add more *at functions. #562

Mic92 · 2017-03-26T10:42:36Z

No description provided.

Mic92 · 2017-03-30T10:38:49Z

That's all folks. I have all methods for my fuse.

kamalmarhubi · 2017-04-08T00:00:14Z

@Mic92 this is heroic. I'm going to take a look over the weekend. Sorry for the delay!

kamalmarhubi · 2017-04-08T00:09:18Z

I retried the failing test and it works now so that's a plus.

Mic92 · 2017-04-08T08:42:30Z

I would also add large file support to all methods in a second pull request (64bit fields on 32-bit systems), to align the with glibc/musl/rust. Ideally this should be done before the next release.

homu · 2017-04-15T21:13:01Z

☔ The latest upstream changes (presumably #536) made this pull request unmergeable. Please resolve the merge conflicts.

posborne · 2017-04-16T01:16:34Z

This is great @Mic92, just needs a rebase by the looks of it (the curse of the changelog). The only thing I would add is that we have been trying to add docs to more of the functions we are adding. I do not think this is a blocker for getting this included but would be a nice bonus for this change or a future PR.

E.g. Docs for fork: https://github.com/nix-rust/nix/blob/master/src/unistd.rs#L47. The main value is providing a quick link to relevant man pages and a quick overview (some of the names of the functions are somewhat cryptic).

lucab · 2017-04-24T14:25:42Z

I have a doubt on this (and on other fn taking RawFd in this crate): are all these functions just taking ownership of (a copy of) the fd or borrowing it? I think these all do an implicit copy+own of the RawFd (as it is just an alias to an int), and I fear this doesn't really guarantee that the fd has not been closed/dropped in the meanwhile by a close().

I'm not sure if it is a valid concern or not, but shouldn't all these functions take a &RawFd (and perhaps RawFd be made non-Copy)?

kamalmarhubi · 2017-04-24T15:26:29Z

@lucab this is a real concern, and I agree that we should do what you suggest. We cannot change RawFd as it's part of the standard library. But we should change all RawFd parameters to take references, or something at least that safe. I've opened #594 to track it. Thanks for raising this!

Mic92 · 2017-06-03T21:55:46Z

A reference does not buy you anything: https://play.rust-lang.org/?gist=1739781faf61fcec4afaf65bbaadd968&version=stable&backtrace=0
An not copyable type would destroy the semantics of some system calls like renameat, as it would no longer allow to pass the same directory twice.

Mic92 · 2017-06-03T21:57:53Z

Merge conflict resolved. Man page references added.

Susurrus · 2017-06-03T22:00:57Z

src/unistd.rs

@@ -356,6 +375,13 @@ pub fn getcwd() -> Result<PathBuf> {
    }
 }

+// According to the POSIX specification, -1 is used to indicate that


This comment doesn't make sense here. Probably want to put it back in the function.

I use this function in four different flavors of chown. Should I copied the comment to each of them? Why it does not make it sense here?

Because you're referring to uid_t and gid_t which aren't used in this function. The description to this should be more broad (not referring to those specific use cases) and then you can put comments stating why you need to use this function at the call site that refers to specific variables.

Susurrus · 2017-06-03T22:01:16Z

src/sys/stat.rs

@@ -53,6 +55,7 @@ bitflags! {
    }
 }

+


This extra line should be removed.

Susurrus · 2017-06-03T22:03:00Z

src/unistd.rs

@@ -601,6 +672,69 @@ pub fn lseek64(fd: RawFd, offset: libc::off64_t, whence: Whence) -> Result<libc:
    Errno::result(res).map(|r| r as libc::off64_t)
 }

+/// call the link function to create a link to a file


Please capitalize the first word and add a period to the end of the first line of all these doc comments. If you run cargo doc --no-deps --open you can easily review their appearance to make sure things look good there.

Susurrus · 2017-06-03T22:10:18Z

src/sys/stat.rs

+
+    /// Change file timestamps with nanosecond precision
+    /// (see [utimensat(2)](http://man7.org/linux/man-pages/man2/utimensat.2.html))
+    pub fn utimensat<P: ?Sized + NixPath>(dirfd: RawFd,


dirfd and pathname actually combo together to form a ternary input type: one allowing a RawFd, another a relative path from the current working directory, and a third using an absolute path. I think we should use a ternary enum type here instead of the dirfd and pathname inputs to make this function call safer.

