summaryrefslogtreecommitdiffstats
path: root/third_party/rust/byteorder/src
diff options
context:
space:
mode:
authorMatt A. Tobin <mattatobin@localhost.localdomain>2018-02-02 04:16:08 -0500
committerMatt A. Tobin <mattatobin@localhost.localdomain>2018-02-02 04:16:08 -0500
commit5f8de423f190bbb79a62f804151bc24824fa32d8 (patch)
tree10027f336435511475e392454359edea8e25895d /third_party/rust/byteorder/src
parent49ee0794b5d912db1f95dce6eb52d781dc210db5 (diff)
downloadUXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar
UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.gz
UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.lz
UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.tar.xz
UXP-5f8de423f190bbb79a62f804151bc24824fa32d8.zip
Add m-esr52 at 52.6.0
Diffstat (limited to 'third_party/rust/byteorder/src')
-rw-r--r--third_party/rust/byteorder/src/lib.rs802
-rw-r--r--third_party/rust/byteorder/src/new.rs269
2 files changed, 1071 insertions, 0 deletions
diff --git a/third_party/rust/byteorder/src/lib.rs b/third_party/rust/byteorder/src/lib.rs
new file mode 100644
index 000000000..a25bab6e0
--- /dev/null
+++ b/third_party/rust/byteorder/src/lib.rs
@@ -0,0 +1,802 @@
+/*!
+This crate provides convenience methods for encoding and decoding numbers
+in either big-endian or little-endian order.
+
+The organization of the crate is pretty simple. A trait, `ByteOrder`, specifies
+byte conversion methods for each type of number in Rust (sans numbers that have
+a platform dependent size like `usize` and `isize`). Two types, `BigEndian`
+and `LittleEndian` implement these methods. Finally, `ReadBytesExt` and
+`WriteBytesExt` provide convenience methods available to all types that
+implement `Read` and `Write`.
+
+# Examples
+
+Read unsigned 16 bit big-endian integers from a `Read` type:
+
+```rust
+use std::io::Cursor;
+use byteorder::{BigEndian, ReadBytesExt};
+
+let mut rdr = Cursor::new(vec![2, 5, 3, 0]);
+// Note that we use type parameters to indicate which kind of byte order
+// we want!
+assert_eq!(517, rdr.read_u16::<BigEndian>().unwrap());
+assert_eq!(768, rdr.read_u16::<BigEndian>().unwrap());
+```
+
+Write unsigned 16 bit little-endian integers to a `Write` type:
+
+```rust
+use byteorder::{LittleEndian, WriteBytesExt};
+
+let mut wtr = vec![];
+wtr.write_u16::<LittleEndian>(517).unwrap();
+wtr.write_u16::<LittleEndian>(768).unwrap();
+assert_eq!(wtr, vec![5, 2, 0, 3]);
+```
+*/
+
+#![crate_name = "byteorder"]
+#![doc(html_root_url = "http://burntsushi.net/rustdoc/byteorder")]
+
+#![cfg_attr(not(feature = "std"), no_std)]
+
+#![deny(missing_docs)]
+
+#[cfg(feature = "std")]
+extern crate core;
+
+use core::mem::transmute;
+use core::ptr::copy_nonoverlapping;
+
+#[cfg(feature = "std")]
+pub use new::{ReadBytesExt, WriteBytesExt};
+
+#[cfg(feature = "std")]
+mod new;
+
+#[inline]
+fn extend_sign(val: u64, nbytes: usize) -> i64 {
+ let shift = (8 - nbytes) * 8;
+ (val << shift) as i64 >> shift
+}
+
+#[inline]
+fn unextend_sign(val: i64, nbytes: usize) -> u64 {
+ let shift = (8 - nbytes) * 8;
+ (val << shift) as u64 >> shift
+}
+
+#[inline]
+fn pack_size(n: u64) -> usize {
+ if n < 1 << 8 {
+ 1
+ } else if n < 1 << 16 {
+ 2
+ } else if n < 1 << 24 {
+ 3
+ } else if n < 1 << 32 {
+ 4
+ } else if n < 1 << 40 {
+ 5
+ } else if n < 1 << 48 {
+ 6
+ } else if n < 1 << 56 {
+ 7
+ } else {
+ 8
+ }
+}
+
+/// ByteOrder describes types that can serialize integers as bytes.
+///
+/// Note that `Self` does not appear anywhere in this trait's definition!
+/// Therefore, in order to use it, you'll need to use syntax like
+/// `T::read_u16(&[0, 1])` where `T` implements `ByteOrder`.
+///
+/// This crate provides two types that implement `ByteOrder`: `BigEndian`
+/// and `LittleEndian`.
+///
+/// # Examples
+///
+/// Write and read `u32` numbers in little endian order:
+///
+/// ```rust
+/// use byteorder::{ByteOrder, LittleEndian};
+///
+/// let mut buf = [0; 4];
+/// LittleEndian::write_u32(&mut buf, 1_000_000);
+/// assert_eq!(1_000_000, LittleEndian::read_u32(&buf));
+/// ```
+///
+/// Write and read `i16` numbers in big endian order:
+///
+/// ```rust
+/// use byteorder::{ByteOrder, BigEndian};
+///
+/// let mut buf = [0; 2];
+/// BigEndian::write_i16(&mut buf, -50_000);
+/// assert_eq!(-50_000, BigEndian::read_i16(&buf));
+/// ```
+pub trait ByteOrder {
+ /// Reads an unsigned 16 bit integer from `buf`.
+ ///
+ /// Panics when `buf.len() < 2`.
