Artem Belevich

If you can reproduce the issue on [godbolt.org](https://cuda.godbolt.org/z/qY9h9b9Gv), that would be very helpful.

Was there anything else printed out by the compiler before the stack trace? E.g. any assertion failures? The easier it is to reproduce the issue the higher the chance that...

Yup. I've got a moving captcha on my login page now. No more live data use tracking. :-(  So long and thanks for all the fish.

Looks like the issue is more general and may not be limited to implib's use from a shared library. I've updated the title and description

The wrappers are part of clang installation. They should be in the same directory where clang finds its other header files. https://github.com/llvm/llvm-project/tree/main/clang/lib/Headers

One way to find the path is to run `clang -v -x c++ /dev/null -fsyntax-only` and check the list of include paths. The path to clang's will be towards the...

See the details in the body of the bug report: > AFAICT, there's no reliable enough way to reliably control placement of cuda_wrappers in the search path from command line....

> It looks like this was caused by a recent change in libstdc++ here: [gcc-mirror/gcc@dbf8bd3#diff-b358f609a31a4af8af72cc3197566abaa157bb7f8681b45580f1e5477540457cR192-R193](https://github.com/gcc-mirror/gcc/commit/dbf8bd3c2f2cd2d27ca4f0fe379bd9490273c6d7#diff- > However, this issue is unique to `clang` as `nvcc` compiles the equivalent just fine:...

Right. The `__attribute__((__attribute__((noinline)))) void foo();` gets magically transformed into `__attribute((noinline)) void foo()` by the time it makes it to the final host compilation. 😭 And the magic seems to work...

It's a fairly recent development. I guess it was introduced with a libstdc++ upgrade on godbolt machines. All you need to reproduce it is to compile the default CUDA example...