Unix: Add _PATH_BSHELL and _PATH_DEFPATH where available

[asder8215]

This PR adds _PATH_BSHELL and _PATH_DEFPATH constants from include/paths.h on multiple unix-based platforms (a couple platforms only have _PATH_BSHELL). The reason for why we need these macros could be seen from this Zulip chat #t-libs/crates > Including <paths.h> macros into libc? @ 💬 and PR.

The following are sources to each platform's paths.h:

From BSD:

• Apple: Has _PATH_BSHELL and _PATH_DEFPATH
• Dragonfly: Has _PATH_BSHELL and _PATH_DEFPATH
• FreeBSD: Has _PATH_BSHELL and _PATH_DEFPATH (verified same values on version 11+)
• NetBSD: Has _PATH_BSHELL and _PATH_DEFPATH
• OpenBSD: Has _PATH_BSHELL and _PATH_DEFPATH

From Linux:

• Android: Has _PATH_BSHELL and _PATH_DEFPATH
• glibc: Has _PATH_BSHELL and _PATH_DEFPATH
• musl: Has _PATH_BSHELL and _PATH_DEFPATH
• uClibc: Has _PATH_BSHELL and _PATH_DEFPATH
• emscripten: Has _PATH_BSHELL and _PATH_DEFPATH
• l4re: There's musl and uclibc paths.h file containing _PATH_BSHELL and _PATH_DEFPATH values

Miscellaneous Platforms:

• Haiku: Has _PATH_BSHELL and _PATH_DEFPATH
• Hurd: Doesn't have_PATH_BSHELL and _PATH_DEFPATH in paths.h (wouldn't hurt to confirm this though in terminal)
• Cygwin: Has _PATH_BSHELL and _PATH_DEFPATH
• newlib: Has _PATH_BSHELL
• Redox: Has _PATH_BSHELL

Platforms I'm unsure if they have _PATH_BSHELL/_PATH_DEFPATH:

• Solarish: I'm unsure how to get this info since I believe Solaris is closed source?
• AIX: I think this is also closed source? However, I saw this archive repo of AIX 4.13 that has _PATH_BSHELL.
• nuttx: Couldn't find a include/paths.h file.
• NTO/QNX: Couldn't find a include/paths.h file.
  Would be helpful if someone could go on terminal and print out _PATH_BSHELL and see if that produces results?

Other questions I have are:

• For NetBSD, I noticed that _PATH_DEFPATH concatenates the RESCUEDIR macro in front of its string literal if it exists, should we also have that in our _PATH_DEFPATH macro? I wasn't sure (I also don't know what the value of RESCUEDIR is), and I was advised by Jonas to reach out to @he32 and @0323pin since they're experienced with NetBSD.

This is also my first time contributing to libc. I've read through the CONTRIBUTING.md file and hopefully I got everything done correctly. I know the md file also mentions about reorganization effort for moving from the hierarchial structure to the source-mapped structure in src/new. If you want me to restructure the constants I'm introducing to src/new, I can try for that; I may need more clarity on how that process works.

cc @joboet
@rustbot label stable-nominated

[rustbot]

Some changes occurred in a NetBSD-like module

cc @semarie

Some changes occurred in an OpenBSD module

cc @semarie

Some changes occurred in an Android module

cc @maurer

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[tgross35]

You'll need to use byte arrays to meet the MSRV. Add a function to

libc/src/types.rs

Lines 69 to 140 in b6d2732

	/// Used to avoid `overflowing_literals` when the value is in-range for the unsigned number but
	/// out-of-range for signed.
	#[allow(unused)]
	pub(crate) const fn u16_cast_short(x: u16) -> c_short {
	assert!(size_of::<u16>() <= size_of::<c_short>()); // Should always be true
	x as i16
	}
	/// Used to avoid `overflowing_literals` when the value is in-range for the unsigned number but
	/// out-of-range for signed.
	#[allow(unused)]
	pub(crate) const fn u32_cast_int(x: u32) -> c_int {
	// May not be true on 16-bit platforms, but should be everywhere this is used.
	assert!(size_of::<u32>() <= size_of::<c_int>());
	x as i32
	}
	/// Used to avoid `overflowing_literals` when the value is in-range for the unsigned number but
	/// out-of-range for signed.
	#[allow(unused)]
	pub(crate) const fn u32_cast_long(x: u32) -> c_long {
	assert!(size_of::<u32>() <= size_of::<c_long>()); // Should always be true
	x as c_long
	}
	/// Checked casting from `unsigned long` to `int`.
	#[allow(unused)]
	pub(crate) const fn ulong_cast_int(x: c_ulong) -> c_int {
	assert!(x <= (c_int::MAX as c_ulong));
	x as c_int
	}
	/// Checked casting from `unsigned long` to `unsigned int`.
	#[allow(unused)]
	pub(crate) const fn ulong_cast_uint(x: c_ulong) -> c_uint {
	assert!(x <= (c_uint::MAX as c_ulong));
	x as c_uint
	}
	/// Used to avoid `overflowing_literals` when the value is in-range for the unsigned number but
	/// out-of-range for signed.
	#[allow(unused)]
	#[cfg(any(target_os = "linux", target_os = "android", target_os = "l4re"))]
	pub(crate) const fn u32_cast_ioctl(x: u32) -> crate::Ioctl {
	assert!(size_of::<u32>() <= size_of::<crate::Ioctl>()); // Should always be true
	x as crate::Ioctl
	}
	#[allow(unused)]
	pub(crate) const fn u8_slice_cast_char_slice(x: &[u8]) -> &[c_char] {
	assert!(size_of::<u8>() == size_of::<c_char>());
	// SAFETY: same repr, possibly just a sign cast
	unsafe { mem::transmute::<&[u8], &[c_char]>(x) }
	}
	/// Replace bytes in an array with those from a slice. This is a polyfill for `[T]::copy_from_slice`
	/// in `const`.
	// FIXME(msrv): we can switch to `copy_from_slice` in 1.87.
	#[must_use]
	#[allow(dead_code)]
	pub const fn replace_array_items<T: Copy, const N: usize>(
	mut dst: [T; N],
	src: &[T],
	start: usize,
	) -> [T; N] {
	let mut i = 0;
	while i < src.len() {
	dst[i + start] = src[i];
	i += 1;
	}
	dst
	}

like fn cstr(bytes: &[u8]) -> *const c_char that asserts there's a single nul at the end and does the cast, with a FIXME(msrv) comment.

View changes since this review

[tgross35]

Drop the doc comments, libc expects you to read the manpages

[tgross35]

Simplify this by using if cfg!(...) { ... } else { ... } rather than #[cfg]

[tgross35]

Please wrap long lines with \

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[tgross35]

Two other things:

• Please add the source list to the commit message (this repo uses rebase merging)
• You'll need to update libc-test/build.rs to include the relevant headers

[asder8215]

Going to update commit with permalinks

[asder8215]

@tgross35 I'm unsure why aarch64-apple-darwin is failing. It mentions:

error[E0432]: unresolved import `libc::_PATH_DEFPATH`
    --> /Users/runner/work/libc/libc/target/aarch64-apple-darwin/debug/build/libc-test/62cbf2747c9d1d64/out/semver.rs:2534:5
     |
2534 |     _PATH_DEFPATH,
     |     ^^^^^^^^^^^^^ no `_PATH_DEFPATH` in the root

I know paths.h isn't defined in:
https://github.com/apple-oss-distributions/xnu/tree/f6217f891ac0bb64f3d375211650a4c1ff8ca1ea

But it is defined in their Libc repo.