* Pin setuptools to lowest supported @ PEP 517 test
This allows catching the behavior of builds under old setuptools.
* Stop invoking `setup.py install` in tests
This is not the part we care about since it involves dealing with the
external runtime dependencies rather than building our source
distribution.
This patch modifies the in-tree build backend to build sdists that swap
out pointers to it in the `pyproject.toml`'s `[build-system]` section.
The effect of this is that the first build from source (for example,
from a Git checkout) uses our PEP 517 in-tree build backend. But the
produced tarball has `build-backend` set to `setuptools.build_meta`
which is the native build backend of `setuptools`. So any following
builds from that sdist will skip using the in-tree build backend,
calling the setuptools' one.
The good news is that if the first build generated the manpages, they
will be included and won't go anywhere even though, a different build
system is in place.
Combined with #80253, this will make sure not to modify the current
source checkout on that first build.
Co-authored-by: Matt Clay <matt@mystile.com>
* 🧪 Switch macOS 13.2 to 12.0 in CI
The former revealed unexpected flakiness while the latter is the
previous value that was used to be stable. This is a temporary revert.
* Skip lookup_url integration test under macOS 12.0
* ✨ Add macOS 13.2 to `ansible-test`
* 🧪 Replace macOS 12.0 with 13.2 in the CI matrix
* Skip `lookup_url` under macOS 13.2
This is due to https://wefearchange.org/2018/11/forkmacos.rst.html
that manifests itself as follows:
TASK [lookup_url : Test that retrieving a url works] ***************************
objc[15394]: +[__NSCFConstantString initialize] may have been in progress in another thread when fork() was called.
objc[15394]: +[__NSCFConstantString initialize] may have been in progress in another thread when fork() was called. We cannot safely call it or ignore it in t
he fork() child process. Crashing instead. Set a breakpoint on objc_initializeAfterForkError to debug.
ERROR! A worker was found in a dead state
* 📝 Extend ansible-test change note w/ macOS 13.2
* 🐛 Make integration tests compatible w/ modern Git
This patch makes use of the `init.defaultBranch` setting to unify
the test across new and old Git versions since one defaults to
`master` and the other uses `main` for the default branch.
Where possible, it uses the `HEAD` committish to avoid having to
normalize the branch name.
The change fixes the following integration tests:
* `ansible-galaxy`
* `ansible-galaxy-collection-scm` (recursive collection)
* `git`
* 🐛Replace `git-symbolic-ref` with a repo template
This custom Git repository template emulates the `init.defaultBranch` setting
on Git versions below 2.28. Ref: https://superuser.com/a/1559582.
Other workarounds mentioned there, like invoking
`git symbolic-ref HEAD refs/heads/main` after each `git init` turned
out to have mysterious side effects that break the tests in surprising ways.
* 🎨 Make Git integration test non-destructive
This patch makes use of the `$HOME` environment variable to trick Git
into using a user-global config generated in the temporary directory.
* unarchive - properly handle relative path for dest
* Add integration test
* Return output of underlying commands with increased verbosity
* Revert "Return output of underlying commands with increased verbosity"
This reverts commit a2790c8275cdc5697b65670a0beffdc74b741bf6.
* Warn when a relative destination path was provided
* Create a queue per WorkerProcess to receive intra-task updates
* Update `pause` action to use the worker queue
* Deprecate ConnectionBase()._new_stdin
* Add new `Display` convenience method `prompt_until` to manage both controller- and worker-sourced prompting without cross-fork stdin sharing, in-worker mechanism to handle request-response over new worker queue.
This patch creates a thin wrapper around the `setuptools`' PEP 517
build backend in-tree. It features an ability to request generating
the manpage files in the process of building a source distribution.
This toggle is implemented using the `config_settings` mechanism of
PEP 517.
One must explicitly pass it a CLI option to the build front-end to
trigger said behavior. The packagers are expected to use the
following call:
python -m build --config-setting=--build-manpages
This option has no effect on building wheels.
🧪 The change includes integration tests
This test runs building and re-building sdists and wheels with and
without the `--build-manpages` config setting under the
oldest-supported and new `setuptools` pinned.
It is intended to preserve the interoperability of the packaging setup
across Python runtimes.
An extra smoke test also verifies that non PEP 517 interfaces remain functional.
PR #79606
Co-authored-by: Matt Clay <matt@mystile.com>
The previous change to the default mode of operation for the plugin is now limited to collections.
This enables easier testing of the plugin in core, so the previously removed tests have been restored.
* Add condition that causes a when to skip a task
* Fix up tests
* Use false_condition instead of failed_condition
* Remove formatting accidentially added
* Fix sanity
* ansible-inventory, add --limit option
* also graph
* optimize the when not providing limit
* added tests and clog
* avoid empty groups, fix tests
* i swear there as an ignore_errors there already!!?!?!?
* Fix stdout test fallback
Co-authored-by: Sloane Hertel <19572925+s-hertel@users.noreply.github.com>
* simpler approach using subset against inventory directly
* tyupose
* renamed funciton to what it actuall does
got yaml in line with others about removing empty groups
* have graph ignore limts, also note same for --host
* fixed typo long line
* better test
* cause hosvars are not a thing
---------
Co-authored-by: Matt Davis <6775756+nitzmahone@users.noreply.github.com>
Co-authored-by: Sloane Hertel <19572925+s-hertel@users.noreply.github.com>
* Change ansible_job_id format
...to be something that does not resemble float or other type which
value could be changed by literal_eval that jinja2_native uses.
Specifically the format of '%d.%d' is converted from str to float and
then back to float which may result in truncating the number resulting
in the job not being found because the job id does not exist.
* Fix detection of available hashlib algorithms
Detection of hashlib algorithms now works on Python 3.x.
The new implementation works on Python 2.7 and later.
Test coverage is provided by both integration and unit tests.
