On Thu, Feb 15, 2024 at 10:40 PM Anton Voloshin <a.volos...@postgrespro.ru> wrote: > On 19/01/2024 01:35, Thomas Munro wrote: > > I don't yet have an opinion on the best way to > > do it though. Would it be enough to add emit_message($node, 0) after > > advance_out_of_record_splitting_zone()? > > Yes, indeed that seems to be enough. At least I could not produce any > more "xl_tot_len zero" failures with that addition. > > I like this solution the best.
Oh, it looks like this new build farm animal "skimmer" might be reminding us about this issue: https://buildfarm.postgresql.org/cgi-bin/show_history.pl?nm=skimmer&br=HEAD I don't know why it changed, but since this is an LSN/page alignment thing, it could be due to external things like an OS upgrade adding more locales or something that affects initdb. Will look soon and fix.
