[3.13] gh-144872: fix heap buffer overflow _PyTokenizer_ensure_utf8 (GH-144807) (#145441)

StanFromIreland · AdamKorcz · web-flow · commit 5577bd1298d5 · 2026-03-02T23:00:59.000Z
(cherry picked from commit 3fc945d) Co-authored-by: AdamKorcz <44787359+AdamKorcz@users.noreply.github.com>
diff --git a/Lib/test/test_source_encoding.py b/Lib/test/test_source_encoding.py
@@ -64,6 +64,23 @@ def test_issue7820(self):
         # two bytes in common with the UTF-8 BOM
         self.assertRaises(SyntaxError, eval, b'\xef\xbb\x20')
 
+    def test_truncated_utf8_at_eof(self):
+        # Regression test for https://issues.oss-fuzz.com/issues/451112368
+        # Truncated multi-byte UTF-8 sequences at end of input caused an
+        # out-of-bounds read in Parser/tokenizer/helpers.c:valid_utf8().
+        truncated = [
+            b'\xc2',              # 2-byte lead, missing 1 continuation
+            b'\xdf',              # 2-byte lead, missing 1 continuation
+            b'\xe0',              # 3-byte lead, missing 2 continuations
+            b'\xe0\xa0',          # 3-byte lead, missing 1 continuation
+            b'\xf0\x90',          # 4-byte lead, missing 2 continuations
+            b'\xf0\x90\x80',      # 4-byte lead, missing 1 continuation
+            b'\xf3',              # 4-byte lead, missing 3 (the oss-fuzz reproducer)
+        ]
+        for seq in truncated:
+            with self.subTest(seq=seq):
+                self.assertRaises(SyntaxError, compile, seq, '<test>', 'exec')
+
     @requires_subprocess()
     def test_20731(self):
         sub = subprocess.Popen([sys.executable,
diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2026-02-16-12-28-43.gh-issue-144872.k9_Q30.rst b/Misc/NEWS.d/next/Core_and_Builtins/2026-02-16-12-28-43.gh-issue-144872.k9_Q30.rst
@@ -0,0 +1 @@
+Fix heap buffer overflow in the parser found by OSS-Fuzz.
diff --git a/Parser/tokenizer/helpers.c b/Parser/tokenizer/helpers.c
@@ -481,9 +481,11 @@ valid_utf8(const unsigned char* s)
         return 0;
     }
     length = expected + 1;
-    for (; expected; expected--)
-        if (s[expected] < 0x80 || s[expected] >= 0xC0)
+    for (int i = 1; i <= expected; i++) {
+        if (s[i] < 0x80 || s[i] >= 0xC0) {
             return 0;
+        }
+    }
     return length;
 }
 

Original file line number	Diff line number	Diff line change
`@@ -0,0 +1 @@`
	`1`	`+Fix heap buffer overflow in the parser found by OSS-Fuzz.`
Original file line number	Diff line number	Diff line change
`@@ -481,9 +481,11 @@ valid_utf8(const unsigned char* s)`
`481`	`481`	`return 0;`
`482`	`482`	`}`
`483`	`483`	`length = expected + 1;`
`484`		`- for (; expected; expected--)`
`485`		`- if (s[expected] < 0x80 \|\| s[expected] >= 0xC0)`
	`484`	`+ for (int i = 1; i <= expected; i++) {`
	`485`	`+ if (s[i] < 0x80 \|\| s[i] >= 0xC0) {`
`486`	`486`	`return 0;`
	`487`	`+ }`
	`488`	`+ }`
`487`	`489`	`return length;`
`488`	`490`	`}`
`489`	`491`