In the case of a free style strategy, it is possible to end up with
multiple hosts trying to include from the same role, however the tasks
being included may be different with the use of tasks_from. Previously
if you had two hosts that were included the same role when the
process_include_results function tries to determine if a included needs
to be run on a specific host, it would end up merging two different
tasks into which ever one was processed first.
This change updates the equality check to also check if the task uuid
associated with the IncludedFile is the same. The previous check only
checked if the task's parent uuid was the same. This breaks down when
both includes have the same parent.
- hosts: all
strategy: free
gather_facts: false
tasks:
- include_role:
name: random_sleep
- block:
- name: set a fact (1)
include_role:
name: set_a_fact
tasks_from: fact1.yml
- name: set a fact (2)
include_role:
name: set_a_fact
tasks_from: fact2.yml
- name: include didn't run
fail:
msg: >
set_a_fact didn't run
fact1: {{ fact1 | default('not defined')}}
fact2: {{ fact2 | default('not defined') }}"
when: (fact1 is not defined or fact2 is not defined)
Closes#69521
* do not return the body even if it failed
* add some tests for this and rebase
* import test task
* ignore_errors when fails
Co-authored-by: Jack Zhang <jack.zhang@aspiraconnect.com>
* Enable installing collections from git repositories
* Add tests for installing individual and multiple collections from git repositories
* Test to make sure recursive dependencies with different syntax are deduplicated
* Add documentation
* add a changelog
* Skip Python 2.6
* Only fail if no collections are located in a git repository
Add support for a 'type' key for collections in requirement.yml files.
Update the changelog and document the supported keys and allowed values for the type.
Add a note that the collection(s) in the repo must contain a galaxy.yml
* Add a warning about embedding credentials in SCM URLs
* Update with review suggestions
* suppress sanity compile failure for Python 2.6
Change:
Rather than only using config, have base connection plugins fall back to
play_context.
Test Plan:
- Tested ansible-connection logic against an IOS device
- Tested -k against a VM
- CI
Signed-off-by: Rick Elrod <rick@elrod.me>
* Track collection for version_added.
Validate *all* version numbers in validate-modules.
For tagged version numbers (i.e. version_added), consider source collection to chose validation.
* Make tagging/untagging functions more flexible.
* Tag all versions in doc fragments.
* Tag all deprecation versions issued by code.
* Make Display.deprecated() understand tagged versions.
* Extend validation to enforce tagged version numbers.
* Tag versions in tests.
* Lint and fix test.
* Mention collection name in collection loader's deprecation/removal messages.
* Fix error IDs.
* Handle tagged dates in Display.deprecated().
* Also require that removed_at_date and deprecated_aliases.date are tagged.
* Also automatically tag/untag removed_at_date; fix sanity module removal version check.
* Improve error message when invalid version number is used (like '2.14' in collections).
PR #66461 introduced a regression that resulted in an in correct block in the file
if the block to be inserted did not end with a line separator. Fix this bug and add
tests to cover this scenario.
Fixes#64966
* Fix "TypeError: splitlines() takes no keyword arguments" on Python2.7
* Add changelog fragment
* Don't use `grep -P` for BSD/macOS compatibility
* Fix sanity checks complaining about test fixtures with mixed line endings
* Update changelogs/fragments/66461-blockinfile_preserve_line_endings.yaml
Change:
Allows the user to configure sshpass (1.06+) to look for a different
substring than the default "assword" that it comes with.
Test Plan:
Set a custom ssh password prompt on a VM with PAM and tried connecting to
it. Without `ansible_sshpass_prompt` set in inventory: experienced hang.
With `ansible_sshpass_prompt` in inventory: connected successfully.
Tried setting `ansible_sshpass_prompt` with an older `sshpass` in PATH
and got a loud error, as expected.
Tickets:
Fixes#34722, fixes#54743, refs #11565.
Signed-off-by: Rick Elrod <rick@elrod.me>
* Allow to deprecate options and aliases by date instead of only by version.
* Update display.deprecate().
* Adjust behavior to conform to tested behavior, extend tests, and improve C# style.
* Parse date and fail on invalid date.
This is mainly to make sure that people start using invalid dates, and we eventually have a mess to clean up.
* C# code: improve validation and update/extend tests.
* Make sure that deprecate() is not called with both date and version.
* Forgot to remove no longer necessary formatting.
* Adjust order of warnings in C# code.
* Adjust unrelated test.
* Fix grammar (and make that test pass).
* Don't parse date, and adjust message to be same as in #67684.
* Sanity tests: disable date in past test.
* Validate-modules: validate ISO 8601 date format.
* Validate-modules: switch schema declaration for deprecated_aliases to improve error messages for invalid dates.
* Use DateTime instead of string for date deprecation.
* Validate that date in deprecated_aliases is actually a DateTime.
* Fix tests.
* Fix rebasing error.
