More instantiation problems under hpux

[Loren James Rittle]

> This description is a perfect transcript of the philosophy behing the
> limit-thingies. 

;-)

> | If you see the same failure mode (thus confirming my theory beyond
> | i386 and one OS),

> I think architectures where register and memory precisions differ can be all
> affected.  In the same theory, SPARCs should not be affected.  Any
> confirmation or infirmation?

OK, I think you found a hole in my theory (at least in terms of
applying as widely as I thought it might).

As of mainline today, I can report that sparc-sun-solaris2.7 is
passing 18_support/numeric_limits.cc but still failing
27_io/istream_extractor_arith.cc.  The exact assertion failure is the
same as FreeBSD and hpux:

assertion "is.fail()" failed: file
"[...]27_io/istream_extractor_arith.cc", line 539

But Dave's report that the assertion is failing for a different reason
than FreeBSD/i386 in conjunction with your statement makes me think
that Solaris might be as well...

I am building a gdb that can debug the situation more fully under
Solaris.

Regards,
Loren