Re: BUG #15592: Memory overuse with subquery containing unnest() and set operations (11.x regression)

Hi,

On 2019-02-09 09:34:41 -0500, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> >> ... Given that, I think it's ok
> >> to not explicitly shutdown the expr context.
> 
> Uh, what?

Well, you explicitly removed the surrounding reasoning. What's the issue
you see here? The generated expression cannot contain anything that uses
shutdown callbacks (it can only contain *_FETCHSOME, *_VAR,
*_NOT_DISTINCT, *_QUAL expression steps). No user defined expressions
have access to the ExprContext.  I'd written about this a few weeks ago,
nobody responded with a better suggestion.


> > Dmitry, the fix will be included in the next minor release, which is
> > scheduled for next week.
> 
> This doesn't really seem like the kind of patch to push on a release
> weekend.  At this point you can't even be confident of getting a readout
> from every active buildfarm member.

Yea, I'd hoped to push this earlier, but unfortune family issues +
related travel + work travel prevented me from getting to this
earlier. The memory leak is significant, the patch hasn't materially
changed in the two weeks since it has been posted, and there's nothing
operating system / architecture related, which I think makes this
acceptable.

Greetings,

Andres Freund