"smart" shutdown is broken in HEAD

Tom Lane wrote:

"pg_ctl stop" seems to be causing a fast stop today. Anybody remember
touching logic that might affect that?

What's the evidence for that? I can't see any on the buildfarm.

cheers

andrew

Andrew Dunstan <andrew@dunslane.net> writes:

Tom Lane wrote:

"pg_ctl stop" seems to be causing a fast stop today.

What's the evidence for that? I can't see any on the buildfarm.

I don't think the buildfarm does anything that would prove it one way or
the other. But I find that if I start the postmaster, start a session,
issue "pg_ctl stop" (with or without "-m smart"), the stop completes
immediately and I'm left with a dead psql:

regression=# select 1;
FATAL: terminating connection due to administrator command
server closed the connection unexpectedly
This probably means the server terminated abnormally
before or while processing the request.
The connection to the server was lost. Attempting reset: Failed.

Once the postmaster has been up for a little bit this seems not to happen
anymore, which means it could have been that way for awhile without
anyone noticing. I'm not sure yet how long is "a little bit" --- seems
to be more than a minute, which destroys my first theory that the first
autovac launch was involved...

regards, tom lane

I wrote:

Once the postmaster has been up for a little bit this seems not to happen
anymore, which means it could have been that way for awhile without
anyone noticing. I'm not sure yet how long is "a little bit" --- seems
to be more than a minute, which destroys my first theory that the first
autovac launch was involved...

Ah, found it ... not timing sensitive, but uninitialized-memory
sensitive :-(

regards, tom lane