Failed to launch debugger in Eclipse

I am working with Eclipse 2022 and CUDA-11.6 plugin and I am able to run a simple code. However, when I start the debugger, I get this error:

Error in services launch sequence
Timed out trying to launch GDB.

An internal error occurred during: "Start GDB Process Job".

'java.lang.String[] com.nvidia.cuda.debug.ui.launch.CudaBackend.getDebuggerCommandLine()'

Maybe the cuda-gdb has not been properly defined although I have installed the plugin.
In the following setting page, when I click “Make cuda-gdb and NVIDIA profiler as default launchers”, nothing happens (no feedback). So, I assume the cuda-gdb is correctly attached to Eclipse.

Any idea about that?

Same problem here. I’m doing a proof of concept trying to run Eclipse in WSL2 - so a brand new install. Also have moved from the old integrated NSight Eclipse to the newer “standard Eclipse with plugins”.

I’m wondering if it is an issue with the latest build of Eclipse?

I’ve checked the directory for cuda-gdb, and that is correct. Initially I was missing this directory from my path.

Watching for replies :-)

1 Like

The CUDA debugger class com.nvidia.cuda.debug.ui.launch.CudaBackend is a child class of org.eclipse.cdt.dsf.gdb.service.GDBBackend that uses a deprecated method ‘getDebuggerCommandLine()’. This method was removed in Eclipse 2022-12 (“Eliminate deprecated GDBBackend methods for CDT 11”: Eliminate deprecated GDBBackend methods for CDT 11 (#112) · eclipse-cdt/cdt@1590791 · GitHub).

Therefore, the latest usable version for the Nsight plugin is Eclipse 2022-09 so far (CUDA 12.1).

1 Like

Thanks for the detailed bug report, Arnis. We’ve filed a defect against our plugins to address the use of the now-dropped getDebuggerCommandLine() method.

Hi! Is there a recommended version for CUDA 12.2? I’m running into the same problem with Eclipse IDE + Nsight Plugin in WSL2 on Windows 11. My Eclipse IDE version is Version: 2023-06 (4.28.0), Build id: 20230608-1333.

CUDA 12.2 was validated with Eclipse 4.16, 4.19, 4.24 (2022-06) and 4.25 (2022-09). We were not able to fix the above bug in time for the 12.2 release, but we have fixed the issue and validated 4.26/2022-12 with our to-be-released 12.3 plugins.

We are in the process of validating 4.27/2023-03 and 4.28/2023-06, and expect to be able to announce support for those versions in CUDA 12.3.

Hi, I struggled with this trouble as well when I updated to Eclipse Version: 2023-09 (4.29.0) and had Cuda 12.1 installed. But after Cuda update to version 12.3 this problem disappeared and seems to be resolved.


Thanks for letting us know it works now !

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.