feat(rust): port block entropy statistics

Continue the incremental zstd_compress.c migration with its sequence
statistics and seqStore entropy-compression layer.  The following now
live in rust/src/zstd_compress_stats.rs:

- ZSTD_seqToCodes(), exported under its original name because the C
  dictionary builder (zdict.c) and decodecorpus link against it,
- ZSTD_buildSequencesStatistics() and its dummy variant, whose result
  struct no longer crosses the language boundary,
- ZSTD_entropyCompressSeqStore_internal(), _wExtLitBuffer(), and
  ZSTD_entropyCompressSeqStore(),
- ZSTD_buildBlockEntropyStats() with its literals/sequences helpers,
- ZSTD_copyBlockSequences() and the ZSTD_updateRep() rules it shares
  with the superblock writer.

Boundary: ZSTD_CCtx and ZSTD_CCtx_params stay private to C.  These
paths read exactly two parameter fields, so the C shims keep the
original static/exported function names and forward the strategy and
ZSTD_literalsCompressionIsDisabled() as int scalars alongside the
seqStore and entropy-table leaves.  The leaf layouts (SeqDef,
SeqStore_t, ZSTD_hufCTables_t, ZSTD_fseCTables_t, entropy metadata,
SeqCollector, ZSTD_Sequence) are pinned by compile-time asserts in
zstd_compress.c and by both-pointer-width layout tests in Rust.
ZSTD_buildSeqStore, block dispatch/splitting, and the block-size
estimation helpers remain C for a later slice.

The superblock module previously round-tripped through the C export of
ZSTD_buildBlockEntropyStats and mirrored the entropy leaf structs
privately.  It now calls the crate-internal builder directly, and the
shared struct definitions moved to zstd_compress_stats; consequently
ZSTD_rust_compressSuperBlock() takes the two parameter scalars instead
of an opaque ZSTD_CCtx_params pointer, extracted by its C shim.  Its
layout and repcode tests moved with the definitions.

One C helper family gets no shim: the static
ZSTD_entropyCompressSeqStore_wExtLitBuffer() had a single caller and
was folded into the Rust implementation.

Byte-identity was verified against the pre-change compressor: COPYING,
datagen -g5000000 -s7, and datagen -g300000 -s21 -P90, each at levels
1/3/9/19 and --fast=5, plus a superblock-heavy pass at level 19 with
--target-compressed-block-size=1024; all 18 frames are byte-identical,
covering the repeat-mode state machine, longOffsets, RLE/raw fallback,
and dstSize_tooSmall paths through the block splitter and superblock.

Test plan:
- cd rust && cargo fmt --check && cargo clippy --all-targets -- -D
  warnings && cargo test --all-targets (131 tests: new reference
  vectors for seqToCodes, RLE table headers, empty-seqStore repeat
  copies, literal-stats type selection, and repcode resolution in
  copyBlockSequences)
- cargo build --release --no-default-features --features compression
  and --features decompression
- make -C tests fuzzer && ./tests/fuzzer -i1 --no-big-tests (covers
  ZSTD_generateSequences and ZDICT training over the Rust seqToCodes)
- make -C tests test-rust-lib-smoke
- ./tests/zstreamtest -i1 and ./tests/zstreamtest --newapi -t1 -i1
- ./tests/decodecorpus -t -T1s (decodecorpus links Rust seqToCodes)
- /tmp/ref vs /tmp/got frame diff as described above: byte-identical

rust/README.md

@@ -31,6 +31,13 @@ zstd ABI:
   - `zstd_presplit` chooses split points for full compression blocks.
   - `zstd_compress_literals` emits raw, RLE, and Huffman literal sections
     while preserving the compressor's Huffman-table repeat state.
+  - `zstd_compress_stats` converts stored sequences into symbol codes,
+    selects each block's symbol encoding types, compresses a seqStore's
+    literals and sequences into a compressed-block body, builds the block
+    entropy statistics shared with the superblock writer and the block
+    splitter, and exports collected sequences in the public `ZSTD_Sequence`
+    format.  Its C shims extract the sequence store, the entropy-table
+    leaves, and the two `ZSTD_CCtx_params` scalars these paths read.
   - `zstd_compress_frame` serializes frame headers, skippable frames, and the
     last empty block; it takes scalar frame parameters so the C-owned
     `ZSTD_CCtx_params` layout never crosses the language boundary.

rust/src/lib.rs

@@ -32,6 +32,8 @@ pub mod zstd_compress_params;
 #[cfg(feature = "compression")]
 pub mod zstd_compress_sequences;
 #[cfg(feature = "compression")]
+pub mod zstd_compress_stats;
+#[cfg(feature = "compression")]
 pub mod zstd_compress_superblock;
 #[cfg(feature = "decompression")]
 pub mod zstd_ddict;

rust/src/zstd_compress_literals.rs

@@ -65,7 +65,7 @@ unsafe fn copy_bytes(dst: *mut u8, src: *const u8, size: usize) {
 }
 
 #[inline]
-unsafe fn copy_huf_tables(dst: *mut ZSTD_hufCTables_t, src: *const ZSTD_hufCTables_t) {
+pub(crate) unsafe fn copy_huf_tables(dst: *mut ZSTD_hufCTables_t, src: *const ZSTD_hufCTables_t) {
     // Copy as bytes, not as a Rust struct: C's memcpy preserves the four tail
     // padding bytes that this layout has on 64-bit targets.  Some C callers
     // copy or compare the complete entropy state, so leaving those bytes from
@@ -192,8 +192,10 @@ fn min_literals_to_compress(strategy: c_int, huf_repeat: c_int) -> usize {
     }
 }
 
+/// C's `ZSTD_minGain()`: minimum compression gain required to emit a
+/// compressed block or compressed literals section.
 #[inline]
-fn min_gain(src_size: usize, strategy: c_int) -> usize {
+pub(crate) fn min_gain(src_size: usize, strategy: c_int) -> usize {
     let min_log = if strategy >= ZSTD_BTULTRA {
         strategy.saturating_sub(1) as u32
     } else {

