feat(compress): move sequence API orchestration into Rust

Move the high-level orchestration of ZSTD_compressSequences() and
ZSTD_compressSequencesAndLiterals() into Rust. Rust now owns the validation
precedence, CCtx initialization handoff, frame-header and checksum ordering,
block-loop dispatch, and output accounting for both public sequence APIs.

Keep the private CCtx, sequence-store, block-state, checksum, and conversion
layouts in C. C supplies scalar block projections and callbacks for private
initialization, frame-header emission, checksum operations, and sequence-state
preparation, preserving the ABI boundary and the existing codec leaves.

Test Plan:
- focused sequence API policy tests: 4 passed
- cargo test --manifest-path rust/Cargo.toml --all-targets -- --test-threads=1:
  533 passed
- cargo test --manifest-path rust/cli/Cargo.toml --all-targets --
  --test-threads=1: 169 passed
- legacy compression/decompression/dictionary-builder feature matrix:
  588 passed
- six library/CLI clippy gates with -D warnings
- capped serial native library/program rebuilds, 41 CLI tests, Rust library
  smoke, and full test-zstd round trips
- capped serial stress gates: 278 fuzzer cases, 84+129+143 zstream cases,
  and 1,601 decode-corpus cases
- every heavyweight command used CARGO_BUILD_JOBS=1 or make -j1 and
  ulimit -v 41943040; no worker/native process remained afterward

Commit is intentionally unsigned because GPG pinentry hangs in this
non-interactive environment.
This commit is contained in:
2026-07-19 10:12:21 +02:00
parent fa37b1d533
commit 3b989ab3ee
2 changed files with 626 additions and 169 deletions
+218 -162
View File
@@ -1021,6 +1021,72 @@ typedef char ZSTD_rust_sequence_literals_state_layout[
== 6 * sizeof(void*) + 4 * sizeof(int) + 3 * sizeof(void*))
? 1 : -1];
/* The public sequence APIs keep their CCtx initialization, frame header, and
* checksum state in C, but Rust owns the ordering and output accounting. The
* two block-loop projections above are populated by the initialization
* callback after the private CCtx has been initialized. */
typedef struct ZSTD_rust_sequenceApiState_s ZSTD_rust_sequenceApiState;
typedef size_t (*ZSTD_rust_sequenceApiInit_f)(
void* context, size_t pledgedSrcSize,
ZSTD_rust_sequenceApiState* state);
typedef size_t (*ZSTD_rust_sequenceApiWriteFrameHeader_f)(
void* context, void* dst, size_t dstCapacity, size_t pledgedSrcSize);
typedef void (*ZSTD_rust_sequenceApiUpdateChecksum_f)(
void* context, const void* src, size_t srcSize);
typedef U32 (*ZSTD_rust_sequenceApiDigestChecksum_f)(void* context);
typedef void (*ZSTD_rust_sequenceApiWriteChecksum_f)(
void* context, void* dst, U32 checksum);
struct ZSTD_rust_sequenceApiState_s {
void* callbackContext;
ZSTD_rust_sequenceCompressionState* sequenceState;
ZSTD_rust_sequenceLiteralsState* sequenceLiteralsState;
ZSTD_rust_sequenceApiInit_f init;
ZSTD_rust_sequenceApiWriteFrameHeader_f writeFrameHeader;
ZSTD_rust_sequenceApiUpdateChecksum_f updateChecksum;
ZSTD_rust_sequenceApiDigestChecksum_f digestChecksum;
ZSTD_rust_sequenceApiWriteChecksum_f writeChecksum;
int checksumFlag;
int blockDelimiters;
int validateSequences;
};
size_t ZSTD_rust_compressSequences(
ZSTD_rust_sequenceApiState* state,
void* dst, size_t dstCapacity,
const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
const void* src, size_t srcSize);
size_t ZSTD_rust_compressSequencesAndLiterals(
ZSTD_rust_sequenceApiState* state,
void* dst, size_t dstCapacity,
