Hi,
We are able to run the script on r32.1 + the lib. Could you check with the sensor vendor? Probably it can give some clues in checking the sensor driver.
You ran it with TWO cameras and you didn’t get an EOS after the second play? Can you please provide me with GST_DEBUG=4 logs?
EDIT: It’s not about just running the script - it’s about proper pipeline behavior.
Hi,
We modify the resolution to 2592x1944 and then run the script. Don’t see any error message in running it. Could you make the error print more noticeable like print(“!!!@@@ ERROR_MESSAGE !!!@@@”)
I can do this but honestly, that’s your job and/or the developers.
The bottom line is this: Why is EOS being sent on a pipeline restart? It really is that simple.
Pipeline starts in the NULL state.
Switches to the PLAYING state.
Switches to the NULL state on a manual EOS sent and caught
Switches back to the PLAYING state → immediate EOS is delivered to application, why??? Check my script, I don’t do that.
Use GST_DEBUG=4 and you can see the logs for yourself. Post them here please so we can look at them.
Btw, nvargus-daemon occasionally crashes randomly when two simultaneously streams are started at the same time. It also crashes immediately if the pipeline syntax is wrong. In both cases, the daemon has to be restarted to get output.
Another thing I noticed: With this patch, the frame rate is no longer being honored. I am recording at 60 fps when I explicitly wanted 30 fps?? (it was 30fps before patch)
I still see occasional segvs when two pipes are started at the same time, i.e. I think you still have a timing bug of some sort.
Is there an ETA when these fixes are going to get into the latest release?
r32.3.1 gst_element_set_state set to NULL sometimes hangs and looks like only nvargus-daemon is running
any debugging I can do for you?
Terry
I need to see if this exists in the latest release still. I hope this was all fixed by now because nvargus-daemon was royally broke for months.
Hi,
We don’t observe the issue in running the default camera Ov5693. If you use the camera module from our camera partners, please let us know the brand and model ID. We will check with the partners.
Leopard Imaging IMX-577. At least on R32.1 it hangs on stop immediately using more than one camera.
Also has any of the fixes in the patch above put into R32.2.1+?
Finally, did you see my blurb about frame rate not being honored?
This is a MAJOR roadblock in our deployment of Jetson.
Can you confirm that this issue is fixed in R32.2.1 or R32.3.1? @DaneLL
Hi,
We don’t observe the issue on r32.2.1 with default camera board(onboard camera is ov5693). Also checking with Leopard Imageing on updating IMX-577 driver to new version.
You need at least two cameras to observe this and you have to use gstreamer API.
The patch you posted previously fixes the stop issue but creates others (frame rate is no longer honored and nvargus-daemon still crashes sometimes).
Leopard doesn’t have a driver for R32.2.1+ which is why I’m kinda stuck right now.
NOTE: I sent you guys a script that demonstrates the problem with two cameras. Please take a look at it again once you have two cameras.
r32.3.1 gst_element_set_state set to NULL sometimes hangs and looks like only nvargus-daemon is running
any debugging I can do for you?
I am running a new lumenera r32.3.1 camera. System hangs when setting the state to NULL, before I kill the pipeline and try to bring it back up.
Hi terrysu50z,
For clearness, this thread is for R32.1. We have delivered a prebuilt lib in #32 that works well with default camera board. However, it does not work well with IMX577 and we have reported it to our camera partners.
We would suggest you make a new post. Share steps for reproducing the issue. Some issues are in Argus. Some issues are in camera driver. We would like to track these one by one.
hello terrysu50z,
we had recently confirm it’s a regression issue of l4t-r32.3.1,
since this discussion thread is focus-on r32.1, let’s moving to Topic 1070455 for further supports.
thanks
This is exactly the problem that was in R32.1. I suspect R32.3.1 has the same issue - nvargus-daemon hangs/segvs during pipeline state transition to NULL.
Note: Right now we are confused on what we can do because there seems to be a major issue in both pieces of software preventing our deployment.
Hi All
Please have a try the attached lib to fix the problem.
libgstnvarguscamerasrc_tx2_xaver.so.txt (81.5 KB)