* Add additional details about hashlib in docs
* quiet default ansible-doc integration test output
* typical non-verbose output was exceeding 27k lines per fun
* disables `set -x` unless `-v` is passed to the script (eg when ansible-test is called with `-v` or `--retry-on-error` adds it on the second try
* added simple progress echoes
* suppress some grep output
* only enable `set -x` for >= `-vvv`
* fix shellcheck default complaint
* Improving the documentation on how we generate the default value of the filename parameter
* fix pep8
* removing unnecessary documentation and improving the module's return
* making the RETURN docs
* pep8
* version_added and changelog
* module._diff
* module._diff fix
* add rudimentary tests for new outputs
---------
Co-authored-by: Matt Davis <mrd@redhat.com>
* Update `collections.abc` imports
- Use `six.moves` for modules and module_utils
- Use `collections.abc` for controller code
This avoids using `ansible.module_utils.common._collections_compat`,
which was added before the vendored `six` was updated to provide these
imports.
* Update _collections_compat to use six.moves
Also update the custom pylint rule to reflect this change.
* add null 'manifest' key to metadata for git repo collections containing MANIFEST.json
changelog
* set to Sentinel instead of None
* Test installing a collection in a git repo that contains a MANIFEST.json
* fix test
* Update changelogs/fragments/ansible-galaxy-install-git-src-manifest.yml
* validate-modules - Remove `__future__` limits
Limits on specific `__future__` imports are handled by other sanity tests.
* Add integration test for module/plugin imports.
* Normalize deprecation records.
* Fix alias deprecations in suboptions.
* Report in which option an alias warning happened for suboptions.
* Add deprecation tests for suboptions.
* Also test deprecation in list of dicts.
* Adjust unit tests for toplevel alias deprecation field name change.
* Add support for importlib.resources
* Remove the importlib.resources imports
* return the correct data
* Some code comments, and re-order for consistency
* Disallow traversing packages below an individual collection
* Add a traversable class for namespaces
* Re-use variable
* Utilize itertools.chain.from_iterable
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* Simplify logic to check for packages from ansible loaders
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* Just a generator expression, instead of a generator
* docstrings
* Add comment about find_spec for our namespaces
* Add some initial unit tests for importlib.resources
* normalize
* Utilize importlib.resources for listing collections
* collections_path is already in config, just use config
* install uses a different default for collections_path
* Remove unused import
* Remove duplicate __truediv__
* Bring back TraversableResources
* Apply some small suggestions from code review
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
Co-authored-by: Matt Davis <6775756+nitzmahone@users.noreply.github.com>
* Remove cross contamination between plugin loader code and CLI code
* Remove unused import
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
Co-authored-by: Matt Davis <6775756+nitzmahone@users.noreply.github.com>
* validate-modules: don't fail on invalid YAML
When validate-modules encounters invalid YAML (e.g. in the EXAMPLES
section), it tries to reformat the exception to include the line number
in the Python file instead of the line number of the embedded YAML
document. However, PyYAML doesn't allow modification of the Mark object
(anymore) which leads to a new exception being raised, instead of
reporting the original exception.
As the original exception is not needed in other places anymore, we
don't have to modify it at all and can just compute the right line
number when reporting the error via ansible-test.
Fixes: #75837
* Add test for invalid module doc YAML syntax.
Co-authored-by: Matt Clay <matt@mystile.com>
Also improve the ansible-test-container integration test:
- Add coverage for the no-probe code path.
- Add work-arounds for centos6 containers (to support backporting).
- Avoid systemd debug when the container doesn't use cgroup.
* background threads writing to stdout/stderr can cause children to deadlock if a thread in the parent holds the internal lock on the BufferedWriter wrapper
* prevent writes to std handles during fork by monkeypatching stdout/stderr during display startup to require a mutex lock with fork(); this ensures no background threads can hold the lock during a fork operation
* add integration test that fails reliably on Linux without this fix
* galaxy: Add license_file to manifest directives
* ag collection build: Test license handling
This adds tests to ensure that
- REUSE licensing files: .reuse/dep5, LICENSES/*, anyfile.license
- galaxy.yml license_file
are always included in the manifest.
This to match the structure with the tests for the user module. When having the tests available as stand-alone tests, it will be easier to add more integration tests in the future.
* Log `runme.sh` execution in integration tests
This patch adds `set -x` where it's missing in the integration tests.
It also enables `pipefail` in `runme.sh` scripts that use pipes.
* Add a change note for PR #79263
* Restrict `wheel` below v0.38.0 under Pythons < 3.7
* Add a change note for PR #79187
* Update changelogs/fragments/79187--wheel-0.38.0.yml
Co-authored-by: Matt Clay <matt@mystile.com>
* Use constraints file when installing wheel.
Co-authored-by: Matt Clay <matt@mystile.com>
Use GalaxyAPI for isinstance check instead of RoleDistributionServer, since the latter is defined in __main__ sometimes (when running integration tests or ansible-galaxy from source) and importing from ansible.cli.galaxy won't reference the same object.
* copy module - fix copying directories containing modified subdirs with remote_src=True. Previously, the first changed subdir would prevent recursively checking for changes for in subdirs at the same level.
* Fix reporting changed for copying empty directories with remote_src=True. If a directory is created on the remote but nothing else, changed is True.
* Custom salt for ansible-vault encrypt
add VAULT_ENCRYPT_SALT config
add salt testing
Co-authored-by: Abhijeet Kasurde <akasurde@redhat.com>
Co-authored-by: Piotr Stawarski <p@stawarski.pl>
* Fix installation from source transforms symlinks of dirs to empty dirs
* Add test to check symlinks to dirs are respected when installing from source
* Add changelog for collection install from source symlink to dirs issue
* Ensure that we do not squash keywords in validate. Fixes#79021
* become_user: nobody should only apply to the test tasks, not the setup_test_user role
* Update how become_user is specified
* Add test to ensure keyword inheritance is working for become
* Add clog frag
* Cache fattributes to prevent re-calculation
* ci_complete
* Remove unnecessary getattr
* Do not crash templating when filter/test name is not a valid Ansible plugin name.
