Andres, all,
* Andres Freund (andres@anarazel.de) wrote:
> On 2017-12-06 21:38:42 +0100, Tomas Vondra wrote:
> > FWIW I've constructed the data sets for two reasons - to convince myself
> > that my understanding of the simplehash code is correct, and to provide
> > a data set triggering the other growth condition in simplehash code. My
> > understanding is that if we stop growing the table after the load factor
> > drops below some threshold (as TL proposed earlier in this thread), it
> > should address both of these cases.
>
> Yea, I'm not adverse to adding a few stopgaps that break in a less
> annoying manner. WAll I'm saying is that I don't think we need to be
> super concerned about this specific way of breaking things.
I think I'm agreeing with Tomas here when I say that we shouldn't have a
system that breaks in the "tries to allocate infinite memory" way when
the load factor drops far below some reasonable threshold and that we
*should* be concerned that we have this problem in the existing
implementation. It's bad, and we should care about that and fix it.
Running forever in a loop isn't great either, but at least you only
expend at most one CPU to it and can go kill it, assuming we have an
appropriate CHECK_FOR_INTERRUPTS() placement. Running the box out of
memory can lead to other things *failing* (even entirely reasonable
things!), not just being slower. That's an important difference.
Thanks!
Stephen