIPPED = -2
+RESOURCE_DENIED = -3
+INTERRUPTED = -4
+CHILD_ERROR = -5   # error in a child process
+
+_FORMAT_TEST_RESULT = {
+    PASSED: '%s passed',
+    FAILED: '%s failed',
+    ENV_CHANGED: '%s failed (env changed)',
+    SKIPPED: '%s skipped',
+    RESOURCE_DENIED: '%s skipped (resource denied)',
+    INTERRUPTED: '%s interrupted',
+    CHILD_ERROR: '%s crashed',
+}
+
+# Minimum duration of a test to display its duration or to mention that
+# the test is running in background
+PROGRESS_MIN_TIME = 30.0   # seconds
+
+# small set of tests to determine if we have a basically functioning interpreter
+# (i.e. if any of these fail, then anything else is likely to follow)
+STDTESTS = [
+    'test_grammar',
+    'test_opcodes',
+    'test_dict',
+    'test_builtin',
+    'test_exceptions',
+    'test_types',
+    'test_unittest',
+    'test_doctest',
+    'test_doctest2',
+    'test_support'
+]
+
+# set of tests that we don't want to be executed when using regrtest
+NOTTESTS = set()
+
+
+def format_test_result(test_name, result):
+    fmt = _FORMAT_TEST_RESULT.get(result, "%s")
+    return fmt % test_name
+
+
+def findtests(testdir=None, stdtests=STDTESTS, nottests=NOTTESTS):
+    """Return a list of all applicable test modules."""
+    testdir = findtestdir(testdir)
+    names = os.listdir(testdir)
+    tests = []
+    others = set(stdtests) | nottests
+    for name in names:
+        mod, ext = os.path.splitext(name)
+        if mod[:5] == "test_" and ext in (".py", "") and mod not in others:
+            tests.append(mod)
+    return stdtests + sorted(tests)
+
+
+def runtest(ns, test):
+    """Run a single test.
+
+    ns -- regrtest namespace of options
+    test -- the name of the test
+
+    Returns the tuple (result, test_time), where result is one of the
+    constants:
+
+        INTERRUPTED      KeyboardInterrupt when run under -j
+        RESOURCE_DENIED  test skipped because resource denied
+        SKIPPED          test skipped for some other reason
+        ENV_CHANGED      test failed because it changed the execution environment
+        FAILED           test failed
+        PASSED           test passed
+    """
+
+    output_on_failure = ns.verbose3
+
+    use_timeout = (ns.timeout is not None)
+    if use_timeout:
+        faulthandler.dump_traceback_later(ns.timeout, exit=True)
+    try:
+        support.match_tests = ns.match_tests
+        if ns.failfast:
+            support.failfast = True
+        if output_on_failure:
+            support.verbose = True
+
+            # Reuse the same instance to all calls to runtest(). Some
+            # tests keep a reference to sys.stdout or sys.stderr
+            # (eg. test_argparse).
+            if runtest.stringio is None:
+                stream = io.StringIO()
+                runtest.stringio = stream
+            else:
+                stream = runtest.stringio
+                stream.seek(0)
+                stream.truncate()
+
+            orig_stdout = sys.stdout
+            orig_stderr = sys.stderr
+            try:
+                sys.stdout = stream
+                sys.stderr = stream
+                result = runtest_inner(ns, test, display_failure=False)
+                if result[0] != PASSED:
+                    output = stream.getvalue()
+                    orig_stderr.write(output)
+                    orig_stderr.flush()
+            finally:
+                sys.stdout = orig_stdout
+                sys.stderr = orig_stderr
+        else:
+            support.verbose = ns.verbose  # Tell tests to be moderately quiet
+            result = runtest_inner(ns, test, display_failure=not ns.verbose)
+        return result
+    finally:
+        if use_timeout:
+            faulthandler.cancel_dump_traceback_later()
+        cleanup_test_droppings(test, ns.verbose)
+runtest.stringio = None
+
+
+def runtest_inner(ns, test, display_failure=True):
+    support.unload(test)
+
+    test_time = 0.0
+    refleak = False  # True if the test leaked references.
