gh-103824: fix use-after-free error in Parser/tokenizer.c (#103993)

Lib/test/test_tokenize.py

@@ -11,7 +11,7 @@
 from test.test_grammar import (VALID_UNDERSCORE_LITERALS,
                                INVALID_UNDERSCORE_LITERALS)
 from test.support import os_helper
-from test.support.script_helper import run_test_script, make_script
+from test.support.script_helper import run_test_script, make_script, run_python_until_end
 import os
 import token
 
@@ -1470,6 +1470,19 @@ def test_comment_at_the_end_of_the_source_without_newline(self):
             self.assertEqual(tok_name[tokens[i + 1].exact_type], tok_name[expected_tokens[i]])
         self.assertEqual(tok_name[tokens[-1].exact_type], tok_name[token.ENDMARKER])
 
+    def test_invalid_character_in_fstring_middle(self):
+        # See gh-103824
+        script = b'''F"""
+        \xe5"""'''
+
+        with os_helper.temp_dir() as temp_dir:
+            filename = os.path.join(temp_dir, "script.py")
+            with open(filename, 'wb') as file:
+                file.write(script)
+            rs, _ = run_python_until_end(filename)
+            self.assertIn(b"SyntaxError", rs.err)
+
+
 class UntokenizeTest(TestCase):
 
     def test_bad_input_order(self):

Parser/tokenizer.c

@@ -2552,6 +2552,10 @@ tok_get_fstring_mode(struct tok_state *tok, tokenizer_mode* current_tok, struct
     while (end_quote_size != current_tok->f_string_quote_size) {
         int c = tok_nextc(tok);
         if (c == EOF || (current_tok->f_string_quote_size == 1 && c == '\n')) {
+            if (tok->decoding_erred) {
+                return MAKE_TOKEN(ERRORTOKEN);
+            }
+
             assert(tok->multi_line_start != NULL);
             // shift the tok_state's location into
             // the start of string, and report the error