def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
if len(self.parameters) <= 0:
num_processes = 1
else:
num_processes = self.parameters[0]
indices = range(num_processes)
# Create sandboxes and FIFOs
sandbox_mgr = create_sandbox(file_cacher)
sandbox_user = [create_sandbox(file_cacher) for i in indices]
fifo_dir = [tempfile.mkdtemp(dir=config.temp_dir) for i in indices]
fifo_in = [os.path.join(fifo_dir[i], "in%d" % i) for i in indices]
fifo_out = [os.path.join(fifo_dir[i], "out%d" % i) for i in indices]
for i in indices:
os.mkfifo(fifo_in[i])
os.mkfifo(fifo_out[i])
os.chmod(fifo_dir[i], 0o755)
os.chmod(fifo_in[i], 0o666)
os.chmod(fifo_out[i], 0o666)
# First step: we start the manager.
manager_filename = "manager"
manager_command = ["./%s" % manager_filename]
for i in indices:
manager_command.append(fifo_in[i])
manager_command.append(fifo_out[i])
manager_executables_to_get = {
manager_filename:
job.managers[manager_filename].digest
}
manager_files_to_get = {
"input.txt": job.input
}
manager_allow_dirs = fifo_dir
for filename, digest in manager_executables_to_get.iteritems():
sandbox_mgr.create_file_from_storage(
filename, digest, executable=True)
for filename, digest in manager_files_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename, digest)
manager = evaluation_step_before_run(
sandbox_mgr,
manager_command,
num_processes * job.time_limit,
0,
allow_dirs=manager_allow_dirs,
writable_files=["output.txt"],
stdin_redirect="input.txt")
# Second step: we start the user submission compiled with the
# stub.
executable_filename = job.executables.keys()[0]
executables_to_get = {
executable_filename:
job.executables[executable_filename].digest
}
processes = [None for i in indices]
for i in indices:
command = ["./%s" % executable_filename, fifo_out[i], fifo_in[i]]
if num_processes != 1:
command.append(str(i))
user_allow_dirs = [fifo_dir[i]]
for filename, digest in executables_to_get.iteritems():
sandbox_user[i].create_file_from_storage(
filename, digest, executable=True)
processes[i] = evaluation_step_before_run(
sandbox_user[i],
command,
job.time_limit,
job.memory_limit,
allow_dirs=user_allow_dirs)
# Consume output.
wait_without_std(processes + [manager])
# TODO: check exit codes with translate_box_exitcode.
user_results = [evaluation_step_after_run(s) for s in sandbox_user]
success_user = all(r[0] for r in user_results)
plus_user = reduce(merge_evaluation_results,
[r[1] for r in user_results])
success_mgr, unused_plus_mgr = \
evaluation_step_after_run(sandbox_mgr)
if plus_user['exit_status'] == Sandbox.EXIT_OK and \
plus_user["execution_time"] >= job.time_limit:
plus_user['exit_status'] = Sandbox.EXIT_TIMEOUT
# Merge results.
job.sandboxes = [s.path for s in sandbox_user] + [sandbox_mgr.path]
job.plus = plus_user
# If at least one evaluation had problems, we report the
# problems.
if not success_user or not success_mgr:
success, outcome, text = False, None, None
# If the user sandbox detected some problem (timeout, ...),
# the outcome is 0.0 and the text describes that problem.
elif not is_evaluation_passed(plus_user):
success = True
outcome, text = 0.0, human_evaluation_message(plus_user)
# Otherwise, we use the manager to obtain the outcome.
else:
success = True
outcome, text = extract_outcome_and_text(sandbox_mgr)
# If asked so, save the output file, provided that it exists
if job.get_output:
if sandbox_mgr.file_exists("output.txt"):
job.user_output = sandbox_mgr.get_file_to_storage(
"output.txt",
"Output file in job %s" % job.info)
else:
job.user_output = None
# Whatever happened, we conclude.
job.success = success
job.outcome = "%s" % outcome if outcome is not None else None
job.text = text
delete_sandbox(sandbox_mgr)
for s in sandbox_user:
delete_sandbox(s)
if not config.keep_sandbox:
for d in fifo_dir:
rmtree(d)
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
# Create the sandbox
sandbox = create_sandbox(file_cacher, job.multithreaded_sandbox)
# Prepare the execution
assert len(job.executables) == 1
executable_filename = next(iterkeys(job.executables))
language = get_language(job.language)
commands = language.get_evaluation_commands(
executable_filename,
main="grader" if self._uses_grader() else executable_filename)
executables_to_get = {
executable_filename: job.executables[executable_filename].digest
}
input_filename, output_filename = self.parameters[1]
stdin_redirect = None
stdout_redirect = None
files_allowing_write = []
if len(input_filename) == 0:
input_filename = "input.txt"
stdin_redirect = input_filename
if len(output_filename) == 0:
output_filename = "output.txt"
stdout_redirect = output_filename
else:
files_allowing_write.append(output_filename)
files_to_get = {input_filename: job.input}
# Put the required files into the sandbox
for filename, digest in iteritems(executables_to_get):
sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in iteritems(files_to_get):
sandbox.create_file_from_storage(filename, digest)
# Actually performs the execution
success, plus = evaluation_step(sandbox,
commands,
job.time_limit,
job.memory_limit,
writable_files=files_allowing_write,
stdin_redirect=stdin_redirect,
stdout_redirect=stdout_redirect)
job.sandboxes = [sandbox.path]
job.plus = plus
outcome = None
text = []
# Error in the sandbox: nothing to do!
if not success:
pass
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(plus):
outcome = 0.0
text = human_evaluation_message(plus)
if job.get_output:
job.user_output = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not sandbox.file_exists(output_filename):
outcome = 0.0
text = [
N_("Evaluation didn't produce file %s"), output_filename
]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage
if job.get_output:
job.user_output = sandbox.get_file_to_storage(
output_filename,
"Output file in job %s" % job.info,
trunc_len=100 * 1024)
# If just asked to execute, fill text and set dummy
# outcome.
if job.only_execution:
outcome = 0.0
text = [N_("Execution completed successfully")]
# Otherwise evaluate the output file.
else:
# Put the reference solution into the sandbox
sandbox.create_file_from_storage("res.txt", job.output)
# Check the solution with white_diff
if self.parameters[2] == "diff":
outcome, text = white_diff_step(
sandbox, output_filename, "res.txt")
# Check the solution with a comparator
elif self.parameters[2] == "comparator":
manager_filename = "checker"
if manager_filename not in job.managers:
logger.error(
"Configuration error: missing or "
"invalid comparator (it must be "
"named 'checker')",
extra={"operation": job.info})
success = False
else:
sandbox.create_file_from_storage(
manager_filename,
job.managers[manager_filename].digest,
executable=True)
# Rewrite input file. The untrusted
# contestant program should not be able to
# modify it; however, the grader may
# destroy the input file to prevent the
# contestant's program from directly
# accessing it. Since we cannot create
# files already existing in the sandbox,
# we try removing the file first.
try:
sandbox.remove_file(input_filename)
except OSError as e:
# Let us be extra sure that the file
# was actually removed and we did not
# mess up with permissions.
assert not sandbox.file_exists(input_filename)
sandbox.create_file_from_storage(
input_filename, job.input)
# Allow using any number of processes (because e.g.
# one may want to write a bash checker who calls
# other processes). Set to a high number because
# to avoid fork-bombing the worker.
sandbox.max_processes = 1000
success, _ = evaluation_step(
sandbox, [[
"./%s" % manager_filename, input_filename,
"res.txt", output_filename
]])
if success:
try:
outcome, text = \
extract_outcome_and_text(sandbox)
except ValueError as e:
logger.error(
"Invalid output from "
"comparator: %s",
e.message,
extra={"operation": job.info})
success = False
else:
raise ValueError("Unrecognized third parameter"
" `%s' for Batch tasktype." %
self.parameters[2])
# Whatever happened, we conclude.
job.success = success
job.outcome = "%s" % outcome if outcome is not None else None
job.text = text
delete_sandbox(sandbox, job.success)
def evaluate_testcase(self, test_number):
"""See TaskType.evaluate_testcase."""
