Add HTTP/2 HPACK (RFC 7541) and LHA -lh2- codecs

CHANGELOG.md

@@ -7,6 +7,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+
+- **HTTP/2 HPACK header compression** (RFC 7541) behind the new `hpack`
+  feature. `compcol::hpack::{HpackEncoder, HpackDecoder}` implement the full
+  header codec — static + dynamic indexing tables, N-bit-prefix integers,
+  string literals, and all field representations (indexed, literal
+  with/without indexing, never-indexed, dynamic-table size update). Validated
+  byte-for-byte against the RFC 7541 Appendix C worked examples. The §5.2
+  string Huffman primitive is also exposed as the `Http2Huffman` codec
+  (name `h2-huffman`) through the uniform `Encoder`/`Decoder` traits.
+- **LHA `-lh2-`** added to the `lha` feature: 8 KiB-window LZSS with adaptive
+  (dynamic) Huffman for both literals/lengths and match positions. Like `lh1`
+  it is continuous and size-terminated, so its decoder takes the uncompressed
+  length via `DecoderConfig::with_len`. Clean-room, round-trip validated.
+
 ## [0.6.0](https://github.com/KarpelesLab/compcol/compare/v0.5.1...v0.6.0) - 2026-06-03
 
 ### Other

Cargo.toml

@@ -50,6 +50,7 @@ all = [
     "lha",
     "bcj", "bcj2", "delta",
     "arc_crunch", "arc_squeeze", "arc_squash",
+    "hpack",
 ]
 # Enables `alloc`-backed conveniences (e.g. the `factory` module, the
 # `compcol::vec` one-shot helpers). Pulled in automatically by features
@@ -239,6 +240,12 @@ arc_squeeze = ["alloc"]
 # block-mode CLEAR code and no header byte (no RLE pre-pass). Encoder and
 # decoder both implemented and validated by round-trip.
 arc_squash = ["alloc"]
+# HTTP/2 HPACK header compression (RFC 7541): static + dynamic indexing
+# tables, integer/string coding, and the §5.2 string Huffman code. The
+# Huffman primitive is also exposed as the `Http2Huffman` codec
+# (name `"h2-huffman"`). The full header codec lives behind its own
+# `compcol::hpack` API (it is stateful over header lists, not a byte stream).
+hpack   = ["alloc"]
 # `compcol::tokio_io` — async mirrors of compcol::io for the tokio
 # runtime. Pulls the tokio dependency for its AsyncRead/AsyncWrite
 # trait definitions; the rest of the crate stays dep-free.

