I think this condition is inaccurate.
we should be building the following description when there is a base type in a discriminated set.
the fact whether this model is abstract or not is implementation details.
in downstream emitters, they might not still model discriminated base type "abstract" (such as the provisioning emitter)
Originally posted by @ArcturusZhang in #10988
Originally posted by @ArcturusZhang in #10988