blob: 1e3c18295fba5fb7e387a5864a4466c8ae7f8800 [file] [edit]
/*
* Copyright (c) Meta Platforms, Inc. and affiliates.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
/*-************************************
* Compiler specific
**************************************/
#ifdef _MSC_VER /* Visual Studio */
# define _CRT_SECURE_NO_WARNINGS /* fgets */
# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */
# pragma warning(disable : 4204) /* disable: C4204: non-constant aggregate initializer */
#endif
/*-************************************
* Includes
**************************************/
#include <stdlib.h> /* free */
#include <stdio.h> /* fgets, sscanf */
#include <string.h> /* strcmp */
#include <time.h> /* time(), time_t */
#undef NDEBUG /* always enable assert() */
#include <assert.h>
#define ZSTD_STATIC_LINKING_ONLY /* ZSTD_compressContinue, ZSTD_compressBlock */
#include "debug.h" /* DEBUG_STATIC_ASSERT */
#include "fse.h"
#include "huf.h"
#define ZSTD_DISABLE_DEPRECATE_WARNINGS /* No deprecation warnings, we still test some deprecated functions */
#include "zstd.h" /* ZSTD_VERSION_STRING */
#include "zstd_errors.h" /* ZSTD_getErrorCode */
#define ZDICT_STATIC_LINKING_ONLY
#include "zdict.h" /* ZDICT_trainFromBuffer */
#include "mem.h"
#include "datagen.h" /* RDG_genBuffer */
#define XXH_STATIC_LINKING_ONLY /* XXH64_state_t */
#include "xxhash.h" /* XXH64 */
#include "timefn.h" /* SEC_TO_MICRO, UTIL_time_t, UTIL_TIME_INITIALIZER, UTIL_clockSpanMicro, UTIL_getTime */
/* must be included after util.h, due to ERROR macro redefinition issue on Visual Studio */
#include "zstd_internal.h" /* ZSTD_WORKSPACETOOLARGE_MAXDURATION, ZSTD_WORKSPACETOOLARGE_FACTOR, KB, MB */
#include "threading.h" /* ZSTD_pthread_create, ZSTD_pthread_join */
#include "pool.h" /* POOL_sizeof */
#include "compress/hist.h" /* HIST_count_wksp */
#include "compress/zstd_compress_internal.h" /* ZSTD_get1BlockSummary */
#include "decompress/zstd_decompress_internal.h" /* ZSTD_DCtx */
#include "decompress/zstd_ddict.h" /* empty DDict storage invariants */
#include "bitstream.h" /* BIT_reloadDStream boundary tests */
#include "dictBuilder/cover.h" /* COVER_selectDict */
/*-************************************
* Macros
**************************************/
#define COUNTOF(array) (sizeof(array) / sizeof(*(array)))
/*-************************************
* Constants
**************************************/
#define GB *(1U<<30)
static const int FUZ_compressibility_default = 50;
static const int nbTestsDefault = 30000;
/*-************************************
* Display Macros
**************************************/
#define DISPLAY(...) fprintf(stderr, __VA_ARGS__)
#define DISPLAYLEVEL(l, ...) if (g_displayLevel>=l) { DISPLAY(__VA_ARGS__); }
static U32 g_displayLevel = 2;
static const U64 g_refreshRate = SEC_TO_MICRO / 6;
static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
#define DISPLAYUPDATE(l, ...) \
if (g_displayLevel>=l) { \
if ((UTIL_clockSpanMicro(g_displayClock) > g_refreshRate) || (g_displayLevel>=4)) \
{ g_displayClock = UTIL_getTime(); DISPLAY(__VA_ARGS__); \
if (g_displayLevel>=4) fflush(stderr); } \
}
/*-*******************************************************
* Compile time test
*********************************************************/
#undef MIN
#undef MAX
/* Declaring the function, to avoid -Wmissing-prototype */
void FUZ_bug976(void);
void FUZ_bug976(void)
{ /* these constants shall not depend on MIN() macro */
DEBUG_STATIC_ASSERT(ZSTD_HASHLOG_MAX < 31);
DEBUG_STATIC_ASSERT(ZSTD_CHAINLOG_MAX < 31);
}
/*-*******************************************************
* Internal functions
*********************************************************/
#define MIN(a,b) ((a)<(b)?(a):(b))
#define MAX(a,b) ((a)>(b)?(a):(b))
#define FUZ_rotl32(x,r) ((x << r) | (x >> (32 - r)))
static U32 FUZ_rand(U32* src)
{
static const U32 prime1 = 2654435761U;
static const U32 prime2 = 2246822519U;
U32 rand32 = *src;
rand32 *= prime1;
rand32 += prime2;
rand32 = FUZ_rotl32(rand32, 13);
*src = rand32;
return rand32 >> 5;
}
static U32 FUZ_highbit32(U32 v32)
{
unsigned nbBits = 0;
if (v32==0) return 0;
while (v32) v32 >>= 1, nbBits++;
return nbBits;
}
/*=============================================
* Test macros
=============================================*/
#define CHECK(fn) { if(!(fn)) { DISPLAYLEVEL(1, "Error : test (%s) failed \n", #fn); exit(1); } }
#define CHECK_Z(f) { \
size_t const err = f; \
if (ZSTD_isError(err)) { \
DISPLAY("Error => %s : %s ", \
#f, ZSTD_getErrorName(err)); \
exit(1); \
} }
#define CHECK_VAR(var, fn) var = fn; if (ZSTD_isError(var)) { DISPLAYLEVEL(1, "%s : fails : %s \n", #fn, ZSTD_getErrorName(var)); exit(1); }
#define CHECK_NEWV(var, fn) size_t const CHECK_VAR(var, fn)
#define CHECKPLUS(var, fn, more) { CHECK_NEWV(var, fn); more; }
#define CHECK_OP(op, lhs, rhs) { \
if (!((lhs) op (rhs))) { \
DISPLAY("Error L%u => FAILED %s %s %s ", __LINE__, #lhs, #op, #rhs); \
exit(1); \
} \
}
#define CHECK_EQ(lhs, rhs) CHECK_OP(==, lhs, rhs)
#define CHECK_LT(lhs, rhs) CHECK_OP(<, lhs, rhs)
/*=============================================
* Memory Tests
=============================================*/
#if defined(__APPLE__) && defined(__MACH__)
#include <malloc/malloc.h> /* malloc_size */
typedef struct {
unsigned long long totalMalloc;
size_t currentMalloc;
size_t peakMalloc;
unsigned nbMalloc;
unsigned nbFree;
} mallocCounter_t;
static const mallocCounter_t INIT_MALLOC_COUNTER = { 0, 0, 0, 0, 0 };
static void* FUZ_mallocDebug(void* counter, size_t size)
{
mallocCounter_t* const mcPtr = (mallocCounter_t*)counter;
void* const ptr = malloc(size);
if (ptr==NULL) return NULL;
DISPLAYLEVEL(4, "allocating %u KB => effectively %u KB \n",
(unsigned)(size >> 10), (unsigned)(malloc_size(ptr) >> 10)); /* OS-X specific */
mcPtr->totalMalloc += size;
mcPtr->currentMalloc += size;
if (mcPtr->currentMalloc > mcPtr->peakMalloc)
mcPtr->peakMalloc = mcPtr->currentMalloc;
mcPtr->nbMalloc += 1;
return ptr;
}
static void FUZ_freeDebug(void* counter, void* address)
{
mallocCounter_t* const mcPtr = (mallocCounter_t*)counter;
DISPLAYLEVEL(4, "freeing %u KB \n", (unsigned)(malloc_size(address) >> 10));
mcPtr->nbFree += 1;
mcPtr->currentMalloc -= malloc_size(address); /* OS-X specific */
free(address);
}
static void FUZ_displayMallocStats(mallocCounter_t count)
{
DISPLAYLEVEL(3, "peak:%6u KB, nbMallocs:%2u, total:%6u KB \n",
(unsigned)(count.peakMalloc >> 10),
count.nbMalloc,
(unsigned)(count.totalMalloc >> 10));
}
static int FUZ_mallocTests_internal(unsigned seed, double compressibility, unsigned part,
void* inBuffer, size_t inSize, void* outBuffer, size_t outSize)
{
/* test only played in verbose mode, as they are long */
if (g_displayLevel<3) return 0;
/* Create compressible noise */
if (!inBuffer || !outBuffer) {
DISPLAY("Not enough memory, aborting\n");
exit(1);
}
RDG_genBuffer(inBuffer, inSize, compressibility, 0. /*auto*/, seed);
/* simple compression tests */
if (part <= 1)
{ int compressionLevel;
for (compressionLevel=1; compressionLevel<=6; compressionLevel++) {
mallocCounter_t malcount = INIT_MALLOC_COUNTER;
ZSTD_customMem const cMem = { FUZ_mallocDebug, FUZ_freeDebug, &malcount };
ZSTD_CCtx* const cctx = ZSTD_createCCtx_advanced(cMem);
CHECK_Z( ZSTD_compressCCtx(cctx, outBuffer, outSize, inBuffer, inSize, compressionLevel) );
ZSTD_freeCCtx(cctx);
DISPLAYLEVEL(3, "compressCCtx level %i : ", compressionLevel);
FUZ_displayMallocStats(malcount);
} }
/* streaming compression tests */
if (part <= 2)
{ int compressionLevel;
for (compressionLevel=1; compressionLevel<=6; compressionLevel++) {
