1
1
mirror of https://github.com/kanaka/mal.git synced 2024-08-16 17:20:23 +03:00
mal/runtest.py
Ben Harris 9d7fb0fd8d runtest: Treat test timeouts as fatal
If a test times out, that means it hasn't returned a prompt within
reasonable time.  Continuing to type at it is unlikely to be helpful,
since the lack of a prompt indicates that the subprocess has hung, so
subsequent tests are likely to time out as well.

This is particularly annoying in step 5, where the line that usually
times out, "(def! res2 (sum2 10000 0))" has its output ignored, so a
timeout is treated as success, meaning that runtest proceeds to the next
test, which of course also times out.
2020-06-01 15:08:10 +01:00

351 lines
12 KiB
Python
Executable File

#!/usr/bin/env python
from __future__ import print_function
import os, sys, re
import argparse, time
import signal, atexit
from subprocess import Popen, STDOUT, PIPE
from select import select
# Pseudo-TTY and terminal manipulation
import pty, array, fcntl, termios
IS_PY_3 = sys.version_info[0] == 3
debug_file = None
log_file = None
def debug(data):
if debug_file:
debug_file.write(data)
debug_file.flush()
def log(data, end='\n'):
if log_file:
log_file.write(data + end)
log_file.flush()
print(data, end=end)
sys.stdout.flush()
sep = "\n"
rundir = None
parser = argparse.ArgumentParser(
description="Run a test file against a Mal implementation")
parser.add_argument('--rundir',
help="change to the directory before running tests")
parser.add_argument('--start-timeout', default=10, type=int,
help="default timeout for initial prompt")
parser.add_argument('--test-timeout', default=20, type=int,
help="default timeout for each individual test action")
parser.add_argument('--pre-eval', default=None, type=str,
help="Mal code to evaluate prior to running the test")
parser.add_argument('--no-pty', action='store_true',
help="Use direct pipes instead of pseudo-tty")
parser.add_argument('--log-file', type=str,
help="Write messages to the named file in addition the screen")
parser.add_argument('--debug-file', type=str,
help="Write all test interaction the named file")
parser.add_argument('--hard', action='store_true',
help="Turn soft tests (soft, deferrable, optional) into hard failures")
# Control whether deferrable and optional tests are executed
parser.add_argument('--deferrable', dest='deferrable', action='store_true',
help="Enable deferrable tests that follow a ';>>> deferrable=True'")
parser.add_argument('--no-deferrable', dest='deferrable', action='store_false',
help="Disable deferrable tests that follow a ';>>> deferrable=True'")
parser.set_defaults(deferrable=True)
parser.add_argument('--optional', dest='optional', action='store_true',
help="Enable optional tests that follow a ';>>> optional=True'")
parser.add_argument('--no-optional', dest='optional', action='store_false',
help="Disable optional tests that follow a ';>>> optional=True'")
parser.set_defaults(optional=True)
parser.add_argument('test_file', type=str,
help="a test file formatted as with mal test data")
parser.add_argument('mal_cmd', nargs="*",
help="Mal implementation command line. Use '--' to "
"specify a Mal command line with dashed options.")
parser.add_argument('--crlf', dest='crlf', action='store_true',
help="Write \\r\\n instead of \\n to the input")
class Runner():
def __init__(self, args, no_pty=False, line_break="\n"):
#print "args: %s" % repr(args)
self.no_pty = no_pty
# Cleanup child process on exit
atexit.register(self.cleanup)
self.p = None
env = os.environ
env['TERM'] = 'dumb'
env['INPUTRC'] = '/dev/null'
env['PERL_RL'] = 'false'
if no_pty:
self.p = Popen(args, bufsize=0,
stdin=PIPE, stdout=PIPE, stderr=STDOUT,
preexec_fn=os.setsid,
env=env)
self.stdin = self.p.stdin
self.stdout = self.p.stdout
else:
# provide tty to get 'interactive' readline to work
master, slave = pty.openpty()
# Set terminal size large so that readline will not send
# ANSI/VT escape codes when the lines are long.
buf = array.array('h', [100, 200, 0, 0])
fcntl.ioctl(master, termios.TIOCSWINSZ, buf, True)
self.p = Popen(args, bufsize=0,
stdin=slave, stdout=slave, stderr=STDOUT,
preexec_fn=os.setsid,
env=env)
# Now close slave so that we will get an exception from
# read when the child exits early
# http://stackoverflow.com/questions/11165521
os.close(slave)
self.stdin = os.fdopen(master, 'r+b', 0)
self.stdout = self.stdin
#print "started"
self.buf = ""
self.last_prompt = ""
self.line_break = line_break
def read_to_prompt(self, prompts, timeout):
end_time = time.time() + timeout
while time.time() < end_time:
[outs,_,_] = select([self.stdout], [], [], 1)
if self.stdout in outs:
new_data = self.stdout.read(1)
new_data = new_data.decode("utf-8") if IS_PY_3 else new_data
#print("new_data: '%s'" % new_data)
debug(new_data)
# Perform newline cleanup
self.buf += new_data.replace("\r", "")
for prompt in prompts:
regexp = re.compile(prompt)
match = regexp.search(self.buf)
if match:
end = match.end()
buf = self.buf[0:match.start()]
self.buf = self.buf[end:]
self.last_prompt = prompt
return buf
return None
def writeline(self, str):
