[vtk-developers] Adding ComputeArea() to vtkTriangle

[David Doria]

On Tue, Dec 22, 2009 at 4:54 PM, Bill Lorensen <bill.lorensen at gmail.com>wrote:

> You can do whatever you want to in a test as long as it is portable.
>
> On Tue, Dec 22, 2009 at 4:45 PM, David Doria <daviddoria+vtk at gmail.com<daviddoria%2Bvtk at gmail.com>>
> wrote:
> >  Tue, Dec 22, 2009 at 4:39 PM, David Cole <david.cole at kitware.com>
> wrote:
> >>
> >> On Tue, Dec 22, 2009 at 4:34 PM, David Doria <daviddoria+vtk at gmail.com<daviddoria%2Bvtk at gmail.com>
> >
> >> wrote:
> >>>
> >>> On Tue, Dec 22, 2009 at 4:21 PM, Bill Lorensen <
> bill.lorensen at gmail.com>
> >>> wrote:
> >>>>
> >>>> 1) I would modify the current test for triangle.
> >>>> 2) When you add the TestPolygon test, check vtkPolygon's coverage to
> >>>> see if you can exercise some uncovered code. There are 193 uncovered
> >>>> lines of code:
> >>>>
> >>>>
> http://www.cdash.org/CDash/viewCoverageFile.php?buildid=498540&fileid=10908473
> >>>>
> >>>> Bill
> >>>
> >>> Ok, I'll add more polygon tests. I've never looked at these coverage
> >>> files - I'm assuming that the red 000000 means that line is not tested?
> Can
> >>> you figure out where the line is tested by the number? e.g.
> >>>
> >>> 00037579   void vtkPolygon::ComputeNormal(vtkPoints *p, double *n)
> >>>
> >>> How would you turn 00037579 into a file name?
> >>
> >> The number preceding each line is supposed to be the number of times
> that
> >> line was called. To find callers, grep for the function name, once you
> find
> >> a candidate, set a breakpoint on the line and run a test in the debugger
> to
> >> verify that you can stop there...
> >>
> >
> > A couple more things...
> > 1) Can/should tests use smart pointers?
> > 2) Can this function be placed somewhere so it doesn't need to be copied
> > into many of the tests?
> > // Perform a fuzzy compare of floats/doubles
> > template<class A>
> > bool fuzzyCompare(A a, A b) {
> >   return fabs(a - b) < vtkstd::numeric_limits<A>::epsilon();
> > }
> > Thanks,
> >
> > David
>

I have written tests for several of the vtkPolygon functions (2 versions of
ComputeNormal and IntersectPolygonWithPolygon). Some were very awkward -
taking double* input rather than vtkPoints - maybe I will wrap these into

polygon->ComputeNormal(normal);
and
polygon->IntersectWithPolygon(polygon2);

Anyone have a problem with this?

Is there a way I can verify the coverage increase before committing these?
Or should I just look at the dashboard coverage the next day?

Thanks,

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20091223/1cf99208/attachment.html>