mallocCounter_t malcount = INIT_MALLOC_COUNTER;
ZSTD_customMem const cMem = { FUZ_mallocDebug, FUZ_freeDebug, &malcount };
ZSTD_CCtx* const cstream = ZSTD_createCStream_advanced(cMem);
ZSTD_outBuffer out = { outBuffer, outSize, 0 };
ZSTD_inBuffer in = { inBuffer, inSize, 0 };
CHECK_Z( ZSTD_initCStream(cstream, compressionLevel) );
CHECK_Z( ZSTD_compressStream(cstream, &out, &in) );
CHECK_Z( ZSTD_endStream(cstream, &out) );
ZSTD_freeCStream(cstream);
DISPLAYLEVEL(3, "compressStream level %i : ", compressionLevel);
FUZ_displayMallocStats(malcount);
} }
/* advanced MT API test */
if (part <= 3)
{ int nbThreads;
for (nbThreads=1; nbThreads<=4; nbThreads++) {
int compressionLevel;
for (compressionLevel=1; compressionLevel<=6; compressionLevel++) {
mallocCounter_t malcount = INIT_MALLOC_COUNTER;
ZSTD_customMem const cMem = { FUZ_mallocDebug, FUZ_freeDebug, &malcount };
ZSTD_CCtx* const cctx = ZSTD_createCCtx_advanced(cMem);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, compressionLevel) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, nbThreads) );
CHECK_Z( ZSTD_compress2(cctx, outBuffer, outSize, inBuffer, inSize) );
ZSTD_freeCCtx(cctx);
DISPLAYLEVEL(3, "compress_generic,-T%i,end level %i : ",
nbThreads, compressionLevel);
FUZ_displayMallocStats(malcount);
} } }
/* advanced MT streaming API test */
if (part <= 4)
{ int nbThreads;
for (nbThreads=1; nbThreads<=4; nbThreads++) {
int compressionLevel;
for (compressionLevel=1; compressionLevel<=6; compressionLevel++) {
mallocCounter_t malcount = INIT_MALLOC_COUNTER;
ZSTD_customMem const cMem = { FUZ_mallocDebug, FUZ_freeDebug, &malcount };
ZSTD_CCtx* const cctx = ZSTD_createCCtx_advanced(cMem);
ZSTD_outBuffer out = { outBuffer, outSize, 0 };
ZSTD_inBuffer in = { inBuffer, inSize, 0 };
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, compressionLevel) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, nbThreads) );
CHECK_Z( ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue) );
while ( ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end) ) {}
ZSTD_freeCCtx(cctx);
DISPLAYLEVEL(3, "compress_generic,-T%i,continue level %i : ",
nbThreads, compressionLevel);
FUZ_displayMallocStats(malcount);
} } }
return 0;
}
static int FUZ_mallocTests(unsigned seed, double compressibility, unsigned part)
{
size_t const inSize = 64 MB + 16 MB + 4 MB + 1 MB + 256 KB + 64 KB; /* 85.3 MB */
size_t const outSize = ZSTD_compressBound(inSize);
void* const inBuffer = malloc(inSize);
void* const outBuffer = malloc(outSize);
int result;
/* Create compressible noise */
if (!inBuffer || !outBuffer) {
DISPLAY("Not enough memory, aborting \n");
exit(1);
}
result = FUZ_mallocTests_internal(seed, compressibility, part,
inBuffer, inSize, outBuffer, outSize);
free(inBuffer);
free(outBuffer);
return result;
}
#else
static int FUZ_mallocTests(unsigned seed, double compressibility, unsigned part)
{
(void)seed; (void)compressibility; (void)part;
return 0;
}
#endif
static void FUZ_decodeSequences(BYTE* dst, ZSTD_Sequence* seqs, size_t seqsSize,
BYTE* src, size_t size, ZSTD_SequenceFormat_e format)
{
size_t i;
size_t j;
for(i = 0; i < seqsSize; ++i) {
assert(dst + seqs[i].litLength + seqs[i].matchLength <= dst + size);
assert(src + seqs[i].litLength + seqs[i].matchLength <= src + size);
if (format == ZSTD_sf_noBlockDelimiters) {
assert(seqs[i].matchLength != 0 || seqs[i].offset != 0);
}
memcpy(dst, src, seqs[i].litLength);
dst += seqs[i].litLength;
src += seqs[i].litLength;
size -= seqs[i].litLength;
if (seqs[i].offset != 0) {
for (j = 0; j < seqs[i].matchLength; ++j)
dst[j] = dst[(ptrdiff_t)(j - seqs[i].offset)];
dst += seqs[i].matchLength;
src += seqs[i].matchLength;
size -= seqs[i].matchLength;
}
}
if (format == ZSTD_sf_noBlockDelimiters) {
memcpy(dst, src, size);
}
}
static size_t FUZ_getLitSize(const ZSTD_Sequence* seqs, size_t nbSeqs)
{
size_t n, litSize = 0;
assert(seqs != NULL);
for (n=0; n<nbSeqs; n++) {
litSize += seqs[n].litLength;
}
return litSize;
}
static void
FUZ_transferLiterals(void* dst, size_t dstCapacity,
const void* src, size_t srcSize,
const ZSTD_Sequence* seqs, size_t nbSeqs)
{
size_t n;
const char* ip = (const char*)src;
char* op = (char*)dst;
size_t const litSize = FUZ_getLitSize(seqs, nbSeqs);
assert(litSize <= dstCapacity);
for (n=0; n<nbSeqs; n++) {
size_t const ll = seqs[n].litLength;
memcpy(op, ip, ll);
op += ll;
ip += ll + seqs[n].matchLength;
}
assert((size_t)(ip - (const char*)src) == srcSize);
}
#ifdef ZSTD_MULTITHREAD
typedef struct {
ZSTD_CCtx* cctx;
ZSTD_threadPool* pool;
void* CNBuffer;
size_t CNBuffSize;
void* compressedBuffer;
size_t compressedBufferSize;
void* decodedBuffer;
int err;
} threadPoolTests_compressionJob_payload;
static void* threadPoolTests_compressionJob(void* payload) {
threadPoolTests_compressionJob_payload* args = (threadPoolTests_compressionJob_payload*)payload;
size_t cSize;
if (ZSTD_isError(ZSTD_CCtx_refThreadPool(args->cctx, args->pool))) args->err = 1;
cSize = ZSTD_compress2(args->cctx, args->compressedBuffer, args->compressedBufferSize, args->CNBuffer, args->CNBuffSize);
if (ZSTD_isError(cSize)) args->err = 1;
if (ZSTD_isError(ZSTD_decompress(args->decodedBuffer, args->CNBuffSize, args->compressedBuffer, cSize))) args->err = 1;
return payload;
}
static int threadPoolTests(void) {
int testResult = 0;
size_t err;
size_t const CNBuffSize = 5 MB;
void* const CNBuffer = malloc(CNBuffSize);
size_t const compressedBufferSize = ZSTD_compressBound(CNBuffSize);
void* const compressedBuffer = malloc(compressedBufferSize);
void* const decodedBuffer = malloc(CNBuffSize);
size_t const kPoolNumThreads = 8;
RDG_genBuffer(CNBuffer, CNBuffSize, 0.5, 0.5, 0);
DISPLAYLEVEL(3, "thread pool test : threadPool reuse roundtrips: ");
{
ZSTD_CCtx* cctx = ZSTD_createCCtx();
ZSTD_threadPool* pool = ZSTD_createThreadPool(kPoolNumThreads);
size_t nbThreads = 1;
for (; nbThreads <= kPoolNumThreads; ++nbThreads) {
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, (int)nbThreads);
err = ZSTD_CCtx_refThreadPool(cctx, pool);
if (ZSTD_isError(err)) {
DISPLAYLEVEL(3, "refThreadPool error!\n");
ZSTD_freeCCtx(cctx);
goto _output_error;
}
err = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
if (ZSTD_isError(err)) {
DISPLAYLEVEL(3, "Compression error!\n");
ZSTD_freeCCtx(cctx);
goto _output_error;
}
err = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, err);
if (ZSTD_isError(err)) {
DISPLAYLEVEL(3, "Decompression error!\n");
ZSTD_freeCCtx(cctx);
goto _output_error;
}
}
ZSTD_freeCCtx(cctx);
ZSTD_freeThreadPool(pool);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "thread pool test : threadPool simultaneous usage: ");
{
void* const decodedBuffer2 = malloc(CNBuffSize);
void* const compressedBuffer2 = malloc(compressedBufferSize);
ZSTD_threadPool* pool = ZSTD_createThreadPool(kPoolNumThreads);
ZSTD_CCtx* cctx1 = ZSTD_createCCtx();
ZSTD_CCtx* cctx2 = ZSTD_createCCtx();
ZSTD_pthread_t t1;
ZSTD_pthread_t t2;
threadPoolTests_compressionJob_payload p1 = {cctx1, pool, CNBuffer, CNBuffSize,
compressedBuffer, compressedBufferSize, decodedBuffer, 0 /* err */};
threadPoolTests_compressionJob_payload p2 = {cctx2, pool, CNBuffer, CNBuffSize,
compressedBuffer2, compressedBufferSize, decodedBuffer2, 0 /* err */};
ZSTD_CCtx_setParameter(cctx1, ZSTD_c_nbWorkers, 2);
ZSTD_CCtx_setParameter(cctx2, ZSTD_c_nbWorkers, 2);
ZSTD_CCtx_refThreadPool(cctx1, pool);
ZSTD_CCtx_refThreadPool(cctx2, pool);
ZSTD_pthread_create(&t1, NULL, threadPoolTests_compressionJob, &p1);
ZSTD_pthread_create(&t2, NULL, threadPoolTests_compressionJob, &p2);
ZSTD_pthread_join(t1);
ZSTD_pthread_join(t2);
