* test(yarn): add Alpine Linux support via apk
Install nodejs and yarn via apk on Alpine, sharing the functional
test block with the existing non-Alpine (pre-built binary) path.
Extracts the test block into tests.yml to avoid duplication.
Fixes#4270
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(yarn): skip Node.js runtime warnings in stderr processing
Node.js 24 emits DeprecationWarning lines to stderr (e.g. for url.parse())
that are not JSON, causing _process_yarn_error to fail with "Unexpected
stderr output from Yarn". Skip lines starting with "(node:" before
attempting JSON parsing.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(yarn): add changelog fragment for #11943
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(yarn): only JSON-parse lines starting with '{' in stderr
Node.js 24 emits multi-line DeprecationWarnings to stderr (e.g. the hint
line "(Use `node --trace-deprecation ...`") that are not JSON and were
tripping the "Unexpected stderr output from Yarn" failure. Yarn's
structured output always starts with '{', so skip any line that doesn't.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(yarn): install sqlite on Alpine to fix nodejs 22 symbol error
On Alpine 3.21 nodejs 22 requires SQLite session extension symbols
(sqlite3session_*) that are not present in sqlite-libs; installing
the full sqlite package provides them.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(yarn): refresh apk cache and upgrade sqlite-libs before installing nodejs
The CI Alpine container may have a stale sqlite-libs that lacks the
session extension symbols (sqlite3session_*) required by nodejs 22+.
Force a cache refresh and upgrade sqlite-libs to the latest revision.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(yarn): warn on non-JSON stderr lines instead of silently skipping
Non-JSON lines in stderr (e.g. Node.js runtime DeprecationWarnings) are
surfaced to the user via module.warn() rather than being silently ignored,
since their content and meaning are not known in advance.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* prefix yarn output line
* Update changelogs/fragments/11943-yarn-nodejs-runtime-warnings.yml
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Felix Fontein <felix@fontein.de>
* fix(xml): coerce boolean values to string with a warning
Fixes#7171
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(xml): add integration tests for boolean value handling
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* changelog: add fragment for PR 11959
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* adjustments from review
* test(xml): update boolean-value integration tests to expect failure
Now that xml fails on non-string values, replace the old success-path
tests with failure assertions and add a positive test for quoted strings.
Remove the no-longer-needed result XML fixtures.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* adjustments from review
* fix(xml): correct boolean test assertions to match actual error message format
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* multiple module utils: flatten directories
* adjust pritunl tests
* adjust lxca and keycloak tests
* adjust botmeta
* rename test files correctly
* and an import fix
* rename pritunl api mod utils test
* fix typo in test filename
* rename references to pritunl api test
* rename keycloak mod utils test
* gem - fix --user-install conflict with OS-injected --install-dir
Some distributions (e.g. Fedora) inject --install-dir via operating_system.rb
as a platform default. Combining that with --user-install causes a gem CLI
parser error. Resolve the user install directory at install time and pass
--install-dir instead, which is semantically equivalent and avoids the conflict.
Uninstall is intentionally left unscoped so gem can find gems regardless of
where they were originally installed.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* gem - add changelog fragment for #11873
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* gem - fix user_install handling for install and uninstall
Two issues found in CI:
1. `gem environment user_gemhome` is not supported on older RubyGems (e.g.
Ubuntu 20.04 ships 3.1.2). Simplify get_user_install_dir() to always parse
the full `gem environment` output for "USER INSTALLATION DIRECTORY", which
is stable across all supported versions.
2. On Fedora, `gem uninstall` without flags only searches the system gem path
(set by operating_system.rb), so it cannot find gems installed to the user
dir via --install-dir. Add user_install to the uninstall args_order so that
gem uninstall --user-install is passed when user_install=True. The OS
defaults conflict only applies to gem install, not gem uninstall.
The integration test is updated to be consistent: the user_install:false
install/remove block now also specifies user_install:false on removal.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* gem - use --install-dir for both install and uninstall of user gems
gem uninstall --user-install does not reliably find gems on Fedora/RHEL when
running as root, because those systems may disable user gem home for root and
Gem.user_dir may differ from the path resolved via 'gem environment'.
