Update nitrokey crate to 0.4.0

This change finally updates the version of the nitrokey crate that we
consume to 0.4.0. Along with that we update rand_core, one of its
dependencies, to 0.5.1. Further more we add cfg-if in version 0.1.10 and
getrandom in version 0.1.13, both of which are now new (non-development)
dependencies.

Import subrepo nitrokey/:nitrokey at e81057037e9b4f370b64c0a030a725bc6bdfb870
Import subrepo cfg-if/:cfg-if at 4484a6faf816ff8058088ad857b0c6bb2f4b02b2
Import subrepo getrandom/:getrandom at d661aa7e1b8cc80b47dabe3d2135b3b47d2858af
Import subrepo rand/:rand at d877ed528248b52d947e0484364a4e1ae59ca502

3 files changed, 157 insertions, 67 deletions

diff --git a/rand/rand_pcg/src/lib.rs b/rand/rand_pcg/src/lib.rs
index 9648e85..22ba4a0 100644
--- a/rand/rand_pcg/src/lib.rs
+++ b/rand/rand_pcg/src/lib.rs
@@ -17,11 +17,12 @@
 //! -   `Pcg32` aka `Lcg64Xsh32`, officially known as `pcg32`, a general
 //!     purpose RNG. This is a good choice on both 32-bit and 64-bit CPUs
 //!     (for 32-bit output).
-//! -   `Pcg64Mcg` aka `Mcg128Xsl64`, officially known as `mcg_xsl_rr_128_64`,
+//! -   `Pcg64` aka `Lcg128Xsl64`, officially known as `pcg64`, a general
+//!     purpose RNG. This is a good choice on 64-bit CPUs.
+//! -   `Pcg64Mcg` aka `Mcg128Xsl64`, officially known as `pcg64_fast`,
 //!     a general purpose RNG using 128-bit multiplications. This has poor
 //!     performance on 32-bit CPUs but is a good choice on 64-bit CPUs for
-//!     both 32-bit and 64-bit output. (Note: this RNG is only available using
-//!     Rust 1.26 or later.)
+//!     both 32-bit and 64-bit output.
 //!
 //! Both of these use 16 bytes of state and 128-bit seeds, and are considered
 //! value-stable (i.e. any change affecting the output given a fixed seed would
@@ -34,15 +35,15 @@
 #![deny(missing_docs)]
 #![deny(missing_debug_implementations)]
 
-#![no_std]
-
-pub extern crate rand_core;
+#![allow(clippy::unreadable_literal)]
 
-#[cfg(feature="serde1")] extern crate serde;
-#[cfg(feature="serde1")] #[macro_use] extern crate serde_derive;
+#![no_std]
 
 mod pcg64;
-#[cfg(all(rustc_1_26, not(target_os = "emscripten")))] mod pcg128;
+#[cfg(not(target_os = "emscripten"))] mod pcg128;
 
 pub use self::pcg64::{Pcg32, Lcg64Xsh32};
-#[cfg(all(rustc_1_26, not(target_os = "emscripten")))] pub use self::pcg128::{Pcg64Mcg, Mcg128Xsl64};
+#[cfg(not(target_os = "emscripten"))] pub use self::pcg128::{
+        Pcg64, Lcg128Xsl64,
+        Pcg64Mcg, Mcg128Xsl64,
+};
diff --git a/rand/rand_pcg/src/pcg128.rs b/rand/rand_pcg/src/pcg128.rs
index 9aff506..311a41b 100644
--- a/rand/rand_pcg/src/pcg128.rs
+++ b/rand/rand_pcg/src/pcg128.rs
@@ -14,8 +14,109 @@
 const MULTIPLIER: u128 = 0x2360_ED05_1FC6_5DA4_4385_DF64_9FCC_F645;
 
