Issue in installing v3.0 in NERSC Perlmutter CPU nodes

[biqar]

Describe the bug

I have tried to install Opencilk v3.0 in NERSC Perlmutter CPU nodes and getting the following error:

/usr/bin/ranlib: ../libLLVMDemangle.a: error reading MicrosoftDemangleNodes.cpp.o: file truncated
/usr/bin/ranlib: ../libLLVMDemangle.a: error reading MicrosoftDemangle.cpp.o: file truncated
gmake[6]: *** [lib/Demangle/CMakeFiles/LLVMDemangle.dir/build.make:178: lib/libLLVMDemangle.a] Error 1
gmake[6]: *** Deleting file 'lib/libLLVMDemangle.a'
gmake[5]: *** [CMakeFiles/Makefile2:10898: lib/Demangle/CMakeFiles/LLVMDemangle.dir/all] Error 2
gmake[4]: *** [CMakeFiles/Makefile2:11202: utils/TableGen/CMakeFiles/llvm-min-tblgen.dir/rule] Error 2
gmake[3]: *** [Makefile:501: llvm-min-tblgen] Error 2
gmake[2]: *** [utils/TableGen/CMakeFiles/llvm-min-tblgen-host.dir/build.make:73: NATIVE/bin/llvm-min-tblgen] Error 2
gmake[1]: *** [CMakeFiles/Makefile2:16769: utils/TableGen/CMakeFiles/llvm-min-tblgen-host.dir/all] Error 2
gmake[1]: *** Waiting for unfinished jobs....
gmake[6]: *** [lib/Demangle/CMakeFiles/LLVMDemangle.dir/build.make:178: lib/libLLVMDemangle.a] Error 1
gmake[5]: *** [CMakeFiles/Makefile2:10898: lib/Demangle/CMakeFiles/LLVMDemangle.dir/all] Error 2
gmake[4]: *** [CMakeFiles/Makefile2:32007: tools/clang/utils/TableGen/CMakeFiles/clang-tblgen.dir/rule] Error 2
gmake[3]: *** [Makefile:9601: clang-tblgen] Error 2
gmake[2]: *** [tools/clang/utils/TableGen/CMakeFiles/clang-tblgen-host.dir/build.make:73: NATIVE/bin/clang-tblgen] Error 2
gmake[1]: *** [CMakeFiles/Makefile2:40657: tools/clang/utils/TableGen/CMakeFiles/clang-tblgen-host.dir/all] Error 2

Expected behavior

Successfully installing Opencilk v3.0

OpenCilk version

Built from source:

• opencilk-project: tag "opencilk/v3.0"
• infrastructure: commit "e5cd3ca3fd88295e5c7b53f2e165e458508cb09c", tag "opencilk/v3.0"

System information

• OS: SUSE Linux Enterprise Server 15 SP5
• CPU: AMD EPYC 7713 64-Core Processor

Steps to reproduce (include relevant output)

> git clone https://github.com/OpenCilk/infrastructure
> infrastructure/tools/get -t opencilk/v3.0 $(pwd)/opencilk
> infrastructure/tools/build $(pwd)/opencilk $(pwd)/build

Here you will able to find the full installation log

[VoxSciurorum]

The "file truncated" warning suggests either insufficient disk space or a race condition in a parallel build. If you have sufficient disk space, does a serial build work? Add a third argument "1" to the tools/build command to build with only one job.

[biqar]

@VoxSciurorum The disk have enough space.

I tried with lower parallel threads (e.g., 64) and still getting the same issue. Single thread is too slow for interactive jobs. I will try to submit it with a longer time limit.

