tr: Accept non UTF-8 arguments for sets

[RenjiSann]

Fixes #6343

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/tail/inotify-dir-recreate (fails in this run but passes in the 'main' branch)

[RenjiSann]

I'm having an issue here.
In order to accept non UTF-8 sets, I use OsString, which at some points, I need to transform into bytes.

To do this, I can either use OsStr::as_encoded_bytes, which is cross-platform, but only available from Rust 1.74, so this would need a bump of MSRV from 1.70 to 1.74.

The other solution is to use OsStrExt::as_bytes which is available in Rust 1.70 but only for Unix and not for windows, so there wouldn't be support for windows I guess.

Is there another possible solution ? Which one shall we proceed with ? @tertsdiepraam

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/tail/inotify-dir-recreate (fails in this run but passes in the 'main' branch)

[BenWiederhake]

That's an issue indeed.

It seems @sylvestre doesn't want to raise the MSRV, and I don't know enough on the topic to have an opinion. So let's not do that, for now.

This isn't the first time that uutils needs to convert an OsString to bytes, so I suggest taking any existing implementation (e.g. in cut: https://github.com/uutils/coreutils/pull/6037/files#diff-939c8f8ce2deb6faec07f4ce55505aec484179070d0c817da4f931a676c38cf4R355 ), moving it to some "canonical" place (perhaps src/uucore/src/lib/lib.rs?), and then just calling it. I hope this isn't too much work, and it would have the benefit of not copy-pasting the code all over the project.

So in short I'm suggesting to use OsStrExt::as_bytes, but without the code duplication.

[sylvestre]

It seems @sylvestre doesn't want to raise the MSRV

wat ? :)

[github-actions]

GNU testsuite comparison:

Congrats! The gnu test tests/timeout/timeout is no longer failing!

[RenjiSann]

I moved os_string_to_bytes from cut to the uucore crate (and transformed it into os_str_as_bytes in the meantime), and used it inside of tr.

[BenWiederhake]

The changes look good, the test is fine, lovely!

CI is red for a reason, please fix those issues first:

• Missing "use" on Windows: https://github.com/uutils/coreutils/actions/runs/9955233780/job/27502537737?pr=6563
• Still too high MSRV: https://github.com/uutils/coreutils/actions/runs/9955233780/job/27502538603?pr=6563
  I recommend doing rustup default 1.70 or something like that, which should switch your toolchain down to version 1.70. Then you'll automatically notice any problems!
• Remaining CI didn't even run because of that.

[RenjiSann]

My bad, thanks for noticing

[BenWiederhake]

Yup, thank you! Remaining CI failures are known issues: #6534 #6570 (both)

I hope I'm not being too nitpicky, but can you change the first commit's subject line to follow the convention a bit closer? I suggest:

• cut: move os_string_as_bytes utility to uucore (mention util name)
• tr: accept non utf8 arguments for sets (lowercase first letter)

Thanks for fixing tr! :)

[RenjiSann]

• cut: move os_string_as_bytes utility to uucore (mention util name)
• tr: accept non utf8 arguments for sets (lowercase first letter)

Thanks for fixing tr! :)

Done ! 😁