Auto merge of #36807 - brson:pal, r=brson

Restrict where in the tree platform-specific cfgs may be mentioned

With the ports of Rust never ending, it's important that we keep things tidy. The main thing this PR does is introduce  a new "pal" (platform abstraction layer) tidy check that limits where platform-specific CFGs may appear.

This is intended to maintain existing standards of code organization
in hopes that the standard library will continue to be refactored to
isolate platform-specific bits, making porting easier; where "standard
library" roughly means "all the dependencies of the std and test
crates".

This generally means placing restrictions on where `cfg(unix)`,
`cfg(windows)`, `cfg(target_os)` and `cfg(target_env)` may appear,
the basic objective being to isolate platform-specific code to the
platform-specific `std::sys` modules, and to the allocation,
unwinding, and libc crates.

Following are the basic rules, though there are currently
exceptions:

- core may not have platform-specific code
- liballoc_system may have platform-specific code
- liballoc_jemalloc may have platform-specific code
- libpanic_abort may have platform-specific code
- libpanic_unwind may have platform-specific code
- other crates in the std facade may not
- std may have platform-specific code in the following places
  - sys/unix/
  - sys/windows/
  - os/

There are plenty of exceptions today though, noted in the whitelist.

The end-state, IMO, is for the standard library to be portable by porting only `std::sys` (possibly extracted to its own crate), an allocator crate, an unwinder crate, and possibly a libc crate (if std depends on it); but that outcome is far off and independent of the utility of enforcing where such code lives today.

cc @rust-lang/libs

4 files changed, 234 insertions, 103 deletions

diff --git a/src/libstd/sys/common/args.rs b/src/libstd/sys/common/args.rs
deleted file mode 100644
index b5330463e30..00000000000
--- a/src/libstd/sys/common/args.rs
+++ /dev/null
@@ -1,100 +0,0 @@
-// Copyright 2012-2015 The Rust Project Developers. See the COPYRIGHT
-// file at the top-level directory of this distribution and at
-// http://rust-lang.org/COPYRIGHT.
-//
-// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
-// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
-// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
-// option. This file may not be copied, modified, or distributed
-// except according to those terms.
-
-//! Global storage for command line arguments
-//!
-//! The current incarnation of the Rust runtime expects for
-//! the processes `argc` and `argv` arguments to be stored
-//! in a globally-accessible location for use by the `os` module.
-//!
-//! Only valid to call on Linux. Mac and Windows use syscalls to
-//! discover the command line arguments.
-//!
-//! FIXME #7756: Would be nice for this to not exist.
-
-#![allow(dead_code)] // different code on OSX/linux/etc
-
-/// One-time global initialization.
-pub unsafe fn init(argc: isize, argv: *const *const u8) { imp::init(argc, argv) }
-
-/// One-time global cleanup.
-pub unsafe fn cleanup() { imp::cleanup() }
-
-/// Make a clone of the global arguments.
-pub fn clone() -> Option<Vec<Vec<u8>>> { imp::clone() }
-
-#[cfg(any(target_os = "linux",
-          target_os = "android",
-          target_os = "freebsd",
-          target_os = "dragonfly",
-          target_os = "bitrig",
-          target_os = "netbsd",
-          target_os = "openbsd",
-          target_os = "solaris",
-          target_os = "emscripten",
-          target_os = "haiku"))]
-mod imp {
-    use libc::c_char;
-    use mem;
-    use ffi::CStr;
-
-    use sys_common::mutex::Mutex;
-
-    static mut GLOBAL_ARGS_PTR: usize = 0;
-    static LOCK: Mutex = Mutex::new();
-
-    pub unsafe fn init(argc: isize, argv: *const *const u8) {
-        let args = (0..argc).map(|i| {
-            CStr::from_ptr(*argv.offset(i) as *const c_char).to_bytes().to_vec()
-        }).collect();
-
-        LOCK.lock();
-        let ptr = get_global_ptr();
-        assert!((*ptr).is_none());
-        (*ptr) = Some(box args);
-        LOCK.unlock();
-    }
-
-    pub unsafe fn cleanup() {
-        LOCK.lock();
-        *get_global_ptr() = None;
-        LOCK.unlock();
-    }
-
-    pub fn clone() -> Option<Vec<Vec<u8>>> {
-        unsafe {
-            LOCK.lock();
-            let ptr = get_global_ptr();
-            let ret = (*ptr).as_ref().map(|s| (**s).clone());
-            LOCK.unlock();
-            return ret
-        }
-    }
-
-    fn get_global_ptr() -> *mut Option<Box<Vec<Vec<u8>>>> {
-        unsafe { mem::transmute(&GLOBAL_ARGS_PTR) }
-    }
-
-}
-
-#[cfg(any(target_os = "macos",
-          target_os = "ios",
-          target_os = "windows"))]
-mod imp {
-    pub unsafe fn init(_argc: isize, _argv: *const *const u8) {
-    }
-
-    pub fn cleanup() {
-    }
-
-    pub fn clone() -> Option<Vec<Vec<u8>>> {
-        panic!()
-    }
-}
diff --git a/src/libstd/sys/common/io.rs b/src/libstd/sys/common/io.rs
index 47cec4ef5c2..0483725dd83 100644
--- a/src/libstd/sys/common/io.rs
+++ b/src/libstd/sys/common/io.rs
@@ -50,7 +50,8 @@ pub unsafe fn read_to_end_uninitialized(r: &mut Read, buf: &mut Vec<u8>) -> io::
     }
 }
 
