mmapi on Kirkstone-l4t-r32.7.x

[SGerm]

Hi,
I try to use the mmapi on a Tx2i.
My app works well under the Nvidia OS installed with SDKManager 1.8.1.10392.
With Kirkstone-l4t-r32.7.x, the video_decoder sample is also working well, my code is generally inspired by that sample.

But, with my app, on Kirkstone-l4t-r32.7.x, the open, setOutputPlaneFormat, setupPlane, setStreamStatus, ... seem to work.
But at VIDIOC_QBUF of the V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE, I got systematically a "Device or resource busy" (errno 16)

One thing that I suspect is the fact that I reuse the Common file that was coming with the sample (NvVideoDecoder.cpp, ...) coming from the my Nvidia furnished OS.
But I cannot find any evidence that this file has been modified in the meantime.

Also, I am suspecting compilation / linkage invocation, but I cannot figure out what could be wrong here too.

Thanks,
Stephane

[SGerm]

Sorry, it is not systematic like I sed.
I setup my output plane with 12 buffer( I also try also with more), so I can VIDIOC_QBUF 12 buffers, but I never receive the V4L2_EVENT_RESOLUTION_CHANGE and at the 13eme sample, I got the "Device or resource busy"...

[SGerm]

Here some more information.
It is like:
https://forums.developer.nvidia.com/t/mm-api-v4l2-event-resolution-change-event-never-received/173216
But I can now confirm that all the .h used to compile my app are the same that are in .../build/tmp/work/armv8a_tegra186-poky-linux/tegra-mmapi-samples/32.7.2-20220417024839-r0/tegra-mmapi/usr/src/jetson_multimedia_api/include

[SGerm]

Ok sorry, let me close that issue. 30 buffers was still not enough. Sometime, Gst 1.20, (compared to 1.14 from Nvidia Os ?) send a lot of small buffers (< 6 Bytes, metadata? SPS? PPS? so if you not receive an image payload in that buffer, you hang there without V4L2_EVENT_RESOLUTION_CHANGE