[PATCH] D26649: [CMake] Support lld with LTO bootstrap

[Petr Hosek via cfe-commits]

phosek added a comment.

In https://reviews.llvm.org/D26649#595356, @mehdi_amini wrote:

> In https://reviews.llvm.org/D26649#595296, @phosek wrote:
>
> > It's sufficient, I just tested it.
>
>
> How did you check it? I don't understand how LLVM_ENABLE_LLD is propagated to stage-2?


Sufficient as in Clang looks for lld in the same directory where `clang`/`clang++` binary is first so we don't need to explicitly pass the path to lld to later stages.



================
Comment at: CMakeLists.txt:534
+        add_dependencies(LLVMgold)
+      endif()
       set(LTO_AR -DCMAKE_AR=${LLVM_RUNTIME_OUTPUT_INTDIR}/llvm-ar)
----------------
mehdi_amini wrote:
> What if not by the way? Should we error out here? What is the expected behavior?
I assume that's platform dependent; if some platforms (other than Darwin) use linker other than gold or lld for LTO build, erroring out here might break them. I don't think there are any such platforms though because they'd be broken already as we currently always try to use LLVMgold which causes CMake error in case we're not building with Binutils (that's how I discovered this issue, since we're using lld rather than gold on our platform).


Repository:
  rL LLVM

https://reviews.llvm.org/D26649