Use --install-dir <user_dir> for uninstall as well, since that is the exact
path used during install, making the operation consistent across platforms.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* gem - add override_platform_install_dir option and type hints
- add type hints to all functions
- fix misleading comment about --install-dir scoping for uninstall
- add override_platform_install_dir option (default=false) to opt in to
resolving and passing the user gem dir explicitly to both gem install
and gem uninstall, working around OS-injected platform defaults on
distributions such as Fedora
- reclassify changelog fragment as minor_changes (new parameter, not
backport-eligible)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(gem): add integration test for override_platform_install_dir
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(gem): skip default user_install test on RedHat family
OS-injected --install-dir on RHEL/Fedora makes the default user_install: true
case fail. The override_platform_install_dir block already covers the correct
path on those platforms.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
* test(odbc): add Arch Linux support via AUR psqlodbc
Fixes#4267
* test(setup_postgresql_db): guard Arch Linux initdb with creates
* test(odbc): add setup_remote_tmp_dir dependency
* pfexec become plugin: fix broken defaults for illumos/SmartOS
The pfexec become plugin has had incorrect defaults since it was
migrated from Ansible core, making it unusable on illumos without
manual workarounds:
1. become_flags defaulted to '-H -S -n' which are sudo flags.
pfexec does not accept any of these options, causing:
'exec: illegal option -- H'
2. wrap_exe defaulted to false. Unlike sudo, pfexec does not
interpret shell constructs internally. Since Ansible generates
compound commands (echo BECOME-SUCCESS-xxx ; python3), these
must be wrapped in /bin/sh -c for pfexec to execute them.
These issues were originally reported in 2016 (ansible/ansible#15642),
migrated to community.general as #3671, and partially fixed by PR #3889
in 2022 (which corrected quoting but not the defaults). Users have had
to work around this with explicit inventory settings ever since.
Changes:
- become_flags default: '-H -S -n' -> '' (empty)
- wrap_exe default: false -> true
- build_become_command: handle empty flags cleanly
- Updated tests to match corrected defaults
- Added test for custom flags
- Improved wrap_exe description to explain why it should be enabled
* Update changelog fragment with PR number
* Fix ruff formatting in test_pfexec.py
* Address review feedback from russoz
Remove redundant 'should generally be left enabled' description line
and simplify become command return by removing unnecessary flags
conditional.
* Fix unit test regexes for empty default flags
Match double space in test assertions when become_flags defaults to
empty string, consistent with doas, dzdo, and pbrun test patterns.
* pfexec become plugin: deprecate wrap_exe default rather than flipping
Changing the wrap_exe default from false to true is a breaking change
for the narrow case (e.g. ansible.builtin.raw) where the current default
does work, so deprecate instead: remove the default, emit a deprecation
warning when the option is unset, and treat that as false for now.
Build the become command with " ".join() so an empty become_flags no
longer produces a stray double space.
Tests set wrap_exe explicitly so the deprecation warning does not fire
during unit runs.
* pfexec become plugin: target 14.0.0 for wrap_exe deprecation
Per felixfontein's review, switch the deprecation target for the wrap_exe
default from community.general 15.0.0 to 14.0.0, and reword the option
description to mark the current default as deprecated rather than just
"changing in a future release".
* Make all doc fragments private.
* Make all plugin utils private.
* Make all module utils private.
* Reformat.
* Changelog fragment.
* Update configs and ignores.
* Adjust unit test names.
* Bump version to 13.0.0.
* Remove deprecated modules and plugins.
* Remove deprecated module utils.
* Remove leftovers.
* Remove mode=compatibility.
* Change default of is_pre740 from true to false.
* Change default of force_defaults from true to false.
* Remove support for ubuntu_legacy mechanism.
* Remove cpanm compatibility tests.
* udm_user - replace crypt/legacycrypt with passlib
The stdlib crypt module was removed in Python 3.13. Replace the
crypt/legacycrypt import chain with passlib (already used elsewhere
in the collection) and use CryptContext.verify() for password
comparison.
Fixes#4690
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add changelog fragment for PR 11860
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* remove redundant ignore file entries
* udm_user, homectl - replace crypt/legacycrypt with _crypt module utils
Add a new _crypt module_utils that abstracts password hashing and
verification. It uses passlib when available, falling back to the
stdlib crypt or legacycrypt, and raises ImportError if none of them
can be imported. Both udm_user and homectl now use this shared
utility, fixing compatibility with Python 3.13+.
