kadircet added a comment. In D141467#4045190 <https://reviews.llvm.org/D141467#4045190>, @tra wrote:
> Given that it's indeed unused, I'm fine with removing it. > > That said, it's somewhat odd that in your setup clang was able to find > everything but the library directory. You generally would need to have > `--cuda-path=` among compiler options and that should be pointing to a valid > CUDA installation. > > Can you elaborate on where/how clang manages to find CUDA headers in your > case? We actually use a pipeline that runs compilations by recording & replaying past snapshots of our codebase. The record logic we have don't retain access to directories in its required inputs set, as clang stats lots of directories when performing include search etc, which are not ~never needed for the compilation (apart from the one that contains the actual files). Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D141467/new/ https://reviews.llvm.org/D141467 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
