lp://staging/~rcj/simplestreams/trunk

Branch merges

Propose for merging

No branches dependent on this one.

Merged into lp://staging/simplestreams at revision 456

Scott Moser (community): Approve on 2018-01-16

Server Team CI bot: Approve (continuous-integration) on 2018-01-16

Philip Roche (community): Approve on 2018-01-16

Diff: 10 lines (+1/-0)

1 file modified

tools/make-test-data (+1/-0)

api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

456. By Robert C Jennings on 2018-01-16

Add China's cn-northwest-1 to make-test-data

455. By Scott Moser on 2017-11-02

OpenStack: support uploading squash images for nova-lxd.

Previously, populating a nova-lxd cloud was possible by using
root.tar.gz. A filter like:
  ftype~(root.tar.gz|root.tar.xz)
would cause simplestreams to upload an image with 'disk-format' of
root-tar.

However, Ubuntu 17.04 and newer do not have root.tar.gz or root.tar.xz
images available. Currently here is what is available:
 14.04: root.tar.gz root.tar.xz
 16.04: root.tar.gz root.tar.xz squashfs
 17.10: squashfs

If we simply expected the user to change their filter to include
  root.tar.xz|squashfs
Then they would get two lxd images imported for 16.04 each version.

The change here is to not do anything for an item insert, but instead
insert when the version's insert is called. Then, all the information
about what images there are is available, and it can "pick"
one or the other. Currently preference is given to the .tar.xz format.

The end result is that now users can specify an ftype filter of:
  ftype~(root.tar.gz|root.tar.xz|squashfs)
and the right thing will be done.

Also here is simple knowledge that the squashfs type should be
uploaded to glance with a 'disk_format' of 'squashfs'.

454. By David Ames on 2017-10-31

Support using older swift clients (older than Ocata).

The swift client lagged behind other openstack client libraries in
gaining support for v3 auth. The fix here is to try the v3 auth, and
catch a TypeError and fallback older.

453. By David Ames on 2017-10-31

Fix glance sync always uploading new image with keystone v3 auth.

When using keystone v3 auth, a 'sync' against a populated glance
mirror would provide None as the tenant_id when querying glance for
existing images.

The fix is just to get the tenant_id from the auth client in that case.

452. By Scott Moser on 2017-10-24

tox.ini: Pin versions of style check tools used in flake8 env.

As upstream makes changes, new errors can come up causing breakage
in daily c-i that are unrelated to project errors. The solution here
is to pin the versions of the upstream projects so that style checks
are done on a specific version.

In other projects we have also added a 'flake8-tip' that could be run
but was not run by default. We can do that at a later date if deemed
necessary.

451. By Scott Moser on 2017-10-24

flake8: remove usage of bare except.

Fix an E722 (bare except) issue raised in pycodestyle.

450. By David Ames on 2017-09-14

Keystone v3 Support

Get the correct environment variables for Keystone v3.
Use keystoneauth1 sessions when available.

This is backwards compatible with older versions.
In order to use the v3 support OS_IDENTITY_API_VERSION must be
set in the environment, which is typical of 'novarc' files for
keystone v3 clouds.

449. By Robert C Jennings on 2017-04-25

tools/ubuntu_versions.py: Exclude old versions by version not name

Exclusion from ubuntu versions (and streams) has been done by a
name comparison with 'hardy' will is incorrect for 'artful' and later.
This patch changes to use the version number to provide the same
comparison.

448. By Scott Moser on 2016-11-28

fix flake8

447. By Scott Moser on 2016-11-28

tests: change to having http server select its own port

The big change here is to let the http server in tests/httpserver.py
select its own server by passing port 0 through as the port to HTTPServer.
Then, have it report what port it got.

The tests now run fewer of these servers also, as we run one per class
instead of one per unit test, which makes it significantly faster.
Instead of ~28 python web server processes running during test, it runs
just 5.