fix: return None from CompositeGCS.get_stored_object_md5 for missing objects

The resumable upload_url path calls get_stored_object_md5(filepath) to dedup
against a stored object's md5. On CompositeGCS this delegated to
_get_readable_backend(name), which raises FileNotFoundError when the object is
in no backend -- i.e. every not-yet-uploaded file -- 500ing upload_url for all
new resumable uploads. Catch FileNotFoundError and return None, matching
GoogleCloudStorage.get_stored_object_md5 for a missing blob.

Add regression tests for the found and not-found cases; the not-found test
fails against the pre-fix code (FileNotFoundError).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019ineqU2EGLAcWW2WPWR3mE

Author: rtibbles

contentcuration/contentcuration/tests/test_gcs_storage.py

@@ -8,6 +8,7 @@
 from google.cloud.storage.blob import Blob
 from mixer.main import mixer
 
+from contentcuration.utils.files import hex_to_base64
 from contentcuration.utils.gcs_storage import CompositeGCS
 from contentcuration.utils.gcs_storage import GoogleCloudStorage
 
@@ -232,3 +233,21 @@ def test_get_created_time(self):
             self.storage.get_created_time("blob"),
             self.blob_cls.return_value.time_created,
         )
+
+    def test_get_stored_object_md5(self):
+        mock_blob = self.blob_cls("blob", "blob")
+        mock_blob.md5_hash = hex_to_base64("d41d8cd98f00b204e9800998ecf8427e")
+        self.mock_default_bucket.get_blob.return_value = mock_blob
+        self.assertEqual(
+            self.storage.get_stored_object_md5("blob"),
+            "d41d8cd98f00b204e9800998ecf8427e",
+        )
+
+    def test_get_stored_object_md5__returns_none_if_not_found(self):
+        # Regression: a not-yet-uploaded object is in no backend, so
+        # _get_readable_backend raises FileNotFoundError. get_stored_object_md5
+        # must swallow that and return None (else the resumable upload_url
+        # endpoint 500s on every new file), not propagate the error.
+        self.mock_default_bucket.get_blob.return_value = None
+        self.mock_anon_bucket.get_blob.return_value = None
+        self.assertIsNone(self.storage.get_stored_object_md5("blob"))

contentcuration/contentcuration/utils/gcs_storage.py

@@ -302,4 +302,11 @@ def create_resumable_upload_session(self, name, md5_b64, size):
         )
 
     def get_stored_object_md5(self, name):
-        return self._get_readable_backend(name).get_stored_object_md5(name)
+        # A not-yet-uploaded object exists in no backend; treat that as "no
+        # stored md5" (None) rather than raising, matching
+        # GoogleCloudStorage.get_stored_object_md5 for a missing blob.
+        try:
+            backend = self._get_readable_backend(name)
+        except FileNotFoundError:
+            return None
+        return backend.get_stored_object_md5(name)