gh-142539: Fix traceback caret location calculation for SyntaxErrors with wide chars (#142540)

Author: StanFromIreland

Lib/test/test_traceback.py

@@ -88,6 +88,12 @@ def syntax_error_bad_indentation2(self):
     def tokenizer_error_with_caret_range(self):
         compile("blech  (  ", "?", "exec")
 
+    def syntax_error_with_caret_wide_char(self):
+        compile("女女女=1; 女女女/", "?", "exec")
+
+    def syntax_error_with_caret_wide_char_range(self):
+        compile("f(x, 女女女 for 女女女 in range(30), z)", "?", "exec")
+
     def test_caret(self):
         err = self.get_exception_format(self.syntax_error_with_caret,
                                         SyntaxError)
@@ -125,6 +131,20 @@ def test_caret(self):
         self.assertEqual(err[1].find("("), err[2].find("^"))  # in the right place
         self.assertEqual(err[2].count("^"), 1)
 
+    def test_caret_wide_char(self):
+        err = self.get_exception_format(self.syntax_error_with_caret_wide_char,
+                                        SyntaxError)
+        self.assertIn("^", err[2])
+        # "女女女=1; 女女女/" has display width 17
+        self.assertEqual(err[2].find("^"), 4 + 17)
+
+        err = self.get_exception_format(self.syntax_error_with_caret_wide_char_range,
+                                        SyntaxError)
+        self.assertIn("^", err[2])
+        self.assertEqual(err[2].find("^"), 4 + 5)
+        # "女女女 for 女女女 in range(30)" has display width 30
+        self.assertEqual(err[2].count("^"), 30)
+
     def test_nocaret(self):
         exc = SyntaxError("error", ("x.py", 23, None, "bad syntax"))
         err = traceback.format_exception_only(SyntaxError, exc)

Lib/traceback.py

@@ -1464,10 +1464,11 @@ def _format_syntax_error(self, stype, **kwargs):
                 # Convert 1-based column offset to 0-based index into stripped text
                 colno = offset - 1 - spaces
                 end_colno = end_offset - 1 - spaces
-                caretspace = ' '
                 if colno >= 0:
-                    # non-space whitespace (likes tabs) must be kept for alignment
-                    caretspace = ((c if c.isspace() else ' ') for c in ltext[:colno])
+                    # Calculate display width to account for wide characters
+                    dp_colno = _display_width(ltext, colno)
+                    highlighted = ltext[colno:end_colno]
+                    caret_count = _display_width(highlighted) if highlighted else (end_colno - colno)
                     start_color = end_color = ""
                     if colorize:
                         # colorize from colno to end_colno
@@ -1480,9 +1481,9 @@ def _format_syntax_error(self, stype, **kwargs):
                         end_color = theme.reset
                     yield '    {}\n'.format(ltext)
                     yield '    {}{}{}{}\n'.format(
-                        "".join(caretspace),
+                        ' ' * dp_colno,
                         start_color,
-                        ('^' * (end_colno - colno)),
+                        '^' * caret_count,
                         end_color,
                     )
                 else:

Misc/NEWS.d/next/Library/2025-12-10-21-19-10.gh-issue-142539._8Vzr0.rst

@@ -0,0 +1,2 @@
+:mod:`traceback`: Fix location of carets in :exc:`SyntaxError`\s when the
+source contains wide characters.