moto versions >= 1.0.0 have changed the way the mocked connections through
boto are handled with the @mock_ec2 decorator. They use the boto3 connection
method. However, since we are still using boto in many places, we need to use
the new @mock_ec2_deprecated decorator instead to handle the boto connection
functions for the unit tests.
Versions of moto < 1.0.0 are not Python 3 compatible, so salt-jenkins should
be installing newer versions of moto for those tests. Unfortunately, we cannot
install an older version of moto for Python2 that use the original @mock_ec2 call
and also import the @mock_ec2_deprecated function for newer versions of moto
simultaneously as the @mock_ec2_deprecated function doesn't exist in older
versions of moto.
PyYAML works with and without bytestrings on PY3, and Python 3 can read
all of the test data as regular strings, so bytestrings are unnecessary
here.
This also adds use of textwrap.dedent to make the YAML easier to read,
so it can be indented away from the far left side of the line.
moto versions >= 1.0.0 have changed the way the mocked connections through
boto are handled with the @mock_ec2 decorator. They use the boto3 connection
method. However, since we are still using boto in many places, we need to use
the new @mock_ec2_deprecated decorator instead to handle the boto connection
functions for the unit tests.
Versions of moto < 1.0.0 are not Python 3 compatible, so salt-jenkins should
be installing newer versions of moto for those tests. Unfortunately, we cannot
install an older version of moto for Python2 that use the original @mock_ec2 call
and also import the @mock_ec2_deprecated function for newer versions of moto
simultaneously as the @mock_ec2_deprecated function doesn't exist in older
versions of moto.
Without this, we stacktrace because it does not appear that setting
'gather_facts' to False prevents the library from assuming the presence
of facts. I believe this to be an upstream bug with jnpr.
Because they have listed this as being a deprecated option in the future
this may re-break in the future.
These test the scenarios where another network with a similar name
already exists, verifying that absent doesn't attempt to remove a
network which isn't specified, and present still attempts to create the
specified network despite a similarly named network already being present.
This started failing following commit 515c612, which relied on the
'Name' key being present in the return value of docker.networks -
as the mock didn't have this set the test started failing.
For Python 2 tests, we can use an older version. But when running
these tests of Python 3, we need a newer version of moto that supports
Python 3.
This gates the tests if the expected version of moto is missing.
We can't rely on lists having the same order in Python3 the same
way we rely on them in Python2. If we sort them first, and then
compare them, this test will be more reliable.
017fbdb changed the function calls in such a way that tests in 2017.7
now fail. This updates the mocking to account for new code paths
introduced in this commit.
The mode arg was removed in PR #41966 to get this test to pass on
Windows. However, this caused the same test to fail on Arch Linux
when running Python 3.
Let's keep the test the way it was for non-windows boxes, and only
remove the mode for windows.
These test the scenarios where another network with a similar name
already exists, verifying that network_absent doesn't attempt to remove
a network which isn't specified, and network_present still attempts to
create the specified network despite a similarly named network already
being present.
This started failing following commit 515c612, which relied on the
'Name' key being present in the return value of dockerng.networks -
as the mock didn't have this set the test started failing.
This will keep future IP address changes from impacting this test. It
also tests a host which returns an IPv6 address, and a host which cannot
be resolved.
When there are multiple matches (such as when both an os_family and os
grain both match the minion), the order in which they are processed is
determined by the iteration order of the class' __dict__ attribute.
Since this would be different in different versions of Python, behavior
would be inconsistent depending on which version of Python is being used.
This adds a priority attribute to the Map subclass which, if present,
will re-order the match groups such that the results will be predictable
irrespective of Python version.
