search

luxonis / depthai-python

DepthAI Python Library
MIT License
359 stars 193 forks source link

rgb_preview.py crash my guest machine in vmware workstation #293

Open wureka opened 3 years ago

wureka commented 3 years ago

I know this could be the problem of vmware workstation, but I still report my situation.

Environment 1:

Environment 2:

Reproduce steps: git clone https://github.com/luxonis/depthai-python cd depthai-python python3 examples/install_requirements.py cd examples python3 rgb_preview.py

In Environment 1, It previews without any problem. But when I press 'q', my guest is crashed and is closed (poweroff). so I could not collect log nor error messages. In Environment 2, however, rgb_preview.py runs OK, Even when I press 'q', it stops normally.

I believe this is not the problem of rgb_preview.py. because I tried other python examples and same result.

Luxonis-Brandon commented 3 years ago

Thanks for reporting. Strange. I wonder if it's something to do with how the USB3 routing is done between the environment and Windows. The USB drivers are quite different between Windows and Ubuntu, and I'm wondering if there's some timing thing upon termination that doesn't go well.

One thing to try in Environment 1 would be to force USB2 mode when running the example, with Device(pipeline, usb2Mode=True). That would let us know if the issue is specific to USB3, or not.

wureka commented 3 years ago

@Luxonis-Brandon I just gave it a shot. I used both two case as below: with dai.Device(pipeline, True) as device:

with dai.Device(pipeline, False) as device:

Both above two all still crash my guest OS in Environment 1.

However, if I change the USB compatibility to 2.0 for my guest, and use with dai.Device(pipeline, True) as device:

And rgb_preview.py can sometime successfully stop without crashing my guest OS, and crashes sometime.

Luxonis-Brandon commented 3 years ago

Interesting. I think this is actually an underlying issue with how we are ending the pipeline. As a fellow using DepthAI in java saw the same (or similar) behavior in Java on a Pi (but not in Java on macOS).

Likely we will go away from XLINK completely, as it seems to have strange compatibility/stability issues depending on the OS.

In the meantime, we'll try to figure out what is going on here.

Thanks for the thorough reporting.