Gentoo Archives: gentoo-science

From: "François Bissey" <f.r.bissey@×××××××××.nz>
To: Thomas Kahle <tom111@×××.de>
Cc: gentoo-science@l.g.o
Subject: [gentoo-science] Re: [Bug 320669] dev-python/matplotlib-0.99.1.1-r1 doesn't build because numpy lapack support is broken
Date: Wed, 30 Jun 2010 09:00:57
Message-Id: 201006302100.46178.f.r.bissey@massey.ac.nz
> On 06/30/2010 10:47 AM, bugzilla-daemon@g.o wrote: > > ------- Comment #6 from f.r.bissey@×××××××××.nz 2010-06-30 08:47 0000 > > ------- that could explain a lot of things since I suffer from a broken > > atlas as well. I will try switching to a working lapack and see what > > happens. > > By the way, do we (does sage) really need this atlas implementation, or > would reference work in principle? This self-tuning business makes me > shiver. >
sage itself would probably be fine with reference I think. However linbox and historically ghmm (now gone) need clapack which is provided by lapack-atlas. Actually _you can_ build lapack-atlas using blas-reference not very efficient but doable. For the rest I need to check if linbox is linked in sage - if it is then you run the risk of linking with multiple versions of lapack. Which actually given how sage is built may not be such a big problem. Francois