[Numpy-discussion] Branching 1.1.x and starting 1.2.x development

[Charles R Harris]

On Tue, May 20, 2008 at 8:32 PM, David Cournapeau <
david at ar.media.kyoto-u.ac.jp> wrote:

> Charles R Harris wrote:
> > Curious bug on Stefan's Ubuntu build client:
> >
> > ImportError: /usr/lib/atlas/libblas.so.3gf: undefined symbol:
> _gfortran_st_write_done
> > make[1]: *** [test] Error 1
>
> Unfortunately not curious and very well known: I bet the client is
> configured to build with g77; libblas.so.3gf uses the gfortran ABI
> (which is incompatible with the g77 ABI). Any library with 3gf is the
> gfortran ABI.
>
> Two possible solutions:
>    - forget about gfortran, and install atlas3* (g77 ABI) instead of
> libatlas* (gfortran ABI)
>    - use gfortran in the build
>
> The first one is safer: I cannot find any concrete information on the
> Ubuntu side of things, but I believe g77 is still the default ABI.
> (Debian) Lenny uses gfortran:
>
> http://wiki.debian.org/GfortranTransition
>

Ah... Is this going to be a problem with Debian installs? I just ran
Stefan's buildbot too see if it was working yet.

Chuck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20080520/cfa0afbb/attachment.html>