gh-113626: Add allow_code parameter in marshal functions (GH-113648)

Passing allow_code=False prevents serialization and de-serialization of
code objects which is incompatible between Python versions.

Author: serhiy-storchaka

Doc/library/marshal.rst

@@ -23,7 +23,11 @@ transfer of Python objects through RPC calls, see the modules :mod:`pickle` and
 :mod:`shelve`.  The :mod:`marshal` module exists mainly to support reading and
 writing the "pseudo-compiled" code for Python modules of :file:`.pyc` files.
 Therefore, the Python maintainers reserve the right to modify the marshal format
-in backward incompatible ways should the need arise.  If you're serializing and
+in backward incompatible ways should the need arise.
+The format of code objects is not compatible between Python versions,
+even if the version of the format is the same.
+De-serializing a code object in the incorrect Python version has undefined behavior.
+If you're serializing and
 de-serializing Python objects, use the :mod:`pickle` module instead -- the
 performance is comparable, version independence is guaranteed, and pickle
 supports a substantially wider range of objects than marshal.
@@ -40,7 +44,8 @@ Not all Python object types are supported; in general, only objects whose value
 is independent from a particular invocation of Python can be written and read by
 this module.  The following types are supported: booleans, integers, floating
 point numbers, complex numbers, strings, bytes, bytearrays, tuples, lists, sets,
-frozensets, dictionaries, and code objects, where it should be understood that
+frozensets, dictionaries, and code objects (if *allow_code* is true),
+where it should be understood that
 tuples, lists, sets, frozensets and dictionaries are only supported as long as
 the values contained therein are themselves supported.  The
 singletons :const:`None`, :const:`Ellipsis` and :exc:`StopIteration` can also be
@@ -54,27 +59,32 @@ bytes-like objects.
 The module defines these functions:
 
 
-.. function:: dump(value, file[, version])
+.. function:: dump(value, file, version=version, /, *, allow_code=True)
 
    Write the value on the open file.  The value must be a supported type.  The
    file must be a writeable :term:`binary file`.
 
    If the value has (or contains an object that has) an unsupported type, a
    :exc:`ValueError` exception is raised --- but garbage data will also be written
    to the file.  The object will not be properly read back by :func:`load`.
+   :ref:`Code objects <code-objects>` are only supported if *allow_code* is true.
 
    The *version* argument indicates the data format that ``dump`` should use
    (see below).
 
    .. audit-event:: marshal.dumps value,version marshal.dump
 
+   .. versionchanged:: 3.13
+      Added the *allow_code* parameter.
 
-.. function:: load(file)
+
+.. function:: load(file, /, *, allow_code=True)
 
    Read one value from the open file and return it.  If no valid value is read
    (e.g. because the data has a different Python version's incompatible marshal
-   format), raise :exc:`EOFError`, :exc:`ValueError` or :exc:`TypeError`.  The
-   file must be a readable :term:`binary file`.
+   format), raise :exc:`EOFError`, :exc:`ValueError` or :exc:`TypeError`.
+   :ref:`Code objects <code-objects>` are only supported if *allow_code* is true.
+   The file must be a readable :term:`binary file`.
 
    .. audit-event:: marshal.load "" marshal.load
 
@@ -88,24 +98,32 @@ The module defines these functions:
       This call used to raise a ``code.__new__`` audit event for each code object. Now
       it raises a single ``marshal.load`` event for the entire load operation.
 
+   .. versionchanged:: 3.13
+      Added the *allow_code* parameter.
+
 
-.. function:: dumps(value[, version])
+.. function:: dumps(value, version=version, /, *, allow_code=True)
 
    Return the bytes object that would be written to a file by ``dump(value, file)``.  The
    value must be a supported type.  Raise a :exc:`ValueError` exception if value
    has (or contains an object that has) an unsupported type.
+   :ref:`Code objects <code-objects>` are only supported if *allow_code* is true.
 
    The *version* argument indicates the data format that ``dumps`` should use
    (see below).
 