* Store and re-raise KeyError if there was one.
Co-authored-by: s-hertel <19572925+s-hertel@users.noreply.github.com>
* omit keyword should reset to context
ensure we use context/inheritance when calculating value,
using default only when context is unavailable.
fixes#75692
- correct 'vars:' precedence to allow phasing out of include_params
- actually merge vars and always include role_vars
- avoided dupe deps from giving wrong vars
- use 'first' instance of dep as others are from previous instances/invocations
and can have diff values for vars
- ensured deps only provide exportable vars themselves
- added COMMENTS
- added tests
- apply export restrictions setting to defaults
- use 'public' as cutoff
Co-authored-by: Sloane Hertel <19572925+s-hertel@users.noreply.github.com>
ci_complete
* Ensure that meta/runtime.yml redirects are FQCRs.
* Avoid crash when YAML errors without context mark happen, for example if file starts with 'foo---' instead of '---'.
- Allow disabled, unsupported, unstable and destructive integration test targets to be selected using their respective prefixes.
- Allow unstable tests to run when targeted changes are made and the ``--allow-unstable-changed`` option is specified (resolves https://github.com/ansible/ansible/issues/74213).
* Add --offline option to 'ansible-galaxy collection install' to prevent querying distribution servers
This allows installing/upgrading individual tarfiles to have dependency resolution.
Previously needed to be done manually with --no-deps or else all collections and dependencies needed to be included in the requirements.
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* new _fqcn attribute to plugin objects
* unbreak plugins in subdirs
* Fix inadventent changes to _load_name and use existing vars
* add plugin aliases and name property, and replace plugin._load_name where incorrect
* Fix listing plugin names
Fix listing legacy and builtin together
test deprecated plugin documentation
fix doc extensions
remove sometimes inaccurate _load_name handling from plugin.name
* Add tests for REJECT_EXTS and doc extensions
Fix unpredictable collection redirects so non-fqcns in the redirect list are guaranteed to be legacy (instead of determined by the collections keyword)
Move aliases and name properties to _update_object so all plugin types, including doc fragments, can use them
* make legacy plugin names internally consistent
* rename attributes to ansible_name and ansible_aliases
* Recover unix load average into an ansible_loadavg fact
The average load is available out of the box with os.getloadavg().
It adds useful context to the existing ansible_memory_mb and ansible_mounts
facts that provide size and utilization or memory and disk space.
* Add an integration test
* Limit Galaxy API calls during ansible-galaxy collection dependency resolution when possible
Installing a tarfile with a dependency from a Galaxy server (e.g. dependencies: {'ns.coll': '>=1.0.0'}) does not get the available versions of the dependency from the galaxy server if a sufficient version is already installed.
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* refactor and remove redundant code in documentation
allow location and building api to be more accessible
fix issues with displaying ansible.legacy and ansible.builtin
ensure we don't x2 process tokens (some modules reference them also) fixes#77764
move to constants vs hardcoded
more informative errors and comments
now have actual filter/test plugins, which expose the filter/test functions
moved filter/test loading/finding logic into jinja2pluginloader, removed dupe implementations
added tests for case in which we unique by basename when listing
Update lib/ansible/utils/plugin_docs.py
Co-authored-by: Sloane Hertel <19572925+s-hertel@users.noreply.github.com>
* Add a warning for collections that are attempting to be autoloaded to no effect
* Deprecate REQUIRES_WHITELIST and add support for REQUIRES_ENABLED so the docs are accurate
* Fix documentation
* add more vars plugin tests
* Simplify code and add a FIXME for another bug
* fix precedence
* Make setting the class attr at all a warning, even if it's True
* Add fun parsing for _load_name
* include _load_name in messages
This specific integration test gets stuck periodically causing the
Galaxy jobs to be killed on timeout wasting an hour of runtime. The
module that gets stuck waiting on Pulp is an in-test one, called
`setup_collections`. When it works, the task is complete in around 70
seconds but when it doesn't, it just freezes the whole play.
This patch attempts to make it fail faster by putting a reasonable
timeout value of 2 minutes.
On BusyBox systems such as Alpine, chattr on a tmpfs fails with a status of 0 and output only on stderr.
This change updates the test to not assume output on stdout.
* add required file to collection skeleton
This file is required to be able to upload a collection.
It is present in
https://github.com/ansible-collections/collection_template/blob/main/meta/runtime.yml
but that does not get used by default.
Without this, if you use the "ansible-galaxy collection init" command
and you try and publish that collection without adding this file, you
get the error:
"ERROR! Galaxy import process failed: 'requires_ansible' in
meta/runtime.yml is mandatory, but no meta/runtime.yml found (Code:
UNKNOWN)"
Also updates relevant test and adds a changelog fragment
* Relocate venv-pythons.py script.
* Split out unsupported-directory test.
* Split out sanity-import test.
* Split out sanity-validate-modules test.
* Split out units test.
* Split out integration test.
* Split out units-constraints test.
* Split out integration-constraints test.
* Split out coverage test.
* Split out sanity test.
* Split out git test.
* Update test groups.
* Don't actually update the cache in check mode
Add tests for updating the cache in check mode
* Don't mark packages as manually installed in check mode and add a test for it
* changelog
* Update test/integration/targets/apt/tasks/apt.yml
* fix test
* update attributes of files that are symlinks
* update attributes of files that are hard links
* fix default state in documentation
* remove unnecessary suppression
* add to changelog
* `context/target` tests must be in groups 1 - 2.
* `context/controller` tests must be in groups 3 - 5.
This makes it easier to efficiently organize groups and balance test runtimes.
* Rethread pr/70185 through the dependency resolver
Hang optional metadata toggle on the ConcreteArtifactsManager instead of threading it through whole list codepath
Don't error while listing collections if a collection's metadata is missing keys required for building a collection.
Give an informative warning if metadata has been badly formatted.