-#[cfg(all(test, not(target_os = "emscripten")))]
+#[cfg(test)]
+#[allow(dead_code)] // not used on emscripten
 pub mod test {
     use path::{Path, PathBuf};
     use env;
diff --git a/src/libstd/sys/common/memchr.rs b/src/libstd/sys/common/memchr.rs
new file mode 100644
index 00000000000..3824a5fb528
--- /dev/null
+++ b/src/libstd/sys/common/memchr.rs
@@ -0,0 +1,230 @@
+// Copyright 2015 The Rust Project Developers. See the COPYRIGHT
+// file at the top-level directory of this distribution and at
+// http://rust-lang.org/COPYRIGHT.
+//
+// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
+// option. This file may not be copied, modified, or distributed
+// except according to those terms.
+//
+// Original implementation taken from rust-memchr
+// Copyright 2015 Andrew Gallant, bluss and Nicolas Koch
+
+#[allow(dead_code)]
+pub mod fallback {
+    use cmp;
+    use mem;
+
+    const LO_U64: u64 = 0x0101010101010101;
+    const HI_U64: u64 = 0x8080808080808080;
+
+    // use truncation
+    const LO_USIZE: usize = LO_U64 as usize;
+    const HI_USIZE: usize = HI_U64 as usize;
+
+    /// Return `true` if `x` contains any zero byte.
+    ///
+    /// From *Matters Computational*, J. Arndt
+    ///
+    /// "The idea is to subtract one from each of the bytes and then look for
+    /// bytes where the borrow propagated all the way to the most significant
+    /// bit."
+    #[inline]
+    fn contains_zero_byte(x: usize) -> bool {
+        x.wrapping_sub(LO_USIZE) & !x & HI_USIZE != 0
+    }
+
+    #[cfg(target_pointer_width = "32")]
+    #[inline]
+    fn repeat_byte(b: u8) -> usize {
+        let mut rep = (b as usize) << 8 | b as usize;
+        rep = rep << 16 | rep;
+        rep
+    }
+
+    #[cfg(target_pointer_width = "64")]
+    #[inline]
+    fn repeat_byte(b: u8) -> usize {
+        let mut rep = (b as usize) << 8 | b as usize;
+        rep = rep << 16 | rep;
+        rep = rep << 32 | rep;
+        rep
+    }
+
+    /// Return the first index matching the byte `a` in `text`.
+    pub fn memchr(x: u8, text: &[u8]) -> Option<usize> {
+        // Scan for a single byte value by reading two `usize` words at a time.
+        //
+        // Split `text` in three parts
+        // - unaligned initial part, before the first word aligned address in text
+        // - body, scan by 2 words at a time
+        // - the last remaining part, < 2 word size
+        let len = text.len();
+        let ptr = text.as_ptr();
+        let usize_bytes = mem::size_of::<usize>();
+
+        // search up to an aligned boundary
+        let align = (ptr as usize) & (usize_bytes- 1);
+        let mut offset;
+        if align > 0 {
+            offset = cmp::min(usize_bytes - align, len);
+            if let Some(index) = text[..offset].iter().position(|elt| *elt == x) {
+                return Some(index);
+            }
+        } else {
+            offset = 0;
+        }
+
+        // search the body of the text
+        let repeated_x = repeat_byte(x);
+
+        if len >= 2 * usize_bytes {
+            while offset <= len - 2 * usize_bytes {
+                unsafe {
+                    let u = *(ptr.offset(offset as isize) as *const usize);
+                    let v = *(ptr.offset((offset + usize_bytes) as isize) as *const usize);
+
+                    // break if there is a matching byte
+                    let zu = contains_zero_byte(u ^ repeated_x);
+                    let zv = contains_zero_byte(v ^ repeated_x);
+                    if zu || zv {
+                        break;
+                    }
+                }
+                offset += usize_bytes * 2;
+            }
+        }
+
+        // find the byte after the point the body loop stopped
+        text[offset..].iter().position(|elt| *elt == x).map(|i| offset + i)
+    }
+
+    /// Return the last index matching the byte `a` in `text`.
+    pub fn memrchr(x: u8, text: &[u8]) -> Option<usize> {
+        // Scan for a single byte value by reading two `usize` words at a time.
+        //
+        // Split `text` in three parts