Susurrus · 2017-06-03T22:12:13Z

src/sys/stat.rs

+    use sys::time::TimeSpec;
+
+    pub enum UtimeSpec {
+        Now,


This enum is public and as such needs docs.

Susurrus · 2017-06-04T22:14:34Z

There a common theme through a lot of these functions of having a dirfd and a path that is a tristate:

The dirfd is ignored if path is absolute
If dirfd is AT_FDCWD and path is relative it's resolved relative to the CWD
If path is relative and dirfd isn't AT_FDCWD then the path is resolved relative to that path.

I'm thinking an enum encompassing these states would be better instead of having two separate variables. I think an ideal solution would be to have NixPath have an Absolute and Relative mode or something to that effect, but for this PR I think just having an enum with different types is good enough.

Susurrus · 2017-06-04T22:14:52Z

test/test_stat.rs

 use std::os::unix::prelude::AsRawFd;
+use tempfile::NamedTempFile;


Please put non-std/core imports down in their own group.

Susurrus · 2017-06-04T22:16:09Z

src/sys/stat.rs

+        Now,
+        /// The corresponding file timestamp is left unchanged.
+        Omit,
+        Time(TimeSpec)


This one needs a comment too.

Susurrus · 2017-06-04T22:17:06Z

src/sys/stat.rs

+        Time(TimeSpec)
+    }
+
+    fn to_timespec(time: &UtimeSpec) -> libc::timespec {


Can you add a doccomment for this function? Nix devs should know if they can or should use it elsewhere in this function.

Susurrus · 2017-06-04T22:18:40Z

CHANGELOG.md

- Added `nix::unistd::{openat, fstatat, readlink, readlinkat, rename, renameat, mknodat}`
-  ([#551](https://github.com/nix-rust/nix/pull/551))
+- Added `nix::unistd::{openat, fstatat, readlink, readlinkat, rename, renameat, mknodat, unlinkat}`
+  ([#497](https://github.com/nix-rust/nix/pull/551))


I don't think the PR numbering change he was intentional.

Mic92 · 2017-06-05T05:51:55Z

@Susurrus do you want to implement that enum for these functions send me a pull request? I don't see how Rust's typesystem is expressive enough to ensure a path is always relative/absolute without runtime checks.

Susurrus · 2017-06-05T14:40:51Z

My suggestion is to have a:

pub enum AtPath {
  AbsPath(NixPath),
  RelativeTo(RawFd, NixPath),
  RelativeCwd(NixPath),
}

Now we could add runtime checks through having some conversion path from NixPaths to this enum and I think that's the most ergonomic, which would be ideal, and if you'd like to implement something like that we can spec it out here. But at least with this enum it's more clear to the user what combinations of inputs are valid.

Susurrus · 2017-06-05T14:54:22Z

I wanted to point out that I also preferred the arguments @bugaevc who was working on access/faccessat in #605. There they came up with a dual-state enum for use with those functions that looks like:

pub enum AccessFlags {
  FileExists,
  FileExistsAndPermissions(PermFlags),
}

where PermFlags are the R_OK, etc. constants. I'd prefer having that than the regular libc-like constants.

Susurrus · 2017-08-22T16:13:13Z

@Mic92 I wanted to circle back to this and see if we could get this merged? I'm working on some stuff in fcntl but it should get merged within a day or so. Then I'd like to get this PR merged if possible. Still willing to push this through?

Mic92 · 2017-08-22T19:18:32Z

I am back from holiday on Thursday.

lucab · 2017-09-06T12:01:50Z

What happened to this?

Mic92 · 2017-09-06T18:00:29Z

I made multiple pull requests out of it.

Mic92 force-pushed the atfamily branch 2 times, most recently from 0c43c15 to b7fa4c0 Compare March 27, 2017 12:44

Mic92 force-pushed the atfamily branch 5 times, most recently from 98eb8a0 to 6077eef Compare April 5, 2017 15:03

kamalmarhubi mentioned this pull request Apr 24, 2017

Taking RawFd by value is not sound #594

Open

Mic92 force-pushed the atfamily branch from 6077eef to 7eac89f Compare June 3, 2017 21:57

Susurrus reviewed Jun 3, 2017

View reviewed changes

src/sys/stat.rs Outdated

@@ -53,6 +55,7 @@ bitflags! {

}

}

Copy link

Contributor

Susurrus Jun 3, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This extra line should be removed.