# f stand for first, s for second.
first_sandbox = create_sandbox(self)
second_sandbox = create_sandbox(self)
fifo_dir = tempfile.mkdtemp(dir=config.temp_dir)
fifo = os.path.join(fifo_dir, "fifo")
os.mkfifo(fifo)
# First step: we start the first manager.
first_filename = "manager"
first_command = ["./%s" % first_filename, "0", fifo]
first_executables_to_get = {
first_filename:
self.job.executables[first_filename].digest
}
first_files_to_get = {
"input.txt": self.job.testcases[test_number].input
}
first_allow_path = ["input.txt", fifo]
# Put the required files into the sandbox
for filename, digest in first_executables_to_get.iteritems():
first_sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in first_files_to_get.iteritems():
first_sandbox.create_file_from_storage(filename, digest)
first = evaluation_step_before_run(
first_sandbox,
first_command,
self.job.time_limit,
self.job.memory_limit,
first_allow_path,
stdin_redirect="input.txt",
wait=False)
# Second step: we start the second manager.
second_filename = "manager"
second_command = ["./%s" % second_filename, "1", fifo]
second_executables_to_get = {
second_filename:
self.job.executables[second_filename].digest
}
second_files_to_get = {}
second_allow_path = [fifo, "output.txt"]
# Put the required files into the second sandbox
for filename, digest in second_executables_to_get.iteritems():
second_sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in second_files_to_get.iteritems():
second_sandbox.create_file_from_storage(filename, digest)
second = evaluation_step_before_run(
second_sandbox,
second_command,
self.job.time_limit,
self.job.memory_limit,
second_allow_path,
stdout_redirect="output.txt",
wait=False)
# Consume output.
wait_without_std([second, first])
# TODO: check exit codes with translate_box_exitcode.
success_first, first_plus = \
evaluation_step_after_run(first_sandbox)
success_second, second_plus = \
evaluation_step_after_run(second_sandbox)
self.job.evaluations[test_number] = {'sandboxes': [first_sandbox.path,
second_sandbox.path],
'plus': second_plus}
outcome = None
text = None
evaluation = self.job.evaluations[test_number]
success = True
# Error in the sandbox: report failure!
if not success_first or not success_second:
success = False
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(first_plus) or \
not is_evaluation_passed(second_plus):
outcome = 0.0
if not is_evaluation_passed(first_plus):
text = human_evaluation_message(first_plus)
else:
text = human_evaluation_message(second_plus)
if self.job.get_output:
evaluation['output'] = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not second_sandbox.file_exists('output.txt'):
outcome = 0.0
text = "Execution didn't produce file output.txt"
if self.job.get_output:
evaluation['output'] = None
else:
# If asked so, put the output file into the storage
if self.job.get_output:
evaluation['output'] = second_sandbox.get_file_to_storage(
"output.txt",
"Output file for testcase %d in job %s" %
(test_number, self.job.info))
# If not asked otherwise, evaluate the output file
if not self.job.only_execution:
# Put the reference solution into the sandbox
second_sandbox.create_file_from_storage(
"res.txt",
self.job.testcases[test_number].output)
outcome, text = white_diff_step(
second_sandbox, "output.txt", "res.txt")
# Whatever happened, we conclude.
evaluation['success'] = success
evaluation['outcome'] = outcome
evaluation['text'] = text
delete_sandbox(first_sandbox)
delete_sandbox(second_sandbox)
return success
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
# Create sandboxes and FIFOs
sandbox_mgr = create_sandbox(file_cacher)
sandbox_user = create_sandbox(file_cacher)
fifo_dir = tempfile.mkdtemp(dir=config.temp_dir)
fifo_in = os.path.join(fifo_dir, "in")
fifo_out = os.path.join(fifo_dir, "out")
os.mkfifo(fifo_in)
os.mkfifo(fifo_out)
os.chmod(fifo_dir, 0o755)
os.chmod(fifo_in, 0o666)
os.chmod(fifo_out, 0o666)
# First step: we start the manager.
manager_filename = "manager"
manager_command = ["./%s" % manager_filename, fifo_in, fifo_out]
manager_executables_to_get = {
manager_filename:
job.managers[manager_filename].digest
}
manager_files_to_get = {
"input.txt": job.input
}
manager_allow_dirs = [fifo_dir]
for filename, digest in manager_executables_to_get.iteritems():
sandbox_mgr.create_file_from_storage(
filename, digest, executable=True)
for filename, digest in manager_files_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename, digest)
manager = evaluation_step_before_run(
sandbox_mgr,
manager_command,
job.time_limit,
0,
allow_dirs=manager_allow_dirs,
stdin_redirect="input.txt")
# Second step: we start the user submission compiled with the
# stub.
executable_filename = job.executables.keys()[0]
command = ["./%s" % executable_filename, fifo_out, fifo_in]
executables_to_get = {
executable_filename:
job.executables[executable_filename].digest
}
user_allow_dirs = [fifo_dir]
for filename, digest in executables_to_get.iteritems():
sandbox_user.create_file_from_storage(
filename, digest, executable=True)
process = evaluation_step_before_run(
sandbox_user,
command,
job.time_limit,
job.memory_limit,
allow_dirs=user_allow_dirs)
# Consume output.
wait_without_std([process, manager])
# TODO: check exit codes with translate_box_exitcode.
success_user, plus_user = \
evaluation_step_after_run(sandbox_user)
success_mgr, plus_mgr = \
evaluation_step_after_run(sandbox_mgr)
job.sandboxes = [sandbox_user.path,
sandbox_mgr.path]
job.plus = plus_user
# If at least one evaluation had problems, we report the
# problems.
if not success_user or not success_mgr:
success, outcome, text = False, None, None
# If the user sandbox detected some problem (timeout, ...),
# the outcome is 0.0 and the text describes that problem.
elif not is_evaluation_passed(plus_user):
success = True
outcome, text = 0.0, human_evaluation_message(plus_user)
# Otherwise, we use the manager to obtain the outcome.
else:
success = True
outcome, text = extract_outcome_and_text(sandbox_mgr)
# If asked so, save the output file, provided that it exists
if job.get_output:
if sandbox_mgr.file_exists("output.txt"):
job.user_output = sandbox_mgr.get_file_to_storage(
"output.txt",
"Output file in job %s" % job.info)
else:
job.user_output = None
# Whatever happened, we conclude.
job.success = success
job.outcome = str(outcome) if outcome is not None else None
job.text = text
delete_sandbox(sandbox_mgr)
delete_sandbox(sandbox_user)
shutil.rmtree(fifo_dir)
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
# Create sandboxes and FIFOs
sandbox_mgr = create_sandbox(file_cacher)
sandbox_user = create_sandbox(file_cacher)
fifo_dir = tempfile.mkdtemp(dir=config.temp_dir)
fifo_in = os.path.join(fifo_dir, "in")
fifo_out = os.path.join(fifo_dir, "out")
os.mkfifo(fifo_in)
os.mkfifo(fifo_out)
os.chmod(fifo_dir, 0o755)
os.chmod(fifo_in, 0o666)
os.chmod(fifo_out, 0o666)
# First step: we start the manager.
manager_filename = "manager"
manager_command = ["./%s" % manager_filename, fifo_in, fifo_out]
manager_executables_to_get = {
manager_filename: job.managers[manager_filename].digest
}
manager_files_to_get = {"input.txt": job.input}
manager_allow_dirs = [fifo_dir]
for filename, digest in manager_executables_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in manager_files_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename, digest)
manager = evaluation_step_before_run(sandbox_mgr,
manager_command,
job.time_limit,
0,
allow_dirs=manager_allow_dirs,
writable_files=["output.txt"],
stdin_redirect="input.txt")
# Second step: we start the user submission compiled with the
# stub.
executable_filename = job.executables.keys()[0]
command = ["./%s" % executable_filename, fifo_out, fifo_in]
executables_to_get = {
executable_filename: job.executables[executable_filename].digest
}
user_allow_dirs = [fifo_dir]
for filename, digest in executables_to_get.iteritems():
sandbox_user.create_file_from_storage(filename,
digest,
executable=True)
process = evaluation_step_before_run(sandbox_user,
command,
job.time_limit,
job.memory_limit,
allow_dirs=user_allow_dirs)
# Consume output.
wait_without_std([process, manager])
# TODO: check exit codes with translate_box_exitcode.
success_user, plus_user = \
evaluation_step_after_run(sandbox_user)
success_mgr, unused_plus_mgr = \
evaluation_step_after_run(sandbox_mgr)
job.sandboxes = [sandbox_user.path, sandbox_mgr.path]
job.plus = plus_user
# If at least one evaluation had problems, we report the
# problems.
if not success_user or not success_mgr:
success, outcome, text = False, None, None
# If the user sandbox detected some problem (timeout, ...),
# the outcome is 0.0 and the text describes that problem.
elif not is_evaluation_passed(plus_user):
success = True
outcome, text = 0.0, human_evaluation_message(plus_user)
# Otherwise, we use the manager to obtain the outcome.
else:
success = True
outcome, text = extract_outcome_and_text(sandbox_mgr)
# If asked so, save the output file, provided that it exists
if job.get_output:
if sandbox_mgr.file_exists("output.txt"):
job.user_output = sandbox_mgr.get_file_to_storage(
"output.txt", "Output file in job %s" % job.info)
else:
job.user_output = None
# Whatever happened, we conclude.
job.success = success
job.outcome = "%s" % outcome if outcome is not None else None
job.text = text
delete_sandbox(sandbox_mgr)
delete_sandbox(sandbox_user)
if not config.keep_sandbox:
rmtree(fifo_dir)
def evaluate_testcase(self, test_number):
"""See TaskType.evaluate_testcase."""