+    try:
+        if test.startswith('test.') or ns.testdir:
+            abstest = test
+        else:
+            # Always import it from the test package
+            abstest = 'test.' + test
+        clear_caches()
+        with saved_test_environment(test, ns.verbose, ns.quiet, pgo=ns.pgo) as environment:
+            start_time = time.time()
+            the_module = importlib.import_module(abstest)
+            # If the test has a test_main, that will run the appropriate
+            # tests.  If not, use normal unittest test loading.
+            test_runner = getattr(the_module, "test_main", None)
+            if test_runner is None:
+                def test_runner():
+                    loader = unittest.TestLoader()
+                    tests = loader.loadTestsFromModule(the_module)
+                    for error in loader.errors:
+                        print(error, file=sys.stderr)
+                    if loader.errors:
+                        raise Exception("errors while loading tests")
+                    support.run_unittest(tests)
+            test_runner()
+            if ns.huntrleaks:
+                refleak = dash_R(the_module, test, test_runner, ns.huntrleaks)
+            test_time = time.time() - start_time
+    except support.ResourceDenied as msg:
+        if not ns.quiet and not ns.pgo:
+            print(test, "skipped --", msg, flush=True)
+        return RESOURCE_DENIED, test_time
+    except unittest.SkipTest as msg:
+        if not ns.quiet and not ns.pgo:
+            print(test, "skipped --", msg, flush=True)
+        return SKIPPED, test_time
+    except KeyboardInterrupt:
+        raise
+    except support.TestFailed as msg:
+        if not ns.pgo:
+            if display_failure:
+                print("test", test, "failed --", msg, file=sys.stderr,
+                      flush=True)
+            else:
+                print("test", test, "failed", file=sys.stderr, flush=True)
+        return FAILED, test_time
+    except:
+        msg = traceback.format_exc()
+        if not ns.pgo:
+            print("test", test, "crashed --", msg, file=sys.stderr,
+                  flush=True)
+        return FAILED, test_time
+    else:
+        if refleak:
+            return FAILED, test_time
+        if environment.changed:
+            return ENV_CHANGED, test_time
+        return PASSED, test_time
+
+
+def cleanup_test_droppings(testname, verbose):
+    import shutil
+    import stat
+    import gc
+
+    # First kill any dangling references to open files etc.
+    # This can also issue some ResourceWarnings which would otherwise get
+    # triggered during the following test run, and possibly produce failures.
+    gc.collect()
+
+    # Try to clean up junk commonly left behind.  While tests shouldn't leave
+    # any files or directories behind, when a test fails that can be tedious
+    # for it to arrange.  The consequences can be especially nasty on Windows,
+    # since if a test leaves a file open, it cannot be deleted by name (while
+    # there's nothing we can do about that here either, we can display the
+    # name of the offending test, which is a real help).
+    for name in (support.TESTFN,
+                 "db_home",
+                ):
+        if not os.path.exists(name):
+            continue
+
+        if os.path.isdir(name):
+            kind, nuker = "directory", shutil.rmtree
+        elif os.path.isfile(name):
+            kind, nuker = "file", os.unlink
+        else:
+            raise SystemError("os.path says %r exists but is neither "
+                              "directory nor file" % name)
+
+        if verbose:
+            print("%r left behind %s %r" % (testname, kind, name))
+        try:
+            # if we have chmod, fix possible permissions problems
+            # that might prevent cleanup
+            if (hasattr(os, 'chmod')):
+                os.chmod(name, stat.S_IRWXU | stat.S_IRWXG | stat.S_IRWXO)
+            nuker(name)
+        except Exception as msg:
+            print(("%r left behind %s %r and it couldn't be "
+                "removed: %s" % (testname, kind, name, msg)), file=sys.stderr)
+
+
+def findtestdir(path=None):
+    return path or os.path.dirname(os.path.dirname(__file__)) or os.curdir
diff --git a/Lib/test/libregrtest/runtest_mp.py b/Lib/test/libregrtest/runtest_mp.py
new file mode 100644
index 0000000000..9604c16600
--- /dev/null
+++ b/Lib/test/libregrtest/runtest_mp.py
@@ -0,0 +1,245 @@
+import faulthandler
+import json
+import os
+import queue
+import sys
+import time
+import traceback
+import types
+from test import support
+try:
+    import threading
+except ImportError:
+    print("Multiprocess option requires thread support")
+    sys.exit(2)
+
+from test.libregrtest.runtest import (
+    runtest, INTERRUPTED, CHILD_ERROR, PROGRESS_MIN_TIME,
+    format_test_result)
+from test.libregrtest.setup import setup_tests
+
+
+# Display the running tests if nothing happened last N seconds
+PROGRESS_UPDATE = 30.0   # seconds
+
+# If interrupted, display the wait progress every N seconds
+WAIT_PROGRESS = 2.0   # seconds
+
+
+def run_test_in_subprocess(testname, ns):
+    """Run the given test in a subprocess with --slaveargs.
