Scott Moser has proposed merging 
~smoser/cloud-init:bug/1751051-subp-encode-with-utf8 into cloud-init:master.

Commit message:
subp: Fix subp usage with non-ascii characters when no system locale.

If python starts up without a locale set, then its default encoding
ends up set as ascii.  That is not easily changed with the likes of
setlocale.  In order to avoid UnicodeDecodeErrors cloud-init will
decode python3 string (python2 basestring) so that the values
passed to Popen are already bytes.

LP: #1751051

Requested reviews:
  cloud-init commiters (cloud-init-dev)
Related bugs:
  Bug #1751051 in cloud-init: "UnicodeEncodeError when creating user with 
non-ascii chars"

For more details, see:

see commit message
Your team cloud-init commiters is requested to review the proposed merge of 
~smoser/cloud-init:bug/1751051-subp-encode-with-utf8 into cloud-init:master.
diff --git a/cloudinit/ b/cloudinit/
index 338fb97..42fea64 100644
--- a/cloudinit/
+++ b/cloudinit/
@@ -1865,6 +1865,8 @@ def subp(args, data=None, rcs=None, env=None, capture=True, shell=False,
         if not isinstance(data, bytes):
             data = data.encode()
+    args = [x if isinstance(x, six.binary_type) else x.encode("utf-8")
+            for x in args]
         sp = subprocess.Popen(args, stdout=stdout,
                               stderr=stderr, stdin=stdin,
diff --git a/tests/unittests/ b/tests/unittests/
index 4a92e74..167008d 100644
--- a/tests/unittests/
+++ b/tests/unittests/
@@ -8,7 +8,9 @@ import shutil
 import stat
 import tempfile
+import json
 import six
+import sys
 import yaml
 from cloudinit import importer, util
@@ -733,6 +735,43 @@ class TestSubp(helpers.CiTestCase):
                          util.target_path("/target/", "///my/path/"))
+    def test_c_lang_can_take_utf8_args(self):
+        """Independent of system LC_CTYPE, args can contain utf-8 strings.
+        When python starts up, its default encoding gets set based on
+        the value of LC_CTYPE.  If no system locale is set, the default
+        encoding for both python2 and python3 in some paths will end up
+        being ascii.
+        Attempts to use setlocale or patching (or changing) os.environ
+        in the current environment seem to not be effective.
+        This test starts up a python with LC_CTYPE set to C so that
+        the default encoding will be set to ascii.  In such an environment
+        supb(['command', 'non-ascii-arg']) will cause a UnicodeDecodeError.
+        """
+        python_prog = '\n'.join([
+            'import json, sys',
+            'from cloudinit.util import subp',
+            'data =',
+            'cmd = json.loads(data)',
+            'sys.stderr.write("cmd=%s\\n" % cmd)',
+            'subp(cmd, capture=False)',
+            ''])
+        cmd = [BASH, '-c', 'echo -n "$@"', '--',
+               self.utf8_valid.decode("utf-8")]
+        python_subp = [sys.executable, '-c', python_prog]
+        out, _err = util.subp(
+            ['cat'], update_env={'LC_CTYPE': 'C'},
+            data=json.dumps(cmd).encode("utf-8"),
+            decode=False)
+        out, _err = util.subp(
+            python_subp, update_env={'LC_CTYPE': 'C'},
+            data=json.dumps(cmd).encode("utf-8"),
+            decode=False)
+        self.assertEqual(self.utf8_valid, out)
 class TestEncode(helpers.TestCase):
     """Test the encoding functions"""
Mailing list:
Post to     :
Unsubscribe :
More help   :

Reply via email to