Re: BUG #16125: Crash of PostgreSQL's wal sender during logicalreplication

On Wed, Nov 27, 2019 at 11:57:00AM -0300, Alvaro Herrera wrote:
>On 2019-Nov-27, Alvaro Herrera wrote:
>
>> I think pg_waldump should print out the speculativeness of an insertion.
>
>Ah, in pg12 we do print the "flags" for the insertion.  But earlier
>releases don't.  IMO we should have backpatched that change, as I said
>in https://postgr.es/m/20180423160216.m2yjhx67grofg7h5@alvherre.pgsql
>but I was outvoted.
>

Hmmmm, yeah. I still think the decision not to backpatch was reasonable,
although it seems it might have been useful in this particular case.

But hey, it should not be that difficult to build a custom pg_waldump
version with that patch applied, and decode the WAL segment. It can even
be done on a completely separate system, it's just a single binary.

regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services