[new check] file path too long

[jameslamb]

What should be checked?

pydistcheck should warn about the presence of files with paths longer than n characters.

That check should be enforced based on the length of complete paths, e.g.

lightgbm/basic.py

Should be treated as 17 characters.

What should n be?

Up to whoever implements this feature.

The chosen default value should have some basis in the published limitations on file paths in different systems.

For example:

• https://learn.microsoft.com/en-us/windows/win32/fileio/maximum-file-path-limitation?tabs=registry
• https://serverfault.com/questions/9546/filename-length-limits-on-linux
• https://en.wikipedia.org/wiki/Comparison_of_file_systems#Limits
• https://stat.ethz.ch/pipermail/r-package-devel/2015q4/000511.html
• https://github.com/wch/r-source/blob/29559f9bf4df2c55ef5eace203cbe335bbd03f2f/src/library/tools/R/check.R#L839-LL841

It should also consider the fact that the files in a Python distribution are highly likely to be installed in paths that themselves might be fairly nested.

So, for example, on a filesystem with a maximum filepath length of 255 characters, if users install into

C:/Users/david.ortiz/Program Files/python3.9/site-packages/

, then that's already 59 characters gone.

What should the name of this check be?

contains-long-filepaths

Will this check introduce any additional configuration?

yes

Details on additional configuration required.

--max-filepath-length should be added, which controls the maximum filepath length allowed in distributions.

Distribution type

• [X] source (e.g. .tar.gz)
• [X] built (e.g. .whl)

Notes

Everything in R CMD check's check_file_names() could be useful: https://github.com/wch/r-source/blob/29559f9bf4df2c55ef5eace203cbe335bbd03f2f/src/library/tools/R/check.R#L700

[jameslamb]

I just ran into this in LightGBM's CI!

error: [Errno 63] File name too long: '/Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2-config.h /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2-impl.h /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2.h /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2b-load-sse2.h /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2b-load-sse41.h /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2b-ref.c /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2b-round.h /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2b.c /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2s-load-sse2.h /Users/runner/miniforge3/conda-bld/python-split_1686419725558/work/Modules/_blake2/impl/blake2s-load-sse41.h '

[jameslamb]

here's another good reference: https://blog.r-project.org/2023/03/07/path-length-limit-on-windows/index.html