gh-149277: Fix error position for invalid numeric literals

Doc/faq/programming.rst

@@ -854,7 +854,7 @@ a :exc:`SyntaxError` because the period is seen as a decimal point::
    >>> 1.__class__
      File "<stdin>", line 1
      1.__class__
-      ^
+       ^
    SyntaxError: invalid decimal literal
 
 The solution is to separate the literal from the period

Lib/test/test_exceptions.py

@@ -296,14 +296,14 @@ def testSyntaxErrorOffset(self):
         check('try:\n  pass\nexcept*:\n  pass\nexcept* ValueError:\n  pass', 3, 8)
 
         # Errors thrown by the tokenizer
-        check('(0x+1)', 1, 3)
-        check('x = 0xI', 1, 6)
+        check('(0x+1)', 1, 4)
+        check('x = 0xI', 1, 7)
         check('0010 + 2', 1, 1)
-        check('x = 32e-+4', 1, 8)
+        check('x = 32e-+4', 1, 9)
         check('x = 0o9', 1, 7)
-        check('\u03b1 = 0xI', 1, 6)
-        check(b'\xce\xb1 = 0xI', 1, 6)
-        check(b'# -*- coding: iso8859-7 -*-\n\xe1 = 0xI', 2, 6,
+        check('\u03b1 = 0xI', 1, 7)
+        check(b'\xce\xb1 = 0xI', 1, 7)
+        check(b'# -*- coding: iso8859-7 -*-\n\xe1 = 0xI', 2, 7,
               encoding='iso8859-7')
         check(b"""if 1:
             def foo():
@@ -322,15 +322,15 @@ def baz():
         check("""f'''
             {
             (123_a)
-            }'''""", 3, 17)
+            }'''""", 3, 18)
         check("""f'''
             {
             f\"\"\"
             {
             (123_a)
             }
             \"\"\"
-            }'''""", 5, 17)
+            }'''""", 5, 18)
         check('''f"""
 
 

Lib/test/test_grammar.py

@@ -21,6 +21,7 @@
     import_helper,
     skip_emscripten_stack_overflow,
     skip_wasi_stack_overflow,
+    subTests,
 )
 from test.support.numbers import (
     VALID_UNDERSCORE_LITERALS,
@@ -180,6 +181,35 @@ def check(test, error=False):
         check("[0x1for x in ()]")
         check("[0xfor x in ()]")
 
+    @subTests('source,offset,msg',
+              [("0xfg", 4, "hexadecimal"),
+               ("0x9g", 4, "hexadecimal"),
+               ("0b1z", 4, "binary"),
+               ("0o7q", 4, "octal"),
+               ("9spam", 2, " decimal"),
+               ("0xfspam", 4, "hexadecimal"),
+               ("1.0x", 4, " decimal"),
+               ("1e3w", 4, " decimal"),
+               ("1jz", 3, "imaginary"),
+               ("0xI", 3, "hexadecimal"),
+               ("0bz", 3, "binary"),
+               ("100_a", 5, " decimal"),
+               ("1.__class__", 3, " decimal"),
+               ("1e+-3", 4, " decimal"),
+               # SyntaxWarning's currently:
+               ("1or 0", 2, " decimal"),
+               ("0o1or 0", 4, "octal"),
+               ])
+    def test_end_of_numerical_literals_offset(self, source, offset, msg):
+        # gh-149277: verify the error caret points at the first invalid
+        # character, not the last valid digit.
+        with warnings.catch_warnings():
+            warnings.simplefilter("error", SyntaxWarning)
+            with self.assertRaises(SyntaxError) as cm:
+                compile(source, "<test>", "eval")
+        self.assertEqual(cm.exception.offset, offset)
+        self.assertIn(msg, cm.exception.msg)
+
     def test_string_literals(self):
         x = ''; y = ""; self.assertTrue(len(x) == 0 and x == y)
         x = '\''; y = "'"; self.assertTrue(len(x) == 1 and x == y and ord(x) == 39)

