Fixes#74255
* Fix call to 'unique(case_sensitive=False)' triggering error when falling back to Ansible's version which **is** case-sensitive
* Test multiple situations of 'unique' filter errors with fallback not handling specific parameters
Signed-off-by: Rick Elrod <rick@elrod.me>
Co-authored-by: Rick Elrod <rick@elrod.me>
* Add yaml utils file and use it
* Linting fix and missing import
* Abstract a few more details
* Parser imports
* Don't use CSafeDumper in AnsibleDumper
* Move and use convert_yaml_objects_to_native when libyaml is present
* yaml_load and yaml_dump, instead of safe_*
* re-use HAS_LIBYAML from utils.yaml
* add changelog fragment
* Address recent changes
* Use representer instead of recursive type converter
* Restore needed import
* move yaml utils to module_utils
* Properly guard imports
* Update from_yaml(_all)? to handle text wrappers with CSafeLoader
* Use yaml utils for legacy_collection_loader
* Add HAS_YAML, and ignore pylint issue
* oops
* GPL->BSD
Change:
- Newer Solaris drops setfacl. Add a fallback for its chmod ACL syntax.
Test Plan:
- New units
Tickets:
- Fixes#74282
Signed-off-by: Rick Elrod <rick@elrod.me>
Change:
- Regression introduced in #70785
- When macOS chmod ACL syntax is used, Solaris-derived chmods return
with a status of 5. This is also used for our sshpass handling,
because sshpass will return 5 on auth failure. This means on Solaris,
we incorrectly assume auth failure when we reach this branch of logic
and try to run chmod with macOS syntax.
- We now wrap this specific use of chmod in an exception handler that
looks for AnsibleAuthenticationFailure and skips over it. This adds
another authentication attempt (something we normally avoid to prevent
account lockout), but seems better than the regression of not allowing
other fallbacks to be used.
- Without this patch, if setfacl fails on Solaris (and sshpass is used),
we do not try common_remote_group or world-readable tmpdir fallbacks.
Test Plan:
- New unit
Signed-off-by: Rick Elrod <rick@elrod.me>
* Handle netconf plugin ncclient import error when running in FIPS mode
* While running in FIPS mode importing ncclient result in
InternalError raised by cryptography
* Refer https://github.com/ansible/ansible/pull/65477
This simplifies rendering the hostname (or hostname+delegated host) in
the default callback module, and reduces code duplication
I've chosen not move where in each handler the host label is rendered,
in case subsequent operations has side effects. However I'm happy to
change that if considered safe.
I've chosen not to change the formatting operator used (%), to avoid
changes in rendering that might result.
Signed-off-by: Alex Willmer <alex@moreati.org.uk>
Change:
- We were passing a directory name directly to re.compile().
If the directory isn't valid regex (or is) this can have odd side
effects, such as crashing.
- Fix a few other similar cases, but less likely to be a real issue.
Test Plan:
- New test
Signed-off-by: Rick Elrod <rick@elrod.me>
* all lookups to support config system
- added get_options to get full dict with all opts
- fixed tests to match new error messages
- kept inline string k=v parsing methods for backwards compat
- placeholder depredation for inline string k=v parsing
- updated tests and examples to also show new way
- refactored and added comments to most custom k=v parsing
- added missing docs for template_vars to template
- normalized error messages and exception types
- fixed constants default
- better details value errors
Co-authored-by: Felix Fontein <felix@fontein.de>
* Support omitting the trailing separator when a dictionary key's value is an empty string
* Support a default value when the value used in the group name is an empty string
* Add tests
* change log
Change:
- Tests for advanced_host_list; it is now at 100% coverage.
- There was a typo (host vs hostnames) where when the host pattern
failed to parse, instead of treating the name as a literal as
intended, it would trigger UnboundLocalError. This didn't fatal
as there's a global "Exception" handler below, but it did lead to
an ugly error and incorrect behavior.
