Hi Jonathan,

Thanks so much for the reply.

I downgraded numpy from 1.7.0 to 1.6.2, and it did help. I first found I
had to recompile pysparse (same software version) before I could even
"import fipy." Then I reran the test and got only 4 errors (
http://pastebin.com/GRagmTgc). I noticed that the errors this time seemed
to be related to scipy, so I went to an interactive python shell and tried
"import scipy.special", and it complained that it scipy was compiled
against a newer version than the currently installed numpy. (I'm currently
using precompiled binaries of numpy/scipy/matplotlib from my distribution's
package manager). So I went back and also got the version of scipy and
matplotlib that were released with numpy-1.6. This left me with
numpy-1.6.2
scipy-0.10.1
matplotlib-1.1.1

With this combination, the full test suite ran with no errors (
http://pastebin.com/fb73KwWZ).

Finally, I noticed that right before numpy 1.7.0 was released, scipy was
already on 0.11.0, so I re-upgraded that one (still pre-combiled binary),
and got no errors (http://pastebin.com/csib0wTK) with
numpy-1.6.2
scipy-0.11.0
matplotlib-1.1.1

I'm also perfectly willing to run different combinations of software if
that would help. For example, I could run numpy-1.6.2 with the newest
version matplotlib (though I'm not sure matplotlib affects the
non-interactive test). I would simply have to compile locally rather than
used pre-compiled binaries, which should not be a problem.

So as far as I can tell, you were exactly right, and numpy-1.7.0 was
causing the issues. If there's anything else I can do that would help
clarify what's going on or help ease the transition to numpy-1.7.0, let me
know.

Thanks again for your help with this.

Ray

On Fri, Feb 22, 2013 at 9:57 AM, Jonathan Guyer <[email protected]> wrote:

>
> On Feb 21, 2013, at 7:43 PM, Raymond Smith wrote:
>
> > I've just installed fipy, and I got a number of errors while running the
> test. My output is located at http://pastebin.com/grn7Maki.
> > I'm using Arch Linux, 64 bit, with fipy-3.0, and I've installed pysparse
> (1.1.1), gmsh (2.6.1), numpy (1.7.0), scipy (0.11.0), matplotlib (1.2.0),
> mayavi (4.2.0).
> >
> > I looked through some of the tests an couldn't find a clear pattern for
> which tests succeeded or failed. I tried going to an interactive python
> prompt and executing the code in the docstrings which is being executed in
> the tests. I got the same errors, but I also got errors trying to do that
> in tests that did not fail, so I didn't seem to be copying the tests
> accurately at the prompt.
> >
> > In particular, in the modules,
> > * models.levelSet.electroChem
> > * meshes.mesh.Mesh
> > * fipy.meshes.gmshMesh.MSHFile
> > * fipy.meshes.abstractMesh.AbstractMesh
> > I got
> > "ValueError: operands could not be broadcast together with shapes (2)
> (2,640) (2)"
> >
> > And I also got a few other errors, including a couple of
> > "IndexError: axis 1 out of bounds [0, 1)"
> > in a few seemingly random other places.
> >
> > Any thoughts?
>
> I don't know for sure, but my guess is that this is due to numpy 1.7.
> We're not running that on any of our slaves, yet. We obviously need to get
> one set up to test that, since it's the official release now, but can you
> try downgrading to numpy 1.6.x and see if you still get the same errors?
>
>
> _______________________________________________
> fipy mailing list
> [email protected]
> http://www.ctcms.nist.gov/fipy
>   [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]
>
_______________________________________________
fipy mailing list
[email protected]
http://www.ctcms.nist.gov/fipy
  [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]

Reply via email to