[BUG]: thrust/system/cuda/detail/util.h missing include header for std::terminate

[rnburn]

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this bug and that I agree to the Code of Conduct

Type of Bug

Compile-time Error

Component

CUB

Describe the bug

#include "cub/cub.cuh" fails with the error message:

In file included from external/local_cuda/cuda/include/cub/cub.cuh:54:
In file included from external/local_cuda/cuda/include/cub/device/device_adjacent_difference.cuh:33:
In file included from external/local_cuda/cuda/include/cub/device/dispatch/dispatch_adjacent_difference.cuh:30:
In file included from external/local_cuda/cuda/include/cub/agent/agent_adjacent_difference.cuh:37:
In file included from external/local_cuda/cuda/include/thrust/system/cuda/detail/core/util.h:32:
external/local_cuda/cuda/include/thrust/system/cuda/detail/util.h:203:29: error: no member named 'terminate' in namespace 'std'; did you mean simply 'terminate'?
  203 |   NV_IF_TARGET(NV_IS_HOST, (std::terminate();), (asm("trap;");));
      |                             ^~~~~
external/local_cuda/cuda/include/nv/detail/__target_macros:504:75: note: expanded from macro 'NV_IF_TARGET'
  504 | #  define NV_IF_TARGET(cond, ...)    _NV_BLOCK_EXPAND(_NV_TARGET_IF(cond, __VA_ARGS__))
      |                                                                           ^~~~~~~~~~~
external/local_cuda/cuda/include/nv/detail/__target_macros:467:74: note: expanded from macro '_NV_TARGET_IF'
  467 | #    define _NV_TARGET_IF(cond, t, ...) _NV_IF( _NV_ARCH_COND_CAT(cond), t, __VA_ARGS__)
      |                                                                          ^
external/local_cuda/cuda/include/nv/detail/__preprocessor:44:61: note: expanded from macro '_NV_IF'
   44 | #define _NV_IF(cond, t, f) _NV_IF1(_NV_EVAL(cond), _NV_EVAL(t), _NV_EVAL(f))
      |                                                             ^
note: (skipping 23 expansions in backtrace; use -fmacro-backtrace-limit=0 to see all)
external/local_cuda/cuda/include/nv/detail/__preprocessor:87:47: note: expanded from macro '_NV_STRIP_PAREN'
   87 | #define _NV_STRIP_PAREN(...) _NV_STRIP_PAREN1(__VA_ARGS__)
      |                                               ^~~~~~~~~~~
external/local_cuda/cuda/include/nv/detail/__preprocessor:86:48: note: expanded from macro '_NV_STRIP_PAREN1'
   86 | #define _NV_STRIP_PAREN1(...) _NV_STRIP_PAREN2 __VA_ARGS__
      |                                                ^~~~~~~~~~~
external/local_cuda/cuda/include/nv/detail/__preprocessor:85:31: note: expanded from macro '_NV_STRIP_PAREN2'
   85 | #define _NV_STRIP_PAREN2(...) __VA_ARGS__
      |                               ^~~~~~~~~~~
external/local_cuda/cuda/include/thrust/system/cuda/detail/util.h:201:1: note: 'terminate' declared here
  201 | terminate()
      | ^
1 error generated when compiling for host.
INFO: Elapsed time: 6.975s, Critical Path: 6.15s
INFO: 4 processes: 4 internal.
FAILED: Build did NOT complete successfully

I was able to workaround the issue by adding #include <exception> before #include "cub/cub.cuh"; it looks like the header file is missing an include for std::terminate.

Note: I'm using a fairly non-standard build environment (clang with libc++ instead of libstdc++).

Here is the file that produces the error: https://github.com/spaceandtimelabs/blitzar/blob/95823fcd5f2cae49bebcdf27da2cbf393ad3a7d1/sxt/algorithm/transform/prefix_scan.cc

How to Reproduce

You can reproduce the issue by following the build instructions here: https://github.com/spaceandtimelabs/blitzar/tree/95823fcd5f2cae49bebcdf27da2cbf393ad3a7d1?tab=readme-ov-file#usage
for the commit spaceandtimelabs/blitzar@95823fc

Expected behavior

No compilation error

Reproduction link

No response

Operating System

No response

nvidia-smi output

No response

NVCC version

nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Tue_Jun_13_19:16:58_PDT_2023
Cuda compilation tools, release 12.2, V12.2

[miscco]

@rnburn thanks a lot for bringing this up. We always appreciate feedback / bug reports especially for slightly nonstandard build environments.

In this case I am happy to inform you that this has already been fixed. This should already be in the 12.3 CTK.