assert(!memcmp(decodedBuffer, decodedBuffer2, CNBuffSize));
free(decodedBuffer2);
free(compressedBuffer2);
ZSTD_freeThreadPool(pool);
ZSTD_freeCCtx(cctx1);
ZSTD_freeCCtx(cctx2);
if (p1.err || p2.err) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "thread pool test : change the pool and nbWorkers of a context: ");
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_threadPool* const pool1 = ZSTD_createThreadPool(1);
ZSTD_threadPool* const pool2 = ZSTD_createThreadPool(1);
CHECK(cctx != NULL && pool1 != NULL && pool2 != NULL);
CHECK(!ZSTD_isError( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, 2) ));
CHECK(!ZSTD_isError( ZSTD_CCtx_refThreadPool(cctx, pool1) ));
CHECK(!ZSTD_isError( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize) ));
/* the next frames must use pool2, then an internal pool */
CHECK(!ZSTD_isError( ZSTD_CCtx_refThreadPool(cctx, pool2) ));
ZSTD_freeThreadPool(pool1);
CHECK(!ZSTD_isError( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize) ));
{ size_t const pool2Size = POOL_sizeof(pool2);
/* the context must not resize a pool it shares */
CHECK(!ZSTD_isError( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, 4) ));
CHECK(!ZSTD_isError( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize) ));
CHECK(POOL_sizeof(pool2) == pool2Size);
}
CHECK(!ZSTD_isError( ZSTD_CCtx_refThreadPool(cctx, NULL) ));
ZSTD_freeThreadPool(pool2);
CHECK(!ZSTD_isError( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize) ));
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
_end:
free(CNBuffer);
free(compressedBuffer);
free(decodedBuffer);
return testResult;
_output_error:
testResult = 1;
DISPLAY("Error detected in Unit tests ! \n");
goto _end;
}
#endif /* ZSTD_MULTITHREAD */
/*=============================================
* Unit tests
=============================================*/
static void test_compressBound(unsigned tnb)
{
DISPLAYLEVEL(3, "test%3u : compressBound : ", tnb);
/* check ZSTD_compressBound == ZSTD_COMPRESSBOUND
* for a large range of known valid values */
DEBUG_STATIC_ASSERT(sizeof(size_t) >= 4);
{ int s;
for (s=0; s<30; s++) {
size_t const w = (size_t)1 << s;
CHECK_EQ(ZSTD_compressBound(w), ZSTD_COMPRESSBOUND(w));
} }
/* Ensure error if srcSize too big */
{ size_t const w = ZSTD_MAX_INPUT_SIZE + 1;
CHECK(ZSTD_isError(ZSTD_compressBound(w))); /* must fail */
CHECK_EQ(ZSTD_COMPRESSBOUND(w), 0);
}
DISPLAYLEVEL(3, "OK \n");
}
static void test_decompressBound(unsigned tnb)
{
DISPLAYLEVEL(3, "test%3u : decompressBound : ", tnb);
/* Simple compression, with size : should provide size; */
{ const char example[] = "abcd";
char cBuffer[ZSTD_COMPRESSBOUND(sizeof(example))];
size_t const cSize = ZSTD_compress(cBuffer, sizeof(cBuffer), example, sizeof(example), 0);
CHECK_Z(cSize);
CHECK_EQ(ZSTD_decompressBound(cBuffer, cSize), (unsigned long long)sizeof(example));
}
/* Simple small compression without size : should provide 1 block size */
{ char cBuffer[ZSTD_COMPRESSBOUND(0)];
ZSTD_outBuffer out = { cBuffer, sizeof(cBuffer), 0 };
ZSTD_inBuffer in = { NULL, 0, 0 };
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
assert(cctx);
CHECK_Z( ZSTD_initCStream(cctx, 0) );
CHECK_Z( ZSTD_compressStream(cctx, &out, &in) );
CHECK_EQ( ZSTD_endStream(cctx, &out), 0 );
CHECK_EQ( ZSTD_decompressBound(cBuffer, out.pos), ZSTD_BLOCKSIZE_MAX );
ZSTD_freeCCtx(cctx);
}
/* Attempt to overflow 32-bit intermediate multiplication result
* This requires dBound >= 4 GB, aka 2^32.
* This requires 2^32 / 2^17 = 2^15 blocks
* => create 2^15 blocks (can be empty, or just 1 byte). */
{ const char input[] = "a";
size_t const nbBlocks = (1 << 15) + 1;
size_t blockNb;
size_t const outCapacity = 1 << 18; /* large margin */
char* const outBuffer = malloc (outCapacity);
ZSTD_outBuffer out = { outBuffer, outCapacity, 0 };
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
assert(cctx);
assert(outBuffer);
CHECK_Z( ZSTD_initCStream(cctx, 0) );
for (blockNb=0; blockNb<nbBlocks; blockNb++) {
ZSTD_inBuffer in = { input, sizeof(input), 0 };
CHECK_Z( ZSTD_compressStream(cctx, &out, &in) );
CHECK_EQ( ZSTD_flushStream(cctx, &out), 0 );
}
CHECK_EQ( ZSTD_endStream(cctx, &out), 0 );
CHECK( ZSTD_decompressBound(outBuffer, out.pos) > 0x100000000ULL /* 4 GB */ );
ZSTD_freeCCtx(cctx);
free(outBuffer);
}
DISPLAYLEVEL(3, "OK \n");
}
/* Size-inspection APIs trust the declared content size without decoding the
* blocks. Keep the frame structurally complete while varying that field. */
static const BYTE FUZ_contentSizeFrame[] = {
0x28, 0xB5, 0x2F, 0xFD, 0xE0,
0, 0, 0, 0, 0, 0, 0, 0,
0x01, 0x00, 0x00
};
static void FUZ_writeContentSizeFrame(BYTE* dst, U64 contentSize)
{
memcpy(dst, FUZ_contentSizeFrame, sizeof(FUZ_contentSizeFrame));
MEM_writeLE64(dst + ZSTD_FRAMEIDSIZE + 1, contentSize);
}
static void test_decompressBoundOverflow(unsigned tnb)
{
static const U64 tails[] = { 0, 1, 0xFC, 0xFD, 0xFE, 0xFF, 0x100, 0x110 };
static const BYTE skippable[8] = { 0x50, 0x2A, 0x4D, 0x18, 0, 0, 0, 0 };
static const BYTE unknownSize[9] = {
0x28, 0xB5, 0x2F, 0xFD, 0x00, 0x00, 0x01, 0x00, 0x00
};
U64 const first = 0xFFFFFFFFFFFFFF00ULL;
U64 const maxBound = ZSTD_CONTENTSIZE_ERROR - 1;
size_t const frameSize = sizeof(FUZ_contentSizeFrame);
size_t const skippableEnd = frameSize + sizeof(skippable);
BYTE frames[2 * sizeof(FUZ_contentSizeFrame) + sizeof(skippable)];
size_t i;
DISPLAYLEVEL(3, "test%3u : decompressBound concatenation overflow : ", tnb);
CHECK_EQ(ZSTD_decompressBound(NULL, 0), 0);
FUZ_writeContentSizeFrame(frames, maxBound);
CHECK_EQ(ZSTD_decompressBound(frames, frameSize), maxBound);
FUZ_writeContentSizeFrame(frames, ZSTD_CONTENTSIZE_ERROR);
CHECK_EQ(ZSTD_decompressBound(frames, frameSize), ZSTD_CONTENTSIZE_ERROR);
FUZ_writeContentSizeFrame(frames, ZSTD_CONTENTSIZE_UNKNOWN);
/* Regardless of how the header parser treats this reserved field value,
* decompressBound() must return a numeric bound or ERROR, never UNKNOWN. */
CHECK(ZSTD_decompressBound(frames, frameSize) != ZSTD_CONTENTSIZE_UNKNOWN);
FUZ_writeContentSizeFrame(frames, first);
CHECK_EQ(ZSTD_decompressBound(frames, frameSize), first);
for (i = 0; i < COUNTOF(tails); ++i) {
U64 const expected = tails[i] <= maxBound - first
? first + tails[i] : ZSTD_CONTENTSIZE_ERROR;
FUZ_writeContentSizeFrame(frames + frameSize, tails[i]);
CHECK_EQ(ZSTD_decompressBound(frames, 2 * frameSize), expected);
}
/* The third addition must use the accumulated total, not the first size. */
{ BYTE threeFrames[3 * sizeof(FUZ_contentSizeFrame)];
FUZ_writeContentSizeFrame(threeFrames, first);
FUZ_writeContentSizeFrame(threeFrames + frameSize, 0x80);
FUZ_writeContentSizeFrame(threeFrames + 2 * frameSize, 0x7D);
CHECK_EQ(ZSTD_decompressBound(threeFrames, sizeof(threeFrames)), maxBound);
FUZ_writeContentSizeFrame(threeFrames + 2 * frameSize, 0x7E);
CHECK_EQ(ZSTD_decompressBound(threeFrames, sizeof(threeFrames)), ZSTD_CONTENTSIZE_ERROR);
}
/* A skippable frame contributes zero, including at the largest result. */
memcpy(frames + frameSize, skippable, sizeof(skippable));
FUZ_writeContentSizeFrame(frames + skippableEnd, maxBound - first);
CHECK_EQ(ZSTD_decompressBound(frames, sizeof(frames)), maxBound);
FUZ_writeContentSizeFrame(frames + skippableEnd, 0x100);
CHECK_EQ(ZSTD_decompressBound(frames, sizeof(frames)), ZSTD_CONTENTSIZE_ERROR);
CHECK_EQ(ZSTD_decompressBound(skippable, sizeof(skippable)), 0);
/* Unknown-size frames have a numeric bound, not an UNKNOWN sentinel. */
CHECK_EQ(ZSTD_getFrameContentSize(unknownSize, sizeof(unknownSize)), ZSTD_CONTENTSIZE_UNKNOWN);
CHECK_EQ(ZSTD_decompressBound(unknownSize, sizeof(unknownSize)), 1024);
