transform_inclusive_scan+counting_iterator="error : calling a host function" in CUDA 4.1 RC + Nsight 2.1

Please post a short, self-contained code sample which reproduces the
problem:

#include ...

int main(){
    thrust::counting_iterator<int> cit(0);
    thrust::device_vector<int> v(10, 0);

    thrust::transform_inclusive_scan(cit, cit+10, v.begin(),
        thrust::negate<int>(), thrust::plus<int>());

    exit(EXIT_SUCCESS);
}

What is the expected output? What do you see instead?

no error expected, but I get

error : calling a host function("thrust::transform_iterator<thrust::negate<int> 
, thrust::counting_iterator<int, thrust::use_default, thrust::use_default, 
thrust::use_default> , int, thrust::use_default> ::transform_iterator") from a 
__device__/__global__ 
function("thrust::detail::backend::cuda::detail::fast_scan::upsweep_intervals_cl
osure<thrust::transform_iterator<thrust::negate<int> , 
thrust::counting_iterator<int, thrust::use_default, thrust::use_default, 
thrust::use_default> , int, thrust::use_default> , int, thrust::plus<int> , 
thrust::detail::backend::uniform_decomposition<unsigned int> , 
thrust::detail::backend::cuda::detail::statically_blocked_thread_array<(unsigned
 int)224u> > ::operator ()") is not allowed

and 7 more similar errors.

What version of Thrust are you using? Which version of nvcc?  Which host
compiler?  On what operating system?  Which GPU?

thrust rev. 1571, nvcc 4.1 NSIGHT, MSVC 2010, Windows 7, GTX 480

Please provide any additional information below.

This code snippet compiles normally with nvcc 4.1 but without Nsight 2.1 
installed, as well as with nvcc 4.0 and Nsight 2.1

Original issue reported on code.google.com by JKoma...@gmail.com on 29 Nov 2011 at 3:53

lion03 / thrust

transform_inclusive_scan+counting_iterator="error : calling a host function" in CUDA 4.1 RC + Nsight 2.1 #416