On Mon, Sep 15, 2025 at 02:06:03PM +0200, Matthias van de Meent wrote:
> Presumably this was `PgAioHandle io_handles[]` at some point, but now
> that it's a pointer it's a proper part of the struct's own size, and
> should be treated as such for memory accounting.
I would bet on a FLEXIBLE_ARRAY_MEMBER from a previous version..
>> Please find a patch to account for io_handles member of PgAioCtl.
>> The patch uses sizeof(PgAioCtl) instead of offsetof(PgAioCtl, io_handles) in AioCtlShmemSize.
>
> LGTM.
Yep, will fix. Thanks for the report, the patch and the review, to
both of you.
--
Michael