← Back to team overview

dolfin team mailing list archive

Re: [HG DOLFIN] Added SubFunction to dolfin_function.h so swig interface compile\nRename of function.in to function._in\nAdded __contains__ in FunctionSpace so one can write u in V.\nAdded % import ufc.h in dolfin_dunction_pre.i as dolfi

 

On Mon, Nov 03, 2008 at 02:35:07PM +0100, Martin Sandve Alnæs wrote:
> 2008/11/3 Johan Hake <hake@xxxxxxxxx>:
> > On Monday 03 November 2008 14:25:45 Johannes Ring wrote:
> >> On Mon, November 3, 2008 14:07, Johan Hake wrote:
> >> > On Monday 03 November 2008 13:31:29 Martin Sandve Alnæs wrote:
> >> >> 2008/11/3 Johan Hake <hake@xxxxxxxxx>:
> >> >> > On Monday 03 November 2008 13:07:38 Garth N. Wells wrote:
> >> >> >> DOLFIN wrote:
> >> >> >> > One or more new changesets pushed to the primary dolfin repository.
> >> >> >> > A short summary of the last three changesets is included below.
> >> >> >> >
> >> >> >> > changeset:   5069:fbbadf8f8b7ac1d0320109bb0c1e0f3a90a32f8a
> >> >> >> > tag:         tip
> >> >> >> > user:        "Johan Hake <hake@xxxxxxxxx>"
> >> >> >> > date:        Sun Nov 02 16:16:54 2008 +0100
> >> >> >> > files:       dolfin/function/dolfin_function.h dolfin/scons.cfg
> >> >> >> > dolfin/swig/dolfin_docstrings.i dolfin/swig/dolfin_function_post.i
> >> >> >> > dolfin/swig/dolfin_function_pre.i dolfin/swig/dolfin_headers.i
> >> >> >> > description:
> >> >> >> > Added SubFunction to dolfin_function.h so swig interface
> >> >> >> > compile\nRename of function.in to function._in\nAdded __contains__
> >> >>
> >> >> in
> >> >>
> >> >> >> > FunctionSpace so one can write u in V.\nAdded % import ufc.h in
> >> >> >> > dolfin_dunction_pre.i as dolfin::function inherits
> >> >>
> >> >> ufc::function\nand
> >> >>
> >> >> >> > last added ufc-1 as swig dependencies so swig find ufc.h
> >> >> >>
> >> >> >> Related to this change, I'm getting the swig error
> >> >> >>
> >> >> >>      dolfin/swig/dolfin_function_pre.i:2: Error: Unable to find
> >> >>
> >> >> 'ufc.h'
> >> >>
> >> >> > Hmmm, this should work...
> >> >> >
> >> >> > I added 'ufc-1' to SwigDependencies so simula_scons would pick up the
> >> >> > dependency. I guess that ufc.h is located in a directory that is
> >> >>
> >> >> included
> >> >>
> >> >> > anyway on my computer, and thats why I did not get the error.
> >> >> >
> >> >> > Johannes any clue why this is not working?
> >> >> >
> >> >> > Johan
> >> >>
> >> >> My previous experience is that pkg-config usually skips printing
> >> >> default include directories like /usr/include and /usr/local/include,
> >> >> while SWIG does not include those by default.
> >> >> If UFC is in one of those, just asking pkg-config won't work.
> >> >
> >> > It's only when something is installed under /usr that pkg-config does not
> >> > give
> >> > any output. This beaviour can be configured by setting the environment
> >> > variable
> >> >
> >> >   PKG_CONFIG_ALLOW_SYSTEM_CFLAGS
> >> >
> >> > before calling pkg-config.
> >> >
> >> > ufc.h is installed under /usr if not given any prefix. We should probably
> >> > set
> >> > default installation directory to /usr/local instead of /usr when not set
> >> > by
> >> > user while running ufc/setup.py.
> >> >
> >> > If no one have any objection maybee Johannes can do that?
> >>
> >> I guess someone with write access to the UFC hg repository should fix
> >> that.
> >
> > Ahh, I thought you had access to all sort of places :)
> >
> >> However, it won't help if someone installs UFC under /usr
> >> deliberately. A solution might be to force SWIG to always look under
> >> /usr/include by adding it to the default SWIG flags. It shouldn't do any
> >> harm i guess.
> >
> > I think we should change the installation dir _and_ set the environment
> > variable PKG_CONFIG_ALLOW_SYSTEM_CFLAGS when calling pkg-config in
> > simula_scons.
> >
> > Johan
> 
> How can the installation dir be set in a portable way then?
> Currently ufc/setup.py uses sys.prefix, which is portable.
> 
> And should instant set PKG_CONFIG_ALLOW_SYSTEM_CFLAGS=1
> for all its pkg-config calls? It's typically intended for use with SWIG,
> so I guess that would be ok?

Yes, at long as it works for me I'm happy. :-)

-- 
Anders

Attachment: signature.asc
Description: Digital signature


References