[vtk-developers] Long failing test

[Marcus D. Hanwell]

On Tue, Jan 6, 2015 at 2:48 PM, Ken Martin <ken.martin at kitware.com> wrote:
> Nothing to do with OpenGL2, but TestChartXYZ test fails on various platforms
> and it has been bugging me.  It turns out it is failing due to floating
> point issues. Basically
>
> 1)      By default the plot data points define the bounds of the axes for
> the chart
>
> 2)      the axes define the plot area (clip stuff outside that)
>
> 3)      which defines the clipping plane equations
>
> 4)      which can in turn clip the original data points subject to floating
> point precision
>
> This is what is happening (FYI, the point that disappears in the middle is
> actual on the boundary in the depth axis). So ... I can “fix” the test by
> defining axes that are slightly bigger than the data and replacing the valid
> image. This is what I am planning to do unless someone hollers. I’m not sure
> what the more general fix is.
>
That seems reasonable to me, I hadn't spotted the floating point issue
you describe but it makes sense.

Thanks,

Marcus