Merge lp:~abentley/juju-release-tools/image-index into lp:juju-release-tools

[Aaron Bentley]

Use index.json, not index2.json for images.

This branch switches image streams to use index.json as their index.

index2.json is a workaround needed for agents because we now use multiple content-ids, and early jujus couldn't cope with that. But it was not a change to the simplestreams format. Juju considers index.json a "legacy" file, but it still uses it, and this is still the only file generated by cloud-images.

Early jujus could break if they encountered multiple image-ids in simplestreams, but they didn't access streams.canonical.com/images/releases by default. Setting image-streams changes both content-id and URL, so there's no reason for image streams to have multiple content-ids. So streams.canonical.com/images/releases is unlikely to ever have multiple content-ids, but even if it did, it would not break old jujus.

Therefore, image streams should follow the format used by cloud-images.

[Curtis Hovey (sinzui)]

Thank you.

review: Approve (code)

=== modified file 'make_aws_image_streams.py'
--- make_aws_image_streams.py	2016-05-10 19:03:36 +0000
+++ make_aws_image_streams.py	2016-05-20 15:27:55 +0000
@@ -11,6 +11,7 @@
 
 from boto import ec2
 from simplestreams.generate_simplestreams import (
+    FileNamer,
     generate_index,
     items2content_trees,
     json_dump,
@@ -18,11 +19,17 @@
 from simplestreams.json2streams import (
     Item,
     JujuFileNamer,
-    write_release_index,
     )
 from simplestreams import util
 
 
+class WindowsFriendlyNamer(JujuFileNamer):
+
+    @classmethod
+    def get_index_path(cls):
+        return FileNamer.get_index_path()
+
+
 def get_parameters(argv=None):
     """Return streams, creds_filename for this invocation.
 
@@ -205,7 +212,7 @@
     # Based on simplestreams.json2streams.write_juju_streams +
     # simplestreams.generate_simplestreams.write_streams,
     # but allows sticky to be specified.
-    namer = JujuFileNamer
+    namer = WindowsFriendlyNamer
     index = generate_index(trees, updated, namer)
     to_write = [(namer.get_index_path(), index,)]
     # Don't let products_condense modify the input
@@ -221,7 +228,6 @@
         util.mkdir_p(os.path.dirname(filef))
         json_dump(data, filef)
         out_filenames.append(filef)
-    out_filenames.append(write_release_index(out_d))
     return out_filenames
 
 

=== modified file 'tests/test_make_aws_image_streams.py'
--- tests/test_make_aws_image_streams.py	2016-05-10 17:20:21 +0000
+++ tests/test_make_aws_image_streams.py	2016-05-20 15:27:55 +0000
@@ -282,13 +282,13 @@
                     with patch('sys.stderr'):
                         write_streams(credentials, china_credentials, now,
                                       streams)
+            self.assertFalse(
+                os.path.exists(os.path.join(streams, 'streams', 'v1',
+                                            'index2.json')))
             index = load_json(streams, 'index.json')
-            index2 = load_json(streams, 'index2.json')
             releases = load_json(streams, 'com.ubuntu.cloud.released-aws.json')
             irc_mock.assert_called_once_with(credentials, china_credentials)
         self.assertEqual(
-            {'format': 'index:1.0', 'updated': 'now', 'index': {}}, index)
-        self.assertEqual(
             {'format': 'index:1.0', 'updated': 'now', 'index': {
                 'com.ubuntu.cloud.released:aws': {
                     'format': 'products:1.0',
@@ -297,7 +297,7 @@
                     'path': 'streams/v1/com.ubuntu.cloud.released-aws.json',
                     'products': ['com.ubuntu.cloud:server:centos7:amd64'],
                     }
-                }}, index2)
+                }}, index)
         expected = {
             'content_id': 'com.ubuntu.cloud.released:aws',
             'format': 'products:1.0',