"Ed L." <pgsql@bluepolka.net> writes:
> One idea for handling the round-to-localtime issue from the other end of the
> problem: optionally rotate logs upon an *interpolated* filename change.
The code-as-committed wants to be able to predict the time change in
advance. I'm not totally wedded to that, but I don't want to do a
strftime every time through the main loop, either ...
> I also noticed pg_tm.tm_gmtoff is apparently not set, at least not for my
> local (US MT).
It works for me. You want to look into why not for you?
regards, tom lane