Nvdsosd segmentation fault

I have a lot of sigsegvs when trying to use deepstream on Nvidia Jetson Nano. It is not the topic here but it can be related.

The simplest pipeline with nvdsosd does not work:

gst pipeline

gst-launch-1.0 filesrc location=/opt/nvidia/deepstream/deepstream/samples/streams/sample_720p.mjpeg !
jpegparse !
nvv4l2decoder mjpeg=1 !
nvvidconv !
nvdsosd display-clock=1 ! # Works with 0, does not work with 1
fakesink

Output:

Setting pipeline to PAUSED …
Opening in BLOCKING MODE
Opening in BLOCKING MODE
Pipeline is PREROLLING …
NvMMLiteOpen : Block : BlockType = 277
NVMEDIA: Reading vendor.tegra.display-size : status: 6
NvMMLiteBlockCreate : Block : BlockType = 277
Caught SIGSEGV
#0 0x0000007f9bcd6e28 in __GI___poll (fds=0x558019a090, nfds=548075881352, timeout=) at …/sysdeps/unix/sysv/linux/poll.c:41
#1 0x0000007f9bde3f58 in () at /usr/lib/aarch64-linux-gnu/libglib-2.0.so.0
#2 0x000000557f8d1310 in ()
Spinning. Please run ‘gdb gst-launch-1.0 8329’ to continue debugging, Ctrl-C to quit, or Ctrl-\ to dump core.

Backtrace:
https://www.hobrasoft.cz/files/backtrace.txt

System info:

/etc/nv_tegra_release
R32 (release), REVISION: 5.1, GCID: 26202423, BOARD: t210ref, EABI: aarch64, DATE: Fri Feb 19 16:45:52 UTC 2021

  • NVIDIA Jetson Nano (Developer Kit Version)
    • Jetpack 4.5.1 [L4T 32.5.1]
    • NV Power Mode: MAXN - Type: 0
  • Libraries:
    • CUDA: 10.2.89
    • cuDNN: 8.0.0.180
    • TensorRT: 7.1.3.0
    • Visionworks: 1.6.0.501
    • OpenCV: 4.1.1 compiled CUDA: NO
    • VPI: ii libnvvpi1 1.0.15 arm64 NVIDIA Vision Programming Interface library
    • Deepstream: 5.1.0-1

The same behaviour with Jetpak 4.4.

nvvidconv is not deepstream plugin, it can not be used with deepstream pipeline. DeepStream SDK FAQ - Intelligent Video Analytics / DeepStream SDK - NVIDIA Developer Forums