This works fine with the current compiler. (Probably since it forced
everything to soft float for the quad-long conversions).

Make sure you have a compiler from the last month or so installed (new kernel
shouldn't hurt as well as Andrey points out but with the current compiler
nothing should be using fpu emulation anymore)

James

>
>On Sun, Apr 28, 2002 at 10:22:16AM +1000, matthew green wrote:
>> 
>> 
>>    Hi, while trying to get Perl-5.8.0-to-be running on NetBSD sparc64,
>>    we ran into this oddity (this is a severely cut down version of a
>>    test run by Perl's Configure script):
>> 
>>    [ .. ]
>>    % uname -a
>>    NetBSD rfhs8036 1.5ZA NetBSD 1.5ZA (GENERIC) #0: Sun Dec 23 03:12:29 PST 2001     chs@ultra2:/build/obj/build/src/sys/arch/sparc64/compile/GENERIC sparc64
>> 
>> 
>> there have been a whole bunch of double floating point fixes made
>> since this time, if i recall correctly.  i don't have my U5 turned
>> on currently so i can't check if this test is fixed now, but can
>> someone with a modern compiler test.  (jmc?)
>
>That's fpu emulation bug which was fixed in more current -current.
>The test work for me with the same compiler.
>
>So update kernel.
>
>	Andrey
>
>
>
>