Re: literal 1000. of type double is out of range error on ppc32 soft-float

[Andreas Tobler <toa at pop dot agri dot ch>]

David Daney wrote:
> Andreas Tobler wrote:
> > Martin Michlmayr wrote:
> > > * Tom Tromey <tromey@redhat.com> [2007-10-25 15:14]:
> > > > I don't have time but we could help you track it down if you want to
> > > > try debugging.
> > >
> > > Andrew Haley tracked it down to the lack of libffi support for powerpc
> > > soft-float.  Michael Koch thought that Andreas Tobler (CC:ed) might have
> > > an idea or even some patches.
> >
> > Do you have a real use case for this? I can try to add soft float 
> > support to libffi ppc (32-bit) But I need a motivating factor ;)
>
> How about: It would be fun to implement this?

Sure, no problem and a good motivating factor. I did it for darwin 
ppc-32, x86_64 and hpux-32, just for fun.
But I do not feel very happy if people do the work twice in the hidden 
place. See OS-X 10.5, they have a libffi implementation, FAT, means for 
ppc-32/64 and i686/x86_64 w/o telling the community about. I tried to 
implement a ppc-64 variant for darwin as well. But when I saw the 
libffi.dylib in /usr/lib/ of leopard I got really up-set! I asked in 
advance if people are working on...
 .....

> Not that it is the canonical libffi implementation, but the MIPS port 
> supports both hard and soft float ABIs.  Feel free use it as a model for 
> the PPC soft-float implementation.

I'll have a look at.

Thanks David!

Andreas