Error while installing DriveOS 5.2.0 from SDK Manager 1.6.1

Please provide the following info (check/uncheck the boxes after creating this topic):
Software Version
DRIVE OS Linux 5.2.6
[*] DRIVE OS Linux 5.2.0
DRIVE OS Linux 5.2.0 and DriveWorks 3.5
NVIDIA DRIVE™ Software 10.0 (Linux)
NVIDIA DRIVE™ Software 9.0 (Linux)
other DRIVE OS version
other

Target Operating System
[*] Linux
QNX
other

Hardware Platform
NVIDIA DRIVE™ AGX Xavier DevKit (E3550)
NVIDIA DRIVE™ AGX Pegasus DevKit (E3550)
other

SDK Manager Version
[*] 1.6.1.8175
1.6.0.8170
other

Host Machine Version
native Ubuntu 18.04
other

Hi All - I see this error -

11:37:11 INFO: NVIDIA Nsight Systems - host: Preparing to unpack .../13-libcusparse-dev-10-2_10.3.1.311-1_amd64.deb ...

11:37:11 INFO: NVIDIA Nsight Systems - host: Unpacking libcusparse-dev-10-2 (10.3.1.311-1) ...

11:37:16 ERROR: NVIDIA Nsight Systems - host: No apport report written because MaxReports is reached already

11:37:16 INFO: NVIDIA Nsight Systems - host: Preparing to unpack .../14-libnpp-10-2_10.2.1.311-1_amd64.deb ...

11:37:16 INFO: NVIDIA Nsight Systems - host: Unpacking libnpp-10-2 (10.2.1.311-1) ...

11:37:21 INFO: NVIDIA Nsight Systems - host: trying to overwrite '/usr/local/cuda-10.2/lib64', which is also in package cuda-cudart-dev-10-2 10.2.311-1

11:37:21 INFO: NVIDIA Nsight Systems - host: Preparing to unpack .../15-libnpp-dev-10-2_10.2.1.311-1_amd64.deb ...

11:37:21 INFO: NVIDIA Nsight Systems - host: Unpacking libnpp-dev-10-2 (10.2.1.311-1) ...

11:37:26 ERROR: NVIDIA Nsight Systems - host: No apport report written because MaxReports is reached already

11:37:26 INFO: NVIDIA Nsight Systems - host: Preparing to unpack .../16-libnvjpeg-10-2_10.3.1.311-1_amd64.deb ...

11:37:26 INFO: NVIDIA Nsight Systems - host: Unpacking libnvjpeg-10-2 (10.3.1.311-1) ...

11:37:26 ERROR: NVIDIA Nsight Systems - host: No apport report written because MaxReports is reached already

11:37:26 INFO: NVIDIA Nsight Systems - host: Preparing to unpack .../17-libnvjpeg-dev-10-2_10.3.1.311-1_amd64.deb ...

11:37:26 INFO: NVIDIA Nsight Systems - host: Unpacking libnvjpeg-dev-10-2 (10.3.1.311-1) ...

11:37:26 ERROR: NVIDIA Nsight Systems - host: No apport report written because MaxReports is reached already

11:37:26 INFO: NVIDIA Nsight Systems - host: Preparing to unpack .../18-cuda-nvml-dev-10-2_10.2.311-1_amd64.deb ...

11:37:26 INFO: NVIDIA Nsight Systems - host: Unpacking cuda-nvml-dev-10-2 (10.2.311-1) ...

11:37:26 ERROR: NVIDIA Nsight Systems - host: No apport report written because MaxReports is reached already

11:37:26 INFO: NVIDIA Nsight Systems - host: /tmp/apt-dpkg-install-yZsSue/18-cuda-nvml-dev-10-2_10.2.311-1_amd64.deb

11:37:27 ERROR: NVIDIA Nsight Systems - host: command terminated with error

11:37:27 SUMMARY: NVIDIA Nsight Systems - host: APT system is broken and requires manual fix

What could be the reason for errors like -

ERROR Cuda on Host - host: No apport report written because MaxReports is reached already

Thanks

Hi, @ai.trucker

Please compress and upload ~/.nvsdkm on your host system for our investigation. Thanks.

Please also see if Installation failing on Host helps.

As hinted in the link you gave, I had some apt stability issue. I wiped out everything (i.e. uninstalled prev installations from sdk manager, reinstalled sdk manager again) and then tried to install 5.2.0 again. Now I see following, which still seems related to apt…

15:28:56 ERROR: CUDA Cross-Compile for Linux - host: E: Packages were downgraded and -y was used without --allow-downgrades.

15:28:56 ERROR: CUDA Cross-Compile for Linux - host: command terminated with error

15:28:56 SUMMARY: TensorRT Cross-Compile for Linux - host: Depends on failed component

Please find .nvsdkm data attached.

Thanks,nvsdkm-9-30.tar.gz (334.7 KB)

Please try to clarify the issue with the below bold command. Maybe you need to remove the installed version first.

15:28:56.018 - error: NV_CUDA_HOST_CROSS_LINUX_COMP@host: [error]: Error when apt install failed; [exec_command]: sudo apt-get update; sleep 0.5; sudo apt-get -y install cuda-cross-aarch64-10-2=10.2.187* ; [error]: exit status 100; [deb_path]: /home/pankaj.chaurasia/Downloads/nvidia/sdkm_downloads/cuda-repo-cross-aarch64-10-2-local-10.2.187_1.0-1_all.deb

sudo apt-get -y install cuda-cross-aarch64-10-2=10.2.187
Reading package lists... Done
Building dependency tree       
Reading state information... Done
E: Version '10.2.187' for 'cuda-cross-aarch64-10-2' was not found

I did the uninstall for all components from SDK Manager (repair/uninstall tab).

Thanks

Now please try to install with SDK Manager again. Thanks.

I did a shut down of SDK manger . Ran it again and tried installing 5.2.0 - I still see same issue.
Thanks

17:02:08 ERROR: CUDA Cross-Compile for Linux - host: E: Packages were downgraded and -y was used without --allow-downgrades.

17:02:08 ERROR: CUDA Cross-Compile for Linux - host: command terminated with error

17:02:08 SUMMARY: TensorRT Cross-Compile for Linux - host: Depends on failed component

Please try if uninstalling the relevant components of DRIVE OS 5.2.6 on your host system first helps.

Hi VickNV,
I have already uninstalled 5.2.6 from sdk manager. That part went well. Still if I can remove components from cmd line - please let me know what are those components ? I am looking for more of a grep text string which could be used with "find " based cmd to find all components and remove them ?
Thanks

Because it failed to downgrade, I suspect you have already installed cuda-cross-aarch64-10-2 of DRIVE OS 5.2.6 (with high version).
After 5.2.6 was uninstalled, do you still see the same error during the 5.2.0 installation?

Yes I am running installation of 5.2.0 AFTER 5.2.6 was uninstalled from sdk manager.

Please try to remove some packages with the below command prior to installation with sdkmanager. Thanks.

$ sudo apt purge "cuda*" "*cufft*" "*curand*" "*npp*" "*cusparse*"

@ai.trucker Does it fix your issue?

VickNV,
Yes it does. Thanks for your help. Could this be part of repair / uninstall button ?
Thanks