+
+    ns is the option Namespace parsed from command-line arguments. regrtest
+    is invoked in a subprocess with the --slaveargs argument; when the
+    subprocess exits, its return code, stdout and stderr are returned as a
+    3-tuple.
+    """
+    from subprocess import Popen, PIPE
+
+    ns_dict = vars(ns)
+    slaveargs = (ns_dict, testname)
+    slaveargs = json.dumps(slaveargs)
+
+    cmd = [sys.executable, *support.args_from_interpreter_flags(),
+           '-X', 'faulthandler',
+           '-m', 'test.regrtest',
+           '--slaveargs', slaveargs]
+    if ns.pgo:
+        cmd += ['--pgo']
+
+    # Running the child from the same working directory as regrtest's original
+    # invocation ensures that TEMPDIR for the child is the same when
+    # sysconfig.is_python_build() is true. See issue 15300.
+    popen = Popen(cmd,
+                  stdout=PIPE, stderr=PIPE,
+                  universal_newlines=True,
+                  close_fds=(os.name != 'nt'),
+                  cwd=support.SAVEDCWD)
+    with popen:
+        stdout, stderr = popen.communicate()
+        retcode = popen.wait()
+    return retcode, stdout, stderr
+
+
+def run_tests_slave(slaveargs):
+    ns_dict, testname = json.loads(slaveargs)
+    ns = types.SimpleNamespace(**ns_dict)
+
+    setup_tests(ns)
+
+    try:
+        result = runtest(ns, testname)
+    except KeyboardInterrupt:
+        result = INTERRUPTED, ''
+    except BaseException as e:
+        traceback.print_exc()
+        result = CHILD_ERROR, str(e)
+
+    print()   # Force a newline (just in case)
+    print(json.dumps(result), flush=True)
+    sys.exit(0)
+
+
+# We do not use a generator so multiple threads can call next().
+class MultiprocessIterator:
+
+    """A thread-safe iterator over tests for multiprocess mode."""