+ fn read_u16(buf: &[u8]) -> u16;
+
+ /// Reads an unsigned 32 bit integer from `buf`.
+ ///
+ /// Panics when `buf.len() < 4`.
+ fn read_u32(buf: &[u8]) -> u32;
+
+ /// Reads an unsigned 64 bit integer from `buf`.
+ ///
+ /// Panics when `buf.len() < 8`.
+ fn read_u64(buf: &[u8]) -> u64;
+
+ /// Reads an unsigned n-bytes integer from `buf`.
+ ///
+ /// Panics when `nbytes < 1` or `nbytes > 8` or
+ /// `buf.len() < nbytes`
+ fn read_uint(buf: &[u8], nbytes: usize) -> u64;
+
+ /// Writes an unsigned 16 bit integer `n` to `buf`.
+ ///
+ /// Panics when `buf.len() < 2`.
+ fn write_u16(buf: &mut [u8], n: u16);
+
+ /// Writes an unsigned 32 bit integer `n` to `buf`.
+ ///
+ /// Panics when `buf.len() < 4`.
+ fn write_u32(buf: &mut [u8], n: u32);
+
+ /// Writes an unsigned 64 bit integer `n` to `buf`.
+ ///
+ /// Panics when `buf.len() < 8`.
+ fn write_u64(buf: &mut [u8], n: u64);
+
+ /// Writes an unsigned integer `n` to `buf` using only `nbytes`.
+ ///
+ /// If `n` is not representable in `nbytes`, or if `nbytes` is `> 8`, then
+ /// this method panics.
+ fn write_uint(buf: &mut [u8], n: u64, nbytes: usize);
+
+ /// Reads a signed 16 bit integer from `buf`.
+ ///
+ /// Panics when `buf.len() < 2`.
+ #[inline]
+ fn read_i16(buf: &[u8]) -> i16 {
+ Self::read_u16(buf) as i16
+ }
+
+ /// Reads a signed 32 bit integer from `buf`.
+ ///
+ /// Panics when `buf.len() < 4`.
+ #[inline]
+ fn read_i32(buf: &[u8]) -> i32 {
+ Self::read_u32(buf) as i32
+ }
+
+ /// Reads a signed 64 bit integer from `buf`.
+ ///
+ /// Panics when `buf.len() < 8`.
+ #[inline]
+ fn read_i64(buf: &[u8]) -> i64 {
+ Self::read_u64(buf) as i64
+ }
+
+ /// Reads a signed n-bytes integer from `buf`.
+ ///
+ /// Panics when `nbytes < 1` or `nbytes > 8` or
+ /// `buf.len() < nbytes`
+ #[inline]
+ fn read_int(buf: &[u8], nbytes: usize) -> i64 {
+ extend_sign(Self::read_uint(buf, nbytes), nbytes)
+ }
+
+ /// Reads a IEEE754 single-precision (4 bytes) floating point number.
+ ///
+ /// Panics when `buf.len() < 4`.
+ #[inline]
+ fn read_f32(buf: &[u8]) -> f32 {
+ unsafe { transmute(Self::read_u32(buf)) }
+ }
+
+ /// Reads a IEEE754 double-precision (8 bytes) floating point number.
+ ///
+ /// Panics when `buf.len() < 8`.
+ #[inline]
+ fn read_f64(buf: &[u8]) -> f64 {
+ unsafe { transmute(Self::read_u64(buf)) }
+ }
+
+ /// Writes a signed 16 bit integer `n` to `buf`.
+ ///
+ /// Panics when `buf.len() < 2`.
+ #[inline]
+ fn write_i16(buf: &mut [u8], n: i16) {
+ Self::write_u16(buf, n as u16)
+ }
+
+ /// Writes a signed 32 bit integer `n` to `buf`.
+ ///
+ /// Panics when `buf.len() < 4`.
+ #[inline]
+ fn write_i32(buf: &mut [u8], n: i32) {
+ Self::write_u32(buf, n as u32)
+ }
+
+ /// Writes a signed 64 bit integer `n` to `buf`.
+ ///
+ /// Panics when `buf.len() < 8`.
+ #[inline]
+ fn write_i64(buf: &mut [u8], n: i64) {
+ Self::write_u64(buf, n as u64)
+ }
+
+ /// Writes a signed integer `n` to `buf` using only `nbytes`.
+ ///
+ /// If `n` is not representable in `nbytes`, or if `nbytes` is `> 8`, then
+ /// this method panics.
+ #[inline]
+ fn write_int(buf: &mut [u8], n: i64, nbytes: usize) {
+ Self::write_uint(buf, unextend_sign(n, nbytes), nbytes)
+ }
+
+ /// Writes a IEEE754 single-precision (4 bytes) floating point number.
+ ///
+ /// Panics when `buf.len() < 4`.
+ #[inline]
+ fn write_f32(buf: &mut [u8], n: f32) {
+ Self::write_u32(buf, unsafe { transmute(n) })
+ }
+
+ /// Writes a IEEE754 double-precision (8 bytes) floating point number.
+ ///
+ /// Panics when `buf.len() < 8`.
+ #[inline]
+ fn write_f64(buf: &mut [u8], n: f64) {
+ Self::write_u64(buf, unsafe { transmute(n) })
+ }
+}
+
+/// Defines big-endian serialization.
+///
+/// Note that this type has no value constructor. It is used purely at the
+/// type level.
+#[allow(missing_copy_implementations)] pub enum BigEndian {}
+
+/// Defines little-endian serialization.
+///
+/// Note that this type has no value constructor. It is used purely at the
+/// type level.
+#[allow(missing_copy_implementations)] pub enum LittleEndian {}
+
+/// Defines network byte order serialization.
