On 12. mars 2011, at 02:01, "Garth N. Wells" <[email protected]> wrote:
> > > On 11/03/11 15:51, Johan Hake wrote: >> On Mar 11, 2011, at 3:10, "Garth N. Wells" <[email protected]> wrote: >> >>> >>> >>> On 11/03/11 00:24, Anders Logg wrote: >>>> On Thu, Mar 10, 2011 at 04:20:51PM -0800, Johan Hake wrote: >>>>> On Thursday March 10 2011 16:17:49 Anders Logg wrote: >>>>>> On Fri, Mar 11, 2011 at 12:53:16AM +0100, Anders Logg wrote: >>>>>>> On Thu, Mar 10, 2011 at 11:39:52PM +0000, Garth N. Wells >>>>>>> wrote: >>>>>>>> On 10/03/11 23:38, Anders Logg wrote: >>>>>>>>> I'm having trouble passing a MeshFunction from the >>>>>>>>> Python interface to >>>>>>>>> >>>>>>>>> the following function in the C++ Form class: void >>>>>>>>> set_cell_domains(boost::shared_ptr<const >>>>>>>>> MeshFunction<uint> > cell_domains); >>>>>>>> >>>>>>>> Change this to >>>>>>>> >>>>>>>> void set_cell_domains(boost::shared_ptr<const >>>>>>>> MeshFunction<unsigned >>>>>>>> >>>>>>>> int> > cell_domains); >>>>>>>> >>>>>>>> (uint -> unsigned int) >>>>>>> >>>>>>> Same problem (but with uint --> unsigned int): >>>>>>> >>>>>>> TypeError: in method 'Form_set_cell_domains', argument 2 of >>>>>>> type 'boost::shared_ptr< dolfin::MeshFunction< unsigned int >>>>>>>> const >' >>>>>>> >>>>>>> Printing out type() for the argument just prior to the call >>>>>>> gives >>>>>>> >>>>>>> <class 'dolfin.cpp.MeshFunctionUInt'> >>>>>>> >>>>>>> So somehow it is not recognized that >>>>>>> dolfin.cpp.MeshFunctionUint can be used as a >>>>>>> 'boost::shared_ptr< dolfin::MeshFunction< unsigned int > >>>>>>> const >. >>>>>> >>>>>> I have temporarily added pure pointer versions of the >>>>>> set_foo_domains functions in the Form class. These work fine >>>>>> from Python. These should be removed as soon as we figure out >>>>>> how to pass the arguments as shared ptrs. >>>>> >>>>> I need some time to get into the changes Garth has done. If it >>>>> is not fixed by then I can take a look at it, weekendish... >>>> >>> >>> I looks like shared_ptrs are not declared for MeshFunctions in the >>> SWIG version < 2 part of the interface. >> >> That should be easily fixed. >> >>> When do we want to move to SWIG 2 only? >> >> I think it would be nice to try this for the next release. If we >> decide for it I can take a shot at it sometime next week. >> > > I say that we do it now - I'm tired of wasting time making the old > version of SWIG work (I need SWIG 2.0 for Trilinos). 2.0 is *much* > better for shared pointers. > > We can make the change now and developers can install the SWIG 2. We can > figure out the swig2.0 executable issue during the next week. > > If anyone has an objection, please speak up quickly. > Could we wait till over the weekend? -- Marie > Garth > > >> Johan >> >> >>> Garth >>> >>>> ok, thanks. >>>> >>>> -- Anders >>> >>> >>> _______________________________________________ Mailing list: >>> https://launchpad.net/~dolfin Post to : >>> [email protected] Unsubscribe : >>> https://launchpad.net/~dolfin More help : >>> https://help.launchpad.net/ListHelp > > _______________________________________________ > Mailing list: https://launchpad.net/~dolfin > Post to : [email protected] > Unsubscribe : https://launchpad.net/~dolfin > More help : https://help.launchpad.net/ListHelp _______________________________________________ Mailing list: https://launchpad.net/~dolfin Post to : [email protected] Unsubscribe : https://launchpad.net/~dolfin More help : https://help.launchpad.net/ListHelp
