Add malformed XML test and fix verify.py regression

openverifiablellm/verify.py

@@ -593,6 +593,7 @@ def verify_preprocessing(
                     actual=reproduced_manifest.get("preprocessing_version"),
                     detail="Preprocessing version tag",
                 )
+
             if "chunk_size_bytes" in manifest:
                 _check_field(
                     report,
@@ -601,6 +602,14 @@ def verify_preprocessing(
                     actual=reproduced_manifest.get("chunk_size_bytes"),
                     detail="Merkle chunk size used during preprocessing",
                 )
+            else:
+                report.add(
+                    CheckResult(
+                        name="manifest_chunk_size_bytes",
+                        status=CheckStatus.SKIP,
+                        detail="Field absent from manifest (older version)",
+                    )
+                )
         else:
             report.add(
                 CheckResult(

tests/test_util.py

@@ -316,3 +316,27 @@ def test_export_and_load_merkle_proof(tmp_path):
         chunk_data=chunk,
         expected_root=root,
     )
+
+
+def test_extract_text_from_xml_malformed_xml(tmp_path, monkeypatch):
+    import defusedxml.ElementTree as ET
+
+    malformed_xml_content = """<?xml version="1.0"?>
+    <mediawiki>
+      <page>
+        <revision>
+          <text>Hello [[Malformed]]
+        </revision>
+      </page>
+    </mediawiki>
+    """
+
+    input_file = tmp_path / "simplewiki-20260201-pages-malformed.xml"
+
+    with open(input_file, "w", encoding="utf-8") as f:
+        f.write(malformed_xml_content)
+
+    monkeypatch.chdir(tmp_path)
+
+    with pytest.raises(ET.ParseError):
+        utils.extract_text_from_xml(input_file)

tests/test_verify.py

@@ -367,10 +367,11 @@ def setUp(self):
 
     def test_merkle_checks_are_skipped(self):
         r = verify_preprocessing(self.dump, project_root=self.tmp)
-        for name in ("raw_merkle_root", "processed_merkle_root"):
+        for name in ("raw_merkle_root", "processed_merkle_root", "manifest_chunk_size_bytes"):
             c = next((x for x in r.checks if x.name == name), None)
             self.assertIsNotNone(c, f"check '{name}' not found")
             self.assertEqual(c.status, CheckStatus.SKIP)
+            self.assertIn("Field absent from manifest (older version)", c.detail)
 
     def test_other_checks_still_pass(self):
         r = verify_preprocessing(self.dump, project_root=self.tmp)