When we enable GPG Verification for pull request reviews, we should
make sure there is information available in our Contributing docs
about how to sign commits.
This allows for the runner to clear the git lock if any of the
per-remote-only params was present in the configuration.
This also adds PER_REMOTE_ONLY attributes to the winrepo runner and
git_pillar external pillar, and fixes all refs to GitBase's init_remotes
function to include PER_REMOTE_ONLY arguments. This will insulate us
against future headaches if either git_pillar or winrepo gain extra
per-remote-only params, as we will only need to make the change to the
attribute in their respective modules.
Fixes#43650
The merge-forward process needs a more prominent position and explanation
in the contributing documentation. This change attempts to explain this
process a little more fully and incorporates some changes to how we are
handling "main" release branches, "dot" release branches, and "develop"
in a more complete context with merge-forwards.
Taken the develop version of win_pkg.py and
- import sys to get INT max size for compare_version()
- enhanced compare_version to support versions of 'Not Found' and 'latest'
which can occur in windows.
- comment out import salt.utils.platform which is not in 2016.11
-- salt.utils.platform.is_windows() renamed back to salt.utils.is_windows()
- added back the warning about win_repo_source_dir removal as this is
required for 2016.11 version.
Added on top of the backport is
- Additional comments into the code
-- Indicate pkg_list can return a 'version number' and also 'Not Found'
version number
-- _get_latest_pkg_version() can return 'version number' or 'latest'
-- latest_version() can return blank if the latest version is installed or no
version at all is installed
- upgrade_available did not require to version() and latest_version() as
latest_version() does everything required.
- list_upgrades() added comments to explain code code, simply code
to latest_version() rather than calling latest_version()
and compare_versions()
- list_available() is not required to force a refresh by default
- version() added comments and questions as the return value
is non standard
- genrepo() add questions around the open mode and
get_repo_data open mode
- _repo_process_pkg_sls()
-- file param was renamed to filename as file variable name classes
with file()
-- rename var version to version_str so as not to import
version() function
-- remove a section of code which was never executed and never used
- install()
-- simplied an if statements
-- add more code comments
Because requisite resolution doesn't happen until we run call_high, and
state.sls_id doesn't run call_high (but rather calls a single low chunk
from the compiled low chunks), the use/use_in requisites are ignored.
This fixes that oversight by resolving requisites in state.sls_id before
we compile the high data.
If a VPC or Subnet ID is not found when calling functions that are
supposed to be checking for vpc/subnet ID existence, the return
should be consistent by returning booleans/None instead of returning
the NotFound error from AWS.
The surrounding code blocks indicate that this is expected as well
as unit test assertions. The moto library had a bug in it where
it wasn't raising "x.NotFound" errors when it should have been. The
latest version of moto has fixed this bug, causing our tests to fail
since the boto_vpc module is not handling the "x.NotFound" errors
separately from the generic BotoServerErrors.
This fixes the test failures in the branch tests that were caused by
upgrading the moto version to the latest release.
* Removed redundant import.
* No longer inheriting from LogSettingsParserTests.
* Replaced test class description with somethin meaninful.
* Fixed identation.
I've also moved the class to the bottom, since all the classes inheriting
from LogSettingsParserTests are in the block above.
This is a follow up on this PR:
https://github.com/saltstack/salt/pull/43366
Since we can get an OSError durin PIDfile deletion with non-root users, it
would make sense to also test for this. So here are the two test cases.
One with an OSError and the other one without.
