Bartosz Sawicki wrote:
> Marie Rognes wrote:
>> Bartosz Sawicki wrote:
>>> Hi,
>>>
>>> I have few months break in working with Fenics. Today I've just
>>> downloaded the current version from hg repository, ad realized that my
>>> previously working code fails.
>>>
>>> It seems that the problem is in FFC, which produce no-warning-free C++
>>> code, especially evalute_dof() method.
>>>
>>> The error is very easy to repeat. Just try to compile any form with
>>> Nedelec elements. For example:
>>>
>>> # File first.form
>>> eN = FiniteElement("Nedelec", "tetrahedron", 0)
>>> eL = VectorElement("Lagrange","tetrahedron", 1)
>>> v = TestFunction(eN)
>>> u = TrialFunction(eN)
>>> source = Function(eL)
>>> a= dot(rot(v), rot(u))*dx
>>> L = dot(v,source)*dx
>>>
>>> $ ffc -l dolfin first.form
>>>
>>> $ `pkg-config --variable=compiler dolfin` `pkg-config --cflags dolfin`
>>> -c main.cpp
>>>
>>> first.h: In member function ‘virtual double
>>> UFC_firstBilinearForm_finite_element_1::evaluate_dof(unsigned int,
>>> const ufc::function&, const ufc::cell&) const’:
>>> first.h:1394: warning: unused variable ‘Jinv_00’
>>> first.h:1395: warning: unused variable ‘Jinv_01’
>>> first.h:1396: warning: unused variable ‘Jinv_02’
>>> first.h:1397: warning: unused variable ‘Jinv_10’
>>> first.h:1398: warning: unused variable ‘Jinv_11’
>>> first.h:1399: warning: unused variable ‘Jinv_12’
>>> first.h:1400: warning: unused variable ‘Jinv_20’
>>> first.h:1401: warning: unused variable ‘Jinv_21’
>>> first.h:1402: warning: unused variable ‘Jinv_22’
>>> (...)
>>>
>>> If I comment out these lines, and everything seems to work properly,
>>> so it seems that this is just a tiny bug :)
>>>
>>>
>> See if it works better now.
>>
>
> Yes, know it works perfectly :) Thank you.
>
>> PS: ffc/compiler/format/codesnippets.py needs a clean-up (especially now
>> ;) ).
>
> I realized that there is no single demo in dolfin which use Nedelec.
> Don't you think, that it won't be such simple bugs when buildbot has a
> chance to find it?
> I can prepare simple demo catalog, if you like.
>
A Python example was added recently,
sandbox/electromagnetics/demos
but hasn't been moved to the usual demo folder yet. A C++ version of the
same problem would be good. I suspect that JIT may use less stringent
g++ compiler flags than dolfin.
If you have something, send a hg bundle to the dolfin-dev list
Garth
> regards,
> BArtek
>
> _______________________________________________
> FFC-dev mailing list
> [EMAIL PROTECTED]
> http://www.fenics.org/mailman/listinfo/ffc-dev
_______________________________________________
DOLFIN-dev mailing list
[email protected]
http://www.fenics.org/mailman/listinfo/dolfin-dev
