[Numpy-discussion] long double woes on win32

[Charles R Harris]

On Sat, Aug 16, 2008 at 5:46 PM, David Cournapeau <cournape at gmail.com>wrote:

> On Sat, Aug 16, 2008 at 5:18 PM, Charles R Harris
> <charlesr.harris at gmail.com> wrote:
> >
>
> > could have expl replaced by exp. But there are likely other problems that
> > will need fixing.
>
> I think this is red-herring. Does it really make sense to force
> configuring long double as double if the C runtime and compiler do
> support long double as a different type than double ? The problem is
> only a concern on windows because of compilers bugs (or more exactly
> compiler/runtime mismatch): on windows, long double are double, and so
> long double functions are given 8 bytes double, which is why it should
> work.
>

Yeah, I think that is right. Although it would be nice if things were set up
so a few simple defines would make everything work. There is really no
reason it couldn't work, just that numpy wasn't designed that way.

Anyway, where does mingw get its math.h include file? I suspect it might be
a problem if it doesn't match the windows version because the testcode might
not correctly detect the existence, or lack thereof, of various functions.

Chuck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20080816/cc36f689/attachment.html>