    .. audit-event:: marshal.dumps value,version marshal.dump
 
+   .. versionchanged:: 3.13
+      Added the *allow_code* parameter.
 
-.. function:: loads(bytes)
+
+.. function:: loads(bytes, /, *, allow_code=True)
 
    Convert the :term:`bytes-like object` to a value.  If no valid value is found, raise
-   :exc:`EOFError`, :exc:`ValueError` or :exc:`TypeError`.  Extra bytes in the
-   input are ignored.
+   :exc:`EOFError`, :exc:`ValueError` or :exc:`TypeError`.
+   :ref:`Code objects <code-objects>` are only supported if *allow_code* is true.
+   Extra bytes in the input are ignored.
 
    .. audit-event:: marshal.loads bytes marshal.load
 
@@ -114,6 +132,9 @@ The module defines these functions:
       This call used to raise a ``code.__new__`` audit event for each code object. Now
       it raises a single ``marshal.loads`` event for the entire load operation.
 
+   .. versionchanged:: 3.13
+      Added the *allow_code* parameter.
+
 
 In addition, the following constants are defined:
 

Doc/whatsnew/3.13.rst

@@ -247,6 +247,14 @@ ipaddress
 * Add the :attr:`ipaddress.IPv4Address.ipv6_mapped` property, which returns the IPv4-mapped IPv6 address.
   (Contributed by Charles Machalow in :gh:`109466`.)
 
+marshal
+-------
+
+* Add the *allow_code* parameter in module functions.
+  Passing ``allow_code=False`` prevents serialization and de-serialization of
+  code objects which are incompatible between Python versions.
+  (Contributed by Serhiy Storchaka in :gh:`113626`.)
+
 mmap
 ----
 

Include/internal/pycore_global_objects_fini_generated.h

@@ -276,6 +276,7 @@ struct _Py_global_strings {
         STRUCT_FOR_ID(after_in_child)
         STRUCT_FOR_ID(after_in_parent)
         STRUCT_FOR_ID(aggregate_class)
+        STRUCT_FOR_ID(allow_code)
         STRUCT_FOR_ID(append)
         STRUCT_FOR_ID(argdefs)
         STRUCT_FOR_ID(arguments)

Include/internal/pycore_global_strings.h

@@ -129,6 +129,32 @@ def test_different_filenames(self):
         self.assertEqual(co1.co_filename, "f1")
         self.assertEqual(co2.co_filename, "f2")
 
+    def test_no_allow_code(self):
+        data = {'a': [({0},)]}
+        dump = marshal.dumps(data, allow_code=False)
+        self.assertEqual(marshal.loads(dump, allow_code=False), data)
+
+        f = io.BytesIO()
+        marshal.dump(data, f, allow_code=False)
+        f.seek(0)
+        self.assertEqual(marshal.load(f, allow_code=False), data)
+
+        co = ExceptionTestCase.test_exceptions.__code__
+        data = {'a': [({co, 0},)]}
+        dump = marshal.dumps(data, allow_code=True)
+        self.assertEqual(marshal.loads(dump, allow_code=True), data)
+        with self.assertRaises(ValueError):
+            marshal.dumps(data, allow_code=False)
+        with self.assertRaises(ValueError):
+            marshal.loads(dump, allow_code=False)
+
+        marshal.dump(data, io.BytesIO(), allow_code=True)
+        self.assertEqual(marshal.load(io.BytesIO(dump), allow_code=True), data)
+        with self.assertRaises(ValueError):
+            marshal.dump(data, io.BytesIO(), allow_code=False)
+        with self.assertRaises(ValueError):
+            marshal.load(io.BytesIO(dump), allow_code=False)
+
     @requires_debug_ranges()
     def test_minimal_linetable_with_no_debug_ranges(self):
         # Make sure when demarshalling objects with `-X no_debug_ranges`

Include/internal/pycore_runtime_init_generated.h

@@ -0,0 +1,3 @@
+Add support for the *allow_code* argument in the :mod:`marshal` module.
+Passing ``allow_code=False`` prevents serialization and de-serialization of
+code objects which is incompatible between Python versions.