Co-authored-by: Sam Doran <sdoran@redhat.com>
The current language pack fails to install on Ubuntu 22.04.
However, since the langauge only needs to be non-English, changing it preserves the test functionality.
* wait_for - Read file and perform comparisons using bytes to avoid decode errors. Fixes#78214
* Write non-ascii via script instead of static file
* Use contexlib.closing to support py27 context manager
* Use executable from task, instead of shebang
* Update encoded bytes to utf16
* Don't set become_flags in become integration test.
Setting the flags in the test overrides the default value in the become plugins, which was not intentional.
* Remove hack for broken implicit tilde support.
It was only broken because the tests were overriding the default become flags for sudo.
* Simplify assertions.
`group` and `setup_cron` integration tests currently install `libuser` and `faketime`
packages respectively, from the non-default edge/testing repositories. It is no
longer necessary to do so for the former so this patch drops the repo option. As
for the latter, this patch includes a note explaining when the same could be done
for it too.
This patch also adds a note on installing `faketime` on Alpine in CI.
Co-Authored-By: Matt Clay <matt@mystile.com>
* Add setup_test_user integration target.
This integration target creates a temporary test user and removes
it when the current integration target finishes. This user is
configured with a password so it can be used with become, even
when the current user is unprivileged.
* Fix slurp integration test.
* Fix unarchive integration test.
* Fix module_utils integration test.
* Move undefined check from concat to finalize
In the classic Jinja2's Environment str() is called on the return value of the
finalize method to potentially trigger the undefined error. That is not
the case in NativeEnvironment where string conversion of the return value is
not desired. We workaround that by checking for Undefined in all of our concat
functions. It seems simpler to do it earlier in the finalize method(s) instead.
As a side-effect it fixes an undefined variable detection in imported templates.
Fixes#78156
ci_complete
* Fix sanity
* ...
* sigh
* [dnf] Fix skip_broken, add test coverage
Change:
- skip_broken was set in config but not actually used in calls to
base.install()
- added a lot of test cases with specialized repo
- got rid of external (docker repo) nobest test cases since the
specialized repo works well for those too
- Slight cleanup and adding comments in dnf module
Test Plan:
- ci_complete
Tickets:
- Fixes#73072
Original-author: Rick Elrod <rick@elrod.me>
* Use a better test for checking results list
ci_complete
Signed-off-by: Rick Elrod <rick@elrod.me>
Co-authored-by: Rick Elrod <rick@elrod.me>
* Relocate update-ignore.py for easier re-use.
* Add script to ease collection testing.
* Skip ignore rewrite if file does not exist.
* Add integration test for the shebang sanity test.
* Fix ansible-test-no-tty integration test.
Previously the test only verified a TTY was not used if a TTY already existed.
This prevented the test from verifying behavior when run in CI.
Now the test creates a PTY before invoking ansible-test.
* Clean up ansible-test-docker integration test.
Add a new handler class ZipZArchive to use unzip -Z as an alternative to zipinfo
Run 'unzip -Z' in can_handle_archive so we fall back to the next handler if it's not available (failing in is_unarchived is too late)
* Add a test for unzip -Z when zipinfo is not available
* Update test for missing binary altogether by removing /usr/bin from the PATH
* ansible-galaxy - support resolvelib versions >= 0.5.3, <= 0.8.1
Test incompatibilities are removed for resolvelib >= 0.6.0
Test against the latest 0.8.x version and fix requirements
* Fix tests - use a venv for testing the range of resolvelib versions
* Update temporary hardcoded fallback for ansible-test
* Update hardcoded upperbound for sanity tests
* Make error check more flexible
checksum can also accept a checksum only file (no filename beside the checksum).
fixes#54390
Co-authored-by: Baptiste Mille-Mathias <baptiste.millemathias@gmail.com>
* Add RHEL 9.0 to CI.
* Restrict network manager inspection to RHEL8
* Skip module tests when astream_name is undefined, undefine it for RHEL9 until 9.1
* Remove redundant test.
Co-authored-by: Matt Martz <matt@sivel.net>
* Add find test to assert patterns with regex
* Add test to assert that nested includes with relative path work
* Add test to assert symmetric_difference supports unhashable types
* ci_complete ci_coverage
* Add test to assert JSON values are converted to Python
* Add test to assert how unexpected module failures behave
* Add test to verify remote_is_local
* ci_complete ci_coverage
* Get real path of local tmp
* ci_complete ci_coverage
* Expanded symmetric_difference tests
* ci_complete ci_coverage
* I don't even know
* ci_complete ci_coverage
* Revert mathstuff changes
* ci_complete ci_coverage
* apt: fix virtual package install version detection
Change 4a62c4e3e4 introduced version
matching in installation.
The problem stems from
if version_installable or version:
pkg_list.append("'%s=%s'" % (name, version_installable or version))
When the package is a virtual-package, package_status() is returning
the "version_installable" of the package *satisfying* the
virtual-package; but then this is trying to install the
virtual-package with this version pin.
For example, "yaml-mode" is a virtual package satisifed by
"elpa-yaml-mode" (currently 0.0.14-1) and trying to install it fails
with
$ usr/bin/apt-get -y ... install 'yaml-mode=0.0.14-1'
... failed: E: Version '0.0.14-1' for 'yaml-mode' was not found ...
In the case of a virtual-package with nothing installed to satisfy it,
we should just return blank values to allow apt-get to do it's thing.
The tests are updated to install and remove this package.
Fixes: #76779
* Fix traceback when a supported version of resolvelib is not installed
Try to read the supported version range from the package distribution info and fall back to a hardcoded lowerbound/upperbound (>=0.5.3,<0.6.0).
* Add tests for unsupported resolvelib versions
* Resolve remaining import sanity test issues.
