Re: 18/numeric_limits fail on solaris 2.8

[Loren James Rittle <rittle at latour dot rsch dot comm dot mot dot com>]

In article <3CA969A6.4070807@unitus.it>,
Paolo Carlini <pcarlini@unitus.it> writes:

>> Closely reading the above comment (which I wrote) makes me think that
>> it should be...  I am almost sure that when I posted the patch that
>> added that specialization, I begged a solaris person to enable it
>> after verifying the situation I found.  I am sorry if I dropped the
>> ball on a port I barely use

> Please don't!

Please don't what?  ;-)  I didn't enable the slack FP path to relax the
numeric test related to long double on Solaris for exactly this reason
(that I don't fully understand their FP model).  I will not ever enable it.

> Really, there is a big mistery here! Most probably not strictly a 
> library issue, however.
> Have you tried Benjamin's "C" testcase?

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

OK, I just tried it on the latest gcc 3.1 prerelease I have installed
for sparc-sun-solaris2.7 (3.1 20020125).

> What do you think about it?

Sorry, I do not see the behavior that was reported.  The program runs
to completion without asserting.  When run under gdb, I see this after
stepping to the end of main():

(gdb) print ldmin
$2 = 3.3621031431120935062626778173217526e-4932
(gdb) print ldmax
$3 = 1.189731495357231765085759326628007e+4932
(gdb) p min
$4 = 3.3621031431120935062626778173217526e-4932
(gdb) p max
$5 = 1.189731495357231765085759326628007e+4932