+
+    def __init__(self, tests):
+        self.interrupted = False
+        self.lock = threading.Lock()
+        self.tests = tests
+
+    def __iter__(self):
+        return self
+
+    def __next__(self):
+        with self.lock:
+            if self.interrupted:
+                raise StopIteration('tests interrupted')
+            return next(self.tests)
+
+
+class MultiprocessThread(threading.Thread):
+    def __init__(self, pending, output, ns):
+        super().__init__()
+        self.pending = pending
+        self.output = output
+        self.ns = ns
+        self.current_test = None
+        self.start_time = None
+
+    def _runtest(self):
+        try:
+            test = next(self.pending)
+        except StopIteration:
+            self.output.put((None, None, None, None))
+            return True
+
+        try:
+            self.start_time = time.monotonic()
+            self.current_test = test
+
+            retcode, stdout, stderr = run_test_in_subprocess(test, self.ns)
+        finally:
+            self.current_test = None
+
+        stdout, _, result = stdout.strip().rpartition("\n")
+        if retcode != 0:
+            result = (CHILD_ERROR, "Exit code %s" % retcode)
+            self.output.put((test, stdout.rstrip(), stderr.rstrip(),
+                             result))
+            return True
+
+        if not result:
+            self.output.put((None, None, None, None))
+            return True
+
+        result = json.loads(result)
+        self.output.put((test, stdout.rstrip(), stderr.rstrip(),
+                         result))
+        return False
+
+    def run(self):
+        try:
+            stop = False
+            while not stop:
+                stop = self._runtest()
+        except BaseException:
+            self.output.put((None, None, None, None))
+            raise
+
+
+def run_tests_multiprocess(regrtest):
+    output = queue.Queue()
+    pending = MultiprocessIterator(regrtest.tests)
+    test_timeout = regrtest.ns.timeout
+    use_timeout = (test_timeout is not None)
+
+    workers = [MultiprocessThread(pending, output, regrtest.ns)
+               for i in range(regrtest.ns.use_mp)]
+    print("Run tests in parallel using %s child processes"
+          % len(workers))
+    for worker in workers:
+        worker.start()
+
+    def get_running(workers):
+        running = []
+        for worker in workers:
+            current_test = worker.current_test
+            if not current_test:
+                continue
+            dt = time.monotonic() - worker.start_time
+            if dt >= PROGRESS_MIN_TIME:
+                running.append('%s (%.0f sec)' % (current_test, dt))
+        return running
+
+    finished = 0
+    test_index = 1
+    get_timeout = max(PROGRESS_UPDATE, PROGRESS_MIN_TIME)
+    try:
+        while finished < regrtest.ns.use_mp:
+            if use_timeout:
+                faulthandler.dump_traceback_later(test_timeout, exit=True)
+
+            try:
+                item = output.get(timeout=get_timeout)
+            except queue.Empty:
+                running = get_running(workers)
+                if running and not regrtest.ns.pgo:
+                    print('running: %s' % ', '.join(running))
+                continue
+
+            test, stdout, stderr, result = item
+            if test is None:
+                finished += 1
+                continue
+            regrtest.accumulate_result(test, result)
+
+            # Display progress
+            ok, test_time = result
+            text = format_test_result(test, ok)
+            if (ok not in (CHILD_ERROR, INTERRUPTED)
+                and test_time >= PROGRESS_MIN_TIME
+                and not regrtest.ns.pgo):
+                text += ' (%.0f sec)' % test_time
+            running = get_running(workers)
+            if running and not regrtest.ns.pgo:
+                text += ' -- running: %s' % ', '.join(running)
+            regrtest.display_progress(test_index, text)
+
+            # Copy stdout and stderr from the child process
+            if stdout:
+                print(stdout, flush=True)
+            if stderr and not regrtest.ns.pgo:
+                print(stderr, file=sys.stderr, flush=True)
+
+            if result[0] == INTERRUPTED:
+                raise KeyboardInterrupt
+            if result[0] == CHILD_ERROR:
+                msg = "Child error on {}: {}".format(test, result[1])
+                raise Exception(msg)
+            test_index += 1
+    except KeyboardInterrupt:
+        regrtest.interrupted = True
+        pending.interrupted = True
+        print()
+    finally:
+        if use_timeout:
+            faulthandler.cancel_dump_traceback_later()
+
+    # If tests are interrupted, wait until tests complete
+    wait_start = time.monotonic()
+    while True:
+        running = [worker.current_test for worker in workers]
+        running = list(filter(bool, running))
+        if not running:
+            break
+
+        dt = time.monotonic() - wait_start
+        line = "Waiting for %s (%s tests)" % (', '.join(running), len(running))
+        if dt >= WAIT_PROGRESS:
+            line = "%s since %.0f sec" % (line, dt)
+        print(line)
+        for worker in workers:
+            worker.join(WAIT_PROGRESS)
diff --git a/Lib/test/libregrtest/save_env.py b/Lib/test/libregrtest/save_env.py
new file mode 100644
index 0000000000..96ad3af8df
--- /dev/null
+++ b/Lib/test/libregrtest/save_env.py
@@ -0,0 +1,285 @@
+import builtins
+import locale
+import logging
+import os
+import shutil
+import sys
+import sysconfig
+import warnings
+from test import support
+try:
+    import threading
+except ImportError:
+    threading = None
+try:
+    import _multiprocessing, multiprocessing.process
+except ImportError:
+    multiprocessing = None
+
+
+# Unit tests are supposed to leave the execution environment unchanged
+# once they complete.  But sometimes tests have bugs, especially when
+# tests fail, and the changes to environment go on to mess up other
+# tests.  This can cause issues with buildbot stability, since tests
+# are run in random order and so problems may appear to come and go.