Co-authored-by: Matt Clay <matt@mystile.com>
Co-authored-by: Matt Martz <matt@sivel.net>
* If there is a platform specific handler, prefer the resolved module over the resolved action when loading module_defaults
Add a toggle for action plugins to prefer the resolved module when loading module_defaults
Allow moving away from modules intercepted as actions pattern
Fixes#77059
* Add intentional test coverage for lib/ansible/plugins/inventory/__init__.py
* Add intentional (integration) test coverage for lib/ansible/module_utils/common/network.py
* Add ansible_release info test from https://github.com/ansible/ansible/pull/74673
ci_complete
ci_coverage
Co-authored-by: Rick Elrod <rick@elrod.me>
* ansible-test - Add a Ubuntu 22.04 container.
* ansible-test - Add a Fedora 36 container.
* ansible-test - Update distro containers.
* Fix dnf test on Fedora 36.
* Work around scp test issues.
* Add new templating target, test splitter
* Add test for host:port parsing in parse_address via add_host
* Test already notified listening handler
* Add test for gathering bridge network facts
* Add veth to bridge
* ci_complete ci_coverage
* make the pre-flight sniffing more robust to different failure conditions (was failing on Ubuntu 22.04 VMs)
* remove skip aliases (the test needs to function everywhere to assert that the selinux facts bits behave properly when it's N/A)
* Run code-smell sanity tests in UTF-8 Mode.
* Update subprocess use in sanity test programs.
* Use raw_command instead of run_command with always=True set.
* Add more capture=True usage.
* Don't expose stdin to subprocesses.
* Capture more output. Warn on retry.
* Add more captures.
* Capture coverage cli output.
* Capture windows and network host checks.
* Be explicit about interactive usage.
* Use a shell for non-captured, non-interactive subprocesses.
* Add integration test to assert no TTY.
* Add unit test to assert no TTY.
* Require blocking stdin/stdout/stderr.
* Use subprocess.run in ansible-core sanity tests.
* Remove unused arg.
* Be explicit with subprocess.run check=False.
* Add changelog.
* Use a Python subprocess instead of a shell.
* Use InternalError instead of Exception.
* Require capture argument.
* Check for invalid raw_command arguments.
* Removed pointless communicate=True usage.
* Relocate stdout w/o capture check.
* Use threads instead of a subprocess for IO.
* Expand ansible-doc to tests/filters and fix existing issues
enable filter/test docs if in single file or companion yaml
add docs for several filters/tests plugins
allow .yml companion for docs for other plugins, must be colocated
verify plugins are valid (not modules, cannot)
fix 'per collection' filtering
limit old style deprecation (_ prefix) to builtin/legacy
start move to pathlib for saner path handling
moved some funcitons, kept backwards compat shims with deprecation notice
Co-authored-by: Abhijeet Kasurde <akasurde@redhat.com>
Co-authored-by: Felix Fontein <felix@fontein.de>
Co-authored-by: Sandra McCann <samccann@redhat.com>
* Prevent losing unsafe from lookups
This patch fixes a bug which under certain conditions results in data
returned from lookups not being marked as unsafe.
Each time Templar.do_template is invoked a new AnsibleContext is
created and stored effectively at two places:
1) as an instance variable in templar_obj.cur_context
2) as a local variable called new_context in do_template method of Templar
Due to custom functionality in Ansible's Context that allows for nested
templating it is possible that during resolving variable's value
template/do_template method is called recursively again, again creating
a new context. At that point the problem manifests itself because as
mentioned in 1) above the context is overwriten on the templar object
which means that any subsequent calls to _lookup will use the new
context to mark it as unsafe which is now different to the local
new_context which is used for testing for unsafe property.
The solution to the problem appears to be to restore the original
context inside do_template and also to eliminate the local variable
new_context to prevent problems in the future.
It appears that we don't have a better way of storing the context other
than as some form of global variable and so this appears to be the
"best" solution possible at this point. Hopefully data tagging will be
the solution here.
For more examples see unit and integration tests included in this patch.
Fixes#77535
* Run code-smell sanity tests in UTF-8 Mode.
* Update subprocess use in sanity test programs.
* Use raw_command instead of run_command with always=True set.
* Add more capture=True usage.
* Don't expose stdin to subprocesses.
* Capture more output. Warn on retry.
* Add more captures.
* Capture coverage cli output.
* Capture windows and network host checks.
* Be explicit about interactive usage.
* Use a shell for non-captured, non-interactive subprocesses.
* Add integration test to assert no TTY.
* Add unit test to assert no TTY.
* Require blocking stdin/stdout/stderr.
* Use subprocess.run in ansible-core sanity tests.
* Remove unused arg.
* Be explicit with subprocess.run check=False.
* Add changelog.
* Proper error on missing jinja2 override separator
Properly catch the case when no (or a wrong) separator is used in a
jinja2 override, and return a useful error message to the user.
* Support colons in jinja2 override value
By limiting the split to 1, any colons in the value are preserved and
passed on.
* Fall back to implicit namespace.name from the path if the metadata is invalid
* Test listing a collection with null namespace/name/version fields in its galaxy.yml
* Add option --no-fail-on-errors to return errors for ansible-doc --metadata-dump in JSON result instead of failing.
* Adjust changelog fragment.
* Add basic tests.
* Support ignoring of certificates for ansible-galaxy during SCM cloning
* Add integration tests installing a role from an untrusted repository
Test installing the role without --ignore-certs fails
Test installing the role with --ignore-certs is successful
* Add a toggle to control the number of signatures required to verify the authenticity of a collection
* Make the default number of required valid signatures 1
* Add option to make signature verification strict and fail if there are no valid signatures (e.g. "+1")
* Use a regex to validate --required-valid-signature-count
* Add a toggle to limit the gpg status codes that are considered a failure
* Update documentation and changelog
* Add unit and integration tests for the new options
* Fixes#77146
Fix using user-provided signatures when running 'ansible-galaxy collection verify ns.coll --offline'
Add a test for a user-provided signature when running ansible-galaxy collection verify with --offline
Fix displaying overall gpg failure without extra verbosity
Add a test for displaying gpg failure without verbosity
Improve documentation to be more clear that signature verification only currently applies to collections directly sourced from Galaxy servers
* apply owner/group permissions to top folder
* remove unused var unarchive30
* fix permissions for top folders if the tarball include multiple top folders
* added test for top folder ownership
* fix facts delegation loop overwrite
partial revert of change to allow facts to be present in each loop iteration
was not needed in final results as result processing alreayd had the disctiontion
and ended up breaking the assumptions in the calling code.
fixes#76676
* ansible-galaxy collection install|verify:
- Support verifying the origin of the MANIFEST.json when the Galaxy server has provided signatures.
