Rollup merge of #94503 - joshtriplett:core-ffi-c, r=Amanieu

Provide C FFI types via core::ffi, not just in std

Tracking issue: #94501

The ability to interoperate with C code via FFI is not limited to crates
using std; this allows using these types without std.

The existing types in `std::os::raw` become type aliases for the ones in
`core::ffi`. This uses type aliases rather than re-exports, to allow the
std types to remain stable while the core types are unstable.

This also moves the currently unstable `NonZero_` variants and
`c_size_t`/`c_ssize_t`/`c_ptrdiff_t` types to `core::ffi`, while leaving
them unstable.

Historically, we didn't do this because these types are target-dependent.
However, `core` itself is also target-dependent. `core` should not call
any OS services, but it knows the target and the target's ABI.

Author: matthiaskrgr

Diff for: library/std/src/os/raw/char.md renamed to library/core/src/ffi/c_char.md

@@ -2,8 +2,7 @@ Equivalent to C's `char` type.
 
 [C's `char` type] is completely unlike [Rust's `char` type]; while Rust's type represents a unicode scalar value, C's `char` type is just an ordinary integer. On modern architectures this type will always be either [`i8`] or [`u8`], as they use byte-addresses memory with 8-bit bytes.
 
-C chars are most commonly used to make C strings. Unlike Rust, where the length of a string is included alongside the string, C strings mark the end of a string with the character `'\0'`. See [`CStr`] for more information.
+C chars are most commonly used to make C strings. Unlike Rust, where the length of a string is included alongside the string, C strings mark the end of a string with the character `'\0'`. See `CStr` for more information.
 
 [C's `char` type]: https://en.wikipedia.org/wiki/C_data_types#Basic_types
 [Rust's `char` type]: char
-[`CStr`]: crate::ffi::CStr

Diff for: library/std/src/os/raw/double.md renamed to library/core/src/ffi/c_double.md

@@ -0,0 +1,16 @@
+Equivalent to C's `void` type when used as a [pointer].
+
+In essence, `*const c_void` is equivalent to C's `const void*`
+and `*mut c_void` is equivalent to C's `void*`. That said, this is
+*not* the same as C's `void` return type, which is Rust's `()` type.
+
+To model pointers to opaque types in FFI, until `extern type` is
+stabilized, it is recommended to use a newtype wrapper around an empty
+byte array. See the [Nomicon] for details.
+
+One could use `std::os::raw::c_void` if they want to support old Rust
+compiler down to 1.1.0. After Rust 1.30.0, it was re-exported by
+this definition. For more information, please read [RFC 2521].
+
+[Nomicon]: https://doc.rust-lang.org/nomicon/ffi.html#representing-opaque-structs
+[RFC 2521]: https://github.com/rust-lang/rfcs/blob/master/text/2521-c_void-reunification.md

Diff for: library/std/src/os/raw/float.md renamed to library/core/src/ffi/c_float.md

@@ -1,28 +1,157 @@
+//! Platform-specific types, as defined by C.
+//!
+//! Code that interacts via FFI will almost certainly be using the
+//! base types provided by C, which aren't nearly as nicely defined
+//! as Rust's primitive types. This module provides types which will
+//! match those defined by C, so that code that interacts with C will
+//! refer to the correct types.
+
 #![stable(feature = "", since = "1.30.0")]
 #![allow(non_camel_case_types)]
 
-//! Utilities related to foreign function interface (FFI) bindings.
-
 use crate::fmt;
 use crate::marker::PhantomData;
+use crate::num::*;
 use crate::ops::{Deref, DerefMut};
 
