add recipe for finch-tensor

[lucascolley]

https://github.com/finch-tensor/finch-tensor

Checklist

• Title of this PR is meaningful: e.g. "Adding my_nifty_package", not "updated meta.yaml".
• License file is packaged (see here for an example).
• Source is from official source.
• Package does not vendor other packages. (If a package uses the source of another package, they should be separate packages or the licenses of all packages need to be packaged).
• If static libraries are linked in, the license of the static library is packaged.
• Package does not ship static libraries. If static libraries are needed, follow CFEP-18.
• Build number is 0.
• A tarball (url) rather than a repo (e.g. git_url) is used in your recipe (see here for more details).
• GitHub users listed in the maintainer section have posted a comment confirming they are willing to be listed there.
• When in trouble, please check our knowledge base documentation before pinging a team.

[github-actions]

Hi! This is the staged-recipes linter and your PR looks excellent! 🚀

[conda-forge-admin]

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 (recipes/finch-tensor/recipe.yaml) and found it was in an excellent condition.

[lucascolley]

@conda-forge/help-python-c ready for review, thanks!

[lucascolley]

FYI @mtsokol @willow-ahrens

[willow-ahrens]

thanks Lucas! do you need the other PR merged to merge this one?

[lucascolley]

thanks Lucas! do you need the other PR merged to merge this one?

nope I think this is ready and standalone! Just waiting for a review from @conda-forge/help-python-c.

[mtsokol]

@lucascolley Can you add Willow and me as recipe maintainers?

[traversaro]

Can you explain why this is not noarch: python and why it can't depend on c-compiler?

[lucascolley]

Can you explain why this is not noarch: python and why it can't depend on c-compiler?

There were some test failures on Windows, let me make that change now for demonstration

[traversaro]

If the package itself does not support Windows (that seems the case to me) you can also just add a run dependency on __unix to prevent it being installed on Windows, and skip the build. Windows support can always be added later in the feedstock.

[lucascolley]

and skip the build

this requires reverting the noarch: python, correct?

[lucascolley]

If the package itself does not support Windows (that seems the case to me) you can also just add a run dependency on __unix to prevent it being installed on Windows, and skip the build. Windows support can always be added later in the feedstock.

can you advise @willow-ahrens @mtsokol ? Indeed we are seeing some windows failures over in finch-tensor/finch-tensor#314

[willow-ahrens]

the windows failure is related to the optional c-compiler dep right? It sounds like we should make the c-compiler tests opt-in somehow so to reflect the fact that the default numba backend is cross-platform.

[lucascolley]

It sounds like we should make the c-compiler tests opt-in somehow

I'll mark this as draft until that is done

[willow-ahrens]

The c compiler tests are opt-in now. we have different test groups for with or without c

[lucascolley]

@willow-ahrens @mtsokol ready for a look

[lucascolley]

OSX failure for the Julia backend:

 │ self = Julia: Main, expr = 'using HDF5'
 │     def seval(self, expr):
 │ >       return self._jl_callmethod($(pyjl_methodnum(pyjlmodule_seval)), expr)
 │                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 │ E       juliacall.JuliaError: InitError: could not load library "/Users/runner/bld/test/test_finch-tensor4BG0Bn/test/.julia/artifacts/1b50039ea42f7ba809ca23f081b4dbdc358a4346/lib/libhdf5_fortran.320.1.1.dylib"
 │ E       dlopen(/Users/runner/bld/test/test_finch-tensor4BG0Bn/test/.julia/artifacts/1b50039ea42f7ba809ca23f081b4dbdc358a4346/lib/libhdf5_fortran.320.1.1.dylib, 0x0001): Library not loaded: libgcc_s.1.1.dylib
 │ E         Referenced from: <9C4BFE20-CC3C-3A5B-B7E4-0B98C12875A6> /Users/runner/bld/test/test_finch-tensor4BG0Bn/test/.julia/artifacts/1b50039ea42f7ba809ca23f081b4dbdc358a4346/lib/libhdf5_fortran.320.1.1.dylib
 │ E         Reason: tried: 'libgcc_s.1.1.dylib' (no such file), '/System/Volumes/Preboot/Cryptexes/OSlibgcc_s.1.1.dylib' (no such file), 'libgcc_s.1.1.dylib' (no such file), '/usr/local/lib/libgcc_s.1.1.dylib' (no such file), '/usr/lib/libgcc_s.1.1.dylib' (no such file, not in dyld cache), '$PREFIX/etc/conda/test-files/finch-tensor/1/libgcc_s.1.1.dylib' (no such file), '/System/Volumes/Preboot/Cryptexes/OS$PREFIX/etc/conda/test-files/finch-tensor/1/libgcc_s.1.1.dylib' (no such file), '$PREFIX/etc/conda/test-files/finch-tensor/1/libgcc_s.1.1.dylib' (no such file), '/usr/local/lib/libgcc_s.1.1.dylib' (no such file), '/usr/lib/libgcc_s.1.1.dylib' (no such file, not in dyld cache)

