pythongh-111089: Add PyUnicode_AsUTF8Safe() function

Revert PyUnicode_AsUTF8() change: it no longer rejects embedded null
characters: the PyUnicode_AsUTF8Safe() function should be used
instead.

Doc/c-api/unicode.rst

@@ -979,6 +979,15 @@ These are the UTF-8 codec APIs:
    responsible for deallocating the buffer. The buffer is deallocated and
    pointers to it become invalid when the Unicode object is garbage collected.
 
+   If *size* is NULL and the *unicode* string contains null characters, the
+   UTF-8 encoded string contains embedded null bytes and the caller is not
+   aware since the string size is not stored. C functions processing null
+   terminated ``char*`` truncate the string at the first embedded null byte, and
+   so ignore bytes after the null byte. The :c:func:`PyUnicode_AsUTF8` function
+   can be used to raise an exception rather than truncating the string. Or
+   :c:func:`PyUnicode_AsUTF8(unicode, &size) <PyUnicode_AsUTF8AndSize>` can be
+   used to store the size.
+
    .. versionadded:: 3.3
 
    .. versionchanged:: 3.7
@@ -990,12 +999,13 @@ These are the UTF-8 codec APIs:
 
 .. c:function:: const char* PyUnicode_AsUTF8(PyObject *unicode)
 
-   As :c:func:`PyUnicode_AsUTF8AndSize`, but does not store the size.
+   Similar to :c:func:`PyUnicode_AsUTF8AndSize`, but does not store the size.
 
-   Raise an exception if the *unicode* string contains embedded null
-   characters. To accept embedded null characters and truncate on purpose
-   at the first null byte, ``PyUnicode_AsUTF8AndSize(unicode, NULL)`` can be
-   used instead.
+   If the *unicode* string contains null characters, the UTF-8 encoded string
+   contains embedded null bytes. C functions processing null terminated ``char*``
+   truncate the string at the first embedded null byte, and so ignore bytes
+   after the null byte. The :c:func:`PyUnicode_AsUTF8` function can be used to
+   raise an exception rather than truncating the string.
 
    .. versionadded:: 3.3
 
@@ -1005,6 +1015,20 @@ These are the UTF-8 codec APIs:
    .. versionchanged:: 3.13
       Raise an exception if the string contains embedded null characters.
 
+.. c:function:: const char* PyUnicode_AsUTF8Safe(PyObject *unicode)
+
+   Similar to :c:func:`PyUnicode_AsUTF8`, but raise :exc:`ValueError` if the
+   string contains embedded null characters.
+
+   The Unicode Character Set contains characters which can cause bugs or even
+   security issues depending on how they are proceed. See for example `Unicode
+   Technical Report #36: Unicode Security Considerations
+   <https://unicode.org/reports/tr36/>`_. This function implements a single
+   check: only test if the string contains null characters. Additional checks
+   are needed to prevent further issues cause by Unicode characters.
+
+   .. versionadded:: 3.13
+
 
 UTF-32 Codecs
 """""""""""""

Doc/whatsnew/3.13.rst

@@ -1137,6 +1137,11 @@ New Features
 * Add :c:func:`PyUnicode_AsUTF8` function to the limited C API.
   (Contributed by Victor Stinner in :gh:`111089`.)
 
+* Add :c:func:`PyUnicode_AsUTF8Safe` function: similar to
+  :c:func:`PyUnicode_AsUTF8`, but raise :exc:`ValueError` if the string
+  contains embedded null characters.
+  (Contributed by Victor Stinner in :gh:`111089`.)
+
 
 Porting to Python 3.13
 ----------------------
@@ -1207,12 +1212,6 @@ Porting to Python 3.13
   Note that ``Py_TRASHCAN_BEGIN`` has a second argument which
   should be the deallocation function it is in.
 
