On Thu, Nov 12, 2026 at 4:10 AM Masahiko Sawada <[email protected]> wrote: > > On Mon, Nov 9, 2026 at 5:21 PM Masahiko Sawada <[email protected]> wrote: > > > > I've attached the new patches. I just updated the commit message for > > the 0001 patch, and added some fixes and regression tests for the 0002 > > patch. Please review them. > > I think the 0001 patch can be pushed independently. So I'm going to > push it to master, barring objections. >
Few comments on 0001: * As the parameter to stream_message callback could be a sub-transaction which is different from other similar callbacks like truncate and change, it would be better to add that in docs where we describe LogicalDecodeStreamMessageCB. The one more thing which we may consider is how to get the top txn if required by callback as the patch did in test_decoding. * As we are planning this as a master-only change, will it be better if we just pass subxid as a separate parameter to keep the interface consistent with other similar interfaces and avoid the need for plugins to retrieve toptxn separately. * Similar to pg_decode_stream_message, shouldn't there be a call to stream_start if (!txndata->stream_wrote_changes). This is just a code observation, I haven't tested it. -- With Regards, Amit Kapila.
