chore: audit cleanup, CI, cargo-deny, encode fuzz target (0.3.4)

Simplify encoder exit branch (−41 B code size), add GitHub Actions CI
with MSRV matrix, cargo-deny with permissive-only license policy,
encode-only fuzz target, and feature coverage across test gates.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.github/workflows/ci.yml

@@ -0,0 +1,69 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+env:
+  CARGO_TERM_COLOR: always
+
+jobs:
+  test:
+    name: Test (${{ matrix.rust }})
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        rust: [stable, "1.93"]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dtolnay/rust-toolchain@master
+        with:
+          toolchain: ${{ matrix.rust }}
+      - run: cargo test
+
+  test-legacy-msrv:
+    name: Test (1.83, legacy-msrv)
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dtolnay/rust-toolchain@master
+        with:
+          toolchain: "1.83"
+      - run: cargo test --features legacy-msrv
+
+  clippy:
+    name: Clippy
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dtolnay/rust-toolchain@stable
+        with:
+          components: clippy
+      - run: cargo clippy -- -D warnings
+
+  fmt:
+    name: Format
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dtolnay/rust-toolchain@stable
+        with:
+          components: rustfmt
+      - run: cargo fmt -- --check
+
+  doc:
+    name: Doc
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: dtolnay/rust-toolchain@stable
+      - run: cargo doc --no-deps
+
+  deny:
+    name: Cargo Deny
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: EmbarkStudios/cargo-deny-action@v2

CHANGELOG.md

@@ -1,5 +1,36 @@
 # Changelog
 
+## 0.3.4 — 2026-04-01
+
+### Changed
+
+- **Encode shrunk by 41 bytes** — simplified branch structure in encoder exit
+  path. 388 B encode, 392 B decode, 780 B total (was 821 B). Zero benchmark
+  regressions.
+- **Clarified decode invariant** — added inline comment explaining the subtle
+  trailing-0x01 COBS rule (end-of-data marker vs zero byte).
+- **Clippy cleanup** — `saturating_sub` in proptest (cosmetic, test-only).
+
+### Added
+
+- **GitHub Actions CI** — tests on stable + MSRV 1.93 + 1.83/legacy-msrv,
+  clippy, fmt, doc, and cargo-deny.
+- **`cargo-deny` integration** — `deny.toml` with permissive-only license
+  policy, advisory checks, source auditing. Added `just test-deny` gate.
+- **Encode-only fuzz target** — `fuzz_encode` exercises buffer-too-small
+  paths and validates output properties independently of decode.
+- **Feature coverage** — `just test-unit` and `just test-clippy` now test
+  both default and `legacy-msrv` features. `just test-doc` builds with
+  `--all-features`.
+
+### Fixed
+
+- **README accuracy** — overhead description changed from "exactly" to
+  "at most ... plus 1", added miri to test mentions, updated code size
+  numbers.
+- **CLAUDE.md rewrite** — updated stale API signatures, LOC count, roadmap,
+  dev-deps list, and development commands.
+
 ## 0.3.3 — 2026-04-01
 
 ### Fixed

CLAUDE.md

@@ -10,9 +10,9 @@
 
 1. **Correct** — 100% COBS spec-compliant (Cheshire & Baker 1999). No shortcuts, no deviations.
 2. **Interoperable** — proven byte-for-byte compatibility with `corncobs`, `cobs` crates and Python `cobs` packages.
-3. **Insanely tested** — canonical vectors, property-based tests (proptest), fuzz targets (cargo-fuzz), cross-crate interop, randomized payloads. If you think testing is done, you're 10% done.
+3. **Insanely tested** — canonical vectors, property-based tests (proptest), fuzz targets (cargo-fuzz), cross-crate interop, miri, mutation testing, randomized payloads. If you think testing is done, you're 10% done.
-4. **Fast** — benchmark against every competing implementation. Must be at least as fast, ideally fastest. Benchmarks are CI-enforced.
+4. **Fast** — benchmark against every competing implementation. Must be at least as fast, ideally fastest. Benchmarks run via `just bench`.
-5. **Tiny** — `no_std`, zero-alloc, zero runtime dependencies. ~120 lines of implementation. Runs on 8-bit MCUs and servers alike.
+5. **Tiny** — `no_std`, zero-alloc, zero runtime dependencies. ~140 lines of implementation. Runs on 8-bit MCUs and servers alike.
 6. **Trustworthy** — small enough to audit by hand. No unsafe. No panics in release. Fuzz-hardened.
 
