New FreeBSD failures (somewhat) debugged

[Loren James Rittle]

> [...] I think the cases are failing on various architectures due to
> latent issues (I dare not call them bugs since I think they exist
> somewhat by design).

Given the quick followup and many open questions, I feel I better post
that I am away from e-mail for two weeks starting right now. ;-)

Other than knowing about the mismatch between expressed limits and
real hardware limits on various platforms (I remember we last
discussed this issue back when 18_support/numeric_limits.cc started
failing on FreeBSD); I can offer no other input.

To better ensure that timely reports are posted from ref5 (only
FreeBSD machine automatically bootstrapping gcc mainline daily), I
have forced it to use GNU make exclusively for the time being.

Regards,
Loren