# Create sandboxes and FIFOs
sandbox_mgr = create_sandbox(self)
sandbox_user = create_sandbox(self)
fifo_dir = tempfile.mkdtemp(dir=config.temp_dir)
fifo_in = os.path.join(fifo_dir, "in")
fifo_out = os.path.join(fifo_dir, "out")
os.mkfifo(fifo_in)
os.mkfifo(fifo_out)
# First step: we start the manager.
manager_filename = "manager"
manager_command = ["./%s" % manager_filename, fifo_in, fifo_out]
manager_executables_to_get = {
manager_filename: self.job.managers[manager_filename].digest
}
manager_files_to_get = {
"input.txt": self.job.testcases[test_number].input
}
manager_allow_path = ["input.txt", "output.txt", fifo_in, fifo_out]
for filename, digest in manager_executables_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in manager_files_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename, digest)
manager = evaluation_step_before_run(sandbox_mgr,
manager_command,
self.job.time_limit,
0,
manager_allow_path,
stdin_redirect="input.txt")
# Second step: we start the user submission compiled with the
# stub.
executable_filename = self.job.executables.keys()[0]
command = ["./%s" % executable_filename, fifo_out, fifo_in]
executables_to_get = {
executable_filename:
self.job.executables[executable_filename].digest
}
allow_path = [fifo_in, fifo_out]
for filename, digest in executables_to_get.iteritems():
sandbox_user.create_file_from_storage(filename,
digest,
executable=True)
process = evaluation_step_before_run(sandbox_user, command,
self.job.time_limit,
self.job.memory_limit, allow_path)
# Consume output.
wait_without_std([process, manager])
# TODO: check exit codes with translate_box_exitcode.
success_user, plus_user = \
evaluation_step_after_run(sandbox_user)
success_mgr, plus_mgr = \
evaluation_step_after_run(sandbox_mgr)
self.job.evaluations[test_number] = \
{'sandboxes': [sandbox_user.path,
sandbox_mgr.path],
'plus': plus_user}
evaluation = self.job.evaluations[test_number]
# If at least one evaluation had problems, we report the
# problems. (TODO: shouldn't outcome and text better be None
# and None?)
if not success_user or not success_mgr:
success, outcome, text = False, None, None
# If outcome_user is not None, it is 0.0 and it means that
# there has been some errors in the user solution, and outcome
# and text are meaningful, so we use them.
elif not is_evaluation_passed(plus_user):
success = True
outcome, text = 0.0, human_evaluation_message(plus_user)
# Otherwise, we use the manager to obtain the outcome.
else:
success = True
outcome, text = extract_outcome_and_text(sandbox_mgr)
# If asked so, save the output file, provided that it exists
if self.job.get_output:
if sandbox_mgr.file_exists("output.txt"):
evaluation['output'] = sandbox_mgr.get_file_to_storage(
"output.txt", "Output file for testcase %d in job %s" %
(test_number, self.job.info))
else:
evaluation['output'] = None
# Whatever happened, we conclude.
evaluation['success'] = success
evaluation['outcome'] = str(outcome) if outcome is not None else None
evaluation['text'] = text
delete_sandbox(sandbox_mgr)
delete_sandbox(sandbox_user)
return success
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
if len(job.executables) != 1:
raise ValueError("Unexpected number of executables (%s)" %
len(job.executables))
# Create the sandbox
sandbox = create_sandbox(
file_cacher,
multithreaded=job.multithreaded_sandbox,
name="evaluate")
# Prepare the execution
executable_filename = next(iterkeys(job.executables))
language = get_language(job.language)
main = Batch.GRADER_BASENAME \
if self._uses_grader() else executable_filename
commands = language.get_evaluation_commands(
executable_filename, main=main)
executables_to_get = {
executable_filename:
job.executables[executable_filename].digest
}
stdin_redirect = None
stdout_redirect = None
files_allowing_write = []
if len(self.input_filename) == 0:
self.input_filename = Batch.DEFAULT_INPUT_FILENAME
stdin_redirect = self.input_filename
if len(self.output_filename) == 0:
self.output_filename = Batch.DEFAULT_OUTPUT_FILENAME
stdout_redirect = self.output_filename
else:
files_allowing_write.append(self.output_filename)
files_to_get = {
self.input_filename: job.input
}
# Put the required files into the sandbox
for filename, digest in iteritems(executables_to_get):
sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in iteritems(files_to_get):
sandbox.create_file_from_storage(filename, digest)
# Actually performs the execution
success, plus = evaluation_step(
sandbox,
commands,
job.time_limit,
job.memory_limit,
writable_files=files_allowing_write,
stdin_redirect=stdin_redirect,
stdout_redirect=stdout_redirect)
job.sandboxes = [sandbox.path]
job.plus = plus
outcome = None
text = []
# Error in the sandbox: nothing to do!
if not success:
pass
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(plus):
outcome = 0.0
text = human_evaluation_message(plus)
if job.get_output:
job.user_output = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not sandbox.file_exists(self.output_filename):
outcome = 0.0
text = [N_("Evaluation didn't produce file %s"),
self.output_filename]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage.
if job.get_output:
job.user_output = sandbox.get_file_to_storage(
self.output_filename,
"Output file in job %s" % job.info,
trunc_len=100 * 1024)
# If just asked to execute, fill text and set dummy outcome.
if job.only_execution:
outcome = 0.0
text = [N_("Execution completed successfully")]
# Otherwise evaluate the output file.
else:
# Create a brand-new sandbox just for checking. Only admin
# code runs in it, so we allow multithreading and many
# processes (still with a limit to avoid fork-bombs).
checkbox = create_sandbox(
file_cacher,
multithreaded=True,
name="check")
checkbox.max_processes = 1000
checker_success, outcome, text = self._eval_output(
checkbox, job, sandbox.get_root_path())
success = success and checker_success
# Whatever happened, we conclude.
job.success = success
job.outcome = "%s" % outcome if outcome is not None else None
job.text = text
delete_sandbox(sandbox, job.success)
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
# Create the sandbox
sandbox = create_sandbox(file_cacher)
# Prepare the execution
executable_filename = job.executables.keys()[0]
language = job.language
commands = get_evaluation_commands(language, executable_filename)
executables_to_get = {
executable_filename: job.executables[executable_filename].digest
}
input_filename, output_filename = self.parameters[1]
stdin_redirect = None
stdout_redirect = None
if input_filename == "":
input_filename = "input.txt"
stdin_redirect = input_filename
if output_filename == "":
output_filename = "output.txt"
stdout_redirect = output_filename
files_to_get = {input_filename: job.input}
# Put the required files into the sandbox
for filename, digest in executables_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in files_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest)
# Actually performs the execution
success, plus = evaluation_step(sandbox,
commands,
job.time_limit,
job.memory_limit,
stdin_redirect=stdin_redirect,
stdout_redirect=stdout_redirect)
job.sandboxes = [sandbox.path]
job.plus = plus
outcome = None
text = None
# Error in the sandbox: nothing to do!
if not success:
pass
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(plus):
outcome = 0.0
text = human_evaluation_message(plus)
if job.get_output:
job.user_output = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not sandbox.file_exists(output_filename):
outcome = 0.0
text = [
N_("Evaluation didn't produce file %s"), output_filename
]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage
if job.get_output:
job.user_output = sandbox.get_file_to_storage(
output_filename,
"Output file in job %s" % job.info,
trunc_len=100 * 1024)
# If not asked otherwise, evaluate the output file
if not job.only_execution:
# Put the reference solution into the sandbox
sandbox.create_file_from_storage("res.txt", job.output)
# Check the solution with white_diff
if self.parameters[2] == "diff":
outcome, text = white_diff_step(
sandbox, output_filename, "res.txt")
# Check the solution with a comparator
elif self.parameters[2] == "comparator":
manager_filename = "checker"
if not manager_filename in job.managers:
logger.error(
"Configuration error: missing or "
"invalid comparator (it must be "
"named 'checker')",
extra={"operation": job.info})
success = False
else:
sandbox.create_file_from_storage(
manager_filename,
job.managers[manager_filename].digest,
executable=True)
success, _ = evaluation_step(
sandbox, [[
"./%s" % manager_filename, input_filename,
"res.txt", output_filename
]])
if success:
try:
outcome, text = \
extract_outcome_and_text(sandbox)
except ValueError, e:
logger.error("Invalid output from "
"comparator: %s" % (e.message, ),
extra={"operation": job.info})
success = False
else:
raise ValueError("Unrecognized third parameter"
" `%s' for Batch tasktype." %
self.parameters[2])
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
# Create the sandbox
sandbox = create_sandbox(file_cacher)
# Prepare the execution
executable_filename = job.executables.keys()[0]
command = [os.path.join(".", executable_filename)]
executables_to_get = {
executable_filename:
job.executables[executable_filename].digest
}
input_filename, output_filename = self.parameters[1]
stdin_redirect = None
stdout_redirect = None
if input_filename == "":
input_filename = "input.txt"
stdin_redirect = input_filename
if output_filename == "":
output_filename = "output.txt"
stdout_redirect = output_filename
files_to_get = {
input_filename: job.input
}
# Put the required files into the sandbox
for filename, digest in executables_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in files_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest)
# Actually performs the execution
success, plus = evaluation_step(
sandbox,
command,
job.time_limit,
job.memory_limit,
stdin_redirect=stdin_redirect,
stdout_redirect=stdout_redirect)
job.sandboxes = [sandbox.path]
