14539: Escape backslash & space chars on file/dir names, with tests.
authorLucas Di Pentima <ldipentima@veritasgenetics.com>
Thu, 20 Dec 2018 19:32:39 +0000 (16:32 -0300)
committerLucas Di Pentima <ldipentima@veritasgenetics.com>
Fri, 21 Dec 2018 21:59:44 +0000 (18:59 -0300)
Arvados-DCO-1.1-Signed-off-by: Lucas Di Pentima <ldipentima@veritasgenetics.com>

sdk/python/arvados/_normalize_stream.py
sdk/python/arvados/collection.py
sdk/python/tests/test_collections.py

index 47b66c82da000d840bdb7221575019cf6396981e..9caef764edd5af6b6f1c39149ee5f7bf64433fd7 100644 (file)
@@ -5,6 +5,11 @@
 from __future__ import absolute_import
 from . import config
 
+import re
+
+def escape(path):
+    return re.sub('\\\\([0-3][0-7][0-7])', lambda m: '\\134'+m.group(1), path).replace(' ', '\\040')
+
 def normalize_stream(stream_name, stream):
     """Take manifest stream and return a list of tokens in normalized format.
 
@@ -16,7 +21,7 @@ def normalize_stream(stream_name, stream):
 
     """
 
-    stream_name = stream_name.replace(' ', '\\040')
+    stream_name = escape(stream_name)
     stream_tokens = [stream_name]
     sortedfiles = list(stream.keys())
     sortedfiles.sort()
@@ -38,7 +43,7 @@ def normalize_stream(stream_name, stream):
     for streamfile in sortedfiles:
         # Add in file segments
         current_span = None
-        fout = streamfile.replace(' ', '\\040')
+        fout = escape(streamfile)
         for segment in stream[streamfile]:
             # Collapse adjacent segments
             streamoffset = blocks[segment.locator] + segment.segment_offset
index c2517c618923c6f6e2cd36961f2d358a5daf792b..7ad07cc607206fe32f46fe0c94cf9ea34e115224 100644 (file)
@@ -26,7 +26,7 @@ from stat import *
 from .arvfile import split, _FileLikeObjectBase, ArvadosFile, ArvadosFileWriter, ArvadosFileReader, WrappableFile, _BlockManager, synchronized, must_be_writable, NoopLock
 from .keep import KeepLocator, KeepClient
 from .stream import StreamReader
-from ._normalize_stream import normalize_stream
+from ._normalize_stream import normalize_stream, escape
 from ._ranges import Range, LocatorAndRange
 from .safeapi import ThreadSafeApiCache
 import arvados.config as config
@@ -562,6 +562,7 @@ class RichCollectionBase(CollectionBase):
     def stream_name(self):
         raise NotImplementedError()
 
+
     @synchronized
     def has_remote_blocks(self):
         """Recursively check for a +R segment locator signature."""
@@ -600,9 +601,6 @@ class RichCollectionBase(CollectionBase):
 
         pathcomponents = path.split("/", 1)
         if pathcomponents[0]:
-            # Don't allow naming files/dirs \\056
-            if pathcomponents[0] == "\\056":
-                raise IOError(errno.EINVAL, "Invalid name", pathcomponents[0])
             item = self._items.get(pathcomponents[0])
             if len(pathcomponents) == 1:
                 if item is None:
@@ -1842,7 +1840,8 @@ class Subcollection(RichCollectionBase):
     def _get_manifest_text(self, stream_name, strip, normalize, only_committed=False):
         """Encode empty directories by using an \056-named (".") empty file"""
         if len(self._items) == 0:
-            return "%s %s 0:0:\\056\n" % (stream_name, config.EMPTY_BLOCK_LOCATOR)
+            return "%s %s 0:0:\\056\n" % (
+                escape(stream_name), config.EMPTY_BLOCK_LOCATOR)
         return super(Subcollection, self)._get_manifest_text(stream_name,
                                                              strip, normalize,
                                                              only_committed)
index 55dd0838f253f0b9bf97703eee6c472568e73893..3a4dabfeae7877d9b8a4a2def8b713886f1149f6 100644 (file)
@@ -952,23 +952,36 @@ class NewCollectionTestCase(unittest.TestCase, CollectionTestMixin):
         self.assertIs(c.find("./nonexistant.txt"), None)
         self.assertIs(c.find("./nonexistantsubdir/nonexistant.txt"), None)
 
+    def test_escaped_paths_dont_get_unescaped_on_manifest(self):
+        # Dir & file names are literally '\056' (escaped form: \134056)
+        manifest = './\\134056\\040Test d41d8cd98f00b204e9800998ecf8427e+0 0:0:\\134056\n'
+        c = Collection(manifest)
+        self.assertEqual(c.portable_manifest_text(), manifest)
+
+    def test_escaped_paths_do_get_unescaped_on_listing(self):
+        # Dir & file names are literally '\056' (escaped form: \134056)
+        manifest = './\\134056\\040Test d41d8cd98f00b204e9800998ecf8427e+0 0:0:\\134056\n'
+        c = Collection(manifest)
+        self.assertIn('\\056 Test', c.keys())
+        self.assertIn('\\056', c['\\056 Test'].keys())
+
+    def test_make_empty_dir_with_escaped_chars(self):
+        c = Collection()
+        c.mkdirs('./Empty\\056Dir')
+        self.assertEqual(c.portable_manifest_text(),
+                         './Empty\\134056Dir d41d8cd98f00b204e9800998ecf8427e+0 0:0:\\056\n')
+
+    def test_make_empty_dir_with_spaces(self):
+        c = Collection()
+        c.mkdirs('./foo bar/baz waz')
+        self.assertEqual(c.portable_manifest_text(),
+                         './foo\\040bar/baz\\040waz d41d8cd98f00b204e9800998ecf8427e+0 0:0:\\056\n')
+
     def test_remove_in_subdir(self):
         c = Collection('. 781e5e245d69b566979b86e28d23f2c7+10 0:10:count1.txt\n./foo 781e5e245d69b566979b86e28d23f2c7+10 0:10:count2.txt\n')
         c.remove("foo/count2.txt")
         self.assertEqual(". 781e5e245d69b566979b86e28d23f2c7+10 0:10:count1.txt\n./foo d41d8cd98f00b204e9800998ecf8427e+0 0:0:\\056\n", c.portable_manifest_text())
 
-    def test_create_dot_file(self):
-        c = Collection()
-        with self.assertRaises(IOError):
-            with c.open("./dir/\\056", "wb") as f:
-                f.write("Should not be allowed")
-
-    def test_create_file_inside_dot_dir(self):
-        c = Collection()
-        with self.assertRaises(IOError):
-            with c.open("./dir/\\056/foo", "wb") as f:
-                f.write("Should not be allowed")
-
     def test_remove_empty_subdir(self):
         c = Collection('. 781e5e245d69b566979b86e28d23f2c7+10 0:10:count1.txt\n./foo 781e5e245d69b566979b86e28d23f2c7+10 0:10:count2.txt\n')
         c.remove("foo/count2.txt")