From interaction with openshift devs using 4.x vs. 3.x based CI
per agreement with Nalin, opening a bug to track
See slack thread https://coreos.slack.com/archives/C014MHHKUSF/p1591216952054700 for the triage / diagnosis
Realized I did the wrong product/component ... fixing now.
Moving to the Buildah component as assigning to Nalin.
I'm able to reproduce this, but only in OpenShift, as the buildah CLI seems to be getting the right result. The incorrect value could be coming from ContainerConfig, or it could be the hardwired default that we use when no value is set, since in the error case they appear to have the same value.
I no longer believe that ContainerConfig/Config confusion is the root cause. https://github.com/openshift/imagebuilder/pull/169 should fix the problem once it's merged and we integrate a version with the change.
We've encountered a regression in the upstream fix (https://github.com/openshift/imagebuilder/issues/173) which I thought would resolve this, so we need to do some more work there.
Regression in imagebuilder should be fixed as of https://github.com/openshift/imagebuilder/pull/174, so we need to tag a release there and then we can pull it in to the builder.
The release in ImageBuilder was just recently created and needs to now be vendored through Buildah and Podman. That won't happen today, so I'm moving to Upcoming Sprint.