Re: some thoughts on inf and nan

-----Original Message----- 
From: Father Chrysostomos
Sent: Thursday, August 21, 2014 5:51 AM

> And it would prevent those of us who reach for POSIX::strtod to work> around perl's broken number parsing from doing so, wouldn't it?

Yes - which might then prompt someone to fix perl's broken number parsing 
;-)

AIUI the main point of having perl do it's own number parsing is to avoid 
reliance on such things as POSIX::strtod - and perhaps with good reason.
For some floating point values, POSIX::strtod sets wrong values for perls 
built with the mingw.org ports of gcc. These are the only POSIX::strtod 
failings I've come across, but who is to say that there aren't others.

And this sort of reliance on POSIX::strtod doesn't help at all if nvtype is 
long double. POSIX::strtod cannot be relied upon to provide the correct 
values in that case, and there is no POSIX::strtold.
Because of this, I wrote Math::NV::nv() which calls either strtod or strtold 
(as appropriate) - but Math::NV is not a core module. It also doesn't fix 
the problem with mingw.org compilers - which is the reason I now use 
mingw.64 compilers on Windows.

This thread is perhaps not the right place to raise it, but I think that if 
perl's number parsing is not going to be fixed, we should at least provide a 
POSIX::strtold function.

Cheers,
Rob