Rebuild for CUDA 11.8 w/arch support with additional fix for out of space issue

[traversaro]

Checklist

• [x] Used a personal fork of the feedstock to propose changes
• [x] Bumped the build number (if the version is unchanged)
• [ ] Reset the build number to 0 (if the version changed)
• [x] Re-rendered with the latest conda-smithy (Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• [x] Ensured the license file is being packaged.

[conda-forge-webservices[bot]]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[traversaro]

@conda-forge-admin, please rerender

[traversaro]

Now it fails with:

2023-09-28T14:31:02.8474028Z clang.exe -DHAVE_CONFIG_H -I. -I../../include/private/autogen -I../../include/hwloc/autogen  -I"/d/bld/libhwloc_1695910688784/work"/include -I"/d/bld/libhwloc_1695910688784/work"/include -I../../utils/hwloc  -DLSTOPO_HAVE_GRAPHICS -ID:/bld/libhwloc_1695910688784/_h_env/Library/include -O2 -D_CRT_SECURE_NO_WARNINGS -D_MT -D_DLL -nostdlib -Xclang --dependent-lib=msvcrt -fuse-ld=lld -Dputenv=_putenv -Dmktemp=_mktemp -Dopen=_open -Dunlink=_unlink -Dclose=_close -Dstrdup=_strdup  -mwindows -ID:/bld/libhwloc_1695910688784/_h_env/Library/include -O2 -D_CRT_SECURE_NO_WARNINGS -D_MT -D_DLL -nostdlib -Xclang --dependent-lib=msvcrt -fuse-ld=lld -c -o lstopo_win-lstopo.o `test -f 'lstopo.c' || echo './'`lstopo.c
2023-09-28T14:31:03.2085724Z clang: error: unsupported option '-mwindows' for target 'x86_64-pc-windows-msvc'

An error that was not there in https://github.com/conda-forge/libhwloc-feedstock/pull/67 .

[traversaro]

clang: error: unsupported option '-mwindows'

Probably this is due to the new clang 17: https://github.com/llvm/llvm-project/issues/64464 .

[jakirkham]

Looks like that should be fixed in Clang 17.0.2: https://github.com/llvm/llvm-project-release-prs/commit/f6cf58eed9737cb8718465c4fb6917ca854b1bc3

There's already a feedstock update PR: https://github.com/conda-forge/clangdev-feedstock/pull/245

[traversaro]

@conda-forge-admin, please rerender

[traversaro]

Looks like that should be fixed in Clang 17.0.2: llvm/llvm-project-release-prs@f6cf58e

There's already a feedstock update PR: conda-forge/clangdev-feedstock#245

Perfect, thanks! I removed the workaround.

[jakirkham]

Hmm...I guess since we are still on 16 (and presumably that fix hasn't been backported), we still need a workaround

Edit: From the previous CI build, have attached the log for posterity.

[traversaro]

Hmm...I guess since we are still on 16 (and presumably that fix hasn't been backported), we still need a workaround

But the workaround was to use clang 16, see https://github.com/conda-forge/libhwloc-feedstock/pull/69/commits/6173f7c2fc1739c14bec931017b563981b5f37a7 . Anyhow, it was working, let's try to put it back.

[traversaro]

Hmm...I guess since we are still on 16 (and presumably that fix hasn't been backported), we still need a workaround

Actually the CI is installing 17, and somehow there is the -mwindows failure also on clang 17.0.2 . I don't think we are fixing clang on Windows? Anyhow, fixing clang to 16 was working, let's try again that.

[traversaro]

@conda-forge-admin, please rerender

[github-actions[bot]]

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you, but it looks like there was nothing to do.

This message was generated by GitHub actions workflow run https://github.com/conda-forge/libhwloc-feedstock/actions/runs/6566680601.

[traversaro]

Probably this is due to the new clang 17: https://github.com/llvm/llvm-project/issues/64464 .

Looking more at that, I am not sure if that is related, as that seems to be related to llvm with the MinGW toolchain (i.e. the one used in https://github.com/mstorsjo/llvm-mingw, whose mantainer is indeed the author of https://github.com/llvm/llvm-project-release-prs/commit/f6cf58eed9737cb8718465c4fb6917ca854b1bc3), while here we use the MSVC toolchain. I am not sure if we should open an issue in llvm for our problem.

[jakirkham]

Think that is unrelated. Our clang builds happen here, which just uses the LLVM sources

It looks like we were pulling in Clang 17 before when it was unpinned on CI

    clang:                 17.0.2-h3dc180e_0         conda-forge
    clang-17:              17.0.2-default_heb8d277_0 conda-forge
    clangxx:               17.0.2-default_heb8d277_0 conda-forge

Am now wondering why that wasn't working 🤔

In any event it should be ok to use Clang 16. Just a bit confused by the issue is all

[traversaro]

Think that is unrelated. Our clang builds happen here, which just uses the LLVM sources

Not sure what you were referring with this sentence.

Am now wondering why that wasn't working 🤔

I think there is an issue similar to https://github.com/llvm/llvm-project/issues/64464, but related to the MSVC toolchain, so it is not affected by the fix in https://github.com/llvm/llvm-project-release-prs/commit/f6cf58eed9737cb8718465c4fb6917ca854b1bc3, that is mingw toolchain specific. However, the problem could be on the libhwloc side, as -mwindows may be a flag that does not make sense for the msvc toolchain of clang, and just by chance this was working on clang <= 16 .

[jakirkham]

Maybe I'm misunderstanding something, but it seems like there are a bunch of MinGW components pulled into this build. So thought we might be trying to build with MinGW (hence the MinGW bug looked appropriate). Again I might be misunderstanding things

[jakirkham]

Filed issue ( https://github.com/llvm/llvm-project/issues/69964 ) for the MSVC case

[jakirkham]

From upstream:

However, the problem could be on the libhwloc side, as -mwindows may be a flag that does not make sense for the msvc toolchain of clang, and just by chance this was working on clang <= 16 .

That's exactly the case.

Sounds like a libhwloc issue then. Can see they add -mwindows on this line. Maybe it needs a guard to check whether MinGW or MSVC is used

Edit: Filed upstream issue ( https://github.com/open-mpi/hwloc/issues/631 )

[traversaro]

Thanks for digging on this @jakirkham !

[jakirkham]

@traversaro ofc! Feel free to chime in on issue ( https://github.com/llvm/llvm-project/issues/69964 ) if you have more info that would help

Sounds like we can merge this and follow up on anything else after ( cc @conda-forge/libhwloc )

[RaulPPelaez]

Hi guys, thanks for the good work! could we merge this?

[jakirkham]

Thanks all! 🙏