Tom Lane wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
> > On Tue, May 8, 2018 at 4:48 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> >> Really the only thing here that jumps out as being unduly expensive for
> >> what it's doing is select_default_timezone. That is, and always has been,
> >> a brute-force algorithm; I wonder if there's a way to do better?
>
> > Who says we need a portable way? If we had something that worked on
> > Linux and macOS, it would cover most developer environments. I wonder
> > if readlink("/etc/localtime", buf, sz) might be a viable approach.
>
> I wondered about that, but I'm afraid it's often a hardlink not a
> symlink. Still, we could try it.
In Debian systems, it's a symlink. Apparently in RHEL6 and older it's a
copy or hardlink, and the file /etc/sysconfig/clock contains a ZONE
variable that points to the right zone. Maybe if we add enough
platform-dependent hacks, we would use the slow fallback only for rare
cases. (Maybe have initdb emit a warning when the fallback is used, so
that we know what else to look for.)
This comment is insightful:
https://github.com/HowardHinnant/date/issues/269#issuecomment-353792132
It's talking about this code:
https://github.com/HowardHinnant/date/blob/master/src/tz.cpp#L3652
--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services