>> Yeah, or something we didn't anticipate, like Linux swapping the RAM out
>> when we're still using it.
>
> OK, my question is whether it deserves special mention as a testing
> area.
Well, I'll admit to having an ulterior motive here. The SHMMAX/SHMALL
thing has been a constant irritation for both sysadmins and developers
installing on their laptops (especially Mac laptops) for so long that I
want to give it special mention, but it's not really a "major feature".
This was the compromise we worked out on -advocacy.
Also, we don't *know* that the switch to mmap() won't introduce new
memory issues we haven't thought of, and if it does, they'll be on
offbeat platforms/configurations we don't test. So I do think there's a
call for extra testing around this area.
BTW, please remember to credit Agent M along with Robert Haas for that
feature.
--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com