New FreeBSD failures (somewhat) debugged

[Benjamin Kosnik]

> You're right in that right now <limits> is trying to match what the
> compiler says limits are, not what the host "C" library says they
> are. After Loren's analysis, I agree that we should match the host "C"
> library views.  Does everybody agree?

Sure. You're the boss.

;)

This appears to be a non-optimal solution, however: why can't the "C" 
library headers and the compiler (gcc/g++) agree? This mis-match seems odd 
to me and a potential source of bugs later.

I confess I'm busy and I don't really understand the ins and outs of this 
particular issue though. I long for the day of the old limits 
script...

Let's at least try this, and hopefully the numerics failure will also go 
away for BSD. (In any case, let's try to resolve that first, before the 
27_io/*.cc FAILS.)

thanks,
benjamin