Various fixes

.github/workflows/auto-approve.yml

@@ -0,0 +1,12 @@
+name: Auto approve
+
+on: pull_request_target
+
+jobs:
+  auto-approve:
+    runs-on: ubuntu-latest
+    permissions:
+      pull-requests: write
+    if: github.actor == 'dependabot[bot]' || github.actor == 'jmesmon'
+    steps:
+      - uses: hmarr/auto-approve-action@v3

.github/workflows/auto-merge.yml

@@ -0,0 +1,12 @@
+name: Auto-Merge
+on:
+  pull_request_target:
+
+jobs:
+  enable-auto-merge:
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: alexwilson/enable-github-automerge-action@main
+      with:
+        github-token: "${{ secrets.GITHUB_TOKEN }}"

.github/workflows/ci.yml

@@ -14,7 +14,7 @@ jobs:
           - stable
           - beta
           - nightly
-          - 1.47.0
+          - 1.64.0
 
     steps:
       - uses: actions/checkout@v2
@@ -26,7 +26,8 @@ jobs:
           override: true
 
       - name: Cache Rust dependencies
-        uses: Swatinem/rust-cache@v1
+        uses: Swatinem/[email protected]
+
 
       - name: Run all tests
         uses: actions-rs/cargo@v1
@@ -47,7 +48,7 @@ jobs:
           components: rustfmt, clippy
 
       - name: Cache Rust dependencies
-        uses: Swatinem/rust-cache@v1
+        uses: Swatinem/rust-cache@v2.4.0
 
       - uses: actions-rs/cargo@v1
         with:

Cargo.toml

@@ -7,7 +7,7 @@ keywords = [ "hash", "rolling", "incremental", "split" , "cdc"]
 license = "AGPL-3.0-or-later"
 repository = "https://github.com/jmesmon/hash-roll.git"
 documentation = "https://docs.rs/hash-roll"
-include = ["Cargo.toml", "**/*.rs", "README.md", "COMPARE.md"]
+include = ["**/*.rs", "README.md", "COMPARE.md"]
 edition = "2018"
 
 [features]
@@ -21,7 +21,8 @@ default = [
 	"pigz",
 	"ram",
 	"zpaq",
-	"zstd"
+	"zstd",
+
 ]
 
 bup = []
@@ -35,18 +36,19 @@ ram = []
 zpaq = []
 zstd = []
 
+slow_tests = []
+
 [dependencies]
 fmt-extra = "0.2"
-#circbuf = "0.1.4"
 
 [dev-dependencies]
-rand = "0.7.3"
-histogram = "0.6"
-quickcheck = "0.9"
+rand = "0.8"
+histogram = "0.7"
+quickcheck = "1"
 rollsum = "0.3"
-criterion = "0.3"
-rand_pcg = "0.2.1"
-proptest = "0.10.0"
+criterion = "0.5"
+rand_pcg = "0.3"
+proptest = "1"
 
 [[bench]]
 name = "compare"

src/buzhash.rs

@@ -332,7 +332,7 @@ impl fmt::Debug for BuzHashTableBufHash {
     }
 }
 
-impl<'a> From<Box<[u32; 256]>> for BuzHashTableBufHash {
+impl From<Box<[u32; 256]>> for BuzHashTableBufHash {
     fn from(table: Box<[u32; 256]>) -> Self {
         Self { table }
     }

src/gear.rs

@@ -72,14 +72,8 @@ impl<'a> ToChunkIncr for Gear32<'a> {
     type Incr = GearIncr32<'a>;
 
     fn to_chunk_incr(&self) -> Self::Incr {
-        self.into()
-    }
-}
-
-impl<'a> From<&Gear32<'a>> for GearIncr32<'a> {
-    fn from(params: &Gear32<'a>) -> Self {
-        Self {
-            params: params.clone(),
+        Self::Incr {
+            params: self.clone(),
             state: Default::default(),
         }
     }

src/lib.rs

@@ -110,10 +110,10 @@ pub mod gzip;
 pub mod mii;
 pub mod pigz;
 pub mod ram;
-pub mod range;
 pub mod zpaq;
 pub mod zstd;
 
+mod range;
 pub(crate) use range::RangeExt;
 
