Bash scripts fail but GH Actions continue

[kimwnasptd]

Bug Description

Observed it firstly in the following code

bundle-kubeflow/.github/workflows/scan-images.yaml

Line 56 in b226918

IMAGES=$(./kubeflow-ci/scripts/images/get-all-images.sh ${{ matrix.bundle }}/bundle.yaml ${RELEASE}-${RISK})

In the above case the script there failed, because of some misconfiguration, but the GH Job step didn't fail at all. The above is a bug since we can't accurately understand what went wrong and need to further look at all steps to find the one that failed.

To Reproduce

Try and call a bash script, that doesn't set -xe, and one of its sub-commands fail. In this case the overall bash script will just continue.

Environment

Any GH action

Relevant Log Output

.

Additional Context

To avoid this, we should:

1. Use python everywhere possible instead of bash scripts
2. Set set -xe in the bash scripts, to ensure they exit on failure

[syncronize-issues-to-jira]

Thank you for reporting us your feedback!

The internal ticket has been created: https://warthogs.atlassian.net/browse/KF-6139.

This message was autogenerated

[kimwnasptd]

After merging canonical/kubeflow-ci#142 the scanning actions are indeed erroring out in the correct step
https://github.com/canonical/bundle-kubeflow/actions/runs/10523106840/job/29157141080

> Get images
...
++ bash ./tools/get-images.sh
Error: open ./net-istio.yaml: no such file or directory
rm: cannot remove 'net-istio.yaml': No such file or directory

Will go on and close this issue and scope it for the scanning scripts, but the same should apply for all bash scripts