dconf: parse GVariant values to check for equality whenever possible
Direct string comparisons are an inaccurate way to compare two
GVariant representations. For example, 'foo' and "foo" (including the
quote marks, which are part of the representation) are equal GVariants
but if you just do a string compare (remember, including the quotes)
they'll be interpreted.
We therefore want to use the `gi.repository` Python library to parse
GVariant representations before comparing them whenever possible.
However, we don't want to assume that this library will always be
available or require it for Ansible to function, so we use a straight
string comparison as a fallback when the library isn't available. This
may result in some false positives, i.e., Ansible thinking a value is
changing when it actually isn't, but will not result in incorrect
values being written into `dconf`.
Co-authored-by: Jonathan Kamens <jik@jik5.kamens.us>
The fix ensures that no 'KeyError' is raised, when 'access_level' is not provided as module parameter or when 'access_level_on_creation' is false.
Signed-off-by: Christoph Fiehe <c.fiehe@eurodata.de>
Co-authored-by: Christoph Fiehe <c.fiehe@eurodata.de>
* add persistent option for modprobe
* add suggested changes + fix broken test
* change modprobe module path in tests due to rebase
* change persistent option type from bool to str with choices
* fix unused import
* add example with persistent option
* fix some minor issues after review
- move regexps compiling to __init__
- move AnsibleModule to build_module function and use this function in tests instead of AnsibleModule
- fix terminlogy issue in documentation
* fix unused-import
* Add new project features to API
* add changelog fragment
* remove extra line from changelog
* Update changelog formatting
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: Felix Fontein <felix@fontein.de>
* sefcontext: add path substitution support (#1193)
First commit for feedback, missing docs and tests.
* sefcontext: add documentation
* Add changelog fragment
* Documentation formatting
* Delete extra newline
* pep8 fixes
Fix indentation
* Add version_added to arg docs
* Add examples
* Don't delete non-matching path substitutions
* Add integration tests
* Delete only substitutions if such arg passed
Don't delete existing regular file context mappings if deletion of
a path substitution was requested with the presence of the
`equal` arg - delete only path substitutions in such case.
Path substitutions and regular mappings may overlap.
* Can only add args in minor releases
:(
* Cleanup before tests
* Fix deletion using substitution
Was comparing wrong var.
* Fix test checking wrong var
* Improve args documentation and examples
List the default values for selevel, seuser.
Add example for deleting path substitutions only.
* Add attributes documentation block
Not sure if should add become/delegate/async,
shouldn't those work just like that without any
specific code added for them?
* and fix indentation on attribute block
* Consistent indentation for attributes
Confusing, most plugins indent with 4 spaces.
But some use 2 like the rest of the code, so use 2.
* Add missing ref for attribute block
* Use correct c.g version in doc block
Co-authored-by: Felix Fontein <felix@fontein.de>
* Add full stop to changelog fragment
Co-authored-by: Felix Fontein <felix@fontein.de>
* Streamline documentation
Co-authored-by: Alexei Znamensky <103110+russoz@users.noreply.github.com>
* Support limiting deletion to setype
Deleting file context mappings may be limited by
passing setype or equal, if neither arg is passed
then delete either setype/equal mappings that match.
* Change arg name, diff mode output fix
Change arg name from equal to substitute.
Print target = subsitute in diff mode same way as
semanage does.
Also put back platform attribute, try to improve
clumsy language in the substitute arg docs.
* Delete even if arg setype not match existing
Test 5 indicates that deletion is supposed to not check that
the arg setype passed when deleting matches the setype
of the mapping to delete.
Delete any mapping that matches target, regardless of
setype arg value.
* Update arg name in tests
* Too eager replacing
Accidentally replaced seobject function names so fix them back
* 4564: Fix invalid setype in doc example
Change from httpd_git_rw_content_t which
does not exist to httpd_sys_rw_content_t
Fixes#4564
* Fix documentation attributes
Additional fragment
Co-authored-by: Felix Fontein <felix@fontein.de>
* Update version_added in docs
Bumping minor to 6.4.0 since it didn't make 6.3.0.
* Add more description to the new arg docs
Try to improve discoverability of the new feature and make it easier to understand without deep SELinux understanding.
* Update platform to Linux in documentation
* Add equal as alias for the new argument
Improve discoverability of the new feature by adding an alias to the new module argument. The argument name "equal" will be easy to find for users who are not familiar with SELinux and who just try to match to the CLI tool `semanage`.
* And add alias argument properly
Previous commit missed actually adding the alias (added to docs only).
---------
Co-authored-by: Felix Fontein <felix@fontein.de>
Co-authored-by: Alexei Znamensky <103110+russoz@users.noreply.github.com>
* cloudflare_dns: Fix setting SRV records with a root level entry
* cloudflare_dns: Remove the part which deletes the zone from the SRV record name
The cloudflare API accepts the record name + zone name to be sent. Removing that, will guarantee the module to be idempotent even though that line was added ~7 years ago for that specific reason: 7477fe5141
It seems the most logical explanition is that Cloudflare changed their API response somewhere over the last 7 years.