job.plus = plus
outcome = None
text = None
# Error in the sandbox: nothing to do!
if not success:
pass
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(plus):
outcome = 0.0
text = human_evaluation_message(plus)
if job.get_output:
job.user_output = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not sandbox.file_exists(output_filename):
outcome = 0.0
text = [N_("Evaluation didn't produce file %s"),
output_filename]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage
if job.get_output:
job.user_output = sandbox.get_file_to_storage(
output_filename,
"Output file in job %s" % job.info,
trunc_len=100 * 1024)
# If not asked otherwise, evaluate the output file
if not job.only_execution:
# Put the reference solution into the sandbox
sandbox.create_file_from_storage(
"res.txt",
job.output)
# Check the solution with white_diff
if self.parameters[2] == "diff":
outcome, text = white_diff_step(
sandbox, output_filename, "res.txt")
# Check the solution with a comparator
elif self.parameters[2] == "comparator":
manager_filename = "checker"
if not manager_filename in job.managers:
logger.error("Configuration error: missing or "
"invalid comparator (it must be "
"named 'checker')",
extra={"operation": job.info})
success = False
else:
sandbox.create_file_from_storage(
manager_filename,
job.managers[manager_filename].digest,
executable=True)
success, _ = evaluation_step(
sandbox,
["./%s" % manager_filename,
input_filename, "res.txt", output_filename])
if success:
try:
outcome, text = \
extract_outcome_and_text(sandbox)
except ValueError, e:
logger.error("Invalid output from "
"comparator: %s" % (e.message,),
extra={"operation": job.info})
success = False
else:
raise ValueError("Unrecognized third parameter"
" `%s' for Batch tasktype." %
self.parameters[2])
def evaluate_testcase(self, test_number):
"""See TaskType.evaluate_testcase."""
# Create sandboxes and FIFOs
sandbox_mgr = create_sandbox(self)
sandbox_user = create_sandbox(self)
fifo_dir = tempfile.mkdtemp(dir=config.temp_dir)
fifo_in = os.path.join(fifo_dir, "in")
fifo_out = os.path.join(fifo_dir, "out")
os.mkfifo(fifo_in)
os.mkfifo(fifo_out)
# First step: we start the manager.
manager_filename = "manager"
manager_command = ["./%s" % manager_filename, fifo_in, fifo_out]
manager_executables_to_get = {
manager_filename:
self.job.managers[manager_filename].digest
}
manager_files_to_get = {
"input.txt": self.job.testcases[test_number].input
}
manager_allow_path = ["input.txt", "output.txt", fifo_in, fifo_out]
for filename, digest in manager_executables_to_get.iteritems():
sandbox_mgr.create_file_from_storage(
filename, digest, executable=True)
for filename, digest in manager_files_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename, digest)
manager = evaluation_step_before_run(
sandbox_mgr,
manager_command,
self.job.time_limit,
0,
manager_allow_path,
stdin_redirect="input.txt")
# Second step: we start the user submission compiled with the
# stub.
executable_filename = self.job.executables.keys()[0]
command = ["./%s" % executable_filename, fifo_out, fifo_in]
executables_to_get = {
executable_filename:
self.job.executables[executable_filename].digest
}
allow_path = [fifo_in, fifo_out]
for filename, digest in executables_to_get.iteritems():
sandbox_user.create_file_from_storage(
filename, digest, executable=True)
process = evaluation_step_before_run(
sandbox_user,
command,
self.job.time_limit,
self.job.memory_limit,
allow_path)
# Consume output.
wait_without_std([process, manager])
# TODO: check exit codes with translate_box_exitcode.
success_user, plus_user = \
evaluation_step_after_run(sandbox_user)
success_mgr, plus_mgr = \
evaluation_step_after_run(sandbox_mgr)
self.job.evaluations[test_number] = \
{'sandboxes': [sandbox_user.path,
sandbox_mgr.path],
'plus': plus_user}
evaluation = self.job.evaluations[test_number]
# If at least one evaluation had problems, we report the
# problems. (TODO: shouldn't outcome and text better be None
# and None?)
if not success_user or not success_mgr:
success, outcome, text = False, None, None
# If outcome_user is not None, it is 0.0 and it means that
# there has been some errors in the user solution, and outcome
# and text are meaningful, so we use them.
elif not is_evaluation_passed(plus_user):
success = True
outcome, text = 0.0, human_evaluation_message(plus_user)
# Otherwise, we use the manager to obtain the outcome.
else:
success = True
outcome, text = extract_outcome_and_text(sandbox_mgr)
# If asked so, save the output file, provided that it exists
if self.job.get_output:
if sandbox_mgr.file_exists("output.txt"):
evaluation['output'] = sandbox_mgr.get_file_to_storage(
"output.txt",
"Output file for testcase %d in job %s" %
(test_number, self.job.info))
else:
evaluation['output'] = None
# Whatever happened, we conclude.
evaluation['success'] = success
evaluation['outcome'] = str(outcome) if outcome is not None else None
evaluation['text'] = text
delete_sandbox(sandbox_mgr)
delete_sandbox(sandbox_user)
return success
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
# f stand for first, s for second.
first_sandbox = create_sandbox(file_cacher,
multithreaded=job.multithreaded_sandbox,
name="first_evaluate")
second_sandbox = create_sandbox(
file_cacher,
multithreaded=job.multithreaded_sandbox,
name="second_evaluate")
fifo_dir = tempfile.mkdtemp(dir=config.temp_dir)
fifo = os.path.join(fifo_dir, "fifo")
os.mkfifo(fifo)
os.chmod(fifo_dir, 0o755)
os.chmod(fifo, 0o666)
# First step: we start the first manager.
first_filename = "manager"
first_command = ["./%s" % first_filename, "0", fifo]
first_executables_to_get = {
first_filename: job.executables[first_filename].digest
}
first_files_to_get = {"input.txt": job.input}
first_allow_path = [fifo_dir]
# Put the required files into the sandbox
for filename, digest in iteritems(first_executables_to_get):
first_sandbox.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in iteritems(first_files_to_get):
first_sandbox.create_file_from_storage(filename, digest)
first = evaluation_step_before_run(first_sandbox,
first_command,
job.time_limit,
job.memory_limit,
first_allow_path,
stdin_redirect="input.txt",
wait=False)
# Second step: we start the second manager.
second_filename = "manager"
second_command = ["./%s" % second_filename, "1", fifo]
second_executables_to_get = {
second_filename: job.executables[second_filename].digest
}
second_files_to_get = {}
second_allow_path = [fifo_dir]
# Put the required files into the second sandbox
for filename, digest in iteritems(second_executables_to_get):
second_sandbox.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in iteritems(second_files_to_get):
second_sandbox.create_file_from_storage(filename, digest)
second = evaluation_step_before_run(second_sandbox,
second_command,
job.time_limit,
job.memory_limit,
second_allow_path,
stdout_redirect="output.txt",
wait=False)
# Consume output.
wait_without_std([second, first])
# TODO: check exit codes with translate_box_exitcode.
success_first, first_plus = \
evaluation_step_after_run(first_sandbox)
success_second, second_plus = \
evaluation_step_after_run(second_sandbox)
job.sandboxes = [first_sandbox.path, second_sandbox.path]
job.plus = second_plus
success = True
outcome = None
text = []
# Error in the sandbox: report failure!
if not success_first or not success_second:
success = False
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(first_plus) or \
not is_evaluation_passed(second_plus):
outcome = 0.0
if not is_evaluation_passed(first_plus):
text = human_evaluation_message(first_plus)
else:
text = human_evaluation_message(second_plus)
if job.get_output:
job.user_output = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not second_sandbox.file_exists('output.txt'):
outcome = 0.0
text = [N_("Evaluation didn't produce file %s"), "output.txt"]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage
if job.get_output:
job.user_output = second_sandbox.get_file_to_storage(
"output.txt", "Output file in job %s" % job.info)
# If not asked otherwise, evaluate the output file
if not job.only_execution:
# Put the reference solution into the sandbox
second_sandbox.create_file_from_storage(
"res.txt", job.output)
# If a checker is not provided, use white-diff
if self.parameters[0] == "diff":
outcome, text = white_diff_step(
second_sandbox, "output.txt", "res.txt")
elif self.parameters[0] == "comparator":
if TwoSteps.CHECKER_FILENAME not in job.managers:
logger.error(
"Configuration error: missing or "
"invalid comparator (it must be "
"named `checker')",
extra={"operation": job.info})
success = False
else:
second_sandbox.create_file_from_storage(
TwoSteps.CHECKER_FILENAME,
job.managers[TwoSteps.CHECKER_FILENAME].digest,
executable=True)
# Rewrite input file, as in Batch.py
try:
second_sandbox.remove_file("input.txt")
except OSError as e:
assert not second_sandbox.file_exists(
"input.txt")
second_sandbox.create_file_from_storage(
"input.txt", job.input)
success, _ = evaluation_step(
second_sandbox, [[
"./%s" % TwoSteps.CHECKER_FILENAME,
"input.txt", "res.txt", "output.txt"
]])
if success:
try:
outcome, text = extract_outcome_and_text(
second_sandbox)
except ValueError as e:
logger.error(
"Invalid output from "
"comparator: %s",
e,
extra={"operation": job.info})
success = False
else:
raise ValueError("Uncrecognized first parameter"
" `%s' for TwoSteps tasktype." %
self.parameters[0])
# Whatever happened, we conclude.
job.success = success
job.outcome = str(outcome) if outcome is not None else None
job.text = text
delete_sandbox(first_sandbox, job.success)
delete_sandbox(second_sandbox, job.success)
def evaluate_testcase(self, test_number):
"""See TaskType.evaluate_testcase."""
