Nsight Systems not generating detailed traces

I am trying to get started using Nsight System. I am using AGX Xavier with L4T R32.1.
I have JetPack 4.2 and Nsight System 2019.3.2.

Profiling an NVIDIA example (e.g. boxFilter) seems to work.

When I try to profile my own application (Qt5 based, attached to existing process), I get the following error:

RuntimeError (120) {
ErrorText: Throw location unknown (consider using BOOST_THROW_EXCEPTION)
Dynamic exception type: boost::exception_detail::clone_implQuadDDaemon::ConvertEventError
std::exception::what: ConvertEventError
[QuadDDaemon::tag_error_code*] = 55

}

It still does profiling, but I am missing lost of data such as the stack trace list and the CPU load is shown empty.
What does the error above mean? How can I fix it?

Sorry for the delay. I’m more on the x86 side, so I have pinged a Tegra engineer to respond to this.

Hello,
We are looking at the problem.

Hi,

mandre,
Yes, it is reproduced on 2019.3.2.XXX releases with some QT5 applications.
This issue was already fixed in the latest releases (verified on 2019.4.1.10 build).
Please update Nsight Systems.

Thanks,
Igor.

Hey, was there a solution for this error? I`m experiencing the same problem with Nsight 2020.5.3 remotely connected to Jetson Xavier via Docker

@Andrey_Trachenko