test(hooks): isolate test_hooks_cli from test_cli side effect (#1510)

[igorls]

Problem

Nine save/log/precompact tests in test_hooks_cli.py only pass in the full suite because test_cli.py (alphabetically earlier) creates ~/.mempalace in the session tmp HOME as a side effect, satisfying the _palace_root_exists() kill-switch. Run in isolation (pytest tests/test_hooks_cli.py) they short-circuit: 9 failed, 80 passed, 1 skipped.

Fix

Module-scoped autouse fixture that points PALACE_ROOT/STATE_DIR at a per-test palace root that exists, so every test is robust standalone and future tests don't inherit the same trap. Tests that exercise the absent-root kill-switch call _redirect_palace_root (or set their own root) after the fixture; monkeypatch's last-write-wins keeps their absent/file root and teardown still restores the real module value. Chose this over decorating the 9 tests (fragile — new tests re-trap) and over mkdir-ing real ~/.mempalace (the coupling the issue warned against).

Verification

• Issue repro: pytest tests/test_hooks_cli.py → 100 passed, 1 skipped (was 9 failed).
• No ordering regression: pytest tests/test_cli.py tests/test_hooks_cli.py → 165 passed, 1 skipped.
• ruff 0.15.9 check + format clean.

Closes #1510

[Copilot]

Pull request overview

Adds a module-level autouse fixture in tests/test_hooks_cli.py that points PALACE_ROOT/STATE_DIR at a per-test existing directory, eliminating the hidden ordering dependency on test_cli.py creating ~/.mempalace (issue #1510).

Changes:

• New _isolated_existing_palace_root autouse fixture sets PALACE_ROOT, STATE_DIR, and resets _state_dir_initialized.
• Relies on monkeypatch last-write-wins so absent-root kill-switch tests calling _redirect_palace_root after the fixture still work.

---

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

[gemini-code-assist]

Code Review

This pull request introduces an autouse pytest fixture in tests/test_hooks_cli.py to isolate the PALACE_ROOT and STATE_DIR environments, preventing tests from relying on or affecting the user's local filesystem. The reviewer identified that _MINE_PID_DIR also needs to be patched to avoid filesystem leaks and recommended explicitly creating the state directory to match the fixture's documentation.

[gemini-code-assist]

The fixture patches PALACE_ROOT and STATE_DIR, but it misses _MINE_PID_DIR. Since _MINE_PID_DIR is initialized at module import time (line 277 of hooks_cli.py), it retains its original value pointing to the real ~/.mempalace even after STATE_DIR is patched. This causes tests that spawn mine processes (like _spawn_mine or _ingest_transcript) to create PID files in the real user directory, leading to filesystem leaks and potential conflicts with a running instance.

Additionally, the docstring promises an existing STATE_DIR, but the code only creates the PALACE_ROOT. I've updated the suggestion to create both and ensure _MINE_PID_DIR is also isolated.

Suggested change

	root = tmp_path / ".mempalace"
	root.mkdir(exist_ok=True)
	monkeypatch.setattr(hooks_cli_mod, "PALACE_ROOT", root)
	monkeypatch.setattr(hooks_cli_mod, "STATE_DIR", root / "hook_state")
	monkeypatch.setattr(hooks_cli_mod, "_state_dir_initialized", False)
	root = tmp_path / ".mempalace"
	state_dir = root / "hook_state"
	state_dir.mkdir(parents=True, exist_ok=True)
	monkeypatch.setattr(hooks_cli_mod, "PALACE_ROOT", root)
	monkeypatch.setattr(hooks_cli_mod, "STATE_DIR", state_dir)
	monkeypatch.setattr(hooks_cli_mod, "_MINE_PID_DIR", state_dir / "mine_pids")
	monkeypatch.setattr(hooks_cli_mod, "_state_dir_initialized", False)