+///
+/// Network byte order is defined by [RFC 1700][1] to be big-endian, and is
+/// referred to in several protocol specifications. This type is an alias of
+/// BigEndian.
+///
+/// [1]: https://tools.ietf.org/html/rfc1700
+///
+/// Note that this type has no value constructor. It is used purely at the
+/// type level.
+pub type NetworkEndian = BigEndian;
+
+/// Defines system native-endian serialization.
+///
+/// Note that this type has no value constructor. It is used purely at the
+/// type level.
+#[cfg(target_endian = "little")]
+pub type NativeEndian = LittleEndian;
+
+/// Defines system native-endian serialization.
+///
+/// Note that this type has no value constructor. It is used purely at the
+/// type level.
+#[cfg(target_endian = "big")]
+pub type NativeEndian = BigEndian;
+
+macro_rules! read_num_bytes {
+ ($ty:ty, $size:expr, $src:expr, $which:ident) => ({
+ assert!($size == ::core::mem::size_of::<$ty>());
+ assert!($size <= $src.len());
+ let mut data: $ty = 0;
+ unsafe {
+ copy_nonoverlapping(
+ $src.as_ptr(),
+ &mut data as *mut $ty as *mut u8,
+ $size);
+ }
+ data.$which()
+ });
+}
+
+macro_rules! write_num_bytes {
+ ($ty:ty, $size:expr, $n:expr, $dst:expr, $which:ident) => ({
+ assert!($size <= $dst.len());
+ unsafe {
+ // N.B. https://github.com/rust-lang/rust/issues/22776
+ let bytes = transmute::<_, [u8; $size]>($n.$which());
+ copy_nonoverlapping((&bytes).as_ptr(), $dst.as_mut_ptr(), $size);
+ }
+ });
+}
+
+impl ByteOrder for BigEndian {
+ #[inline]
+ fn read_u16(buf: &[u8]) -> u16 {
+ read_num_bytes!(u16, 2, buf, to_be)
+ }
+
+ #[inline]
+ fn read_u32(buf: &[u8]) -> u32 {
+ read_num_bytes!(u32, 4, buf, to_be)
+ }
+
+ #[inline]
+ fn read_u64(buf: &[u8]) -> u64 {
+ read_num_bytes!(u64, 8, buf, to_be)
+ }
+
+ #[inline]
+ fn read_uint(buf: &[u8], nbytes: usize) -> u64 {
+ assert!(1 <= nbytes && nbytes <= 8 && nbytes <= buf.len());
+ let mut out = [0u8; 8];
+ let ptr_out = out.as_mut_ptr();
+ unsafe {
+ copy_nonoverlapping(
+ buf.as_ptr(), ptr_out.offset((8 - nbytes) as isize), nbytes);
+ (*(ptr_out as *const u64)).to_be()
+ }
+ }
+
+ #[inline]
+ fn write_u16(buf: &mut [u8], n: u16) {
+ write_num_bytes!(u16, 2, n, buf, to_be);
+ }
+
+ #[inline]
+ fn write_u32(buf: &mut [u8], n: u32) {
+ write_num_bytes!(u32, 4, n, buf, to_be);
+ }
+
+ #[inline]
+ fn write_u64(buf: &mut [u8], n: u64) {
+ write_num_bytes!(u64, 8, n, buf, to_be);
+ }
+
+ #[inline]
+ fn write_uint(buf: &mut [u8], n: u64, nbytes: usize) {
+ assert!(pack_size(n) <= nbytes && nbytes <= 8);
+ assert!(nbytes <= buf.len());
+ unsafe {
+ let bytes: [u8; 8] = transmute(n.to_be());
+ copy_nonoverlapping(
+ bytes.as_ptr().offset((8 - nbytes) as isize),
+ buf.as_mut_ptr(),
+ nbytes);
+ }
+ }
+}
+
+impl ByteOrder for LittleEndian {
+ #[inline]
+ fn read_u16(buf: &[u8]) -> u16 {
+ read_num_bytes!(u16, 2, buf, to_le)
+ }
+
+ #[inline]
+ fn read_u32(buf: &[u8]) -> u32 {
+ read_num_bytes!(u32, 4, buf, to_le)
+ }
+
+ #[inline]
+ fn read_u64(buf: &[u8]) -> u64 {
+ read_num_bytes!(u64, 8, buf, to_le)
+ }
+
+ #[inline]
+ fn read_uint(buf: &[u8], nbytes: usize) -> u64 {
+ assert!(1 <= nbytes && nbytes <= 8 && nbytes <= buf.len());
+ let mut out = [0u8; 8];
+ let ptr_out = out.as_mut_ptr();
+ unsafe {
+ copy_nonoverlapping(buf.as_ptr(), ptr_out, nbytes);
+ (*(ptr_out as *const u64)).to_le()
+ }
+ }
+
+ #[inline]
+ fn write_u16(buf: &mut [u8], n: u16) {
+ write_num_bytes!(u16, 2, n, buf, to_le);
+ }
+
+ #[inline]
+ fn write_u32(buf: &mut [u8], n: u32) {
+ write_num_bytes!(u32, 4, n, buf, to_le);
+ }
+
+ #[inline]
+ fn write_u64(buf: &mut [u8], n: u64) {
+ write_num_bytes!(u64, 8, n, buf, to_le);
+ }
+
+ #[inline]
+ fn write_uint(buf: &mut [u8], n: u64, nbytes: usize) {
+ assert!(pack_size(n as u64) <= nbytes && nbytes <= 8);
+ assert!(nbytes <= buf.len());
+ unsafe {
+ let bytes: [u8; 8] = transmute(n.to_le());
+ copy_nonoverlapping(bytes.as_ptr(), buf.as_mut_ptr(), nbytes);
+ }
+ }
+}
+
+#[cfg(test)]
+mod test {
+ extern crate quickcheck;
