Consider special files as empty ones when computing content hashes

Closes T255
Ref. D179

swh/model/git.py

-# Copyright (C) 2015  The Software Heritage developers
+# Copyright (C) 2015-2017  The Software Heritage developers
 # See the AUTHORS file at the top-level directory of this distribution
 # License: GNU General Public License version 3, or any later version
 # See top-level LICENSE file for more information
 
 
 import os
+import stat
 
 from enum import Enum
 
@@ -160,17 +161,27 @@ def compute_blob_metadata(filepath):
 
     Args:
         filepath: absolute pathname of the file.
+                  This could be special files (fifo, character or
+                  block device), they will be considered empty files.
 
     Returns:
         Dictionary of values:
             - name: basename of the file
+            - length: data length
             - perms: git permission for file
             - type: git type for file
             - path: absolute filepath on filesystem
 
     """
-    blob_metadata = hashutil.hash_path(filepath)
-    perms = GitPerm.EXEC if os.access(filepath, os.X_OK) else GitPerm.BLOB
+    mode = os.lstat(filepath).st_mode
+    if not stat.S_ISREG(mode):  # special (block or character device, fifo)
+        perms = GitPerm.BLOB
+        blob_metadata = hashutil.hash_data(b'')
+        blob_metadata['length'] = 0
+    else:
+        perms = GitPerm.EXEC if os.access(filepath, os.X_OK) else GitPerm.BLOB
+        blob_metadata = hashutil.hash_path(filepath)
+
     blob_metadata.update({
         'name': os.path.basename(filepath),
         'perms': perms,

swh/model/tests/test_git.py

-# Copyright (C) 2015  The Software Heritage developers
+# Copyright (C) 2015-2017  The Software Heritage developers
 # See the AUTHORS file at the top-level directory of this distribution
 # License: GNU General Public License version 3, or any later version
 # See top-level LICENSE file for more information
@@ -138,6 +138,41 @@ blah
         self.assertEqual(checksum, self.checksums['tag_sha1_git'])
 
 
+@attr('fs')
+class ComputeBlobMetadata(unittest.TestCase):
+    @istest
+    def compute_blob_metadata__special_file_returns_nothing(self):
+        # prepare
+        tmp_root_path = tempfile.mkdtemp().encode('utf-8')
+        name = b'fifo-file'
+        path = os.path.join(tmp_root_path, name)
+
+        # given
+        os.mkfifo(path)
+
+        # when
+        actual_metadata = git.compute_blob_metadata(path)
+
+        # then
+        expected_metadata = {
+            'sha1': b'\xda9\xa3\xee^kK\r2U\xbf\xef\x95`\x18\x90\xaf\xd8\x07\t',
+            'sha1_git': b'\xe6\x9d\xe2\x9b\xb2\xd1\xd6CK\x8b)\xaewZ\xd8\xc2'
+                        b'\xe4\x8cS\x91',
+            'sha256': b"\xe3\xb0\xc4B\x98\xfc\x1c\x14\x9a\xfb\xf4\xc8\x99o"
+                      b"\xb9$'\xaeA\xe4d\x9b\x93L\xa4\x95\x99\x1bxR\xb8U",
+            'perms': git.GitPerm.BLOB,
+            'path': path,
+            'name': name,
+            'type': git.GitType.BLOB,
+            'length': 0
+        }
+
+        self.assertEquals(actual_metadata, expected_metadata)
+
+        # cleanup
+        shutil.rmtree(tmp_root_path)
+
+
 @attr('fs')
 class GitHashWalkArborescenceTree:
     """Root class to ease walk and git hash testing without side-effecty

[Phabricator Migration user]

Some references in the commit message have been migrated:

• D179 is now !279 (closed)