- Allow supplemental signatures to use during verification on the CLI/requirements file.
* ansible-galaxy collection install:
- Support disabling signature verification. This silences the warning provided by ansible-galaxy if the Galaxy server provided signatures it cannot use because no keyring is configured.
- Store Galaxy server metadata alongside installed collections for provenance. This is used by 'ansible-galaxy collection verify --offline'.
* Add unit tests for method that gets signatures from a Galaxy server
* Add integration tests for user-provided signature sources
- Test CLI option combinations
- Test installing collections with valid/invalid signature sources
- Test disabling GPG verification when installing collections
- Test verifying collections with valid/invalid signature sources
* Make signature verification advisory-by-default if signatures are provided by the Galaxy server
- Make the default keyring None
- Warn if the keyring is None but the Galaxy server provided signatures
- Error if the keyring is None but the user supplied signatures
- Error if the keyring is not None but is invalid
* changelog
* add ansible-galaxy user documentation for new options
Co-authored-by: Matt Martz <matt@sivel.net>
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
Co-authored-by: Martin Krizek <martin.krizek@gmail.com>
Co-authored-by: Sandra McCann <samccann@redhat.com>
Co-authored-by: Andy Mott <amott@redhat.com>
Co-authored-by: John R Barker <john@johnrbarker.com>
* Return rc=0 on success.
Error handling in playbooks generally expects `rc` to be set to 0 when a module has not failed. Playbook authors should not have to check for the existence of `rc` first.
* Use single definition and added changelog
* Fix up tests with new return value
Co-authored-by: Jordan Borean <jborean93@gmail.com>
* add DebianStrategy tests
* ensure hostname can be changed by using become
* use Systemd strat for debian and Base for generic.
* add test to ensure all strategies are available
Signed-off-by: Thomas Sjögren <konstruktoid@users.noreply.github.com>
ssh plugin, use 'correct' information source in all cases
* still fallback to pc
* added inventory to new test
* undef var can still show as parser error on pc
now task_exectuer has a more accurate error handling
* updated tests to conform to new block inheritance
Co-authored-by: Brian Scholer <1260690+briantist@users.noreply.github.com>
Co-authored-by: Sloane Hertel <19572925+s-hertel@users.noreply.github.com>
* minor refactor in other options by pushing common code into functions
* consolidate coll_filter
* more normalizing loader
* dont pass plugin_loader, its global import
* Also dump roles and collections
* adjusted tests to new err msg
* disable namespace filter (unused)
* Make include_role/include_tasks work with any_errors_fatal v2
Fixes#50897
Co-authored-by: Matt Martz <matt@sivel.net>
* Add failed to results in free strategy too
* Fix
* Avoid duplicating results
* ci_complete
Co-authored-by: Matt Martz <matt@sivel.net>
* ansible-test - Improve help for unsupported cwd.
* The `--help` option is now available when an unsupported cwd is in use.
* The `--help` output now shows the same instructions about cwd as would be shown in error messages if the cwd is unsupported.
* Add `--version` support to show the ansible-core version.
* The explanation about cwd usage has been improved to explain more clearly what is required.
Resolves https://github.com/ansible/ansible/issues/64523
Resolves https://github.com/ansible/ansible/issues/67551
* group2 - switch to setup_remote_tmp_dir instead of using output_dir
* output_file needs to be in the tmp dir
* Use comparison that should work with macos
* remove unused dep
* allow inputting 0 for password_expire_{min|max}
0 is meaningful for min days (any time)
0 is technically valid for max_days
* add test for setting both min and max expiry
* [0] return result of execute_command from set_password_expire*
* [1] better return for set_password_expire
* [2] handle returns from set_password_expire*
* only set password expiry if user exists
* collect return-handling code
* combine password min and max into one execution
* handle case where spwd is not present like on macOS and FreeBSD
Co-authored-by: Sam Doran <sdoran@redhat.com>
* Update dnf tests to reflect new behavior.
Previously dnf would report there was nothing to do when trying to install a package from the cache when it was not present.
A recent update to dnf has changed this behavior to match yum, resulting in a failure instead.
* Allow dnf to fail or report no changes.
This adds the ability to specify a package version using >=. This will ensure the package is at the specified version or above.
* If the package is not installed, the latest version will be installed.
* If the package is installed and less than the specified version, it will be upgraded.
* If the package is installed and greater than or equal to the specified version, it will be left alone.
The version selection is handled by Apt itself, so things like the system policy, pinning, etc, are considered.
* Add additional tests for check_mode with shell/command/script
* update check_mode documentation: if a module/action is not skipped in check mode and accurately reflects whether a change is made to the remote support should be 'full'
* Make reporting skipped in check mode mutually exclusive with 'changed: True'
* Add missing documented attributes
* Fix tests to expect skipped=True and changed=True do not occur together
* Fix script check_mode support documentation
* Fix earlier changelog
* document platforms attribute
* Use tasks's check mode since the value from PlayContext does not reflect loop items
Co-authored-by: Brian Coca <bcoca@users.noreply.github.com>
* Ensure the version is valid for directories and collections in git repos before installing
Fix the error message for invalid semantic versions
* Make requested changes
* Add a test case for unhandled ValueError exception
* Add changelog
* Update lib/ansible/galaxy/collection/galaxy_api_proxy.py
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* Reword error message
Include link to learn how to compose a semver version
* Move version validation into the caller, find_matches
* Add tests for more invalid version types
* Remove unused import
Fix raising unexpected error
* Update lib/ansible/galaxy/collection/__init__.py
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* Update lib/ansible/galaxy/dependency_resolution/providers.py
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* Update lib/ansible/galaxy/dependency_resolution/providers.py
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* fixes#28977
* Include the original error message when pick_handler fails
* Add a test to unarchive a tar file with an invalid extra option
* add a changelog
* Disable the _distutils_hack in newer setuptools. Doesn't fix the underlying issue of the venv finding the _distutils_hack of a setuptools that is not its own.
