Re: pgwin32_safestat weirdness

Magnus Hagander <magnus@hagander.net> writes:
> Tom - was there a reason it now runs in FRONTEND as well, or was that
> an oversight?

I did do that intentionally because I was worried about "frontend"
code maybe expecting stat to work fully.  Like pg_standby for example.

I think the immediate problem is that libpq uses stat() as well,
and depending on your link rules that might mean that safestat
actually has to be bound into libpq.

I would not have a problem with assuming that libpq will never care
about st_size being right, but I'm a lot more nervous about making
that presumption for all FRONTEND code.
        regards, tom lane