rust/src/zstd_compress_stats.rs

@@ -0,0 +1,1549 @@
+#![allow(non_camel_case_types)]
+#![allow(non_snake_case)]
+
+//! Sequence statistics and seqStore entropy compression.
+//!
+//! This module translates the block entropy-statistics layer of
+//! `zstd_compress.c`: `ZSTD_seqToCodes()`, `ZSTD_buildSequencesStatistics()`,
+//! `ZSTD_entropyCompressSeqStore*()`, `ZSTD_buildBlockEntropyStats*()`, and
+//! `ZSTD_copyBlockSequences()`.  The `ZSTD_CCtx` and `ZSTD_CCtx_params`
+//! layouts stay private to C: the C shims extract the sequence store, the
+//! entropy-table leaves, and the two parameter scalars these paths read
+//! (the compression strategy and the literals-compression switch).  Block
+//! dispatch, block splitting, and the block-size estimation heuristics also
+//! remain in C for a later slice.
+
+use crate::bits::ZSTD_highbit32;
+use crate::common::{
+    DEFAULT_MAX_OFF, LL_DEFAULT_NORM, LL_DEFAULT_NORM_LOG, LL_FSE_LOG, LONGNBSEQ, MAX_LL, MAX_ML,
+    MAX_OFF, MAX_SEQ, MINMATCH, ML_DEFAULT_NORM, ML_DEFAULT_NORM_LOG, ML_FSE_LOG, OFF_FSE_LOG,
+    OF_DEFAULT_NORM, OF_DEFAULT_NORM_LOG, ZSTD_MAX_FSE_HEADERS_SIZE, ZSTD_MAX_HUF_HEADER_SIZE,
+    ZSTD_REP_NUM,
+};
+use crate::errors::{ERR_isError, ZstdErrorCode, ERROR};
+use crate::hist::{HIST_countFast_wksp, HIST_count_wksp};
+use crate::huf_compress::{
+    HUF_buildCTable_wksp, HUF_estimateCompressedSize, HUF_optimalTableLog, HUF_validateCTable,
+    HUF_writeCTable_wksp,
+};
+use crate::mem::{MEM_32bits, MEM_writeLE16};
+use crate::zstd_compress_literals::{
+    copy_huf_tables, min_gain, ZSTD_compressLiterals, ZSTD_hufCTables_t,
+};
+use crate::zstd_compress_sequences::{
+    SeqDef, ZSTD_buildCTable, ZSTD_encodeSequences, ZSTD_selectEncodingType,
+};
+use std::ffi::c_void;
+use std::mem::{size_of, size_of_val};
+use std::os::raw::{c_int, c_uint};
+use std::ptr;
+
+const SET_BASIC: c_int = 0;
+const SET_RLE: c_int = 1;
+const SET_COMPRESSED: c_int = 2;
+const SET_REPEAT: c_int = 3;
+
+const FSE_REPEAT_NONE: c_int = 0;
+
+const HUF_REPEAT_NONE: c_int = 0;
+const HUF_REPEAT_CHECK: c_int = 1;
+const HUF_REPEAT_VALID: c_int = 2;
+
+const HUF_FLAGS_OPTIMAL_DEPTH: c_int = 1 << 1;
+const HUF_OPTIMAL_DEPTH_THRESHOLD: c_int = 8; /* ZSTD_btultra */
+
+const ZSTD_DEFAULT_DISALLOWED: c_int = 0;
+const ZSTD_DEFAULT_ALLOWED: c_int = 1;
+
+const HUF_SYMBOLVALUE_MAX: c_uint = 255;
+const LIT_HUF_LOG: c_uint = 11;
+const COMPRESS_LITERALS_SIZE_MIN: usize = 63;
+const SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO: usize = 20;
+
+const STREAM_ACCUMULATOR_MIN_32: u32 = 25;
+const STREAM_ACCUMULATOR_MIN_64: u32 = 57;
+
+const LL_DELTA_CODE: u32 = 19;
+const ML_DELTA_CODE: u32 = 36;
+
+pub(crate) const ZSTD_LLT_LITERAL_LENGTH: c_int = 1;
+pub(crate) const ZSTD_LLT_MATCH_LENGTH: c_int = 2;
+
+/// `FSE_CTABLE_SIZE_U32(OffFSELog, MaxOff)` from the C headers.
+pub const OFF_CTABLE_SIZE: usize = 1 + (1 << (OFF_FSE_LOG - 1)) + ((MAX_OFF + 1) * 2);
+/// `FSE_CTABLE_SIZE_U32(MLFSELog, MaxML)` from the C headers.
+pub const ML_CTABLE_SIZE: usize = 1 + (1 << (ML_FSE_LOG - 1)) + ((MAX_ML + 1) * 2);
+/// `FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL)` from the C headers.
+pub const LL_CTABLE_SIZE: usize = 1 + (1 << (LL_FSE_LOG - 1)) + ((MAX_LL + 1) * 2);
+
+/// ABI-compatible `SeqStore_t` leaf layout from `zstd_compress_internal.h`.
+///
+/// The C context itself remains opaque; its shims pass the `seqStore` member
+/// directly, so this small, stable hot-path structure is the only sequence
+/// storage representation crossing into Rust.
+#[repr(C)]
+pub struct SeqStore_t {
+    pub sequencesStart: *mut SeqDef,
+    pub sequences: *mut SeqDef,
+    pub litStart: *mut u8,
+    pub lit: *mut u8,
+    pub llCode: *mut u8,
+    pub mlCode: *mut u8,
+    pub ofCode: *mut u8,
+    pub maxNbSeq: usize,
+    pub maxNbLit: usize,
+    pub longLengthType: c_int,
+    pub longLengthPos: u32,
+}
+
+/// C's `ZSTD_fseCTables_t`.  The table element type is `FSE_CTable`, an
+/// `unsigned`, and the table lengths are the header macros expanded above.
+#[repr(C)]
+pub struct ZSTD_fseCTables_t {
+    pub offcodeCTable: [u32; OFF_CTABLE_SIZE],
+    pub matchlengthCTable: [u32; ML_CTABLE_SIZE],
+    pub litlengthCTable: [u32; LL_CTABLE_SIZE],
+    pub offcode_repeatMode: c_int,
+    pub matchlength_repeatMode: c_int,
+    pub litlength_repeatMode: c_int,
+}
+
+/// C's `ZSTD_entropyCTables_t`.
+#[repr(C)]
+pub struct ZSTD_entropyCTables_t {
+    pub huf: ZSTD_hufCTables_t,
+    pub fse: ZSTD_fseCTables_t,
+}
+
+/// C's `ZSTD_compressedBlockState_t`.
+#[repr(C)]
+pub struct ZSTD_compressedBlockState_t {
+    pub entropy: ZSTD_entropyCTables_t,
+    pub rep: [u32; ZSTD_REP_NUM],
+}
+
+/// C's `ZSTD_hufCTablesMetadata_t`.
+#[repr(C)]
+pub struct ZSTD_hufCTablesMetadata_t {
+    pub hType: c_int,
+    pub hufDesBuffer: [u8; ZSTD_MAX_HUF_HEADER_SIZE],
+    pub hufDesSize: usize,
+}
+
+/// C's `ZSTD_fseCTablesMetadata_t`.
+#[repr(C)]
+pub struct ZSTD_fseCTablesMetadata_t {
+    pub llType: c_int,
+    pub ofType: c_int,
+    pub mlType: c_int,
+    pub fseTablesBuffer: [u8; ZSTD_MAX_FSE_HEADERS_SIZE],
+    pub fseTablesSize: usize,
+    pub lastCountSize: usize,
+}
+
+/// C's `ZSTD_entropyCTablesMetadata_t`.
+#[repr(C)]
+pub struct ZSTD_entropyCTablesMetadata_t {
+    pub hufMetadata: ZSTD_hufCTablesMetadata_t,
+    pub fseMetadata: ZSTD_fseCTablesMetadata_t,
+}
+
+/// Public `ZSTD_Sequence` from `zstd.h`.
+#[repr(C)]
+#[derive(Clone, Copy)]
+pub struct ZSTD_Sequence {
+    pub offset: u32,
+    pub litLength: u32,
+    pub matchLength: u32,
+    pub rep: u32,
+}
+
+/// C's `SeqCollector` leaf from `zstd_compress_internal.h`.
+#[repr(C)]
+pub struct SeqCollector {
+    pub collectSequences: c_int,
+    pub seqStart: *mut ZSTD_Sequence,
+    pub seqIndex: usize,
+    pub maxSequences: usize,
+}
+
+/// Result of `build_sequences_statistics()`; mirrors C's
+/// `ZSTD_symbolEncodingTypeStats_t`, which no longer crosses the boundary.
+pub(crate) struct SymbolEncodingTypeStats {
+    pub ll_type: u32,
+    pub off_type: u32,
+    pub ml_type: u32,
+    pub size: usize,
+    /// Accounts for a bug in zstd 1.3.4; see
+    /// `entropy_compress_seq_store_internal()`.
+    pub last_count_size: usize,
+    pub long_offsets: c_int,
+}
+
+/// C's `ZSTD_updateRep()` repeat-offset history update.
+#[inline]
+pub(crate) fn update_rep(reps: &mut [u32; 3], off_base: u32, literal_length_is_zero: bool) {
+    if off_base > 3 {
+        reps[2] = reps[1];
+        reps[1] = reps[0];
+        reps[0] = off_base - 3;
+        return;
+    }
+    let rep_code = off_base - 1 + u32::from(literal_length_is_zero);
+    if rep_code == 0 {
+        return;
+    }
+    let current_offset = if rep_code == 3 {
+        reps[0].wrapping_sub(1)
+    } else {
+        reps[rep_code as usize]
+    };
+    reps[2] = if rep_code >= 2 { reps[1] } else { reps[2] };
+    reps[1] = reps[0];
+    reps[0] = current_offset;
+}
+
+#[inline]
+fn ll_code(lit_length: u32) -> u32 {
+    const LL_CODE: [u8; 64] = [
+        0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 16, 17, 17, 18, 18, 19, 19, 20,
+        20, 20, 20, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22, 22, 22, 23, 23, 23, 23, 23, 23, 23, 23,
+        24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+    ];
+    if lit_length > 63 {
+        ZSTD_highbit32(lit_length) + LL_DELTA_CODE
+    } else {
+        LL_CODE[lit_length as usize] as u32
+    }
+}
+
+#[inline]
+fn ml_code(ml_base: u32) -> u32 {
+    const ML_CODE: [u8; 128] = [
+        0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24,
+        25, 26, 27, 28, 29, 30, 31, 32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 36, 36, 37, 37, 37, 37,
+        38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39, 40, 40, 40, 40, 40, 40, 40,
+        40, 40, 40, 40, 40, 40, 40, 40, 40, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
+        41, 41, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
+        42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
+    ];
+    if ml_base > 127 {
+        ZSTD_highbit32(ml_base) + ML_DELTA_CODE
+    } else {
+        ML_CODE[ml_base as usize] as u32
+    }
+}
+
+#[inline]
+unsafe fn copy_fse_tables(dst: *mut ZSTD_fseCTables_t, src: *const ZSTD_fseCTables_t) {
+    unsafe {
+        ptr::copy_nonoverlapping(
+            src.cast::<u8>(),
+            dst.cast::<u8>(),
+            size_of::<ZSTD_fseCTables_t>(),
+        )
+    };
+}
+
+/// Converts the stored lengths and distances into symbol codes.
+///
+/// Retains its original exported name: the C dictionary builder and the
+/// decodecorpus test tool link against it directly.
+#[no_mangle]
+pub unsafe extern "C" fn ZSTD_seqToCodes(seq_store: *const SeqStore_t) -> c_int {
+    let store = unsafe { &*seq_store };
+    let sequences = store.sequencesStart;
+    let nb_seq = unsafe { store.sequences.offset_from(sequences) } as usize;
+    let mut long_offsets = 0;
+    debug_assert!(nb_seq <= store.maxNbSeq);
+
+    for index in 0..nb_seq {
+        let sequence = unsafe { sequences.add(index).read() };
+        let of_code = ZSTD_highbit32(sequence.offBase);
+        unsafe {
+            *store.llCode.add(index) = ll_code(sequence.litLength as u32) as u8;
+            *store.ofCode.add(index) = of_code as u8;
+            *store.mlCode.add(index) = ml_code(sequence.mlBase as u32) as u8;
+        }
+        debug_assert!(MEM_32bits() || of_code < STREAM_ACCUMULATOR_MIN_64);
+        if MEM_32bits() && of_code >= STREAM_ACCUMULATOR_MIN_32 {
+            long_offsets = 1;
+        }
+    }
+    if store.longLengthType == ZSTD_LLT_LITERAL_LENGTH {
+        unsafe { *store.llCode.add(store.longLengthPos as usize) = MAX_LL as u8 };
+    }
+    if store.longLengthType == ZSTD_LLT_MATCH_LENGTH {
+        unsafe { *store.mlCode.add(store.longLengthPos as usize) = MAX_ML as u8 };
+    }
+    long_offsets
+}
+
+/// C's `ZSTD_buildSequencesStatistics()`.
+///
+/// Returns the statistics with a zstd error code in the `size` field on
+/// failure, and updates `next_entropy` as a side effect.  `nb_seq` must be
+/// greater than zero, and the entropy workspace must hold at least
+/// `ENTROPY_WORKSPACE_SIZE - (MaxSeq + 1) * sizeof(U32)` bytes.
+#[allow(clippy::too_many_arguments)]
+pub(crate) unsafe fn build_sequences_statistics(
+    seq_store: *const SeqStore_t,
+    nb_seq: usize,
+    prev_entropy: *const ZSTD_fseCTables_t,
+    next_entropy: *mut ZSTD_fseCTables_t,
+    dst: *mut u8,
+    dst_capacity: usize,
+    strategy: c_int,
+    count_workspace: *mut c_uint,
+    entropy_workspace: *mut c_void,
+    entropy_wksp_size: usize,
+) -> SymbolEncodingTypeStats {
+    let store = unsafe { &*seq_store };
+    let mut stats = SymbolEncodingTypeStats {
+        ll_type: 0,
+        off_type: 0,
+        ml_type: 0,
+        size: 0,
+        last_count_size: 0,
+        long_offsets: 0,
+    };
+    let mut written = 0usize;
