rsmith added a comment. I have to admit to not understanding the motivation for this change. It claims that this is removing a reliance on having an existing C++ standard library, but it doesn't appear to affect that: this test explicitly passes `-nostdinc++` to avoid using a system standard library, and instead adds the (currently-being-configured) libc++ headers to the include path.
If that approach is a problem somehow, how about this alternative: instead of indirectly trying to get the compiler to emit a use of a function provided by libatomic, we create a (C++) compilation test that declares a function from libatomic and calls it. That should tell us whether the C++ compiler implicitly links in libatomic or a suitable alternative. https://reviews.llvm.org/D23719 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
