Fix secondary tag encoding. See #449449_secondary_tag_url_enc

The second exclamation mark was getting url encoded, when it
should be preserved. Added a regression test.

1 files changed, 10 insertions, 3 deletions

diff --git a/scanner.py b/scanner.py
index 61cae63..86f7caf 100644
--- a/scanner.py
+++ b/scanner.py
@@ -1098,6 +1098,13 @@ class Scanner:
         srp = self.reader.peek
         start_mark = self.reader.get_mark()
         ch = srp(1)
+        short_handle = '!'
+        if ch == '!':
+            short_handle = '!!'
+            self.reader.forward()
+            srp = self.reader.peek
+            ch = srp(1)
+
         if ch == '<':
             handle = None
             self.reader.forward(2)
@@ -1112,7 +1119,7 @@ class Scanner:
             self.reader.forward()
         elif ch in _THE_END_SPACE_TAB:
             handle = None
-            suffix = '!'
+            suffix = short_handle
             self.reader.forward()
         else:
             length = 1
@@ -1123,11 +1130,11 @@ class Scanner:
                     break
                 length += 1
                 ch = srp(length)
-            handle = '!'
+            handle = short_handle
             if use_handle:
                 handle = self.scan_tag_handle('tag', start_mark)
             else:
-                handle = '!'
+                handle = short_handle
                 self.reader.forward()
             suffix = self.scan_tag_uri('tag', start_mark)
         ch = srp()