+
+    /* convert length/distances into codes */
+    stats.long_offsets = unsafe { ZSTD_seqToCodes(seq_store) };
+    debug_assert!(nb_seq != 0); /* ZSTD_selectEncodingType() divides by nb_seq */
+
+    /* build CTable for Literal Lengths */
+    {
+        let mut max = MAX_LL as c_uint;
+        let most_frequent = unsafe {
+            HIST_countFast_wksp(
+                count_workspace,
+                &mut max,
+                store.llCode.cast::<c_void>(),
+                nb_seq,
+                entropy_workspace,
+                entropy_wksp_size,
+            )
+        }; /* can't fail */
+        unsafe { (*next_entropy).litlength_repeatMode = (*prev_entropy).litlength_repeatMode };
+        stats.ll_type = unsafe {
+            ZSTD_selectEncodingType(
+                ptr::addr_of_mut!((*next_entropy).litlength_repeatMode),
+                count_workspace,
+                max,
+                most_frequent,
+                nb_seq,
+                LL_FSE_LOG as c_uint,
+                ptr::addr_of!((*prev_entropy).litlengthCTable).cast::<u32>(),
+                LL_DEFAULT_NORM.as_ptr(),
+                LL_DEFAULT_NORM_LOG,
+                ZSTD_DEFAULT_ALLOWED,
+                strategy,
+            )
+        } as u32;
+        let count_size = unsafe {
+            ZSTD_buildCTable(
+                dst.add(written).cast::<c_void>(),
+                dst_capacity - written,
+                ptr::addr_of_mut!((*next_entropy).litlengthCTable).cast::<u32>(),
+                LL_FSE_LOG as u32,
+                stats.ll_type as c_int,
+                count_workspace,
+                max,
+                store.llCode,
+                nb_seq,
+                LL_DEFAULT_NORM.as_ptr(),
+                LL_DEFAULT_NORM_LOG,
+                MAX_LL as u32,
+                ptr::addr_of!((*prev_entropy).litlengthCTable).cast::<u32>(),
+                size_of::<[u32; LL_CTABLE_SIZE]>(),
+                entropy_workspace,
+                entropy_wksp_size,
+            )
+        };
+        if ERR_isError(count_size) {
+            stats.size = count_size;
+            return stats;
+        }
+        if stats.ll_type == SET_COMPRESSED as u32 {
+            stats.last_count_size = count_size;
+        }
+        written += count_size;
+    }
+    /* build CTable for Offsets */
+    {
+        let mut max = MAX_OFF as c_uint;
+        let most_frequent = unsafe {
+            HIST_countFast_wksp(
+                count_workspace,
+                &mut max,
+                store.ofCode.cast::<c_void>(),
+                nb_seq,
+                entropy_workspace,
+                entropy_wksp_size,
+            )
+        }; /* can't fail */
+        /* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */
+        let default_policy = if max <= DEFAULT_MAX_OFF as c_uint {
+            ZSTD_DEFAULT_ALLOWED
+        } else {
+            ZSTD_DEFAULT_DISALLOWED
+        };
+        unsafe { (*next_entropy).offcode_repeatMode = (*prev_entropy).offcode_repeatMode };
+        stats.off_type = unsafe {
+            ZSTD_selectEncodingType(
+                ptr::addr_of_mut!((*next_entropy).offcode_repeatMode),
+                count_workspace,
+                max,
+                most_frequent,
+                nb_seq,
+                OFF_FSE_LOG as c_uint,
+                ptr::addr_of!((*prev_entropy).offcodeCTable).cast::<u32>(),
+                OF_DEFAULT_NORM.as_ptr(),
+                OF_DEFAULT_NORM_LOG,
+                default_policy,
+                strategy,
+            )
+        } as u32;
+        let count_size = unsafe {
+            ZSTD_buildCTable(
+                dst.add(written).cast::<c_void>(),
+                dst_capacity - written,
+                ptr::addr_of_mut!((*next_entropy).offcodeCTable).cast::<u32>(),
+                OFF_FSE_LOG as u32,
+                stats.off_type as c_int,
+                count_workspace,
+                max,
+                store.ofCode,
+                nb_seq,
+                OF_DEFAULT_NORM.as_ptr(),
+                OF_DEFAULT_NORM_LOG,
+                DEFAULT_MAX_OFF as u32,
+                ptr::addr_of!((*prev_entropy).offcodeCTable).cast::<u32>(),
+                size_of::<[u32; OFF_CTABLE_SIZE]>(),
+                entropy_workspace,
+                entropy_wksp_size,
+            )
+        };
+        if ERR_isError(count_size) {
+            stats.size = count_size;
+            return stats;
+        }
+        if stats.off_type == SET_COMPRESSED as u32 {
+            stats.last_count_size = count_size;
+        }
+        written += count_size;
+    }
+    /* build CTable for MatchLengths */
+    {
+        let mut max = MAX_ML as c_uint;
+        let most_frequent = unsafe {
+            HIST_countFast_wksp(
+                count_workspace,
+                &mut max,
+                store.mlCode.cast::<c_void>(),
+                nb_seq,
+                entropy_workspace,
+                entropy_wksp_size,
+            )
+        }; /* can't fail */
+        unsafe { (*next_entropy).matchlength_repeatMode = (*prev_entropy).matchlength_repeatMode };
+        stats.ml_type = unsafe {
+            ZSTD_selectEncodingType(
+                ptr::addr_of_mut!((*next_entropy).matchlength_repeatMode),
+                count_workspace,
+                max,
+                most_frequent,
+                nb_seq,
+                ML_FSE_LOG as c_uint,
+                ptr::addr_of!((*prev_entropy).matchlengthCTable).cast::<u32>(),
+                ML_DEFAULT_NORM.as_ptr(),
+                ML_DEFAULT_NORM_LOG,
+                ZSTD_DEFAULT_ALLOWED,
+                strategy,
+            )
+        } as u32;
+        let count_size = unsafe {
+            ZSTD_buildCTable(
+                dst.add(written).cast::<c_void>(),
+                dst_capacity - written,
+                ptr::addr_of_mut!((*next_entropy).matchlengthCTable).cast::<u32>(),
+                ML_FSE_LOG as u32,
+                stats.ml_type as c_int,
+                count_workspace,
+                max,
+                store.mlCode,
+                nb_seq,
+                ML_DEFAULT_NORM.as_ptr(),
+                ML_DEFAULT_NORM_LOG,
+                MAX_ML as u32,
+                ptr::addr_of!((*prev_entropy).matchlengthCTable).cast::<u32>(),
+                size_of::<[u32; ML_CTABLE_SIZE]>(),
+                entropy_workspace,
+                entropy_wksp_size,
+            )
+        };
+        if ERR_isError(count_size) {
+            stats.size = count_size;
+            return stats;
+        }
+        if stats.ml_type == SET_COMPRESSED as u32 {
+            stats.last_count_size = count_size;
+        }
+        written += count_size;
+    }
+    stats.size = written;
+    stats
+}
+
+/// C's `ZSTD_entropyCompressSeqStore_internal()`: compresses literals and
+/// sequences into one compressed-block body.
+#[allow(clippy::too_many_arguments)]
+unsafe fn entropy_compress_seq_store_internal(
+    dst: *mut c_void,
+    dst_capacity: usize,
+    literals: *const c_void,
+    lit_size: usize,
+    seq_store: *const SeqStore_t,
+    prev_entropy: *const ZSTD_entropyCTables_t,
+    next_entropy: *mut ZSTD_entropyCTables_t,
+    strategy: c_int,
+    disable_literal_compression: c_int,
+    entropy_workspace: *mut c_void,
+    entropy_wksp_size: usize,
+    bmi2: c_int,
+) -> usize {
+    let count = entropy_workspace.cast::<c_uint>();
+    let entropy_workspace = unsafe { count.add(MAX_SEQ + 1) }.cast::<c_void>();
+    let entropy_wksp_size = entropy_wksp_size.wrapping_sub((MAX_SEQ + 1) * size_of::<c_uint>());
+    let store = unsafe { &*seq_store };
+    let sequences = store.sequencesStart;
+    let nb_seq = unsafe { store.sequences.offset_from(sequences) } as usize;
+    let ostart = dst.cast::<u8>();
+    let mut written = 0usize;
+    let last_count_size;
+    let long_offsets;
+
+    /* Compress literals */
+    {
+        /* Base suspicion of uncompressibility on ratio of literals to sequences */
+        let suspect_uncompressible =
+            c_int::from(nb_seq == 0 || lit_size / nb_seq >= SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO);
+        let c_size = unsafe {
+            ZSTD_compressLiterals(
+                dst,
+                dst_capacity,
+                literals,
+                lit_size,
+                entropy_workspace,
+                entropy_wksp_size,
+                ptr::addr_of!((*prev_entropy).huf),
+                ptr::addr_of_mut!((*next_entropy).huf),
+                strategy,
+                disable_literal_compression,
+                suspect_uncompressible,
+                bmi2,
+            )
+        };
+        if ERR_isError(c_size) {
+            return c_size;
+        }
+        debug_assert!(c_size <= dst_capacity);
+        written += c_size;
+    }
+
+    /* Sequences Header */
+    if (dst_capacity as isize).wrapping_sub(written as isize) < 3 /* max nbSeq Size */ + 1
+    /* seqHead */
+    {
+        return ERROR(ZstdErrorCode::DstSizeTooSmall);
+    }
+    unsafe {
+        if nb_seq < 128 {
+            *ostart.add(written) = nb_seq as u8;
+            written += 1;
+        } else if nb_seq < LONGNBSEQ as usize {
+            *ostart.add(written) = ((nb_seq >> 8) + 0x80) as u8;
+            *ostart.add(written + 1) = nb_seq as u8;
+            written += 2;
+        } else {
+            *ostart.add(written) = 0xFF;
+            MEM_writeLE16(
+                ostart.add(written + 1).cast::<c_void>(),
+                (nb_seq - LONGNBSEQ as usize) as u16,
+            );
+            written += 3;
+        }
+    }
+    if nb_seq == 0 {
+        /* Copy the old tables over as if we repeated them */
+        unsafe {
+            copy_fse_tables(
+                ptr::addr_of_mut!((*next_entropy).fse),
+                ptr::addr_of!((*prev_entropy).fse),
+            )
+        };
+        return written;
+    }
+    {
+        let seq_head = written;
+        written += 1;
+        /* build stats for sequences */
+        let stats = unsafe {
+            build_sequences_statistics(
+                seq_store,
+                nb_seq,
+                ptr::addr_of!((*prev_entropy).fse),
+                ptr::addr_of_mut!((*next_entropy).fse),
+                ostart.add(written),
+                dst_capacity - written,
+                strategy,
+                count,
+                entropy_workspace,
+                entropy_wksp_size,
+            )
+        };
+        if ERR_isError(stats.size) {
+            return stats.size;
+        }
+        unsafe {
+            *ostart.add(seq_head) =
+                ((stats.ll_type << 6) + (stats.off_type << 4) + (stats.ml_type << 2)) as u8;
+        }
+        last_count_size = stats.last_count_size;
+        written += stats.size;
+        long_offsets = stats.long_offsets;
+    }
+
+    {
