[llvm-dev] building llvm_Rel400 on Scientific Linux (RHEL) 7.3 x86_64

Tim Northover via llvm-dev llvm-dev at lists.llvm.org
Sat Jun 3 16:04:57 PDT 2017

Hi Adam,

On 3 June 2017 at 14:29, Adam Jensen via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> I am trying to build the LLVM suite on a RedHat Enterprise Linux clone (Scientific Linux <https://www.scientificlinux.org/>). In the latest attempt, the build seems to complete without any explicit failures but the `check-all` process fails. Any ideas about what is wrong or suggestions for how to proceed would be much appreciated.

I can't reproduce this exact failure (my build dies earlier), but I
think what's happening is that you're finding a libstdc++.a from the
GCC toolchain (i.e. a static rather than dynamic library).

For static libraries the order of objects on the command-line matters
libclang_rt.xray-fdr-x86_64.a is after libstdc++, so libstdc++ isn't
examined again and the link step fails.

I think you should be able to fix it by changing the
compiler-rt/lib/xray/test/CMakeLists.txt. If you find the
"add_compiler_rt_test" call and move "-lstdc++" to the end, just after
"-lrt" it should work.

I'm not sure if this is the real fix though we'd want to adopt into
compiler-rt though.



More information about the llvm-dev mailing list