[anujsingh-cse]

Proposed Commit Message

chore: add type annotations to DataSourceConfigDrive

Removing cloudinit.sources.DataSourceConfigDrive from the mypy
override list in pyproject.toml surfaces four errors.

The base class declares self.ec2_metadata = UNSET (a str sentinel), so
mypy types it as str; ConfigDrive assigns None in __init__ and
replaces it with an optional dict from results.get("ec2-metadata").
Annotate the base declaration as Optional[Any], matching the adjacent
self.network_json: Optional[str] = UNSET which already carries the
same sentinel-or-data pattern. The only reader compares against UNSET,
and the only other writer (DataSourceOpenStack) stays in the override
list until its own slice.

_get_subplatform calls self.source.startswith("/dev") and
self.source is Optional; the method is only invoked after get_data()
succeeded and set source, so raise ValueError on the impossible None
branch instead of crashing with AttributeError.

read_config_drive builds a list of (reader method, args, kwargs)
tuples that mypy cannot infer from an empty-containing literal;
annotate finders.

Refs GH-5445

Additional Context

Refs GH-5445. Claimed in #5445 (comment).

Removing the module from the override list surfaces four errors:

DataSourceConfigDrive.py:51: error: Incompatible types in assignment
                (expression has type "None", variable has type "str")  [assignment]
DataSourceConfigDrive.py:138: error: Incompatible types in assignment
                (expression has type "Any | None", variable has type "str")  [assignment]
DataSourceConfigDrive.py:197: error: Item "None" of "Any | None"
                has no attribute "startswith"  [union-attr]
DataSourceConfigDrive.py:206: error: Need type annotation for "finders"  [var-annotated]

• The first two come from the base class typing
  self.ec2_metadata = UNSET ("_unset", a str). The attribute
  genuinely holds the sentinel, a metadata dict, or None across
  datasources, so the declaration in cloudinit/sources/__init__.py
  is annotated Optional[Any] = UNSET, the same sentinel-or-data
  pattern the adjacent self.network_json: Optional[str] = UNSET
  already uses. The only reader (__init__.py:551) compares against
  UNSET, and the only other writer (DataSourceOpenStack.py) remains
  in the override list until its own slice.
• _get_subplatform reads self.source.startswith("/dev") where
  source is Optional. The method is only invoked after a successful
  get_data(), which always sets source, so the impossible None
  branch raises ValueError with a clear message instead of an
  uncaught AttributeError.
• finders is a list of (reader method, args, kwargs) tuples that
  mypy cannot infer; it is annotated.

Test Steps

$ python -m mypy cloudinit/ tests/ tools/
(0 errors in DataSourceConfigDrive and cloudinit/sources/__init__;
 the 140 remaining errors in the local Windows run are the
 pre-existing environment noise that also appears on untouched main)

$ pytest tests/unittests/sources/test_configdrive.py tests/unittests/sources/test_openstack.py -q
61 passed, 5 failed — the same 5 failures occur on untouched main
(Windows sandboxing/network tests); no regressions from this change.

$ ruff check / black --check / isort --check-only / pylint (changed files)
All green

tox -e py3 and tox -e check_format to be confirmed by CI.

Merge type

• Squash merge using "Proposed Commit Message"
• Rebase and merge unique commits. Requires commit messages per-commit each referencing the pull request number (#<PR_NUM>)

[nryanl]

See review comment in #7118, and also this commentary in #5445.

[anujsingh-cse]

Rebased onto current main by merge — head 4d3d3c9, no force-push after review. The duplicate tests/unittests/sources/test_azure_helper.py change is gone (empty diff against main).

All 13 functions in DataSourceConfigDrive.py are annotated. Any remains only in the parsed-data dicts (sys_cfg, md, results, the config-drive payload), which are heterogeneous config/metadata; the finder list is now List[Callable[[], Dict[str, Any]]] rather than Callable[..., Any], and read_config_drive returns Dict[str, Any] for the same reason as the other datasource PRs.

The one supporting change outside the datasource is DataSource.ec2_metadata: Optional[Any] in cloudinit/sources/__init__.py: the base initializes the attribute to the UNSET string sentinel while subclasses assign parsed dicts (here Optional[Dict[str, Any]]), so the base annotation must admit both shapes.

Gates on this head: black 26.5.1, ruff and mypy clean. test_configdrive.py: 23 passed; the 3 failures are the known Windows symlink ones — the failing test set is identical on unmodified main. The fresh CI runs will need approval to start.