Fixes#4690
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Add BOTMETA entry for _crypt module utils
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* _crypt - fix mypy errors and handle complete unavailability
Replace CryptContext = object fallback (rejected by mypy) with a
proper dummy class definition. Add has_crypt_context flag so modules
can detect when no backend is available. Update both modules to
import and check has_crypt_context instead of testing for None.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* adjsutments from review
* Update plugins/modules/homectl.py
Co-authored-by: Felix Fontein <felix@fontein.de>
* Update plugins/modules/udm_user.py
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Felix Fontein <felix@fontein.de>
* remove pacemaker wait arg and fix race condition
* fix up pacemaker resource and stonith polling
* add changelog for pacemaker timeout bug
* remove env from test case and fix changelog file name
* Update changelogs/fragments/11750-pacemaker-wait-race-condition.yml
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: Felix Fontein <felix@fontein.de>
* parted: add unit_preserve_case option to fix unit case in return value
Adds O(unit_preserve_case) feature flag (bool, default None) to control
the case of the ``unit`` field in the module return value.
Previously the unit was always lowercased (e.g. ``kib``), making it
impossible to feed ``disk.unit`` back as the ``unit`` parameter without
a validation error. With O(unit_preserve_case=true) the unit is returned
in its original mixed case (e.g. ``KiB``), matching the accepted input
values.
The default (None) emits a deprecation notice; the default will become
V(true) in community.general 14.0.0.
Fixes#1860
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* parted: add changelog fragment for PR #11813
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* adjustments from review
* Comment 15.0.0 deprecation in option decription.
* parted: fix unit test calls to parse_partition_info after signature change
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* parted: fix unit_preserve_case - parted outputs lowercase units in machine mode
Parted's machine-parseable output always uses lowercase unit suffixes
(e.g. ``kib``, ``mib``) regardless of what was passed to the ``unit``
parameter. Removing the explicit ``.lower()`` call was therefore not
enough to preserve case.
Add a ``canonical_unit()`` helper that maps a unit string to its canonical
mixed-case form using ``parted_units`` as the reference, and use it
instead of a bare identity when ``unit_preserve_case=true``.
Also fix a yamllint violation in the DOCUMENTATION block (missing space
after ``#`` in inline comments).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Update plugins/modules/parted.py
Co-authored-by: Felix Fontein <felix@fontein.de>
* Update plugins/modules/parted.py
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Felix Fontein <felix@fontein.de>
* fix(logrotate): add missing defaults and parameter validation declarations
- Add default="present" to state parameter
- Add default="/etc/logrotate.d" to config_dir parameter
- Add required_by declarations for shred and compression parameters
* fix(logrotate): fix runtime validation bugs, remove duplicate checks
- Fix shred_cycles TypeError when value is None
- Fix enabled=None handling in get_config_path
- Remove duplicate runtime mutually_exclusive checks
- Add runtime boolean truthiness checks
- Add 'create' parameter format validation
- Remove stale test method
* fix(logrotate): restructure file operations, validate before write
- Write content to tmpdir temp file, validate, then atomic move to destination.
- Wrap all os.remove() calls in try/except with fail_json on error
- Wrap all module.atomic_move() calls in try/except with fail_json on error
- Also add self.mock_module.tmpdir = self.test_dir to test setUp for new code path
* docs(logrotate): update DOCUMENTATION block
- Add 'default: present' to state option
- Add 'default: /etc/logrotate.d' to config_dir option
* feat(logrotate): add optional backup parameter
* chore: add logrotate fixes changelog fragment
* chore(changelog/logrotate): use present tense singular
* fix(logrotate): handle trailing spaces in create param
Co-authored-by: Alexei Znamensky <103110+russoz@users.noreply.github.com>
* refactor(logrotate): remove redundant checks
These are already handled by `required_if` statements in the module spec
* refactor(logrotate): use tempfile to create temporary file
* refactor(logrotate): remove redundant `bool()` casts on `target_enabled`
`target_enabled` is guaranteed to be bool by this point. It's either the module param (typed bool) or falls back to `current_enabled` (also bool). The `bool()` wraps are no-ops.
