On Mon, Feb 10, 2014 at 9:38 PM, Anders Logg <[email protected]> wrote:

> On Mon, Feb 10, 2014 at 07:58:51AM +0100, Johan Hake wrote:
> > On Sun, Feb 9, 2014 at 10:46 PM, Anders Logg <[email protected]> wrote:
> >
> >     On Sun, Feb 09, 2014 at 10:32:53PM +0100, Johan Hake wrote:
> >     > Wasn't the deal that we should merge UFC into FFC, and not the
> other way
> >     > around?
> >
> >     Yes, of course. At least I got it right in the issue...
> >
> >     > Before we switch from cmake to distutils in ufc I suggest we make
> the
> >     > shared_ptr switch in UFC/DOLFIN. I can take responsibility for this
> >     switch in
> >     > the SWIG layer.
> >
> >     Great, that's the hard part. I could do the switch in UFC and DOLFIN
> >     but I'm not sure it would help. I expect many things will break until
> >     the SWIG layer is fixed.
> >
> >
> > Sure no Python tests will run until we have the SWIG layer in place.
> >
> >
> >     Would it help if I switched in UFC + DOLFIN (except for SWIG) and
> >     pushed to new branches, or is it just as easy to run dolfinreplace?
> >
> >
> > A lot of things can be done with dolfiunreplace, but things are bound to
> go
> > wrong. So if you make the switch in the DOLFIN/UFC/FFC C++ layer first
> and at
> > least make everything compile. I also suggest that you get the C++ code
> > generation in-place too before we fix the SWIG layer.
>
> I have made the switch on the C++ side in UFC and DOLFIN branches
>
>   logg/topic-c++11
>

Ok, I will have a look at the SWIG layer later this week.

Johan


Things seem to build and run well on the C++ side, but SWIG is failing
> in UFC, and likely also DOLFIN which I never tested, with 'shared_ptr'
> is not a member of 'std'.
>
> >     > Also the build.py has nothing to do the with the UFC build system.
> That
> >     is a
> >     > module responsible for JIT-compile FFC generated code UFC code.
> >
> >     It uses Instant. I assumed somewhere along the chain some CMake file
> >     generated by UFC would be invoked.
> >
> >
> > That is true, and I suggest we keep this. The reason we switched to
> CMake in
> > instant was due to inferior compiler configuration support in distutils
> > compared to CMake. We can still install UFC with distutils and generate
> and
> > install a UFCConfig.cmake file for compiling agains UFC using distutils.
>
> ok.
>
> --
> Anders
>
_______________________________________________
fenics mailing list
[email protected]
http://fenicsproject.org/mailman/listinfo/fenics

Reply via email to