Merge branch 'dev' into feature-ai

src/documents/management/commands/document_consumer.py

@@ -82,6 +82,13 @@ def _is_ignored(filepath: Path) -> bool:
 
 
 def _consume(filepath: Path) -> None:
+    # Check permissions early
+    try:
+        filepath.stat()
+    except (PermissionError, OSError):
+        logger.warning(f"Not consuming file {filepath}: Permission denied.")
+        return
+
     if filepath.is_dir() or _is_ignored(filepath):
         return
 
@@ -323,7 +330,12 @@ class Command(BaseCommand):
 
                         # Also make sure the file exists still, some scanners might write a
                         # temporary file first
-                        file_still_exists = filepath.exists() and filepath.is_file()
+                        try:
+                            file_still_exists = filepath.exists() and filepath.is_file()
+                        except (PermissionError, OSError):  # pragma: no cover
+                            # If we can't check, let it fail in the _consume function
+                            file_still_exists = True
+                            continue
 
                         if waited_long_enough and file_still_exists:
                             _consume(filepath)

src/documents/management/commands/document_fuzzy_match.py

@@ -92,6 +92,9 @@ class Command(MultiProcessMixin, ProgressBarMixin, BaseCommand):
                 # doc to doc is obviously not useful
                 if first_doc.pk == second_doc.pk:
                     continue
+                # Skip empty documents (e.g. password-protected)
+                if first_doc.content.strip() == "" or second_doc.content.strip() == "":
+                    continue
                 # Skip matching which have already been matched together
                 # doc 1 to doc 2 is the same as doc 2 to doc 1
                 doc_1_to_doc_2 = (first_doc.pk, second_doc.pk)