FUZ_writeContentSizeFrame(frames, maxBound - 1024);
memcpy(frames + frameSize, unknownSize, sizeof(unknownSize));
CHECK_EQ(ZSTD_decompressBound(frames, frameSize + sizeof(unknownSize)), maxBound);
FUZ_writeContentSizeFrame(frames, maxBound - 1023);
CHECK_EQ(ZSTD_decompressBound(frames, frameSize + sizeof(unknownSize)), ZSTD_CONTENTSIZE_ERROR);
/* Zero-sized frames remain valid; malformed trailing data does not. */
FUZ_writeContentSizeFrame(frames, 0);
memcpy(frames + frameSize, skippable, sizeof(skippable));
CHECK_EQ(ZSTD_decompressBound(frames, skippableEnd), 0);
CHECK_EQ(ZSTD_decompressBound(frames, skippableEnd - 1), ZSTD_CONTENTSIZE_ERROR);
frames[skippableEnd] = 0;
CHECK_EQ(ZSTD_decompressBound(frames, skippableEnd + 1), ZSTD_CONTENTSIZE_ERROR);
DISPLAYLEVEL(3, "OK \n");
}
/* Write a compressed block with Huffman literals and no sequences. An empty
* table header selects repeat mode, otherwise this installs a new table. */
static size_t FUZ_writeHuffmanLiteralBlock(BYTE* block, size_t capacity,
const BYTE* literals, size_t literalSize, const HUF_CElt* ctable,
const BYTE* tableHeader, size_t tableHeaderSize, int fourStreams)
{
size_t hufSize;
size_t cLiteralSize;
CHECK((tableHeader == NULL) == (tableHeaderSize == 0));
CHECK(capacity > 4 + tableHeaderSize);
if (tableHeaderSize != 0)
memcpy(block + 3, tableHeader, tableHeaderSize);
hufSize = fourStreams
? HUF_compress4X_usingCTable(block + 3 + tableHeaderSize,
capacity - 4 - tableHeaderSize, literals, literalSize, ctable, 0)
: HUF_compress1X_usingCTable(block + 3 + tableHeaderSize,
capacity - 4 - tableHeaderSize, literals, literalSize, ctable, 0);
CHECK_Z(hufSize);
CHECK(hufSize != 0);
cLiteralSize = tableHeaderSize + hufSize;
CHECK(literalSize < 1024 && cLiteralSize < 1024);
MEM_writeLE24(block, (U32)(tableHeaderSize ? set_compressed : set_repeat)
| ((U32)fourStreams << 2)
| ((U32)literalSize << 4) | ((U32)cLiteralSize << 14));
block[3 + cLiteralSize] = 0; /* no sequences */
return 4 + cLiteralSize;
}
/* The malformed cases regress the forced-X2 dispatch bug. Default and
* forced-X1 builds already reject them and serve as compatibility controls. */
static void test_decompressHuffmanTable(unsigned tnb)
{
enum { dictionaryHeaderSize = 8, fseTableCount = 3, fseNCountSize = 2,
dictionaryContentSize = 16 };
BYTE literals[32];
BYTE output[128];
BYTE tableHeader[HUF_CTABLEBOUND];
BYTE block[128];
BYTE repeatBlock[128];
BYTE frame[256];
/* Magic/id, Huffman table, three FSE tables, repcodes, and content. */
BYTE dictionary[dictionaryHeaderSize + HUF_CTABLEBOUND
+ fseTableCount * fseNCountSize + ZSTD_REP_NUM * 4
+ dictionaryContentSize];
unsigned counts[HUF_SYMBOLVALUE_MAX + 1] = { 0 };
HUF_CREATE_STATIC_CTABLE(ctable, HUF_SYMBOLVALUE_MAX);
HUF_CREATE_STATIC_DTABLEX2(dtable, HUF_TABLELOG_MAX);
U64 workspace[HUF_WORKSPACE_SIZE_U64];
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
ZSTD_DDict* ddict;
size_t tableLog;
size_t tableHeaderSize;
size_t dictionarySize;
size_t repeatSize;
size_t i;
unsigned fourStreams;
DISPLAYLEVEL(3, "test%3u : require a table for compressed Huffman literals : ", tnb);
CHECK(dctx != NULL);
for (i = 0; i < sizeof(literals); ++i) {
literals[i] = (BYTE)(i & 1);
++counts[literals[i]];
}
CHECK_VAR(tableLog, HUF_buildCTable_wksp(ctable, counts, 1,
HUF_TABLELOG_DEFAULT, workspace, sizeof(workspace)));
CHECK_VAR(tableHeaderSize, HUF_writeCTable_wksp(tableHeader,
sizeof(tableHeader), ctable, 1, (unsigned)tableLog,
workspace, sizeof(workspace)));
CHECK(tableHeaderSize != 0);
/* A formatted dictionary with this Huffman table, minimal FSE tables,
* repcodes of 4, and 16 bytes of content. */
MEM_writeLE32(dictionary, ZSTD_MAGIC_DICTIONARY);
MEM_writeLE32(dictionary + 4, 1);
memcpy(dictionary + dictionaryHeaderSize, tableHeader, tableHeaderSize);
dictionarySize = dictionaryHeaderSize + tableHeaderSize;
for (i = 0; i < fseTableCount; ++i) {
/* FSE NCount: tableLog 5, normalized count 32 for symbol 0. */
MEM_writeLE16(dictionary + dictionarySize, 0x03F0);
dictionarySize += fseNCountSize;
}
for (i = 0; i < ZSTD_REP_NUM; ++i) {
MEM_writeLE32(dictionary + dictionarySize, 4);
dictionarySize += 4;
}
memset(dictionary + dictionarySize, 0, dictionaryContentSize);
dictionarySize += dictionaryContentSize;
ddict = ZSTD_createDDict(dictionary, dictionarySize);
CHECK(ddict != NULL);
repeatSize = FUZ_writeHuffmanLiteralBlock(repeatBlock, sizeof(repeatBlock),
literals, sizeof(literals), ctable, NULL, 0, 0);
for (i = 0; i < 2; ++i) {
size_t const cLiteralSize = i == 0 ? 1 : 16;
size_t const blockSize = 4 + cLiteralSize;
size_t frameSize = 6;
unsigned withDict;
/* These are the generic HUF decoder's RLE and raw shortcuts, but
* neither is a valid Zstd compressed-literals Huffman table. */
MEM_writeLE24(block, set_compressed | (16U << 4)
| ((U32)cLiteralSize << 14));
memset(block + 3, 'A', cLiteralSize);
block[3 + cLiteralSize] = 0;
for (withDict = 0; withDict < 2; ++withDict) {
CHECK_Z(ZSTD_decompressBegin_usingDDict(dctx, withDict ? ddict : NULL));
CHECK_EQ(ZSTD_getErrorCode(ZSTD_decompressBlock(dctx, output,
sizeof(output), block, blockSize)), ZSTD_error_corruption_detected);
}
/* The originally reported two-block frame must stop at the malformed
* block instead of reaching its repeat block with invalid state. */
MEM_writeLE32(frame, ZSTD_MAGICNUMBER);
frame[4] = 0;
frame[5] = 0;
MEM_writeLE24(frame + frameSize, (U32)(blockSize << 3) | (bt_compressed << 1));
frameSize += 3;
memcpy(frame + frameSize, block, blockSize);
frameSize += blockSize;
MEM_writeLE24(frame + frameSize, (U32)(repeatSize << 3) | (bt_compressed << 1) | 1);
frameSize += 3;
memcpy(frame + frameSize, repeatBlock, repeatSize);
frameSize += repeatSize;
CHECK_EQ(ZSTD_getErrorCode(ZSTD_decompressDCtx(dctx, output,
sizeof(output), frame, frameSize)), ZSTD_error_corruption_detected);
}
/* Valid one- and four-stream tables must still install reusable state.
* Also reuse the context after the preceding errors and test a DDict's
* caller-owned table without installing an in-frame Huffman table. */
for (fourStreams = 0; fourStreams < 2; ++fourStreams) {
size_t const blockSize = FUZ_writeHuffmanLiteralBlock(block, sizeof(block),
literals, sizeof(literals), ctable, tableHeader, tableHeaderSize, fourStreams);
repeatSize = FUZ_writeHuffmanLiteralBlock(repeatBlock, sizeof(repeatBlock),
literals, sizeof(literals), ctable, NULL, 0, fourStreams);
CHECK_Z(ZSTD_decompressBegin(dctx));
CHECK_EQ(ZSTD_decompressBlock(dctx, output, sizeof(output), block, blockSize), sizeof(literals));
CHECK_EQ(memcmp(output, literals, sizeof(literals)), 0);
CHECK_EQ(ZSTD_decompressBlock(dctx, output, sizeof(output), repeatBlock, repeatSize), sizeof(literals));
CHECK_EQ(memcmp(output, literals, sizeof(literals)), 0);
CHECK_Z(ZSTD_decompressBegin_usingDDict(dctx, ddict));
CHECK_EQ(ZSTD_decompressBlock(dctx, output, sizeof(output), repeatBlock, repeatSize), sizeof(literals));
CHECK_EQ(memcmp(output, literals, sizeof(literals)), 0);
}
/* The generic HUF API still promises its raw/RLE shortcuts. */
CHECK_EQ(HUF_decompress1X_DCtx_wksp(dtable, output, sizeof(literals),
"A", 1, workspace, sizeof(workspace), 0), sizeof(literals));
for (i = 0; i < sizeof(literals); ++i) CHECK_EQ(output[i], 'A');
CHECK_EQ(HUF_decompress1X_DCtx_wksp(dtable, output, sizeof(literals),
literals, sizeof(literals), workspace, sizeof(workspace), 0), sizeof(literals));
CHECK_EQ(memcmp(output, literals, sizeof(literals)), 0);
ZSTD_freeDDict(ddict);
ZSTD_freeDCtx(dctx);
DISPLAYLEVEL(3, "OK \n");
}
/* Size of the literals of the block built by FUZ_writeSplitLitBlock().
