[3.12] gh-155999: tarfile: handle a member that leaves the destination but comes back (GH-156000) (#156043)

gh-155999: `tarfile`: handle a member that leaves the destination but comes back (GH-156000)
(cherry picked from commit 9768834)

Co-authored-by: Stan Ulbrych <stan@python.org>

Author: miss-islington

Doc/library/tarfile.rst

@@ -1035,6 +1035,10 @@ reused in custom filters:
     paths (in case the name is absolute
     even after stripping slashes, e.g. ``C:/foo`` on Windows).
     This raises :class:`~tarfile.AbsolutePathError`.
+  - Normalize filenames (:attr:`TarInfo.name`) that contain ``..`` components
+    using :func:`os.path.normpath`.
+    Note that this removes internal ``..`` components, which may change the
+    meaning of the name if it traverses symbolic links.
   - :ref:`Refuse <tarfile-extraction-refuse>` to extract files whose absolute
     path (after following symlinks) would end up outside the destination.
     This raises :class:`~tarfile.OutsideDestinationError`.
@@ -1043,6 +1047,10 @@ reused in custom filters:
 
   Return the modified ``TarInfo`` member.
 
+  .. versionchanged:: next
+
+     Filenames containing ``..`` components are now normalized.
+
 .. function:: data_filter(member, path)
 
   Implements the ``'data'`` filter.

Lib/tarfile.py

@@ -779,6 +779,13 @@ def _get_filtered_attrs(member, dest_path, for_data=True):
         # For example, 'C:/foo' on Windows.
         raise AbsolutePathError(member)
     # Ensure we stay in the destination
+    if '..' in name.replace(os.sep, '/').split('/'):
+        # Directories are created from the name as given, so a name that
+        # leaves the destination part-way through would create them
+        # outside it even if the resolved path stays inside.
+        normalized = os.path.normpath(name)
+        if normalized != name:
+            name = new_attrs['name'] = normalized
     target_path = os.path.realpath(os.path.join(dest_path, name),
                                    strict=os.path.ALLOW_MISSING)
     if os.path.commonpath([target_path, dest_path]) != dest_path:

Lib/test/test_tarfile.py

@@ -3716,6 +3716,20 @@ def test_absolute(self):
                         tarfile.AbsolutePathError,
                         """['"].*escaped.evil['"] has an absolute path""")
 
+    def test_parent_dir_out_and_back(self):
+        # Test a member that leaves the destination and comes back.
+        # The containment check looks at the resolved path, which stays
+        # inside, but the intermediate directories are created from the
+        # name as given, which does not.
+        with ArchiveMaker() as arc:
+            arc.add(f'../escaped.evil/../{self.destdir.name}/sub/file',
+                    content='content')
+
+        for filter in 'tar', 'data':
+            with self.subTest(filter):
+                with self.check_context(arc.open(), filter):
+                    self.expect_file('sub/file', content='content')
+
     @symlink_test
     def test_parent_symlink(self):
         # Test interplaying symlinks

Misc/NEWS.d/next/Security/2026-08-13-13-08-11.gh-issue-155999.Xt4rWq.rst

@@ -0,0 +1,5 @@
+Fix the :mod:`tarfile` ``tar`` and ``data`` extraction filters creating
+directories outside the destination for members whose name leaves the
+destination and returns to it, such as ``../evil/../dest/sub/file``. The
+containment check used the resolved path, but intermediate directories were
+created from the name as given.