+        let bitstream_size = unsafe {
+            ZSTD_encodeSequences(
+                ostart.add(written).cast::<c_void>(),
+                dst_capacity - written,
+                ptr::addr_of!((*next_entropy).fse.matchlengthCTable).cast::<u32>(),
+                store.mlCode,
+                ptr::addr_of!((*next_entropy).fse.offcodeCTable).cast::<u32>(),
+                store.ofCode,
+                ptr::addr_of!((*next_entropy).fse.litlengthCTable).cast::<u32>(),
+                store.llCode,
+                sequences,
+                nb_seq,
+                long_offsets,
+                bmi2,
+            )
+        };
+        if ERR_isError(bitstream_size) {
+            return bitstream_size;
+        }
+        written += bitstream_size;
+        /* zstd versions <= 1.3.4 mistakenly report corruption when
+         * FSE_readNCount() receives a buffer < 4 bytes.
+         * Fixed by https://github.com/facebook/zstd/pull/1146.
+         * This can happen when the last set_compressed table present is 2
+         * bytes and the bitstream is only one byte.
+         * In this exceedingly rare case, we will simply emit an uncompressed
+         * block, since it isn't worth optimizing.
+         */
+        if last_count_size != 0 && last_count_size + bitstream_size < 4 {
+            /* last_count_size >= 2 && bitstream_size > 0 ==> last_count_size == 3 */
+            debug_assert!(last_count_size + bitstream_size == 3);
+            return 0;
+        }
+    }
+
+    written
+}
+
+/// C's `ZSTD_entropyCompressSeqStore_wExtLitBuffer()`: adds the
+/// incompressibility fallbacks around the internal entropy pass.
+#[allow(clippy::too_many_arguments)]
+unsafe fn entropy_compress_seq_store_w_ext_lit_buffer(
+    dst: *mut c_void,
+    dst_capacity: usize,
+    literals: *const c_void,
+    lit_size: usize,
+    block_size: usize,
+    seq_store: *const SeqStore_t,
+    prev_entropy: *const ZSTD_entropyCTables_t,
+    next_entropy: *mut ZSTD_entropyCTables_t,
+    strategy: c_int,
+    disable_literal_compression: c_int,
+    entropy_workspace: *mut c_void,
+    entropy_wksp_size: usize,
+    bmi2: c_int,
+) -> usize {
+    let c_size = unsafe {
+        entropy_compress_seq_store_internal(
+            dst,
+            dst_capacity,
+            literals,
+            lit_size,
+            seq_store,
+            prev_entropy,
+            next_entropy,
+            strategy,
+            disable_literal_compression,
+            entropy_workspace,
+            entropy_wksp_size,
+            bmi2,
+        )
+    };
+    if c_size == 0 {
+        return 0;
+    }
+    /* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block.
+     * Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block.
+     */
+    if c_size == ERROR(ZstdErrorCode::DstSizeTooSmall) && block_size <= dst_capacity {
+        return 0; /* block not compressed */
+    }
+    if ERR_isError(c_size) {
+        return c_size;
+    }
+
+    /* Check compressibility */
+    let max_c_size = block_size.wrapping_sub(min_gain(block_size, strategy));
+    if c_size >= max_c_size {
+        return 0; /* block not compressed */
+    }
+    c_size
+}
+
+/// C ABI wrapper for the external-literal-buffer entropy pass; the C shim
+/// keeps the `ZSTD_entropyCompressSeqStore_internal()` name and extracts the
+/// two `ZSTD_CCtx_params` scalars.
+#[allow(clippy::too_many_arguments)]
+#[no_mangle]
+pub unsafe extern "C" fn ZSTD_rust_entropyCompressSeqStore_internal(
+    dst: *mut c_void,
+    dst_capacity: usize,
+    literals: *const c_void,
+    lit_size: usize,
+    seq_store: *const SeqStore_t,
+    prev_entropy: *const ZSTD_entropyCTables_t,
+    next_entropy: *mut ZSTD_entropyCTables_t,
+    strategy: c_int,
+    disable_literal_compression: c_int,
+    entropy_workspace: *mut c_void,
+    entropy_wksp_size: usize,
+    bmi2: c_int,
+) -> usize {
+    unsafe {
+        entropy_compress_seq_store_internal(
+            dst,
+            dst_capacity,
+            literals,
+            lit_size,
+            seq_store,
+            prev_entropy,
+            next_entropy,
+            strategy,
+            disable_literal_compression,
+            entropy_workspace,
+            entropy_wksp_size,
+            bmi2,
+        )
+    }
+}
+
+/// C ABI implementation behind the `ZSTD_entropyCompressSeqStore()` shim.
+#[allow(clippy::too_many_arguments)]
+#[no_mangle]
+pub unsafe extern "C" fn ZSTD_rust_entropyCompressSeqStore(
+    seq_store: *const SeqStore_t,
+    prev_entropy: *const ZSTD_entropyCTables_t,
+    next_entropy: *mut ZSTD_entropyCTables_t,
+    strategy: c_int,
+    disable_literal_compression: c_int,
+    dst: *mut c_void,
+    dst_capacity: usize,
+    src_size: usize,
+    entropy_workspace: *mut c_void,
+    entropy_wksp_size: usize,
+    bmi2: c_int,
+) -> usize {
+    let store = unsafe { &*seq_store };
+    let lit_size = unsafe { store.lit.offset_from(store.litStart) } as usize;
+    unsafe {
+        entropy_compress_seq_store_w_ext_lit_buffer(
+            dst,
+            dst_capacity,
+            store.litStart.cast::<c_void>(),
+            lit_size,
+            src_size,
+            seq_store,
+            prev_entropy,
+            next_entropy,
+            strategy,
+            disable_literal_compression,
+            entropy_workspace,
+            entropy_wksp_size,
+            bmi2,
+        )
+    }
+}
+
+/// C's `ZSTD_buildBlockEntropyStats_literals()`.
+///
+/// Stores the literals block type and Huffman description table into
+/// `huf_metadata` and returns the description size, or an error code.
+#[allow(clippy::too_many_arguments)]
+unsafe fn build_block_entropy_stats_literals(
+    src: *const c_void,
+    src_size: usize,
+    prev_huf: *const ZSTD_hufCTables_t,
+    next_huf: *mut ZSTD_hufCTables_t,
+    huf_metadata: *mut ZSTD_hufCTablesMetadata_t,
+    literals_compression_is_disabled: c_int,
+    workspace: *mut c_void,
+    wksp_size: usize,
+    huf_flags: c_int,
+) -> usize {
+    let count_wksp = workspace.cast::<c_uint>();
+    let count_wksp_size = (HUF_SYMBOLVALUE_MAX as usize + 1) * size_of::<c_uint>();
+    let node_wksp = unsafe { workspace.cast::<u8>().add(count_wksp_size) }.cast::<c_void>();
+    let node_wksp_size = wksp_size.wrapping_sub(count_wksp_size);
+    let mut max_symbol_value = HUF_SYMBOLVALUE_MAX;
+    let mut huff_log = LIT_HUF_LOG;
+    let mut repeat = unsafe { (*prev_huf).repeatMode };
+
+    /* Prepare nextEntropy assuming reusing the existing table */
+    unsafe { copy_huf_tables(next_huf, prev_huf) };
+
+    if literals_compression_is_disabled != 0 {
+        unsafe { (*huf_metadata).hType = SET_BASIC };
+        return 0;
+    }
+
+    /* small ? don't even attempt compression (speed opt) */
+    {
+        let min_lit_size = if repeat == HUF_REPEAT_VALID {
+            6
+        } else {
+            COMPRESS_LITERALS_SIZE_MIN
+        };
+        if src_size <= min_lit_size {
+            unsafe { (*huf_metadata).hType = SET_BASIC };
+            return 0;
+        }
+    }
+
+    /* Scan input and build symbol stats */
+    {
+        let largest = unsafe {
+            HIST_count_wksp(
+                count_wksp,
+                &mut max_symbol_value,
+                src,
+                src_size,
+                workspace,
+                wksp_size,
+            )
+        };
+        if ERR_isError(largest) {
+            return largest;
+        }
+        if largest == src_size {
+            /* only one literal symbol */
+            unsafe { (*huf_metadata).hType = SET_RLE };
+            return 0;
+        }
+        if largest <= (src_size >> 7) + 4 {
+            /* heuristic: likely not compressible */
+            unsafe { (*huf_metadata).hType = SET_BASIC };
+            return 0;
+        }
+    }
+
+    /* Validate the previous Huffman table */
+    if repeat == HUF_REPEAT_CHECK
+        && unsafe {
+            HUF_validateCTable(
+                ptr::addr_of!((*prev_huf).CTable).cast::<usize>(),
+                count_wksp,
+                max_symbol_value,
+            )
+        } == 0
+    {
+        repeat = HUF_REPEAT_NONE;
+    }
+
+    /* Build Huffman Tree */
+    unsafe {
+        ptr::write_bytes(
+            ptr::addr_of_mut!((*next_huf).CTable).cast::<u8>(),
+            0,
+            size_of_val(&(*next_huf).CTable),
+        );
+    }
+    huff_log = unsafe {
+        HUF_optimalTableLog(
+            huff_log,
+            src_size,
+            max_symbol_value,
+            node_wksp,
+            node_wksp_size,
+            ptr::addr_of_mut!((*next_huf).CTable).cast::<usize>(),
+            count_wksp,
+            huf_flags,
+        )
+    };
+    debug_assert!(huff_log <= LIT_HUF_LOG);
+    {
+        let max_bits = unsafe {
+            HUF_buildCTable_wksp(
+                ptr::addr_of_mut!((*next_huf).CTable).cast::<usize>(),
+                count_wksp,
+                max_symbol_value,
+                huff_log,
+                node_wksp,
+                node_wksp_size,
+            )
+        };
+        if ERR_isError(max_bits) {
+            return max_bits;
+        }
+        huff_log = max_bits as c_uint;
+    }
+    {
+        /* Build and write the CTable */
+        let new_c_size = unsafe {
+            HUF_estimateCompressedSize(
+                ptr::addr_of!((*next_huf).CTable).cast::<usize>(),
+                count_wksp,
+                max_symbol_value,
+            )
+        };
+        let h_size = unsafe {
+            HUF_writeCTable_wksp(
+                ptr::addr_of_mut!((*huf_metadata).hufDesBuffer).cast::<c_void>(),
+                size_of::<[u8; ZSTD_MAX_HUF_HEADER_SIZE]>(),
+                ptr::addr_of!((*next_huf).CTable).cast::<usize>(),
+                max_symbol_value,
+                huff_log,
+                node_wksp,
+                node_wksp_size,
+            )
+        };
+        /* Check against repeating the previous CTable */
+        if repeat != HUF_REPEAT_NONE {
+            let old_c_size = unsafe {
+                HUF_estimateCompressedSize(
+                    ptr::addr_of!((*prev_huf).CTable).cast::<usize>(),
+                    count_wksp,
+                    max_symbol_value,
+                )
+            };
+            if old_c_size < src_size
+                && (old_c_size <= h_size.wrapping_add(new_c_size)
+                    || h_size.wrapping_add(12) >= src_size)
+            {
+                unsafe {
+                    copy_huf_tables(next_huf, prev_huf);
+                    (*huf_metadata).hType = SET_REPEAT;
+                }
+                return 0;
+            }
+        }
+        if new_c_size.wrapping_add(h_size) >= src_size {
+            unsafe {
+                copy_huf_tables(next_huf, prev_huf);
+                (*huf_metadata).hType = SET_BASIC;
+            }
+            return 0;
+        }
+        unsafe {
