On Mon, Apr 20, 2020 at 07:24:28PM +0900, Fujii Masao wrote:
> I was misreading this as something like "any other blocking than
> the blocking in WalSndLoop()". Ok, I have no more comments on
> the patch.
Patch looks rather sane to me at quick glance. I can see that WAL
senders are now not stuck at 100% CPU per process when sitting idle,
for both physical and logical replication. Thanks.
--
Michael