On Sun, Oct 06, 2013 at 09:48:04AM +0200, Fabien COELHO wrote:
> Also, I'm not sure of any system used with pgbench that do not have
> threads, so ISTM that the thread fork-emulation hack is more or less
> useless, and it is pretty masochistic to maintain when adding
> features.
Fair point. When added, the code pertaining to fork-emulation was
well-isolated, and that remained the case as recently as 9.3. Your recent
--progress patch was the first to suffer contortions for the benefit of that
mode. (The per-statement latencies feature just declined to support it.)
> >For the time being, I propose the attached comment patch.
>
> It comment seems very clear to me. I do not understand why it starts
> with XXX, though.
PostgreSQL code uses that notation regularly. When I add it, I typically have
in mind "the following is not fully satisfying, but it's not bad enough to
make a point of improving".
I've committed the comment patch.
--
Noah Misch
EnterpriseDB http://www.enterprisedb.com