The jinja2 query() function (or lookup with wantslist=True, which is
the same thing) should always return a list.
However, if you combine a query with errors='ignore' and take the
error path, the current code returns a None value. This is important
in a case such as
- name: Conditional include file
import_tasks: '{{ item }}'
vars:
params:
files:
- path/file1.yaml
- path/file2.yaml
loop: "{{ q('first_found', params, errors='ignore') }}"
If neither file1.yaml or file2.yaml exist, this should do nothing by
returning an empty list to the loop. Currently if you run the above
task you'll get a rather unhelpful:
Invalid data passed to 'loop', it requires a list, got this instead: .
This change ensures that when a query ignores an error, it returns a
empty list. The errors='ignore' case is tested in several variants
with first_found. The extant (but deprecated) "skip: True" for
first_found doesn't seem to be explicitly tested; a test is added here
to avoid regressions before removal in 2.12.
This fixes a regression you'll hit if you follow the suggestion in the
deprecation message included with
e17a2b502d to use errors=ignore over
"skip: True" for first_found. This change adds an example that points
out the query/lookup difference and also fixes the error message to
not mention the now deprecated "skip: True".
Closes#56775
* Rewrite idempotency check
- Check now operates recursively and works on multiple types
- Order of lists matter
* Remove blank line for lint
* Fixed idempotency checks in meraki_ssid
- New sanitize() method for finding keys unique in compared dicts
- Fixed bug in meraki_ssid where SSID specified by number breaks
- This will require a backport
- Converted ignored_keys from tuple to list
* Made changes required for idempotency
* Add changelog fragment
* Add unidirectional option for testing
* Disable option 1 check
* General fixes for is_update_required testing
- Added commented out debug statements in method
- Fixed ignored_keys modifications
* Remove old commented algorithm
Add check for all policy fields (name, apply_to, pattern, tags,
priority) to have correct changed state. Previosly changed state was
based on policy name only.
* Add support for rabbitmq 3.7
* Fix exec args for rabbitmqctl status
* Add changelog and fix description for Ansible 2.9
* Return results even when the cache is disabled
By default the cache is disabled and so the results of the API call
are not placed in there for the return statement to fetch.
* Always update self._cache to return
It is possible for provider.authorize to exist, but set to None. Rather
then loading this value blindly, assume the default is False.
File "/home/zuul/src/github.com/ansible/ansible/lib/ansible/executor/task_executor.py", line 145, in run
res = self._execute()
File "/home/zuul/src/github.com/ansible/ansible/lib/ansible/executor/task_executor.py", line 635, in _execute
result = self._handler.run(task_vars=variables)
File "/home/zuul/src/github.com/ansible/ansible/lib/ansible/plugins/action/ios.py", line 50, in run
provider = load_provider(ios_provider_spec, self._task.args)
File "/home/zuul/src/github.com/ansible/ansible/lib/ansible/module_utils/network/common/utils.py", line 424, in load_provider
provider['authorize'] = boolean(provider['authorize'])
File "/home/zuul/src/github.com/ansible/ansible/lib/ansible/module_utils/parsing/convert_bool.py", line 26, in boolean
raise TypeError("The value '%s' is not a valid boolean. Valid booleans include: %s" % (to_text(value), ', '.join(repr(i) for i in BOOLEANS)))
Signed-off-by: Paul Belanger <pabelanger@redhat.com>
This was required to be installed by a human, out side of our testing.
Add it so ansible-test can now manage.
Signed-off-by: Paul Belanger <pabelanger@redhat.com>
ios_user requires escalated privledges, rather then doing this in the
inventory, we can just add it into the playbook tasks.
Also add missing provider settings for authorize.
Signed-off-by: Paul Belanger <pabelanger@redhat.com>
* Add support to start a vm and put it on hold in one_vm
* Add version_added to vm_start_on_hold option for one_vm
* Add version_added to vm_start_on_hold option for one_vm#2
* Add version_added to vm_start_on_hold option for one_vm#3
* Fix indentation for one_vm
* route53_facts: add check mode support
* route53_facts: add changelog fragment mentioning check mode support
* route53_facts: alter changelog fragment type from `minor_changes` to `bugfixes`
* Update changelogs/fragments/56900-route53-facts-check-mode.yaml
Co-Authored-By: Felix Fontein <felix@fontein.de>
* ios_ntp: Switch interface to GigabitEthernet0/1 for tests
Lets use an interface we know that exists for testing, in some cases
Loopback0 maybe not be configured.
Signed-off-by: Paul Belanger <pabelanger@redhat.com>
* ios_ntp: properly register result variable for testing
This is currently broken, and tests don't work. Fix this by properly
registering the result variable.
Signed-off-by: Paul Belanger <pabelanger@redhat.com>
dict.items() in pytho2 returns a list of tuples which can be iterated
while modifying the dict. In python 3 it returns a view which is tied to
the underlying dict, meaning the modifications to the dict while
iterating are unsafe.
This commit generates new list containing the tuples from the iterator
in python 3 which breaks the link to the dict, allowing the dict to be
modified while iterating the list.
In python 2 it would simply copy the list.
Fixes#56902
