On Tue, Apr 28, 2026 at 02:13:31PM +0900, Michael Paquier wrote:
> In short, it took me some time to put some order into all that,
> finishing with the attached patch set. 0001 is the minimum for v19,
> that reverts 6dcfac9696cb so as we have more *GetDatum() matching with
> the types in the SQL functions. That would take care of the open item
> on top of my head.
This was not completely right after a second look. 6dcfac9696cb did
not get things completely wrong, either, the brin and gist parts of
the changes were right. I have undone the incorrect bits for now to
address the open item.
> 0002 is a set of fixes that I have spotted while investigating this
> set of issues in depth. These spots are actually wrong, some of them
> for a long time. I would be slightly tempted to do something about
> these in v19 rather than wait for v20, as these are somewhat latent
> bugs, to have more consistency across the board. Has anybody from the
> RMT an opinion to offer? There is not much urgency in it, still..
> Added the RMT in CC for opinions.
This deserves a different discussion, unrelated to DatumGetInt8().
I'll post that on a separate thread, for v20.
--
Michael
