rune/tests/test_testing_mode.py

"""test_testing_mode.py — T19 §19L/19M/19N/19O/19P held-outs."""

from __future__ import annotations

import sys
from pathlib import Path

import pytest

REPO = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(REPO / "src"))

from rune import behaviour_ledger as bl  # noqa: E402
from rune import forge  # noqa: E402
from rune import testing_mode as tm  # noqa: E402


@pytest.fixture()
def env(tmp_path):
    beh = bl.BehaviourLedger(tmp_path)
    beh.define(behaviour_id="T-1", owning_project="p",
               intended_behaviour="first behaviour",
               procedure="run the thing; expect output X",
               pass_criteria="output == X", fail_criteria="output != X",
               owner_component="component:shared",
               dependencies=["component:shared"])
    beh.define(behaviour_id="T-2", owning_project="p",
               intended_behaviour="second behaviour")
    beh.save()
    return beh


def test_19n_session_is_coordination_not_authority(tmp_path, env):
    s = tm.TestingSession(tmp_path, project="p")
    item = s.next_item(env)
    assert item["behaviour_id"] == "behaviour:T-1"  # FAIL items none yet;
    # UNVALIDATED sorts after — both outstanding, T-1 first alphabetically
    s.record(env, "T-1", passed=True, detail="checked by hand")
    assert env.get("T-1")["state"] == bl.USER_CONFIRMED
    s.pause()
    assert s.next_item(env) is None  # paused
    s.resume()
    assert s.next_item(env)["behaviour_id"] == "behaviour:T-2"
    # session file holds coordination, NOT spec truth
    import json
    sess = json.loads((tmp_path / "testing_session.json").read_text())
    assert "intended_behaviour" not in str(sess)
    # durable truth intact in the ledger
    assert env.get("T-1")["intended_behaviour"] == "first behaviour"


def test_19m_fix_reopens_dependent_confirmation(tmp_path, env):
    """Test-B-fix-invalidates-Test-A (19S-9): A confirmed; B fails; the
    B fix touches a dependency shared with A → A re-enters the queue."""
    env.record_auto_verification("T-1", passed=True, generation="g1")
    env.record_real_world("T-1", passed=True, generation="g1")  # A hardened
    # B fails in the field
    env.record_real_world("T-2", passed=False, detail="broke")
    assert env.get("T-2")["state"] == bl.TEST_FAILED
    # the fix structurally touches A's dependency
    invalidated = env.invalidate_for_change(
        changed_entities={"component:shared"}, generation="g2",
        reasons={"behaviour:T-1": "fix for T-2 rewrote shared component"})
    assert [r["behaviour_id"] for r in invalidated] == ["behaviour:T-1"]
    assert env.get("T-1")["state"] == bl.RETEST_REQUIRED
    # dynamic queue now contains BOTH
    ids = [b["behaviour_id"] for b in env.outstanding()]
    assert "behaviour:T-1" in ids and "behaviour:T-2" in ids
    # A's generation-1 confirmation remains queryable
    hist = env.retest_history("T-1")
    assert any(h["generation"] == "g1" and h["event"].endswith(
        "USER_CONFIRMED") for h in hist)


def test_19s10_no_blanket_invalidation(tmp_path, env):
    env.record_real_world("T-1", passed=True)
    env.record_real_world("T-2", passed=True)
    # a change to an unrelated entity invalidates NOTHING
    invalidated = env.invalidate_for_change(
        changed_entities={"component:unrelated-thing"}, generation="g")
    assert invalidated == []
    assert env.get("T-1")["state"] == bl.USER_CONFIRMED
    assert env.get("T-2")["state"] == bl.USER_CONFIRMED


def test_19o_observation_to_closure(tmp_path):
    obs = tm.record_observation(
        tmp_path, observed="button does nothing",
        expected="button saves the form",
        behaviour_id="behaviour:T-1", environment="production",
        reproduction="click save with empty field",
        candidate_areas=["component:form-handler"])  # hypothesis only
    assert obs["candidate_areas"] == ["component:form-handler"]
    # premature closure (no user confirm) is recorded as NOT closed
    r = tm.close_observation(tmp_path, obs["observation_id"],
                             behaviour_retest_passed=True,
                             user_confirmed=False)
    assert r["closure"]["closed"] is False  # automated green ≠ closure
    # full closure requires retest AND user confirmation
    r2 = tm.close_observation(tmp_path, obs["observation_id"],
                              behaviour_retest_passed=True,
                              user_confirmed=True)
    assert r2["closure"]["closed"] is True


def test_19p_forge_projection_never_rewrites_truth(tmp_path, env):
    """19S-14: forge unavailable → projection fails loudly; ledger truth
    unchanged."""
    obs = tm.record_observation(tmp_path, observed="x", expected="y",
                                behaviour_id="behaviour:T-1")
    before = env.get("T-1")["state"]
    bad = forge.GiteaForge(server="https://invalid.invalid")
    with pytest.raises(forge.ForgeUnavailable):
        tm.project_failure_to_forge(bad, "conductor/rune",
                                    behaviour_id="T-1",
                                    observation=obs)
    assert env.get("T-1")["state"] == before  # truth untouched


def test_19l_aggregate_across_projects(tmp_path):
    reg = tm.ProjectRegistry(tmp_path)
    p1 = tmp_path / "p1"
    p1.mkdir()
    beh = bl.BehaviourLedger(p1)
    beh.define(behaviour_id="X-1", owning_project="p1",
               intended_behaviour="x")
    beh.save()
    reg.register(project_id="p1", scaffold_location=str(p1))
    reg.register(project_id="p2", scaffold_location=str(tmp_path / "p2"))
    rows = tm.aggregate(reg, tmp_path)
    by_p = {r["project"]: r for r in rows}
    assert by_p["p1"]["outstanding"] == 1
    assert by_p["p2"]["outstanding"] is None  # no ledger yet