fix: avoid needlessly blocking import of session minutes (ietf-tools#3761)

Fixes ietf-tools#3558

Author: jennifer-richards

ietf/meeting/tests_views.py

@@ -6109,17 +6109,47 @@ def test_refuses_identical_import(self):
                          kwargs={'num': self.meeting.number, 'session_id': self.session.pk})
 
         self.client.login(username='secretary', password='secretary+password')
-        r = self.client.post(url, {'markdown_text': 'original markdown text'})  # create a rev
-        self.assertEqual(r.status_code, 302)
         with requests_mock.Mocker() as mock:
             mock.get(f'https://notes.ietf.org/{self.session.notes_id()}/download', text='original markdown text')
             mock.get(f'https://notes.ietf.org/{self.session.notes_id()}/info',
                      text=json.dumps({"title": "title", "updatetime": "2021-12-02T11:22:33z"}))
+            # Create a revision. Run the original text through the preprocessing done when importing
+            # from the notes site.
+            r = self.client.get(url)  # let GET do its preprocessing
+            q = PyQuery(r.content)
+            r = self.client.post(url, {'markdown_text': q('input[name="markdown_text"]').attr['value']})
+            self.assertEqual(r.status_code, 302)
+
             r = self.client.get(url)  # try to import the same text
             self.assertContains(r, "This document is identical", status_code=200)
             q = PyQuery(r.content)
             self.assertEqual(len(q('button:disabled[type="submit"]')), 1)
-            self.assertEqual(len(q('button:not(:disabled)[type="submit"]')), 0)
+            self.assertEqual(len(q('button:enabled[type="submit"]')), 0)
+
+    def test_allows_import_on_existing_bad_unicode(self):
+        """Should not be able to import text identical to the current revision"""
+        url = urlreverse('ietf.meeting.views.import_session_minutes',
+                         kwargs={'num': self.meeting.number, 'session_id': self.session.pk})
+
+        self.client.login(username='secretary', password='secretary+password')
+        r = self.client.post(url, {'markdown_text': 'replaced below'})  # create a rev
+        with open(
+                self.session.sessionpresentation_set.filter(document__type="minutes").first().document.get_file_name(),
+                'wb'
+        ) as f:
+            # Replace existing content with an invalid Unicode byte string. The particular invalid
+            # values here are accented characters in the MacRoman charset (see ticket #3756).
+            f.write(b'invalid \x8e unicode \x99\n')
+        self.assertEqual(r.status_code, 302)
+        with requests_mock.Mocker() as mock:
+            mock.get(f'https://notes.ietf.org/{self.session.notes_id()}/download', text='original markdown text')
+            mock.get(f'https://notes.ietf.org/{self.session.notes_id()}/info',
+                     text=json.dumps({"title": "title", "updatetime": "2021-12-02T11:22:33z"}))
+            r = self.client.get(url)  # try to import the same text
+            self.assertNotContains(r, "This document is identical", status_code=200)
+            q = PyQuery(r.content)
+            self.assertEqual(len(q('button:enabled[type="submit"]')), 1)
+            self.assertEqual(len(q('button:disabled[type="submit"]')), 0)
 
     def test_handles_missing_previous_revision_file(self):
         """Should still allow import if the file for the previous revision is missing"""

ietf/meeting/views.py

@@ -4169,11 +4169,11 @@ def import_session_minutes(request, session_id, num):
     if current_minutes:
         try:
             with open(current_minutes.get_file_name()) as f:
-                if import_contents == Note.preprocess_source(f.read()):
+                if import_contents == f.read():
                     contents_changed = False
                     messages.warning(request, 'This document is identical to the current revision, no need to import.')
-        except FileNotFoundError:
-            pass  # allow import if the file is missing
+        except Exception:
+            pass  # Do not let a failure here prevent minutes from being updated.
 
     return render(
         request,