* Fix IndexError in free strategy when hosts become unreachable
Fixes an IndexError crash in the free strategy plugin when hosts
become unreachable during playbook execution.
The bug occurred when:
- last_host index persists across outer loop iterations
- hosts_left is regenerated each iteration via get_hosts_left()
- Some hosts become unreachable between iterations
- hosts_left shrinks but last_host retains its previous value
- Accessing hosts_left[last_host] raises IndexError
The fix adds a bounds check after regenerating hosts_left to reset
last_host to 0 if it's out of bounds. This preserves the round-robin
fairness algorithm while preventing the IndexError.
Assisted-by: Claude Sonnet 4.5 <noreply@anthropic.com>
* integration test
* review comment fixes
* ci_complete
* Replace gpg in rpm_key with librpm
* Manually compute primary key ID and fingerprint for older librpm
* Code cleanup
* Add some v6 tests
* Add rhel 10.1 to CI matrix and changelog
* Remove RHEL 10.0
* consolidate common code and rename unused vars
* remove unnecessary exception raise
* Fix return type
* review comments: change import and use of Optional
* ci_complete ci_coverage
* address review comments
* When checking for existing keys, account for short from key ID and revert test change that hid this
* Support RPM version 6+. Allow 'key' to be fingerprint.
* Replace v6 test key with non-PQC algo version and enable Fedora v6 tests
* modify changelog, mod doc, and del 10.1 test req
* Code refactor and new tests
Refactor code to not need to use librpm API to get list of installed
keys. Also add new tests to verify deleting by fingerprint.
* Refactor drop_key() by rpm version, fully type hint code
* use hexdigest()
The local connection plugin is incorrectly passing a bytearray to methods of become plugins that expect bytes. A recent change to the su plugin exposed this by breaking the become plugin for different locale settings (the bytearray was not properly converted to a str for comparison operations). This changes the local connection plugin to send bytes.
* support separate role argspec file in ansible-doc
* support separate role argspec file in ansible-core
* support both .yml and .yaml extensions on argspec file in ansible-doc
* fix filename building bug and rename some argspec files to test variations
* use yaml extensions from constants
* add superfluous meta/main.yml files to tests
* Update lib/ansible/cli/doc.py
Co-authored-by: Sam Doran <sdoran@redhat.com>
* update docs
* ci_complete
* add changelog and allow for main.yml variations
* add collection role testing
Co-authored-by: Sam Doran <sdoran@redhat.com>
* Initial import of modified version of alikins' code
* Add unit testing for new Role methods
* Fix validate_arg_spec module for sanity test. Add test_include_role_fails.yml integration test from orig PR.
* Add testing of suboptions
* Use new ArgumentSpecValidator class instead of AnsibleModule
* fix for roles with no tasks, use FQ name of new plugin
* Add role dep warning
* Change role argspec file used by ansible-doc
This changes the file used for role argument specs from meta/argument_specs.yml
to meta/main.yml. The argument specs are expected to be in that file under the
top-level entry of `argument_spec`.
* Switch to argument_specs
* Support listing roles in text and JSON
* Change tests for unfrack'd playbook_dir var
These tests were using '/tmp' for testing the setting of the playbook_dir
var. Now that we unfrack that var, MacOS will change this to '/private/tmp'
causing the tests to fail. We can choose a path that does not exist (since
unfrack does not validate existence) so that we can guarantee unfracking
will not change the value.
* check run state of current block only
* Add changelog and test
* Add test for issue 29047
* Fix for both tests
* blerg
* Change test messages
* fix tests
* Add multi-level block in rescue test case
* Add recursive rescue check and multi-level test
* Should probably run the new test
* ci_complete
* Merge new tests
* ci_complete
* Clarify blockinfile docs for insertafter/insertbefore
It's not clear from the docs that these options take effect
only when no marker lines are found in the document.
* Add changelog fragment
When using 'remote_src: yes' and 'mode: preserve', the code handling
the file modes has to be handled on the remote node because it's
the one that has access to the source files. This means that the
copy module itself must handle this, rather than the copy action
plugin (which is where all that logic exists). The copy module
handles this when we copy a single file over. But when it is a
directory as the src parameter value, the mode of the files
beneath it are not considered. Subdirectories are copied with
shutil.copytree() which will preserve permissions automatically.
Individual files are copied with shutil.copyfile() which does NOT
preserve permissions. We need to add some calls to shutil.copymode()
to correct that.
Note: This *always* retains individial file permissions. Specifying
a 'mode' other than 'preserve' when giving a source directory for
the 'src' param does not make sense so will be ignored in that case
only.
Fixes#69783
* Add changelog and test
* Do not pass file mode during recursive copy on symlink files.
The 'file' module cannot deal with mode=preserve. Do not pass that
mode to the module when 'preserve' is used.
* Fix changelog fragment filename
* Make sure collection is a list if a str is given
* Call field validation early on collections
Because we are doing work on modifying the collections value before
it is actually validated, we can validate it ourselves early to make
sure the user supplies either a string or list. Dicts are not valid.
The new validation allows us to simplify the _ensure_default_collection()
function. And since the field is now static, we no longer need to specify
a default for it, which also allows us to simplify the function. Since
the default is now removed, we can also remove the sanity/ignore.txt entry
for collectionsearch.py.
New unit tests are added (and the existing one modified) that allow us to
make sure that we throw a parser error if a user specifies something other
than a string or list for the collections value everywhere it can be specified.
* Revert removing the collection default
The default is actually used, so restore it.
* Fix unit tests in test_helpers.py affected by early collection validation
* Force collections to be static
Templating of collection names does not work at all. Force them to
be static so that a warning is generated for the user.
* Add collectionsearch unit test and fix for reviews
New unit test validates the new _load_collections() code and moves
the new check to the end of the method.
* Change unit test to pytest
* Adjust unit test to use capsys instead of monkeypatch
* Fix pep8 error
* Add changelog fragment
Closes#68704
A value for the project_id parameter to shade's create_network()
call was always being sent, even if no value for 'project' was
supplied. This was breaking folks with older versions of shade
(< 1.6).
Fixes PR https://github.com/ansible/ansible-modules-core/issues/3567
The shade update_router() call will return None if the router is
not actually updated. This will cause the module to fail if we
do not protect against that.
The os_server module could automatically generate a floating IP for
the user with auto_ip=true, but we didn't allow for this FIP to be
automatically deleted when deleting the instance, which is a bug.
Add a new option called delete_fip that enables this.
A change was merged to the main Ansible core code that can cause
a potential hang if any libraries are called that use threading.
This change was:
https://github.com/ansible/ansible/commit/4b0aa1214ca989c71bb87f7a805b1e3673faefa9
This affected the os_object module by causing a hang on the shade
create_object() API call (which in turn calls swiftclient which
uses threading). The fix is to make sure all modules have a main()
that is wrapped with an "if __name__ == '__main__'" check.
We want to NOT consider the async task as failed if the result is
not parsed, which was the intent of:
https://github.com/ansible/ansible/pull/16458
However, the logic doesn't actually do that because we default
the 'parsed' value to True. It should default to False so that
we continue waiting, as intended.