 /// Accept incrimental input and provide indexes of split points
@@ -287,9 +287,12 @@ pub trait Chunk {
 
     /// Find the next "chunk" in `data` to emit
     ///
-    /// The return value is a pair of a range representing the start and end of the chunk being
-    /// emitted, and the offset from which subsequent `data` subsets should be passed to the next
-    /// call to `find_chunk_edge`.
+    /// The return value is a tuple of:
+    ///
+    ///  1. an optional offset in `data` which is just after a chunk to be emitted (this is called
+    ///     the "cut point")
+    ///  2. the number of bytes to remove from the start of `data` before passing `data` to
+    ///     `find_chunk_edge()` again (this is called the "discard count")
     ///
     /// `state` is mutated so that it does not rexamine previously examined data, even when a chunk
     /// is not emitted.
@@ -312,9 +315,9 @@ pub trait Chunk {
     /// let mut prev_cut = 0;
     ///
     /// loop {
-    ///    let (chunk, discard_ct) = chunk.find_chunk_edge(&mut ss, data);
+    ///    let (cut_point_in_data, discard_ct) = chunk.find_chunk_edge(&mut ss, data);
     ///
-    ///    match chunk {
+    ///    match cut_point_in_data {
     ///        Some(cut_point) => {
     ///            // map `cut_point` from the current slice back into the original slice so we can
     ///            // have consistent indexes

src/mii.rs

@@ -51,7 +51,12 @@ impl crate::Chunk for Mii {
     ) -> (Option<usize>, usize) {
         match state.push(data) {
             Some(v) => {
-                state.reset();
+                // NOTE: these check that we don't carry state across chunk boundaries. The
+                // original paper is unclear if this is _exactly_ the case, and there are no test
+                // vectors for MII. That said: given the goals, it's unlikely that `prev` should
+                // persist across chunks (as doing so would cause inconsistent chunking).
+                debug_assert_eq!(state.incr.prev, 0xff);
+                debug_assert_eq!(state.incr.increment, 0);
                 (Some(v), v)
             }
             None => (None, data.len()),
@@ -78,12 +83,6 @@ pub struct MiiSearchState {
     incr: MiiIncr,
 }
 
-impl MiiSearchState {
-    fn reset(&mut self) {
-        self.incr.reset();
-    }
-}
-
 impl From<MiiIncr> for MiiSearchState {
     fn from(incr: MiiIncr) -> Self {
         Self { incr }
@@ -127,7 +126,7 @@ impl ChunkIncr for MiiIncr {
                 if self.increment == self.w {
                     // this is a split
                     self.increment = 0;
-                    self.prev = 0;
+                    self.prev = 0xff;
                     return Some(i + 1);
                 }
             } else {
@@ -139,10 +138,3 @@ impl ChunkIncr for MiiIncr {
         None
     }
 }
-
-impl MiiIncr {
-    fn reset(&mut self) {
-        self.prev = 0xff;
-        self.increment = 0;
-    }
-}

src/range.rs

@@ -24,12 +24,12 @@ impl<T, R: std::ops::RangeBounds<T>> RangeExt<T> for R {
         T: PartialOrd<T>,
     {
         match self.end_bound() {
-            Included(ref i) => {
+            Included(i) => {
                 if item > i {
                     return true;
                 }
             }
-            Excluded(ref i) => {
+            Excluded(i) => {
                 if item >= i {
                     return true;
                 }
@@ -45,12 +45,12 @@ impl<T, R: std::ops::RangeBounds<T>> RangeExt<T> for R {
         T: PartialOrd<T>,
     {
         match self.start_bound() {
-            Included(ref i) => {
+            Included(i) => {
                 if item < i {
                     return true;
                 }
             }
-            Excluded(ref i) => {
+            Excluded(i) => {
                 if item <= i {
                     return true;
                 }

src/zpaq.rs

@@ -207,7 +207,7 @@ impl Chunk for Zpaq {
     ) -> (Option<usize>, usize) {
         for (i, v) in data.iter().enumerate() {
             let h = state.feed(*v);
-            if self.split_here(h, (state.idx + 1) as u64) {
+            if self.split_here(h, state.idx + 1) {
                 *state = self.to_search_state();
                 return (Some(i + 1), i + 1);
             }