[lucascolley]

hey @conda-forge/help-julia, should we be including a dependency on the julia package perhaps?

[mkitti]

I think these dependencies are sufficient unless you have specific issue.

[lucascolley]

I think these dependencies are sufficient unless you have specific issue.

there was an issue: #32124 (comment)

[lucascolley]

hey @traversaro, could you possibly take another look?

[traversaro]

Besides the Windows CI failures, the recipe seems good to me.

[lucascolley]

Besides the Windows CI failures, the recipe seems good to me.

Any idea what's up there?

[traversaro]

2026-08-01T19:29:58.3822960Z  │ (base) %SRC_DIR%_run_env\etc\conda\test-files\finch-tensor\1>set "RB_SECTION_ERRORLEVEL=0" 
2026-08-01T19:29:58.3834717Z  │ (base) %SRC_DIR%_run_env\etc\conda\test-files\finch-tensor\1>popd
2026-08-01T19:29:58.3859611Z  │ (base) %SRC_DIR%>if 0 EQU 0 if 0 NEQ 0 set "RB_SECTION_ERRORLEVEL=0" 
2026-08-01T19:29:58.3877139Z  │ (base) %SRC_DIR%>endlocal   & if 0 NEQ 0 exit /b 0 
2026-08-01T19:29:58.3879844Z  │
2026-08-01T19:29:58.3880414Z  ╰─────────────────── (took 4 minutes)
2026-08-01T19:30:07.7033200Z Error:   × Test failed
2026-08-01T19:30:07.7035934Z   ╰─▶ failed to remove directory `D:\bld\test\test_finch-tensorQIfCAc`: Access
2026-08-01T19:30:07.7037292Z       is denied. (os error 5)
2026-08-01T19:30:07.7038425Z 
2026-08-01T19:30:07.7561051Z Traceback (most recent call last):
2026-08-01T19:30:07.7562124Z   File "D:\a\1\s\.ci_support\build_all.py", line 409, in <module>
2026-08-01T19:30:07.7567178Z     build_all(os.path.join(root_dir, "recipes"), args.arch)
2026-08-01T19:30:07.7569244Z   File "D:\a\1\s\.ci_support\build_all.py", line 206, in build_all
2026-08-01T19:30:07.7572597Z     build_folders_rattler_build(recipes_dir, platform, arch, channel_urls)
2026-08-01T19:30:07.7573265Z   File "D:\a\1\s\.ci_support\build_all.py", line 340, in build_folders_rattler_build
2026-08-01T19:30:07.7576054Z     subprocess.run(args + ["--variant-config", fp.name], check=True)
2026-08-01T19:30:07.7577202Z   File "D:\Miniforge\Lib\subprocess.py", line 571, in run
2026-08-01T19:30:07.7581447Z     raise CalledProcessError(retcode, process.args,

I guess that perhaps some process started by the tests is not properly closed at the end of the tests?

[lucascolley]

I guess that perhaps some process started by the tests is not properly closed at the end of the tests?

@willow-ahrens @mtsokol @hameerabbasi any guesses?

[mkitti]

You are in luck because I also happen to be the HDF5.jl maintainer. At the moment, I cannot figure out why libhdf5 would need to load the Fortran library for Julia. I suspect that more than one library may be trying to access HDF5. We have a few options, including pointing to the conda-forge HDF5 libraries.

[hameerabbasi]

I guess that perhaps some process started by the tests is not properly closed at the end of the tests?

@willow-ahrens @mtsokol @hameerabbasi any guesses?

Could be the C compiler/linker; I can't think of any other external processes.

[lucascolley]

let's return to this once prefix-dev/rattler-build#2729 is released