Eclipse: Edit toolchain (remove build step to create a shared library for CUDA) - c++

Question

I am trying to configure Eclipse such that it compiles a shared library in one project and uses it in another.
The problem is, that using the CUDA plugin for Eclipse one can only choose an executable generating project type.
So what I want to do is creating such a project and modify that toolchain such that Eclipse does not execute anything else than nvcc.
As you can see compiling the library is not a problem:
18:27:25 **** Incremental Build of configuration Default for project cudamath ****
make all
Building file: ../test.cu
Invoking: CUDA NVCC Compiler
nvcc --shared -Xcompiler -fPIC -o "cu_test.o" "../test.cu" && \
echo -n 'cu_test.d' ./ > 'cu_test.d' && \
nvcc -M "../test.cu" >> 'cu_test.d'
nvcc warning : The 'compute_10' and 'sm_10' architectures are deprecated, and may be removed in a future release.
nvcc warning : The 'compute_10' and 'sm_10' architectures are deprecated, and may be removed in a future release.
Finished building: ../test.cu
The problem is that Eclipse then calls g++ which is that step of the toolchain I want to cut off:
Building target: cudamath
Invoking: C++ Linker
g++ -L/opt/cuda/lib64 -o "cudamath" ./cu_test.o -lcuda -lcublas -lcudart
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.9.0/../../../../lib/crt1.o: In function `_start':
(.text+0x20): undefined reference to `main'
collect2: error: ld returned 1 exit status
makefile:32: recipe for target 'cudamath' failed
make: *** [cudamath] Error 1
Is there a way I can do this? I've been crawling through my project settings but I can't seem to find what I'm looking for.

Answer 1

Here's what I did using Nsight Eclipse Edition:
File... new CUDA C/C++ project
In the next dialog, select Shared Library...Empty Project, and give the project a name (let's say it is testlib)
Finish that wizard/dialog. A new testlib project is created in the project explorer on the left
In the project explorer on the left, right click on the project name and create a new folder for your source files
Open that folder and create your new source file. For this, I selected a CUDA C/C++ source file using the CUDA bitreverse "template". This creates a new source file with the bitreverse code in it.
change int main() { in your source file to int myfunc(){
save the source file and build the project. A new libtestlib.so is successfully built.