# Create the sandbox
sandbox = create_sandbox(self)
# Prepare the execution
executable_filename = self.job.executables.keys()[0]
command = [os.path.join(".", executable_filename)]
executables_to_get = {
executable_filename:
self.job.executables[executable_filename].digest
}
input_filename, output_filename = self.job.task_type_parameters[1]
stdin_redirect = None
stdout_redirect = None
if input_filename == "":
input_filename = "input.txt"
stdin_redirect = input_filename
if output_filename == "":
output_filename = "output.txt"
stdout_redirect = output_filename
files_to_get = {input_filename: self.job.testcases[test_number].input}
# Put the required files into the sandbox
for filename, digest in executables_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in files_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest)
# Actually performs the execution
success, plus = evaluation_step(sandbox,
command,
self.job.time_limit,
self.job.memory_limit,
stdin_redirect=stdin_redirect,
stdout_redirect=stdout_redirect)
self.job.evaluations[test_number] = {
'sandboxes': [sandbox.path],
'plus': plus
}
outcome = None
text = None
evaluation = self.job.evaluations[test_number]
# Error in the sandbox: nothing to do!
if not success:
pass
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(plus):
outcome = 0.0
text = human_evaluation_message(plus)
if self.job.get_output:
evaluation['output'] = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not sandbox.file_exists(output_filename):
outcome = 0.0
text = "Execution didn't produce file %s" % \
(output_filename)
if self.job.get_output:
evaluation['output'] = None
else:
# If asked so, put the output file into the storage
if self.job.get_output:
evaluation['output'] = sandbox.get_file_to_storage(
output_filename,
"Output file for testcase %d in job %s" %
(test_number, self.job.info),
trunc_len=100 * 1024)
# If not asked otherwise, evaluate the output file
if not self.job.only_execution:
# Put the reference solution into the sandbox
sandbox.create_file_from_storage(
"res.txt", self.job.testcases[test_number].output)
# Check the solution with white_diff
if self.job.task_type_parameters[2] == "diff":
outcome, text = white_diff_step(
sandbox, output_filename, "res.txt")
# Check the solution with a comparator
elif self.job.task_type_parameters[2] == "comparator":
manager_filename = "checker"
if not manager_filename in self.job.managers:
logger.error("Configuration error: missing or "
"invalid comparator (it must be "
"named 'checker')")
success = False
else:
sandbox.create_file_from_storage(
manager_filename,
self.job.managers[manager_filename].digest,
executable=True)
success, _ = evaluation_step(
sandbox, [
"./%s" % manager_filename, input_filename,
"res.txt", output_filename
])
if success:
try:
outcome, text = \
extract_outcome_and_text(sandbox)
except ValueError, e:
logger.error("Invalid output from "
"comparator: %s" % (e.message, ))
success = False
else:
raise ValueError("Unrecognized third parameter"
" `%s' for Batch tasktype." %
self.job.task_type_parameters[2])
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
# Create the sandbox
sandbox = create_sandbox(file_cacher, job.multithreaded_sandbox)
# Prepare the execution
executable_filename = job.executables.keys()[0]
language = get_language(job.language)
commands = language.get_evaluation_commands(
executable_filename,
main="grader" if self._uses_grader() else executable_filename)
executables_to_get = {
executable_filename:
job.executables[executable_filename].digest
}
input_filename, output_filename = self.parameters[1]
stdin_redirect = None
stdout_redirect = None
files_allowing_write = []
if input_filename == "":
input_filename = "input.txt"
stdin_redirect = input_filename
if output_filename == "":
output_filename = "output.txt"
stdout_redirect = output_filename
else:
files_allowing_write.append(output_filename)
files_to_get = {
input_filename: job.input
}
# Put the required files into the sandbox
for filename, digest in executables_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in files_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest)
# Actually performs the execution
success, plus = evaluation_step(
sandbox,
commands,
job.time_limit,
job.memory_limit,
writable_files=files_allowing_write,
stdin_redirect=stdin_redirect,
stdout_redirect=stdout_redirect)
job.sandboxes = [sandbox.path]
job.plus = plus
outcome = None
text = None
# Error in the sandbox: nothing to do!
if not success:
pass
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(plus):
outcome = 0.0
text = human_evaluation_message(plus)
if job.get_output:
job.user_output = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not sandbox.file_exists(output_filename):
outcome = 0.0
text = [N_("Evaluation didn't produce file %s"),
output_filename]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage
if job.get_output:
job.user_output = sandbox.get_file_to_storage(
output_filename,
"Output file in job %s" % job.info,
trunc_len=100 * 1024)
# If just asked to execute, fill text and set dummy
# outcome.
if job.only_execution:
outcome = 0.0
text = [N_("Execution completed successfully")]
# Otherwise evaluate the output file.
else:
# Put the reference solution into the sandbox
sandbox.create_file_from_storage(
"res.txt",
job.output)
# Check the solution with white_diff
if self.parameters[2] == "diff":
outcome, text = white_diff_step(
sandbox, output_filename, "res.txt")
# Check the solution with a comparator
elif self.parameters[2] == "comparator":
manager_filename = "checker"
if manager_filename not in job.managers:
logger.error("Configuration error: missing or "
"invalid comparator (it must be "
"named 'checker')",
extra={"operation": job.info})
success = False
else:
sandbox.create_file_from_storage(
manager_filename,
job.managers[manager_filename].digest,
executable=True)
# Rewrite input file. The untrusted
# contestant program should not be able to
# modify it; however, the grader may
# destroy the input file to prevent the
# contestant's program from directly
# accessing it. Since we cannot create
# files already existing in the sandbox,
# we try removing the file first.
try:
sandbox.remove_file(input_filename)
except OSError as e:
# Let us be extra sure that the file
# was actually removed and we did not
# mess up with permissions.
assert not sandbox.file_exists(input_filename)
sandbox.create_file_from_storage(
input_filename,
job.input)
# Allow using any number of processes (because e.g.
# one may want to write a bash checker who calls
# other processes). Set to a high number because
# to avoid fork-bombing the worker.
sandbox.max_processes = 1000
success, _ = evaluation_step(
sandbox,
[["./%s" % manager_filename,
input_filename, "res.txt", output_filename]])
if success:
try:
outcome, text = \
extract_outcome_and_text(sandbox)
except ValueError, e:
logger.error("Invalid output from "
"comparator: %s", e.message,
extra={"operation": job.info})
success = False
else:
raise ValueError("Unrecognized third parameter"
" `%s' for Batch tasktype." %
self.parameters[2])
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
if len(self.parameters) <= 0:
num_processes = 1
else:
num_processes = self.parameters[0]
indices = range(num_processes)
# Create sandboxes and FIFOs
sandbox_mgr = create_sandbox(
file_cacher,
multithreaded=job.multithreaded_sandbox,
name="manager_evaluate")
sandbox_user = [
create_sandbox(
file_cacher,
multithreaded=job.multithreaded_sandbox,
name="user_evaluate")
for i in indices]
fifo_dir = [tempfile.mkdtemp(dir=config.temp_dir) for i in indices]
fifo_in = [os.path.join(fifo_dir[i], "in%d" % i) for i in indices]
fifo_out = [os.path.join(fifo_dir[i], "out%d" % i) for i in indices]
for i in indices:
os.mkfifo(fifo_in[i])
os.mkfifo(fifo_out[i])
os.chmod(fifo_dir[i], 0o755)
os.chmod(fifo_in[i], 0o666)
os.chmod(fifo_out[i], 0o666)
# First step: prepare the manager.
manager_filename = "manager"
manager_command = ["./%s" % manager_filename]
for i in indices:
manager_command.append(fifo_in[i])
manager_command.append(fifo_out[i])
manager_executables_to_get = {
manager_filename:
job.managers[manager_filename].digest
}
manager_files_to_get = {
"input.txt": job.input
}
manager_allow_dirs = fifo_dir
for filename, digest in iteritems(manager_executables_to_get):
sandbox_mgr.create_file_from_storage(
filename, digest, executable=True)
for filename, digest in iteritems(manager_files_to_get):
sandbox_mgr.create_file_from_storage(filename, digest)
# Second step: load the executables for the user processes
# (done before launching the manager so that it does not
# impact its wall clock time).
assert len(job.executables) == 1
executable_filename = next(iterkeys(job.executables))
executables_to_get = {
executable_filename:
job.executables[executable_filename].digest
}
for i in indices:
for filename, digest in iteritems(executables_to_get):
sandbox_user[i].create_file_from_storage(
filename, digest, executable=True)
# Third step: start the manager.
manager = evaluation_step_before_run(
sandbox_mgr,
manager_command,
num_processes * job.time_limit,
0,
allow_dirs=manager_allow_dirs,
writable_files=["output.txt"],
stdin_redirect="input.txt")
