Re: VACUUM FULL, CLUSTER, and REPACK block on other sessions' temp tables

On 2026-Mar-26, Jim Jones wrote:

>
> On 26/03/2026 12:25, Antonin Houska wrote:
> > Chao Li <li(dot)evan(dot)chao(at)gmail(dot)com> wrote:
> >
> >> I don't think such a TAP test is necessary.
> > +1
>
> I've kept the tests in a separate file so the committer can easily skip
> them if needed.

Thanks for noticing and patching this issue. I have pushed the 0001
patch just now.

I decided against pushing the other patch. Although I would have
preferred to add a test, its cost seems not trivial: there are three
full-database scans in it (one for each command), and that seemed a bit
excessive. (There's also one extra initdb, but I'm not sure that part
is too bad since we've optimized that particular part.)

I also considered backpatching, since the code has been like this
essentially forever (i.e. at least since pg14). However, I don't
remember any complaints about this and I would hate to destabilize
things for people without an excellent reason. Maybe we can reconsider
after this month's minors, if somebody shows up with vehement opinions
about it.

Thanks again,

--
Álvaro Herrera PostgreSQL Developer — https://www.EnterpriseDB.com/
"Crear es tan difícil como ser libre" (Elsa Triolet)