"""pytest conftest for RecreationBench AT-SPI tests (Tier-1).

Bridges pytest → the legacy scoring contract so nothing downstream changes:
each test's outcome is written into ``<module>_results.json`` in the exact shape
``{"module": m, "tests": [{"name", "passed", "message"}]}`` that
``atspi_test_runner`` / ``build_manifest`` / ``score_against_manifest`` and the
VLM half already consume. Only test *authoring* (pytest) and *execution* (run
pytest instead of ``python3 test_file``) change; the manifest, fixed-denominator
scorer, and ``vlm_assertions`` are untouched.

Provides:
- ``app``  — the AT-SPI Accessible for the app under test (``rb_atspi.get_app``),
  session-scoped (the runner launches one app per module and runs pytest on that
  module's file, so a session == one module).
- ``results_dir`` — where ``screenshot_assert`` and results land.
- pytest-check is available to tests for soft assertions (many checks per test,
  never abort on the first failure); the test == one scored entry (name = the
  test function name → manifest key ``<module>::<test_name>``).
"""

from __future__ import annotations

import json
import os
from pathlib import Path

import pytest

try:  # kit is co-located on the VM (PYTHONPATH has the tools dir); pure helpers.
    import rb_testgen_kit as _kit
except Exception:  # pragma: no cover - kit absent -> record the legacy fields only
    _kit = None

# nodeid -> {"name", "passed", "message", depth, jump_kind, expect};
# the call phase (or a failing setup) wins.
_RESULTS: dict = {}


def _results_dir() -> str:
    return os.environ.get("RB_RESULTS_DIR") or os.environ.get("RESULTS_DIR") or "."


def pytest_configure(config):
    config.addinivalue_line(
        "markers",
        "rb(jump_kind, expect, depth): declare a behavioral test's interaction tier "
        "(none/liveness/reach/transition/value), the specific observed target, and "
        "optional navigation depth (else measured from rb_enter/rb_reach calls).",
    )


def _rb_meta(item):
    """(jump_kind, expect, declared_depth) from the @pytest.mark.rb(...) marker."""
    m = item.get_closest_marker("rb")
    if m is None:
        return "none", "", None
    jk = m.kwargs.get("jump_kind") or (m.args[0] if m.args else "none") or "none"
    return jk, m.kwargs.get("expect", "") or "", m.kwargs.get("depth", None)


class _RBDepth:
    """Per-test navigation-depth tracker. A test calls ``rb.enter("<surface>")`` on each
    real surface crossing (dialog/page/submenu); ``reached`` is the MAX depth — a runtime
    count, so it cannot be faked (no source AST audit needed)."""

    def __init__(self) -> None:
        self.depth = 0
        self.reached = 0

    def enter(self, _surface: str = "") -> int:
        self.depth += 1
        self.reached = max(self.reached, self.depth)
        return self.depth

    def back(self) -> int:
        self.depth = max(0, self.depth - 1)
        return self.depth


@pytest.fixture
def rb(request):
    """Depth tracker: ``def test_x(app, rb): rb.enter("prefs")``. Fresh per test;
    pytest_runtest_makereport reads its ``reached`` into the record's ``depth``."""
    tracker = _RBDepth()
    request.node._rb = tracker
    return tracker


@pytest.fixture(scope="session")
def results_dir() -> str:
    return _results_dir()


@pytest.fixture(scope="session")
def app():
    """Attach to the app under test on the AT-SPI bus (by the detected RB_APP_NAME)."""
    import rb_atspi

    a = rb_atspi.get_app()
    if a is None:
        pytest.fail(
            f"app not found on the AT-SPI bus (RB_APP_NAME={os.environ.get('RB_APP_NAME')!r})",
            pytrace=False,
        )
    # Focus the app's window before any test drives it: a cold-relaunched window
    # (run_baseline) is often not WM-focused, so xdotool input misses — this makes the
    # cold path behave like the warm run_live instance. Best-effort, no-op if already focused.
    rb_atspi.focus_app_window(a)
    return a


@pytest.fixture(scope="session")
def frame(app):
    """The app's main top-level frame/window."""
    import rb_atspi

    return rb_atspi.get_main_frame(app)


def _shorten(s: str, n: int = 400) -> str:
    s = " ".join(str(s).split())
    return s[-n:] if len(s) > n else s


@pytest.hookimpl(hookwrapper=True)
def pytest_runtest_makereport(item, call):
    outcome = yield
    report = outcome.get_result()
    # Score the call phase; also capture a setup failure (fixture/app error) as a
    # failed test so a broken precondition scores as fail (never silently skipped).
    if report.when == "call" or (report.when == "setup" and report.outcome != "passed"):
        name = getattr(item, "originalname", None) or item.name
        passed = report.outcome == "passed"
        msg = "" if passed else _shorten(report.longreprtext or str(report.longrepr))
        # Superset fields: the frozen-manifest scorer reads only name/passed; depth/
        # jump_kind/expect are consumed by rb_audit. depth = the marker's declared value
        # if given, else the MEASURED max from the `rb` fixture (rb.enter() calls).
        jk, expect, decl_depth = _rb_meta(item)
        tracker = getattr(item, "_rb", None)  # the `rb` fixture, if the test used it
        depth = (
            decl_depth
            if decl_depth is not None
            else (tracker.reached if tracker else 0)
        )
        prev = _RESULTS.get(item.nodeid)
        if prev is None or report.when == "call":
            _RESULTS[item.nodeid] = {
                "name": name,
                "passed": passed,
                "message": msg,
                "depth": int(depth or 0),
                "jump_kind": jk,
                "expect": str(expect),
            }


def pytest_sessionfinish(session, exitstatus):
    """Write per-module <module>_results.json in the legacy contract format."""
    rd = _results_dir()
    by_module: dict[str, list] = {}
    for nodeid, rec in _RESULTS.items():
        fstem = Path(nodeid.split("::", 1)[0]).stem
        module = fstem.replace("test_atspi_", "")
        by_module.setdefault(module, []).append(rec)
    for module, tests in by_module.items():
        payload = {
            "module": module,
            "tests": tests,
            "passed": sum(1 for t in tests if t.get("passed")),
            "total": len(tests),
        }
        try:
            with open(os.path.join(rd, f"{module}_results.json"), "w") as f:
                json.dump(payload, f, indent=1)
        except Exception:
            pass


def pytest_terminal_summary(terminalreporter, exitstatus, config):
    """Finalize self-audit (REPORT-ONLY, never a gate): print the depth/rigor quota
    gaps + portability issues in pytest's own summary section, so the orchestrator can
    target a deeper reach/value test before it finalizes."""
    if _kit is None:
        return
    rd = _results_dir()
    tg = os.environ.get("RB_TESTGEN_DIR") or str(getattr(config, "rootdir", "")) or rd
    for fn, args in (
        (_kit.rb_audit, (rd,)),
        (_kit.rb_lint, (rd,)),
        (_kit.rb_lint_sources, (tg,)),
    ):
        try:
            fn(*args)
        except Exception:
            pass