# Fourth step: start the user submissions compiled with the stub.
language = get_language(job.language)
processes = [None for i in indices]
for i in indices:
args = [fifo_out[i], fifo_in[i]]
if num_processes != 1:
args.append(str(i))
commands = language.get_evaluation_commands(
executable_filename,
main="stub",
args=args)
user_allow_dirs = [fifo_dir[i]]
# Assumes that the actual execution of the user solution
# is the last command in commands, and that the previous
# are "setup" that doesn't need tight control.
if len(commands) > 1:
evaluation_step(sandbox_user[i], commands[:-1], 10, 256)
processes[i] = evaluation_step_before_run(
sandbox_user[i],
commands[-1],
job.time_limit,
job.memory_limit,
allow_dirs=user_allow_dirs)
# Consume output.
wait_without_std(processes + [manager])
# TODO: check exit codes with translate_box_exitcode.
user_results = [evaluation_step_after_run(s) for s in sandbox_user]
success_user = all(r[0] for r in user_results)
plus_user = reduce(merge_evaluation_results,
[r[1] for r in user_results])
success_mgr, unused_plus_mgr = \
evaluation_step_after_run(sandbox_mgr)
if plus_user['exit_status'] == Sandbox.EXIT_OK and \
plus_user["execution_time"] >= job.time_limit:
plus_user['exit_status'] = Sandbox.EXIT_TIMEOUT
# Merge results.
job.sandboxes = [s.path for s in sandbox_user] + [sandbox_mgr.path]
job.plus = plus_user
# If at least one evaluation had problems, we report the
# problems.
if not success_user or not success_mgr:
success, outcome, text = False, None, []
# If the user sandbox detected some problem (timeout, ...),
# the outcome is 0.0 and the text describes that problem.
elif not is_evaluation_passed(plus_user):
success = True
outcome, text = 0.0, human_evaluation_message(plus_user)
# Otherwise, we use the manager to obtain the outcome.
else:
success = True
outcome, text = extract_outcome_and_text(sandbox_mgr)
# If asked so, save the output file, provided that it exists
if job.get_output:
if sandbox_mgr.file_exists("output.txt"):
job.user_output = sandbox_mgr.get_file_to_storage(
"output.txt",
"Output file in job %s" % job.info)
else:
job.user_output = None
# Whatever happened, we conclude.
job.success = success
job.outcome = "%s" % outcome if outcome is not None else None
job.text = text
delete_sandbox(sandbox_mgr, job.success)
for s in sandbox_user:
delete_sandbox(s, job.success)
if not config.keep_sandbox:
for d in fifo_dir:
rmtree(d)
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
if len(self.parameters) <= 0:
num_processes = 1
else:
num_processes = self.parameters[0]
indices = range(num_processes)
# Create sandboxes and FIFOs
sandbox_mgr = create_sandbox(file_cacher, job.multithreaded_sandbox)
sandbox_user = [
create_sandbox(file_cacher, job.multithreaded_sandbox)
for i in indices
]
fifo_dir = [tempfile.mkdtemp(dir=config.temp_dir) for i in indices]
fifo_in = [os.path.join(fifo_dir[i], "in%d" % i) for i in indices]
fifo_out = [os.path.join(fifo_dir[i], "out%d" % i) for i in indices]
for i in indices:
os.mkfifo(fifo_in[i])
os.mkfifo(fifo_out[i])
os.chmod(fifo_dir[i], 0o755)
os.chmod(fifo_in[i], 0o666)
os.chmod(fifo_out[i], 0o666)
# First step: we start the manager.
manager_filename = "manager"
manager_command = ["./%s" % manager_filename]
for i in indices:
manager_command.append(fifo_in[i])
manager_command.append(fifo_out[i])
manager_executables_to_get = {
manager_filename: job.managers[manager_filename].digest
}
manager_files_to_get = {"input.txt": job.input}
manager_allow_dirs = fifo_dir
for filename, digest in manager_executables_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in manager_files_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename, digest)
manager = evaluation_step_before_run(sandbox_mgr,
manager_command,
num_processes * job.time_limit,
0,
allow_dirs=manager_allow_dirs,
writable_files=["output.txt"],
stdin_redirect="input.txt")
# Second step: we start the user submission compiled with the
# stub.
language = get_language(job.language)
executable_filename = job.executables.keys()[0]
executables_to_get = {
executable_filename: job.executables[executable_filename].digest
}
processes = [None for i in indices]
for i in indices:
args = [fifo_out[i], fifo_in[i]]
if num_processes != 1:
args.append(str(i))
commands = language.get_evaluation_commands(executable_filename,
main="stub",
args=args)
user_allow_dirs = [fifo_dir[i]]
for filename, digest in executables_to_get.iteritems():
sandbox_user[i].create_file_from_storage(filename,
digest,
executable=True)
# Assumes that the actual execution of the user solution
# is the last command in commands, and that the previous
# are "setup" that doesn't need tight control.
if len(commands) > 1:
evaluation_step(sandbox_user[i], commands[:-1], 10, 256)
processes[i] = evaluation_step_before_run(
sandbox_user[i],
commands[-1],
job.time_limit,
job.memory_limit,
allow_dirs=user_allow_dirs)
# Consume output.
wait_without_std(processes + [manager])
# TODO: check exit codes with translate_box_exitcode.
user_results = [evaluation_step_after_run(s) for s in sandbox_user]
success_user = all(r[0] for r in user_results)
plus_user = reduce(merge_evaluation_results,
[r[1] for r in user_results])
success_mgr, unused_plus_mgr = \
evaluation_step_after_run(sandbox_mgr)
if plus_user['exit_status'] == Sandbox.EXIT_OK and \
plus_user["execution_time"] >= job.time_limit:
plus_user['exit_status'] = Sandbox.EXIT_TIMEOUT
# Merge results.
job.sandboxes = [s.path for s in sandbox_user] + [sandbox_mgr.path]
job.plus = plus_user
# If at least one evaluation had problems, we report the
# problems.
if not success_user or not success_mgr:
success, outcome, text = False, None, None
# If the user sandbox detected some problem (timeout, ...),
# the outcome is 0.0 and the text describes that problem.
elif not is_evaluation_passed(plus_user):
success = True
outcome, text = 0.0, human_evaluation_message(plus_user)
# Otherwise, we use the manager to obtain the outcome.
else:
success = True
outcome, text = extract_outcome_and_text(sandbox_mgr)
# If asked so, save the output file, provided that it exists
if job.get_output:
if sandbox_mgr.file_exists("output.txt"):
job.user_output = sandbox_mgr.get_file_to_storage(
"output.txt", "Output file in job %s" % job.info)
else:
job.user_output = None
# Whatever happened, we conclude.
job.success = success
job.outcome = "%s" % outcome if outcome is not None else None
job.text = text
delete_sandbox(sandbox_mgr, job.success)
for s in sandbox_user:
delete_sandbox(s, job.success)
if not config.keep_sandbox:
for d in fifo_dir:
rmtree(d)
def evaluate_testcase(self, test_number):
"""See TaskType.evaluate_testcase."""
# f stand for first, s for second.
first_sandbox = create_sandbox(self)
second_sandbox = create_sandbox(self)
fifo_dir = tempfile.mkdtemp(dir=config.temp_dir)
fifo = os.path.join(fifo_dir, "fifo")
os.mkfifo(fifo)
# First step: we start the first manager.
first_filename = "manager"
first_command = ["./%s" % first_filename, "0", fifo]
first_executables_to_get = {
first_filename: self.job.executables[first_filename].digest
}
first_files_to_get = {
"input.txt": self.job.testcases[test_number].input
}
first_allow_path = ["input.txt", fifo]
# Put the required files into the sandbox
for filename, digest in first_executables_to_get.iteritems():
first_sandbox.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in first_files_to_get.iteritems():
first_sandbox.create_file_from_storage(filename, digest)
first = evaluation_step_before_run(first_sandbox,
first_command,
self.job.time_limit,
self.job.memory_limit,
first_allow_path,
stdin_redirect="input.txt",
wait=False)
# Second step: we start the second manager.
second_filename = "manager"
second_command = ["./%s" % second_filename, "1", fifo]
second_executables_to_get = {
second_filename: self.job.executables[second_filename].digest
}
second_files_to_get = {}
second_allow_path = [fifo, "output.txt"]
# Put the required files into the second sandbox
for filename, digest in second_executables_to_get.iteritems():
second_sandbox.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in second_files_to_get.iteritems():
second_sandbox.create_file_from_storage(filename, digest)
second = evaluation_step_before_run(second_sandbox,
second_command,
self.job.time_limit,
self.job.memory_limit,
second_allow_path,
stdout_redirect="output.txt",
wait=False)
# Consume output.
wait_without_std([second, first])
# TODO: check exit codes with translate_box_exitcode.
success_first, first_plus = \
evaluation_step_after_run(first_sandbox)
success_second, second_plus = \
evaluation_step_after_run(second_sandbox)
self.job.evaluations[test_number] = {
'sandboxes': [first_sandbox.path, second_sandbox.path],
'plus': second_plus
}
outcome = None
text = None
evaluation = self.job.evaluations[test_number]