* refactor(logrotate): remove unused `self.config_file` attribute
* refactor(logrotate): remove dead `any_state` parameter from `read_existing_config`
* fix(logrotate): raise error instead of falling through on enabled-state rename failures
* refactor(logrotate): tighten `get_config_path` sig to bool
`None` callers are removed now so this is safe
* test(logrotate): remove stale open mock assertion after tempfile refactor
* style(logrotate): format file
* chore(logrotate): add missing `version_added` attribute
Co-authored-by: Felix Fontein <felix@fontein.de>
* fix(logrotate): clean up temp file
Co-authored-by: Felix Fontein <felix@fontein.de>
* fix(logrotate): remove redundant temp file cleanup
Co-authored-by: Felix Fontein <felix@fontein.de>
* refactor(logrotate): Use dict subscript to access required backup param
Co-authored-by: Felix Fontein <felix@fontein.de>
* fix(logrotate): fix: only remove old config file when path differs from target
Co-authored-by: Felix Fontein <felix@fontein.de>
* fix(logrotate): update logrotate_bin type hint to str
* feat(logrotate): add backup file handling when removing old config
* style(logrotate): format file
* test(logrotate): add missing backup default to `_setup_module_params`
* test(logrotate): fix incorrect `os.remove` assertion in update test
* refactor(logrotate): remove unnecessary `to_native()` call
Co-authored-by: Felix Fontein <felix@fontein.de>
* refactor(logrotate): replace str quotes with !r
Co-authored-by: Felix Fontein <felix@fontein.de>
* fix(logrotate): change backup default back to true
* fix(logrotate): raise error when `shred_cycle`s is set with `shred=false`
* docs(logrotate): clarify `shred_cycles` behaviour
* fix(logrotate): remove to_native calls for exception messages
* docs(logrotate): improve `config_dir` param description
* refactor(logrotate): simplify backup file assignment logic
* style(logrotate): format file
* docs(logrotate): improve config_map description
---------
Co-authored-by: Alexei Znamensky <103110+russoz@users.noreply.github.com>
Co-authored-by: Felix Fontein <felix@fontein.de>
* iso_extract: strip leading path separator from file entries
Fixes#5283
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* iso_extract: add changelog fragment for issue 5283
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* incus, machinectl, run0 - fix become over pty connections
Four small fixes across three plugins, all discovered while trying to
use community.general.machinectl (and later community.general.run0)
as become methods over the community.general.incus connection.
Core bug: machinectl and run0 both set require_tty = True, but the
incus connection plugin was ignoring that hint and invoking
'incus exec' without -t. Honor require_tty by passing -t, mirroring
what the OpenSSH plugin does with -tt.
Once the pty is in place, both become plugins emit terminal control
sequences (window-title OSC, ANSI reset) around the child command
that land in captured stdout alongside the module JSON and trip the
result parser with "Module invocation had junk after the JSON data".
Suppress that decoration at the source by prefixing the constructed
shell command with SYSTEMD_COLORS=0. TERM=dumb would work too but
has a wider blast radius (it also affects interactive tools inside
the become-user session); SYSTEMD_COLORS is the documented
systemd-scoped knob.
run0 was also missing pipelining = False. When run0 is used over a
connection that honors require_tty, ansible's pipelining sends the
module source on stdin to remote python3, which cannot be forwarded
cleanly through the pty chain and hangs indefinitely. Disable
pipelining the same way community.general.machinectl already does.
Also add tests/unit/plugins/become/test_machinectl.py mirroring the
existing test_run0.py. machinectl had no unit test coverage before,
which is why CI did not catch the SYSTEMD_COLORS=0 prefix change
when the equivalent run0 change broke test_run0_basic/test_run0_flags.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Update changelogs/fragments/11771-incus-machinectl-run0-become-pty.yml
Co-authored-by: Alexei Znamensky <103110+russoz@users.noreply.github.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: Alexei Znamensky <103110+russoz@users.noreply.github.com>
* flatpak: add from_url parameter, deprecate URLs in name
Adds a new `from_url` parameter for installing flatpaks from a
.flatpakref URL, using `flatpak install --from <url>`. The `name`
parameter then carries the reverse DNS application ID, enabling
reliable idempotency checks.
Passing URLs directly in `name` is now deprecated and will be
removed in community.general 14.0.0.
Fixes#4000
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* flatpak: add changelog fragment for PR #11748
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* flatpak: remove deprecation, adjust docs tone
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* flatpak: add integration tests for from_url parameter
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* ensure standard locale in run_command (group5-batch1)
Adds ``LANGUAGE=C`` and ``LC_ALL=C`` to ``run_command()`` calls in modules
that parse command output, to prevent locale-dependent parsing failures on
non-C-locale systems.
