18/numeric_limits fail on solaris 2.8

[Paolo Carlini]

Benjamin Kosnik wrote:

>>Really. And what is wrong has /nothing/ to do with libstdc++-v3 !
>>This is with Solaris8, and the head compiler?
>>
>Yes.
>
I see. And can't be that gcc on solaris must be configured in some way 
in order to actually have 128 bit long doubles, since you have checked 
that they are indeed 16 bytes long...

Any Solaris-Sparc expert out there? (*)

By the way, thank you for teaching me the "C trick": before working for 
v3 it never occurred to me that writing a corresponding simplified "C" 
testcase may help so much in diagnosing a library problem...

Ciao, P.

(*) I'm CC-ing Jakub, which perhaps may be interested in your testcase. 
In case, Jakub, you can find it here:

    http://gcc.gnu.org/ml/libstdc++/2002-04/msg00018.html