const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
const void* literals, size_t litSize, size_t litCapacity,
size_t decompressedSize);
typedef char ZSTD_rust_sequence_api_state_layout[
(offsetof(ZSTD_rust_sequenceApiState, callbackContext) == 0
&& offsetof(ZSTD_rust_sequenceApiState, sequenceState)
== sizeof(void*)
&& offsetof(ZSTD_rust_sequenceApiState, sequenceLiteralsState)
== 2 * sizeof(void*)
&& offsetof(ZSTD_rust_sequenceApiState, init)
== 3 * sizeof(void*)
&& offsetof(ZSTD_rust_sequenceApiState, writeFrameHeader)
== 4 * sizeof(void*)
&& offsetof(ZSTD_rust_sequenceApiState, updateChecksum)
== 5 * sizeof(void*)
&& offsetof(ZSTD_rust_sequenceApiState, digestChecksum)
== 6 * sizeof(void*)
&& offsetof(ZSTD_rust_sequenceApiState, writeChecksum)
== 7 * sizeof(void*)
&& offsetof(ZSTD_rust_sequenceApiState, checksumFlag)
== 8 * sizeof(void*)
&& offsetof(ZSTD_rust_sequenceApiState, blockDelimiters)
== 8 * sizeof(void*) + sizeof(int)
&& offsetof(ZSTD_rust_sequenceApiState, validateSequences)
== 8 * sizeof(void*) + 2 * sizeof(int)
&& sizeof(ZSTD_rust_sequenceApiState)
== (sizeof(void*) == 8 ? 80 : 44))
? 1 : -1];
typedef char ZSTD_rust_stats_seqdef_layout[(sizeof(SeqDef) == 8) ? 1 : -1];
typedef char ZSTD_rust_stats_block_summary_layout[
(sizeof(BlockSummary) == 3 * sizeof(size_t)) ? 1 : -1];
@@ -5387,100 +5453,6 @@ ZSTD_transferSequences_wBlockDelim(ZSTD_CCtx* cctx,
ZSTD_hasExtSeqProd(&cctx->appliedParams));
}
/* Compress all provided sequences, block-by-block.
*
* Returns the cumulative size of all compressed blocks (including their headers),
* otherwise a ZSTD error.
*/
static size_t
ZSTD_compressSequences_internal(ZSTD_CCtx* cctx,
void* dst, size_t dstCapacity,
const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
const void* src, size_t srcSize)
{
U32 dictSize;
ZSTD_rust_sequenceCompressionState state;
if (cctx->cdict) {
dictSize = (U32)cctx->cdict->dictContentSize;
} else if (cctx->prefixDict.dict) {
dictSize = (U32)cctx->prefixDict.dictSize;
} else {
dictSize = 0;
}
state.seqStore = &cctx->seqStore;
state.prevCBlock = &cctx->blockState.prevCBlock;
state.nextCBlock = &cctx->blockState.nextCBlock;
state.tmpWorkspace = cctx->tmpWorkspace;
state.tmpWkspSize = cctx->tmpWkspSize;
state.blockSizeMax = cctx->blockSizeMax;
state.bmi2 = cctx->bmi2;
state.blockDelimiters = (int)cctx->appliedParams.blockDelimiters;
state.strategy = (int)cctx->appliedParams.cParams.strategy;
state.disableLiteralCompression = ZSTD_literalsCompressionIsDisabled(&cctx->appliedParams);
state.searchForExternalRepcodes = (int)cctx->appliedParams.searchForExternalRepcodes;
state.validateSequences = cctx->appliedParams.validateSequences;
state.minMatch = cctx->appliedParams.cParams.minMatch;
state.windowLog = cctx->appliedParams.cParams.windowLog;
state.dictSize = dictSize;
state.useSequenceProducer = ZSTD_hasExtSeqProd(&cctx->appliedParams);
state.isFirstBlock = &cctx->isFirstBlock;
DEBUGLOG(4, "ZSTD_compressSequences_internal srcSize: %zu, inSeqsSize: %zu", srcSize, inSeqsSize);
return ZSTD_rust_compressSequencesInternal(&state, dst, dstCapacity,
inSeqs, inSeqsSize, src, srcSize);
}
size_t ZSTD_compressSequences(ZSTD_CCtx* cctx,
void* dst, size_t dstCapacity,
const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
const void* src, size_t srcSize)
{
BYTE* op = (BYTE*)dst;
size_t cSize = 0;
/* Transparent initialization stage, same as compressStream2() */
DEBUGLOG(4, "ZSTD_compressSequences (nbSeqs=%zu,dstCapacity=%zu)", inSeqsSize, dstCapacity);
