[vtk-developers] Failing test rendering problem

[David Cole]

You may notice from git blame, those comments were added only
recently, in October 2014 in commit e136e01e...

I didn't even know when I did that commit that there was a python
equivalent to vtkTesting.cxx. I wonder why vtkTesting.cxx isn't simply
wrapped and used directly from python...?


David C.



On Thu, Jul 2, 2015 at 10:45 AM, Joachim Pouderoux
<joachim.pouderoux at kitware.com> wrote:
> David,
>
> We discovered that interesting comment in vtkTesting.cxx (line 775):
>
>     // Typically, the image testing is done using the back buffer
>     // to avoid accidentally capturing overlapping window artifacts
>     // in the image when using the front buffer. However, some graphics
>     // drivers do not have up to date contents in the back buffer,
>     // causing "failed" tests even though, upon visual inspection, the
>     // front buffer looks perfectly valid... So:
>     //
>     // If the test failed using the back buffer, re-test using the
>     // front buffer. This way, more tests pass on dashboards run with
>     // the Intel HD built-in graphics drivers.
>
> We gonna change Testing.py to do the same - question: why this has never
> been done? :)
>
> J.
>
> Joachim Pouderoux
>
> PhD, Technical Expert
> Kitware SAS
>
>
> 2015-07-02 15:37 GMT+02:00 David Thompson <david.thompson at kitware.com>:
>>
>> Hi Mathieu,
>>
>> Thanks for looking into this. I've been noticing failures like that in VTK
>> and in some projects that use it. However, given the number of image tests
>> that VTK runs, adding 2s of delay to each one would slow the dashboards down
>> too much to be usable. However, it is a great start on figuring out how to
>> fix the issue. I don't see anywhere that
>> vtkRenderWindow::WaitForCompletion() (which calls glFinish) gets called when
>> grabbing the framebuffer. Perhaps adding this would be faster than a timer.
>> Can you verify that it works? If so, maybe we can figure out where to add it
>> to avoid slowing production code down for testing.
>>
>>         Thanks,
>>         David
>>
>> > I have some problem with tests wich render and compare to an image.
>> >
>> > Lots of test do the rendering which can be seen on screen, but the
>> > generated image is completelly black ( or white ) and the test fail.
>> >
>> > Most of these failing tests are python test but not only.
>> > The list of failing tests is attached.
>> >
>> > I've tried to understood what was going on, so i've tried to fix
>> > sync3dAll.py
>> >
>> > Adding :
>> >         import time
>> >         time.sleep(1)
>> >         renWin.Render()
>> >         time.sleep(1)
>> >
>> > after the first render() did the trick. It look like the renderring is
>> > not done fast enough ans the produced image is empty as it has not yet been
>> > rendered.
>> >
>> > Any help appreciated.
>> >
>> > Mathieu Westphal
>> > <tests>_______________________________________________
>> > Powered by www.kitware.com
>> >
>> > Visit other Kitware open-source projects at
>> > http://www.kitware.com/opensource/opensource.html
>> >
>> > Search the list archives at:
>> > http://markmail.org/search/?q=vtk-developers
>> >
>> > Follow this link to subscribe/unsubscribe:
>> > http://public.kitware.com/mailman/listinfo/vtk-developers
>> >
>>
>> _______________________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Search the list archives at: http://markmail.org/search/?q=vtk-developers
>>
>> Follow this link to subscribe/unsubscribe:
>> http://public.kitware.com/mailman/listinfo/vtk-developers
>>
>
>
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Search the list archives at: http://markmail.org/search/?q=vtk-developers
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/vtk-developers
>
>