-/// Equivalent to C's `void` type when used as a [pointer].
-///
-/// In essence, `*const c_void` is equivalent to C's `const void*`
-/// and `*mut c_void` is equivalent to C's `void*`. That said, this is
-/// *not* the same as C's `void` return type, which is Rust's `()` type.
+macro_rules! type_alias_no_nz {
+    {
+      $Docfile:tt, $Alias:ident = $Real:ty;
+      $( $Cfg:tt )*
+    } => {
+        #[doc = include_str!($Docfile)]
+        $( $Cfg )*
+        #[unstable(feature = "core_ffi_c", issue = "94501")]
+        pub type $Alias = $Real;
+    }
+}
+
+// To verify that the NonZero types in this file's macro invocations correspond
+//
+//  perl -n < library/std/src/os/raw/mod.rs -e 'next unless m/type_alias\!/; die "$_ ?" unless m/, (c_\w+) = (\w+), NonZero_(\w+) = NonZero(\w+)/; die "$_ ?" unless $3 eq $1 and $4 eq ucfirst $2'
+//
+// NB this does not check that the main c_* types are right.
+
+macro_rules! type_alias {
+    {
+      $Docfile:tt, $Alias:ident = $Real:ty, $NZAlias:ident = $NZReal:ty;
+      $( $Cfg:tt )*
+    } => {
+        type_alias_no_nz! { $Docfile, $Alias = $Real; $( $Cfg )* }
+
+        #[doc = concat!("Type alias for `NonZero` version of [`", stringify!($Alias), "`]")]
+        #[unstable(feature = "raw_os_nonzero", issue = "82363")]
+        $( $Cfg )*
+        pub type $NZAlias = $NZReal;
+    }
+}
+
+type_alias! { "c_char.md", c_char = c_char_definition::c_char, NonZero_c_char = c_char_definition::NonZero_c_char;
+// Make this type alias appear cfg-dependent so that Clippy does not suggest
+// replacing `0 as c_char` with `0_i8`/`0_u8`. This #[cfg(all())] can be removed
+// after the false positive in https://github.com/rust-lang/rust-clippy/issues/8093
+// is fixed.
+#[cfg(all())]
+#[doc(cfg(all()))] }
+type_alias! { "c_schar.md", c_schar = i8, NonZero_c_schar = NonZeroI8; }
+type_alias! { "c_uchar.md", c_uchar = u8, NonZero_c_uchar = NonZeroU8; }
+type_alias! { "c_short.md", c_short = i16, NonZero_c_short = NonZeroI16; }
+type_alias! { "c_ushort.md", c_ushort = u16, NonZero_c_ushort = NonZeroU16; }
+type_alias! { "c_int.md", c_int = i32, NonZero_c_int = NonZeroI32; }
+type_alias! { "c_uint.md", c_uint = u32, NonZero_c_uint = NonZeroU32; }
+type_alias! { "c_long.md", c_long = i32, NonZero_c_long = NonZeroI32;
+#[doc(cfg(all()))]
+#[cfg(any(target_pointer_width = "32", windows))] }
+type_alias! { "c_ulong.md", c_ulong = u32, NonZero_c_ulong = NonZeroU32;
+#[doc(cfg(all()))]
+#[cfg(any(target_pointer_width = "32", windows))] }
+type_alias! { "c_long.md", c_long = i64, NonZero_c_long = NonZeroI64;
+#[doc(cfg(all()))]
+#[cfg(all(target_pointer_width = "64", not(windows)))] }
+type_alias! { "c_ulong.md", c_ulong = u64, NonZero_c_ulong = NonZeroU64;
+#[doc(cfg(all()))]
+#[cfg(all(target_pointer_width = "64", not(windows)))] }
+type_alias! { "c_longlong.md", c_longlong = i64, NonZero_c_longlong = NonZeroI64; }
+type_alias! { "c_ulonglong.md", c_ulonglong = u64, NonZero_c_ulonglong = NonZeroU64; }
+type_alias_no_nz! { "c_float.md", c_float = f32; }
+type_alias_no_nz! { "c_double.md", c_double = f64; }
+
+/// Equivalent to C's `size_t` type, from `stddef.h` (or `cstddef` for C++).
 ///
-/// To model pointers to opaque types in FFI, until `extern type` is
-/// stabilized, it is recommended to use a newtype wrapper around an empty
-/// byte array. See the [Nomicon] for details.
+/// This type is currently always [`usize`], however in the future there may be
+/// platforms where this is not the case.
+#[unstable(feature = "c_size_t", issue = "88345")]