* It is larger than the decoder's 64 KB extra literal buffer, so the first
* kSplitLitInDst literals are stored at the end of the block's area in dst. */
#define kSplitLitSize (64 KB + 112)
#define kSplitLitInDst (kSplitLitSize - 64 KB)
#define kSplitLitBlockMaxSize (kSplitLitSize + 3 + 9)
/* Writes a compressed block (without block header) with raw literals and 2 sequences,
* which decodes to 48 literals, a match of 65539+seq1MLExtra bytes at offset 1,
* 49 literals, a match of 3 bytes at offset 1, and the remaining literals.
* The block decodes to more than ZSTD_BLOCKSIZE_MAX bytes, so the first match runs
* up to (or past) the literals that are stored in dst.
* Writes the expected output to expected, which must hold
* FUZ_splitLitBlockDecodedSize(seq1MLExtra) bytes.
* @return : size of the block */
static size_t FUZ_writeSplitLitBlock(BYTE* block, BYTE* expected, U32 seq1MLExtra)
{
U32 const seq1ML = 65539 + seq1MLExtra;
BYTE* op = block;
size_t i;
/* Literals section header: raw literals, 3 byte header */
MEM_writeLE24(op, (U32)(kSplitLitSize << 4) | (3 << 2));
op += 3;
for (i = 0; i < kSplitLitSize; i++) {
*op++ = i < kSplitLitInDst ? (BYTE)(i + 1) : 0xCC;
}
/* Sequences section header: 2 sequences, LL & OF RLE, ML predefined.
* LL code 24 (base 48, 4 extra bits), OF code 0 (repcode 1 = offset 1). */
*op++ = 2;
*op++ = 0x50;
*op++ = 24;
*op++ = 0;
/* Bitstream, read from the top down: ML initial state 57 (ML code 52: base 65539,
* 16 extra bits), sequence 1 ML extra bits, LL extra bits 0 (LL 48),
* ML state update to state 0 (ML 3), sequence 2 LL extra bits 1 (LL 49). */
{ U64 const bitStream = (1ULL << 36) | (57ULL << 30) | ((U64)seq1MLExtra << 14) | 1;
MEM_writeLE32(op, (U32)bitStream);
op[4] = (BYTE)(bitStream >> 32);
op += 5;
}
{ BYTE* eop = expected;
memcpy(eop, block + 3, 48);
eop += 48;
memset(eop, 48, seq1ML);
eop += seq1ML;
memcpy(eop, block + 3 + 48, 49);
eop += 49;
memset(eop, 97, 3);
eop += 3;
memcpy(eop, block + 3 + 97, kSplitLitSize - 97);
}
assert((size_t)(op - block) <= kSplitLitBlockMaxSize);
return (size_t)(op - block);
}
static size_t FUZ_splitLitBlockDecodedSize(U32 seq1MLExtra)
{
return kSplitLitSize + 65539 + seq1MLExtra + 3;
}
static void test_decompressSplitLitBuffer(unsigned tnb)
{
/* The literals are split, rather than stored after the block's output, because
* dstCapacity <= ZSTD_BLOCKSIZE_MAX + litSize + 2 * WILDCOPY_OVERLENGTH. */
size_t const dstCapacity = ZSTD_BLOCKSIZE_MAX + kSplitLitSize + 64;
size_t const srcCapacity = ZSTD_FRAMEHEADERSIZE_MAX + ZSTD_blockHeaderSize + kSplitLitBlockMaxSize;
/* seq1MLExtra = 65371 : The first match ends 18 bytes before the next literals,
* so the literals are copied to an overlapping position.
* seq1MLExtra = 65460 : The first match overwrites the next literals. */
U32 const seq1MLExtras[] = { 65371, 65460 };
BYTE* const src = (BYTE*)malloc(srcCapacity);
BYTE* const dst = (BYTE*)malloc(dstCapacity);
BYTE* const expected = (BYTE*)malloc(dstCapacity);
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
size_t i;
int anySucceeded = 0;
CHECK(src != NULL && dst != NULL && expected != NULL && dctx != NULL);
DISPLAYLEVEL(3, "test%3u : decompress block with split literals overrun by output : ", tnb);
for (i = 0; i < COUNTOF(seq1MLExtras); i++) {
size_t const decodedSize = FUZ_splitLitBlockDecodedSize(seq1MLExtras[i]);
size_t blockSize;
size_t r;
CHECK(decodedSize <= dstCapacity);
/* The block API has no ZSTD_BLOCKSIZE_MAX limit on the decoded size,
* so the decoder must either reject the block or decode it correctly. */
blockSize = FUZ_writeSplitLitBlock(src, expected, seq1MLExtras[i]);
CHECK_Z( ZSTD_decompressBegin(dctx) );
r = ZSTD_decompressBlock(dctx, dst, dstCapacity, src, blockSize);
if (!ZSTD_isError(r)) {
CHECK_EQ(r, decodedSize);
CHECK(!memcmp(dst, expected, decodedSize));
anySucceeded = 1;
}
/* Wrap the block in a frame with windowLog 17, so blockSizeMax = ZSTD_BLOCKSIZE_MAX.
* The frame must be rejected, because the block decodes to more than blockSizeMax. */
MEM_writeLE32(src, ZSTD_MAGICNUMBER);
src[4] = 0x00; /* Frame header descriptor: no content size, no checksum, no dictID */
src[5] = 0x38; /* Window descriptor: windowLog 17 */
blockSize = FUZ_writeSplitLitBlock(src + 9, expected, seq1MLExtras[i]);
MEM_writeLE24(src + 6, 1 /* last block */ | ((U32)bt_compressed << 1) | (U32)(blockSize << 3));
r = ZSTD_decompress(dst, dstCapacity, src, 9 + blockSize);
CHECK(ZSTD_isError(r));
}
CHECK(anySucceeded); /* Must not reject all blocks */
DISPLAYLEVEL(3, "OK \n");
ZSTD_freeDCtx(dctx);
free(expected);
free(dst);
free(src);
}
static void test_decompressContinueWithoutBegin(unsigned tnb)
{
char src[1000];
char compressed[ZSTD_COMPRESSBOUND(sizeof(src))];
char decoded[sizeof(src)];
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
size_t cSize, cPos = 0, dPos = 0;
CHECK(dctx != NULL);
DISPLAYLEVEL(3, "test%3u : ZSTD_decompressContinue() without ZSTD_decompressBegin() : ", tnb);
memset(src, 'a', sizeof(src));
CHECK_VAR(cSize, ZSTD_compress(compressed, sizeof(compressed), src, sizeof(src), 1));
/* new context */
CHECK_EQ(ZSTD_nextSrcSizeToDecompress(dctx), 0);
CHECK_EQ(ZSTD_getErrorCode(ZSTD_decompressContinue(dctx, decoded, sizeof(decoded), NULL, 0)), ZSTD_error_stage_wrong);
/* finished frame */
CHECK_Z(ZSTD_decompressBegin(dctx));
while (ZSTD_nextSrcSizeToDecompress(dctx) != 0) {
size_t const inSize = ZSTD_nextSrcSizeToDecompress(dctx);
CHECK_NEWV(outSize, ZSTD_decompressContinue(dctx, decoded + dPos, sizeof(decoded) - dPos, compressed + cPos, inSize));
cPos += inSize;
dPos += outSize;
}
CHECK_EQ(cPos, cSize);
CHECK_EQ(dPos, sizeof(src));
CHECK_EQ(ZSTD_getErrorCode(ZSTD_decompressContinue(dctx, decoded, sizeof(decoded), NULL, 0)), ZSTD_error_stage_wrong);
DISPLAYLEVEL(3, "OK \n");
ZSTD_freeDCtx(dctx);
}
typedef struct {
const void* prefixStart;
const void* virtualStart;
const void* dictEnd;
const void* previousDstEnd;
} FUZ_DCtxHistory;
static FUZ_DCtxHistory FUZ_getDCtxHistory(const ZSTD_DCtx* dctx)
{
FUZ_DCtxHistory history;
history.prefixStart = dctx->prefixStart;
history.virtualStart = dctx->virtualStart;
history.dictEnd = dctx->dictEnd;
history.previousDstEnd = dctx->previousDstEnd;
return history;
}
static void FUZ_checkDCtxHistory(const ZSTD_DCtx* dctx, FUZ_DCtxHistory history)
{
CHECK_EQ(dctx->prefixStart, history.prefixStart);
CHECK_EQ(dctx->virtualStart, history.virtualStart);
CHECK_EQ(dctx->dictEnd, history.dictEnd);
CHECK_EQ(dctx->previousDstEnd, history.previousDstEnd);
}
typedef enum {
/* Keep the three dictionary-configuration cases first. */
historyNone,
historyDictionary,
historyDDict,
historyInsertedPrefix,
historyDecodedPrefix,
historyCount
} FUZ_HistoryKind;
static void FUZ_beginHistory(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict,
FUZ_HistoryKind historyKind,
const void* dictionary, size_t dictionarySize)
{
if (historyKind == historyDDict) {
CHECK_Z(ZSTD_decompressBegin_usingDDict(dctx, ddict));
} else if (historyKind == historyDictionary) {
CHECK_Z(ZSTD_decompressBegin_usingDict(dctx, dictionary, dictionarySize));
} else {
CHECK_Z(ZSTD_decompressBegin(dctx));
}
}
static void test_decompressZeroByteHistory(unsigned tnb)
{
enum {
emptyInsertion,
emptyDecompression,
emptyOperationCount
} emptyOperation;
/* Magic, no checksum/content size/dictionary ID, and a 1 KiB window. */
static const BYTE frameHeader[] = { 0x28, 0xB5, 0x2F, 0xFD, 0, 0 };
static const struct {
BYTE bytes[5];
size_t size;
} emptyBlocks[] = {
{ { 0, 0, 0 }, 3 }, /* empty raw block */
{ { 2, 0, 0, 'A' }, 4 }, /* empty RLE block */
{ { 0x14, 0, 0, 0, 0 }, 5 } /* compressed block: no literals or sequences */
};
static const BYTE matchHeader[] = { 0x3D, 0, 0 };
/* No literals, one four-byte match at repeat offset 4. */
static const BYTE matchBlock[] = { 0, 1, 0x54, 0, 0, 1, 1 };
static const BYTE emptyCompressedBlock[] = { 0, 0 };
static const BYTE literalsBlock[] = { 0x20, 'w', 'x', 'y', 'z', 0 };
static const char dictionary[] = "0123456789abcdef";
static const char insertedPrefix[] = "wxyz";
BYTE output1[64];
BYTE output2[64];
BYTE decodedPrefix[64];
void* const emptyDsts[] = { NULL, output1, output2 };
void* const outputDsts[] = { output1, output2 };
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
ZSTD_DDict* const ddict = ZSTD_createDDict(dictionary, sizeof(dictionary) - 1);
size_t blockType, emptyDstNb, outputDstNb;
FUZ_HistoryKind historyKind;
CHECK(dctx != NULL && ddict != NULL);
DISPLAYLEVEL(3, "test%3u : zero-byte decompression preserves history : ", tnb);
/* Bufferless frame decoding. Headers and empty raw/RLE/compressed blocks
* may be supplied with any zero-capacity destination, including NULL.