Lib/test/test_tokenize.py

@@ -2455,7 +2455,7 @@ def test_extra_tokens_relaxes_lexer_errors(self):
         cases = [
             (
                 "2sin(x)",
-                ("invalid decimal literal", (1, 1)),
+                ("invalid decimal literal", (1, 2)),
                 [
                     (token.NUMBER, "2", (1, 0), (1, 1)),
                     (token.NAME, "sin", (1, 1), (1, 4)),

Misc/NEWS.d/next/Core_and_Builtins/2026-08-11-08-16-20.gh-issue-149277.dYRtaF.rst

@@ -0,0 +1,2 @@
+Fix the :exc:`SyntaxError` caret position for invalid numeric literals to
+point at the first invalid character.

Parser/lexer/number.c

@@ -47,11 +47,14 @@ verify_end_of_number(struct tok_state *tok, int c, const char *kind) {
      * other keyword or identifier.
      */
     int r = 0;
+    int in_exp = 0; /* do we handle exponent now? */
+
     if (c == 'a') {
         r = lookahead(tok, "nd");
     }
     else if (c == 'e') {
         r = lookahead(tok, "lse");
+        in_exp = 1;
     }
     else if (c == 'f') {
         r = lookahead(tok, "or");
@@ -69,21 +72,24 @@ verify_end_of_number(struct tok_state *tok, int c, const char *kind) {
     else if (c == 'n') {
         r = lookahead(tok, "ot");
     }
+    if (in_exp) {
+        c = tok_nextc(tok);
+    }
     if (r) {
-        tok_backup(tok, c);
         if (_PyTokenizer_parser_warn(tok, PyExc_SyntaxWarning,
                 "invalid %s literal", kind))
         {
             return 0;
         }
-        tok_nextc(tok);
     }
     else /* In future releases, only error will remain. */
     if (c < 128 && is_potential_identifier_char(c)) {
-        tok_backup(tok, c);
         _PyTokenizer_syntaxerror(tok, "invalid %s literal", kind);
         return 0;
     }
+    if (in_exp) {
+        tok_backup(tok, c);
+    }
     return 1;
 }
 
@@ -101,7 +107,6 @@ tok_decimal_tail(struct tok_state *tok)
         }
         c = tok_nextc(tok);
         if (!Py_ISDIGIT(c)) {
-            tok_backup(tok, c);
             _PyTokenizer_syntaxerror(tok, "invalid decimal literal");
             return 0;
         }
@@ -130,7 +135,6 @@ _PyLexer_scan_number(struct tok_state *tok, struct token *token, int c,
                     c = tok_nextc(tok);
                 }
                 if (!Py_ISXDIGIT(c)) {
-                    tok_backup(tok, c);
                     return MAKE_TOKEN(_PyTokenizer_syntaxerror(tok, "invalid hexadecimal literal"));
                 }
                 do {
@@ -154,7 +158,6 @@ _PyLexer_scan_number(struct tok_state *tok, struct token *token, int c,
                                 "invalid digit '%c' in octal literal", c));
                     }
                     else {
-                        tok_backup(tok, c);
                         return MAKE_TOKEN(_PyTokenizer_syntaxerror(tok, "invalid octal literal"));
                     }
                 }
@@ -182,7 +185,6 @@ _PyLexer_scan_number(struct tok_state *tok, struct token *token, int c,
                         return MAKE_TOKEN(_PyTokenizer_syntaxerror(tok, "invalid digit '%c' in binary literal", c));
                     }
                     else {
-                        tok_backup(tok, c);
                         return MAKE_TOKEN(_PyTokenizer_syntaxerror(tok, "invalid binary literal"));
                     }
                 }
@@ -275,7 +277,6 @@ _PyLexer_scan_number(struct tok_state *tok, struct token *token, int c,
                 if (c == '+' || c == '-') {
                     c = tok_nextc(tok);
                     if (!Py_ISDIGIT(c)) {
-                        tok_backup(tok, c);
                         return MAKE_TOKEN(_PyTokenizer_syntaxerror(tok, "invalid decimal literal"));
                     }
                 } else if (!Py_ISDIGIT(c)) {