A first answer doesn't use in a second prompt when check_all is false

[sky-joker]

SUMMARY

Maybe I don’t understand correctly, so can you please let me know?

In the document is written the following like.

In the following example, the second answer would be ignored and y would be the answer given to both prompts.

https://docs.ansible.com/ansible/latest/network/user_guide/network_working_with_command_output.html#handling-prompts-in-network-modules

In my perception, the first answer should be used even in the second prompt when check_all is false.
I tried that but the first answer wasn't used and a timeout occurs.

ISSUE TYPE

• Bug Report

COMPONENT NAME

plugins/connection/network_cli.py

ANSIBLE VERSION

$ ansible --version
ansible 2.10.1

CONFIGURATION

$ ansible-config dump --only-changed
HOST_KEY_CHECKING(/path/ansible.cfg) = False

OS / ENVIRONMENT

Cisco csr1000

STEPS TO REPRODUCE

pre condition in csr1000

I set that the following was displayed to prompt when executing reload command.

• System configuration has been modified. Save? [yes/no]:
• Proceed with reload? [confirm]

After completing the above work, I created the following playbook and execute it.

---
- name: Test Playbook
  hosts: cisco1
  gather_facts: false
  become: true
  tasks:
    - name: Execute command
      cli_command:
        command: reload
        prompt:
           - 'System configuration has been modified'
           - 'confirm'
        answer:
           - 'yes'
           - 'yes'

EXPECTED RESULTS

The first answer also used in the second prompt and reboot OS.

ACTUAL RESULTS

The first answer doesn't use in the second prompt and the time-out error occurs.

fatal: [ping]: FAILED! => {"ansible_facts": {"discovered_interpreter_python": "/usr/bin/python"}, "changed": false, "msg": "timeout value 30 seconds reached while trying to send command: b'reload'"}

As far as I'm looking at the following codes, I seem that the first answer is ignored in the second prompt if check_all is false.

ansible.netcommon/plugins/connection/network_cli.py

Lines 923 to 978 in 3aca2f0

	def _handle_prompt(
	self,
	resp,
	prompts,
	answer,
	newline,
	prompt_retry_check=False,
	check_all=False,
	):
	"""
	Matches the command prompt and responds
	:arg resp: Byte string containing the raw response from the remote
	:arg prompts: Sequence of byte strings that we consider prompts for input
	:arg answer: Sequence of Byte string to send back to the remote if we find a prompt.
	A carriage return is automatically appended to this string.
	:param prompt_retry_check: Bool value for trying to detect more prompts
	:param check_all: Bool value to indicate if all the values in prompt sequence should be matched or any one of
	given prompt.
	:returns: True if a prompt was found in ``resp``. If check_all is True
	will True only after all the prompt in the prompts list are matched. False otherwise.
	"""
	single_prompt = False
	if not isinstance(prompts, list):
	prompts = [prompts]
	single_prompt = True
	if not isinstance(answer, list):
	answer = [answer]
	prompts_regex = [re.compile(to_bytes(r), re.I) for r in prompts]
	for index, regex in enumerate(prompts_regex):
	match = regex.search(resp)
	if match:
	self._matched_cmd_prompt = match.group()
	self._log_messages(
	"matched command prompt: %s" % self._matched_cmd_prompt
	)
	# if prompt_retry_check is enabled to check if same prompt is
	# repeated don't send answer again.
	if not prompt_retry_check:
	prompt_answer = (
	answer[index] if len(answer) > index else answer[0]
	)
	self._ssh_shell.sendall(b"%s" % prompt_answer)
	if newline:
	self._ssh_shell.sendall(b"\r")
	prompt_answer += b"\r"
	self._log_messages(
	"matched command prompt answer: %s" % prompt_answer
	)
	if check_all and prompts and not single_prompt:
	prompts.pop(0)
	answer.pop(0)
	return False
	return True
	return False

ansible.netcommon/plugins/connection/network_cli.py

Lines 654 to 723 in 3aca2f0

	while True:
	if command_prompt_matched:
	try:
	signal.signal(
	signal.SIGALRM, self._handle_buffer_read_timeout
	)
	signal.setitimer(
	signal.ITIMER_REAL, self._buffer_read_timeout
	)
	data = self._ssh_shell.recv(256)
	signal.alarm(0)
	self._log_messages(
	"response-%s: %s" % (self._window_count + 1, data)
	)
	# if data is still received on channel it indicates the prompt string
	# is wrongly matched in between response chunks, continue to read
	# remaining response.
	command_prompt_matched = False
	# restart command_timeout timer
	signal.signal(signal.SIGALRM, self._handle_command_timeout)
	signal.alarm(self._command_timeout)
	except AnsibleCmdRespRecv:
	# reset socket timeout to global timeout
	self._ssh_shell.settimeout(cache_socket_timeout)
	return self._command_response
	else:
	data = self._ssh_shell.recv(256)
	self._log_messages(
	"response-%s: %s" % (self._window_count + 1, data)
	)
	# when a channel stream is closed, received data will be empty
	if not data:
	break
	recv.write(data)
	offset = recv.tell() - 256 if recv.tell() > 256 else 0
	recv.seek(offset)
	window = self._strip(recv.read())
	self._last_recv_window = window
	self._window_count += 1
	if prompts and not handled:
	handled = self._handle_prompt(
	window, prompts, answer, newline, False, check_all
	)
	self._matched_prompt_window = self._window_count
	elif (
	prompts
	and handled
	and prompt_retry_check
	and self._matched_prompt_window + 1 == self._window_count
	):
	# check again even when handled, if same prompt repeats in next window
	# (like in the case of a wrong enable password, etc) indicates
	# value of answer is wrong, report this as error.
	if self._handle_prompt(
	window,
	prompts,
	answer,
	newline,
	prompt_retry_check,
	check_all,
	):
	raise AnsibleConnectionFailure(
	"For matched prompt '%s', answer is not valid"
	% self._matched_cmd_prompt
	)

Is this correct behavior?

[Qalthos]

Yeah, the documentation seems... wrong. Presently, check_all: false will stop after the first prompt, and check_all: true will not. There may at one point have been an intention for check_all: true to require all prompts, but that does not seem to be the case presently.

I will get a PR together to clean up the various documentation to agree with the current behavior, then we probably have to have a rethink about what this option is doing if it's being described so conflictingly.