Test: share the click-muting fixture via tests/unittests/cli/utils/conftest.py

[AmaadMartin]

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: #issue_number
   Related: #issue_number
2. Or, if no issue exists, describe the change:
   Problem: Six modules under tests/unittests/cli/utils/ each carried their own private copy of an autouse _mute_click fixture that monkeypatches click output functions to no-ops. Five of those copies offered no way to opt out, so any new test in those modules that needs to assert on click output had to shadow or work around the fixture. The unmute_click marker escape hatch existed in exactly one module.

Solution: Hoist a single shared fixture — the variant that supports the unmute_click marker opt-out — into a new tests/unittests/cli/utils/conftest.py, and delete the five redundant copies. Net result: duplication drops from six copies to two intentional definitions, and every module in the directory gains a supported opt-out. Net -3 lines.

Two deliberate deviations, both because the fixtures were not behaviourally identical:

• test_cli_tools_click.py keeps its own _mute_click. It mutes only echo and deliberately leaves secho live. Several of its tests assert on error text written with click.secho — e.g. test_cli_deploy_cloud_run_failure asserts "Deploy failed: boom" in result.output, emitted by click.secho(f"Deploy failed: {e}", fg="red", err=True) in src/google/adk/cli/cli_tools_click.py. Because a module-level fixture shadows a same-named conftest.py fixture, keeping the name _mute_click preserves today's behaviour exactly. Its docstring now records why it deviates, and the commented-out secho line that the docstring replaces is gone.
• test_cli_eval_pretty_print.py opts out with pytestmark = pytest.mark.unmute_click. It had no mute fixture before and asserts on click.echo output through capsys; a directory-scoped autouse fixture would otherwise silently break it. Module-level pytestmark rather than a per-test decorator, because every test in that module exists to assert on printed output.

Why a directory-level conftest.py is safe for the other 11 modules in the directory: none of them reference click, capsys, or CliRunner, so the newly directory-wide muting is a no-op for them.

Scope notes: no file under src/ is touched, no test is added, renamed, deleted, skipped, or xfailed, and pyproject.toml is untouched. The unmute_click marker is intentionally not registered here — it is already unregistered at HEAD (9 existing usages) and emits only a PytestUnknownMarkWarning, never an error. This change adds one new occurrence of that pre-existing warning, not a new kind of warning; registering the marker is a separate change so that the two cannot conflict in either merge order.

Collision check (required by our workflow): ran gh pr list --repo AmaadMartin/adk-python --state open --limit 100 and scanned the --name-only file list of every one of the 67 open PRs, plus grepped all 67 diffs for _mute_click. No open PR creates tests/unittests/cli/utils/conftest.py and no open PR adds or removes a _mute_click line. The nearest neighbour is #52 ("register the unmute_click pytest marker"), which touches only pyproject.toml — complementary, zero file overlap, and safe in either merge order for the reason given above. Several open PRs edit test_cli_tools_click.py (#51, #53, #62, #64), but none of them touch the fixture block at the top of the file. Branched from current main (923dee79); not stacked.

Testing Plan

Please describe the tests that you ran to verify your changes. This is required for all PRs that are not small documentation or typo fixes.

This change ships no production code, so the plan is a behaviour-preservation proof rather than new test authoring.

Unit Tests:
[x] I have added or updated unit tests for my change.
[x] All unit tests pass locally.

Behaviour preservation — identical pass counts before and after. pytest tests/unittests/cli -q -p no:randomly:

	result
Baseline (clean main, 923dee79)	1 failed, 569 passed, 5 skipped, 155 warnings
After this change	1 failed, 569 passed, 5 skipped, 156 warnings

The single failure is pre-existing on unmodified main and unrelated: test_cli_tools_click.py::test_telemetry_cli_commands, a click >= 8.2 bare-group exit-code issue (installed click is 8.4.2; PRs #51/#53/#64 address it). The 155 -> 156 warning delta is exactly the one predicted new occurrence of the pre-existing PytestUnknownMarkWarning: Unknown pytest.mark.unmute_click, raised at test_cli_eval_pretty_print.py:32.

Targeted regression proofs, each guarding one hazard this design exists to avoid — all pass:

• pytest "tests/unittests/cli/utils/test_cli_tools_click.py::test_cli_deploy_cloud_run_failure" -> 1 passed (proves click.secho is still live in that module).
• pytest tests/unittests/cli/utils/test_cli_eval_pretty_print.py -> 1 passed (proves the unmute_click opt-out keeps click.echo live).
• pytest tests/unittests/cli/utils/test_cli_deploy_ignore.py -> 3 passed (proves the shared conftest fixture is picked up by a module that no longer defines one).

Coverage of the new file: 100% line and branch.

Name                                     Stmts   Miss Branch BrPart  Cover
tests/unittests/cli/utils/conftest.py        9      0      2      0   100%

Mutation proofs — coverage is a floor, so each behaviour was proven to be pinned by a test that actually fails without it. Every mutation was reverted afterwards and the test re-run green.