+# There are a few things we can save and restore to mitigate this, and
+# the following context manager handles this task.
+
+class saved_test_environment:
+    """Save bits of the test environment and restore them at block exit.
+
+        with saved_test_environment(testname, verbose, quiet):
+            #stuff
+
+    Unless quiet is True, a warning is printed to stderr if any of
+    the saved items was changed by the test.  The attribute 'changed'
+    is initially False, but is set to True if a change is detected.
+
+    If verbose is more than 1, the before and after state of changed
+    items is also printed.
+    """
+
+    changed = False
+
+    def __init__(self, testname, verbose=0, quiet=False, *, pgo=False):
+        self.testname = testname
+        self.verbose = verbose
+        self.quiet = quiet
+        self.pgo = pgo
+
+    # To add things to save and restore, add a name XXX to the resources list
+    # and add corresponding get_XXX/restore_XXX functions.  get_XXX should
+    # return the value to be saved and compared against a second call to the
+    # get function when test execution completes.  restore_XXX should accept
+    # the saved value and restore the resource using it.  It will be called if
+    # and only if a change in the value is detected.
+    #
+    # Note: XXX will have any '.' replaced with '_' characters when determining
+    # the corresponding method names.
+
+    resources = ('sys.argv', 'cwd', 'sys.stdin', 'sys.stdout', 'sys.stderr',
+                 'os.environ', 'sys.path', 'sys.path_hooks', '__import__',
+                 'warnings.filters', 'asyncore.socket_map',
+                 'logging._handlers', 'logging._handlerList', 'sys.gettrace',
+                 'sys.warnoptions',
+                 # multiprocessing.process._cleanup() may release ref
+                 # to a thread, so check processes first.
+                 'multiprocessing.process._dangling', 'threading._dangling',
+                 'sysconfig._CONFIG_VARS', 'sysconfig._INSTALL_SCHEMES',
+                 'files', 'locale', 'warnings.showwarning',
+                 'shutil_archive_formats', 'shutil_unpack_formats',
+                )
+
+    def get_sys_argv(self):
+        return id(sys.argv), sys.argv, sys.argv[:]
+    def restore_sys_argv(self, saved_argv):
+        sys.argv = saved_argv[1]
+        sys.argv[:] = saved_argv[2]
+
+    def get_cwd(self):
+        return os.getcwd()
+    def restore_cwd(self, saved_cwd):
+        os.chdir(saved_cwd)
+
+    def get_sys_stdout(self):
+        return sys.stdout
+    def restore_sys_stdout(self, saved_stdout):
+        sys.stdout = saved_stdout
+
+    def get_sys_stderr(self):
+        return sys.stderr
+    def restore_sys_stderr(self, saved_stderr):
+        sys.stderr = saved_stderr
+
+    def get_sys_stdin(self):
+        return sys.stdin
+    def restore_sys_stdin(self, saved_stdin):
+        sys.stdin = saved_stdin
+
+    def get_os_environ(self):
+        return id(os.environ), os.environ, dict(os.environ)
+    def restore_os_environ(self, saved_environ):
+        os.environ = saved_environ[1]
+        os.environ.clear()
+        os.environ.update(saved_environ[2])
+
+    def get_sys_path(self):
+        return id(sys.path), sys.path, sys.path[:]
+    def restore_sys_path(self, saved_path):
+        sys.path = saved_path[1]
+        sys.path[:] = saved_path[2]
+
+    def get_sys_path_hooks(self):
+        return id(sys.path_hooks), sys.path_hooks, sys.path_hooks[:]
+    def restore_sys_path_hooks(self, saved_hooks):
+        sys.path_hooks = saved_hooks[1]
+        sys.path_hooks[:] = saved_hooks[2]
+
+    def get_sys_gettrace(self):
+        return sys.gettrace()
+    def restore_sys_gettrace(self, trace_fxn):
+        sys.settrace(trace_fxn)
+
+    def get___import__(self):
+        return builtins.__import__
+    def restore___import__(self, import_):
+        builtins.__import__ = import_
+
+    def get_warnings_filters(self):
+        return id(warnings.filters), warnings.filters, warnings.filters[:]
+    def restore_warnings_filters(self, saved_filters):
+        warnings.filters = saved_filters[1]
+        warnings.filters[:] = saved_filters[2]
+
+    def get_asyncore_socket_map(self):
+        asyncore = sys.modules.get('asyncore')
+        # XXX Making a copy keeps objects alive until __exit__ gets called.
