fix: Return empty list instead of None from _get_required_fields for schemas without properties

[ItsMacto]

Please ensure you have read the contribution guide before creating a pull request.

Link to Issue or Description of Change

1. Link to an existing issue (if applicable):

• Closes: BUG: _get_required_fields returns None instead of [] for empty properties, causing TypeError #5920
• Related: fix(_function_parameter_parse_util): return [] not None for empty schema properties #5921 (earlier PR proposing the same change, stalled on an unsigned CLA since May; this is independent work with regression tests)

2. Or, if no issue exists, describe the change:

Problem:
_get_required_fields in tools/_function_parameter_parse_util.py is annotated -> list[str] but has a bare return when the schema has no properties, so it returns None. Any caller that relies on the declared contract (e.g. iterating the result, as reported in #5920) gets TypeError: 'NoneType' object is not iterable. A maintainer confirmed the bare return as the culprit in the issue thread and endorsed return [].

Solution:
Return [] instead, honoring the declared return type. One-line change.

One honest scoping note: the two current callers in _automatic_function_calling_util.py guard the call behind a truthy properties dict, so today the empty branch is only reachable by calling the util directly (a function whose sole parameter is tool_context produces parameters=None rather than required=None). The fix corrects the contract violation at its source and hardens the util for present and future callers; the regression test pins the contract at the util level, and an added declaration-path test documents the tool_context-only behavior.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

Three tests added to tests/unittests/tools/test_build_function_declaration.py:

• _get_required_fields returns [] (not None) for a schema without properties — fails on the unfixed code (assert None == []), passes with the fix.
• _get_required_fields returns the correct non-empty list for a schema with properties (guards against over-correction).
• A tool_context-only function's declaration has parameters is None (documents the actual public-path behavior).

# before fix:
FAILED test_get_required_fields_no_properties_returns_empty_list - assert None == []

# with fix:
$ pytest tests/unittests/tools/test_build_function_declaration.py -q
41 passed

pyink --check and isort --check-only are clean on both touched files.

Manual End-to-End (E2E) Tests:

Not applicable beyond the unit level — the change is a pure return-value correction in a private utility with no runtime surface of its own; the declaration-path test covers the integrated behavior.

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.

Additional context

None.

[ItsMacto]

/gemini review

[adk-bot]

Response from ADK Triaging Agent

Hello @ItsMacto, thank you for creating this PR!

This PR is extremely well-structured and fully complies with our contribution guidelines:

• It links to the associated issue (BUG: _get_required_fields returns None instead of [] for empty properties, causing TypeError #5920).
• It provides a thorough testing plan along with unit test results.
• The CLA check has already succeeded.

I have labeled this PR with the tools label because it addresses tool function parameter parsing (tools/_function_parameter_parse_util.py). I have also assigned @xuanyang15 as the shepherd for this component.

Thank you for contributing!

[ItsMacto]

Note on the Mypy check failures: the flagged "new" errors are in src/google/adk/models/lite_llm.py, which this PR does not touch. The baseline-vs-PR error diff appears to be unstable for that file (it also only fails on 3.10–3.13, passing on 3.14). All unit tests, the linter, and A2A checks pass. Happy to rebase/retrigger if that helps.