Re: isnan/isinf problems still exist

[Han-Wen Nienhuys]

Graham Percival wrote:
> On 19-Sep-05, at 3:34 AM, Han-Wen Nienhuys wrote:
>
> > Graham Percival wrote:
> >
> > > On 16-Sep-05, at 8:50 AM, Johannes Schindelin wrote:
> > >
> > > > My workaround for now is to append this after "#include <cmath>":
> > > >
> > > > #ifndef isinf
> > > > #define isinf(x) __isinfd(x)
> > > > #define isnan(x) __isnand(x)
> > > > #endif
> > >
> > > Thanks, that solves that problem.  Han-Wen, can I commit this fix to 
> > > CVS?
> >
> >
> > Please don't; fixes like these can easily lead to problems on other 
> > platforms.  I have successfully built 2.7.10 on MacOS 10.3, so this 
> > should no longer be necessary. The (inelegant) fix was to use <math.h> 
> > everywhere.
>
>
> Everywhere?  As in replacing it in 30 different source files?  That's a 
> lot of extra
> stuff to put into the fink patches... could we find an elegant fix?  

No, I have committed it to the mainline branch, because GCC 3.3 has the 
same problem on all platforms.

Now that you mention it, I remember that I installed GCC 3.4 by hand, 
separately to be able to build Lily. However,  it seems to work with 
/usr/bin/g++ fine as well.

> I 
> haven't
> been able to build lily for almost three weeks now, and there's a couple 
> of doc
> updates that might be nice.

Can you check if current CVS works for you, and file a report if it doesn't?

--
 Han-Wen Nienhuys - address@hidden - http://www.xs4all.nl/~hanwen