ci_complete
* re-throw blanket
* check if service is indirect status / add integration test
* Update changelogs/fragments/76453-indirect-systemd-status.yml
Co-authored-by: Abhijeet Kasurde <akasurde@redhat.com>
* simplify indirect check / remove integration test restriction on rhel9
* add dummy indirect service for integration tests
* unarchive: fix non-english locales
For GNU Gettext, the LANGUAGE environment variable takes precedence over LANG or LC_ALL. On systems where LANGUAGE was set to a non-english locale, the output of the tar command therefore not understood and the module failed silently ("changed": false, but the archive was not extracted).
* add tests
* changelog
Ref #44741
Ref ansible/proposals#140
* Replace select filter with a more portable thing
* Add context
This is needed for split controller/remote
Co-authored-by: Sam Doran <sdoran@redhat.com>
* Add changelog fragment and integration tests
* Break up function signature into multiple lines
It passed the 160 character width. Change it to multi-line for easier reading
and better diffs in the future.
* Remove unused import
* Update version added for new feature
* Move changelog fragment to correct location
Co-authored-by: Thomas Stähle <thomas.staehle@check24.de>
Co-authored-by: Sam Doran <sdoran@redhat.com>
* ansible-test - Add support for RHEL 8.5 instances.
* Clear libdnf checksum cache from unsigned package in dnf integration test.
Co-authored-by: Matt Martz <matt@sivel.net>
* First pass of adding yaml result format output to default callback
* Add clog frag
* flow style always false
* Further normalization of results across distros
* no lossy, now pretty
* Fix env var in runme.sh
* Rename variable to better self document
* include NativeJinjaUnsafeText
* Linting cleanup
* Add tests specific to the prior yaml callback
* Make data munging more exact, following the checks in libyaml/pyyaml
* Remove unused import
* Extend comment
* more correct and some comments
* Fix consistency of tab vs space in this section of the file
* Add str representer
* be cooler
* faster character filtering
* Clarify None as a sentinel for default behavior
* Flip filtering logic to avoid CRLF issues with 'space break' filtering
* Py3 change and comment about the use of __call__
Co-authored-by: Matt Davis <mrd@redhat.com>
This task was previously added in https://github.com/ansible/ansible/pull/25243 as a work-around for a setuptools bug.
The pinned version does not work with Python 3.10, and the task should no longer be needed.
* Move RHEL pip package install into function.
* Add RHEL 9 support to bootstrap script.
* ansible-test - Add RHEL 9 support.
* Add install and uninstall instructions for RHEL9 to paramiko tests
* Skip incidental_inventory_docker_swarm on RHEL 9.
* Update dnf tests for rhel9
* Fix prepare_http_tests on RHEL 9.
* fix RHEL9 Python discovery tests to match distro map
* Skip subversion test on RHEL 9.0b.
Co-authored-by: Matt Martz <matt@sivel.net>
Co-authored-by: Matt Davis <mrd@redhat.com>
* Skip recursive suboption validation if sub_parameters is not a dictionary
* Ensure sub parameter elements is a sequence to prevent iterating over string characters and causing duplicate error messages for the same param
* Try to find any valid vault id before raising an exception
* Apply the same logic during vault password file loading
* Also. Improve the warning msg and tests
When we introduced an include parameter to the unarchive module, we
inadvertenly flipped the exclusion logic. This flip meant that the
unarchive module started rejecting files that should be extracted.
This commit flips the bad logic and adds some tests that will make
sure things do not go bad again.
* Use NativeEnvironment for all templating
ci_complete
* Keep Templar.copy_with_new_env for backwards compat
* Mention that AnsibleUndefined.__repr__ changed in the porting guide
* Templar.copy_with_new_env backwards compat
* ci_complete
* Overhaul ansible-test integration tests.
* ansible-test - Fix import test pyyaml usage.
* ansible-test - Remove unused import.
* ansible-test - Fix traceback when pip is unavailable.
* ansible-test - Fix typo in port forwarding message.
* ansible-test - Fix controller logic in requirements install.
* Fix unit tests in ansible-test integration test.
Unit tests are now run for available Python versions which
provide `virtualenv` (Python 2.x) or `venv` (Python 3.x).
* Don't use output_dir in user tests
* Move blockinfile tests from using output_dir to depending on setup_remote_tmp_dir
* Don't use output_dir in git tests
* Don't use output_dir in uri tests
Co-authored-by: Matt Clay <matt@mystile.com>
Co-authored-by: Matt Davis <mrd@redhat.com>
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* Use the task dependency chain to compile the role instead of the parent role's _parents list
* keep existing role inheritance, but limit it to roles in the current task dependency chain
* Test getting role parents recursively by using the current task dependency chain
* changelog
* add tests for fail filter
also tests that fail does not block inspectability
* add fail filter
fallback message is a bit clunky,
since you can't invoke a filter without specifying an input.
