Re: Issue in postgres_fdw causing unnecessary wait for cancel request reply

On Thu, Apr 13, 2023 at 2:04 PM Fujii Masao <masao(dot)fujii(at)oss(dot)nttdata(dot)com> wrote:
> >> To clarify, are you suggesting that PQgetCancel() should
> >> only parse the parameters for TCP connections
> >> if cancel->raddr.addr.ss_family != AF_UNIX?
> >> If so, I think that's a good idea.
> >
> > You're right. I used connip in the diff because I thought it provided
> > the same condition, but in a simpler way.
>
> I made a modification to the 0001 patch. It will now allow PQgetCancel() to parse and interpret TCP connection parameters only when the connection is not made through a Unix-domain socket.

I don't really like this change. It seems to me that what this does is
decide that it's not an error to set tcp_user_timeout='a' when making
a cancel request if the connection doesn't actually use TCP. I agree
that we shouldn't try to *use* the values if they don't apply, but I'm
not sure it's a good idea to skip *sanity-checking* them when they
don't apply. For instance you can't set work_mem=ssdgjsjdg in
postgresql.conf even if you never run a query that needs work_mem.

--
Robert Haas
EDB: http://www.enterprisedb.com