+pub type c_size_t = usize;
+
+/// Equivalent to C's `ptrdiff_t` type, from `stddef.h` (or `cstddef` for C++).
 ///
-/// One could use `std::os::raw::c_void` if they want to support old Rust
-/// compiler down to 1.1.0. After Rust 1.30.0, it was re-exported by
-/// this definition. For more information, please read [RFC 2521].
+/// This type is currently always [`isize`], however in the future there may be
+/// platforms where this is not the case.
+#[unstable(feature = "c_size_t", issue = "88345")]
+pub type c_ptrdiff_t = isize;
+
+/// Equivalent to C's `ssize_t` (on POSIX) or `SSIZE_T` (on Windows) type.
 ///
-/// [Nomicon]: https://doc.rust-lang.org/nomicon/ffi.html#representing-opaque-structs
-/// [RFC 2521]: https://github.com/rust-lang/rfcs/blob/master/text/2521-c_void-reunification.md
+/// This type is currently always [`isize`], however in the future there may be
+/// platforms where this is not the case.
+#[unstable(feature = "c_size_t", issue = "88345")]
+pub type c_ssize_t = isize;
+
+mod c_char_definition {
+    cfg_if! {
+        // These are the targets on which c_char is unsigned.
+        if #[cfg(any(
+            all(
+                target_os = "linux",
+                any(
+                    target_arch = "aarch64",
+                    target_arch = "arm",
+                    target_arch = "hexagon",
+                    target_arch = "powerpc",
+                    target_arch = "powerpc64",
+                    target_arch = "s390x",
+                    target_arch = "riscv64",
+                    target_arch = "riscv32"
+                )
+            ),
+            all(target_os = "android", any(target_arch = "aarch64", target_arch = "arm")),
+            all(target_os = "l4re", target_arch = "x86_64"),
+            all(
+                target_os = "freebsd",
+                any(
+                    target_arch = "aarch64",
+                    target_arch = "arm",
+                    target_arch = "powerpc",
+                    target_arch = "powerpc64",
+                    target_arch = "riscv64"
+                )
+            ),
+            all(
+                target_os = "netbsd",
+                any(target_arch = "aarch64", target_arch = "arm", target_arch = "powerpc")
+            ),
+            all(target_os = "openbsd", target_arch = "aarch64"),
+            all(
+                target_os = "vxworks",
+                any(
+                    target_arch = "aarch64",
+                    target_arch = "arm",
+                    target_arch = "powerpc64",
+                    target_arch = "powerpc"
+                )
+            ),
+            all(target_os = "fuchsia", target_arch = "aarch64")
+        ))] {
+            pub type c_char = u8;
+            pub type NonZero_c_char = crate::num::NonZeroU8;
+        } else {
+            // On every other target, c_char is signed.
+            pub type c_char = i8;
+            pub type NonZero_c_char = crate::num::NonZeroI8;
+        }
+    }
+}
+
 // N.B., for LLVM to recognize the void pointer type and by extension
 //     functions like malloc(), we need to have it represented as i8* in
 //     LLVM bitcode. The enum used here ensures this and prevents misuse
@@ -31,6 +160,7 @@ use crate::ops::{Deref, DerefMut};
 //     otherwise and we need at least one variant as otherwise the enum
 //     would be uninhabited and at least dereferencing such pointers would
 //     be UB.
+#[doc = include_str!("c_void.md")]
 #[repr(u8)]
 #[stable(feature = "core_c_void", since = "1.30.0")]
 pub enum c_void {

