Re: CREATE SUBSCRIPTION ... SERVER vs. pg_dump, etc.

On Sun, 2026-08-09 at 22:15 -0400, Tom Lane wrote:
> 874      if (!OidIsValid(serverid))
> > > >     CID 1699896:         Null pointer dereferences 
> > > > (FORWARD_NULL)
> > > >     Passing null pointer "conninfo" to "cstring_to_text", which
> > > > dereferences it.
> 875      values[Anum_pg_subscription_subconninfo - 1]
> =
> 876      CStringGetTextDatum(conninfo);
>
> AFAICS, it's right: if stmt->servername is set while opts.connect is
> not, we'll arrive at this step with serverid filled in but conninfo
> still NULL.  Even if there's some upstream reason why that
> combination
> can't occur, this is pretty fragile-looking code.

If serverid is filled, that branch won't be taken.

The code relies on the grammar setting either stmt->servername or stmt-
>conninfo, but not both. I agree that the control flow shouldn't rely
on that, and the code could be more clear anyway. Patch attached.

Regards,
Jeff Davis