On 27.11.22 02:39, Tom Lane wrote:
I got confused about how we were managing EquivalenceClass pointers in the copy/equal infrastructure, and it took me awhile to remember that the reason it works is that gen_node_support.pl has hard-wired knowledge about that. I think that's something we'd be best off dropping in favor of explicit annotations on affected fields. Hence, I propose the attached. This results in zero change in the generated copy/equal code.
I suppose the question is whether this behavior is something that is a property of the EquivalenceClass type as such or something that is specific to each individual field.