Test Plan:
- New tests
- Local experimenting
Signed-off-by: Rick Elrod <rick@elrod.me>
* changelog
Signed-off-by: Rick Elrod <rick@elrod.me>
* fix tests
Signed-off-by: Rick Elrod <rick@elrod.me>
use load_name of valid plugin to find config, will handle the renamed loaded plugin that collections do:
`ansible_collection.<ns>.<collname>.plugins.<type>.<name>`
* Begin using ArgumentSpecValidator in AnsibleModule
* Add check parameters to ArgumentSpecValidator
Add additional parameters for specifying required and mutually exclusive parameters.
Add code to the .validate() method that runs these additional checks.
* Make errors related to unsupported parameters match existing behavior
Update the punctuation in the message slightly to make it more readable.
Add a property to ArgumentSpecValidator to hold valid parameter names.
* Set default values after performining checks
* FIx sanity test failure
* Use correct parameters when checking sub options
* Use a dict when iterating over check functions
Referencing by key names makes things a bit more readable IMO.
* Fix bug in comparison for sub options evaluation
* Add options_context to check functions
This allows the parent parameter to be added the the error message if a validation
error occurs in a sub option.
* Fix bug in apply_defaults behavior of sub spec validation
* Accept options_conext in get_unsupported_parameters()
If options_context is supplied, a tuple of parent key names of unsupported parameter will be
created. This allows the full "path" to the unsupported parameter to be reported.
* Build path to the unsupported parameter for error messages.
* Remove unused import
* Update recursive finder test
* Skip if running in check mode
This was done in the _check_arguments() method. That was moved to a function that has no
way of calling fail_json(), so it must be done outside of validation.
This is a silght change in behavior, but I believe the correct one.
Previously, only unsupported parameters would cause a failure. All other checks would not be executed
if the modlue did not support check mode. This would hide validation failures in check mode.
* The great purge
Remove all methods related to argument spec validation from AnsibleModule
* Keep _name and kind in the caller and out of the validator
This seems a bit awkward since this means the caller could end up with {name} and {kind} in
the error message if they don't run the messages through the .format() method
with name and kind parameters.
* Double moustaches work
I wasn't sure if they get stripped or not. Looks like they do. Neat trick.
* Add changelog
* Update unsupported parameter test
The error message changed to include name and kind.
* Remove unused import
* Add better documentation for ArgumentSpecValidator class
* Fix example
* Few more docs fixes
* Mark required and mutually exclusive attributes as private
* Mark validate functions as private
* Reorganize functions in validation.py
* Remove unused imports in basic.py related to argument spec validation
* Create errors is module_utils
We have errors in lib/ansible/errors/ but those cannot be used by modules.
* Update recursive finder test
* Move errors to file rather than __init__.py
* Change ArgumentSpecValidator.validate() interface
Raise AnsibleValidationErrorMultiple on validation error which contains all AnsibleValidationError
exceptions for validation failures.
Return the validated parameters if validation is successful rather than True/False.
Update docs and tests.
* Get attribute in loop so that the attribute name can also be used as a parameter
* Shorten line
* Update calling code in AnsibleModule for new validator interface
* Update calling code in validate_argument_spec based in new validation interface
* Base custom exception class off of Exception
* Call the __init__ method of the base Exception class to populate args
* Ensure no_log values are always updated
* Make custom exceptions more hierarchical
This redefines AnsibleError from lib/ansible/errors with a different signature since that cannot
be used by modules. This may be a bad idea. Maybe lib/ansible/errors should be moved to
module_utils, or AnsibleError defined in this commit should use the same signature as the original.
* Just go back to basing off Exception
* Return ValidationResult object on successful validation
Create a ValidationResult class.
Return a ValidationResult from ArgumentSpecValidator.validate() when validation is successful.
Update class and method docs.
Update unit tests based on interface change.