Susurrus reviewed Jun 3, 2017

View reviewed changes

src/sys/stat.rs Outdated

use sys::time::TimeSpec;

pub enum UtimeSpec {

Now,

Copy link

Contributor

Susurrus Jun 3, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This enum is public and as such needs docs.

Mic92 force-pushed the atfamily branch 3 times, most recently from 2449a0c to b3a3a25 Compare June 3, 2017 22:40

Susurrus reviewed Jun 4, 2017

View reviewed changes

Mic92 force-pushed the atfamily branch 3 times, most recently from 61a42c0 to 6aa502e Compare June 5, 2017 05:47

Susurrus mentioned this pull request Jun 5, 2017

Add unistd::{access, faccessat} #605

Closed

Mic92 added 10 commits September 1, 2017 10:49

add support for mknodat

099628e

add support for rename/renameat

007a2b7

add support for unlinkat

3c2315e

add support for mkdirat

ba25de1

add support for link|linkat

3d56d3a

add support for symlink|symlinkat

7925ab8

add support for access/faccessat

4ecc4a1

add support for chmod/fchmod/fchmodat

df45fe7

add support for chown/lchown/fchown/fchownat

ad170f9

add support for futimens/utimensat

1c84ae3

Mic92 force-pushed the atfamily branch from 6aa502e to 1c84ae3 Compare September 1, 2017 10:22

Mic92 closed this Sep 1, 2017

oblique mentioned this pull request Dec 12, 2018

About *at functions #998

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add more *at functions. #562

add more *at functions. #562

Mic92 commented Mar 26, 2017

Mic92 commented Mar 30, 2017

kamalmarhubi commented Apr 8, 2017

kamalmarhubi commented Apr 8, 2017

Mic92 commented Apr 8, 2017 •

edited

Loading

homu commented Apr 15, 2017

posborne commented Apr 16, 2017 •

edited

Loading

lucab commented Apr 24, 2017

kamalmarhubi commented Apr 24, 2017

Mic92 commented Jun 3, 2017

Mic92 commented Jun 3, 2017

Susurrus Jun 3, 2017

Mic92 Jun 3, 2017

Susurrus Jun 3, 2017

Susurrus Jun 3, 2017

Susurrus Jun 3, 2017

Susurrus Jun 3, 2017

Susurrus Jun 3, 2017

Susurrus commented Jun 4, 2017

Susurrus Jun 4, 2017

Susurrus Jun 4, 2017

Susurrus Jun 4, 2017

Susurrus Jun 4, 2017

Mic92 commented Jun 5, 2017

Susurrus commented Jun 5, 2017

Susurrus commented Jun 5, 2017

Susurrus commented Aug 22, 2017

Mic92 commented Aug 22, 2017

lucab commented Sep 6, 2017

Mic92 commented Sep 6, 2017

		use std::os::unix::prelude::AsRawFd;
		use tempfile::NamedTempFile;

add more *at functions. #562

add more *at functions. #562

Conversation

Mic92 commented Mar 26, 2017

Mic92 commented Mar 30, 2017

kamalmarhubi commented Apr 8, 2017

kamalmarhubi commented Apr 8, 2017

Mic92 commented Apr 8, 2017 • edited Loading

homu commented Apr 15, 2017

posborne commented Apr 16, 2017 • edited Loading

lucab commented Apr 24, 2017

kamalmarhubi commented Apr 24, 2017

Mic92 commented Jun 3, 2017

Mic92 commented Jun 3, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Susurrus commented Jun 4, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Mic92 commented Jun 5, 2017

Susurrus commented Jun 5, 2017

Susurrus commented Jun 5, 2017

Susurrus commented Aug 22, 2017

Mic92 commented Aug 22, 2017

lucab commented Sep 6, 2017

Mic92 commented Sep 6, 2017

Mic92 commented Apr 8, 2017 •

edited

Loading

posborne commented Apr 16, 2017 •

edited

Loading