+            (*huf_metadata).hType = SET_COMPRESSED;
+            (*next_huf).repeatMode = HUF_REPEAT_CHECK;
+        }
+        h_size
+    }
+}
+
+/// C's `ZSTD_buildDummySequencesStatistics()`: marks every encoding type as
+/// basic and resets the FSE repeat modes.
+unsafe fn build_dummy_sequences_statistics(
+    next_entropy: *mut ZSTD_fseCTables_t,
+) -> SymbolEncodingTypeStats {
+    unsafe {
+        (*next_entropy).litlength_repeatMode = FSE_REPEAT_NONE;
+        (*next_entropy).offcode_repeatMode = FSE_REPEAT_NONE;
+        (*next_entropy).matchlength_repeatMode = FSE_REPEAT_NONE;
+    }
+    SymbolEncodingTypeStats {
+        ll_type: SET_BASIC as u32,
+        off_type: SET_BASIC as u32,
+        ml_type: SET_BASIC as u32,
+        size: 0,
+        last_count_size: 0,
+        long_offsets: 0,
+    }
+}
+
+/// C's `ZSTD_buildBlockEntropyStats_sequences()`.
+#[allow(clippy::too_many_arguments)]
+unsafe fn build_block_entropy_stats_sequences(
+    seq_store: *const SeqStore_t,
+    prev_entropy: *const ZSTD_fseCTables_t,
+    next_entropy: *mut ZSTD_fseCTables_t,
+    strategy: c_int,
+    fse_metadata: *mut ZSTD_fseCTablesMetadata_t,
+    workspace: *mut c_void,
+    wksp_size: usize,
+) -> usize {
+    let store = unsafe { &*seq_store };
+    let nb_seq = unsafe { store.sequences.offset_from(store.sequencesStart) } as usize;
+    let count_workspace = workspace.cast::<c_uint>();
+    let entropy_workspace = unsafe { count_workspace.add(MAX_SEQ + 1) }.cast::<c_void>();
+    let entropy_workspace_size = wksp_size.wrapping_sub((MAX_SEQ + 1) * size_of::<c_uint>());
+
+    let stats = if nb_seq != 0 {
+        unsafe {
+            build_sequences_statistics(
+                seq_store,
+                nb_seq,
+                prev_entropy,
+                next_entropy,
+                ptr::addr_of_mut!((*fse_metadata).fseTablesBuffer).cast::<u8>(),
+                ZSTD_MAX_FSE_HEADERS_SIZE,
+                strategy,
+                count_workspace,
+                entropy_workspace,
+                entropy_workspace_size,
+            )
+        }
+    } else {
+        unsafe { build_dummy_sequences_statistics(next_entropy) }
+    };
+    if ERR_isError(stats.size) {
+        return stats.size;
+    }
+    unsafe {
+        (*fse_metadata).llType = stats.ll_type as c_int;
+        (*fse_metadata).ofType = stats.off_type as c_int;
+        (*fse_metadata).mlType = stats.ml_type as c_int;
+        (*fse_metadata).lastCountSize = stats.last_count_size;
+    }
+    stats.size
+}
+
+/// C's `ZSTD_buildBlockEntropyStats()`: builds entropy statistics for one
+/// block.  Requires an `ENTROPY_WORKSPACE_SIZE` workspace.
+#[allow(clippy::too_many_arguments)]
+pub(crate) unsafe fn build_block_entropy_stats(
+    seq_store: *const SeqStore_t,
+    prev_entropy: *const ZSTD_entropyCTables_t,
+    next_entropy: *mut ZSTD_entropyCTables_t,
+    strategy: c_int,
+    disable_literal_compression: c_int,
+    entropy_metadata: *mut ZSTD_entropyCTablesMetadata_t,
+    workspace: *mut c_void,
+    wksp_size: usize,
+) -> usize {
+    let store = unsafe { &*seq_store };
+    let lit_size = unsafe { store.lit.offset_from(store.litStart) } as usize;
+    let huf_flags = if strategy >= HUF_OPTIMAL_DEPTH_THRESHOLD {
+        HUF_FLAGS_OPTIMAL_DEPTH
+    } else {
+        0
+    };
+
+    let huf_des_size = unsafe {
+        build_block_entropy_stats_literals(
+            store.litStart.cast::<c_void>(),
+            lit_size,
+            ptr::addr_of!((*prev_entropy).huf),
+            ptr::addr_of_mut!((*next_entropy).huf),
+            ptr::addr_of_mut!((*entropy_metadata).hufMetadata),
+            disable_literal_compression,
+            workspace,
+            wksp_size,
+            huf_flags,
+        )
+    };
+    unsafe { (*entropy_metadata).hufMetadata.hufDesSize = huf_des_size };
+    if ERR_isError(huf_des_size) {
+        return huf_des_size;
+    }
+
+    let fse_tables_size = unsafe {
+        build_block_entropy_stats_sequences(
+            seq_store,
+            ptr::addr_of!((*prev_entropy).fse),
+            ptr::addr_of_mut!((*next_entropy).fse),
+            strategy,
+            ptr::addr_of_mut!((*entropy_metadata).fseMetadata),
+            workspace,
+            wksp_size,
+        )
+    };
+    unsafe { (*entropy_metadata).fseMetadata.fseTablesSize = fse_tables_size };
+    if ERR_isError(fse_tables_size) {
+        return fse_tables_size;
+    }
+    0
+}
+
+/// C ABI wrapper behind the `ZSTD_buildBlockEntropyStats()` shim, which is
+/// still exported from C for the block splitter's size estimation.
+#[allow(clippy::too_many_arguments)]
+#[no_mangle]
+pub unsafe extern "C" fn ZSTD_rust_buildBlockEntropyStats(
+    seq_store: *const SeqStore_t,
+    prev_entropy: *const ZSTD_entropyCTables_t,
+    next_entropy: *mut ZSTD_entropyCTables_t,
+    strategy: c_int,
+    disable_literal_compression: c_int,
+    entropy_metadata: *mut ZSTD_entropyCTablesMetadata_t,
+    workspace: *mut c_void,
+    wksp_size: usize,
+) -> usize {
+    unsafe {
+        build_block_entropy_stats(
+            seq_store,
+            prev_entropy,
+            next_entropy,
+            strategy,
+            disable_literal_compression,
+            entropy_metadata,
+            workspace,
+            wksp_size,
+        )
+    }
+}
+
+/// C's `ZSTD_copyBlockSequences()`: exports a block's sequences in the public
+/// `ZSTD_Sequence` format, appending the last-literals block delimiter.
+#[no_mangle]
+pub unsafe extern "C" fn ZSTD_rust_copyBlockSequences(
+    seq_collector: *mut SeqCollector,
+    seq_store: *const SeqStore_t,
+    prev_repcodes: *const u32,
+) -> usize {
+    let collector = unsafe { &mut *seq_collector };
+    let store = unsafe { &*seq_store };
+    let in_seqs = store.sequencesStart;
+    let nb_in_sequences = unsafe { store.sequences.offset_from(in_seqs) } as usize;
+    let nb_in_literals = unsafe { store.lit.offset_from(store.litStart) } as usize;
+    let out_seqs = if collector.seqIndex == 0 {
+        collector.seqStart
+    } else {
+        unsafe { collector.seqStart.add(collector.seqIndex) }
+    };
+    let nb_out_sequences = nb_in_sequences + 1;
+    let mut nb_out_literals = 0usize;
+
+    /* Bounds check that we have enough space for every input sequence
+     * and the block delimiter
+     */
+    debug_assert!(collector.seqIndex <= collector.maxSequences);
+    if nb_out_sequences > collector.maxSequences.wrapping_sub(collector.seqIndex) {
+        return ERROR(ZstdErrorCode::DstSizeTooSmall);
+    }
+
+    let mut repcodes = [0u32; ZSTD_REP_NUM];
+    unsafe { ptr::copy_nonoverlapping(prev_repcodes, repcodes.as_mut_ptr(), ZSTD_REP_NUM) };
+    for index in 0..nb_in_sequences {
+        let in_seq = unsafe { in_seqs.add(index).read() };
+        let out = unsafe { out_seqs.add(index) };
+        let mut lit_length = in_seq.litLength as u32;
+        let mut match_length = in_seq.mlBase as u32 + MINMATCH as u32;
+        let mut rep = 0u32;
+
+        /* Handle the possible single length >= 64K
+         * There can only be one because we add MINMATCH to every match length,
+         * and blocks are at most 128K.
+         */
+        if index == store.longLengthPos as usize {
+            if store.longLengthType == ZSTD_LLT_LITERAL_LENGTH {
+                lit_length += 0x10000;
+            } else if store.longLengthType == ZSTD_LLT_MATCH_LENGTH {
+                match_length += 0x10000;
+            }
+        }
+
+        /* Determine the raw offset given the offBase, which may be a repcode. */
+        let raw_offset = if (1..=ZSTD_REP_NUM as u32).contains(&in_seq.offBase) {
+            let repcode = in_seq.offBase;
+            debug_assert!(repcode > 0);
+            rep = repcode;
+            if lit_length != 0 {
+                repcodes[repcode as usize - 1]
+            } else if repcode == 3 {
+                debug_assert!(repcodes[0] > 1);
+                repcodes[0].wrapping_sub(1)
+            } else {
+                repcodes[repcode as usize]
+            }
+        } else {
+            in_seq.offBase.wrapping_sub(ZSTD_REP_NUM as u32)
+        };
+        unsafe {
+            (*out).litLength = lit_length;
+            (*out).matchLength = match_length;
+            (*out).offset = raw_offset;
+            (*out).rep = rep;
+        }
+
+        /* Update repcode history for the sequence */
+        update_rep(&mut repcodes, in_seq.offBase, in_seq.litLength == 0);
+
+        nb_out_literals += lit_length as usize;
+    }
+    /* Insert last literals (if any exist) in the block as a sequence with ml == off == 0.
+     * If there are no last literals, then we'll emit (of: 0, ml: 0, ll: 0), which is a marker
+     * for the block boundary, according to the API.
+     */
+    debug_assert!(nb_in_literals >= nb_out_literals);
+    {
+        let last_ll_size = nb_in_literals.wrapping_sub(nb_out_literals);
+        unsafe {
+            let last = out_seqs.add(nb_in_sequences);
+            (*last).litLength = last_ll_size as u32;
+            (*last).matchLength = 0;
+            (*last).offset = 0;
+        }
+    }
+    collector.seqIndex += nb_out_sequences;
+    debug_assert!(collector.seqIndex <= collector.maxSequences);
+
+    0
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::mem::{align_of, offset_of};
+
+    const HUF_CTABLE_SIZE_ST: usize = HUF_SYMBOLVALUE_MAX as usize + 2;
+    const ENTROPY_WORKSPACE_SIZE: usize = (8 << 10) + 512 + (MAX_SEQ + 2) * size_of::<u32>();
+
+    fn empty_huf_tables(repeat: c_int) -> ZSTD_hufCTables_t {
+        ZSTD_hufCTables_t {
+            CTable: [0; HUF_CTABLE_SIZE_ST],
+            repeatMode: repeat,
+        }
+    }
+
+    fn empty_fse_tables(repeat: c_int) -> ZSTD_fseCTables_t {
+        ZSTD_fseCTables_t {
+            offcodeCTable: [0; OFF_CTABLE_SIZE],
+            matchlengthCTable: [0; ML_CTABLE_SIZE],
+            litlengthCTable: [0; LL_CTABLE_SIZE],
+            offcode_repeatMode: repeat,
+            matchlength_repeatMode: repeat,
+            litlength_repeatMode: repeat,
+        }
+    }
+
+    struct TestSeqStore {
+        sequences: Vec<SeqDef>,
+        literals: Vec<u8>,
+        ll_codes: Vec<u8>,
+        ml_codes: Vec<u8>,
+        of_codes: Vec<u8>,
+    }
+
+    impl TestSeqStore {
+        fn new(sequences: Vec<SeqDef>, literals: Vec<u8>) -> Self {
+            let nb_seq = sequences.len();
+            TestSeqStore {
+                sequences,
+                literals,
+                ll_codes: vec![0; nb_seq],
+                ml_codes: vec![0; nb_seq],
+                of_codes: vec![0; nb_seq],
+            }
+        }
+
+        fn seq_store(&mut self) -> SeqStore_t {
+            let nb_seq = self.sequences.len();
+            let nb_lit = self.literals.len();
+            SeqStore_t {
+                sequencesStart: self.sequences.as_mut_ptr(),
