feat: add missing lch* syscalls to *BSD platforms

[ngie-eign]

This change adds the following system calls to various 4.4BSD derivatives:

• FreeBSD(-like)
  • lchmod ([DragonFlyBSD][1], [FreeBSD][2])
• macOS
  • [lchflags][3]
  • [lchmod][4]
• NetBSD
  • [lchmod][5]

These system calls are available on all ancient versions of *BSD (4.4BSD, FreeBSD 3.0, etc). More specific historical information about when the functions were added is available in the manpages and headers referenced.

1: https://leaf.dragonflybsd.org/cgi/web-man?command=lchmod
2: https://man.freebsd.org/cgi/man.cgi?query=lchmod
3: https://github.com/apple-oss-distributions/xnu/blob/f6217f891ac0bb64f3d375211650a4c1ff8ca1ea/bsd/sys/stat.h#L597
4: https://github.com/apple-oss-distributions/xnu/blob/f6217f891ac0bb64f3d375211650a4c1ff8ca1ea/bsd/sys/stat.h#L598
5: https://man.netbsd.org/lchmod.2

[rustbot]

Some changes occurred in a NetBSD-like module

cc @semarie

[tgross35]

Diff looks reasonable but could you please add source permalinks to the commit message? https://github.com/rust-lang/libc/blob/main/CONTRIBUTING.md#source-links

[ngie-eign]

Diff looks reasonable but could you please add source permalinks to the commit message? https://github.com/rust-lang/libc/blob/main/CONTRIBUTING.md#source-links

I added links to the BSD manpages (they have historical notes for when syscalls were added) instead of source on GitHub--which may or may not change in the future or might be pointing to out-of-date or unofficial readonly mirrors. macOS ... is another beast entirely (partly because they don't publish their documentation as much), so I've linked the actual header source as mentioned in CONTRIBUTING.md .

[tgross35]

I added links to the BSD manpages (they have historical notes for when syscalls were added) instead of source on GitHub--which may or may not change in the future or might be pointing to out-of-date or unofficial readonly mirrors. macOS ... is another beast entirely (partly because they don't publish their documentation as much), so I've linked the actual header source as mentioned in CONTRIBUTING.md .

Manpages are of course fine to include but I always need to check against headers (since sometimes they show more details or don't actually line up), so linking them is really just a way to save me from going hunting. I'll make that more clear in the docs.

As far as I'm aware, the BSD mirrors are official, and asking specifically for permalinks is how we know they're accurate at some point of time. If there's a non-mirror web viewer then that's fine too but they seem to often give problems (e.g. sourceware is flaky, https://svnweb.freebsd.org/ is currently a 503 for me, I'm not even sure if DragonFly has a viewer for their repo...).

Anyway, I already peeked so:

• https://github.com/freebsd/freebsd-src/blob/3aac283613bd3fd0228a06d6c854ca0bf190ecfb/sys/sys/stat.h#L385
• https://github.com/NetBSD/src/blob/5a7765109d6feb3a897b349b8d58572a0b7812f2/sys/sys/stat.h#L268

[ngie-eign]

I added links to the BSD manpages (they have historical notes for when syscalls were added) instead of source on GitHub--which may or may not change in the future or might be pointing to out-of-date or unofficial readonly mirrors. macOS ... is another beast entirely (partly because they don't publish their documentation as much), so I've linked the actual header source as mentioned in CONTRIBUTING.md .

Manpages are of course fine to include but I always need to check against headers (since sometimes they show more details or don't actually line up), so linking them is really just a way to save me from going hunting. I'll make that more clear in the docs.

As far as I'm aware, the BSD mirrors are official, and asking specifically for permalinks is how we know they're accurate at some point of time. If there's a non-mirror web viewer then that's fine too but they seem to often give problems (e.g. sourceware is flaky, https://svnweb.freebsd.org/ is currently a 503 for me, I'm not even sure if DragonFly has a viewer for their repo...).

FreeBSD moved to cgit over a year ago [1]. There was recent discussion about shutting down svnweb on the internal developers list, but there's no timeframe for when that might happen.

Anyway, I already peeked so:

• https://github.com/freebsd/freebsd-src/blob/3aac283613bd3fd0228a06d6c854ca0bf190ecfb/sys/sys/stat.h#L385
• https://github.com/NetBSD/src/blob/5a7765109d6feb3a897b349b8d58572a0b7812f2/sys/sys/stat.h#L268

1. https://docs.freebsd.org/en/books/handbook/mirrors/

[ngie-eign]

• DragonFly hosts their primary git mirror.
• NetBSD uses cvsweb as their primary source of truth.
  • I use the readonly GitHub mirror for my own sanity, but doublecheck against my CVS checkout when upstreaming changes and vendoring changes from NetBSD in FreeBSD.
• OpenBSD uses cvsweb as their primary source of truth, but they also advertise their git mirrors on the front page.

PS My FreeBSD.org handle is: ngie, FYI.
PPS I'm working on this in order to get the pjdfstest rust rewrite done by a GSoC 2022 student into mainline.

[tgross35]

Thanks for all the info, I'll pull that all into our docs at the next update. Also looks like a rather cool project.

[tgross35]

Code LGTM as well, thanks!

View changes since this review

[ngie-eign]

(I see @tgross35 already added the stable-nominated label; whoops -- heh)

Ad hoc libc-0.2 GHA CI/CD run.
More info:

• The job was cancelled because backport: feat: add missing lch* syscalls to *BSD platforms #5405 was closed.
• The haiku target failed to build due to a warnings issue with (nightly, 2) (stable: haiku fails to build with (nightly, 2) due to warning #5406).

[ngie-eign]

• NetBSD uses cvsweb as their primary source of truth.
  • I use the readonly GitHub mirror for my own sanity, but doublecheck against my CVS checkout when upstreaming changes and vendoring changes from NetBSD in FreeBSD.

I just compared my GitHub fork against NetBSD's, and it look like they rewrote their history sometime in the not so distant past (my trunk has different hashes from their's). So, I would strongly advise not using their GitHub readonly mirror as a valid source of truth :/...