 ## Architecture
@@ -22,7 +22,7 @@ Single-file library: `src/lib.rs`
 ### Public API (3 functions)
 
 ```rust
-pub fn encode(src: &[u8], dest: &mut [u8]) -> Option<usize>
+pub const fn encode(src: &[u8], dest: &mut [u8]) -> Option<usize>
 pub fn decode(src: &[u8], dest: &mut [u8]) -> Option<usize>
 pub const fn max_encoded_len(src_len: usize) -> usize
 ```
@@ -32,26 +32,29 @@ pub const fn max_encoded_len(src_len: usize) -> usize
 ## Development
 
 ```sh
-# Run all tests
+# Run all quality gates (tests, clippy, fmt, doc, fuzz, miri, mutants)
-cargo test
+just test
+
+# Unit + property + interop tests only
+just test-unit
+
+# Benchmarks (iai-callgrind, requires valgrind)
+just bench
 
 # Fuzz (requires nightly + cargo-fuzz)
-cd fuzz
+just test-fuzz
-cargo +nightly fuzz run fuzz_decode -- -max_total_time=60
-cargo +nightly fuzz run fuzz_roundtrip -- -max_total_time=60
 ```
 
 ## Dev Dependencies (test-only)
 
+- `cobs` — cross-validation against the `cobs` 0.5 crate
 - `corncobs` — cross-validation against another COBS implementation
+- `iai-callgrind` — deterministic instruction-count benchmarks
+- `mutants` — mutation testing annotations
 - `proptest` — property-based testing
 - `rand` — randomized test payloads
 
 ## Roadmap
 
-- More interop targets: `cobs` crate, Python `cobs` package
-- Benchmarks (criterion) vs corncobs, cobs, and others
 - quickcheck in addition to proptest
 - Even more fuzz targets and edge-case generators
-- MSRV policy and CI
-- `#[deny(unsafe_code)]` badge

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ucobs"
-version = "0.3.3"
+version = "0.3.4"
 edition = "2021"
 rust-version = "1.93"
 license = "MIT"

README.md

@@ -8,8 +8,8 @@
 
 [Consistent Overhead Byte Stuffing](https://en.wikipedia.org/wiki/Consistent_Overhead_Byte_Stuffing)
 (COBS) encodes arbitrary bytes so that `0x00` never appears in the output,
-allowing `0x00` to serve as an unambiguous frame delimiter. Overhead is exactly
+allowing `0x00` to serve as an unambiguous frame delimiter. Overhead is at most
-1 byte per 254 input bytes (worst case).
+1 byte per 254 input bytes, plus 1.
 
 ## Features
 
@@ -18,7 +18,7 @@ allowing `0x00` to serve as an unambiguous frame delimiter. Overhead is exactly
 - **100% spec-compliant** — Cheshire & Baker, IEEE/ACM Transactions on Networking, 1999
 - **Proven interoperability** — cross-validated against `corncobs`, `cobs`, and Python `cobs`
 - **Insanely tested** — canonical vectors, property-based tests, fuzz targets,
-  cross-crate interop, randomized payloads, and more coming
+  cross-crate interop, miri, mutation testing, randomized payloads
 
 ## Performance
 
@@ -64,16 +64,16 @@ combined). Run `just bench-size` to reproduce.
 
 | Crate          | `encode` | `decode` | Total |
 |----------------|--------:|---------:|------:|
-| μCOBS          |   429 B |    392 B | 821 B |
+| μCOBS          |   388 B |    392 B | 780 B |
 | `cobs` 0.5     |   282 B |    494 B | 776 B |
 | `corncobs` 0.1 |   375 B |    262 B | 637 B |
 
