add a function for the centralised list of shell/command options

[nama1arpit]

SUMMARY

Fixes #73005

This PR

• adds cmd argument to the shell/command options
• moves the list of shell module options to a centralized function for usage in splitter.py
• adds an integration test

ISSUE TYPE

• Bugfix Pull Request

ADDITIONAL INFORMATION

Integration testing:

$ bin/ansible-test integration -v command_shell
...
TASK [command_shell : execute a shell command with parameter-oriented invocation using "cmd"] ***
changed: [testhost] => {"changed": true, "cmd": "echo test", "delta": "0:00:00.002359", "end": "2023-08-22 10:49:54.248467", "msg": "", "rc": 0, "start": "2023-08-22 10:49:54.246108", "stderr": "", "stderr_lines": [], "stdout": "test", "stdout_lines": ["test"]}

TASK [command_shell : Assert the shell with parameter-oriented invocation ran as expected] ***
ok: [testhost] => {
    "changed": false,
    "msg": "All assertions passed"
}
...
PLAY RECAP *********************************************************************
testhost                   : ok=91   changed=39   unreachable=0    failed=0    skipped=3    rescued=0    ignored=4   

Sanity testing:

$ bin/ansible-test sanity
...
Running sanity test "yamllint"
WARNING: Reviewing previous 25 warning(s):
WARNING: The validate-modules sanity test cannot compare against the base commit because it was not detected.
WARNING: Skipping tests disabled by default without --allow-disabled: package-data
WARNING: Skipping sanity test "compile" on Python 2.7 because it could not be found.
WARNING: Skipping sanity test "compile" on Python 3.6 because it could not be found.
WARNING: Skipping sanity test "compile" on Python 3.7 because it could not be found.
WARNING: Skipping sanity test "compile" on Python 3.8 because it could not be found.
WARNING: Skipping sanity test "compile" on Python 3.9 because it could not be found.
WARNING: Skipping sanity test "compile" on Python 3.11 because it could not be found.
WARNING: Skipping sanity test "compile" on Python 3.12 because it could not be found.
WARNING: Skipping sanity test "import" on Python 2.7 because it could not be found.
WARNING: Skipping sanity test "import" on Python 3.6 because it could not be found.
WARNING: Skipping sanity test "import" on Python 3.7 because it could not be found.
WARNING: Skipping sanity test "import" on Python 3.8 because it could not be found.
WARNING: Skipping sanity test "import" on Python 3.9 because it could not be found.
WARNING: Skipping sanity test "import" on Python 3.11 because it could not be found.
WARNING: Skipping sanity test "import" on Python 3.12 because it could not be found.
WARNING: Skipping sanity test "mypy" on Python 2.7 because it is unsupported. Supported Python versions: 3.6, 3.7, 3.8, 3.9, 3.10, 3.11, 3.12
WARNING: Skipping sanity test "mypy" on Python 3.6 because it could not be found.
WARNING: Skipping sanity test "mypy" on Python 3.7 because it could not be found.
WARNING: Skipping sanity test "mypy" on Python 3.8 because it could not be found.
WARNING: Skipping sanity test "mypy" on Python 3.9 because it could not be found.
WARNING: Skipping sanity test "mypy" on Python 3.11 because it could not be found.
WARNING: Skipping sanity test "mypy" on Python 3.12 because it could not be found.
WARNING: Required program "pwsh" not found.
WARNING: Required program "shellcheck" not found.

[nama1arpit]

Added the changelog fragment

[mattclay]

Although the current list of key-value arguments is out-of-date, I don't think that's a good reason to move them into module_utils. The main reason is that the list of valid arguments varies by module. There are currently 6 different modules that make use of check_raw:

ansible/lib/ansible/constants.py

Lines 116 to 117 in 116948c

	MODULE_REQUIRE_ARGS = tuple(add_internal_fqcns(('command', 'win_command', 'ansible.windows.win_command', 'shell', 'win_shell',
	'ansible.windows.win_shell', 'raw', 'script')))

While there are arguments in common between those modules, they're not all the same. For example,
the win_command module does not accept strip_empty_ends.

It seems like we'd be better off checking the actual list of arguments accepted by the module we're parsing for, and use those when calling parse_kv, instead of using check_raw to activate a hard-coded list.

[mattclay]

It looks like the approach in this PR was suggested by @sivel in #73005 (comment) -- although that seems to have been based on a ~9 year old code comment that predates the split to collections:

ansible/lib/ansible/parsing/splitter.py

Lines 90 to 91 in 29aea9f

	# FIXME: make the retrieval of this list of shell/command
	# options a function, so the list is centralized

[nama1arpit]

I understand the concern and I would need some more guidance to resolve this as I am not very familiar with the codebase. I can see that parse_kv is used in a few places. Should I try to get the actual list of arguments accepted by the module here?

ansible/lib/ansible/cli/console.py

Lines 204 to 206 in 29aea9f

	try:
	check_raw = module in C._ACTION_ALLOWS_RAW_ARGS
	task = dict(action=dict(module=module, args=parse_kv(module_args, check_raw=check_raw)), timeout=self.task_timeout)

[mattclay]

Currently there isn't a good way to get the supported action/module args from the action name while parsing args.
Rather than implementing that (a non-trivial task), one option would be to add cmd to the existing hard-coded list used by the splitter when check_raw is enabled.

While that would allow resolving #73005, it would also have the unwanted affect of changing how the raw action is handled.

For example, a playbook with a task such as the following would be affected:

    - raw: some_command -o cmd=something

Currently this results in running: some_command -o cmd=something

With cmd in the list of args used by check_raw, it would instead become: some_command -o

This should be discussed further before we commit to a solution.

[nitzmahone]

Some planned features may make this more generically possible in the not-too-distant future.

[bcoca]

depends on #79720