Hi,
On 2021-04-08 01:16:02 -0400, Tom Lane wrote:
> Michael Paquier <michael@paquier.xyz> writes:
> > On Wed, Apr 07, 2021 at 04:22:17PM -0400, Tom Lane wrote:
> >> Buildfarm suggests this has some issues under force_parallel_mode.
> >> I'm wondering about missed fields in outfuncs/readfuncs, or the like.
>
> > The problem looks a bit more fundamental to me, as there seems to be
> > some confusion with the concept of what should be the query string
> > when it comes to prosqlbody with a parallel run, as it replaces prosrc
> > in some cases where the function uses SQL as language. If the
> > buildfarm cannot be put back to green, could it be possible to revert
> > this patch?
>
> Andres pushed a stopgap fix.
Let's hope that it does fix it on the BF as well. One holdup was that
check-world didn't succeed with force_parallel_mode=regress even after
the fix - but that turned out to be the fault of
commit 5fd9dfa5f50e4906c35133a414ebec5b6d518493 (HEAD)
Author: Bruce Momjian <bruce@momjian.us>
Date: 2021-04-07 13:06:47 -0400
Move pg_stat_statements query jumbling to core.
et al.
> We might end up reverting the patch altogether for v14, but I don't
> want to be hasty. This should be enough to let people take advantage
> of the last few hours before feature freeze.
Yea, I think it'd be good to make that decision after a decent night of
sleep or two. And an actual look at the issues the patch might (or might
not) have.
Independent of this patch, it might be a good idea to have
ExecInitParallelPlan() be robust against NULL querystrings. Places like
executor_errposition() are certainly trying to be...
Greetings,
Andres Freund