implement std::fs::set_permissions_nofollow on unix #142938

lolbinarycat · 2025-06-24T00:07:31Z

implementation of #141607

rustbot · 2025-06-24T00:07:35Z

rustbot has assigned @ibraheemdev.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

ibraheemdev · 2025-07-06T23:19:05Z

I'm not sure we should merge this if it panics on windows.

lolbinarycat · 2025-07-06T23:37:47Z

@ibraheemdev fair, my instinct was just to avoid having a cfg in the portable code layer, wasn't sure how that would look in the generated code.

can library features be marked as incomplete, or is that only for compiler features?

ibraheemdev · 2025-07-06T23:50:50Z

fair, my instinct was just to avoid having a cfg in the portable code layer, wasn't sure how that would look in the generated code.

I'm not sure what you mean, cfg is a compile-time attribute.

can library features be marked as incomplete, or is that only for compiler features?

I think we could do that, but I'm not sure how useful it would be to have a platform-specific implementation in a cross-platform module, especially because the unix implementation is already accessible.

lolbinarycat · 2025-07-07T02:51:23Z

I'm not sure what you mean, cfg is a compile-time attribute.

Generated docs, my bad. I know rustdoc can't generate docs for all platforms at once, and the standard library does some postprocessing to make it look decent. I believe that postprocessing only happens on the std::os module, though.

I think we could do that, but I'm not sure how useful it would be to have a platform-specific implementation in a cross-platform module, especially because the unix implementation is already accessible.

I'm really not sure what you mean by that? It's not intended to stay platform-specific, i'm just personally not able to write the windows impl for various reasons.

Are you saying there's no value in this being in the standard library because it can just be implemented in user code? isn't that the case for all libc-backed i/o? regardless, the ACP has already been approved, and implementing this just for unix was discussed in this zulip thread

ibraheemdev · 2025-07-07T02:57:55Z

I'm really not sure what you mean by that? It's not intended to stay platform-specific, i'm just personally not able to write the windows impl for various reasons.

Sorry, I just meant I don't see the value in merging the unix-only implementation, given that the unix-only implementation is already accessible, and the value of this function is in that it works cross-platform.

That said I guess it's fine to move the feature forward by landing this, I didn't realize it's already been discussed.

ibraheemdev

Just a couple of docs nits.

ibraheemdev · 2025-07-07T02:58:31Z

library/std/src/fs.rs

+///
+/// # Platform-specific behavior
+///
+/// Currently unimplmented on windows.


Suggested change

/// Currently unimplmented on windows.

/// Currently unimplemented on Windows.

ibraheemdev · 2025-07-07T02:59:21Z

library/std/src/fs.rs

+///
+/// Currently unimplmented on windows.
+///
+/// On unix platforms, this results in a [`FilesystemLoop`] error if the last element is a symlink.


Suggested change

/// On unix platforms, this results in a [`FilesystemLoop`] error if the last element is a symlink.

/// On Unix platforms, this results in a [`FilesystemLoop`] error if the last element is a symlink.

ibraheemdev · 2025-07-07T03:00:54Z

library/std/src/sys/fs/mod.rs

+
+#[cfg(not(unix))]
+pub fn set_permissions_nofollow(_path: &Path, _perm: crate::fs::Permissions) -> io::Result<()> {
+    crate::unimplemented!("set_permissions_nofollow is not implmented on non-unix platforms")


Suggested change

crate::unimplemented!("set_permissions_nofollow is not implmented on non-unix platforms")

crate::unimplemented!("`set_permissions_nofollow` is currently only implemented on Unix platforms")

rustbot assigned ibraheemdev Jun 24, 2025

rustbot added S-waiting-on-review T-libs labels Jun 24, 2025