assert(cctx != NULL);
FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, ZSTD_e_end, srcSize), "CCtx initialization failed");
/* Begin writing output, starting with frame header */
{ size_t const frameHeaderSize = ZSTD_writeFrameHeader(op, dstCapacity,
&cctx->appliedParams, srcSize, cctx->dictID);
op += frameHeaderSize;
assert(frameHeaderSize <= dstCapacity);
dstCapacity -= frameHeaderSize;
cSize += frameHeaderSize;
}
if (cctx->appliedParams.fParams.checksumFlag && srcSize) {
XXH64_update(&cctx->xxhState, src, srcSize);
}
/* Now generate compressed blocks */
{ size_t const cBlocksSize = ZSTD_compressSequences_internal(cctx,
op, dstCapacity,
inSeqs, inSeqsSize,
src, srcSize);
FORWARD_IF_ERROR(cBlocksSize, "Compressing blocks failed!");
cSize += cBlocksSize;
assert(cBlocksSize <= dstCapacity);
dstCapacity -= cBlocksSize;
}
/* Complete with frame checksum, if needed */
if (cctx->appliedParams.fParams.checksumFlag) {
U32 const checksum = (U32) XXH64_digest(&cctx->xxhState);
RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum");
DEBUGLOG(4, "Write checksum : %08X", (unsigned)checksum);
MEM_writeLE32((char*)dst + cSize, checksum);
cSize += 4;
}
DEBUGLOG(4, "Final compressed size: %zu", cSize);
return cSize;
}
/* The public symbol remains C-facing for fullbench and the sequence API, but
* the sequence-store conversion itself is Rust-owned. */
size_t ZSTD_convertBlockSequences(ZSTD_CCtx* cctx,
@@ -5505,41 +5477,151 @@ static size_t ZSTD_convertBlockSequencesForRust(
repcodeResolution);
}
BlockSummary ZSTD_get1BlockSummary(const ZSTD_Sequence* seqs, size_t nbSeqs)
static void ZSTD_rust_sequenceApi_prepareStates(
ZSTD_CCtx* cctx,
ZSTD_rust_sequenceCompressionState* sequenceState,
ZSTD_rust_sequenceLiteralsState* sequenceLiteralsState)
{
return ZSTD_rust_get1BlockSummary(seqs, nbSeqs);
}
U32 dictSize;
if (cctx->cdict) {
dictSize = (U32)cctx->cdict->dictContentSize;
} else if (cctx->prefixDict.dict) {
dictSize = (U32)cctx->prefixDict.dictSize;
} else {
dictSize = 0;
}
static size_t
ZSTD_compressSequencesAndLiterals_internal(ZSTD_CCtx* cctx,
void* dst, size_t dstCapacity,
const ZSTD_Sequence* inSeqs, size_t nbSequences,
const void* literals, size_t litSize, size_t srcSize)
{
ZSTD_rust_sequenceLiteralsState state;
int const repcodeResolution =
(cctx->appliedParams.searchForExternalRepcodes == ZSTD_ps_enable);
sequenceState->seqStore = &cctx->seqStore;
sequenceState->prevCBlock = &cctx->blockState.prevCBlock;
sequenceState->nextCBlock = &cctx->blockState.nextCBlock;
sequenceState->tmpWorkspace = cctx->tmpWorkspace;
sequenceState->tmpWkspSize = cctx->tmpWkspSize;
sequenceState->blockSizeMax = cctx->blockSizeMax;
sequenceState->bmi2 = cctx->bmi2;
sequenceState->blockDelimiters = (int)cctx->appliedParams.blockDelimiters;
sequenceState->strategy = (int)cctx->appliedParams.cParams.strategy;
sequenceState->disableLiteralCompression =
ZSTD_literalsCompressionIsDisabled(&cctx->appliedParams);
sequenceState->searchForExternalRepcodes =
(int)cctx->appliedParams.searchForExternalRepcodes;
sequenceState->validateSequences = cctx->appliedParams.validateSequences;
sequenceState->minMatch = cctx->appliedParams.cParams.minMatch;
sequenceState->windowLog = cctx->appliedParams.cParams.windowLog;
sequenceState->dictSize = dictSize;
sequenceState->useSequenceProducer = ZSTD_hasExtSeqProd(&cctx->appliedParams);
sequenceState->isFirstBlock = &cctx->isFirstBlock;
assert(cctx->appliedParams.searchForExternalRepcodes != ZSTD_ps_auto);