+                sequences: unsafe { self.sequences.as_mut_ptr().add(nb_seq) },
+                litStart: self.literals.as_mut_ptr(),
+                lit: unsafe { self.literals.as_mut_ptr().add(nb_lit) },
+                llCode: self.ll_codes.as_mut_ptr(),
+                mlCode: self.ml_codes.as_mut_ptr(),
+                ofCode: self.of_codes.as_mut_ptr(),
+                maxNbSeq: nb_seq,
+                maxNbLit: nb_lit,
+                longLengthType: 0,
+                longLengthPos: 0,
+            }
+        }
+    }
+
+    #[test]
+    fn c_leaf_layouts_match_supported_abis() {
+        assert_eq!(size_of::<SeqDef>(), 8);
+        assert_eq!(align_of::<SeqDef>(), align_of::<u32>());
+        assert_eq!(offset_of!(SeqStore_t, sequencesStart), 0);
+        assert_eq!(
+            offset_of!(SeqStore_t, longLengthPos),
+            9 * size_of::<usize>() + 4
+        );
+        assert_eq!(size_of::<SeqStore_t>(), 9 * size_of::<usize>() + 8);
+        assert_eq!(size_of::<ZSTD_fseCTables_t>(), 3552);
+        assert_eq!(offset_of!(ZSTD_compressedBlockState_t, entropy), 0);
+        assert_eq!(
+            offset_of!(ZSTD_compressedBlockState_t, rep),
+            size_of::<ZSTD_entropyCTables_t>()
+        );
+        assert_eq!(size_of::<ZSTD_Sequence>(), 16);
+        assert_eq!(offset_of!(SeqCollector, seqStart), size_of::<usize>());
+        assert_eq!(offset_of!(SeqCollector, seqIndex), 2 * size_of::<usize>());
+        assert_eq!(size_of::<SeqCollector>(), 4 * size_of::<usize>());
+        if size_of::<usize>() == 8 {
+            assert_eq!(size_of::<ZSTD_hufCTablesMetadata_t>(), 144);
+            assert_eq!(size_of::<ZSTD_fseCTablesMetadata_t>(), 168);
+            assert_eq!(size_of::<ZSTD_entropyCTablesMetadata_t>(), 312);
+            assert_eq!(size_of::<ZSTD_entropyCTables_t>(), 5616);
+            assert_eq!(size_of::<ZSTD_compressedBlockState_t>(), 5632);
+        } else {
+            assert_eq!(size_of::<ZSTD_hufCTablesMetadata_t>(), 136);
+            assert_eq!(size_of::<ZSTD_fseCTablesMetadata_t>(), 156);
+            assert_eq!(size_of::<ZSTD_entropyCTablesMetadata_t>(), 292);
+            assert_eq!(size_of::<ZSTD_entropyCTables_t>(), 4584);
+            assert_eq!(size_of::<ZSTD_compressedBlockState_t>(), 4596);
+        }
+    }
+
+    #[test]
+    fn repcode_updates_match_the_c_sum_type_rules() {
+        let mut reps = [1, 4, 8];
+        update_rep(&mut reps, 10, false);
+        assert_eq!(reps, [7, 1, 4]);
+        update_rep(&mut reps, 1, true);
+        assert_eq!(reps, [1, 7, 4]);
+        update_rep(&mut reps, 2, false);
+        assert_eq!(reps, [7, 1, 4]);
+        update_rep(&mut reps, 3, false);
+        assert_eq!(reps, [4, 7, 1]);
+    }
+
+    #[test]
+    fn seq_to_codes_matches_the_reference_tables() {
+        let mut store = TestSeqStore::new(
+            vec![
+                SeqDef {
+                    offBase: 4,
+                    litLength: 0,
+                    mlBase: 0,
+                },
+                SeqDef {
+                    offBase: 1,
+                    litLength: 63,
+                    mlBase: 127,
+                },
+                SeqDef {
+                    offBase: 1 << 20,
+                    litLength: 64,
+                    mlBase: 128,
+                },
+            ],
+            vec![0; 127],
+        );
+        let seq_store = store.seq_store();
+        let long_offsets = unsafe { ZSTD_seqToCodes(&seq_store) };
+        assert_eq!(long_offsets, 0);
+        assert_eq!(store.ll_codes, [0, 24, 25]);
+        assert_eq!(store.ml_codes, [0, 42, 43]);
+        assert_eq!(store.of_codes, [2, 0, 20]);
+    }
+
+    #[test]
+    fn rle_sequences_produce_one_byte_table_headers() {
+        // Five identical sequences select set_rle for all three symbol kinds;
+        // each table header is then exactly the one RLE symbol byte.
+        let mut store = TestSeqStore::new(
+            vec![
+                SeqDef {
+                    offBase: 8,
+                    litLength: 2,
+                    mlBase: 3,
+                };
+                5
+            ],
+            vec![b'x'; 10],
+        );
+        let seq_store = store.seq_store();
+        let prev = empty_fse_tables(FSE_REPEAT_NONE);
+        let mut next = empty_fse_tables(FSE_REPEAT_NONE);
+        let mut metadata =
+            unsafe { std::mem::MaybeUninit::<ZSTD_fseCTablesMetadata_t>::zeroed().assume_init() };
+        let mut workspace = vec![0u64; ENTROPY_WORKSPACE_SIZE / size_of::<u64>()];
+
+        let size = unsafe {
+            build_block_entropy_stats_sequences(
+                &seq_store,
+                &prev,
+                &mut next,
+                1,
+                &mut metadata,
+                workspace.as_mut_ptr().cast::<c_void>(),
+                ENTROPY_WORKSPACE_SIZE,
+            )
+        };
+        assert_eq!(size, 3);
+        assert_eq!(metadata.llType, SET_RLE);
+        assert_eq!(metadata.ofType, SET_RLE);
+        assert_eq!(metadata.mlType, SET_RLE);
+        assert_eq!(metadata.lastCountSize, 0);
+        assert_eq!(&metadata.fseTablesBuffer[..3], &[2, 3, 3]);
+        assert_eq!(next.litlength_repeatMode, FSE_REPEAT_NONE);
+        assert_eq!(next.offcode_repeatMode, FSE_REPEAT_NONE);
+        assert_eq!(next.matchlength_repeatMode, FSE_REPEAT_NONE);
+    }
+
+    #[test]
+    fn empty_seq_store_emits_raw_literals_and_repeats_tables() {
+        let mut store = TestSeqStore::new(Vec::new(), b"hello".to_vec());
+        let seq_store = store.seq_store();
+        let prev = ZSTD_entropyCTables_t {
+            huf: empty_huf_tables(HUF_REPEAT_NONE),
+            fse: empty_fse_tables(2),
+        };
+        let mut next = ZSTD_entropyCTables_t {
+            huf: empty_huf_tables(HUF_REPEAT_NONE),
+            fse: empty_fse_tables(FSE_REPEAT_NONE),
+        };
+        let mut workspace = vec![0u64; ENTROPY_WORKSPACE_SIZE / size_of::<u64>()];
+        let mut output = [0xA5u8; 32];
+
+        let size = unsafe {
+            ZSTD_rust_entropyCompressSeqStore_internal(
+                output.as_mut_ptr().cast::<c_void>(),
+                output.len(),
+                seq_store.litStart.cast::<c_void>(),
+                5,
+                &seq_store,
+                &prev,
+                &mut next,
+                1,
+                0,
+                workspace.as_mut_ptr().cast::<c_void>(),
+                ENTROPY_WORKSPACE_SIZE,
+                0,
+            )
+        };
+        // Raw literals header (5 << 3), five literal bytes, nbSeq == 0.
+        assert_eq!(size, 7);
+        assert_eq!(&output[..7], &[0x28, b'h', b'e', b'l', b'l', b'o', 0x00]);
+        // With no sequences, the previous FSE tables are copied unchanged.
+        assert_eq!(next.fse.offcode_repeatMode, 2);
+        assert_eq!(next.fse.matchlength_repeatMode, 2);
+        assert_eq!(next.fse.litlength_repeatMode, 2);
+    }
+
+    #[test]
+    fn literal_stats_pick_rle_and_compressed_types() {
+        let mut workspace = vec![0u64; ENTROPY_WORKSPACE_SIZE / size_of::<u64>()];
+        let prev = empty_huf_tables(HUF_REPEAT_NONE);
+        let mut next = empty_huf_tables(HUF_REPEAT_NONE);
+        let mut metadata =
+            unsafe { std::mem::MaybeUninit::<ZSTD_hufCTablesMetadata_t>::zeroed().assume_init() };
+
+        let rle = [7u8; 128];
+        let size = unsafe {
+            build_block_entropy_stats_literals(
+                rle.as_ptr().cast::<c_void>(),
+                rle.len(),
+                &prev,
+                &mut next,
+                &mut metadata,
+                0,
+                workspace.as_mut_ptr().cast::<c_void>(),
+                ENTROPY_WORKSPACE_SIZE,
+                0,
+            )
+        };
+        assert_eq!(size, 0);
+        assert_eq!(metadata.hType, SET_RLE);
+
+        let mut compressible = [0u8; 1024];
+        for (index, byte) in compressible.iter_mut().enumerate() {
+            *byte = (index % 4) as u8;
+        }
+        let size = unsafe {
+            build_block_entropy_stats_literals(
+                compressible.as_ptr().cast::<c_void>(),
+                compressible.len(),
+                &prev,
+                &mut next,
+                &mut metadata,
+                0,
+                workspace.as_mut_ptr().cast::<c_void>(),
+                ENTROPY_WORKSPACE_SIZE,
+                0,
+            )
+        };
+        assert!(!ERR_isError(size));
+        assert!(size > 0);
+        assert_eq!(metadata.hType, SET_COMPRESSED);
+        assert_eq!(metadata.hufDesSize, 0); // caller stores the returned size
+        assert_eq!(next.repeatMode, HUF_REPEAT_CHECK);
+    }
+
+    #[test]
+    fn copied_sequences_resolve_repcodes_and_append_the_delimiter() {
+        let mut store = TestSeqStore::new(
+            vec![
+                SeqDef {
+                    offBase: 3 + 100, // raw offset 100
+                    litLength: 4,
+                    mlBase: 5,
+                },
+                SeqDef {
+                    offBase: 1, // repcode 1 with literals: rep[0] == 100
+                    litLength: 2,
+                    mlBase: 7,
+                },
+                SeqDef {
+                    offBase: 3, // repcode 3 without literals: rep[0] - 1 == 99
+                    litLength: 0,
+                    mlBase: 9,
+                },
+            ],
+            vec![0; 10],
+        );
+        let seq_store = store.seq_store();
+        let mut out = [ZSTD_Sequence {
+            offset: 0xAA,
+            litLength: 0xAA,
+            matchLength: 0xAA,
+            rep: 0xAA,
+        }; 8];
+        let mut collector = SeqCollector {
+            collectSequences: 1,
+            seqStart: out.as_mut_ptr(),
+            seqIndex: 1,
+            maxSequences: 8,
+        };
+        let reps = [1u32, 4, 8];
+
+        let result =
+            unsafe { ZSTD_rust_copyBlockSequences(&mut collector, &seq_store, reps.as_ptr()) };
+        assert_eq!(result, 0);
+        assert_eq!(collector.seqIndex, 5);
+        let expect = [
+            (100, 4, 5 + 3, 0),
+            (100, 2, 7 + 3, 1),
+            (99, 0, 9 + 3, 3),
+            (0, 10 - 6, 0, 0xAA), // delimiter: last literals; rep untouched
+        ];
+        for (index, &(offset, lit, matched, rep)) in expect.iter().enumerate() {
+            assert_eq!(out[1 + index].offset, offset, "offset {index}");
+            assert_eq!(out[1 + index].litLength, lit, "litLength {index}");
+            assert_eq!(out[1 + index].matchLength, matched, "matchLength {index}");
+            assert_eq!(out[1 + index].rep, rep, "rep {index}");
+        }
+
+        collector.maxSequences = 5;
+        let result =
+            unsafe { ZSTD_rust_copyBlockSequences(&mut collector, &seq_store, reps.as_ptr()) };
+        assert_eq!(result, ERROR(ZstdErrorCode::DstSizeTooSmall));
+    }
+}