+        // - unaligned tail, after the last word aligned address in text
+        // - body, scan by 2 words at a time
+        // - the first remaining bytes, < 2 word size
+        let len = text.len();
+        let ptr = text.as_ptr();
+        let usize_bytes = mem::size_of::<usize>();
+
+        // search to an aligned boundary
+        let end_align = (ptr as usize + len) & (usize_bytes - 1);
+        let mut offset;
+        if end_align > 0 {
+            offset = if end_align >= len { 0 } else { len - end_align };
+            if let Some(index) = text[offset..].iter().rposition(|elt| *elt == x) {
+                return Some(offset + index);
+            }
+        } else {
+            offset = len;
+        }
+
+        // search the body of the text
+        let repeated_x = repeat_byte(x);
+
+        while offset >= 2 * usize_bytes {
+            unsafe {
+                let u = *(ptr.offset(offset as isize - 2 * usize_bytes as isize) as *const usize);
+                let v = *(ptr.offset(offset as isize - usize_bytes as isize) as *const usize);
+
+                // break if there is a matching byte
+                let zu = contains_zero_byte(u ^ repeated_x);
+                let zv = contains_zero_byte(v ^ repeated_x);
+                if zu || zv {
+                    break;
+                }
+            }
+            offset -= 2 * usize_bytes;
+        }
+
+        // find the byte before the point the body loop stopped
+        text[..offset].iter().rposition(|elt| *elt == x)
+    }
+
+    // test fallback implementations on all platforms
+    #[test]
+    fn matches_one() {
+        assert_eq!(Some(0), memchr(b'a', b"a"));
+    }
+
+    #[test]
+    fn matches_begin() {
+        assert_eq!(Some(0), memchr(b'a', b"aaaa"));
+    }
+
+    #[test]
+    fn matches_end() {
+        assert_eq!(Some(4), memchr(b'z', b"aaaaz"));
+    }
+
+    #[test]
+    fn matches_nul() {
+        assert_eq!(Some(4), memchr(b'\x00', b"aaaa\x00"));
+    }
+
+    #[test]
+    fn matches_past_nul() {
+        assert_eq!(Some(5), memchr(b'z', b"aaaa\x00z"));
+    }
+
+    #[test]
+    fn no_match_empty() {
+        assert_eq!(None, memchr(b'a', b""));
+    }
+
+    #[test]
+    fn no_match() {
+        assert_eq!(None, memchr(b'a', b"xyz"));
+    }
+
+    #[test]
+    fn matches_one_reversed() {
+        assert_eq!(Some(0), memrchr(b'a', b"a"));
+    }
+
+    #[test]
+    fn matches_begin_reversed() {
+        assert_eq!(Some(3), memrchr(b'a', b"aaaa"));
+    }
+
+    #[test]
+    fn matches_end_reversed() {
+        assert_eq!(Some(0), memrchr(b'z', b"zaaaa"));
+    }
+
+    #[test]
+    fn matches_nul_reversed() {
+        assert_eq!(Some(4), memrchr(b'\x00', b"aaaa\x00"));
+    }
+
+    #[test]
+    fn matches_past_nul_reversed() {
+        assert_eq!(Some(0), memrchr(b'z', b"z\x00aaaa"));
+    }
+
+    #[test]
+    fn no_match_empty_reversed() {
+        assert_eq!(None, memrchr(b'a', b""));
+    }
+
+    #[test]
+    fn no_match_reversed() {
+        assert_eq!(None, memrchr(b'a', b"xyz"));
+    }
+
+    #[test]
+    fn each_alignment_reversed() {
+        let mut data = [1u8; 64];
+        let needle = 2;
+        let pos = 40;
+        data[pos] = needle;
+        for start in 0..16 {
+            assert_eq!(Some(pos - start), memrchr(needle, &data[start..]));
+        }
+    }
+}
diff --git a/src/libstd/sys/common/mod.rs b/src/libstd/sys/common/mod.rs
index d1ca6765107..2845f895f18 100644
--- a/src/libstd/sys/common/mod.rs
+++ b/src/libstd/sys/common/mod.rs
@@ -25,12 +25,12 @@ macro_rules! rtassert {
     })
 }
 
-pub mod args;
 pub mod at_exit_imp;
 #[cfg(any(not(cargobuild), feature = "backtrace"))]
 pub mod backtrace;
 pub mod condvar;
 pub mod io;
+pub mod memchr;
 pub mod mutex;
 pub mod net;
 pub mod poison;
@@ -91,7 +91,7 @@ pub fn at_exit<F: FnOnce() + Send + 'static>(f: F) -> Result<(), ()> {
 pub fn cleanup() {
     static CLEANUP: Once = Once::new();
     CLEANUP.call_once(|| unsafe {
-        args::cleanup();
+        sys::args::cleanup();
         sys::stack_overflow::cleanup();
         at_exit_imp::cleanup();
     });