Re: Broken HOT chains in system catalogs

"Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov> writes:
> Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> This amounts to assuming that no new indexes get added to system
>> catalogs after initdb, or at least not during concurrent
>> operations wherein indcheckxmin would be important.

> Sounds reasonable, but can we enforce it through locks rather than
> assuming, or isn't there a clear way to do that without risking
> deadlocks?

Well, we already enforce it through the allow_system_table_mods mechanism:

regression=# create index foo on pg_index(indcheckxmin);
ERROR: permission denied: "pg_index" is a system catalog

Yes, you can turn that off if you try hard enough, but it's on your own
head to know the consequences.

regards, tom lane