Michael Paquier <mich...@paquier.xyz> writes:
> And back on that one, I still like better the solution as of the
> attached which skips any relations with their namespace gone missing
> as 246a6c87's intention was only to allow orphaned temp relations to
> be dropped by autovacuum when a backend slot is connected, but not
> using yet its own temp namespace.
Simply skipping the drop looks like basically the right fix to me.
A tiny nit is that using "get_namespace_name(...) != NULL" as a test for
existence of the namespace seems a bit weird/unreadable. I'd be more
inclined to code that as a SearchSysCacheExists test, at least in the
place where you don't actually need the namespace name.
Also, I notice that isTempNamespaceInUse is already detecting the case
where the namespace doesn't exist or isn't really a temp namespace.
I wonder whether it'd be better to teach that to return an indicator about
the namespace not being what you think it is. That would force us to look
at its other callers to see if any of them have related bugs, which seems
like a good thing to check --- and even if they don't, having to think
about the point in future call sites might forestall new bugs.
regards, tom lane