Diff for: library/std/src/os/raw/int.md renamed to library/core/src/ffi/c_int.md

@@ -187,3 +187,96 @@ macro_rules! impl_fn_for_zst {
         )+
     }
 }
+
+/// A macro for defining `#[cfg]` if-else statements.
+///
+/// `cfg_if` is similar to the `if/elif` C preprocessor macro by allowing definition of a cascade
+/// of `#[cfg]` cases, emitting the implementation which matches first.
+///
+/// This allows you to conveniently provide a long list `#[cfg]`'d blocks of code without having to
+/// rewrite each clause multiple times.
+///
+/// # Example
+///
+/// ```
+/// cfg_if! {
+///     if #[cfg(unix)] {
+///         fn foo() { /* unix specific functionality */ }
+///     } else if #[cfg(target_pointer_width = "32")] {
+///         fn foo() { /* non-unix, 32-bit functionality */ }
+///     } else {
+///         fn foo() { /* fallback implementation */ }
+///     }
+/// }
+///
+/// # fn main() {}
+/// ```
+// This is a copy of `cfg_if!` from the `cfg_if` crate.
+// The recursive invocations should use $crate if this is ever exported.
+macro_rules! cfg_if {
+    // match if/else chains with a final `else`
+    (
+        $(
+            if #[cfg( $i_meta:meta )] { $( $i_tokens:tt )* }
+        ) else+
+        else { $( $e_tokens:tt )* }
+    ) => {
+        cfg_if! {
+            @__items () ;
+            $(
+                (( $i_meta ) ( $( $i_tokens )* )) ,
+            )+
+            (() ( $( $e_tokens )* )) ,
+        }
+    };
+
+    // match if/else chains lacking a final `else`
+    (
+        if #[cfg( $i_meta:meta )] { $( $i_tokens:tt )* }
+        $(
+            else if #[cfg( $e_meta:meta )] { $( $e_tokens:tt )* }
+        )*
+    ) => {
+        cfg_if! {
+            @__items () ;
+            (( $i_meta ) ( $( $i_tokens )* )) ,
+            $(
+                (( $e_meta ) ( $( $e_tokens )* )) ,
+            )*
+        }
+    };
+
+    // Internal and recursive macro to emit all the items
+    //
+    // Collects all the previous cfgs in a list at the beginning, so they can be
+    // negated. After the semicolon is all the remaining items.
+    (@__items ( $( $_:meta , )* ) ; ) => {};
+    (
+        @__items ( $( $no:meta , )* ) ;
+        (( $( $yes:meta )? ) ( $( $tokens:tt )* )) ,
+        $( $rest:tt , )*
+    ) => {
+        // Emit all items within one block, applying an appropriate #[cfg]. The
+        // #[cfg] will require all `$yes` matchers specified and must also negate
+        // all previous matchers.
+        #[cfg(all(
+            $( $yes , )?
+            not(any( $( $no ),* ))
+        ))]
+        cfg_if! { @__identity $( $tokens )* }
+
+        // Recurse to emit all other items in `$rest`, and when we do so add all
+        // our `$yes` matchers to the list of `$no` matchers as future emissions
+        // will have to negate everything we just matched as well.
+        cfg_if! {
+            @__items ( $( $no , )* $( $yes , )? ) ;
+            $( $rest , )*
+        }
+    };
+
+    // Internal macro to make __apply work out right for different match types,
+    // because of how macros match/expand stuff.
+    (@__identity $( $tokens:tt )* ) => {
+        $( $tokens )*
+    };
+}

Diff for: library/std/src/os/raw/long.md renamed to library/core/src/ffi/c_long.md

@@ -261,6 +261,7 @@
 #![feature(const_socketaddr)]
 #![feature(const_trait_impl)]
 #![feature(container_error_extra)]
+#![feature(core_ffi_c)]
 #![feature(core_intrinsics)]
 #![feature(core_panic)]
 #![feature(custom_test_frameworks)]
@@ -315,6 +316,7 @@
 #![feature(prelude_import)]
 #![feature(ptr_as_uninit)]
 #![feature(ptr_internals)]
+#![feature(raw_os_nonzero)]
 #![feature(rustc_attrs)]
 #![feature(rustc_private)]
 #![feature(saturating_int_impl)]