dolfin team mailing list archive

[Johan Hake <hake@xxxxxxxxx>]

On Saturday 07 February 2009 14:38:38 Martin Sandve Alnæs wrote:
> On Sat, Feb 7, 2009 at 1:06 PM, Johan Hake <hake@xxxxxxxxx> wrote:
> > On Friday 06 February 2009 16:12:06 Martin Sandve Alnæs wrote:
> >> On Fri, Feb 6, 2009 at 3:50 PM, DOLFIN <dolfin@xxxxxxxxxx> wrote:
> >> >  - The variable dolfin.__version__ now holds the version of DOLFIN.
> >> > This is used when calling the jit compiler, adding this to the
> >> > signature of the form
> >>
> >> By the "jit compiler", do you mean compilation of dolfin functions? If
> >> so, that sounds good.
> >
> > Actually not, but it makes sense to include it. Will do.
>
> Nice. Compilation of dolfin functions do depend on the dolfin version,
> so this is important.

It turned out that is already was included :)

> >> Or do you mean when calling jit from FFC? If so, why?
> >
> > Well, I think that we concluded, in a thread far from now, that we should
> > include the DOLFIN version in the signature of the form compilations, so
> > I threw it in.
>
> I don't recall any such conclusion, and don't understand why?

  <http://www.fenics.org/pipermail/dolfin-dev/2009-January/011895.html>

But it do not tell us why :)

> > First it probably won't heart beeing in, but may not make too much sense,
> > as we compile pure UFC forms. We should maybee include the ufc version
> > instead/too?
>
> Good idea. I'd say instead, unless the compiled module includes dolfin
> wrapper code in which case adding the dolfin version makes sense.

It doesn't.

If we want to include the ufc.version we need to include this in the compiled 
extension module. Any good suggestion on how to do this clean and easy, hard 
code it, as it is in SConstruct now?

Johan