fix(dwh): bind snapshots to validated bytes

2026-07-12 07:31:46 +02:00
parent 2f6daaaea6
commit 4aae6c433d
3 changed files with 136 additions and 15 deletions
@@ -524,6 +524,90 @@ def test_snapshot_root_swap_after_lease_never_reads_replacement(monkeypatch, tmp
    assert (replacement / "sentinel").read_text() == "replacement-secret"


+def test_snapshot_copies_each_validated_artifact_once_without_reopen(monkeypatch, tmp_path):
+    import tht.jobs.dwh_pipeline as module
+
+    pipeline = DwhPreprocessPipeline(
+        workspace_id="demo", workspace_root=tmp_path,
+        config_fingerprint=FP, input_fingerprint=FP,
+        introspect=lambda output: output.write_text("trusted"),
+        build_lsh=lambda physical, output: [
+            (output / name).write_text("trusted")
+            for name in ("demo_lsh.pkl", "demo_minhashes.pkl", "demo_meta.json")
+        ],
+    )
+    assert pipeline.run().status == "succeeded"
+    real_read = module._read_owned_at
+    reads = {}
+
+    def mutate_on_reopen(directory_fd, name, *, readonly):
+        reads[name] = reads.get(name, 0) + 1
+        if name.endswith(".pkl") and reads[name] > 1:
+            return b"MALICIOUS_PICKLE"
+        return real_read(directory_fd, name, readonly=readonly)
+
+    monkeypatch.setattr(module, "_read_owned_at", mutate_on_reopen)
+    with lease_dwh_snapshot(snapshot_config(tmp_path)) as snapshot:
+        assert (snapshot.lsh_dir / "demo_lsh.pkl").read_text() == "trusted"
+        assert "MALICIOUS" not in (snapshot.lsh_dir / "demo_lsh.pkl").read_text()
+    assert all(count == 1 for count in reads.values())
+
+
+def test_reconcile_mismatch_closes_active_generation_fd(monkeypatch, tmp_path):
+    import os
+    from types import SimpleNamespace
+    import pytest
+    import tht.jobs.dwh_pipeline as module
+
+    pipeline = DwhPreprocessPipeline(
+        workspace_id="demo", workspace_root=tmp_path,
+        config_fingerprint=FP, input_fingerprint=FP,
+        introspect=lambda output: output.write_text("catalog"),
+        build_lsh=lambda physical, output: _write_lsh([], physical, output),
+    )
+    report = pipeline.run()
+    run_dir = tmp_path / ".tht-jobs" / "dwh" / "runs" / report.run_id
+    real_active = module._active_generation_fd
+
+    def mismatched_active(root_fd, binding):
+        generation, generation_fd = real_active(root_fd, binding)
+        return "f" * 32, generation_fd
+
+    monkeypatch.setattr(module, "_active_generation_fd", mismatched_active)
+    source = SimpleNamespace(
+        run_id=report.run_id,
+        stages=(SimpleNamespace(
+            status="running", effect_state="intent", name="lsh",
+            artifact_files=("physical.yaml", "demo_lsh.pkl", "demo_minhashes.pkl",
+                            "demo_meta.json"),
+        ),),
+    )
+    before = len(os.listdir("/dev/fd"))
+    with pytest.raises(Exception, match="not ACTIVE"):
+        pipeline._reconcile_effects(source, run_dir)
+    assert len(os.listdir("/dev/fd")) == before
+
+
+def test_pipeline_releases_materialized_snapshot_after_every_run(tmp_path):
+    import tht.jobs.dwh_pipeline as module
+
+    pipeline = DwhPreprocessPipeline(
+        workspace_id="demo", workspace_root=tmp_path,
+        config_fingerprint=FP, input_fingerprint=FP,
+        introspect=lambda output: output.write_text("catalog"),
+        build_lsh=lambda physical, output: _write_lsh([], physical, output),
+    )
+    baseline = set(module._SNAPSHOT_DIRS)
+    for _ in range(3):
+        assert pipeline.run().status == "succeeded"
+        assert set(module._SNAPSHOT_DIRS) == baseline
+        assert pipeline._snapshot_holder is None
+    pipeline.introspect = lambda output: (_ for _ in ()).throw(RuntimeError("injected"))
+    assert pipeline.run().status == "failed"
+    assert set(module._SNAPSHOT_DIRS) == baseline
+    assert pipeline._snapshot_holder is None
+
+
 def test_publish_root_swap_after_lease_never_writes_replacement(monkeypatch, tmp_path):
    import tht.jobs.dwh_pipeline as module