 use core::fmt;
-use core::mem::transmute;
 use rand_core::{RngCore, SeedableRng, Error, le};
+#[cfg(feature="serde1")] use serde::{Serialize, Deserialize};
+
+/// A PCG random number generator (XSL RR 128/64 (LCG) variant).
+///
+/// Permuted Congruential Generator with 128-bit state, internal Linear
+/// Congruential Generator, and 64-bit output via "xorshift low (bits),
+/// random rotation" output function.
+///
+/// This is a 128-bit LCG with explicitly chosen stream with the PCG-XSL-RR
+/// output function. This combination is the standard `pcg64`.
+///
+/// Despite the name, this implementation uses 32 bytes (256 bit) space
+/// comprising 128 bits of state and 128 bits stream selector. These are both
+/// set by `SeedableRng`, using a 256-bit seed.
+#[derive(Clone)]
+#[cfg_attr(feature="serde1", derive(Serialize,Deserialize))]
+pub struct Lcg128Xsl64 {
+    state: u128,
+    increment: u128,
+}
+
+/// `Lcg128Xsl64` is also officially known as `pcg64`.
+pub type Pcg64 = Lcg128Xsl64;
+
+impl Lcg128Xsl64 {
+    /// Construct an instance compatible with PCG seed and stream.
+    ///
+    /// Note that PCG specifies default values for both parameters:
+    ///
+    /// - `state = 0xcafef00dd15ea5e5`
+    /// - `stream = 0xa02bdbf7bb3c0a7ac28fa16a64abf96`
+    pub fn new(state: u128, stream: u128) -> Self {
+        // The increment must be odd, hence we discard one bit:
+        let increment = (stream << 1) | 1;
+        Lcg128Xsl64::from_state_incr(state, increment)
+    }
+
+    #[inline]
+    fn from_state_incr(state: u128, increment: u128) -> Self {
+        let mut pcg = Lcg128Xsl64 { state, increment };
+        // Move away from inital value:
+        pcg.state = pcg.state.wrapping_add(pcg.increment);
+        pcg.step();
+        pcg
+    }
+
+    #[inline]
+    fn step(&mut self) {
+        // prepare the LCG for the next round
+        self.state = self.state
+            .wrapping_mul(MULTIPLIER)
+            .wrapping_add(self.increment);
+    }
+}
+
+// Custom Debug implementation that does not expose the internal state
+impl fmt::Debug for Lcg128Xsl64 {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(f, "Lcg128Xsl64 {{}}")
+    }
+}
+
+/// We use a single 255-bit seed to initialise the state and select a stream.
+/// One `seed` bit (lowest bit of `seed[8]`) is ignored.
+impl SeedableRng for Lcg128Xsl64 {
+    type Seed = [u8; 32];
+
+    fn from_seed(seed: Self::Seed) -> Self {
+        let mut seed_u64 = [0u64; 4];
+        le::read_u64_into(&seed, &mut seed_u64);
+        let state = u128::from(seed_u64[0]) | (u128::from(seed_u64[1]) << 64);
+        let incr = u128::from(seed_u64[2]) | (u128::from(seed_u64[3]) << 64);
+
+        // The increment must be odd, hence we discard one bit:
+        Lcg128Xsl64::from_state_incr(state, incr | 1)
+    }
+}
+
+impl RngCore for Lcg128Xsl64 {
+    #[inline]
+    fn next_u32(&mut self) -> u32 {
+        self.next_u64() as u32
+    }
+
+    #[inline]
+    fn next_u64(&mut self) -> u64 {
+        self.step();
+        output_xsl_rr(self.state)
+    }
+
+    #[inline]
+    fn fill_bytes(&mut self, dest: &mut [u8]) {
+        fill_bytes_impl(self, dest)
+    }
+
+    #[inline]
+    fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), Error> {
+        self.fill_bytes(dest);
+        Ok(())
+    }
+}
+
 
 /// A PCG random number generator (XSL 128/64 (MCG) variant).
 ///
@@ -23,19 +124,18 @@ use rand_core::{RngCore, SeedableRng, Error, le};
 /// Congruential Generator, and 64-bit output via "xorshift low (bits),
 /// random rotation" output function.
 ///
-/// This is a 128-bit MCG with the PCG-XSL-RR output function.
+/// This is a 128-bit MCG with the PCG-XSL-RR output function, also known as
+/// `pcg64_fast`.
 /// Note that compared to the standard `pcg64` (128-bit LCG with PCG-XSL-RR
 /// output function), this RNG is faster, also has a long cycle, and still has
 /// good performance on statistical tests.
-///
-/// Note: this RNG is only available using Rust 1.26 or later.
 #[derive(Clone)]
 #[cfg_attr(feature="serde1", derive(Serialize,Deserialize))]
 pub struct Mcg128Xsl64 {
     state: u128,
 }
 
-/// A friendly name for `Mcg128Xsl64`.
+/// A friendly name for `Mcg128Xsl64` (also known as `pcg64_fast`).
 pub type Pcg64Mcg = Mcg128Xsl64;
 
 impl Mcg128Xsl64 {
@@ -66,8 +166,8 @@ impl SeedableRng for Mcg128Xsl64 {
         // Read as if a little-endian u128 value:
         let mut seed_u64 = [0u64; 2];
         le::read_u64_into(&seed, &mut seed_u64);
-        let state = (seed_u64[0] as u128) |
-                    (seed_u64[1] as u128) << 64;
+        let state = u128::from(seed_u64[0])  |
+                    u128::from(seed_u64[1]) << 64;
         Mcg128Xsl64::new(state)
     }
 }
@@ -80,43 +180,46 @@ impl RngCore for Mcg128Xsl64 {
 
