Do not rely on the md5 hash of the file, compare the file contents when searching for duplicate

boards/models/attachment/__init__.py

@@ -1,3 +1,5 b''
+from itertools import zip_longest
+
 import boards
 from boards.models import STATUS_ARCHIVE
 from django.core.files.images import get_image_dimensions
@@ -12,10 +14,8 b' from boards.utils import get_upload_file'
 class AttachmentManager(models.Manager):
     def create_with_hash(self, file):
         file_hash = utils.get_file_hash(file)
-        existing = self.filter(hash=file_hash)
-        if len(existing) > 0:
-            attachment = existing[0]
-        else:
+        attachment = self._get_existing(file_hash, file)
+        if not attachment:
             # FIXME Use full mimetype here, need to modify viewers too
             file_type = get_extension(file.name)
             attachment = self.create(file=file, mimetype=file_type,
@@ -38,6 +38,23 b' class AttachmentManager(models.Manager):'
             images = images.filter(attachment_posts__threads__tags__in=tags)
         return images.order_by('?')[:count]
 
+    def _get_existing(self, file_hash, file):
+        """
+        Gets an attachment with the same file if one exists.
+        """
+        existing = self.filter(hash=file_hash)
+        attachment = None
+        for existing_attachment in existing:
+            equal = True
+            existing_file = existing_attachment.file
+            for chunk, existing_chunk in zip_longest(file.chunks(), existing_file.chunks()):
+                if chunk != existing_chunk:
+                    equal = False
+                    break
+            if equal:
+                attachment = existing[0]
+                return attachment
+
 
 class Attachment(models.Model):
     objects = AttachmentManager()