+        return asyncore and asyncore.socket_map.copy() or {}
+    def restore_asyncore_socket_map(self, saved_map):
+        asyncore = sys.modules.get('asyncore')
+        if asyncore is not None:
+            asyncore.close_all(ignore_all=True)
+            asyncore.socket_map.update(saved_map)
+
+    def get_shutil_archive_formats(self):
+        # we could call get_archives_formats() but that only returns the
+        # registry keys; we want to check the values too (the functions that
+        # are registered)
+        return shutil._ARCHIVE_FORMATS, shutil._ARCHIVE_FORMATS.copy()
+    def restore_shutil_archive_formats(self, saved):
+        shutil._ARCHIVE_FORMATS = saved[0]
+        shutil._ARCHIVE_FORMATS.clear()
+        shutil._ARCHIVE_FORMATS.update(saved[1])
+
+    def get_shutil_unpack_formats(self):
+        return shutil._UNPACK_FORMATS, shutil._UNPACK_FORMATS.copy()
+    def restore_shutil_unpack_formats(self, saved):
+        shutil._UNPACK_FORMATS = saved[0]
+        shutil._UNPACK_FORMATS.clear()
+        shutil._UNPACK_FORMATS.update(saved[1])
+
+    def get_logging__handlers(self):
+        # _handlers is a WeakValueDictionary
+        return id(logging._handlers), logging._handlers, logging._handlers.copy()
+    def restore_logging__handlers(self, saved_handlers):
+        # Can't easily revert the logging state
+        pass
+
+    def get_logging__handlerList(self):
+        # _handlerList is a list of weakrefs to handlers
+        return id(logging._handlerList), logging._handlerList, logging._handlerList[:]
+    def restore_logging__handlerList(self, saved_handlerList):
+        # Can't easily revert the logging state
+        pass
+
+    def get_sys_warnoptions(self):
+        return id(sys.warnoptions), sys.warnoptions, sys.warnoptions[:]
+    def restore_sys_warnoptions(self, saved_options):
+        sys.warnoptions = saved_options[1]
+        sys.warnoptions[:] = saved_options[2]
+
+    # Controlling dangling references to Thread objects can make it easier
+    # to track reference leaks.
