Hello. At Thu, 14 Jun 2018 16:06:43 -0400, Alvaro Herrera <alvhe...@2ndquadrant.com> wrote in <20180614200643.3my362zmfiwitrni@alvherre.pgsql> > Can somebody (Arseny, Konstantin, horiguti, Sawada) please confirm that > Michaƫl's commit fixes the reported bug?
pg_advance_replication_slots can advance uninitialized physical slots and that might not be good. (Logical slots always have initial invalid values in thw two lsn columns.) About scanning from restart_lsn in the advancing function, I think I confirmed that the value always comes from XLogRecordBuffer.origptr, which comes from ReadRecPtr, not EndRecPtr, which cannot be on page boundary. FWIW, as the result, it looks fine for me also regarding the issue on this thread. regards. -- Kyotaro Horiguchi NTT Open Source Software Center