Should the Nsight CUDA debugger work with OpenGL Interop?

I have a program that calls cudaGraphicsMapResources to allow access to OpenGL resources in a CUDA kernel. When I run “Start CUDA Debugging” from the NSIGHT menu in Visual Studio 2013 (Win32 or x64 targets) I get an exception:

Unhandled exception at 0x53742EF6 (nvcuda.dll) in <Program>.exe: 0xC0000005: Access violation reading location 0x00000018.

If I comment out all the cudaGraphicsMapResources (and subsequent interop calls) I can debug the CUDA kernel without problems.

Should I be able to run the Visual Studio CUDA Nsight debugger in the presence of these graphics interop calls? I expected to be able to because I can run cuda-gdb under linux on this program without problems.

For reference, I’m using the latest Nvidia driver for Windows 8.1 x64: 353.06, and CUDA 7.0. Is there a different driver I should be using?

There are some limitations in debugging graphics interop, for example:

http://stackoverflow.com/questions/26286753/nsight-debugging-with-opengl-interop

I recognize this is not a complete answer. You may get more helpful results by posting your question in the nsight visual studio edition forum:

https://devtalk.nvidia.com/default/board/84/nsight-visual-studio-edition/