[MacPorts] #38571: py27-scipy @0.11.0: scipy.test() failed
MacPorts
noreply at macports.org
Thu Aug 15 07:56:44 PDT 2013
#38571: py27-scipy @0.11.0: scipy.test() failed
-------------------------+--------------------
Reporter: gabor@… | Owner: sean@…
Type: defect | Status: new
Priority: Normal | Milestone:
Component: ports | Version: 2.1.3
Resolution: | Keywords:
Port: py27-scipy |
-------------------------+--------------------
Comment (by wimmer@…):
The scipy test failures are due to two reasons:
- Note that they happen when linked against the Accelerate framework
(which is the default). The Accelerate framework however adheres to the
g77 ABI conventions which are incompatible in some respects with gfortran
(in particular, functions returning complex numbers, and also functions
returning Fortran REALs). See https://github.com/Homebrew/homebrew-
science/issues/7#issuecomment-22174939 for an extended discussion. scipy
0.12 has some wrappers to deal with this problem, but the wrappers are
incomplete. This can be solved by specifying the compilation flag
{{{-ff2c}}} so that gfortran goes back to g77 conventions.
- Arpack has a bug that is triggered by some conventions in the Accelerate
framework: http://forge.scilab.org/index.php/p/arpack-ng/issues/1259/ .
This causes many of the Arpack test failures.
(Future versions of scipy will have both problems fixed probably natively:
https://github.com/scipy/scipy/pull/2695)
I add a Portfile and an Arpack patch to fix both issues.
Let me note that after the applying these fixes, I still get the problem
of #35141, but that I had already before the patches.
--
Ticket URL: <https://trac.macports.org/ticket/38571#comment:7>
MacPorts <http://www.macports.org/>
Ports system for OS X
More information about the macports-tickets
mailing list