+    def get_threading__dangling(self):
+        if not threading:
+            return None
+        # This copies the weakrefs without making any strong reference
+        return threading._dangling.copy()
+    def restore_threading__dangling(self, saved):
+        if not threading:
+            return
+        threading._dangling.clear()
+        threading._dangling.update(saved)
+
+    # Same for Process objects
+    def get_multiprocessing_process__dangling(self):
+        if not multiprocessing:
+            return None
+        # Unjoined process objects can survive after process exits
+        multiprocessing.process._cleanup()
+        # This copies the weakrefs without making any strong reference
+        return multiprocessing.process._dangling.copy()
+    def restore_multiprocessing_process__dangling(self, saved):
+        if not multiprocessing:
+            return
+        multiprocessing.process._dangling.clear()
+        multiprocessing.process._dangling.update(saved)
+
+    def get_sysconfig__CONFIG_VARS(self):
+        # make sure the dict is initialized
+        sysconfig.get_config_var('prefix')
+        return (id(sysconfig._CONFIG_VARS), sysconfig._CONFIG_VARS,
+                dict(sysconfig._CONFIG_VARS))
+    def restore_sysconfig__CONFIG_VARS(self, saved):
+        sysconfig._CONFIG_VARS = saved[1]
+        sysconfig._CONFIG_VARS.clear()
+        sysconfig._CONFIG_VARS.update(saved[2])
+
+    def get_sysconfig__INSTALL_SCHEMES(self):
+        return (id(sysconfig._INSTALL_SCHEMES), sysconfig._INSTALL_SCHEMES,
+                sysconfig._INSTALL_SCHEMES.copy())
+    def restore_sysconfig__INSTALL_SCHEMES(self, saved):
+        sysconfig._INSTALL_SCHEMES = saved[1]
+        sysconfig._INSTALL_SCHEMES.clear()
+        sysconfig._INSTALL_SCHEMES.update(saved[2])
+
+    def get_files(self):
+        return sorted(fn + ('/' if os.path.isdir(fn) else '')
+                      for fn in os.listdir())
+    def restore_files(self, saved_value):
+        fn = support.TESTFN
+        if fn not in saved_value and (fn + '/') not in saved_value:
+            if os.path.isfile(fn):
+                support.unlink(fn)
+            elif os.path.isdir(fn):
+                support.rmtree(fn)
+
+    _lc = [getattr(locale, lc) for lc in dir(locale)
+           if lc.startswith('LC_')]
+    def get_locale(self):
+        pairings = []
+        for lc in self._lc:
+            try:
+                pairings.append((lc, locale.setlocale(lc, None)))
+            except (TypeError, ValueError):
+                continue
+        return pairings
+    def restore_locale(self, saved):
+        for lc, setting in saved:
+            locale.setlocale(lc, setting)
+
+    def get_warnings_showwarning(self):
+        return warnings.showwarning
+    def restore_warnings_showwarning(self, fxn):
+        warnings.showwarning = fxn
+
+    def resource_info(self):
+        for name in self.resources:
+            method_suffix = name.replace('.', '_')
+            get_name = 'get_' + method_suffix
+            restore_name = 'restore_' + method_suffix
+            yield name, getattr(self, get_name), getattr(self, restore_name)
+
+    def __enter__(self):
+        self.saved_values = dict((name, get()) for name, get, restore
+                                                   in self.resource_info())
+        return self
+
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        saved_values = self.saved_values
+        del self.saved_values
+        support.gc_collect()  # Some resources use weak references
+        for name, get, restore in self.resource_info():
+            current = get()
+            original = saved_values.pop(name)
+            # Check for changes to the resource's value
+            if current != original:
+                self.changed = True
+                restore(original)
+                if not self.quiet and not self.pgo:
+                    print(f"Warning -- {name} was modified by {self.testname}",
+                          file=sys.stderr, flush=True)
+                    if self.verbose > 1:
+                        print(f"  Before: {original}\n  After:  {current} ",
+                              file=sys.stderr, flush=True)
+        return False
diff --git a/Lib/test/libregrtest/setup.py b/Lib/test/libregrtest/setup.py
new file mode 100644
index 0000000000..1d24531fc2
--- /dev/null
+++ b/Lib/test/libregrtest/setup.py
@@ -0,0 +1,121 @@
+import atexit
+import faulthandler
+import os
+import signal
+import sys
+import unittest
+from test import support
+try:
+    import gc
+except ImportError:
+    gc = None
+
+from test.libregrtest.refleak import warm_caches
+
+
+def setup_tests(ns):
+    # Display the Python traceback on fatal errors (e.g. segfault)
+    faulthandler.enable(all_threads=True)
+
+    # Display the Python traceback on SIGALRM or SIGUSR1 signal
+    signals = []
+    if hasattr(signal, 'SIGALRM'):
+        signals.append(signal.SIGALRM)
+    if hasattr(signal, 'SIGUSR1'):
+        signals.append(signal.SIGUSR1)
+    for signum in signals:
+        faulthandler.register(signum, chain=True)
+
+    replace_stdout()
+    support.record_original_stdout(sys.stdout)
+
+    if ns.testdir:
+        # Prepend test directory to sys.path, so runtest() will be able
+        # to locate tests
+        sys.path.insert(0, os.path.abspath(ns.testdir))
+
+    # Some times __path__ and __file__ are not absolute (e.g. while running from
+    # Lib/) and, if we change the CWD to run the tests in a temporary dir, some
+    # imports might fail.  This affects only the modules imported before os.chdir().