* They must neither invent history nor discard a real dictionary. */
for (historyKind = historyNone; historyKind < historyInsertedPrefix; ++historyKind) {
for (blockType = 0; blockType < COUNTOF(emptyBlocks); ++blockType) {
BYTE prefix[sizeof(frameHeader) + sizeof(emptyBlocks[0].bytes)];
size_t const prefixSize = sizeof(frameHeader) + emptyBlocks[blockType].size;
memcpy(prefix, frameHeader, sizeof(frameHeader));
memcpy(prefix + sizeof(frameHeader), emptyBlocks[blockType].bytes, emptyBlocks[blockType].size);
for (emptyDstNb = 0; emptyDstNb < COUNTOF(emptyDsts); ++emptyDstNb) {
for (outputDstNb = 0; outputDstNb < COUNTOF(outputDsts); ++outputDstNb) {
FUZ_DCtxHistory history;
size_t pos = 0;
size_t result;
FUZ_beginHistory(dctx, ddict, historyKind, dictionary, sizeof(dictionary) - 1);
history = FUZ_getDCtxHistory(dctx);
while (pos < prefixSize) {
size_t const next = ZSTD_nextSrcSizeToDecompress(dctx);
CHECK(next != 0 && next <= prefixSize - pos);
CHECK_EQ(ZSTD_decompressContinue(dctx, emptyDsts[emptyDstNb], 0,
prefix + pos, next), 0);
pos += next;
}
FUZ_checkDCtxHistory(dctx, history);
CHECK_EQ(ZSTD_nextSrcSizeToDecompress(dctx), sizeof(matchHeader));
CHECK_EQ(ZSTD_decompressContinue(dctx, outputDsts[outputDstNb], sizeof(output1),
matchHeader, sizeof(matchHeader)), 0);
result = ZSTD_decompressContinue(dctx, outputDsts[outputDstNb], sizeof(output1),
matchBlock, sizeof(matchBlock));
if (historyKind != historyNone) {
CHECK_EQ(result, 4);
CHECK(!memcmp(outputDsts[outputDstNb], "cdef", 4));
} else {
CHECK_EQ(ZSTD_getErrorCode(result), ZSTD_error_corruption_detected);
}
}
}
}
}
/* The block APIs have the same invariant. Exercise both empty insertion
* and empty compressed blocks with no history, raw and prepared dictionaries,
* an inserted raw prefix, and a previously decoded prefix. */
for (historyKind = historyNone; historyKind < historyCount; ++historyKind) {
for (emptyOperation = emptyInsertion; emptyOperation < emptyOperationCount; ++emptyOperation) {
for (emptyDstNb = 0; emptyDstNb < COUNTOF(emptyDsts); ++emptyDstNb) {
for (outputDstNb = 0; outputDstNb < COUNTOF(outputDsts); ++outputDstNb) {
FUZ_DCtxHistory history;
size_t result;
FUZ_beginHistory(dctx, ddict, historyKind, dictionary, sizeof(dictionary) - 1);
if (historyKind == historyInsertedPrefix)
CHECK_EQ(ZSTD_insertBlock(dctx, insertedPrefix, sizeof(insertedPrefix) - 1), 4);
if (historyKind == historyDecodedPrefix)
CHECK_EQ(ZSTD_decompressBlock(dctx, decodedPrefix, sizeof(decodedPrefix),
literalsBlock, sizeof(literalsBlock)), 4);
history = FUZ_getDCtxHistory(dctx);
if (emptyOperation == emptyInsertion) {
CHECK_EQ(ZSTD_insertBlock(dctx, emptyDsts[emptyDstNb], 0), 0);
} else {
CHECK_EQ(ZSTD_decompressBlock(dctx, emptyDsts[emptyDstNb], 0,
emptyCompressedBlock, sizeof(emptyCompressedBlock)), 0);
}
FUZ_checkDCtxHistory(dctx, history);
result = ZSTD_decompressBlock(dctx, outputDsts[outputDstNb], sizeof(output1),
matchBlock, sizeof(matchBlock));
if (historyKind != historyNone) {
CHECK_EQ(result, 4);
CHECK(!memcmp(outputDsts[outputDstNb],
historyKind == historyDictionary || historyKind == historyDDict ? "cdef" : "wxyz", 4));
} else {
CHECK_EQ(ZSTD_getErrorCode(result), ZSTD_error_corruption_detected);
}
}
}
}
}
ZSTD_freeDCtx(dctx);
ZSTD_freeDDict(ddict);
DISPLAYLEVEL(3, "OK \n");
}
/* Decodes src with ZSTD_decompressStream() using a single input buffer, whose
* size is first limited to `split` bytes and then extended to srcSize, so the
* second call continues at in.pos == split. Returns the last streaming result.
* Output is written to dst (of capacity dstCap), and its size to *dstSize. */
static size_t FUZ_decompressStreamSplit(ZSTD_DCtx* dctx, const void* src, size_t srcSize,
size_t split, void* dst, size_t dstCap, size_t* dstSize)
{
ZSTD_inBuffer in = { src, split, 0 };
ZSTD_outBuffer out = { dst, dstCap, 0 };
size_t result;
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters));
result = ZSTD_decompressStream(dctx, &out, &in);
if (!ZSTD_isError(result) && split < srcSize) {
in.size = srcSize;
result = ZSTD_decompressStream(dctx, &out, &in);
}
*dstSize = out.pos;
return result;
}
/* Checks that streaming decompression does not depend on where the input is
* split between two calls (in particular, in the middle of the frame header).
* Decodes with every split point and compares against decoding with no split.
* If expected == NULL, the stream must be rejected, else it must decode to expected. */
static void FUZ_checkStreamSplits(const void* src, size_t srcSize,
const void* expected, size_t expectedSize)
{
/* Large enough for the content size claimed by embeddedFrame below, since
* the single-pass fast path is only taken if the output has room for it. */
size_t const outCapacity = 128 KB;
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
BYTE* const refOut = (BYTE*)malloc(outCapacity);
BYTE* const out = (BYTE*)malloc(outCapacity);
size_t refSize, outSize, split, ref;
CHECK(dctx != NULL && refOut != NULL && out != NULL);
ref = FUZ_decompressStreamSplit(dctx, src, srcSize, srcSize, refOut, outCapacity, &refSize);
if (expected == NULL) {
CHECK(ZSTD_isError(ref));
} else {
CHECK_EQ(ref, 0);
CHECK_EQ(refSize, expectedSize);
CHECK(memcmp(refOut, expected, expectedSize) == 0);
}
for (split = 0; split < srcSize; ++split) {
size_t const result = FUZ_decompressStreamSplit(dctx, src, srcSize, split, out, outCapacity, &outSize);
CHECK_EQ(ZSTD_getErrorCode(result), ZSTD_getErrorCode(ref));
if (!ZSTD_isError(result)) CHECK_EQ(result, ref);
CHECK_EQ(outSize, refSize);
CHECK(memcmp(out, refOut, refSize) == 0);
}
free(out);
free(refOut);
ZSTD_freeDCtx(dctx);
}
/* Guards against ZSTD_decompressStream() mishandling a frame header that is
* split across calls. The streaming decoder has fast paths (decode the whole
* frame in one pass, detect legacy frames) that inspect the current input
* buffer from its start. They are only valid when the frame header starts in
* this call. After a split header, they would instead look at the middle of
* the frame, so bytes of a malformed frame that look like the start of another
* frame could be decoded as if they were a frame.