+ extern crate rand;
+
+ use test::rand::thread_rng;
+ use test::quickcheck::{QuickCheck, StdGen, Testable};
+
+ const U64_MAX: u64 = ::std::u64::MAX;
+ const I64_MAX: u64 = ::std::i64::MAX as u64;
+
+ fn qc_sized<A: Testable>(f: A, size: u64) {
+ QuickCheck::new()
+ .gen(StdGen::new(thread_rng(), size as usize))
+ .tests(1_00)
+ .max_tests(10_000)
+ .quickcheck(f);
+ }
+
+ macro_rules! qc_byte_order {
+ ($name:ident, $ty_int:ident, $max:expr,
+ $bytes:expr, $read:ident, $write:ident) => (
+ mod $name {
+ use {BigEndian, ByteOrder, NativeEndian, LittleEndian};
+ use super::qc_sized;
+
+ #[test]
+ fn big_endian() {
+ let max = ($max - 1) >> (8 * (8 - $bytes));
+ fn prop(n: $ty_int) -> bool {
+ let mut buf = [0; 8];
+ BigEndian::$write(&mut buf, n, $bytes);
+ n == BigEndian::$read(&mut buf[..$bytes], $bytes)
+ }
+ qc_sized(prop as fn($ty_int) -> bool, max);
+ }
+
+ #[test]
+ fn little_endian() {
+ let max = ($max - 1) >> (8 * (8 - $bytes));
+ fn prop(n: $ty_int) -> bool {
+ let mut buf = [0; 8];
+ LittleEndian::$write(&mut buf, n, $bytes);
+ n == LittleEndian::$read(&mut buf[..$bytes], $bytes)
+ }
+ qc_sized(prop as fn($ty_int) -> bool, max);
+ }
+
+ #[test]
+ fn native_endian() {
+ let max = ($max - 1) >> (8 * (8 - $bytes));
+ fn prop(n: $ty_int) -> bool {
+ let mut buf = [0; 8];
+ NativeEndian::$write(&mut buf, n, $bytes);
+ n == NativeEndian::$read(&mut buf[..$bytes], $bytes)
+ }
+ qc_sized(prop as fn($ty_int) -> bool, max);
+ }
+ }
+ );
+ ($name:ident, $ty_int:ident, $max:expr,
+ $read:ident, $write:ident) => (
+ mod $name {
+ use std::mem::size_of;
+ use {BigEndian, ByteOrder, NativeEndian, LittleEndian};
+ use super::qc_sized;
+
+ #[test]
+ fn big_endian() {
+ fn prop(n: $ty_int) -> bool {
+ let bytes = size_of::<$ty_int>();
+ let mut buf = [0; 8];
+ BigEndian::$write(&mut buf[8 - bytes..], n);
+ n == BigEndian::$read(&mut buf[8 - bytes..])
+ }
+ qc_sized(prop as fn($ty_int) -> bool, $max - 1);
+ }
+
+ #[test]
+ fn little_endian() {
+ fn prop(n: $ty_int) -> bool {
+ let bytes = size_of::<$ty_int>();
+ let mut buf = [0; 8];
+ LittleEndian::$write(&mut buf[..bytes], n);
+ n == LittleEndian::$read(&mut buf[..bytes])
+ }
+ qc_sized(prop as fn($ty_int) -> bool, $max - 1);
+ }
+
+ #[test]
+ fn native_endian() {
+ fn prop(n: $ty_int) -> bool {
+ let bytes = size_of::<$ty_int>();
+ let mut buf = [0; 8];
+ NativeEndian::$write(&mut buf[..bytes], n);
+ n == NativeEndian::$read(&mut buf[..bytes])
+ }
+ qc_sized(prop as fn($ty_int) -> bool, $max - 1);
+ }
+ }
+ );
+ }
+
+ qc_byte_order!(prop_u16, u16, ::std::u16::MAX as u64, read_u16, write_u16);
+ qc_byte_order!(prop_i16, i16, ::std::i16::MAX as u64, read_i16, write_i16);
+ qc_byte_order!(prop_u32, u32, ::std::u32::MAX as u64, read_u32, write_u32);
+ qc_byte_order!(prop_i32, i32, ::std::i32::MAX as u64, read_i32, write_i32);
+ qc_byte_order!(prop_u64, u64, ::std::u64::MAX as u64, read_u64, write_u64);
+ qc_byte_order!(prop_i64, i64, ::std::i64::MAX as u64, read_i64, write_i64);
+ qc_byte_order!(prop_f32, f32, ::std::u64::MAX as u64, read_f32, write_f32);
+ qc_byte_order!(prop_f64, f64, ::std::i64::MAX as u64, read_f64, write_f64);
+
+ qc_byte_order!(prop_uint_1, u64, super::U64_MAX, 1, read_uint, write_uint);
+ qc_byte_order!(prop_uint_2, u64, super::U64_MAX, 2, read_uint, write_uint);
+ qc_byte_order!(prop_uint_3, u64, super::U64_MAX, 3, read_uint, write_uint);
+ qc_byte_order!(prop_uint_4, u64, super::U64_MAX, 4, read_uint, write_uint);
+ qc_byte_order!(prop_uint_5, u64, super::U64_MAX, 5, read_uint, write_uint);
+ qc_byte_order!(prop_uint_6, u64, super::U64_MAX, 6, read_uint, write_uint);
+ qc_byte_order!(prop_uint_7, u64, super::U64_MAX, 7, read_uint, write_uint);
+ qc_byte_order!(prop_uint_8, u64, super::U64_MAX, 8, read_uint, write_uint);
+
+ qc_byte_order!(prop_int_1, i64, super::I64_MAX, 1, read_int, write_int);
+ qc_byte_order!(prop_int_2, i64, super::I64_MAX, 2, read_int, write_int);
+ qc_byte_order!(prop_int_3, i64, super::I64_MAX, 3, read_int, write_int);