1. Marker opt-out is load-bearing. Deleted the if "unmute_click" in request.keywords: return guard from conftest.py.
   -> test_cli_eval_pretty_print.py FAILED: AssertionError: assert 'Rubric: invocation-rubric' in '' / where '' = CaptureResult(out='', err='').out. Restored -> 1 passed.
2. The test_cli_tools_click.py override is load-bearing. Deleted that module's _mute_click so the conftest fixture (which mutes secho) applied to it.
   -> test_cli_deploy_cloud_run_failure FAILED: AssertionError: assert 'Deploy failed: boom' in ''. Restored -> 1 passed. This is precisely why the override is kept rather than folded into the shared fixture.
3. The mute path genuinely applies. No assertion pins "produces no output", so this was proven observationally: pytest tests/unittests/cli/utils/test_cli_deploy_ignore.py -q -s emits 0 lines matching Copying agent source code|Creating Dockerfile|Deploying to Cloud Run with conftest.py present, and 10 such lines with it removed (Copying agent source code..., Creating Dockerfile..., ...). Restored.

Lint/format, on the exact commit pushed:

• pyink==25.12.0 --config pyproject.toml --check -> 8 files would be left unchanged.
• isort==8.0.1 --settings-path pyproject.toml --check-only -> exit 0.
• No unused-import findings for test_cli_deploy_ignore.py or test_cli_deploy_to_cloud_run.py; import click was removed from exactly those two files (their only click reference was the deleted fixture) and correctly retained in the three that still use it (test_cli.py:151,513, test_cli_create.py click.Abort/click.BadParameter, test_cli_deploy.py click.ClickException/mock.patch("click.echo")).
• No type-checker or linter suppressions were added anywhere in the diff.

Manual End-to-End (E2E) Tests:
Please provide instructions on how to manually test your changes, including any necessary setup or configuration.

Not applicable — this is test-only scaffolding with no runtime behaviour change. As a manual sanity check that the shared fixture is genuinely applying, run a de-duplicated module with output capture off and confirm no click output reaches stdout:

pytest tests/unittests/cli/utils/test_cli_deploy.py -q -s

To see the opposite, temporarily rename tests/unittests/cli/utils/conftest.py and re-run — deploy progress lines will appear.

Checklist

[x] I have read the CONTRIBUTING.md document.
[x] I have performed a self-review of my own code.
[x] I have commented my code, particularly in hard-to-understand areas.
[x] I have added tests that prove my fix is effective or that my feature works.
[x] New and existing unit tests pass locally with my changes.

CI status on this PR (honest classification: failed — both failures pre-existing on main)

The real test jobs ran. Two checks are red, and neither is caused by this diff; both reproduce identically on unrelated branches, so I have deliberately not "fixed" them here (doing so would duplicate live PRs — see below).

1. Unit Tests (all of 3.10 / 3.11 / 3.12 / 3.13 / 3.14) — one pre-existing failure.

FAILED tests/unittests/cli/utils/test_cli_tools_click.py::test_telemetry_cli_commands - assert 2 == 0
= 1 failed, 9358 passed, 68 skipped, 18 xfailed, 1 xpassed (Python 3.10) =

This is the click >= 8.2 bare-group exit-code issue. Evidence it is not mine:

• It reproduces on my local baseline of clean main (923dee79) before I edited a single file: 1 failed, 569 passed, 5 skipped for tests/unittests/cli, same test.
• Unrelated PR Fix: emit LlmResponse.error_code as a plain string from the OpenAI Responses model #67 (an OpenAI Responses error_code change that touches no CLI file) fails on the same single test in CI: FAILED ...::test_telemetry_cli_commands - assert 2 == 0.
• Pass counts confirm no test was lost. PR Fix: emit LlmResponse.error_code as a plain string from the OpenAI Responses model #67 reports 9360 passed on 3.10 versus this PR's 9358. That −2 is exactly accounted for: Fix: emit LlmResponse.error_code as a plain string from the OpenAI Responses model #67 adds one test function decorated with @pytest.mark.parametrize over 2 cases. So main's true baseline is 9358, and this PR reports 9358 — unchanged. Independently, git diff main -U0 | grep -E '^[-+](async )?def test_' returns nothing: this branch adds, removes, renames, skips and xfails zero tests.
• Fixing it here is out of scope and would collide with live PRs Fix: assert help output, not exit code, for bare adk telemetry (Part 1/2) #51, Fix: make bare adk telemetry exit 0 on Click >= 8.2 #53 and Fix: print help and exit 0 for bare CLI command groups on click >= 8.2 #64, which each already address this exact test/behaviour.

2. Pre-commit Linter — pre-existing infrastructure failure, not a lint violation.

The failing hook is update-constraints, which dies with ./scripts/update_constraints.sh: line 103: uv: command not found for all five constraints files. pre-commit/action@v3.0.1 runs --all-files, so this hook fires on pyproject.toml on every PR regardless of its diff — and this PR does not touch pyproject.toml or any constraints-*.txt. Unrelated PR #67 fails on the identical hook with the identical message. Live PRs #49, #54 and #56 exist to fix this hook; duplicating them here would be scope creep.

Every hook that actually applies to my files passes — verified locally on the exact pushed commit:

$ pre-commit run --files $(git diff main --name-only)
fix end of files.........................................................Passed
trim trailing whitespace.................................................Passed
isort....................................................................Passed
pyink....................................................................Passed
addlicense...............................................................Passed
ADK Compliance Checks....................................................Passed
codespell................................................................Passed

Everything else is green: Mypy Check on 3.10 / 3.11 / 3.12 / 3.13 and A2A v0.3 Tests on 3.10 / 3.11 / 3.12 / 3.13 / 3.14 all pass.