*
* Each case is decoded at every split point, and must behave exactly like
* decoding without a split. The malformed frames are crafted so that a wrong
* fast path would accept them, which makes the comparison fail. A valid frame
* checks that normal frames are unaffected. */
static void test_decompressStreamSplitHeader(unsigned tnb)
{
/* Malformed: the header says the frame uses a dictionary (dictID field)
* that is not provided. The 12 bytes after the first 5 bytes (magic + FHD
* 0xE3, which selects 4-byte dictID and 8-byte content size) also encode
* a complete valid empty frame: magic, FHD 0x20, content size 0, and an
* empty last block. Decoding that from a split at byte 5 is wrong. */
static const BYTE embeddedFrame[] = {
0x28, 0xB5, 0x2F, 0xFD, 0xE3,
0x28, 0xB5, 0x2F, 0xFD, 0x20, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00
};
/* Malformed: FHD 0x08 sets a reserved bit, which is reported once the
* 6-byte header is available. The bytes after the first 5 start with the
* magic number of a legacy (v0.7) frame. Only matters when the library
* is built with legacy support, and must not be detected as a legacy
* frame since the frame did not start there. */
static const BYTE embeddedLegacy[] = {
0x28, 0xB5, 0x2F, 0xFD, 0x08,
0x27, 0xB5, 0x2F, 0xFD, 0x00, 0x00, 0x00, 0x00
};
BYTE source[257];
BYTE compressed[ZSTD_COMPRESSBOUND(sizeof(source))];
size_t compressedSize, i;
DISPLAYLEVEL(3, "test%3u : streaming frame header split across calls : ", tnb);
FUZ_checkStreamSplits(embeddedFrame, sizeof(embeddedFrame), NULL, 0);
FUZ_checkStreamSplits(embeddedLegacy, sizeof(embeddedLegacy), NULL, 0);
for (i = 0; i < sizeof(source); ++i) source[i] = (BYTE)(i * 23);
CHECK_VAR(compressedSize, ZSTD_compress(compressed, sizeof(compressed), source, sizeof(source), 1));
FUZ_checkStreamSplits(compressed, compressedSize, source, sizeof(source));
DISPLAYLEVEL(3, "OK \n");
}
static unsigned test_histCountWksp(unsigned seed, unsigned testNb)
{
static const unsigned symLowLimits[] = { 0, 27, 0, 0, 27, 42, 0, 0, 27, 42, 27, 42 };
static const unsigned symHighLimits[] = { 255, 255, 210, 110, 42, 42, 210, 110, 42, 42, 42, 42 };
static const unsigned symMaxLimits[] = { 255, 255, 255, 255, 255, 255, 230, 130, 99, 99, 42, 42 };
static const size_t inputSizes[] = { 3367, 1761, 893, 117 };
unsigned workspace[HIST_WKSP_SIZE_U32];
size_t res, i, is, il;
DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with empty source : ", testNb++);
{
/* With NULL source UBSan of older Clang could fail: applying zero offset to null pointer. */
static const unsigned char source[4] = { 0 };
unsigned count[1] = { 0 };
unsigned maxSym = 0;
res = HIST_count_wksp(count, &maxSym, source, 0, workspace, sizeof(workspace));
CHECK_EQ(res, 0);
CHECK_EQ(maxSym, 0);
CHECK_EQ(count[0], 0);
}
DISPLAYLEVEL(3, "OK \n");
#if HIST_WKSP_SIZE_U32
DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with small workspace : ", testNb++);
{
unsigned count[1] = { 0 };
unsigned maxSym = 0;
res = HIST_count_wksp(count, &maxSym, NULL, 0, workspace, sizeof(workspace) - 1);
CHECK_EQ(res, ERROR(workSpace_tooSmall));
CHECK_EQ(maxSym, 0);
CHECK_EQ(count[0], 0);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with wrong workspace alignment : ", testNb++);
{
unsigned count[1] = { 0 };
unsigned maxSym = 0;
res = HIST_count_wksp(count, &maxSym, NULL, 0, (unsigned*)(void*)((char*)workspace + 1), sizeof(workspace));
CHECK_EQ(res, ERROR(GENERIC));
CHECK_EQ(maxSym, 0);
CHECK_EQ(count[0], 0);
}
DISPLAYLEVEL(3, "OK \n");
#endif
DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with symbol out of range, small size : ", testNb++);
{
/* For less elements HIST_count_parallel_wksp would fail. */
static const unsigned char source[4] = { 1, 4, 0, 2 };
static const unsigned expected[6] = { 0 };
unsigned count[6] = { 0 };
unsigned maxSym = 2;
res = HIST_count_wksp(count, &maxSym, source, sizeof(source), workspace, sizeof(workspace));
CHECK_EQ(res, ERROR(maxSymbolValue_tooSmall));
CHECK_EQ(maxSym, 2);
for (i = 0; i < COUNTOF(expected); ++i) CHECK_EQ(count[i], expected[i]);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3u : HIST_count_wksp with symbol out of range, medium size : ", testNb++);
{
unsigned char source[3407];
unsigned count[6] = { 0 };
unsigned maxSym = 2;
for (i = 0; i < COUNTOF(source); ++i) {
source[i] = (48271 * (i + 1)) & 3;
}
res = HIST_count_wksp(count, &maxSym, source, sizeof(source), workspace, sizeof(workspace));
CHECK_EQ(res, ERROR(maxSymbolValue_tooSmall));
CHECK_EQ(maxSym, 2);
for (i = 0; i < COUNTOF(count); ++i) CHECK_EQ(count[i], 0);
}
DISPLAYLEVEL(3, "OK \n");
for (il = 0; il < COUNTOF(symMaxLimits); ++il) {
unsigned symMax = symMaxLimits[il];
unsigned symLow = symLowLimits[il];
unsigned symHigh = symHighLimits[il];
unsigned symRange = symHigh - symLow + 1;
for (is = 0; is < COUNTOF(inputSizes); ++is) {
unsigned char source[4000];
size_t inputSize = inputSizes[is];
assert(inputSize <= sizeof(source));
DISPLAYLEVEL(3, "test%3u : HIST_count_wksp test in [%u..%u], symMax: %u, inputSize: %u : ",
testNb++, symLow, symHigh, symMax, (unsigned)inputSize);
{
unsigned count[260] = { 0 };
unsigned expected[COUNTOF(count)] = { 0 };
unsigned maxSym = symMax;
unsigned realMaxSym = symMax;
unsigned maxCount = 0;
for (i = 0; i < inputSize; ++i) {
unsigned prng = (48271 * (i + seed)) % symRange + symLow;
source[i] = (unsigned char)prng;
++expected[prng];
}
/* for basic buffer overwrite checks */
for (i = maxSym + 1; i < COUNTOF(count); ++i) expected[i] = count[i] = ~0u;
for (i = 0; i <= maxSym; ++i) maxCount = MAX(maxCount, expected[i]);
for (i = realMaxSym; i > 0; --i) {
if (expected[i]) break;
--realMaxSym;
}
res = HIST_count_wksp(count, &maxSym, source, inputSize, workspace, sizeof(workspace));
CHECK_EQ(res, maxCount);
CHECK_EQ(maxSym, realMaxSym);
for (i = 0; i < COUNTOF(expected); ++i) CHECK_EQ(count[i], expected[i]);
}
DISPLAYLEVEL(3, "OK \n");
}
}
return testNb;
}
static void test_histCountOverflowCase(const BYTE* source, size_t sourceSize,
unsigned maxSymbolValue)
{
unsigned workspace[HIST_WKSP_SIZE_U32];
unsigned expected[260] = { 0 };
unsigned expectedMaxSymbol = 0;
unsigned expectedLargest = 0;
unsigned fast;
size_t index;
for (index = 0; index < sourceSize; ++index) {
unsigned const symbol = source[index];
CHECK(symbol <= maxSymbolValue);
++expected[symbol];
expectedMaxSymbol = MAX(expectedMaxSymbol, symbol);
expectedLargest = MAX(expectedLargest, expected[symbol]);
}
/* Entries past maxSymbolValue must not be written. */
for (index = maxSymbolValue + 1; index < COUNTOF(expected); ++index)
expected[index] = ~0U;
for (fast = 0; fast < 2; ++fast) {
unsigned count[COUNTOF(expected)];
unsigned maxSymbol = maxSymbolValue;
size_t result;
memset(count, 0xFF, sizeof(count));
result = fast
? HIST_countFast_wksp(count, &maxSymbol, source, sourceSize, workspace, sizeof(workspace))
: HIST_count_wksp(count, &maxSymbol, source, sourceSize, workspace, sizeof(workspace));
CHECK_EQ(result, expectedLargest);
CHECK_EQ(maxSymbol, expectedMaxSymbol);
for (index = 0; index < COUNTOF(expected); ++index)
CHECK_EQ(count[index], expected[index]);
}
}
static void test_histCountOverflow(unsigned testNb)
{
/* The SVE2 histogram counts in chunks of 65520 bytes. Cover the sizes
* around 1 and 2 chunks, 16-bit wraparound, and a 4-chunk input. */
static const size_t sizes[] = {
65519, 65520, 65521, 65536, 131040, 131041, 131072, 196625
};
