Re: array_agg(anyarray) silently produces corrupt results with parallel workers when inputs mix NULL and non-NULL array elements

On Sat, Apr 4, 2026 at 5:41 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> Dmytro Astapov <dastapov(at)gmail(dot)com> writes:
>

> The initial-setup path is confused about that too, allocating
> newnitems+1 which is pointless.
>

Yes, using newnitems directly is cleaner than my Max() approach, thank you.

>
> It also troubled me that there's no checks for integer overflow
> when calculating the new sizes.

Good catch, I hadn't considered that.

> So I end with the attached revised patch, where I also made one
> or two cosmetic adjustments like putting the type-comparison checks
> next to the dimension comparisons. Look good to you?
>

Looks good, thank you for improving it.

I built and tested your v2 patch on REL_17_9 and REL_18_3 (with minor
adaptation for the slightly different context lines, like bits8 vs uint8 on
17.x), using the same 10M-row synthetic reproduction from my original
report. They both pass (as expected).

I am attaching the amended patch files for REL_17_9 and REL_18_3 just in
case.

Thank you for the feedback and for the quick turnaround on this!

Best regards, Dmytro