[vtk-developers] Rendering of vtkPolyData broken on Nvidia systems

[Marcus D. Hanwell]

On Mon, Jun 25, 2012 at 5:04 PM, Marcus D. Hanwell
<marcus.hanwell at kitware.com> wrote:
> On Fri, Jun 22, 2012 at 11:58 AM, Oleg Koren <koren at curefab.com> wrote:
>> Hello,
>>
>> I think, now I know the root of the problem. Great thanks for the suggestion
>> to run an OpenGL logger. My log boils down to the following:
>>
>> glColor3f(1.0f, 0.0f, 0.0f);
>> glDeleteLists(1,1);
>> glGenLists(1);
>> glNewList(1,GL_COMPILE);
>> glDisable(GL_LIGHTING);
>> glBegin(GL_LINE_STRIP);
>>        glVertex3f(0.000000,0.000000,0.000000);
>>        glVertex3f(1.000000,0.000000,0.000000);
>>        glVertex3f(1.000000,1.000000,0.000000);
>>        glVertex3f(0.000000,1.000000,0.000000);
>>        glVertex3f(0.000000,0.000000,0.000000);
>> glEnd();
>> glEnable(GL_LIGHTING);
>> glEndList();
>> glCallList(1);
>>
>> glColor3f(1.0f, 1.0f, 0.0f);
>> glDeleteLists(2,1);
>> glGenLists(1);
>> glNewList(2,GL_COMPILE);
>> glDisable(GL_LIGHTING);
>> glBegin(GL_LINE_STRIP);
>>        glVertex3f(0.000000,0.000000,0.000000);
>>        glVertex3f(0.000000,1.000000,0.000000);
>>        glVertex3f(0.000000,1.000000,1.000000);
>>        glVertex3f(0.000000,0.000000,1.000000);
>>        glVertex3f(0.000000,0.000000,0.000000);
>> glEnd();
>> glEnable(GL_LIGHTING);
>> glEndList();
>> glCallList(2);
>>
>> and this actually doesn't work. Which is, I guess, because lists are
>> deprecated since OpenGL 3.0 and should receive very limited support by now,
>> if any. Removing all calls to list functions does the trick.
>>
>> That means, one should always set ImmediateModeRenderingOn on a
>> vtkPolyDataMapper. Immediate rendering mode disables OpenGL lists.
>>
> Hi Oleg,
>
> I have started looking into this issue. The code you gave didn't match
> the images, or the GL log trace you supplied. Pushing the second set
> of lines into z rather than x produced a matching image with the same
> coordinates as in the log.
>
> I am not able to reproduce you issue here using nVidia binary driver
> 302.17 using a Linux 3.4.4 kernel, and an nVidia Quadro 600 card. As
> you did not use SetInputData I guess you are using VTK 5.10 or below,
> I don't think that rendering code has changed very much at all in the
> last few releases and so it is unlikely to affect but it would be good
> to know your VTK version.
>
As another few data points Aashish tried this on a few systems with
Quadro cards, and I was able to verify on a Macbook Pro 13, using
295.49 nVidia driver on Linux 3.3.5 kernel with a consumer line
GeForce 320M I do not see any rendering issues in the supplied example
- this is dual booted into Linux (no VMWare etc).

I think that this is probably a driver bug in one or two versions of
the binary driver (see release notes for some regression fixes related
to display lists), and simply updating the driver version will get rid
of any rendering issues.

Thanks,

Marcus