-* The :c:func:`PyUnicode_AsUTF8` function now raises an exception if the string
-  contains embedded null characters. To accept embedded null characters and
-  truncate on purpose at the first null byte,
-  ``PyUnicode_AsUTF8AndSize(unicode, NULL)`` can be used instead.
-  (Contributed by Victor Stinner in :gh:`111089`.)
-
 * On Windows, ``Python.h`` no longer includes the ``<stddef.h>`` standard
   header file. If needed, it should now be included explicitly. For example, it
   provides ``offsetof()`` function, and ``size_t`` and ``ptrdiff_t`` types.

Include/cpython/unicodeobject.h

@@ -440,6 +440,10 @@ PyAPI_FUNC(PyObject*) PyUnicode_FromKindAndData(
     const void *buffer,
     Py_ssize_t size);
 
+// Similar to PyUnicode_AsUTF8(), but raise ValueError if the string contains
+// embedded null characters.
+PyAPI_FUNC(const char *) PyUnicode_AsUTF8Safe(PyObject *unicode);
+
 
 /* === Characters Type APIs =============================================== */
 

Lib/test/test_capi/test_unicode.py

@@ -905,25 +905,38 @@ def test_fromordinal(self):
         self.assertRaises(ValueError, fromordinal, 0x110000)
         self.assertRaises(ValueError, fromordinal, -1)
 
-    @support.cpython_only
-    @unittest.skipIf(_testcapi is None, 'need _testcapi module')
-    def test_asutf8(self):
-        """Test PyUnicode_AsUTF8()"""
-        from _testcapi import unicode_asutf8
-
+    def check_asutf8(self, unicode_asutf8):
         self.assertEqual(unicode_asutf8('abc', 4), b'abc\0')
         self.assertEqual(unicode_asutf8('абв', 7), b'\xd0\xb0\xd0\xb1\xd0\xb2\0')
         self.assertEqual(unicode_asutf8('\U0001f600', 5), b'\xf0\x9f\x98\x80\0')
 
-        # disallow embedded null characters
-        self.assertRaises(ValueError, unicode_asutf8, 'abc\0', 0)
-        self.assertRaises(ValueError, unicode_asutf8, 'abc\0def', 0)
-
         self.assertRaises(UnicodeEncodeError, unicode_asutf8, '\ud8ff', 0)
         self.assertRaises(TypeError, unicode_asutf8, b'abc', 0)
         self.assertRaises(TypeError, unicode_asutf8, [], 0)
         # CRASHES unicode_asutf8(NULL, 0)
 
+    @support.cpython_only
+    @unittest.skipIf(_testcapi is None, 'need _testcapi module')
+    def test_asutf8(self):
+        """Test PyUnicode_AsUTF8()"""
+        from _testcapi import unicode_asutf8
+        self.check_asutf8(unicode_asutf8)
+
+        # allow embedded null characters
+        self.assertEqual(unicode_asutf8('abc\0', 5), b'abc\0\0')
+        self.assertEqual(unicode_asutf8('abc\0def', 8), b'abc\0def\0')
+
+    @support.cpython_only
+    @unittest.skipIf(_testcapi is None, 'need _testcapi module')
+    def test_asutf8safe(self):
+        """Test PyUnicode_AsUTF8Safe()"""
+        from _testcapi import unicode_asutf8safe
+        self.check_asutf8(unicode_asutf8safe)
+
+        # disallow embedded null characters
+        self.assertRaises(ValueError, unicode_asutf8safe, 'abc\0', 0)
+        self.assertRaises(ValueError, unicode_asutf8safe, 'abc\0def', 0)
+
     @support.cpython_only
     @unittest.skipIf(_testcapi is None, 'need _testcapi module')
     def test_asutf8andsize(self):

Misc/NEWS.d/next/C API/2023-11-03-12-43-35.gh-issue-111089.LWRfqR.rst

@@ -0,0 +1,3 @@
+Add :c:func:`PyUnicode_AsUTF8Safe` function: similar to
+:c:func:`PyUnicode_AsUTF8`, but raise :exc:`ValueError` if the string
+contains embedded null characters. Patch by Victor Stinner.