Commit Graph

5 Commits (853755490b4ee69a7c51b3df39b89328b5256730)

Author SHA1 Message Date
Matt Clay 9f899f9492
Require `from __future__ import annotations` (#81902) 1 year ago
Matt Davis 1368bfa348
PluginLoader now installs module-to-be-imported in sys.modules before exec (as Python import does). (#78364) 2 years ago
Matt Martz 66a83314b9
Modernize install (#76021)
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>
3 years ago
Rick Elrod cdcf0aa42a
Intentional coverage, incidental_azure_rm_resource (#71052)
Change:
- Adds some intentional coverage around PluginLoader for cases that
  incidental_azure_rm_resource covered.
- Specifically, modules starting with an underscore, and starting with
  an underscore but a symlink.

Test Plan:
- CI

Signed-off-by: Rick Elrod <rick@elrod.me>
4 years ago
Toshio Kuratomi 0633f73faf Fix loader for filters (#37748)
* Fix loading of filter and test plugins

Filter and test plugins are different than other plugins in that they
can have many plugins in a single file.  Therefore they need to operate
a little differently.  They need to have all of the potential files
returned.  Then the caller takes care of passing those onto jinja2 in
order for jinja2 to make use of them.

This problem was (most recently) introduced with f921369445

This commit also restructures how we deduplicate plugins to take paths
into account.  If we want to start scoping which set of modules are
loaded (due to roles, for instance) we'll need to hang on to the path
information.

* add integration test for override

* Fix style checks for bcoca code

* Implement jinja2 plugin loader as a subclass

Having a subclass allows us to customize the overriding of jinja
plugins.  We can then move common parts of common code into the Loader.
7 years ago