     #[inline]
     fn next_u64(&mut self) -> u64 {
-        // prepare the LCG for the next round
-        let state = self.state.wrapping_mul(MULTIPLIER);
-        self.state = state;
-
-        // Output function XSL RR ("xorshift low (bits), random rotation")
-        // Constants are for 128-bit state, 64-bit output
-        const XSHIFT: u32 = 64;     // (128 - 64 + 64) / 2
-        const ROTATE: u32 = 122;    // 128 - 6
-
-        let rot = (state >> ROTATE) as u32;
-        let xsl = ((state >> XSHIFT) as u64) ^ (state as u64);
-        xsl.rotate_right(rot)
+        self.state = self.state.wrapping_mul(MULTIPLIER);
+        output_xsl_rr(self.state)
     }
 
     #[inline]
     fn fill_bytes(&mut self, dest: &mut [u8]) {
-        // specialisation of impls::fill_bytes_via_next; approx 3x faster
-        let mut left = dest;
-        while left.len() >= 8 {
-            let (l, r) = {left}.split_at_mut(8);
-            left = r;
-            let chunk: [u8; 8] = unsafe {
-                transmute(self.next_u64().to_le())
-            };
-            l.copy_from_slice(&chunk);
-        }
-        let n = left.len();
-        if n > 0 {
-            let chunk: [u8; 8] = unsafe {
-                transmute(self.next_u64().to_le())
-            };
-            left.copy_from_slice(&chunk[..n]);
-        }
+        fill_bytes_impl(self, dest)
     }
 
     #[inline]
     fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), Error> {
-        Ok(self.fill_bytes(dest))
+        self.fill_bytes(dest);
+        Ok(())
+    }
+}
+
+#[inline(always)]
+fn output_xsl_rr(state: u128) -> u64 {
+    // Output function XSL RR ("xorshift low (bits), random rotation")
+    // Constants are for 128-bit state, 64-bit output
+    const XSHIFT: u32 = 64;     // (128 - 64 + 64) / 2
+    const ROTATE: u32 = 122;    // 128 - 6
+
+    let rot = (state >> ROTATE) as u32;
+    let xsl = ((state >> XSHIFT) as u64) ^ (state as u64);
+    xsl.rotate_right(rot)
+}
+
+#[inline(always)]
+fn fill_bytes_impl<R: RngCore + ?Sized>(rng: &mut R, dest: &mut [u8]) {
+    let mut left = dest;
+    while left.len() >= 8 {
+        let (l, r) = {left}.split_at_mut(8);
+        left = r;
+        let chunk: [u8; 8] = rng.next_u64().to_le_bytes();
+        l.copy_from_slice(&chunk);
+    }
+    let n = left.len();
+    if n > 0 {
+        let chunk: [u8; 8] = rng.next_u64().to_le_bytes();
+        left.copy_from_slice(&chunk[..n]);
     }
 }
diff --git a/rand/rand_pcg/src/pcg64.rs b/rand/rand_pcg/src/pcg64.rs
index 9177ec2..fadc6dc 100644
--- a/rand/rand_pcg/src/pcg64.rs
+++ b/rand/rand_pcg/src/pcg64.rs
@@ -11,8 +11,8 @@
 //! PCG random number generators
 
 use core::fmt;
-use core::mem::transmute;
 use rand_core::{RngCore, SeedableRng, Error, le, impls};
+#[cfg(feature="serde1")] use serde::{Serialize, Deserialize};
 
 // This is the default multiplier used by PCG for 64-bit state.
 const MULTIPLIER: u64 = 6364136223846793005;
@@ -45,7 +45,8 @@ impl Lcg64Xsh32 {
     /// Note that PCG specifies default values for both parameters:
     ///
     /// - `state = 0xcafef00dd15ea5e5`
-    /// - `stream = 721347520444481703`
+    /// - `stream = 0xa02bdbf7bb3c0a7`
+    // Note: stream is 1442695040888963407u64 >> 1
     pub fn new(state: u64, stream: u64) -> Self {
         // The increment must be odd, hence we discard one bit:
         let increment = (stream << 1) | 1;
@@ -115,27 +116,12 @@ impl RngCore for Lcg64Xsh32 {
 
     #[inline]
     fn fill_bytes(&mut self, dest: &mut [u8]) {
-        // specialisation of impls::fill_bytes_via_next; approx 40% faster
-        let mut left = dest;
-        while left.len() >= 4 {
-            let (l, r) = {left}.split_at_mut(4);
-            left = r;
-            let chunk: [u8; 4] = unsafe {
-                transmute(self.next_u32().to_le())
-            };
-            l.copy_from_slice(&chunk);
-        }
-        let n = left.len();
-        if n > 0 {
-            let chunk: [u8; 4] = unsafe {
-                transmute(self.next_u32().to_le())
-            };
-            left.copy_from_slice(&chunk[..n]);
-        }
+        impls::fill_bytes_via_next(self, dest)
     }
 
     #[inline]
     fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), Error> {
-        Ok(self.fill_bytes(dest))
+        self.fill_bytes(dest);
+        Ok(())
     }
 }