That is, "{{ fail }}" doesn't work,
so you have to do "{{ None | fail }}"
* document 'fail' filter
* add changelog fragment
* fail filter uses default message on Undefined or emptystring
makes it slightly easier to use the default message:
```diff
- "{{ None | fail }}"
+ "{{ '' | fail }}"
```
and the user sees a slightly more relevant message
if the message itself is undefined:
```diff
- The error was: {{ failmsg | fail }}: 'failmsg' is undefined
+ The error was: {{ failmsg | fail }}: Mandatory variable has not been overridden
```
* rebuild as the builtin `Undefined`
* harmonise `hint` parameter for make_undefined with jinja
* use code block for documentation item
[ref](https://github.com/ansible/ansible/pull/75435#discussion_r707661035)
* rename to `undef` to expose less Python into the Jinja
[ref](https://github.com/ansible/ansible/pull/75435#pullrequestreview-757799031)
* explicitly instantiate undefined value now that it's possible
see I knew we would break something with reflection
* preserve test coverage of undefined variable
Co-authored-by: Matt Davis <nitzmahone@users.noreply.github.com>
Change:
- Remove only user-facing use of ANSIBLE_ASYNC_DIR.
- Remove two comments saying to change things that, apparently, we
aren't going to change...
Test Plan:
- ci_complete
Tickets:
- Fixes#74139
- Fixes#74138
- Refs #74226
Signed-off-by: Rick Elrod <rick@elrod.me>
* Fix test usage of `which python`.
Also use `command -v` instead of `which` where needed.
* Fix testing_formatter.sh file checking.
The index is longer expected to exist when generating it.
The generation script is run with `set -eux`.
* apt module: add option to allow package downgrades
* Add new option to module so users don't have to force downgrades which
is insecure and dangerous
* Add integration tests similar to upgrade integration tests
* Changelog
* Update changelog fragment
* Update changelogs/fragments/74852-apt-allow-downgrade.yaml
Co-authored-by: Amin Vakil <info@aminvakil.com>
* Update lib/ansible/modules/apt.py
Co-authored-by: Amin Vakil <info@aminvakil.com>
* Update lib/ansible/modules/apt.py
Co-authored-by: Amin Vakil <info@aminvakil.com>
Co-authored-by: Amin Vakil <info@aminvakil.com>
* ensure dump produces json
* clog and tests
* remove library additions since they are not usable
* avoid any/all dirs
* ensure we dont use local dir as playbookdir
* If an exception occurs when getting a collection's metadata, continue to the next in the server list.
* Warn for unknown exceptions when finding versions of a collection
* Test that an invalid server is no longer fatal if a subsequent server has the collection
* Fix server for verify tests - compare checksums against the server from which it was installed
* Add tests for verify and fix that code path to mirror install/download behavior for server errors
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* Make individual dirs that should exist instead of using shutil.copytree
* Test build_ignore when installing collections in git repos
* changelog
* Fix assertion
Fix git directory
* Fix nested content while building the collection dir
Test installing a collection from a git repo and artifact have the same result
Refactor to use variables
* Support templating _from_files with --extra-vars for imported roles
* Add tests for templating the filenames for import_role, import_playbook, and import_tasks
* Add documentation
Co-authored-by: Alicia Cozine <879121+acozine@users.noreply.github.com>
Co-authored-by: Martin Krizek <martin.krizek@gmail.com>
* avoid getting delegated vars w/o inventory host
fixes#75512
In the case of imports, we don't have a host, so getting host vars for
the delegated host makes no sense and should be avoided.
* also avoid error on vars_files with per host vars
* test
* testing given case
* oops
* Add new comment attribute to template plugin
Add comment_start_string and comment_end_string attribute to template
plugin
Co-authored-by: Hossein Zolfi <h.zolfi@inside.sahab.ir>
* Improve error for InconsistentCandidate exceptions
* Add test case for installing a collection with an inconsistent version
* Add test case for installing a collection that has a dependency with an inconsistent version
Co-authored-by: Abhijeet Kasurde <akasurde@redhat.com>
Co-authored-by: Sviatoslav Sydorenko <webknjaz@redhat.com>
* Fix a bug with the dnf module not using all components of a package name when filtering to determine if it's installed
* changelog
* Simplify splitting on the last '.'
* Update lib/ansible/modules/dnf.py
Push /usr/bin/python to almost the bottom of the barrel. This makes the strategy to prefer
specific versions of the "mystery meat" version.
* Change INTERPRETER_PYTHON default to 'auto'
Change description to match change in behavior.
Change deprecation message to a warning.
* Update docs
* Add porting guide entry
* Update unit tests
* Update integration test
* Allow INTERPRETER_PYTHON_FALLBACK to be configure using a variable
* Prefer platform-python above other Python 2 interpreters
* Add Python 3.10 to the list of interpreters
* Make undefined variables in handler names non-fatal if the handler is not used
* If the handler has no way to be notified (i.e. the name can't be templated and the handler has no listen topics), display a warning
* Add tests for variables in handler names
* changelog
test/integration/targets/any_errors_fatal/18602.yml is not run via
runme.sh and is testing the same as
test/integration/targets/any_errors_fatal/always_block.yml which is
actually run. The former file is safe to remove.
* Add packaging to requirement of ansible-test
Fix#75353
After requires_ansible field was added as mandatory to runtime.yml
file, ansible-test fails to check this field if it doesn't have
packaging module.
[1] https://github.com/ansible/galaxy-importer/pull/124
* Canonicalize module_defaults actions and action_groups pre-fork and cache them on the play
* Call get_action_args_with_defaults with the resolved FQCN plugin and don't pass the redirect list
* Add validation for action_group metadata and a toggle to disable the warnings
* Handle groups recursively referring to each other
* Remove special-casing for non-fqcn actions in module_defaults groups
* Error for actions and groups in module_defaults that can't be resolved
* Error for fully templated module_defaults
* Add integration tests for action_groups
* Changelog
Since we moved yum_repository_test_repo.baseurl to file:/// format,
using urlsplit filter on it returned an empty string. Using an empty
string as a left operand of `in` was then always evaluated to True.
This was discovered with native jinja being on where urlsplit returns
None which results in a hard fail as None cannot be a left
operand of `in <string>`.
* The resolved_action is the formatted version of the final plugin in the PluginLoadContext's redirect_list
* Collection plugins are represented as FQCN
* Legacy plugins are represented with only the plugin name
* Add tests
* Changelog
Felix Fontein