I also tried to pull out the latest commits (beyond "opencilk/v3.0" tag) and got the following error:

[  0%] Built target LLVMSupportBlake3
[  0%] Built target LLVMSupportBlake3
[  0%] Built target LLVMSupportBlake3
[  0%] Built target obj.clangSupport
[  0%] Linking CXX static library ../libLLVMDemangle.a
[  0%] Linking CXX static library ../libLLVMDemangle.a
/usr/bin/ranlib: unable to copy file '../libLLVMDemangle.a'; reason: No such file or directory
gmake[6]: *** [lib/Demangle/CMakeFiles/LLVMDemangle.dir/build.make:178: lib/libLLVMDemangle.a] Error 1
gmake[5]: *** [CMakeFiles/Makefile2:10898: lib/Demangle/CMakeFiles/LLVMDemangle.dir/all] Error 2
gmake[5]: *** Waiting for unfinished jobs....
[  0%] Linking CXX static library ../libLLVMDemangle.a
[  0%] Built target obj.LLVMTableGenBasic
gmake[4]: *** [CMakeFiles/Makefile2:11202: utils/TableGen/CMakeFiles/llvm-min-tblgen.dir/rule] Error 2
gmake[3]: *** [Makefile:501: llvm-min-tblgen] Error 2
gmake[2]: *** [utils/TableGen/CMakeFiles/llvm-min-tblgen-host.dir/build.make:73: NATIVE/bin/llvm-min-tblgen] Error 2
gmake[1]: *** [CMakeFiles/Makefile2:16769: utils/TableGen/CMakeFiles/llvm-min-tblgen-host.dir/all] Error 2
gmake[1]: *** Waiting for unfinished jobs....
Error running link command: Bus error
gmake[6]: *** [lib/Demangle/CMakeFiles/LLVMDemangle.dir/build.make:178: lib/libLLVMDemangle.a] Error 1
gmake[6]: *** Deleting file 'lib/libLLVMDemangle.a'
gmake[5]: *** [CMakeFiles/Makefile2:10898: lib/Demangle/CMakeFiles/LLVMDemangle.dir/all] Error 2
gmake[4]: *** [CMakeFiles/Makefile2:32007: tools/clang/utils/TableGen/CMakeFiles/clang-tblgen.dir/rule] Error 2
gmake[3]: *** [Makefile:9601: clang-tblgen] Error 2
gmake[2]: *** [tools/clang/utils/TableGen/CMakeFiles/clang-tblgen-host.dir/build.make:73: NATIVE/bin/clang-tblgen] Error 2
gmake[1]: *** [CMakeFiles/Makefile2:40657: tools/clang/utils/TableGen/CMakeFiles/clang-tblgen-host.dir/all] Error 2
/usr/bin/ranlib: '../libLLVMDemangle.a': No such file
gmake[6]: *** [lib/Demangle/CMakeFiles/LLVMDemangle.dir/build.make:178: lib/libLLVMDemangle.a] Error 1
gmake[5]: *** [CMakeFiles/Makefile2:10898: lib/Demangle/CMakeFiles/LLVMDemangle.dir/all] Error 2
gmake[4]: *** [CMakeFiles/Makefile2:50490: tools/clang/tools/extra/clang-tidy/misc/ConfusableTable/CMakeFiles/clang-tidy-confusable-chars-gen.dir/rule] Error 2
gmake[3]: *** [Makefile:15932: clang-tidy-confusable-chars-gen] Error 2
gmake[2]: *** [tools/clang/tools/extra/clang-tidy/misc/CMakeFiles/genconfusable.dir/build.make:78: NATIVE/bin/clang-tidy-confusable-chars-gen] Error 2
gmake[1]: *** [CMakeFiles/Makefile2:59165: tools/clang/tools/extra/clang-tidy/misc/CMakeFiles/genconfusable.dir/all] Error 2
gmake: *** [Makefile:156: all] Error 2

Any comment on this?

[neboat]

Are you building OpenCilk on a parallel file system? I'm wondering if that might be creating unusual race conditions or other problems in the build process. If you're not already, you might try building OpenCilk on local storage on the compute node and then installing OpenCilk into a directory on the parallel file system.