state.seqStore = &cctx->seqStore;
state.prevCBlock = &cctx->blockState.prevCBlock;
state.nextCBlock = &cctx->blockState.nextCBlock;
state.tmpWorkspace = cctx->tmpWorkspace;
state.tmpWkspSize = cctx->tmpWkspSize;
state.blockSizeMax = cctx->blockSizeMax;
state.bmi2 = cctx->bmi2;
state.strategy = (int)cctx->appliedParams.cParams.strategy;
state.disableLiteralCompression =
sequenceLiteralsState->seqStore = &cctx->seqStore;
sequenceLiteralsState->prevCBlock = &cctx->blockState.prevCBlock;
sequenceLiteralsState->nextCBlock = &cctx->blockState.nextCBlock;
sequenceLiteralsState->tmpWorkspace = cctx->tmpWorkspace;
sequenceLiteralsState->tmpWkspSize = cctx->tmpWkspSize;
sequenceLiteralsState->blockSizeMax = cctx->blockSizeMax;
sequenceLiteralsState->bmi2 = cctx->bmi2;
sequenceLiteralsState->strategy = (int)cctx->appliedParams.cParams.strategy;
sequenceLiteralsState->disableLiteralCompression =
ZSTD_literalsCompressionIsDisabled(&cctx->appliedParams);
state.repcodeResolution = repcodeResolution;
state.isFirstBlock = &cctx->isFirstBlock;
state.callbackContext = cctx;
state.convertBlockSequences = ZSTD_convertBlockSequencesForRust;
sequenceLiteralsState->repcodeResolution =
(cctx->appliedParams.searchForExternalRepcodes == ZSTD_ps_enable);
sequenceLiteralsState->isFirstBlock = &cctx->isFirstBlock;
sequenceLiteralsState->callbackContext = cctx;
sequenceLiteralsState->convertBlockSequences =
ZSTD_convertBlockSequencesForRust;
}
return ZSTD_rust_compressSequencesAndLiteralsInternal(
&state, dst, dstCapacity, inSeqs, nbSequences, literals, litSize,
srcSize);
static size_t ZSTD_rust_sequenceApi_init(
void* context, size_t pledgedSrcSize,
ZSTD_rust_sequenceApiState* state)
{
ZSTD_CCtx* const cctx = (ZSTD_CCtx*)context;
size_t initResult;
if (cctx == NULL || state == NULL
|| state->sequenceState == NULL
|| state->sequenceLiteralsState == NULL) {
return ERROR(GENERIC);
}
initResult = ZSTD_CCtx_init_compressStream2(
cctx, ZSTD_e_end, pledgedSrcSize);
if (ERR_isError(initResult)) {
return initResult;
}
ZSTD_rust_sequenceApi_prepareStates(
cctx, state->sequenceState, state->sequenceLiteralsState);
state->checksumFlag = cctx->appliedParams.fParams.checksumFlag;
state->blockDelimiters = (int)cctx->appliedParams.blockDelimiters;
state->validateSequences = cctx->appliedParams.validateSequences;
return 0;
}
static size_t ZSTD_rust_sequenceApi_writeFrameHeader(
void* context, void* dst, size_t dstCapacity, size_t pledgedSrcSize)
{
ZSTD_CCtx const* const cctx = (ZSTD_CCtx const*)context;
return ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams,
pledgedSrcSize, cctx->dictID);
}
static void ZSTD_rust_sequenceApi_updateChecksum(
void* context, const void* src, size_t srcSize)
{
ZSTD_CCtx* const cctx = (ZSTD_CCtx*)context;
(void)XXH64_update(&cctx->xxhState, src, srcSize);
}
static U32 ZSTD_rust_sequenceApi_digestChecksum(void* context)
{
ZSTD_CCtx const* const cctx = (ZSTD_CCtx const*)context;
return (U32)XXH64_digest(&cctx->xxhState);
}
static void ZSTD_rust_sequenceApi_writeChecksum(
void* context, void* dst, U32 checksum)
{
(void)context;
DEBUGLOG(4, "Write checksum : %08X", (unsigned)checksum);
MEM_writeLE32((char*)dst, checksum);
}
static void ZSTD_rust_sequenceApi_initState(
ZSTD_rust_sequenceApiState* state,
ZSTD_CCtx* cctx,
ZSTD_rust_sequenceCompressionState* sequenceState,
ZSTD_rust_sequenceLiteralsState* sequenceLiteralsState)
{
state->callbackContext = cctx;
state->sequenceState = sequenceState;