* Make it easier to get error objects from AnsibleValidationResultMultiple
This makes the interface cleaner when getting individual error objects contained in a single
AnsibleValidationResultMultiple instance.
* Define custom exception for each type of validation failure
These errors indicate where a validation error occured. Currently they are empty but could
contain specific data for each exception type in the future.
* Update tests based on (yet another) interface change
* Mark several more functions as private
These are all doing rather "internal" things. The ArgumentSpecValidator class is the preferred
public interface.
* Move warnings and deprecations to result object
Rather than calling deprecate() and warn() directly, store them on the result object so the
caller can decide what to do with them.
* Use subclass for module arg spec validation
The subclass uses global warning and deprecations feature
* Fix up docs
* Remove legal_inputs munging from _handle_aliases()
This is done in AnsibleModule by the _set_internal_properties() method. It only makes sense
to do that for an AnsibleModule instance (it should update the parameters before performing
validation) and shouldn't be done by the validator.
Create a private function just for getting legal inputs since that is done in a couple of places.
It may make sense store that on the ValidationResult object.
* Increase test coverage
* Remove unnecessary conditional
ci_complete
* Mark warnings and deprecations as private in the ValidationResult
They can be made public once we come up with a way to make them more generally useful,
probably by creating cusom objects to store the data in more structure way.
* Mark valid_parameter_names as private and populate it during initialization
* Use a global for storing the list of additonal checks to perform
This list is used by the main validate method as well as the sub spec validation.
* Fix a bug adding unrelated candidates to the plugin loader redirect_list
* Add tests for the redirect list
* test redirect list for builtin module
* test redirect list for redirected builtin module
* test redirect list for collection module
* test redirect list for redirected collection module
* test redirect list for legacy module
* changelog
* finish migrating ssh plugin to config system
fixes#72739fixes#57220
* fix connection detection in reset
* correct options for connection meta reset
Co-authored-by: David Shrewsbury <Shrews@users.noreply.github.com>
Adds in the caveat that Python's string modules have different constants depending on the major version (2.x vs 3.x), but preserves the fact that a fair amount of them are common. Other minor docs touchups also included.
* 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
Allow constructed to optionally use vars plugin data
* mostly for those looking to leverage group_vars/ and host_vars/
* limited to already processed sources
When running in verbosity <2, display the file and line number for tasks that fail. This provides
useful information without having to run at increased verbosity.
* Move _print_task_path to CallbackBase class
* Add integration tests
* Add color parameter to _print_task_path()
* Keep color output consistent for now
Currently the path is display with COLOR_DEBUG formatting with verbosity >= 2.
Instead of the color of the path changing based on verbosity level, just keep it at the
currently behavior of COLOR_DEBUG. Having the color of the same information change
based on verbosity level seems incorrect and makes visual parsing of the information
more difficult.
Co-authored-by: tahar.jegham <jeghamseifeddine@gmail.com>
* Fix regression introduced in b77abd0491 causing bug in inventory modules which break functionality in user setting use_contrib_script_compatible_sanitization parameter.
* Add changelog
Co-authored-by: s-hertel <19572925+s-hertel@users.noreply.github.com>
Use correct ssh executable and options in all cases on connection plugin
* Also nicer naming/comments
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
* Consolidate filters/tests handling into JinjaPluginIntercept
ci_complete
* Postpone loading all ansible plugins
* Do we need to create an overlay?
ci_complete
* Typo
ci_complete
* Add FIXME
* conditional.py: use public Environment.parse() method
* Remove remaining occurrences of shared_loader_obj being passed to Templar
* __UNROLLED__ not needed with this change anymore
* Incorrect rebase at some point?
* Allow restricting config values to enumerated list
* dont document internal entries
* also ignore private defs for ansible-config
* remove invalid value from tests
* added porting entry
When the pause module is run in the background and seconds parameter is provided,
do not warn.
* Add tests
* Fix existing tests
The test wasn't failing when it should have.