success = True
# Error in the sandbox: report failure!
if not success_first or not success_second:
success = False
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(first_plus) or \
not is_evaluation_passed(second_plus):
outcome = 0.0
if not is_evaluation_passed(first_plus):
text = human_evaluation_message(first_plus)
else:
text = human_evaluation_message(second_plus)
if self.job.get_output:
evaluation['output'] = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not second_sandbox.file_exists('output.txt'):
outcome = 0.0
text = "Execution didn't produce file output.txt"
if self.job.get_output:
evaluation['output'] = None
else:
# If asked so, put the output file into the storage
if self.job.get_output:
evaluation['output'] = second_sandbox.get_file_to_storage(
"output.txt", "Output file for testcase %d in job %s" %
(test_number, self.job.info))
# If not asked otherwise, evaluate the output file
if not self.job.only_execution:
# Put the reference solution into the sandbox
second_sandbox.create_file_from_storage(
"res.txt", self.job.testcases[test_number].output)
outcome, text = white_diff_step(second_sandbox,
"output.txt", "res.txt")
# Whatever happened, we conclude.
evaluation['success'] = success
evaluation['outcome'] = outcome
evaluation['text'] = text
delete_sandbox(first_sandbox)
delete_sandbox(second_sandbox)
return success
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
if len(job.executables) != 1:
raise ValueError("Unexpected number of executables (%s)" %
len(job.executables))
# Create the sandbox
sandbox = create_sandbox(file_cacher,
multithreaded=job.multithreaded_sandbox,
name="evaluate")
# Prepare the execution
executable_filename = next(iterkeys(job.executables))
language = get_language(job.language)
main = Batch.GRADER_BASENAME \
if self._uses_grader() else executable_filename
commands = language.get_evaluation_commands(executable_filename,
main=main)
executables_to_get = {
executable_filename: job.executables[executable_filename].digest
}
stdin_redirect = None
stdout_redirect = None
files_allowing_write = []
if len(self.input_filename) == 0:
self.input_filename = Batch.DEFAULT_INPUT_FILENAME
stdin_redirect = self.input_filename
if len(self.output_filename) == 0:
self.output_filename = Batch.DEFAULT_OUTPUT_FILENAME
stdout_redirect = self.output_filename
else:
files_allowing_write.append(self.output_filename)
files_to_get = {self.input_filename: job.input}
# Put the required files into the sandbox
for filename, digest in iteritems(executables_to_get):
sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in iteritems(files_to_get):
sandbox.create_file_from_storage(filename, digest)
# Actually performs the execution
success, plus = evaluation_step(sandbox,
commands,
job.time_limit,
job.memory_limit,
writable_files=files_allowing_write,
stdin_redirect=stdin_redirect,
stdout_redirect=stdout_redirect)
job.sandboxes = [sandbox.path]
job.plus = plus
outcome = None
text = []
# Error in the sandbox: nothing to do!
if not success:
pass
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(plus):
outcome = 0.0
text = human_evaluation_message(plus)
if job.get_output:
job.user_output = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not sandbox.file_exists(self.output_filename):
outcome = 0.0
text = [
N_("Evaluation didn't produce file %s"),
self.output_filename
]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage.
if job.get_output:
job.user_output = sandbox.get_file_to_storage(
self.output_filename,
"Output file in job %s" % job.info,
trunc_len=100 * 1024)
# If just asked to execute, fill text and set dummy outcome.
if job.only_execution:
outcome = 0.0
text = [N_("Execution completed successfully")]
# Otherwise evaluate the output file.
else:
# Create a brand-new sandbox just for checking. Only admin
# code runs in it, so we allow multithreading and many
# processes (still with a limit to avoid fork-bombs).
checkbox = create_sandbox(file_cacher,
multithreaded=True,
name="check")
checkbox.max_processes = 1000
checker_success, outcome, text = self._eval_output(
checkbox, job, sandbox.get_root_path())
success = success and checker_success
# Whatever happened, we conclude.
job.success = success
job.outcome = "%s" % outcome if outcome is not None else None
job.text = text
delete_sandbox(sandbox, job.success)
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
if len(self.parameters) <= 0:
num_processes = 1
else:
num_processes = self.parameters[0]
indices = range(num_processes)
# Create sandboxes and FIFOs
sandbox_mgr = create_sandbox(file_cacher, job.multithreaded_sandbox)
sandbox_user = [
create_sandbox(file_cacher, job.multithreaded_sandbox)
for i in indices
]
fifo_dir = [tempfile.mkdtemp(dir=config.temp_dir) for i in indices]
fifo_in = [os.path.join(fifo_dir[i], "in%d" % i) for i in indices]
fifo_out = [os.path.join(fifo_dir[i], "out%d" % i) for i in indices]
for i in indices:
os.mkfifo(fifo_in[i])
os.mkfifo(fifo_out[i])
os.chmod(fifo_dir[i], 0o755)
os.chmod(fifo_in[i], 0o666)
os.chmod(fifo_out[i], 0o666)
# First step: we start the manager.
manager_filename = "manager"
manager_command = ["./%s" % manager_filename]
for i in indices:
manager_command.append(fifo_in[i])
manager_command.append(fifo_out[i])
manager_executables_to_get = {
manager_filename: job.managers[manager_filename].digest
}
manager_files_to_get = {"input.txt": job.input}
manager_allow_dirs = fifo_dir
for filename, digest in manager_executables_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in manager_files_to_get.iteritems():
sandbox_mgr.create_file_from_storage(filename, digest)
manager = evaluation_step_before_run(
sandbox_mgr,
manager_command,
num_processes * job.time_limit,
0,
allow_dirs=manager_allow_dirs,
writable_files=["output.txt"],
stdin_redirect="input.txt",
stdout_redirect="output.txt",
)
# Second step: we start the user submission compiled with the
# stub.
language = get_language(job.language)
executable_filename = job.executables.keys()[0]
executables_to_get = {
executable_filename: job.executables[executable_filename].digest
}
processes = [None for i in indices]
for i in indices:
args = [fifo_out[i], fifo_in[i]]
if num_processes != 1:
args.append(str(i))
commands = language.get_evaluation_commands(executable_filename,
main="grader",
args=args)
user_allow_dirs = [fifo_dir[i]]
for filename, digest in executables_to_get.iteritems():
sandbox_user[i].create_file_from_storage(filename,
digest,
executable=True)
# Assumes that the actual execution of the user solution
# is the last command in commands, and that the previous
# are "setup" that doesn't need tight control.
if len(commands) > 1:
evaluation_step(sandbox_user[i], commands[:-1], 10, 256)
processes[i] = evaluation_step_before_run(
sandbox_user[i],
commands[-1],
job.time_limit,
job.memory_limit,
allow_dirs=user_allow_dirs)
# Consume output.
wait_without_std(processes + [manager])
# TODO: check exit codes with translate_box_exitcode.
user_results = [evaluation_step_after_run(s) for s in sandbox_user]
success_user = all(r[0] for r in user_results)
plus_user = reduce(merge_evaluation_results,
[r[1] for r in user_results])
success_mgr, unused_plus_mgr = \
evaluation_step_after_run(sandbox_mgr)
if plus_user['exit_status'] == Sandbox.EXIT_OK and \
plus_user["execution_time"] >= job.time_limit:
plus_user['exit_status'] = Sandbox.EXIT_TIMEOUT
# Merge results.
job.sandboxes = [s.path for s in sandbox_user] + [sandbox_mgr.path]
job.plus = plus_user
# If at least one evaluation had problems, we report the
# problems.
if not success_user or not success_mgr:
success, outcome, text = False, None, None
# If the user sandbox detected some problem (timeout, ...),
# the outcome is 0.0 and the text describes that problem.
elif not is_evaluation_passed(plus_user):
success = True
outcome, text = 0.0, human_evaluation_message(plus_user)
if job.get_output:
job.user_output = None
# Otherwise, we use the manager to obtain the outcome.
else:
success = True
outcome = None
text = None
input_filename = "input.txt"
output_filename = "output.txt"
# Check that the output file was created
if not sandbox_mgr.file_exists(output_filename):
outcome = 0.0
text = [
N_("Evaluation didn't produce file %s"), output_filename
]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage
if job.get_output:
job.user_output = sandbox_mgr.get_file_to_storage(
output_filename,
"Output file in job %s" % job.info,
trunc_len=1024 * 1024 * 10)
# If just asked to execute, fill text and set dummy
# outcome.
if job.only_execution:
outcome = 0.0
text = [N_("Execution completed successfully")]
# Otherwise evaluate the output file.
else:
# Put the reference solution into the sandbox
sandbox_mgr.create_file_from_storage("res.txt", job.output)
# Check the solution with white_diff
if self.parameters[1] == "diff":
outcome, text = white_diff_step(
sandbox_mgr, output_filename, "res.txt")
# Check the solution with a comparator
elif self.parameters[1] == "comparator":
manager_filename = "checker"
if manager_filename not in job.managers:
logger.error(
"Configuration error: missing or "
"invalid comparator (it must be "
"named 'checker')",
extra={"operation": job.info})
success = False
else:
sandbox_mgr.create_file_from_storage(
manager_filename,
job.managers[manager_filename].digest,
executable=True)
# Rewrite input file. The untrusted
# contestant program should not be able to
# modify it; however, the grader may
# destroy the input file to prevent the
# contestant's program from directly
# accessing it. Since we cannot create
# files already existing in the sandbox,
# we try removing the file first.
try:
sandbox_mgr.remove_file(input_filename)
except OSError as e:
# Let us be extra sure that the file
# was actually removed and we did not
# mess up with permissions.
assert not sandbox_mgr.file_exists(
input_filename)
sandbox_mgr.create_file_from_storage(
input_filename, job.input)
