build ROCKs in CI only once on every PR/push event

[orfeas-k]

What needs to get done

Refactor CI to build a ROCK once and use it across workflows during each event. This is sth like what we could do, except we would pass .rock files.

Concerns

We may hit issues with storage limits on the artefacts for repos with large images (or even repos with many medium sized images)

Why it needs to get done

Context

At the moment, there are 3 workflows that may build a ROCK.

build_and_publish_rock.yaml
build_and_scan_rock.yaml
build_scan_and_test_rock.yaml

These 3 all build ROCKs differently. At some point, we will have them build-scan-tested on_pull_request and then on_push we will have them build-published or maybe build-scan-tested (again) and build-published.

Issue

The problem with the above is that if each of these workflows builds a ROCK on their own

• we will not make sure that what we test is what we publish (sth like what @ca-scribner described for charms here https://github.com/canonical/bundle-kubeflow/issues/766)
• we will waste a lot of resources since ROCK building can be a very resources-intense process

[orfeas-k]

This is also related to https://github.com/canonical/bundle-kubeflow/issues/763 since this discuss the same piece of code but for tox.ini files.

[syncronize-issues-to-jira[bot]]

Thank you for reporting us your feedback!

The internal ticket has been created: https://warthogs.atlassian.net/browse/KF-5126.

This message was autogenerated