test: add batch boundary test for repair_metadata

Verifies that repair_metadata correctly processes multiple wheel
packages across batch flush boundaries after BULK_SIZE reduction.

JIRA: PULP-1573

Author: decko

pulp_python/tests/functional/api/test_repair.py

@@ -331,3 +331,55 @@ def test_metadata_artifact_repair_endpoint(
     # Check deduplication
     assert len(main_artifact_hrefs) == 4
     assert len(metadata_artifact_hrefs) == 3
+
+
+def test_metadata_repair_batch_boundary(
+    create_content_direct,
+    delete_orphans_pre,
+    download_python_file,
+    monitor_task,
+    move_to_repository,
+    python_bindings,
+    python_repo_factory,
+):
+    """
+    Test that repair_metadata correctly handles packages across batch boundaries.
+
+    Verifies that the batch flush (at BULK_SIZE) does not lose or corrupt metadata
+    for packages processed before and after the flush point.
+    """
+    python_repo = python_repo_factory()
+
+    # Create multiple wheel packages with wrong metadata
+    wheel_files = [
+        ("scipy-1.1.0-cp27-none-win32.whl", "ME"),
+        ("scipy-1.1.0-cp27-none-win_amd64.whl", "ME"),
+        ("scipy-1.1.0-cp27-cp27m-manylinux1_x86_64.whl", "ME"),
+    ]
+    content_hrefs = []
+    for filename, author in wheel_files:
+        url = urljoin(urljoin(PYTHON_FIXTURES_URL, "packages/"), filename)
+        file_path = download_python_file(filename, url)
+        data = {
+            "name": "scipy",
+            "version": "1.1.0",
+            "filename": filename,
+            "author": author,
+            "packagetype": "bdist",
+        }
+        content = create_content_direct(file_path, data)
+        assert content.author == "ME"
+        content_hrefs.append(content.pulp_href)
+
+    move_to_repository(python_repo.pulp_href, content_hrefs)
+
+    # Repair
+    response = python_bindings.RepositoriesPythonApi.repair_metadata(python_repo.pulp_href)
+    monitor_task(response.task)
+
+    # All packages should have repaired metadata (author != "ME")
+    for href in content_hrefs:
+        content = python_bindings.ContentPackagesApi.read(href)
+        assert content.author != "ME", f"Package {content.filename} was not repaired"
+        assert content.packagetype == "bdist_wheel"
+        assert content.metadata_sha256 is not None