why atlas for py-scipy?

[Jack Howarth]

On Sat, Jul 23, 2011 at 09:29:08PM +0200, Rainer Müller wrote:
> On 2011-07-23 21:11 , Jack Howarth wrote:
> >    Currently the default py26-scipy can't build on Lion due to the
> > broken atlas build.
> 
> As a workaround use the +no_atlas variant.
> 
> > Why are we defaulting to atlas or the internal
> > blas/lapack? Shouldn't we simply use the system lapack/blas on
> > darwin instead?
> 
> Yep, that came up several times now [1][2], and atlas causes problems on
> specific hardware [3].
> 
> At least we could use
>     default_variants +no_atlas
> for py*-scipy to avoid the dependency from which we do not gain much.
> 
> Rainer
> 
> [1]
> http://lists.macosforge.org/pipermail/macports-dev/2010-April/011682.html
> [2]
> http://lists.macosforge.org/pipermail/macports-users/2010-June/020343.html
> [3]
> http://lists.macosforge.org/pipermail/macports-users/2011-March/023978.html

Actually this brings up a separate question. Are there any plans to depreciate
some of the older gcc4x releases in MacPorts? Over on fink, we are leveraging
the creation of a new 10.7 tree to prune out all of the older gcc releases
before gcc46. FYI, you will be in good shape with gcc 4.6.1 with Lion as all
of the changes to deal with the -pie change are present.
                 Jack

> _______________________________________________
> macports-dev mailing list
> macports-dev at lists.macosforge.org
> http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev