The previous version of the sync script ran `helm dep update` which
would recreate the requirements.lock file. This caused new builds
of charts as a total package with different versions of dependencies
but the same chart version. The package was mutating.
This change works towards our goal of immutable charts for a chart
at a version.
Indentation checking was catching catching out the commenting of
whole sections of indented code in values.yaml files as being
errors. Yet, code editors and other tools automatically indented
to this level. Removing the check.
This commit does a couple things
1. It reformats the layout to have checks via self contained
functions rather than all in one loop.
2 Adds a check that the chart.version was incremented
Ref #2373
The changed.sh script tried to find the merge base but the merge
base branch was not available. This change provides it so that
a common ancestor can be found.
This does not lint yaml templates as they are templates rather
than valid yaml files. They cannot be linted with a normal linter
yamllint is used for linting. This is an existing Python project
https://github.com/adrienverge/yamllint
The rules are not the default rules and are stored in their
entirity so they can be controlled over time
The existance of a Chart.yaml file and values.yaml file is checked
and an error is thrown if one is missing. Helm lint will not
detect a chart if Chart.yaml is missing and if a values.yaml file
is missing it is noted as info.
The run function is introduced to enable running all the linters,
capturing non-zero exit codes, and exiting with a non-zdero code
if any of them fail. This is used instead of exiting when the
first failure happens to provide more feedback to chart developers.
The previous diff was a comparison against master. If master had
changed since the current PR was created the changed charts would
be re-tested. This could create a problem when helm was upgraded
and lint changed causing previously passing charts to fail. Since
they were not associated with the current PR it was an outlier
problem for the PR author.
The new diff does a comparison to the merge-base which ties to look
for changes introduced by the current pull request.
This change does the following:
- Adds a circleci config file
- Has 2 scripts. One for installing tools and a separate one for
executing tests
- Removes lint testing from the changed.sh script since it is
performed elsewhere
- For linting changes the git diff mechanism to look at the merge
base rather than all differences from master. The intent is to
look for changes in this request rather than including those
in PRs that may have already been merged to master while this
one lingered.
Ref #2337
* Fix pvc metadata
The `chart` and `name` keys are in the wrong place.
* Fix indentation for persistent volume
* [stable/kapacitor] bump chart version to 0.2.2
* fix test flake
* Add basic verification of release
* Set BUILD_NUMBER in pr-review.sh
* Set default verification pause if not set
* Work around default pause value with -u
* Fix default pause value in pr-review
* Use kubconfig for tests if possible
* Add basic verification of release
* Add basic verification of release
* Set BUILD_NUMBER in pr-review.sh
* Set default verification pause if not set
* Work around default pause value with -u
* Fix default pause value in pr-review
* Use kubconfig for tests if possible
* Add basic verification of release
* Ensure kubeconfig works standalone
Matt Farina
Josh Dolitsky
Vic Iglesias
Anirudh Ramanathan
Adnan Abdulhussein
Michael Goodness
Dhilip
Anirudh Ramanathan
James Gregory
Adnan Abdulhussein