+ qc_byte_order!(prop_int_4, i64, super::I64_MAX, 4, read_int, write_int);
+ qc_byte_order!(prop_int_5, i64, super::I64_MAX, 5, read_int, write_int);
+ qc_byte_order!(prop_int_6, i64, super::I64_MAX, 6, read_int, write_int);
+ qc_byte_order!(prop_int_7, i64, super::I64_MAX, 7, read_int, write_int);
+ qc_byte_order!(prop_int_8, i64, super::I64_MAX, 8, read_int, write_int);
+
+ macro_rules! qc_bytes_ext {
+ ($name:ident, $ty_int:ident, $max:expr,
+ $bytes:expr, $read:ident, $write:ident) => (
+ mod $name {
+ use std::io::Cursor;
+ use {
+ ReadBytesExt, WriteBytesExt,
+ BigEndian, NativeEndian, LittleEndian,
+ };
+ use super::qc_sized;
+
+ #[test]
+ fn big_endian() {
+ let max = ($max - 1) >> (8 * (8 - $bytes));
+ fn prop(n: $ty_int) -> bool {
+ let mut wtr = vec![];
+ wtr.$write::<BigEndian>(n).unwrap();
+ let mut rdr = Vec::new();
+ rdr.extend(wtr[8 - $bytes..].iter().map(|&x|x));
+ let mut rdr = Cursor::new(rdr);
+ n == rdr.$read::<BigEndian>($bytes).unwrap()
+ }
+ qc_sized(prop as fn($ty_int) -> bool, max);
+ }
+
+ #[test]
+ fn little_endian() {
+ let max = ($max - 1) >> (8 * (8 - $bytes));
+ fn prop(n: $ty_int) -> bool {
+ let mut wtr = vec![];
+ wtr.$write::<LittleEndian>(n).unwrap();
+ let mut rdr = Cursor::new(wtr);
+ n == rdr.$read::<LittleEndian>($bytes).unwrap()
+ }
+ qc_sized(prop as fn($ty_int) -> bool, max);
+ }
+
+ #[test]
+ fn native_endian() {
+ let max = ($max - 1) >> (8 * (8 - $bytes));
+ fn prop(n: $ty_int) -> bool {
+ let mut wtr = vec![];
+ wtr.$write::<NativeEndian>(n).unwrap();
+ let mut rdr = Cursor::new(wtr);
+ n == rdr.$read::<NativeEndian>($bytes).unwrap()
+ }
+ qc_sized(prop as fn($ty_int) -> bool, max);
+ }
+ }
+ );
+ ($name:ident, $ty_int:ident, $max:expr, $read:ident, $write:ident) => (
+ mod $name {
+ use std::io::Cursor;
+ use {
+ ReadBytesExt, WriteBytesExt,
+ BigEndian, NativeEndian, LittleEndian,
+ };
+ use super::qc_sized;
+
+ #[test]
+ fn big_endian() {
+ fn prop(n: $ty_int) -> bool {
+ let mut wtr = vec![];
+ wtr.$write::<BigEndian>(n).unwrap();
+ let mut rdr = Cursor::new(wtr);
+ n == rdr.$read::<BigEndian>().unwrap()
+ }
+ qc_sized(prop as fn($ty_int) -> bool, $max - 1);
+ }
+
+ #[test]
+ fn little_endian() {
+ fn prop(n: $ty_int) -> bool {
+ let mut wtr = vec![];
+ wtr.$write::<LittleEndian>(n).unwrap();
+ let mut rdr = Cursor::new(wtr);
+ n == rdr.$read::<LittleEndian>().unwrap()
+ }
+ qc_sized(prop as fn($ty_int) -> bool, $max - 1);
+ }
+
+ #[test]
+ fn native_endian() {
+ fn prop(n: $ty_int) -> bool {
+ let mut wtr = vec![];
+ wtr.$write::<NativeEndian>(n).unwrap();
+ let mut rdr = Cursor::new(wtr);
+ n == rdr.$read::<NativeEndian>().unwrap()
+ }
+ qc_sized(prop as fn($ty_int) -> bool, $max - 1);
+ }
+ }
+ );
+ }
+
+ qc_bytes_ext!(prop_ext_u16, u16, ::std::u16::MAX as u64, read_u16, write_u16);
+ qc_bytes_ext!(prop_ext_i16, i16, ::std::i16::MAX as u64, read_i16, write_i16);
+ qc_bytes_ext!(prop_ext_u32, u32, ::std::u32::MAX as u64, read_u32, write_u32);
+ qc_bytes_ext!(prop_ext_i32, i32, ::std::i32::MAX as u64, read_i32, write_i32);
+ qc_bytes_ext!(prop_ext_u64, u64, ::std::u64::MAX as u64, read_u64, write_u64);
+ qc_bytes_ext!(prop_ext_i64, i64, ::std::i64::MAX as u64, read_i64, write_i64);
+ qc_bytes_ext!(prop_ext_f32, f32, ::std::u64::MAX as u64, read_f32, write_f32);
+ qc_bytes_ext!(prop_ext_f64, f64, ::std::i64::MAX as u64, read_f64, write_f64);
+
+ qc_bytes_ext!(prop_ext_uint_1, u64, super::U64_MAX, 1, read_uint, write_u64);
+ qc_bytes_ext!(prop_ext_uint_2, u64, super::U64_MAX, 2, read_uint, write_u64);
+ qc_bytes_ext!(prop_ext_uint_3, u64, super::U64_MAX, 3, read_uint, write_u64);
+ qc_bytes_ext!(prop_ext_uint_4, u64, super::U64_MAX, 4, read_uint, write_u64);
+ qc_bytes_ext!(prop_ext_uint_5, u64, super::U64_MAX, 5, read_uint, write_u64);
+ qc_bytes_ext!(prop_ext_uint_6, u64, super::U64_MAX, 6, read_uint, write_u64);
+ qc_bytes_ext!(prop_ext_uint_7, u64, super::U64_MAX, 7, read_uint, write_u64);
