Right now we only buildfetch if we are doing autolock (i.e. basing on mechanical streams). Let's always buildfetch so we can see an RPM diff when the commit/container is built.
Right now there is a bug where `cosa diff` when it's the first build in the stream will fail (i.e. causing `cosa build` to fail), but should be fixed soon. This works around that, but also shows the RPM diff, which is nice too.
We had a case where an RPM build we had fast-tracked finally got submitted as a bodhi update and this test-override job failed because "Input hash unchanged". i.e. it detected there was no changes and didn't do the build, which meant the tests failed.
Let's just force the build here. We know it's not that useful but reporting failure is worse.
We moved the definition of the version/releasever in [1] to build-args.conf so we need to update the logic here to source in from that location instead of the manifest.
seed-github-ci: replace configure block with DSL API
Use the dynamic DSL branchSource{} to configure strategy,traits and build strategies instead of patching raw XML. This also amends traits elements from the previous static github{}.
Based on documentation from: https://jenkins-coreos-ci.apps.ocp.fedoraproject.org/plugin/job-dsl/api-viewer/index.html#path/multibranchPipelineJob-branchSources-branchSource-source-github-traits
In https://github.com/coreos/coreos-assembler/pull/4255 we are proposing to remove followup target build support from `cosa build`; i.e. it won't auto build the `qemu` target when doing a build.
Let's split this part out into a separate step here.
Autodetection of dev builds fails on builds where 'dev' is not part of the filename scheme, like 'rhcos-9.8.20260506-0-live-iso.x86_64.iso'. This causes kola tests to fail when they should run in insecure mode.
The obsolete 'vars/kolaTestIso.groovy' used the insecure flag by default, so this restores that behavior explicitly.
cosaPod: add timeout to prevent pods from running indefinitely
Add a timeout parameter (default 24 hours) to cosaPod to prevent hung or failed pods from blocking the pipeline. Callers can override the default by passing timeout (in hours), or disable it with timeout: 0 for jobs like debug pods that may need extended runtime.
There is a strange issue where probably `shwrapRc("kola help | grep -q testiso")` returns a non-zero exit code despite the `testiso` command being available. As a result, we incorrectly fall back to the newer code path and hit failures like: