Merge pull request #530 from reef-technologies/fix-incremental-mode

Fix continuation for started large files with no fully finished parts

Author: mlech-reef

b2sdk/_internal/transfer/emerge/executor.py

@@ -393,6 +393,7 @@ def _find_matching_unfinished_file(
                 continue
 
             finished_parts = {}
+            conflict_detected = False
 
             for part in self.services.large_file.list_parts(file_.file_id):
                 emerge_part = emerge_parts_dict.get(part.part_number)
@@ -405,33 +406,33 @@ def _find_matching_unfinished_file(
                         file_.file_id,
                         part.part_number,
                     )
-                    finished_parts = None
+                    conflict_detected = True
                     break
 
                 # Compare part sizes
                 if emerge_part.get_length() != part.content_length:
                     logger.debug(
                         'Rejecting %s: part %s size mismatch', file_.file_id, part.part_number
                     )
-                    continue  # part size doesn't match - so we reupload
+                    conflict_detected = True
+                    break  # part size doesn't match - so we reupload
 
                 # Compare part hashes
                 if emerge_part.is_hashable() and emerge_part.get_sha1() != part.content_sha1:
                     logger.debug(
                         'Rejecting %s: part %s sha1 mismatch', file_.file_id, part.part_number
                     )
-                    continue  # part.sha1 doesn't match - so we reupload
+                    conflict_detected = True
+                    break  # part.sha1 doesn't match - so we reupload
 
                 finished_parts[part.part_number] = part
 
-            if finished_parts is None:
+            if conflict_detected:
                 continue
 
             finished_parts_len = len(finished_parts)
 
-            if finished_parts and (
-                best_match_file is None or finished_parts_len > best_match_parts_len
-            ):
+            if best_match_file is None or finished_parts_len > best_match_parts_len:
                 best_match_file = file_
                 best_match_parts = finished_parts
                 best_match_parts_len = finished_parts_len

changelog.d/+fix_large_file_continuation.fixed.md

@@ -0,0 +1 @@
+Fix continuation for started large files with no fully finished parts.

test/unit/bucket/test_bucket.py

@@ -2006,7 +2006,7 @@ def test_upload_large_resume_no_parts(self):
         large_file_id = self._start_large_file('file1')
         progress_listener = StubProgressListener()
         file_info = self.bucket.upload_bytes(data, 'file1', progress_listener=progress_listener)
-        self.assertNotEqual(large_file_id, file_info.id_)  # it's not a match if there are no parts
+        self.assertEqual(large_file_id, file_info.id_)
         self._check_file_contents('file1', data)
         self.assertTrue(progress_listener.is_valid())
 

test/unit/v0/test_bucket.py

@@ -906,7 +906,7 @@ def test_upload_large_resume_no_parts(self):
         large_file_id = self._start_large_file('file1')
         progress_listener = StubProgressListener()
         file_info = self.bucket.upload_bytes(data, 'file1', progress_listener=progress_listener)
-        self.assertNotEqual(large_file_id, file_info.id_)  # it's not a match if there are no parts
+        self.assertEqual(large_file_id, file_info.id_)
         self._check_file_contents('file1', data)
         self.assertTrue(progress_listener.is_valid())