state->sequenceLiteralsState = sequenceLiteralsState;
state->init = ZSTD_rust_sequenceApi_init;
state->writeFrameHeader = ZSTD_rust_sequenceApi_writeFrameHeader;
state->updateChecksum = ZSTD_rust_sequenceApi_updateChecksum;
state->digestChecksum = ZSTD_rust_sequenceApi_digestChecksum;
state->writeChecksum = ZSTD_rust_sequenceApi_writeChecksum;
state->checksumFlag = 0;
state->blockDelimiters = 0;
state->validateSequences = 0;
}
size_t ZSTD_compressSequences(ZSTD_CCtx* cctx,
void* dst, size_t dstCapacity,
const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
const void* src, size_t srcSize)
{
ZSTD_rust_sequenceCompressionState sequenceState;
ZSTD_rust_sequenceLiteralsState sequenceLiteralsState;
ZSTD_rust_sequenceApiState state;
DEBUGLOG(4, "ZSTD_compressSequences (nbSeqs=%zu,dstCapacity=%zu)",
inSeqsSize, dstCapacity);
assert(cctx != NULL);
ZSTD_rust_sequenceApi_initState(
&state, cctx, &sequenceState, &sequenceLiteralsState);
return ZSTD_rust_compressSequences(
&state, dst, dstCapacity, inSeqs, inSeqsSize, src, srcSize);
}
size_t
@@ -5549,49 +5631,23 @@ ZSTD_compressSequencesAndLiterals(ZSTD_CCtx* cctx,
const void* literals, size_t litSize, size_t litCapacity,
size_t decompressedSize)
{
BYTE* op = (BYTE*)dst;
size_t cSize = 0;
ZSTD_rust_sequenceCompressionState sequenceState;
ZSTD_rust_sequenceLiteralsState sequenceLiteralsState;
ZSTD_rust_sequenceApiState state;
/* Transparent initialization stage, same as compressStream2() */
DEBUGLOG(4, "ZSTD_compressSequencesAndLiterals (dstCapacity=%zu)", dstCapacity);
DEBUGLOG(4, "ZSTD_compressSequencesAndLiterals (dstCapacity=%zu)",
dstCapacity);
assert(cctx != NULL);
if (litCapacity < litSize) {
RETURN_ERROR(workSpace_tooSmall, "literals buffer is not large enough: must be at least 8 bytes larger than litSize (risk of read out-of-bound)");
}
FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, ZSTD_e_end, decompressedSize), "CCtx initialization failed");
ZSTD_rust_sequenceApi_initState(
&state, cctx, &sequenceState, &sequenceLiteralsState);
return ZSTD_rust_compressSequencesAndLiterals(
&state, dst, dstCapacity, inSeqs, inSeqsSize, literals, litSize,
litCapacity, decompressedSize);
}
if (cctx->appliedParams.blockDelimiters == ZSTD_sf_noBlockDelimiters) {
RETURN_ERROR(frameParameter_unsupported, "This mode is only compatible with explicit delimiters");
}
if (cctx->appliedParams.validateSequences) {
RETURN_ERROR(parameter_unsupported, "This mode is not compatible with Sequence validation");
}
if (cctx->appliedParams.fParams.checksumFlag) {
RETURN_ERROR(frameParameter_unsupported, "this mode is not compatible with frame checksum");
}
/* Begin writing output, starting with frame header */
{ size_t const frameHeaderSize = ZSTD_writeFrameHeader(op, dstCapacity,
&cctx->appliedParams, decompressedSize, cctx->dictID);
op += frameHeaderSize;
assert(frameHeaderSize <= dstCapacity);
dstCapacity -= frameHeaderSize;
cSize += frameHeaderSize;
}
/* Now generate compressed blocks */
{ size_t const cBlocksSize = ZSTD_compressSequencesAndLiterals_internal(cctx,
op, dstCapacity,
inSeqs, inSeqsSize,
literals, litSize, decompressedSize);
FORWARD_IF_ERROR(cBlocksSize, "Compressing blocks failed!");
cSize += cBlocksSize;
assert(cBlocksSize <= dstCapacity);
dstCapacity -= cBlocksSize;
}
DEBUGLOG(4, "Final compressed size: %zu", cSize);
return cSize;
BlockSummary ZSTD_get1BlockSummary(const ZSTD_Sequence* seqs, size_t nbSeqs)
{
return ZSTD_rust_get1BlockSummary(seqs, nbSeqs);
}
/*====== Finalize ======*/