+ qc_bytes_ext!(prop_ext_uint_8, u64, super::U64_MAX, 8, read_uint, write_u64);
+
+ qc_bytes_ext!(prop_ext_int_1, i64, super::I64_MAX, 1, read_int, write_i64);
+ qc_bytes_ext!(prop_ext_int_2, i64, super::I64_MAX, 2, read_int, write_i64);
+ qc_bytes_ext!(prop_ext_int_3, i64, super::I64_MAX, 3, read_int, write_i64);
+ qc_bytes_ext!(prop_ext_int_4, i64, super::I64_MAX, 4, read_int, write_i64);
+ qc_bytes_ext!(prop_ext_int_5, i64, super::I64_MAX, 5, read_int, write_i64);
+ qc_bytes_ext!(prop_ext_int_6, i64, super::I64_MAX, 6, read_int, write_i64);
+ qc_bytes_ext!(prop_ext_int_7, i64, super::I64_MAX, 7, read_int, write_i64);
+ qc_bytes_ext!(prop_ext_int_8, i64, super::I64_MAX, 8, read_int, write_i64);
+
+ // Test that all of the byte conversion functions panic when given a
+ // buffer that is too small.
+ //
+ // These tests are critical to ensure safety, otherwise we might end up
+ // with a buffer overflow.
+ macro_rules! too_small {
+ ($name:ident, $maximally_small:expr, $zero:expr,
+ $read:ident, $write:ident) => (
+ mod $name {
+ use {BigEndian, ByteOrder, NativeEndian, LittleEndian};
+
+ #[test]
+ #[should_panic]
+ fn read_big_endian() {
+ let buf = [0; $maximally_small];
+ BigEndian::$read(&buf);
+ }
+
+ #[test]
+ #[should_panic]
+ fn read_little_endian() {
+ let buf = [0; $maximally_small];
+ LittleEndian::$read(&buf);
+ }
+
+ #[test]
+ #[should_panic]
+ fn read_native_endian() {
+ let buf = [0; $maximally_small];
+ NativeEndian::$read(&buf);
+ }
+
+ #[test]
+ #[should_panic]
+ fn write_big_endian() {
+ let mut buf = [0; $maximally_small];
+ BigEndian::$write(&mut buf, $zero);
+ }
+
+ #[test]
+ #[should_panic]
+ fn write_little_endian() {
+ let mut buf = [0; $maximally_small];
+ LittleEndian::$write(&mut buf, $zero);
+ }
+
+ #[test]
+ #[should_panic]
+ fn write_native_endian() {
+ let mut buf = [0; $maximally_small];
+ NativeEndian::$write(&mut buf, $zero);
+ }
+ }
+ );
+ ($name:ident, $maximally_small:expr, $read:ident) => (
+ mod $name {
+ use {BigEndian, ByteOrder, NativeEndian, LittleEndian};
+
+ #[test]
+ #[should_panic]
+ fn read_big_endian() {
+ let buf = [0; $maximally_small];
+ BigEndian::$read(&buf, $maximally_small + 1);
+ }
+
+ #[test]
+ #[should_panic]
+ fn read_little_endian() {
+ let buf = [0; $maximally_small];
+ LittleEndian::$read(&buf, $maximally_small + 1);
+ }
+
+ #[test]
+ #[should_panic]
+ fn read_native_endian() {
+ let buf = [0; $maximally_small];
+ NativeEndian::$read(&buf, $maximally_small + 1);
+ }
+ }
+ );
+ }
+
+ too_small!(small_u16, 1, 0, read_u16, write_u16);
+ too_small!(small_i16, 1, 0, read_i16, write_i16);
+ too_small!(small_u32, 3, 0, read_u32, write_u32);
+ too_small!(small_i32, 3, 0, read_i32, write_i32);
+ too_small!(small_u64, 7, 0, read_u64, write_u64);
+ too_small!(small_i64, 7, 0, read_i64, write_i64);
+ too_small!(small_f32, 3, 0.0, read_f32, write_f32);
+ too_small!(small_f64, 7, 0.0, read_f64, write_f64);
+
+ too_small!(small_uint_1, 1, read_uint);
+ too_small!(small_uint_2, 2, read_uint);
+ too_small!(small_uint_3, 3, read_uint);
+ too_small!(small_uint_4, 4, read_uint);
+ too_small!(small_uint_5, 5, read_uint);
+ too_small!(small_uint_6, 6, read_uint);
+ too_small!(small_uint_7, 7, read_uint);
+
+ too_small!(small_int_1, 1, read_int);
+ too_small!(small_int_2, 2, read_int);
+ too_small!(small_int_3, 3, read_int);
+ too_small!(small_int_4, 4, read_int);
+ too_small!(small_int_5, 5, read_int);
+ too_small!(small_int_6, 6, read_int);
+ too_small!(small_int_7, 7, read_int);
+
+ #[test]
+ fn uint_bigger_buffer() {
+ use {ByteOrder, LittleEndian};
+ let n = LittleEndian::read_uint(&[1, 2, 3, 4, 5, 6, 7, 8], 5);
+ assert_eq!(n, 0x0504030201);
+ }
+}
diff --git a/third_party/rust/byteorder/src/new.rs b/third_party/rust/byteorder/src/new.rs
new file mode 100644
index 000000000..b5f37700f
--- /dev/null
+++ b/third_party/rust/byteorder/src/new.rs
@@ -0,0 +1,269 @@
+use std::io::{self, Result};
+
+use ByteOrder;
+
+/// Extends `Read` with methods for reading numbers. (For `std::io`.)
