tests/test_bundle.py

from __future__ import annotations

import os
import subprocess
import sys
from pathlib import Path

import pytest

from autocad_bench.bundle import (
    BUNDLE_FILENAME,
    BUNDLE_VERSION,
    BundleError,
    export_bundle,
    validate_bundle,
)
from autocad_bench.common.paths import BENCHMARK_ROOT
from autocad_bench.common.runtime_contract import DEFAULT_EVALUATOR_VERSION
from autocad_bench.tasks.manifest import load_manifest


def test_run_ready_bundle_is_portable_allowlisted_and_integrity_bound(
    tmp_path: Path,
) -> None:
    output = tmp_path / "bundle"
    exported = export_bundle(
        output,
        evaluator_versions=(DEFAULT_EVALUATOR_VERSION,),
    )

    assert exported.bundle_version == BUNDLE_VERSION
    assert exported.task_count == 50
    assert exported.evaluator_versions == (DEFAULT_EVALUATOR_VERSION,)
    assert len(exported.files) == 204
    assert (output / BUNDLE_FILENAME).is_file()
    assert len(load_manifest(root=output)) == 50
    assert all(not entry.path.startswith("configs/") for entry in exported.files)
    assert all(".env" not in entry.path for entry in exported.files)

    validated = validate_bundle(output)
    assert validated == exported

    subprocess_result = subprocess.run(
        [
            sys.executable,
            "-c",
            (
                "from autocad_bench.tasks.manifest import "
                "BENCHMARK_ROOT, load_manifest; "
                "print(BENCHMARK_ROOT); print(len(load_manifest()))"
            ),
        ],
        cwd=tmp_path,
        env={**os.environ, "AUTOCAD_BENCH_ROOT": str(output)},
        check=True,
        capture_output=True,
        text=True,
    )
    assert subprocess_result.stdout.splitlines() == [str(output), "50"]

    manifest_path = output / "tasks" / "manifest.jsonl"
    original = manifest_path.read_bytes()
    manifest_path.write_bytes(original + b"\n")
    with pytest.raises(BundleError, match="size changed"):
        validate_bundle(output)
    manifest_path.write_bytes(original)

    unexpected = output / "credentials.env"
    unexpected.write_text("SECRET=not-allowed\n")
    with pytest.raises(BundleError, match="unexpected=.*credentials.env"):
        validate_bundle(output)


def test_bundle_export_refuses_to_overlay_an_existing_directory(
    tmp_path: Path,
) -> None:
    output = tmp_path / "existing"
    output.mkdir()

    with pytest.raises(BundleError, match="already exists"):
        export_bundle(output, source_root=BENCHMARK_ROOT)