ffc team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Jan 31, 2008 at 11:27:36AM +0100, Anders Logg wrote:
> On Wed, Jan 30, 2008 at 09:23:42PM -0500, Jake Ostien wrote:
> > Hi,
> > 
> > Something doesn't seem right when I try and define multiple integrals 
> > over subdomains in a form file.
> > 
> > See the attached test.form and test.txt.  Basically I copied the element 
> > tensor section of the generated header file from two different 
> > compilations of test.form, one where
> > 
> > a = v*u*dx  (for a reference)
> > 
> > and one with
> > 
> > dx0 = Integral("cell", 0)
> > dx1 = Integral("cell", 1)
> > 
> > a = v*u*dx0 + 10*v*u*dx1
> > 
> > Now I would expect that the cell_integral_0 from the second compilation 
> > to be identical to the reference compilation, and that cell_integral_1 
> > would be just scaled by 10, but that is not what is happening.  It 
> > appears that in the second compilation the two subdomain integrals are 
> > just added together (see test.txt).
> > 
> > I didn't get too much out of running ffc with the -d1 flag, but I really 
> > don't know what I am looking for.  Any help on where to look?
> > 
> > Thanks,
> > Jake
> 
> Looks like an obvious bug that should be easy to fix. I will take a
> look.

ok, try again and see if it works now. It turned out that the
integrals for the different sub domains were computed correctly but
then completely ignored when generating code.

-- 
Anders