Re: pg_upgrade: optimize replication slot caught-up check

On Tue, Jan 27, 2026 at 3:59 AM Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>
> On Fri, Jan 23, 2026 at 2:04 AM Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com> wrote:
> >
>
> I haven't reviewed v7 in detail but while glancing, I noticed a few
> minor comments:
>
> 1.
> + * Returns the last LSN decodable WAL record's LSN if found, otherwise
> + * returns InvalidXLogRecPtr.
> */
> -bool
> -LogicalReplicationSlotHasPendingWal(XLogRecPtr end_of_wal)
> +XLogRecPtr
> +LogicalReplicationSlotHasPendingWal(XLogRecPtr end_of_wal,
> + XLogRecPtr scan_cutoff_lsn)
>
> The function name suggests that it will return boolean (due to 'Has'
> in its name) but after this change that is not true.
>
> 2.
> We
> + * also use the maximum confirmed_flush_lsn as an early scan
> + * cutoff; finding a decodable WAL record beyond this point
> + * implies that no slot has caught up.
> + *
>
> In this comment, it is not clear if the maximum confirmed_flush_lsn is
> among all logical slots (of current database) or what?
>

Thank you for the comments!

I've incorporated these comments I got so far.

Amit, are you planning to review the patch in detail? If so, I want to
incorporate comments before the push.

Regards,

--
Masahiko Sawada
Amazon Web Services: https://aws.amazon.com