Avoid encoding or decoding checkout paths.

* index.build_index_from_tree will not decode tree paths, but will write to the
  file system using the bytes paths.
* Repo.stage will accept either a bytes or string paths. Only if the path is a
  string with the path be encoded with sys.getfilesystemencoding(), otherwise
  the bytes will be used directly to write to the git tree.

dulwich/index.py

@@ -482,12 +482,13 @@ def build_index_from_tree(prefix, index_path, object_store, tree_id,
     """
 
     index = Index(index_path)
+    if not isinstance(prefix, bytes):
+        prefix = prefix.encode(sys.getfilesystemencoding())
 
     for entry in object_store.iter_tree_contents(tree_id):
         if not validate_path(entry.path, validate_path_element):
             continue
-        full_path = os.path.join(prefix,
-            entry.path.decode(sys.getfilesystemencoding()))
+        full_path = os.path.join(prefix, entry.path)
 
         if not os.path.exists(os.path.dirname(full_path)):
             os.makedirs(os.path.dirname(full_path))
@@ -510,17 +511,13 @@ def blob_from_path_and_stat(path, st):
     :param st: A stat object
     :return: A `Blob` object
     """
+    assert isinstance(path, bytes)
     blob = Blob()
     if not stat.S_ISLNK(st.st_mode):
         with open(path, 'rb') as f:
             blob.data = f.read()
     else:
-        if platform.python_implementation() == 'PyPy':
-            # os.readlink on pypy seems to require bytes
-            # TODO: GaryvdM: test on other pypy configurations,
-            # e.g. windows, pypy3.
-            path = path.encode(sys.getfilesystemencoding())
-        blob.data = os.readlink(path).encode(sys.getfilesystemencoding())
+        blob.data = os.readlink(path)
     return blob
 
 
@@ -532,8 +529,11 @@ def get_unstaged_changes(index, path):
     :return: iterator over paths with unstaged changes
     """
     # For each entry in the index check the sha1 & ensure not staged
+    if not isinstance(path, bytes):
+        path = path.encode(sys.getfilesystemencoding())
+
     for name, entry in index.iteritems():
-        fp = os.path.join(path, name.decode(sys.getfilesystemencoding()))
+        fp = os.path.join(path, name)
         blob = blob_from_path_and_stat(fp, os.lstat(fp))
         if blob.id != entry.sha:
             yield name

dulwich/repo.py

@@ -730,11 +730,14 @@ class Repo(BaseRepo):
         # missing index file, which is treated as empty.
         return not self.bare
 
-    def stage(self, paths, fsencoding=sys.getfilesystemencoding()):
+    def stage(self, paths):
         """Stage a set of paths.
 
         :param paths: List of paths, relative to the repository path
         """
+
+        root_path_bytes = self.path.encode(sys.getfilesystemencoding())
+
         if not isinstance(paths, list):
             paths = [paths]
         from dulwich.index import (
@@ -743,19 +746,21 @@ class Repo(BaseRepo):
             )
         index = self.open_index()
         for path in paths:
-            full_path = os.path.join(self.path, path)
+            if not isinstance(path, bytes):
+                path = path.encode(sys.getfilesystemencoding())
+            full_path = os.path.join(root_path_bytes, path)
             try:
                 st = os.lstat(full_path)
             except OSError:
                 # File no longer exists
                 try:
-                    del index[path.encode(fsencoding)]
+                    del index[path]
                 except KeyError:
                     pass  # already removed
             else:
                 blob = blob_from_path_and_stat(full_path, st)
                 self.object_store.add_object(blob)
-                index[path.encode(fsencoding)] = index_entry_from_stat(st, blob.id, 0)
+                index[path] = index_entry_from_stat(st, blob.id, 0)
         index.write()
 
     def clone(self, target_path, mkdir=True, bare=False,

dulwich/tests/test_index.py

@@ -1,3 +1,4 @@
+# -*- coding: utf-8 -*-
 # test_index.py -- Tests for the git index
 # encoding: utf-8
 # Copyright (C) 2008-2009 Jelmer Vernooij <jelmer@samba.org>
@@ -53,7 +54,6 @@ from dulwich.objects import (
     )
 from dulwich.repo import Repo
 from dulwich.tests import (
-    expectedFailure,
     TestCase,
     skipIf,
     )
@@ -395,7 +395,6 @@ class BuildIndexTests(TestCase):
                 filee.id)
             self.assertFileContents(epath, 'd', symlink=True)
 
-    @expectedFailure
     def test_no_decode_encode(self):
         repo_dir = tempfile.mkdtemp()
         repo_dir_bytes = repo_dir.encode(sys.getfilesystemencoding())

dulwich/tests/test_repository.py

@@ -40,7 +40,6 @@ from dulwich.repo import (
     MemoryRepo,
     )
 from dulwich.tests import (
-    expectedFailure,
     TestCase,
     skipIf,
     )
@@ -54,7 +53,7 @@ missing_sha = b'b91fa4d900e17e99b433218e988c4eb4a3e9a097'
 
 
 def mkdtemp_unicode():
-    suffix = u'déłwíçh'
+    suffix = u'délwíçh'
     return tempfile.mkdtemp(suffix=suffix)
 
 
@@ -746,7 +745,6 @@ class BuildRepoRootTests(TestCase):
         r.stage(['a'])
         r.stage(['a'])  # double-stage a deleted path
 
-    @expectedFailure
     def test_commit_no_encode_decode(self):
         r = self._repo
         repo_path_bytes = r.path.encode(sys.getfilesystemencoding())