ffc team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Mon, Jul 13, 2009 at 01:05:31PM +0100, Garth N. Wells wrote:
> There have been a lot of changes since the last release, and it would be 
> nice to make a release before a full assault on parallelisation, so I 
> suggest that we make a release soon. Regarding the few outstanding 
> issues in the Roadmap for release 0.9.3:
> 
> - Is the move to the new parameter system complete?

Yes, I think so, at least on the C++ side. Johan Hake knows the status
of the Python version better.

There are a few things that are not tested (like parsing command-line
arguments) and not implement (like passing options to PETSc) but that
can be move to 0.9.4.

> - Can command line arguments be parsed with the new system?

Yes, in principle. I have tested it before and it should work, but
haven't in a while. The following should work:

  parameters.parse(argc, argv);

Perhaps we need a way to strip prefixes when reading parameters to
make the following possible:

  krylov_solver.parse(argc, argv, "krylov_solver");

which would pick up options like

  -krylov_solver.relative_tolerance=1e-8

> I suggest moving the target items
> 
> - Break up function Assembler::assemble_system
> - Cleanup Function class logic
> - Extracting vector field from tensor field

Sounds good.

I've been meaning to look at the Function class logic but haven't had
time.

-- 
Anders


> to the next release.
> 
> A new release will also require new releases of UFC, Viper, FFC and 
> possibly UFL.

Attachment: signature.asc
Description: Digital signature