The old example failed with:
fatal: [localhost]: FAILED! => {"msg": "to count backwards make stride negative"}
I fixed this and placed a start parameter to consolidate with the rest of the examples provided.
Co-authored-by: Jorge Rua <jorua@redhat.com>
* Consolidate logic for determining whether or not session is interactive
into a single function, is_interactive()
* Increase test coverage
I wasn't able to find a good way of simulating running a backgrounded test with CI since the
whole test is essentially run not in a TTY, which is similar enough to cause the new is_interactive()
function to always return false.
* Fix 'role_name : tast_name' notation if task contains role name
* Add tests for notifying handler names which contain the role name
Co-authored-by: Thomas Wouters <thomaswouters@gmail.com>
* fixes for collection playbooks
- add fqcn invocation, also handles extensions
- bring import_playbook into new normal
- avoid adding collection playbooks to adjacency
- deal with extensions or lack of em
- added tests
- fix bugs with multiple playbooks supplied
- unicode/bytes/native fixes
- fix modulenotfound for < py3.7
- tests
* Ensure we call action_loader.get with collection_list. Fixes#72170
* Add tests and changelog
* Remove grep, do assertion in playbook. ci_complete
* Skip old jinja2 versions
* ci_complete
* dedupe
* Add support for GSSAPI/Kerberos to urls.py
* Test out changes with the latest test container
* Get remote hosts working
* Fix up httptester_krb5_password reader
* Fix tests for opensuse and macOS
* Hopefully last lot of testing changes
* Dont do CBT on macOS
* Fixes from review
* Fix various sanity errors in plugins.
* Revert callback_type -> type transform.
* Undo paramiko_ssh 'connection' change, since this discrepancy is hardcoded in the plugin loader.
* Standardize on name and type (for callbacks).
* Fix existing author entries.
* Add 'Unknown (!UNKNOWN)' as author when author is missing.
* These are actually parsed as integers in the code.
* Revert "Add 'Unknown (!UNKNOWN)' as author when author is missing."
This reverts commit 29d1438aca.
Change:
- In some cases (always with free strategy, sometimes with linear), the
default callback would not show the task banner for include_tasks.
- This only affects the include_tasks task itself, not the tasks in the
included file.
Test Plan:
- Updated default callback tests
Tickets:
- Fixes#71277
Signed-off-by: Rick Elrod <rick@elrod.me>
* Use InventoryManager for inventory_hostnames lookup. Fixes#17268
* Add test for using patterns
* s/it's/its/
* Allow bypassing parse_sources in InventoryManager
* Remove unneeded method call
Remove "mode: preserve" option from documentation in doc fragments "FILE_COMMON_ARGUMENTS", as it was incorrectly included in the documentation for the `lineinfile`, `unarchive` and other file-related modules.
The `copy` and `templates` modules documentation remains untouched and still contain "mode: preserve", as intended.
* Validate salt when using crypt. Respect salt_size in password lookup. Repair salt for bcrypt. Fixes#71107. Fixes#53750. Fixes#36129.
* Handle algorithms we don't know about, and make sure to return the salt
* Account for old passlib
* Add tests for salt constraints
* Add changelog fragment
* Add test for #36129
* added configurable dir for tree callback
- allows usage in playbooks
- also made errors more specific if bad dir is supplied
- ensure we expand paths
* docfix
This script is mostly unmaintained and relies on the deprecated and
unmaintained `boto` library. Featuring it prominently in the docs
leads to many new users using it instead of the supported `aws_ec2`
inventory plugin.
Change:
- Now sends meta tasks to the task start callback
- Lets callback plugins opt-in to receiving implicit tasks
Test Plan:
- New integration tests
Tickets:
- Indirectly fixes#71007 by allowing custom callbacks with this data
Signed-off-by: Rick Elrod <rick@elrod.me>
skip_missing parameter in subelements lookup plugin is accepted from
inside the dictionary.
Fixes: #38182
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