rust/src/zstd_compress_superblock.rs

@@ -6,13 +6,13 @@
 //! This is the Rust implementation of `zstd_compress_superblock.c`.  Its C
 //! entry-point shim only extracts fields from the opaque `ZSTD_CCtx`; all
 //! sequence partitioning, literal and sequence section writing, entropy
-//! fallback, and repcode repair remain here.  The leaf layouts below are
-//! intentionally kept C-shaped and checked for both supported pointer widths.
+//! fallback, and repcode repair remain here.  The C-shaped leaf layouts are
+//! shared with `zstd_compress_stats`, which also owns the block entropy
+//! statistics this module consumes.
 
 use crate::common::{
     LL_BITS, LL_DEFAULT_NORM, LL_DEFAULT_NORM_LOG, MAX_LL, MAX_ML, MAX_OFF, MINMATCH, ML_BITS,
     ML_DEFAULT_NORM, ML_DEFAULT_NORM_LOG, OF_DEFAULT_NORM, OF_DEFAULT_NORM_LOG,
-    ZSTD_MAX_FSE_HEADERS_SIZE, ZSTD_MAX_HUF_HEADER_SIZE,
 };
 use crate::errors::{ERR_isError, ZstdErrorCode, ERROR};
 use crate::hist::{HIST_countFast_wksp, HIST_count_wksp};
@@ -23,9 +23,17 @@ use crate::mem::{MEM_32bits, MEM_writeLE16, MEM_writeLE24, MEM_writeLE32};
 use crate::zstd_compress_literals::{
     ZSTD_compressRleLiteralsBlock, ZSTD_hufCTables_t, ZSTD_noCompressLiterals,
 };
+use crate::zstd_compress_sequences::{
+    SeqDef, ZSTD_crossEntropyCost, ZSTD_encodeSequences, ZSTD_fseBitCost,
+};
+use crate::zstd_compress_stats::{
+    build_block_entropy_stats, update_rep, SeqStore_t, ZSTD_compressedBlockState_t,
+    ZSTD_entropyCTablesMetadata_t, ZSTD_entropyCTables_t, ZSTD_fseCTablesMetadata_t,
+    ZSTD_fseCTables_t, ZSTD_hufCTablesMetadata_t,
+};
 use std::ffi::c_void;
 use std::mem::MaybeUninit;
-use std::os::raw::{c_int, c_short, c_uint};
+use std::os::raw::c_int;
 use std::ptr;
 
 const SET_BASIC: c_int = 0;
@@ -42,121 +50,6 @@ const LONG_NB_SEQ: usize = 0x7f00;
 const DEFAULT_MAX_OFF: u32 = 28;
 const BYTE_SCALE: usize = 256;
 
