diff --git a/openerp/tests/test_mail.py b/openerp/tests/test_mail.py
index 3dbae1473b0..7d177aba937 100755
--- a/openerp/tests/test_mail.py
+++ b/openerp/tests/test_mail.py
@@ -35,7 +35,7 @@ class TestSanitizer(unittest2.TestCase):
     def test_basic_sanitizer(self):
         cases = [
             ("yop", "<p>yop</p>"),  # simple
-            ("lala<p>yop</p>xxx", "<div><p>lala</p><p>yop</p>xxx</div>"),  # trailing text
+            ("lala<p>yop</p>xxx", "<p>lala</p><p>yop</p>xxx"),  # trailing text
             ("Merci à l'intérêt pour notre produit.nous vous contacterons bientôt. Merci",
                 u"<p>Merci à l'intérêt pour notre produit.nous vous contacterons bientôt. Merci</p>"),  # unicode
         ]
diff --git a/openerp/tools/mail.py b/openerp/tools/mail.py
index 3148083908a..4b686f0af34 100644
--- a/openerp/tools/mail.py
+++ b/openerp/tools/mail.py
@@ -106,6 +106,11 @@ def html_sanitize(src, silent=True, strict=False):
             raise
         logger.warning('unknown error obtained when sanitizing %r', src, exc_info=True)
         cleaned = '<p>Unknown error when sanitizing</p>'
+
+    # this is ugly, but lxml/etree tostring want to put everything in a 'div' that breaks the editor -> remove that
+    if cleaned.startswith('<div>') and cleaned.endswith('</div>'):
+        cleaned = cleaned[5:-6]
+
     return cleaned