← Back to team overview

ffc team mailing list archive

Re: ffc compiler error

 

ok, thanks for spotting this.

Rob, it looks like numbering.py went missing again in 0.3.2. It was missing in 0.3.1 also and was added in 0.3.1-1. Could you make a new release with numbering.py included?

/Anders

Andrey wrote:
Thank you, for advice. The command you mentioned
(# python >>> import FIAT) works - it does not produce any error message.

But it already becomes clear that the trouble was in missed file in
FIAT-0.3.2 package. It is “numbering.py”. I took it from
http://www.fenics.org/pub/software/fiat/snapshot/fiat-snapshot.tar.gz
and put to FIAT installation directory. Now ffc compiler works.

Best regards,
Andrey

Looks like you didn't get FIAT installed correctly. (The FFC error message
is a little cryptic here.)

What if you do the following:

# python
import FIAT

Does that work?

/Anders


Hello, developers of Fenics!

I downloaded and installed the following packages:

numpy-1.0.3.1
FIAT-0.3.2
ufc-1.0
ffc-0.4.2
dolfin-0.7.1

using python-2.5 under Windows XP cygwin environment.
Then tried to compile example:
ffc Poisson.form
and have got the following message:
-------------------------------------------------
*** Warning: Unable to reorder entities. You need to patch or update your
*** Warning: installation of FIAT. Variable numbering_scheme is missing.
*** Warning: Results may be incorrect for higher order elements on tets
Traceback (most recent call last):
  File "/usr/bin/ffc", line 26, in <module>
    from ffc.common.debug import *
  File "/usr/lib/python2.5/site-packages/ffc/__init__.py", line 14, in
<module>
    from ffc.fem.finiteelement import FiniteElement
  File "/usr/lib/python2.5/site-packages/ffc/fem/finiteelement.py", line
14, in <module>
    from FIAT.shapes import *
  File "/usr/lib/python2.5/site-packages/FIAT/shapes.py", line 19, in
<module>
    import numbering
ImportError: No module named numbering
-------------------------------------------------

After that I installed all the set of packages under Debian,
and have got the same result.

This is probably a bug in software.
In any case I could not force ffc to work.

Best regards,
Andrey





References