# Allow using any number of processes (because e.g.
# one may want to write a bash checker who calls
# other processes). Set to a high number because
# to avoid fork-bombing the worker.
sandbox_mgr.max_processes = 1000
success, _ = evaluation_step(
sandbox_mgr, [[
"./%s" % manager_filename, input_filename,
"res.txt", output_filename
]])
if success:
try:
outcome, text = \
extract_outcome_and_text(sandbox_mgr)
except ValueError as e:
logger.error(
"Invalid output from "
"comparator: %s",
e.message,
extra={"operation": job.info})
success = False
else:
raise ValueError("Unrecognized second parameter"
" `%s' for Communication tasktype." %
self.parameters[2])
# Whatever happened, we conclude.
job.success = success
job.outcome = "%s" % outcome if outcome is not None else None
job.text = text
delete_sandbox(sandbox_mgr, job.success)
for s in sandbox_user:
delete_sandbox(s, job.success)
if not config.keep_sandbox:
for d in fifo_dir:
rmtree(d)
def evaluate_testcase(self, test_number):
"""See TaskType.evaluate_testcase."""
# Create the sandbox
sandbox = create_sandbox(self)
# Prepare the execution
executable_filename = self.job.executables.keys()[0]
command = [sandbox.relative_path(executable_filename)]
executables_to_get = {
executable_filename:
self.job.executables[executable_filename].digest
}
input_filename, output_filename = self.job.task_type_parameters[1]
stdin_redirect = None
stdout_redirect = None
if input_filename == "":
input_filename = "input.txt"
stdin_redirect = input_filename
if output_filename == "":
output_filename = "output.txt"
stdout_redirect = output_filename
files_to_get = {
input_filename: self.job.testcases[test_number].input
}
allow_path = [input_filename, output_filename]
# Put the required files into the sandbox
for filename, digest in executables_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest, executable=True)
for filename, digest in files_to_get.iteritems():
sandbox.create_file_from_storage(filename, digest)
# Actually performs the execution
success, plus = evaluation_step(
sandbox,
command,
self.job.time_limit,
self.job.memory_limit,
allow_path,
stdin_redirect=stdin_redirect,
stdout_redirect=stdout_redirect)
self.job.evaluations[test_number] = {'sandboxes': [sandbox.path],
'plus': plus}
outcome = None
text = None
evaluation = self.job.evaluations[test_number]
# Error in the sandbox: nothing to do!
if not success:
pass
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(plus):
outcome = 0.0
text = human_evaluation_message(plus)
if self.job.get_output:
evaluation['output'] = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not sandbox.file_exists(output_filename):
outcome = 0.0
text = "Execution didn't produce file %s" % \
(output_filename)
if self.job.get_output:
evaluation['output'] = None
else:
# If asked so, put the output file into the storage
if self.job.get_output:
evaluation['output'] = sandbox.get_file_to_storage(
output_filename,
"Output file for testcase %d in job %s" %
(test_number, self.job.info),
trunc_len=100 * 1024)
# If not asked otherwise, evaluate the output file
if not self.job.only_execution:
# Put the reference solution into the sandbox
sandbox.create_file_from_storage(
"res.txt",
self.job.testcases[test_number].output)
# Check the solution with white_diff
if self.job.task_type_parameters[2] == "diff":
outcome, text = white_diff_step(
sandbox, output_filename, "res.txt")
# Check the solution with a comparator
elif self.job.task_type_parameters[2] == "comparator":
manager_filename = "checker"
if not manager_filename in self.job.managers:
logger.error("Configuration error: missing or "
"invalid comparator (it must be "
"named 'checker')")
success = False
else:
sandbox.create_file_from_storage(
manager_filename,
self.job.managers[manager_filename].digest,
executable=True)
success, _ = evaluation_step(
sandbox,
["./%s" % manager_filename,
input_filename, "res.txt", output_filename],
allow_path=[input_filename,
"res.txt",
output_filename])
if success:
try:
outcome, text = \
extract_outcome_and_text(sandbox)
except ValueError, e:
logger.error("Invalid output from "
"comparator: %s" % (e.message,))
success = False
# Unknown evaluationg parameter!
else:
raise ValueError("Unrecognized third parameter"
" `%s' for Batch tasktype." %
self.job.task_type_parameters[2])
def evaluate(self, job, file_cacher):
"""See TaskType.evaluate."""
# f stand for first, s for second.
first_sandbox = create_sandbox(file_cacher)
second_sandbox = create_sandbox(file_cacher)
fifo_dir = tempfile.mkdtemp(dir=config.temp_dir)
fifo = os.path.join(fifo_dir, "fifo")
os.mkfifo(fifo)
os.chmod(fifo_dir, 0o755)
os.chmod(fifo, 0o666)
# First step: we start the first manager.
first_filename = "manager"
first_command = ["./%s" % first_filename, "0", fifo]
first_executables_to_get = {
first_filename:
job.executables[first_filename].digest
}
first_files_to_get = {
"input.txt": job.input
}
first_allow_path = [fifo_dir]
# Put the required files into the sandbox
for filename, digest in first_executables_to_get.iteritems():
first_sandbox.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in first_files_to_get.iteritems():
first_sandbox.create_file_from_storage(filename, digest)
first = evaluation_step_before_run(
first_sandbox,
first_command,
job.time_limit,
job.memory_limit,
first_allow_path,
stdin_redirect="input.txt",
wait=False)
# Second step: we start the second manager.
second_filename = "manager"
second_command = ["./%s" % second_filename, "1", fifo]
second_executables_to_get = {
second_filename:
job.executables[second_filename].digest
}
second_files_to_get = {}
second_allow_path = [fifo_dir]
# Put the required files into the second sandbox
for filename, digest in second_executables_to_get.iteritems():
second_sandbox.create_file_from_storage(filename,
digest,
executable=True)
for filename, digest in second_files_to_get.iteritems():
second_sandbox.create_file_from_storage(filename, digest)
second = evaluation_step_before_run(
second_sandbox,
second_command,
job.time_limit,
job.memory_limit,
second_allow_path,
stdout_redirect="output.txt",
wait=False)
# Consume output.
wait_without_std([second, first])
# TODO: check exit codes with translate_box_exitcode.
success_first, first_plus = \
evaluation_step_after_run(first_sandbox)
success_second, second_plus = \
evaluation_step_after_run(second_sandbox)
job.sandboxes = [first_sandbox.path,
second_sandbox.path]
job.plus = second_plus
success = True
outcome = None
text = None
# Error in the sandbox: report failure!
if not success_first or not success_second:
success = False
# Contestant's error: the marks won't be good
elif not is_evaluation_passed(first_plus) or \
not is_evaluation_passed(second_plus):
outcome = 0.0
if not is_evaluation_passed(first_plus):
text = human_evaluation_message(first_plus)
else:
text = human_evaluation_message(second_plus)
if job.get_output:
job.user_output = None
# Otherwise, advance to checking the solution
else:
# Check that the output file was created
if not second_sandbox.file_exists('output.txt'):
outcome = 0.0
text = [N_("Evaluation didn't produce file %s"), "output.txt"]
if job.get_output:
job.user_output = None
else:
# If asked so, put the output file into the storage
if job.get_output:
job.user_output = second_sandbox.get_file_to_storage(
"output.txt",
"Output file in job %s" % job.info)
# If not asked otherwise, evaluate the output file
if not job.only_execution:
# Put the reference solution into the sandbox
second_sandbox.create_file_from_storage(
"res.txt",
job.output)
# If a checker is not provided, use white-diff
if self.parameters[0] == "diff":
outcome, text = white_diff_step(
second_sandbox, "output.txt", "res.txt")
elif self.parameters[0] == "comparator":
if TwoSteps.CHECKER_FILENAME not in job.managers:
logger.error("Configuration error: missing or "
"invalid comparator (it must be "
"named `checker')",
extra={"operation": job.info})
success = False
else:
second_sandbox.create_file_from_storage(
TwoSteps.CHECKER_FILENAME,
job.managers[TwoSteps.CHECKER_FILENAME].digest,
executable=True)
# Rewrite input file, as in Batch.py
try:
second_sandbox.remove_file("input.txt")
except OSError as e:
assert not second_sandbox.file_exists(
"input.txt")
second_sandbox.create_file_from_storage(
"input.txt",
job.input)
success, _ = evaluation_step(
second_sandbox,
[["./%s" % TwoSteps.CHECKER_FILENAME,
"input.txt", "res.txt", "output.txt"]])
if success:
try:
outcome, text = extract_outcome_and_text(
second_sandbox)
except ValueError, e:
logger.error("Invalid output from "
"comparator: %s", e.message,
extra={"operation": job.info})
success = False
else:
raise ValueError("Uncrecognized first parameter"
" `%s' for TwoSteps tasktype." %
self.parameters[0])