def _to_bytes(s):
return bytes(s, "utf-8") if IS_PY_3 else s
self.stdin.write(_to_bytes(str.replace('\r', '\x16\r') + self.line_break))
def cleanup(self):
#print "cleaning up"
if self.p:
try:
os.killpg(self.p.pid, signal.SIGTERM)
except OSError:
pass
self.p = None
class TestReader:
def __init__(self, test_file):
self.line_num = 0
f = open(test_file, newline='') if IS_PY_3 else open(test_file)
self.data = f.read().split('\n')
self.soft = False
self.deferrable = False
self.optional = False
def next(self):
self.msg = None
self.form = None
self.out = ""
self.ret = None
while self.data:
self.line_num += 1
line = self.data.pop(0)
if re.match(r"^\s*$", line): # blank line
continue
elif line[0:3] == ";;;": # ignore comment
continue
elif line[0:2] == ";;": # output comment
self.msg = line[3:]
return True
elif line[0:5] == ";>>> ": # settings/commands
settings = {}
exec(line[5:], {}, settings)
if 'soft' in settings:
self.soft = settings['soft']
if 'deferrable' in settings and settings['deferrable']:
self.deferrable = "\nSkipping deferrable and optional tests"
return True
if 'optional' in settings and settings['optional']:
self.optional = "\nSkipping optional tests"
return True
continue
elif line[0:1] == ";": # unexpected comment
raise Exception("Test data error at line %d:\n%s" % (self.line_num, line))
self.form = line # the line is a form to send
# Now find the output and return value
while self.data:
line = self.data[0]
if line[0:3] == ";=>":
self.ret = line[3:]
self.line_num += 1
self.data.pop(0)
break
elif line[0:2] == ";/":
self.out = self.out + line[2:] + sep
self.line_num += 1
self.data.pop(0)
else:
self.ret = ""
break
if self.ret != None: break
if self.out[-1:] == sep and not self.ret:
# If there is no return value, output should not end in
# separator
self.out = self.out[0:-1]
return self.form
args = parser.parse_args(sys.argv[1:])
# Workaround argparse issue with two '--' on command line
if sys.argv.count('--') > 0:
args.mal_cmd = sys.argv[sys.argv.index('--')+1:]
if args.rundir: os.chdir(args.rundir)
if args.log_file: log_file = open(args.log_file, "a")
if args.debug_file: debug_file = open(args.debug_file, "a")
r = Runner(args.mal_cmd, no_pty=args.no_pty, line_break="\r\n" if args.crlf else "\n")
t = TestReader(args.test_file)
def assert_prompt(runner, prompts, timeout):
# Wait for the initial prompt
header = runner.read_to_prompt(prompts, timeout=timeout)
if not header == None:
if header:
log("Started with:\n%s" % header)
else:
log("Did not receive one of following prompt(s): %s" % repr(prompts))
log(" Got : %s" % repr(r.buf))
sys.exit(1)
# Wait for the initial prompt
try:
assert_prompt(r, ['[^\s()<>]+> '], args.start_timeout)
except:
_, exc, _ = sys.exc_info()
log("\nException: %s" % repr(exc))
log("Output before exception:\n%s" % r.buf)
sys.exit(1)
# Send the pre-eval code if any
if args.pre_eval:
sys.stdout.write("RUNNING pre-eval: %s" % args.pre_eval)
r.writeline(args.pre_eval)
assert_prompt(r, ['[^\s()<>]+> '], args.test_timeout)
test_cnt = 0
pass_cnt = 0
fail_cnt = 0
soft_fail_cnt = 0
failures = []
class TestTimeout(Exception):
pass
while t.next():
if args.deferrable == False and t.deferrable:
log(t.deferrable)
break
if args.optional == False and t.optional:
log(t.optional)
break
if t.msg != None:
log(t.msg)
continue
if t.form == None: continue
log("TEST: %s -> [%s,%s]" % (repr(t.form), repr(t.out), t.ret), end='')
# The repeated form is to get around an occasional OS X issue
# where the form is repeated.
# https://github.com/kanaka/mal/issues/30
expects = [".*%s%s%s" % (sep, t.out, re.escape(t.ret)),
".*%s.*%s%s%s" % (sep, sep, t.out, re.escape(t.ret))]
r.writeline(t.form)
try:
test_cnt += 1
res = r.read_to_prompt(['\r\n[^\s()<>]+> ', '\n[^\s()<>]+> '],
timeout=args.test_timeout)
#print "%s,%s,%s" % (idx, repr(p.before), repr(p.after))
if (res == None):
log(" -> TIMEOUT (line %d)" % t.line_num)
raise TestTimeout("TIMEOUT (line %d)" % t.line_num)
elif (t.ret == "" and t.out == ""):
log(" -> SUCCESS (result ignored)")
pass_cnt += 1
elif (re.search(expects[0], res, re.S) or
re.search(expects[1], res, re.S)):
log(" -> SUCCESS")
pass_cnt += 1
else:
if t.soft and not args.hard:
log(" -> SOFT FAIL (line %d):" % t.line_num)
soft_fail_cnt += 1
fail_type = "SOFT "
else:
log(" -> FAIL (line %d):" % t.line_num)
fail_cnt += 1
fail_type = ""
log(" Expected : %s" % repr(expects[0]))
log(" Got : %s" % repr(res))
failed_test = """%sFAILED TEST (line %d): %s -> [%s,%s]:
Expected : %s
Got : %s""" % (fail_type, t.line_num, t.form, repr(t.out),
t.ret, repr(expects[0]), repr(res))
failures.append(failed_test)
except:
_, exc, _ = sys.exc_info()
log("\nException: %s" % repr(exc))
log("Output before exception:\n%s" % r.buf)
sys.exit(1)
if len(failures) > 0:
log("\nFAILURES:")
for f in failures:
log(f)
results = """
TEST RESULTS (for %s):
%3d: soft failing tests
%3d: failing tests
%3d: passing tests
%3d: total tests
""" % (args.test_file, soft_fail_cnt, fail_cnt,
pass_cnt, test_cnt)
log(results)
debug("\n") # add some separate to debug log
if fail_cnt > 0:
sys.exit(1)
sys.exit(0)