+    # These modules are searched first in sys.path[0] (so '' -- the CWD) and if
+    # they are found in the CWD their __file__ and __path__ will be relative (this
+    # happens before the chdir).  All the modules imported after the chdir, are
+    # not found in the CWD, and since the other paths in sys.path[1:] are absolute
+    # (site.py absolutize them), the __file__ and __path__ will be absolute too.
+    # Therefore it is necessary to absolutize manually the __file__ and __path__ of
+    # the packages to prevent later imports to fail when the CWD is different.
+    for module in sys.modules.values():
+        if hasattr(module, '__path__'):
+            for index, path in enumerate(module.__path__):
+                module.__path__[index] = os.path.abspath(path)
+        if hasattr(module, '__file__'):
+            module.__file__ = os.path.abspath(module.__file__)
+
+    # MacOSX (a.k.a. Darwin) has a default stack size that is too small
+    # for deeply recursive regular expressions.  We see this as crashes in
+    # the Python test suite when running test_re.py and test_sre.py.  The
+    # fix is to set the stack limit to 2048.
+    # This approach may also be useful for other Unixy platforms that
+    # suffer from small default stack limits.
+    if sys.platform == 'darwin':
+        try:
+            import resource
+        except ImportError:
+            pass
+        else:
+            soft, hard = resource.getrlimit(resource.RLIMIT_STACK)
+            newsoft = min(hard, max(soft, 1024*2048))
+            resource.setrlimit(resource.RLIMIT_STACK, (newsoft, hard))
+
+    if ns.huntrleaks:
+        unittest.BaseTestSuite._cleanup = False
+
+        # Avoid false positives due to various caches
+        # filling slowly with random data:
+        warm_caches()
+
+    if ns.memlimit is not None:
+        support.set_memlimit(ns.memlimit)
+
+    if ns.threshold is not None:
+        gc.set_threshold(ns.threshold)
+
+    try:
+        import msvcrt
+    except ImportError:
+        pass
+    else:
+        msvcrt.SetErrorMode(msvcrt.SEM_FAILCRITICALERRORS|
+                            msvcrt.SEM_NOALIGNMENTFAULTEXCEPT|
+                            msvcrt.SEM_NOGPFAULTERRORBOX|
+                            msvcrt.SEM_NOOPENFILEERRORBOX)
+        try:
+            msvcrt.CrtSetReportMode
+        except AttributeError:
+            # release build
+            pass
+        else:
+            for m in [msvcrt.CRT_WARN, msvcrt.CRT_ERROR, msvcrt.CRT_ASSERT]:
+                if ns.verbose and ns.verbose >= 2:
+                    msvcrt.CrtSetReportMode(m, msvcrt.CRTDBG_MODE_FILE)
+                    msvcrt.CrtSetReportFile(m, msvcrt.CRTDBG_FILE_STDERR)
+                else:
+                    msvcrt.CrtSetReportMode(m, 0)
+
+    support.use_resources = ns.use_resources
+
+
+def replace_stdout():
+    """Set stdout encoder error handler to backslashreplace (as stderr error
+    handler) to avoid UnicodeEncodeError when printing a traceback"""
+    stdout = sys.stdout
+    sys.stdout = open(stdout.fileno(), 'w',
+        encoding=stdout.encoding,
+        errors="backslashreplace",
+        closefd=False,
+        newline='\n')
+
+    def restore_stdout():
+        sys.stdout.close()
+        sys.stdout = stdout
+    atexit.register(restore_stdout)