README.md

@@ -61,7 +61,7 @@ flag, and a `compcol` binary turns the library into a Unix-style filter.
 | Microsoft Xpress (plain LZ77) | `xpress` | `.xpress` | full | full (per [MS-XCA] §2.2) | hand-built fixtures |
 | Microsoft Xpress Huffman | `xpress_huffman` | `.xph` | full (LZ77 + canonical Huffman) | full (per [MS-XCA] §2.1; used in WIM / CompactOS NTFS) | hand-built fixtures |
 | LZNT1 (NTFS native compression) | `lznt1` | `.lznt1` | full | full (per [MS-XCA] §2.5; 4 KiB-chunked LZ77, no entropy coding) | hand-built fixtures |
-| LHA / LZH (`-lh1-`/`-lh4-`/`-lh5-`/`-lh6-`/`-lh7-`) | `lha` | `.lzh` | full (lh1 adaptive Huffman; lh4/5/6/7 static Huffman) | full (clean-room from Okumura LZHUF / ar002) | own round-trip (no reference fixture) |
+| LHA / LZH (`-lh1-`/`-lh2-`/`-lh4-`/`-lh5-`/`-lh6-`/`-lh7-`) | `lha` | `.lzh` | full (lh1/lh2 adaptive Huffman; lh4/5/6/7 static Huffman) | full (clean-room from Okumura LZHUF / ar002) | own round-trip (no reference fixture) |
 | BCJ branch filters (x86, ARM, ARMT, ARM64, PPC, SPARC, IA-64, RISC-V) | `bcj` | `bcj-<arch>` | full (reversible filter) | full | round-trip identity (public-domain LZMA SDK transform) |
 | BCJ2 (7z 4-stream x86 filter) | `bcj2` | — | `bcj2::encode` (fn API) | `bcj2::decode` (fn API) | round-trip identity (LZMA SDK algorithm) |
 | Delta filter (distance 1..=256) | `delta` | `delta` | full (reversible filter) | full | round-trip identity |
@@ -76,6 +76,13 @@ flag, and a `compcol` binary turns the library into a Unix-style filter.
 | RAR 2.x | `rar2` | `.rar` | `Unsupported` (license) | full LZ77+Huffman + audio predictor | real rar-2.60 fixtures |
 | RAR 3.x | `rar3` | `.rar` | `Unsupported` (license) | full LZ77+Huffman + E8 filter; PPMd & VM filters refused | libarchive RAR3 fixtures |
 | RAR 5.x | `rar5` | `.rar` | `Unsupported` (license) | full LZ77+Huffman + x86 filter; Delta/ARM refused | RARLAB-CLI fixtures |
+| HTTP/2 HPACK (RFC 7541) | `hpack` | — | full (header codec + `h2-huffman` string codec) | full (static+dynamic tables, integer/string coding) | RFC 7541 Appendix C vectors |
+
+HPACK is HTTP/2's header-compression codec, not a byte-stream codec: it
+operates on `(name, value)` header lists with per-connection dynamic-table
+state, so it lives behind its own `compcol::hpack` API (`HpackEncoder` /
+`HpackDecoder`). The §5.2 string Huffman primitive is also exposed as the
+`Http2Huffman` codec (name `h2-huffman`) through the uniform trait surface.
 
 The RAR encoders are permanently `Unsupported` per RARLAB's unRAR
 license terms (every clean-room RAR reader — libarchive, The

fuzz/Cargo.toml

@@ -199,6 +199,13 @@ test             = false
 doc              = false
 bench            = false
 
+[[bin]]
+name             = "decoder_h2_huffman"
+path             = "fuzz_targets/decoder_h2_huffman.rs"
+test             = false
+doc              = false
+bench            = false
+
 [[bin]]
 name             = "decoder_bcj"
 path             = "fuzz_targets/decoder_bcj.rs"

fuzz/fuzz_targets/decoder_h2_huffman.rs

@@ -0,0 +1,20 @@
+#![no_main]
+use compcol::hpack::{huffman, HpackDecoder};
+use libfuzzer_sys::fuzz_target;
+
+// Smoke property: neither the HPACK header decoder nor the standalone
+// "h2 huffman" string decoder may panic on arbitrary attacker-controlled
+// input. libfuzzer feeds us garbage; any panic/abort trips the harness.
+//
+// Both are pure whole-buffer transforms (the HPACK header block decoder is
+// the primary attack surface — it walks the integer/string/index
+// representations), so we just call them and discard the result.
+fuzz_target!(|data: &[u8]| {
+    // HPACK header block: bounded table so a hostile size update can't grow
+    // state without limit.
+    let mut dec = HpackDecoder::with_max_table_size(4096);
+    let _ = dec.decode(data);
+
+    // The §5.2 Huffman string primitive on its own.
+    let _ = huffman::decode(data);
+});