-All three crates are under 1 KB. μCOBS is the largest because safe
+All three crates are under 1 KB. μCOBS and `cobs` are nearly tied;
-`const fn` encoding with `copy_from_slice`/memcpy requires `split_at`
+`corncobs` is the smallest. μCOBS's safe `const fn` encoding with
-bounds checks that generate panic paths — the cost of combining
+`copy_from_slice`/memcpy requires `split_at` bounds checks that
-compile-time safety with runtime speed. Embedded users targeting size
+generate panic paths — the cost of combining compile-time safety with
-over speed can build with `opt-level = "s"` which reduces μCOBS to
+runtime speed. Embedded users targeting size over speed can build with
-~738 B.
+`opt-level = "s"` to shrink further.
 
 ### Crate properties
 
@@ -105,7 +105,7 @@ over speed can build with `opt-level = "s"` which reduces μCOBS to
 | Dead-simple API (3 functions) | **yes** | yes | more surface area |
 | In-place / streaming encode | no | no | **yes** |
 | `no_std` + zero-alloc by default | **yes** | opt-in | **yes** |
-| Thorough test suite | **111 tests, fuzz, proptest, mutants** | basic | basic |
+| Thorough test suite | **111 tests, fuzz, proptest, miri, mutants** | basic | basic |
 
 **Choose μCOBS** if you want the smallest, most auditable COBS implementation
 with a `const fn` encoder, a dead-simple 3-function API, and leading or tied

deny.toml

@@ -0,0 +1,37 @@
+[graph]
+all-features = true
+
+[advisories]
+yanked = "deny"
+ignore = [
+    # bincode 1.3.3 — transitive dev-dep via iai-callgrind (latest 0.16.1).
+    # Unmaintained but stable and complete; no security issue. Dev-only, never
+    # ships in ucobs. Remove this ignore when iai-callgrind drops bincode.
+    "RUSTSEC-2025-0141",
+]
+
+[licenses]
+# Permissive licenses only. All copyleft (GPL, LGPL, AGPL, MPL, etc.) is denied by omission.
+allow = [
+    "MIT",
+    "Apache-2.0",
+    "BSD-2-Clause",
+    "BSD-3-Clause",
+    "ISC",
+    "Zlib",
+    "Unlicense",
+    "CC0-1.0",
+    "Unicode-3.0",
+    "Unicode-DFS-2016",
+]
+unused-allowed-license = "allow"
+
+[bans]
+multiple-versions = "warn"
+wildcards = "deny"
+
+[sources]
+unknown-registry = "deny"
+unknown-git = "deny"
+allow-registry = ["https://github.com/rust-lang/crates.io-index"]
+allow-git = []

fuzz/Cargo.toml

@@ -30,3 +30,9 @@ name = "fuzz_small_dest"
 path = "fuzz_targets/fuzz_small_dest.rs"
 test = false
 doc = false
+
+[[bin]]
+name = "fuzz_encode"
+path = "fuzz_targets/fuzz_encode.rs"
+test = false
+doc = false

fuzz/fuzz_targets/fuzz_encode.rs

@@ -0,0 +1,33 @@
+//! Fuzz target: encode arbitrary input into buffers of varying sizes.
+//! Exercises encode error paths (buffer too small) and validates output properties.
+#![no_main]
+use libfuzzer_sys::fuzz_target;
+
+fuzz_target!(|data: &[u8]| {
+    if data.len() > 1024 {
+        return;
+    }
+
+    let max = ucobs::max_encoded_len(data.len());
+
+    // Encode into a correctly-sized buffer — must always succeed.
+    let mut buf = vec![0u8; max];
+    let n = ucobs::encode(data, &mut buf)
+        .expect("encode failed with max_encoded_len buffer");
+    let encoded = &buf[..n];
+
+    // Output must never contain 0x00.
+    assert!(!encoded.contains(&0x00), "encoded output contains 0x00");
+
+    // Output length must be bounded by max_encoded_len.
+    assert!(n <= max, "encoded length exceeds max_encoded_len");
+
+    // Output must be at least 1 byte (even empty input encodes to [0x01]).
+    assert!(n >= 1, "encoded length is zero");
+
+    // Encode into every possible undersized buffer — must return None, never panic.
+    for size in 0..n {
+        let mut small = vec![0u8; size];
+        assert_eq!(ucobs::encode(data, &mut small), None);
+    }
+});

