Decode archive paths as UTF-8 first #77

src/extractcode/libarchive2.py

@@ -328,6 +328,26 @@ def __iter__(self):
         return self.iter()
 
 
+def as_unicode_path(path):
+    """
+    Return a unicode string decoded from a path bytes `path` of an archive
+    entry.
+
+    Archive entry paths are bytes. Decode them as UTF-8 first: this is the
+    encoding used by most archives and decoding is deterministic. A charset
+    detector can mistake plain ASCII path segments such as ``+dfsg/`` for
+    UTF-7 and corrupt directory names, extracting files in the wrong
+    directory. Fall back to the historical charset detection for paths that
+    are not valid UTF-8.
+    """
+    if not isinstance(path, bytes):
+        return path
+    try:
+        return path.decode("utf-8")
+    except UnicodeDecodeError:
+        return text.as_unicode(path)
+
+
 @attr.attrs
 class Entry(object):
     """
@@ -416,7 +436,7 @@ def get_path(self, func, func_w):
         if not path:
             path = func_w(self.entry_struct)
         if not isinstance(path, str):
-            path = text.as_unicode(path)
+            path = as_unicode_path(path)
 
         return path
 

tests/test_libarchive2.py

@@ -7,7 +7,9 @@
 # See https://aboutcode.org for more information about nexB OSS projects.
 #
 
+import io
 import os
+import tarfile
 
 from commoncode import fileutils
 
@@ -21,6 +23,46 @@
 """
 
 
+def test_as_unicode_path_does_not_decode_plain_ascii_as_utf7():
+    from extractcode.libarchive2 import as_unicode_path
+
+    path = b"qt6-webengine-6.8.2+dfsg/LICENSES/Qt-GPL-exception-1.0.txt"
+    assert as_unicode_path(path) == ("qt6-webengine-6.8.2+dfsg/LICENSES/Qt-GPL-exception-1.0.txt")
+
+
+def test_as_unicode_path_decodes_utf8():
+    from extractcode.libarchive2 import as_unicode_path
+
+    path = "some/path/äöü.txt".encode("utf-8")
+    assert as_unicode_path(path) == "some/path/äöü.txt"
+
+
+def test_as_unicode_path_handles_missing_paths():
+    from extractcode.libarchive2 import as_unicode_path
+
+    assert as_unicode_path(None) is None
+    assert as_unicode_path(b"") == ""
+
+
+def test_extract_tar_keeps_directories_with_a_plus_sign(tmp_path):
+    from extractcode.libarchive2 import extract
+
+    archive = str(tmp_path / "test.tar")
+    name = "qt6-webengine-6.8.2+dfsg/LICENSES/Qt-GPL-exception-1.0.txt"
+    with tarfile.open(archive, "w") as tar:
+        info = tarfile.TarInfo(name)
+        info.size = 5
+        tar.addfile(info, io.BytesIO(b"hello"))
+
+    target_dir = tmp_path / "target"
+    target_dir.mkdir()
+
+    result = list(extract(archive, str(target_dir)))
+
+    assert result == []
+    assert (target_dir / name).exists()
+
+
 class TestExtractorTest(BaseArchiveTestCase):
     def test_libarchive_extract_can_extract_to_relative_paths(self):
         # The setup is a tad complex because we want to have a relative dir