* Adjust error codes for pylint, and add removed_at_date and deprecated_aliases.date checks to validate-modules.
* Make deprecation date in the past error codes optional.
* Make sure not both version and date are specified for AnsibleModule.deprecate() calls.
* Stop using Python 3.7+ API.
* Make sure errors are actually reported. Re-add 'ansible-' prefix.
* Avoid crashing when 'name' isn't there.
* Linting.
* Update lib/ansible/module_utils/csharp/Ansible.Basic.cs
Co-authored-by: Jordan Borean <jborean93@gmail.com>
* Adjust test to latest change.
* Prefer date over version if both end up in Display.deprecated().
Co-authored-by: Jordan Borean <jborean93@gmail.com>
* [yum] Make package removal confirmation strict
Change:
After removing packages, the yum module does a final check to ensure the
packages are really installed. The check would include packages that
were RPM `Provides:` values of another package.
This means that, for example, if a third-party kernel RPM spec had
`Provides: kernel` in it, removing the stock kernel would be successful
but the check to see if it was really removed would fail and cause
Ansible to report a failure.
Test Plan:
Tested on local CentOS 7 VM with kernel from elrepo which is known to
`Provides: kernel`.
Tickets:
Fixes#69237
Refs #35672
Refs #40723
Signed-off-by: Rick Elrod <rick@elrod.me>
* `meta/` directory in collections
* runtime metadata for redirection/deprecation/removal of plugin loads
* a compatibility layer to keep existing content working on ansible-base + collections
* a Python import redirection layer to keep collections-hosted (and otherwise moved) content importable by things that don't know better
* supported Ansible version validation on collection loads
* Fix `ansible -K` become_pass regression
Change:
- This fixes a breaking change introduced in
2165f9ac40
Test Plan:
- Local VM for now, with plans to add an integration test for -K going
forward.
Tickets:
Refs #69244
* fix delegated interpeter
* allow returning fact if it is 'the right host'
* added note for future fix/efficiency
as it stands we rerun discovery for the delegated host
unless its saving facts to itself
* fixed test lacking delegate_to mock
When mixed with the free strategy (or any custom strategy that does not behave in
a lock-step manner), the linear methodology of _wait_on_handler_results may cause
race conditions with regular task result processing if the strategy uses
_process_pending_results directly. This patch addresses that by splitting the queues
used for results and adding a flag to _process_pending_results to determine which
queue to check.
Fixes#69457
* Add multipart/form-data functionality
* Fix some linting issues
* Fix error message
* Allow filename to be provided with content
* Add integration test
* Update examples
* General improvements to multipart handling
* Use prepare_multipart for galaxy collection publish
* Properly account for py2 vs py3, ensuring no max header length
* Address test assumptions
* Add unit tests
* Add changelog
* Ensure to use CRLF instead of NL
* Ignore line-endings in fixture
* Consolidate code, add comment
* Bump fallaxy container version
* ci_complete
* Unify ansible-galaxy install -r
* Minor nit fixes for docs
* Re-align warnings
* Fix up integration test
* Fix up test where no roles/collections were in file
* fix delegation vars usage and reporting
- just pass delegated host vars + task vars to plugins
and avoid poluting with original host vars
- updated tests
Since Ansible 2.9.8, if the fileglob plugin is passed a path containing
a subdirectory of a non-existent directory, it will fail. For example:
lookup('fileglob', '/'): ok
lookup('fileglob', '/foo'): (non-existent): ok
lookup('fileglob', '/foo/bar'): (non-existent): FAIL
The exact error depends on Python 2 or 3, but here is the error on
Python 2:
AttributeError: 'NoneType' object has no attribute 'endswith'
And on Python 3:
TypeError: expected str, bytes or os.PathLike object, not NoneType
This change fixes the issue by skipping paths that are falsey before
passing them to os.path.join().
Fixes: #69450
Change:
Adds Fedora 32 to shippable and alters tests slightly for new Fedora.
Test Plan:
CI
Tickets:
Fixes#69230
Co-authored-by: Matt Clay <matt@mystile.com>
* Metadata defaults were not being set if only a few fields were missing.
* ansible-doc with no documentation and no status in metadata should
return empty, just like if there was no documentation and no metadata
at all.
* 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
Change:
We were only testing dnf on RHEL previously.
Test on CentOS 8 as well.
Test Plan:
Ran locally in docker.
Signed-off-by: Rick Elrod <rick@elrod.me>
Change:
Extend the logic for custom error handling in the dnf module, so that on
newer DNF (such as DNF that ships with modern Fedora 31 container
images, and ships with RHEL 8.2) we report errors consistently with
older DNF.
Test Plan:
Ran dnf integration tests against an old Fedora 31 container image and a
brand new Fedora 32 container image; tess passed on both.
Signed-off-by: Rick Elrod <rick@elrod.me>
