ldionne wrote: As a side effect, this new way of linking against libc++ instead of passing `-L` also avoids tripping the issue with the compiler-rt build. Indeed, since compiler-rt forcibly passes `-lc++` in its build (weird), Clang ends up passing both `-lc++` and `<toolchain>/lib/libc++.dylib` to the linker. However, since `<toolchain>/lib/libc++.dylib` doesn't contain the right architecture, it gets ignored by the linker and we fall back on the explicitly-specified `-lc++` by compiler-rt.
The compiler-rt build is very wrong for doing that, but if we can avoid having to fix it to unblock this fix, that's a good thing. https://github.com/llvm/llvm-project/pull/170303 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
