* include_role now behaves more like task on error
changes _from errors from syntax to task failures, by default
which makes it more consistent with other existing errors
* also force 'missing role' to behave as syntax error when false
* also error when subdir does not exist, previouslly we ignored missing
file
* add 'rescuable' toggle to allow user to chose error type
Co-authored-by: Abhijeet Kasurde <akasurde@redhat.com>
Co-authored-by: Sloane Hertel <19572925+s-hertel@users.noreply.github.com>
show_origin and variable sources were broken for base config when 'forked' from plugins
---------
Co-authored-by: 🇺🇦 Sviatoslav Sydorenko (Святослав Сидоренко) <wk.cvs.github@sydorenko.org.ua>
Removes the warning emitted when using Add-Type and the cleanup of temp
files fails due to a file still being in use. The cleanup should be
handled by AnsibleModule on exit giving it more time to wait for any
open file handles to close. The exception is still present if calling
`Add-CSharpType` without an `AnsibleModule` object.
Apparently `codecovcli send-notifications` does not have a `--dry-run`
CLI option. This patch stops adding it to the command and implements
an external `dry-run` mode in the wrapper script or this case instead.
This is a follow-up for #85968.
Co-authored-by: Matt Clay <matt@mystile.com>
There were couple of occurrences where the hard 30 seconds limit on
running ssh-agent was not enough for the test to run and the ssh-agent
was killed resulting in the test failing with "Connection refused". This
change just lets the agent run in the background and kills it
manually after the tests finish.
* psrp - ReadTimeout exceptions now mark host as unreachable
* add try to _exec_psrp_script
* fix indent E111
* update raise format
switch to raise Exception from e
Co-authored-by: Jordan Borean <jborean93@gmail.com>
---------
Co-authored-by: Jordan Borean <jborean93@gmail.com>
The logs were displaying a series of numbers in parens like `(66.1.0)`
at the end of each error line. its unintuitive what that means. I had
to look into the source code to confirm my suspicion of it being the
version of `setuptools`. This patch spells it out.
This patch drops unnecessary default for
`CollectionDependencyProvider`'s `concrete_artifacts_manager` argument
as it is always passed, in every place across the code base where the
provider is constructed.
It was also causing MyPy violations on calls to
`_ComputedReqKindsMixin.from_requirement_dict()` in the "strict
optional" mode which is now enforced for $sbj, while remaining
disabled globally.
It is a #85545 follow-up.
This patch is a combination of `pyrefly autotype` and manual
post-processing. Parts of it migrate pre-existing comment-based
annotations, fixing incorrect ones where applicable.
The change also configures MyPy to run checks against actual
`resolvelib` annotations and includes a small tweak of
`ansible.galaxy.collection._resolve_depenency_map` to make it
compatible with those.
Co-Authored-By: Jordan Borean <jborean93@gmail.com>
Co-Authored-By: Matt Clay <matt@mystile.com>
Co-Authored-By: Sloane Hertel <19572925+s-hertel@users.noreply.github.com>
* Remove support for resolvelib < 0.8.0
Remove code handling differences between resolvelib 0.5.3 and 0.8.0
Drop some versions from the test to reduce the time it takes to run
Co-authored-by: Sviatoslav Sydorenko <wk@sydorenko.org.ua>
* Remove type annotation
---------
Co-authored-by: Sviatoslav Sydorenko <wk@sydorenko.org.ua>
* When ssh-keygen fails, return rc and stderr in fail_json
in order to help debugging.
Fixes: #85850
Signed-off-by: Abhijeet Kasurde <Akasurde@redhat.com>
PR #85888
`codecov-cli == 11.0.3` allows `click == 8.3.0` in its deps but the latter causes commit auto-discovery breakage in the former. With https://github.com/getsentry/prevent-cli/pull/95, `codecov-cli == 11.2.3` excludes this version so this patch updates the requirement to that.
To prevent this from happening again, the change also makes use of a pip constraint file that pins the entire dependency tree to concrete versions. The constraint file is managed by `pip-tools`.
Refs:
* https://github.com/getsentry/prevent-cli/pull/95
* https://github.com/pallets/click/issues/3066
ci_coverage
ci_complete
* Prevent run_command output truncation or hang
In cases when the selector used to monitor stdout/stderr activates without
data ready to read (a rare but normal condition), a read from a non-blocking
FD can return `None`, which was being conflated with an empty read (EOF)
condition. This caused the selector to be unregistered prematurely,
sometimes resulting in truncated output or hangs. `None` read results
are now excluded from EOF conditions.
* add changelog
---------
Co-authored-by: Matt Davis <nitzmahone@redhat.com>
* Add new play keyword validate_argspec
Set to True to use the play name as the argument spec identifier. A play name is required (i.e. a host pattern is not supported as an argument spec name).
Alternatively, set to a specific argument spec name.
A valid argument spec for the play is required. Example:
# playbook.meta.yml
argument_specs:
name of the play:
options: {}
* Play argument spec validation runs after fact gathering
Play keywords like tags are inherited and work similarly to fact gathering
* Add test for sidecar documentation for filter plugin in a subdirectory
Fix ansible-doc --list/--list_files/--metadata-dump for relative imports in nested filter/test plugin files
* ansible-galaxy - remove deprecated v2 API
Update unit tests exercising the v2 Galaxy API
Remove v2 API integration tests
* Remove internal test configuration for servers which support API v2 and v3
* Update and simplify verify tests by using the configured server list
On WSL, `os.rename` can't correctly move a file while a handle to that file is still open. It remains half-moved where neither the source or destination seem to exist (according to `os.path.exists`). However the move seems to complete correctly when the open handle is closed.
In `BaseFileCacheModule`, when writing a cache file, a temporary file is created with `mkstemp` that returns an open file descriptor and a filename. Once the cache is written to that file, it is renamed to the correct file name with `os.rename` and then its permissions set with `os.chmod`. On WSL the `os.chmod` fails because it doesn't think the file exists yet because the file descriptor returned by `mkstemp` is still open. This PR fixes this by closing that file descriptor before renaming.
Co-authored-by: Abhijeet Kasurde <akasurde@redhat.com>