justfile

@@ -4,8 +4,8 @@
 default:
     @just --list
 
-# run ALL quality gates (tests, clippy, fmt check, doc, fuzz, miri, mutants)
+# run ALL quality gates (tests, clippy, fmt check, doc, deny, fuzz, miri, mutants)
-test: test-unit test-clippy test-fmt test-doc test-fuzz test-miri test-mutants
+test: test-unit test-clippy test-fmt test-doc test-deny test-fuzz test-miri test-mutants
 
 # quick compile check (no tests)
 check:
@@ -14,11 +14,13 @@ check:
 # unit + integration + proptest tests (via nextest for parallelism)
 test-unit:
     cargo nextest run
+    cargo nextest run --features legacy-msrv
     cargo test --doc
 
 # clippy with all warnings as errors
 test-clippy:
     cargo clippy -- -D warnings
+    cargo clippy --features legacy-msrv -- -D warnings
 
 # check formatting
 test-fmt:
@@ -26,11 +28,16 @@ test-fmt:
 
 # doc build (catches broken doc links)
 test-doc:
-    cargo doc --no-deps
+    cargo doc --no-deps --all-features
+
+# license and advisory audit (requires cargo-deny)
+test-deny:
+    cargo deny check
 
 # fuzz all targets (60s each, requires nightly + cargo-fuzz)
 test-fuzz:
     cd fuzz && cargo +nightly fuzz run fuzz_decode -- -max_total_time=60
+    cd fuzz && cargo +nightly fuzz run fuzz_encode -- -max_total_time=60
     cd fuzz && cargo +nightly fuzz run fuzz_roundtrip -- -max_total_time=60
     cd fuzz && cargo +nightly fuzz run fuzz_small_dest -- -max_total_time=60
 

mise.toml

@@ -1,4 +1,7 @@
 [tools]
-"cargo:cargo-nextest" = "latest"
+rust = "1.93"
+"cargo:cargo-deny" = "latest"
+"cargo:cargo-fuzz" = "latest"
 "cargo:cargo-mutants" = "latest"
+"cargo:cargo-nextest" = "latest"
 "cargo:iai-callgrind-runner" = "latest"

src/lib.rs

@@ -256,14 +256,10 @@ pub const fn encode(src: &[u8], dest: &mut [u8]) -> Option<usize> {
         }
         si += run;
 
-        if full_block {
-            if si >= src.len() {
-                break;
-            }
-        } else {
         if si >= src.len() {
             break;
         }
+        if !full_block {
             si += 1; // skip the zero byte
         }
     }
@@ -315,6 +311,9 @@ pub fn decode(src: &[u8], dest: &mut [u8]) -> Option<usize> {
                 src = rest;
                 count += 1;
             }
+            // If more data follows, every 0x01 represents a zero byte.
+            // If the stream ends here, the final 0x01 is the end-of-data
+            // marker, not a zero — so emit one fewer.
             let zeros = if !src.is_empty() { count } else { count - 1 };
             if di + zeros > dest.len() {
                 return None;
@@ -1259,7 +1258,7 @@ mod proptests {
             shrink in 1usize..256,
         ) {
             let max = max_encoded_len(data.len());
-            let dest_size = if shrink >= max { 0 } else { max - shrink };
+            let dest_size = max.saturating_sub(shrink);
             let mut buf = vec![0u8; dest_size];
             let _ = encode(&data, &mut buf); // may return None, must not panic
         }