Building pyds and ImportError

[l.weingart]

Please provide complete information as applicable to your setup.

• Hardware Platform: Jetson
• DeepStream Version: 6.2
• JetPack Version: 5.1
• TensorRT Version: 2.11.0
• Issue Type: questions

Hello,

I cloned the deepstream_python_apps repository and tried to build the pyds binding following the README here, using DS_VERSION=6.2, PYTHON_MAJOR_VERSION=3, PYTHON_MINOR_VERSION=8 and PIP_PLATFORM=linux_aarch64 as CMAKE options:
deepstream_python_apps/bindings at master · NVIDIA-AI-IOT/deepstream_python_apps · GitHub.

Then, when I run my own python code importing pyds, I get the following error:

Traceback (most recent call last):
  File "main.py", line 11, in <module>
    from pipeline import Pipeline
  File "/home/jetson/git/blimp/jetson-agx/pipeline.py", line 13, in <module>
    import pyds
ImportError: /lib/aarch64-linux-gnu/libGLdispatch.so.0: cannot allocate memory in static TLS block

However, running one of the deepstream_python_apps examples, it runs just fine.
I take it that I messed something up at some point.

Could someone help me understand the issue please?
Thank you in advance for your help

EDIT: It doesn’t solve the root cause, but there was an import pyds in my main.py file and another in the pipeline.py file. By importing pyds before importing the pipeline file, the issue disappeared. :-/

Thanks all

[yuweiw]

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

1. You should use: TensorRT Version:: TRT 8.5.2.2
2. Your code may import some other libraries that affect memory allocation. Could you attach your a minimum executable demo code?