fix(tests): #107 — isolate SYNTHBANSHEE_* env vars in CliRunner tests

[shaypal5]

Problem

Running pytest tests/unit/ with the repo's .envrc sourced silently writes
generated clip artifacts into ~/clones/avdp-synth-corpus/data/he/ instead of
the test's tmp_path. Surfaced during the delivery-003 regen — 6 zero-content
WAVs (192 KB each) leaked into the corpus, one overwriting a legitimate clip.

Root cause: Click options for --output-dir / --cache-dir / --script-cache-dir
in synthbanshee/cli.py declare envvar=SYNTHBANSHEE_*_DIR. When a CliRunner-
backed test does not override the flag, Click reads the parent-process env and
writes outside tmp_path.

Solution

Add an autouse (function-scoped) pytest fixture in tests/conftest.py that
strips SYNTHBANSHEE_DATA_DIR, SYNTHBANSHEE_CACHE_DIR, and
SYNTHBANSHEE_SCRIPT_CACHE_DIR from os.environ for the duration of every
test. No production behaviour change.

Changes per file

File	Change
tests/conftest.py	New autouse fixture _isolate_synthbanshee_env_vars that calls monkeypatch.delenv(...) for each of the three vars
tests/unit/test_cli.py	New TestSynthbansheeEnvVarIsolation class with two regression tests

Test plan

• .venv/bin/pytest tests/unit/test_cli.py::TestSynthbansheeEnvVarIsolation -v — 2/2 passing
• Same tests rerun with SYNTHBANSHEE_DATA_DIR=/tmp/should_not_leak_here and friends set in the parent env — still 2/2 passing (proves the fixture actually defeats the leak vector)
• .venv/bin/pytest tests/unit/ -q — 1698/1698 passing (no regressions)
• ruff check tests/conftest.py tests/unit/test_cli.py — clean
• mypy tests/conftest.py — clean
• Pre-commit hooks (ruff, ruff-format, mypy) — clean at commit time

Tier-3 ASR sanity (local)

Not applicable — this PR is tests-only and touches no audio rendering code
(synthbanshee/tts/, synthbanshee/script/, synthbanshee/augment/, or the
speaker / scene / acoustic / project YAML configs). Per CLAUDE.md "ASR sanity
check policy", this PR is exempt from the local qa-report --asr run.

Fixes #107.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR addresses test isolation for Click envvar= defaults in synthbanshee/cli.py, preventing CliRunner-based unit tests from unintentionally writing generated artifacts to a developer’s corpus directory when SYNTHBANSHEE_*_DIR env vars are set (fixes #107).

Changes:

• Add an autouse pytest fixture that removes SYNTHBANSHEE_DATA_DIR, SYNTHBANSHEE_CACHE_DIR, and SYNTHBANSHEE_SCRIPT_CACHE_DIR from os.environ during each test.
• Add two “regression” tests intended to assert the fixture’s behavior.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
tests/conftest.py	Adds an autouse fixture to delete SYNTHBANSHEE_*_DIR env vars during tests to prevent Click envvar default leakage.
tests/unit/test_cli.py	Adds a new test class intended to regress #107 by asserting env vars are unset and that generate doesn’t leak artifacts.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

pr-agent-context report:

This is a refreshed snapshot of the current PR state.

This run includes unresolved review comments on PR #110 in repository https://github.com/DataHackIL/SynthBanshee

For each unresolved review comment, recommend one of: resolve as irrelevant, accept and implement
the recommended solution, open a separate issue and resolve as out-of-scope for this PR, accept and
implement a different solution, or resolve as already treated by the code.

After I reply with my decision per item, implement the accepted actions, resolve the corresponding
PR comments, and push all of these changes in a single commit.

# Copilot Comments

## COPILOT-1
Location: tests/unit/test_cli.py
URL: https://github.com/DataHackIL/SynthBanshee/pull/110#discussion_r3224080238
Status: outdated
Root author: copilot-pull-request-reviewer

Comment:
    `test_clirunner_generate_does_not_leak_outside_cli_dirs` doesn’t currently validate the #107 leak vector: it never sets any `SYNTHBANSHEE_*_DIR` env var and it passes `--output-dir/--cache-dir/--script-cache-dir` explicitly, so `leak_target` is not a path the CLI would ever choose. As written, this test will pass even if env-var isolation regresses. Consider either (a) removing `leak_target` and instead asserting that every file created by the command is under the explicitly provided dirs, or (b) using a subprocess-based approach (e.g., `pytester` running a fresh pytest process with the env vars set) to prove the autouse fixture defeats parent-process env defaults.

## COPILOT-2
Location: tests/unit/test_cli.py
URL: https://github.com/DataHackIL/SynthBanshee/pull/110#discussion_r3224080282
Status: outdated
Root author: copilot-pull-request-reviewer

Comment:
    `test_env_vars_are_unset_in_test_process` will also pass if the autouse fixture is accidentally removed, as long as the CI/dev environment simply doesn’t have these vars set. If the goal is a true regression test for #107, the test needs to arrange for the vars to be present *before* the code under test runs (typically via a subprocess-based test run with a controlled environment) and then assert they’re stripped inside that run.

Run metadata:

Tool ref: v4
Tool version: 4.0.21
Trigger: schedule
Workflow run: 25736919518 attempt 1
Comment timestamp: 2026-05-12T13:16:20.398958+00:00
PR head commit: 90a0ec4de82badb444d627ec9e05ae755e9f2f9e