* cloudflare_dns: Update the changelog fragment
* feat(module/keycloak_group): add support for ...
... handling subgroups
* added changelog fragment and fixing sanity ...
... test issues
* more sanity fixes
* fix missing version and review issues
* added missing licence header
* fix docu
* fix line beeing too long
* replaced suboptimal string type prefixing ...
... with better subdict based approach
* fix sanity issues
* more sanity fixing
* fixed more review issues
* fix argument list too long
* why is it failing? something wrong with the docu?
* is it this line then?
* undid group attribute removing, it does not ...
... belong into this PR
* fix version_added for parents parameter
---------
Co-authored-by: Mirko Wilhelmi <Mirko.Wilhelmi@sma.de>
* github_webhook: Don't include secret in the config if it's absent
* Add changelogs
* Fix indentation
* Apply suggestion to simplify the check
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: dima1206 <32818228+dima1206@users.noreply.github.com>
Co-authored-by: Felix Fontein <felix@fontein.de>
When setting allow permissions for particular users or groups
there will be circumstances when that user is not known to the
host system.
In that case the output of `zfs allow <pool/dataset>`
looks similar to this:
$ sudo zfs allow tank/test
---- Permissions on tank/test ---------------------------------------
Local+Descendent permissions:
user (unknown: 1002) hold
user zfsuser receive
The fix in this commit removes ' (unknown: '+')' from the output
leaving only the uid/gid.
This allows the current parser to continue even if the uid/gid
is not known.
This situation occurs most often when moving a zpool from one system
to another that may not have the same users/groups. Simply adding
permissions to a user/group and then deleting the user/group
from the system will cause this situation to occur.
* nmcli: Treat order as significant when comparing address lists
Don't sort the old and new values for ipv4.addresses and
ipv6.addresses before comparing them, because order matters in these
parameters: the first address specified is the default source address
for outbound connections.
* Changelog fragment for #6048
* Update changelogs/fragments/6048-nmcli-addres-order.yml
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: Felix Fontein <felix@fontein.de>
* Update osx_defaults documentation examples
* Include stderr in errors from osx_defaults
* Add Changelog Fragment
* Update changelogs/fragments/6011-osx-defaults-errors.yml
Co-authored-by: Felix Fontein <felix@fontein.de>
* Change format of examples
* Update plugins/modules/osx_defaults.py
Co-authored-by: Felix Fontein <felix@fontein.de>
---------
Co-authored-by: Felix Fontein <felix@fontein.de>
* feat(modules/keycloak_user_federation): mapper ...
... provider type should have a default value
* add changelog fragment
---------
Co-authored-by: Mirko Wilhelmi <Mirko.Wilhelmi@sma.de>
* fix(zypper): Added condition to check for transactional-update binary to support microos
closes#5615
* style(changelog): Made zypper-change uppercase
Co-authored-by: Felix Fontein <felix@fontein.de>
* fix(zypper): Removed check for /var/lib/misc/transactional-update.state
* feat(zypper): Aligned transactional-update checks with zypper's
* refactor(zypper): Removed dependency to psutil and made use of parsing /proc/mount
* refactor(zypper): Removed need for regex, plus small refactoring
---------
Co-authored-by: André Dörscheln <ad@itesign.de>
Co-authored-by: Felix Fontein <felix@fontein.de>
* nmcli: Convert current value of wifi.wake-on-wlan before comparing
The new value of wifi.wake-on-wlan is specified as an integer, but in
the nmcli output it's specified as a hex string followed by a textual
description of it. Therefore, to determine properly whether it's being
changed we need to pull the hex string out of the current value,
convert it into an integer, and finally convert the integer back to a
string so that we can compare it to the new specified value. Without
this change, whenever wifi.wake-on-wlan is specified in the module
arguments the module will think the value is being changed even when
it isn't.
* nmcli: Handle wifi options correctly when connection type not specified
When an nmcli task does not specify the connection type and the module
ask nmcli for it, the module needs to convert nmcli's
`802-11-wireless` to `wifi`, the term for this connection type used by
the module.
* nmcli: Correctly detect values changed to the integer 0
If the user specifies a value of 0 (without quotes) in a task, we
should interpret that as an actual value, not empty, when comparing
the new value to the old one. Otherwise we incorrectly conclude that
there was no change.
* Changelog fragment for #5431
* Add attributes to more modules.
* Adjust indentation.
Co-authored-by: Kristian Heljas <11139388+kristianheljas@users.noreply.github.com>
---------
Co-authored-by: Kristian Heljas <11139388+kristianheljas@users.noreply.github.com>
* Add attributes to more modules.
* Apply suggestions from code review.
Co-authored-by: Kristian Heljas <11139388+kristianheljas@users.noreply.github.com>
---------
Co-authored-by: Kristian Heljas <11139388+kristianheljas@users.noreply.github.com>