+///
+/// Most of the methods defined here have an unconstrained type parameter that
+/// must be explicitly instantiated. Typically, it is instantiated with either
+/// the `BigEndian` or `LittleEndian` types defined in this crate.
+///
+/// # Examples
+///
+/// Read unsigned 16 bit big-endian integers from a `Read`:
+///
+/// ```rust
+/// use std::io::Cursor;
+/// use byteorder::{BigEndian, ReadBytesExt};
+///
+/// let mut rdr = Cursor::new(vec![2, 5, 3, 0]);
+/// assert_eq!(517, rdr.read_u16::<BigEndian>().unwrap());
+/// assert_eq!(768, rdr.read_u16::<BigEndian>().unwrap());
+/// ```
+pub trait ReadBytesExt: io::Read {
+ /// Reads an unsigned 8 bit integer from the underlying reader.
+ ///
+ /// Note that since this reads a single byte, no byte order conversions
+ /// are used. It is included for completeness.
+ #[inline]
+ fn read_u8(&mut self) -> Result<u8> {
+ let mut buf = [0; 1];
+ try!(self.read_exact(&mut buf));
+ Ok(buf[0])
+ }
+
+ /// Reads a signed 8 bit integer from the underlying reader.
+ ///
+ /// Note that since this reads a single byte, no byte order conversions
+ /// are used. It is included for completeness.
+ #[inline]
+ fn read_i8(&mut self) -> Result<i8> {
+ let mut buf = [0; 1];
+ try!(self.read_exact(&mut buf));
+ Ok(buf[0] as i8)
+ }
+
+ /// Reads an unsigned 16 bit integer from the underlying reader.
+ #[inline]
+ fn read_u16<T: ByteOrder>(&mut self) -> Result<u16> {
+ let mut buf = [0; 2];
+ try!(self.read_exact(&mut buf));
+ Ok(T::read_u16(&buf))
+ }
+
+ /// Reads a signed 16 bit integer from the underlying reader.
+ #[inline]
+ fn read_i16<T: ByteOrder>(&mut self) -> Result<i16> {
+ let mut buf = [0; 2];
+ try!(self.read_exact(&mut buf));
+ Ok(T::read_i16(&buf))
+ }
+
+ /// Reads an unsigned 32 bit integer from the underlying reader.
+ #[inline]
+ fn read_u32<T: ByteOrder>(&mut self) -> Result<u32> {
+ let mut buf = [0; 4];
+ try!(self.read_exact(&mut buf));
+ Ok(T::read_u32(&buf))
+ }
+
+ /// Reads a signed 32 bit integer from the underlying reader.
+ #[inline]
+ fn read_i32<T: ByteOrder>(&mut self) -> Result<i32> {
+ let mut buf = [0; 4];
+ try!(self.read_exact(&mut buf));
+ Ok(T::read_i32(&buf))
+ }
+
+ /// Reads an unsigned 64 bit integer from the underlying reader.
+ #[inline]
+ fn read_u64<T: ByteOrder>(&mut self) -> Result<u64> {
+ let mut buf = [0; 8];
+ try!(self.read_exact(&mut buf));
+ Ok(T::read_u64(&buf))
+ }
+
+ /// Reads a signed 64 bit integer from the underlying reader.
+ #[inline]
+ fn read_i64<T: ByteOrder>(&mut self) -> Result<i64> {
+ let mut buf = [0; 8];
+ try!(self.read_exact(&mut buf));
+ Ok(T::read_i64(&buf))
+ }
+
+ /// Reads an unsigned n-bytes integer from the underlying reader.
+ #[inline]
+ fn read_uint<T: ByteOrder>(&mut self, nbytes: usize) -> Result<u64> {
+ let mut buf = [0; 8];
+ try!(self.read_exact(&mut buf[..nbytes]));
+ Ok(T::read_uint(&buf[..nbytes], nbytes))
+ }
+
+ /// Reads a signed n-bytes integer from the underlying reader.
+ #[inline]
+ fn read_int<T: ByteOrder>(&mut self, nbytes: usize) -> Result<i64> {
+ let mut buf = [0; 8];
+ try!(self.read_exact(&mut buf[..nbytes]));
+ Ok(T::read_int(&buf[..nbytes], nbytes))
+ }
+
+ /// Reads a IEEE754 single-precision (4 bytes) floating point number from
+ /// the underlying reader.
+ #[inline]
+ fn read_f32<T: ByteOrder>(&mut self) -> Result<f32> {
+ let mut buf = [0; 4];
+ try!(self.read_exact(&mut buf));
+ Ok(T::read_f32(&buf))
+ }
+
+ /// Reads a IEEE754 double-precision (8 bytes) floating point number from
+ /// the underlying reader.
