Remove unicode before sending it to path parser

Fixes https://sentry.io/coreos/backend-production/issues/175929456/

test/test_endpoints.py

@@ -134,6 +134,9 @@ class WebEndpointTestCase(EndpointTestCase):
   def test_repo_view(self):
     self.getResponse('web.repository', path='devtable/simple')
 
+  def test_unicode_repo_view(self):
+    self.getResponse('web.repository', path='%E2%80%8Bcoreos/hyperkube%E2%80%8B')
+
   def test_org_view(self):
     self.getResponse('web.org_view', path='buynlarge')
 

util/names.py

@@ -1,6 +1,8 @@
 import urllib
 import re
 
+import anunidecode # Don't listen to pylint's lies. This import is required for unidecode below.
+
 from uuid import uuid4
 
 REPOSITORY_NAME_REGEX = re.compile(r'^[\.a-zA-Z0-9_-]+$')
@@ -23,6 +25,8 @@ def escape_tag(tag, default='latest'):
 
 
 def parse_namespace_repository(repository, library_namespace, include_tag=False):
+  repository = repository.encode('unidecode', 'ignore')
+
   parts = repository.rstrip('/').split('/', 1)
   if len(parts) < 2:
     namespace = library_namespace