Modules updated: apache2_module, composer, facter_facts, known_hosts module
utils, lvg_rename, macports, modprobe, monit, open_iscsi, pacman_key,
rhsm_release, rpm_ostree_pkg, sysupgrade.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* add changelog fragment for group5-batch1
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Remove lvg_rename from locale fix — superseded by PR #11746
PR #11746 (feat: use CmdRunner for LVM commands) takes priority and
will handle lvg_rename.py via CmdRunner refactor. Removing our
run_command_environ_update change to avoid conflict.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* nmcli: start locale fix - normalize run_command environ to LANGUAGE=C, LC_ALL=C
Work in progress - issue #10384 (UTF-8 conn_name support) requires deeper
investigation beyond simple locale variable normalization.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* nmcli: use get_best_parsable_locale() to support UTF-8 connection names
Fixes issue where UTF-8 connection names (e.g. Chinese characters) were
corrupted to '????' when LC_ALL=C forced ASCII encoding, causing
connection_exists() to always return False for non-ASCII names.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* add changelog fragment for PR #11742
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* consul integration tests: re-enable on macOS
- Update consul version to 1.22.6
- Add arm64/aarch64 architecture support
- Fix macOS Gatekeeper quarantine on downloaded binary
- Add wait_for before ACL bootstrap (race condition fix)
- Update HCL config to use tls stanza (required in 1.22)
- Disable gRPC port (conflicts with tls stanza when not configured)
- Remove skip/macos from aliases
Fixes: https://github.com/ansible-collections/community.general/issues/1016
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* changelogs/fragments: add PR number for consul tests fix
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* remove changelog fragment (test-only PR)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(integration): remove CentOS references
* further simplification
* more removals
* rollback systemd_info for now
* ufw: not trivially used with RHEL9 and RHEL10, simplifying tests
* remove tasks for setup_epel where unused
* adjustments from review
* re-enable monit tests in rhel
* enable EPEL for RHEL<11
* rollback EPEL setup, skip only specific versions
* remove skip entirely
* change download URL in setup_epel, adjusted code to use it
* claude tries to install virtualenv, round 1
* claude tries python3 -m venv instead
* remove outdated centos6 file
* add scripts to clean aliases' skips
* remove legacy skips
* code cosmetics
* add license to ALIASES.md
* Fix typos in ALIASES.md documentation
* rolling back freebsd14.2 and 14.3 in iso_extract
* fix versions and re-run
* composer - make create-project idempotent, add force parameter
Adds a check for an existing composer.json in working_dir before running
create-project, so the task is skipped rather than failing on second run.
A new force parameter allows bypassing this check when needed.
Fixes#725.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* changelog fragment: rename to PR number, add PR URL
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* nsupdate: add unit tests
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix var name to regain sanity
* remove unneeded typing from test file
* formatting
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* pacman: add root, cachedir, and config options
Add three dedicated options -- O(root), O(cachedir), and O(config) --
so that all pacman commands get the corresponding global flags
(--root, --cachedir, --config) prepended, enabling use cases such as
installing packages into a chroot or alternative root directory
(similar to pacstrap).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* add changelog frag
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* etcd3: re-enable and fix tests, add unit tests
- Add unit tests for community.general.etcd3 module (12 tests covering
state=present/absent, idempotency, check mode, and error paths)
- Fix integration test setup: update etcd binary to v3.6.9 (from v3.2.14),
download from GitHub releases, add health-check retry loop after start
- Work around etcd3 Python library incompatibility with protobuf >= 4.x
by setting PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
- Update to FQCNs throughout integration tests
- Re-enable both etcd3 and lookup_etcd3 integration targets
Fixes https://github.com/ansible-collections/community.general/issues/322
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* improve use of multiple context managers
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* osx_defaults: add dict support
* add changelog frag
* osx_defaults: fix dict idempotency by using plutil -extract for type-preserving read
The previous approach piped `defaults read` output (old-style plist text)
through `plutil -convert json`. Old-style plist loses boolean type info
(booleans appear as 1/0, indistinguishable from integers), causing the
comparison to fail and reporting changed=True on every run.
Fix by exporting the domain binary plist to a temp file and using
`plutil -extract key json` which correctly preserves all plist types
(booleans stay true/false, integers stay integers, etc.).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* change param from bool to str
* Apply suggestion from review
* Update plugins/modules/osx_defaults.py
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Felix Fontein <felix@fontein.de>
* Fix setting_types() to properly handle routing_rules as a list type
* Add changelog fragment for ipv6.routing-rules bugfix
* Update changelogs/fragments/11630-nmcli-ipv6-routing-rules.yml
Co-authored-by: Alexei Znamensky <103110+russoz@users.noreply.github.com>
* Add PR URL to changelog fragment
---------
Co-authored-by: Alexei Znamensky <103110+russoz@users.noreply.github.com>