+ #[inline]
+ fn read_f64<T: ByteOrder>(&mut self) -> Result<f64> {
+ let mut buf = [0; 8];
+ try!(self.read_exact(&mut buf));
+ Ok(T::read_f64(&buf))
+ }
+}
+
+/// All types that implement `Read` get methods defined in `ReadBytesExt`
+/// for free.
+impl<R: io::Read + ?Sized> ReadBytesExt for R {}
+
+/// Extends `Write` with methods for writing numbers. (For `std::io`.)
+///
+/// Most of the methods defined here have an unconstrained type parameter that
+/// must be explicitly instantiated. Typically, it is instantiated with either
+/// the `BigEndian` or `LittleEndian` types defined in this crate.
+///
+/// # Examples
+///
+/// Write unsigned 16 bit big-endian integers to a `Write`:
+///
+/// ```rust
+/// use byteorder::{BigEndian, WriteBytesExt};
+///
+/// let mut wtr = vec![];
+/// wtr.write_u16::<BigEndian>(517).unwrap();
+/// wtr.write_u16::<BigEndian>(768).unwrap();
+/// assert_eq!(wtr, vec![2, 5, 3, 0]);
+/// ```
+pub trait WriteBytesExt: io::Write {
+ /// Writes an unsigned 8 bit integer to the underlying writer.
+ ///
+ /// Note that since this writes a single byte, no byte order conversions
+ /// are used. It is included for completeness.
+ #[inline]
+ fn write_u8(&mut self, n: u8) -> Result<()> {
+ self.write_all(&[n])
+ }
+
+ /// Writes a signed 8 bit integer to the underlying writer.
+ ///
+ /// Note that since this writes a single byte, no byte order conversions
+ /// are used. It is included for completeness.
+ #[inline]
+ fn write_i8(&mut self, n: i8) -> Result<()> {
+ self.write_all(&[n as u8])
+ }
+
+ /// Writes an unsigned 16 bit integer to the underlying writer.
+ #[inline]
+ fn write_u16<T: ByteOrder>(&mut self, n: u16) -> Result<()> {
+ let mut buf = [0; 2];
+ T::write_u16(&mut buf, n);
+ self.write_all(&buf)
+ }
+
+ /// Writes a signed 16 bit integer to the underlying writer.
+ #[inline]
+ fn write_i16<T: ByteOrder>(&mut self, n: i16) -> Result<()> {
+ let mut buf = [0; 2];
+ T::write_i16(&mut buf, n);
+ self.write_all(&buf)
+ }
+
+ /// Writes an unsigned 32 bit integer to the underlying writer.
+ #[inline]
+ fn write_u32<T: ByteOrder>(&mut self, n: u32) -> Result<()> {
+ let mut buf = [0; 4];
+ T::write_u32(&mut buf, n);
+ self.write_all(&buf)
+ }
+
+ /// Writes a signed 32 bit integer to the underlying writer.
+ #[inline]
+ fn write_i32<T: ByteOrder>(&mut self, n: i32) -> Result<()> {
+ let mut buf = [0; 4];
+ T::write_i32(&mut buf, n);
+ self.write_all(&buf)
+ }
+
+ /// Writes an unsigned 64 bit integer to the underlying writer.
+ #[inline]
+ fn write_u64<T: ByteOrder>(&mut self, n: u64) -> Result<()> {
+ let mut buf = [0; 8];
+ T::write_u64(&mut buf, n);
+ self.write_all(&buf)
+ }
+
+ /// Writes a signed 64 bit integer to the underlying writer.
+ #[inline]
+ fn write_i64<T: ByteOrder>(&mut self, n: i64) -> Result<()> {
+ let mut buf = [0; 8];
+ T::write_i64(&mut buf, n);
+ self.write_all(&buf)
+ }
+
+ /// Writes an unsigned n-bytes integer to the underlying writer.
+ ///
+ /// If the given integer is not representable in the given number of bytes,
+ /// this method panics. If `nbytes > 8`, this method panics.
+ #[inline]
+ fn write_uint<T: ByteOrder>(
+ &mut self,
+ n: u64,
+ nbytes: usize,
+ ) -> Result<()> {
+ let mut buf = [0; 8];
+ T::write_uint(&mut buf, n, nbytes);
+ self.write_all(&buf[0..nbytes])
+ }
+
+ /// Writes a signed n-bytes integer to the underlying writer.
+ ///
+ /// If the given integer is not representable in the given number of bytes,
+ /// this method panics. If `nbytes > 8`, this method panics.
+ #[inline]
+ fn write_int<T: ByteOrder>(
+ &mut self,
+ n: i64,
+ nbytes: usize,
+ ) -> Result<()> {
+ let mut buf = [0; 8];
+ T::write_int(&mut buf, n, nbytes);
+ self.write_all(&buf[0..nbytes])
+ }
+
+ /// Writes a IEEE754 single-precision (4 bytes) floating point number to
+ /// the underlying writer.
+ #[inline]
+ fn write_f32<T: ByteOrder>(&mut self, n: f32) -> Result<()> {
+ let mut buf = [0; 4];
+ T::write_f32(&mut buf, n);
+ self.write_all(&buf)
+ }
+
+ /// Writes a IEEE754 double-precision (8 bytes) floating point number to
+ /// the underlying writer.
+ #[inline]
+ fn write_f64<T: ByteOrder>(&mut self, n: f64) -> Result<()> {
+ let mut buf = [0; 8];
+ T::write_f64(&mut buf, n);
+ self.write_all(&buf)
+ }
+}
+
+/// All types that implement `Write` get methods defined in `WriteBytesExt`
+/// for free.
+impl<W: io::Write + ?Sized> WriteBytesExt for W {}