Nikita Shulga

> Why is only the x86 macos build failing? Not the arm one and not the linux build? Don't we build linux with WERROR as well? We don't/can't build tests...

MacOS/iOS builds were the sources of frequent reverts in last few weeks anyway, so we either have to enable it on PRs and/or update our Linux clang configs clang-13

@pytorchbot merge -f "MacOS builds and Run MPS is green"

Thank you for opening this! My 5c against using `glog` as a unifying logging facility: same as `protobuf`, it is extremely hard to ship without exposing some of the implementation...

Hmm, it is run with 6 cores: ``` cd /var/lib/jenkins/workspace/third_party/nccl/nccl && env CCACHE_DISABLE=1 SCCACHE_DISABLE=1 make CXX=/opt/cache/bin/c++ CUDA_HOME=/usr/local/cuda NVCC=/usr/local/cuda/bin/nvcc NVCC_GENCODE=-gencode=arch=compute_52,code=sm_52 BUILDDIR=/var/lib/jenkins/cpp-build/caffe2/build/nccl VERBOSE=0 -j 6 ```

Looks like a linker compatibility problem (i.e. when one c++ runtime does not know how to talk to another one or how to parse unwind instructions)

Though it works for me on Ubuntu-18.04, by running the following commands: ``` $ conda create -n py38-torch112-cpu python=3.8 $ conda activate py38-torch112-cpu $ python3 -mpip install --pre torch==1.12 -f...

Just to clarify: ``` $ c++filt _ZNSt19basic_ostringstreamIcSt11char_traitsIcESaIcEEC1Ev std::basic_ostringstream::basic_ostringstream() ```

@nSircombe last time I've tried enabling OpenMP support in OpenBLAS and linking statically, I run into `can not allocate LTS` runtime issue, but let me try again.

Hi @nSircombe, I was about to write, that I can not reproduce the problem, but the reason was that pretty simple: when I've enabled OpenMP in OpenBLAS, pytorch started to...