[Interest] Sporadic rendering problem QtQuick/OpenGl
Gunnar Sletta
gunnar at sletta.org
Fri Oct 21 10:10:05 CEST 2016
Hi Patrick,
The bug is not known to me, but it looks like there is a bug with the geometry being passed to OpenGL by the rendering code. A hint that may help you track down the root of the problem could be too specify the environment variable QSG_RENDERER_DEBUG=upload and look at the vertexdata. There'll be a lot of output there, but it might help to provide some pointer as to where the problem lies.
If you manage to make a testcase that consistently reproduces it, please file the bug to bugreports.qt.io.
cheers,
Gunnar
> On 19 Oct 2016, at 23:22, Patrick.Flaig at continental-corporation.com wrote:
>
> Hi,
> I got the recommodation to send my question/problem to this mailing list.
>
> On our i.MX6 based target device we can observe sporadic rendering problems when using QtQuick.
> As the problem is only happening sporadic we were not able to create a specific code that is causing the rendering problem.
>
> Background information: our system has two framebuffers, Qt is rendering in the foreground framebuffer.
> We are using Qt 5.6 running on a Linux Kernel 3.10.17 (patches from NXP for the iMX6 are applied), including latest opengl libraries from NXP.
>
> We have seen two issues:
> 1. Triangle showing parts of the background framebuffer.
>
> 2. Warped Graphics
>
> We have already started a discussion with NXP regarding that issue.
> The last statement from NXP is:
> 1. Eventually I suspect here an error in the QT framework when passing the mesh to OpenGL. It can be that for example a very simple geometry is used in the vertex buffer (e.g. 2 triangles defining the full frame buffer) but in the end 3 or more triangles are specified due to an application software error, so the OpenGL pipeline tries to render a 3rd triangle. For that it reads the coordinates from the vertex buffer (that contains valid data only for 2 triangles !), and thus invalid data is read therefore random triangles appear on the screen.
> 2. In this example again it looks like a geometry problem. Because the texture containing the gauge appears to be mapped on a strange 2d mesh composed by two incomplete circles. My idea with alpha was just to confirm that some draw primitives are corrupted, which is pretty clear now.
>
> Does anybody know about existing Qt bugs which could cause the issue we can observe?
> Any ideas how to debug this issue?
>
> Thanks a lot.
>
> Patrick
>
> http://www.continental-corporation.com
>
> <Mail Attachment.jpeg>
> ________________________________________________________________________
>
> Continental Automotive GmbH, Vahrenwalder Str. 9, D-30165 Hannover
> Vorsitzender des Aufsichtsrats/Chairman of the Supervisory Board: Helmut Matschi
> Geschaeftsfuehrer/Managing Director: Georg Sistermanns, Harald Stuhlmann
> Sitz der Gesellschaft/Registered Office: Hannover
> Registergericht/Registered Court: Amtsgericht Hannover, HRB 59424, USt.-ID-Nr./VAT-ID-No. DE814950663
> ________________________________________________________________________
>
> Proprietary and confidential. Distribution only by express authority of Continental AG or its subsidiaries.
>
> <issue_1.jpg><issue_2.jpg>_______________________________________________
> Interest mailing list
> Interest at qt-project.org
> http://lists.qt-project.org/mailman/listinfo/interest
>
More information about the Interest
mailing list