ufl team mailing list archive

[Martin Sandve Alnæs <martinal@xxxxxxxxx>]

On 19 October 2011 17:47, Martin Sandve Alnæs <martinal@xxxxxxxxx> wrote:
> On 19 October 2011 17:00, Anders Logg <logg@xxxxxxxxx> wrote:
>> On Wed, Oct 19, 2011 at 04:47:03PM +0200, Martin Sandve Alnæs wrote:
>>> I want to change ('',Vector,Tensor)Constant from DG0 to Real in UFL,
>>> as has been done already in PyDOLFIN. It seems to me that only a
>>> couple of lines need to be changed in UFL, and FFC will be unaffected,
>>> but I haven't tried it yet. Does anyone foresee any trouble with this
>>> change? Is it safe enough to do now? Since a dolfin::Constant is
>>> generated for the wrapper code, and this does not have any
>>> constructors making it possible to actually create DG0 functions from
>>> an ufl.Constant, I don't think it's possible for C++ dolfin users to
>>> be affected by this either.
>>
>> Seems ok to me.
>
> I did the change locally (see lp:~martinal/ufl/work),
> and running the ffc tests results in basically the expected
> regressions, like dofmap changes and signatures. Only a few
> tabulate_tensor entries changed, but the differences are hard to spot
> and seem to be basically FE0 <-> FE1 variable naming changes and some
> floating point precision changes in the values.
>
> The dolfin demos all compile, and running a few of them manually
> everything looks splendid.
>
> I suggest that some of the FFC devs do a
> cd ufl
> bzr pull lp:~martinal/ufl/work
> <install>
>
> cd ffc
> bzr pull lp:~martinal/ffc/work  # just some "triangle"->triangle
> test if you want to
> update references
>
> and then push the ufl change and ffc references simultaneously so as
> not to break stuff.
>
> Going home now.
>
> Martin

All dolfin tests pass locally with this change, so I am confident that
this is safe now.

Martin