-const OFF_CTABLE_SIZE: usize = 1 + (1 << 7) + ((MAX_OFF + 1) * 2);
-const ML_CTABLE_SIZE: usize = 1 + (1 << 8) + ((MAX_ML + 1) * 2);
-const LL_CTABLE_SIZE: usize = 1 + (1 << 8) + ((MAX_LL + 1) * 2);
-
-/// ABI-compatible `SeqDef` from `zstd_compress_internal.h`.
-#[repr(C)]
-#[derive(Clone, Copy, Debug, Default)]
-struct SeqDef {
-    offBase: u32,
-    litLength: u16,
-    mlBase: u16,
-}
-
-/// ABI-compatible `SeqStore_t` leaf layout.
-///
-/// The C context itself remains opaque.  The C shim passes its `seqStore`
-/// member directly, so this small, stable hot-path structure is the only
-/// sequence storage representation crossing into Rust.
-#[repr(C)]
-struct SeqStore_t {
-    sequencesStart: *mut SeqDef,
-    sequences: *mut SeqDef,
-    litStart: *mut u8,
-    lit: *mut u8,
-    llCode: *mut u8,
-    mlCode: *mut u8,
-    ofCode: *mut u8,
-    maxNbSeq: usize,
-    maxNbLit: usize,
-    longLengthType: c_int,
-    longLengthPos: u32,
-}
-
-/// C's `ZSTD_fseCTables_t`.  The table element type is `FSE_CTable`, an
-/// `unsigned`, and the table lengths are the header macros expanded above.
-#[repr(C)]
-struct ZSTD_fseCTables_t {
-    offcodeCTable: [u32; OFF_CTABLE_SIZE],
-    matchlengthCTable: [u32; ML_CTABLE_SIZE],
-    litlengthCTable: [u32; LL_CTABLE_SIZE],
-    offcode_repeatMode: c_int,
-    matchlength_repeatMode: c_int,
-    litlength_repeatMode: c_int,
-}
-
-#[repr(C)]
-struct ZSTD_entropyCTables_t {
-    huf: ZSTD_hufCTables_t,
-    fse: ZSTD_fseCTables_t,
-}
-
-#[repr(C)]
-struct ZSTD_compressedBlockState_t {
-    entropy: ZSTD_entropyCTables_t,
-    rep: [u32; 3],
-}
-
-#[repr(C)]
-struct ZSTD_hufCTablesMetadata_t {
-    hType: c_int,
-    hufDesBuffer: [u8; ZSTD_MAX_HUF_HEADER_SIZE],
-    hufDesSize: usize,
-}
-
-#[repr(C)]
-struct ZSTD_fseCTablesMetadata_t {
-    llType: c_int,
-    ofType: c_int,
-    mlType: c_int,
-    fseTablesBuffer: [u8; ZSTD_MAX_FSE_HEADERS_SIZE],
-    fseTablesSize: usize,
-    lastCountSize: usize,
-}
-
-#[repr(C)]
-struct ZSTD_entropyCTablesMetadata_t {
-    hufMetadata: ZSTD_hufCTablesMetadata_t,
-    fseMetadata: ZSTD_fseCTablesMetadata_t,
-}
-
-unsafe extern "C" {
-    fn ZSTD_buildBlockEntropyStats(
-        seq_store: *const SeqStore_t,
-        prev_entropy: *const ZSTD_entropyCTables_t,
-        next_entropy: *mut ZSTD_entropyCTables_t,
-        cctx_params: *const c_void,
-        entropy_metadata: *mut ZSTD_entropyCTablesMetadata_t,
-        workspace: *mut c_void,
-        wksp_size: usize,
-    ) -> usize;
-
-    fn ZSTD_encodeSequences(
-        dst: *mut c_void,
-        dst_capacity: usize,
-        ctable_match_length: *const u32,
-        ml_code_table: *const u8,
-        ctable_offset_bits: *const u32,
-        of_code_table: *const u8,
-        ctable_lit_length: *const u32,
-        ll_code_table: *const u8,
-        sequences: *const SeqDef,
-        nb_seq: usize,
-        long_offsets: c_int,
-        bmi2: c_int,
-    ) -> usize;
-
-    fn ZSTD_fseBitCost(ctable: *const u32, count: *const u32, max: c_uint) -> usize;
-    fn ZSTD_crossEntropyCost(
-        norm: *const c_short,
-        accuracy_log: c_uint,
-        count: *const c_uint,
-        max: c_uint,
-    ) -> usize;
-}
-
 #[inline]
 fn remaining_capacity(capacity: usize, written: usize) -> Result<usize, usize> {
     capacity
@@ -838,28 +731,6 @@ unsafe fn size_block_sequences(
     nb_seq
 }
 
-#[inline]
-fn update_rep(reps: &mut [u32; 3], off_base: u32, literal_length_is_zero: bool) {
-    if off_base > 3 {
-        reps[2] = reps[1];
-        reps[1] = reps[0];
-        reps[0] = off_base - 3;
-        return;
-    }
-    let rep_code = off_base - 1 + u32::from(literal_length_is_zero);
-    if rep_code == 0 {
-        return;
-    }
-    let current_offset = if rep_code == 3 {
-        reps[0].wrapping_sub(1)
-    } else {
-        reps[rep_code as usize]
-    };
-    reps[2] = if rep_code >= 2 { reps[1] } else { reps[2] };
-    reps[1] = reps[0];
-    reps[0] = current_offset;
-}
-
 #[allow(clippy::too_many_arguments, clippy::manual_checked_ops)]
 unsafe fn compress_subblock_multi(
     seq_store: *const SeqStore_t,
@@ -1098,16 +969,18 @@ unsafe fn compress_subblock_multi(
 
 /// C ABI implementation called by the declaration-only C superblock shim.
 ///
-/// `cctx_params` deliberately remains opaque: only its two required scalar
-/// fields are read by the shim, while the existing C entropy builder receives
-/// the original pointer unchanged.
+/// `ZSTD_CCtx_params` deliberately remains opaque: the C shim extracts the
+/// two scalar fields this path reads (the strategy and the
+/// literals-compression switch) and the crate-internal entropy statistics
+/// builder receives them directly.
 #[allow(clippy::too_many_arguments)]
 #[no_mangle]
 pub unsafe extern "C" fn ZSTD_rust_compressSuperBlock(
     seq_store: *const c_void,
     prev_cblock: *const c_void,
     next_cblock: *mut c_void,
-    cctx_params: *const c_void,
+    strategy: c_int,
+    disable_literal_compression: c_int,
     workspace: *mut c_void,
     wksp_size: usize,
     bmi2: c_int,
@@ -1122,16 +995,17 @@ pub unsafe extern "C" fn ZSTD_rust_compressSuperBlock(
     let seq_store = seq_store.cast::<SeqStore_t>();
     let prev_cblock = prev_cblock.cast::<ZSTD_compressedBlockState_t>();
     let next_cblock = next_cblock.cast::<ZSTD_compressedBlockState_t>();
-    // C only writes the used prefixes of its two metadata byte buffers. Start
-    // with initialized storage so treating the completed C struct as a Rust
-    // value never exposes uninitialized array elements.
+    // The statistics builder only writes the used prefixes of the two
+    // metadata byte buffers. Start with initialized storage so treating the
+    // completed struct as a Rust value never exposes uninitialized elements.
     let mut entropy_metadata = MaybeUninit::<ZSTD_entropyCTablesMetadata_t>::zeroed();
     let entropy_result = unsafe {
-        ZSTD_buildBlockEntropyStats(
+        build_block_entropy_stats(
             seq_store,
             ptr::addr_of!((*prev_cblock).entropy),
             ptr::addr_of_mut!((*next_cblock).entropy),
-            cctx_params,
+            strategy,
+            disable_literal_compression,
             entropy_metadata.as_mut_ptr(),
             workspace,
             wksp_size,
@@ -1163,50 +1037,34 @@ pub unsafe extern "C" fn ZSTD_rust_compressSuperBlock(
 
 #[cfg(test)]
 mod tests {
+    // The shared C leaf layouts and the repcode update rules are asserted in
+    // `zstd_compress_stats`, which owns those definitions.
     use super::*;
-    use std::mem::{align_of, offset_of, size_of};
 
     #[test]
-    fn c_leaf_layouts_match_supported_abis() {
-        assert_eq!(size_of::<SeqDef>(), 8);
-        assert_eq!(align_of::<SeqDef>(), align_of::<u32>());
-        assert_eq!(offset_of!(SeqStore_t, sequencesStart), 0);
-        assert_eq!(
-            offset_of!(SeqStore_t, longLengthPos),
-            9 * size_of::<usize>() + 4
-        );
-        assert_eq!(size_of::<SeqStore_t>(), 9 * size_of::<usize>() + 8);
-        assert_eq!(size_of::<ZSTD_fseCTables_t>(), 3552);
-        assert_eq!(offset_of!(ZSTD_compressedBlockState_t, entropy), 0);
-        assert_eq!(
-            offset_of!(ZSTD_compressedBlockState_t, rep),
-            size_of::<ZSTD_entropyCTables_t>()
-        );
-        if size_of::<usize>() == 8 {
-            assert_eq!(size_of::<ZSTD_hufCTablesMetadata_t>(), 144);
-            assert_eq!(size_of::<ZSTD_fseCTablesMetadata_t>(), 168);
-            assert_eq!(size_of::<ZSTD_entropyCTablesMetadata_t>(), 312);
-            assert_eq!(size_of::<ZSTD_entropyCTables_t>(), 5616);
-            assert_eq!(size_of::<ZSTD_compressedBlockState_t>(), 5632);
-        } else {
-            assert_eq!(size_of::<ZSTD_hufCTablesMetadata_t>(), 136);
-            assert_eq!(size_of::<ZSTD_fseCTablesMetadata_t>(), 156);
-            assert_eq!(size_of::<ZSTD_entropyCTablesMetadata_t>(), 292);
-            assert_eq!(size_of::<ZSTD_entropyCTables_t>(), 4584);
-            assert_eq!(size_of::<ZSTD_compressedBlockState_t>(), 4596);
-        }
-    }
+    fn subblock_sizing_respects_the_budget() {
+        let sequences = [SeqDef {
+            offBase: 4,
+            litLength: 10,
+            mlBase: 10,
+        }; 8];
 
-    #[test]
-    fn repcode_updates_match_the_c_sum_type_rules() {
-        let mut reps = [1, 4, 8];
-        update_rep(&mut reps, 10, false);
-        assert_eq!(reps, [7, 1, 4]);
-        update_rep(&mut reps, 1, true);
-        assert_eq!(reps, [1, 7, 4]);
-        update_rep(&mut reps, 2, false);
-        assert_eq!(reps, [7, 1, 4]);
-        update_rep(&mut reps, 3, false);
-        assert_eq!(reps, [4, 7, 1]);
+        // A generous budget keeps every sequence in one sub-block.
+        let all = unsafe {
+            size_block_sequences(
+                sequences.as_ptr(),
+                sequences.len(),
+                usize::MAX,
+                256,
+                256,
+                true,
+            )
+        };
+        assert_eq!(all, sequences.len());
+
+        // The first sequence always survives, even over budget.
+        let first =
+            unsafe { size_block_sequences(sequences.as_ptr(), sequences.len(), 0, 256, 256, true) };
+        assert_eq!(first, 1);
     }
 }