src/factory.rs

@@ -127,13 +127,19 @@ pub fn encoder_by_name(name: &str) -> Option<Box<dyn Encoder>> {
         #[cfg(feature = "lha")]
         crate::lha::Lh1::NAME => Some(Box::new(<crate::lha::Lh1 as Algorithm>::encoder())),
         #[cfg(feature = "lha")]
+        crate::lha::Lh2::NAME => Some(Box::new(<crate::lha::Lh2 as Algorithm>::encoder())),
+        #[cfg(feature = "lha")]
         crate::lha::Lh4::NAME => Some(Box::new(<crate::lha::Lh4 as Algorithm>::encoder())),
         #[cfg(feature = "lha")]
         crate::lha::Lh5::NAME => Some(Box::new(<crate::lha::Lh5 as Algorithm>::encoder())),
         #[cfg(feature = "lha")]
         crate::lha::Lh6::NAME => Some(Box::new(<crate::lha::Lh6 as Algorithm>::encoder())),
         #[cfg(feature = "lha")]
         crate::lha::Lh7::NAME => Some(Box::new(<crate::lha::Lh7 as Algorithm>::encoder())),
+        #[cfg(feature = "hpack")]
+        crate::hpack::Http2Huffman::NAME => Some(Box::new(
+            <crate::hpack::Http2Huffman as Algorithm>::encoder(),
+        )),
         #[cfg(feature = "bcj")]
         crate::bcj::BcjX86::NAME => Some(Box::new(<crate::bcj::BcjX86 as Algorithm>::encoder())),
         #[cfg(feature = "bcj")]
@@ -359,13 +365,19 @@ pub fn decoder_by_name(name: &str) -> Option<Box<dyn Decoder>> {
         #[cfg(feature = "lha")]
         crate::lha::Lh1::NAME => Some(Box::new(<crate::lha::Lh1 as Algorithm>::decoder())),
         #[cfg(feature = "lha")]
+        crate::lha::Lh2::NAME => Some(Box::new(<crate::lha::Lh2 as Algorithm>::decoder())),
+        #[cfg(feature = "lha")]
         crate::lha::Lh4::NAME => Some(Box::new(<crate::lha::Lh4 as Algorithm>::decoder())),
         #[cfg(feature = "lha")]
         crate::lha::Lh5::NAME => Some(Box::new(<crate::lha::Lh5 as Algorithm>::decoder())),
         #[cfg(feature = "lha")]
         crate::lha::Lh6::NAME => Some(Box::new(<crate::lha::Lh6 as Algorithm>::decoder())),
         #[cfg(feature = "lha")]
         crate::lha::Lh7::NAME => Some(Box::new(<crate::lha::Lh7 as Algorithm>::decoder())),
+        #[cfg(feature = "hpack")]
+        crate::hpack::Http2Huffman::NAME => Some(Box::new(
+            <crate::hpack::Http2Huffman as Algorithm>::decoder(),
+        )),
         #[cfg(feature = "bcj")]
         crate::bcj::BcjX86::NAME => Some(Box::new(<crate::bcj::BcjX86 as Algorithm>::decoder())),
         #[cfg(feature = "bcj")]
@@ -701,6 +713,9 @@ pub const fn extension(name: &str) -> Option<&'static str> {
     if str_eq(name, "lh1") && cfg!(feature = "lha") {
         return Some("lzh");
     }
+    if str_eq(name, "lh2") && cfg!(feature = "lha") {
+        return Some("lzh");
+    }
     if str_eq(name, "lh4") && cfg!(feature = "lha") {
         return Some("lzh");
     }
@@ -861,13 +876,17 @@ pub const fn names() -> &'static [&'static str] {
         #[cfg(feature = "lha")]
         crate::lha::Lh1::NAME,
         #[cfg(feature = "lha")]
+        crate::lha::Lh2::NAME,
+        #[cfg(feature = "lha")]
         crate::lha::Lh4::NAME,
         #[cfg(feature = "lha")]
         crate::lha::Lh5::NAME,
         #[cfg(feature = "lha")]
         crate::lha::Lh6::NAME,
         #[cfg(feature = "lha")]
         crate::lha::Lh7::NAME,
+        #[cfg(feature = "hpack")]
+        crate::hpack::Http2Huffman::NAME,
         #[cfg(feature = "bcj")]
         crate::bcj::BcjX86::NAME,
         #[cfg(feature = "bcj")]