static const unsigned symbols[] = { 0, 63, 64, 159, 160, 255 };
size_t const capacity = sizes[COUNTOF(sizes) - 1];
BYTE* const source = (BYTE*)malloc(capacity);
unsigned workspace[HIST_WKSP_SIZE_U32];
unsigned symbol;
size_t sizeIndex, symbolIndex, index;
CHECK(source != NULL);
DISPLAYLEVEL(3, "test%3u : histogram counter overflow : ", testNb);
/* Every symbol, so every counter lane overflows. */
for (symbol = 0; symbol <= 255; ++symbol) {
memset(source, (int)symbol, 131072);
test_histCountOverflowCase(source, 65536, symbol);
test_histCountOverflowCase(source, 131072, 255);
}
/* Every size, for symbols in each pass, plus a differing tail. */
for (symbolIndex = 0; symbolIndex < COUNTOF(symbols); ++symbolIndex) {
symbol = symbols[symbolIndex];
memset(source, (int)symbol, capacity);
for (sizeIndex = 0; sizeIndex < COUNTOF(sizes); ++sizeIndex) {
test_histCountOverflowCase(source, sizes[sizeIndex], symbol);
test_histCountOverflowCase(source, sizes[sizeIndex], 255);
}
for (index = 65536; index < capacity; ++index)
source[index] = (BYTE)(symbol + 1 + (index - 65536) % 255);
test_histCountOverflowCase(source, capacity, 255);
}
/* Per-chunk max symbols that rise and then fall. */
memset(source, 1, 65520);
memset(source + 65520, 255, 65520);
memset(source + 131040, 0, capacity - 131040);
test_histCountOverflowCase(source, capacity, 255);
/* A symbol out of range in the last chunk is an error. */
{ unsigned count[260];
unsigned maxSymbol = 159;
memset(source, 1, capacity);
source[capacity - 1] = 160;
memset(count, 0xFF, sizeof(count));
CHECK_EQ(HIST_count_wksp(count, &maxSymbol, source, capacity, workspace, sizeof(workspace)),
ERROR(maxSymbolValue_tooSmall));
CHECK_EQ(maxSymbol, 159);
for (index = 160; index < COUNTOF(count); ++index)
CHECK_EQ(count[index], ~0U);
}
free(source);
DISPLAYLEVEL(3, "OK \n");
}
static void test_setCParams(unsigned tnb)
{
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_compressionParameters cparams;
assert(cctx);
DISPLAYLEVEL(3, "test%3u : ZSTD_CCtx_setCParams : ", tnb);
/* valid cparams */
cparams = ZSTD_getCParams(1, 0, 0);
CHECK_Z(ZSTD_CCtx_setCParams(cctx, cparams));
/* invalid cparams (must fail) */
cparams.windowLog = 99;
CHECK(ZSTD_isError(ZSTD_CCtx_setCParams(cctx, cparams)));
free(cctx);
DISPLAYLEVEL(3, "OK \n");
}
static void test_blockSplitter_incompressibleExpansionProtection(unsigned testNb, unsigned seed)
{
DISPLAYLEVEL(3, "test%3i : Check block splitter doesn't oversplit incompressible data (seed %u): ", testNb, seed);
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
size_t const srcSize = 256 * 1024; /* needs to be at least 2 blocks */
void* incompressible = malloc(srcSize);
size_t const dstCapacity = ZSTD_compressBound(srcSize);
void* cBuffer = malloc(dstCapacity);
size_t const chunkSize = 8 KB;
size_t const nbChunks = srcSize / chunkSize;
size_t chunkNb, cSizeNoSplit, cSizeWithSplit;
assert(cctx != NULL);
assert(incompressible != NULL);
assert(cBuffer != NULL);
/* let's fill input with random noise (incompressible) */
RDG_genBuffer(incompressible, srcSize, 0.0, 0.0, seed);
/* this pattern targets the fastest _byChunk variant's sampling (level 3).
* manually checked that, without the @savings protection, it would over-split.
*/
for (chunkNb=0; chunkNb<nbChunks; chunkNb++) {
BYTE* const p = (BYTE*)incompressible + chunkNb * chunkSize;
size_t const samplingRate = 43;
int addOrRemove = chunkNb % 2;
size_t n;
for (n=0; n<chunkSize; n+=samplingRate) {
if (addOrRemove) {
p[n] &= 0x80;
} else {
p[n] |= 0x80;
}
}
}
/* run first without splitting */
ZSTD_CCtx_setParameter(cctx, ZSTD_c_blockSplitterLevel, 1 /* no split */);
cSizeNoSplit = ZSTD_compress2(cctx, cBuffer, dstCapacity, incompressible, srcSize);
/* run with sample43 splitter, check it's still the same */
ZSTD_CCtx_setParameter(cctx, ZSTD_c_blockSplitterLevel, 3 /* sample43, fastest _byChunk variant */);
cSizeWithSplit = ZSTD_compress2(cctx, cBuffer, dstCapacity, incompressible, srcSize);
if (cSizeWithSplit != cSizeNoSplit) {
DISPLAYLEVEL(1, "invalid compressed size: cSizeWithSplit %u != %u cSizeNoSplit \n",
(unsigned)cSizeWithSplit, (unsigned)cSizeNoSplit);
abort();
}
DISPLAYLEVEL(4, "compressed size: cSizeWithSplit %u == %u cSizeNoSplit : ",
(unsigned)cSizeWithSplit, (unsigned)cSizeNoSplit);
free(incompressible);
free(cBuffer);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
}
static void test_targetCBlockSize_dstCapacity(unsigned testNb)
{
static const size_t srcSizes[] = { 1300, 8 KB };
size_t const srcCapacity = 8 KB;
size_t const dstAllocation = 2 * ZSTD_compressBound(srcCapacity);
BYTE* const src = (BYTE*)malloc(srcCapacity);
BYTE* const dst = (BYTE*)malloc(dstAllocation);
BYTE* const decoded = (BYTE*)malloc(srcCapacity);
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
size_t input;
DISPLAYLEVEL(3, "test%3u : targetCBlockSize with short destinations : ", testNb);
CHECK(src != NULL && dst != NULL && decoded != NULL);
CHECK(cctx != NULL && dctx != NULL);
/* Generate Huffman and FSE table descriptions across multiple sub-blocks. */
RDG_genBuffer(src, srcCapacity, 0.5, 0. /*auto*/, 0);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_targetCBlockSize, 1340));
for (input = 0; input < COUNTOF(srcSizes); ++input) {
size_t const srcSize = srcSizes[input];
size_t const bound = ZSTD_compressBound(srcSize);
size_t capacity;
for (capacity = 0; capacity <= bound; ++capacity) {
size_t cSize, i;
memset(dst, 0xA5, dstAllocation);
cSize = ZSTD_compress2(cctx, dst, capacity, src, srcSize);
/* Check before handling errors: the original bug returned
* dstSize_tooSmall after already overwriting the destination. */
for (i = capacity; i < dstAllocation; ++i) {
if (dst[i] != 0xA5) {
DISPLAY("targetCBlockSize wrote past capacity %zu "
"at offset %zu (srcSize=%zu)\n", capacity, i, srcSize);
exit(1);
}
}
if (ZSTD_isError(cSize)) {
CHECK_EQ(ZSTD_getErrorCode(cSize), ZSTD_error_dstSize_tooSmall);
CHECK_LT(capacity, bound);
} else {
CHECK(cSize <= capacity);
CHECK_EQ(ZSTD_decompressDCtx(dctx, decoded, srcCapacity, dst, cSize), srcSize);
CHECK_EQ(memcmp(decoded, src, srcSize), 0);
}
}
}
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtx(cctx);
free(decoded);
free(dst);
free(src);
DISPLAYLEVEL(3, "OK \n");
}
size_t convertSequences_noRepcodes(SeqDef* dstSeqs, const ZSTD_Sequence* inSeqs,
size_t nbSequences);
static size_t convertSequences_noRepcodes_ref(
SeqDef* dstSeqs,
const ZSTD_Sequence* inSeqs,
size_t nbSequences)
{
size_t longLen = 0;
size_t n;
for (n=0; n<nbSequences; n++) {
dstSeqs[n].offBase = OFFSET_TO_OFFBASE(inSeqs[n].offset);
dstSeqs[n].litLength = (U16)inSeqs[n].litLength;
dstSeqs[n].mlBase = (U16)(inSeqs[n].matchLength - MINMATCH);
/* Check for long length > 65535. */
if (UNLIKELY(inSeqs[n].matchLength > 65535+MINMATCH)) {
assert(longLen == 0);
longLen = n + 1;
}
if (UNLIKELY(inSeqs[n].litLength > 65535)) {
assert(longLen == 0);
longLen = n + nbSequences + 1;
}
}
return longLen;
}
static unsigned test_convertSequences_noRepcodes(unsigned seed, unsigned testNb)
{
ZSTD_Sequence nsrc[12];
SeqDef ndst[12], rdst[12];
size_t ref, ret, i, j;
seed += 0xDEADBEEF;
for (i = 0; i < COUNTOF(nsrc); ++i) {
seed = 48271 * ((unsigned)i + seed);
nsrc[i].offset = (seed & 0xFFFF) | 1; /* Offset shall not be zero. */
seed = 48271 * ((unsigned)i + seed);
nsrc[i].litLength = seed & 0xFFFF;
seed = 48271 * ((unsigned)i + seed);
nsrc[i].matchLength = (seed & 0xFFFFFF) % (65536 + MINMATCH);
seed = 48271 * ((unsigned)i + seed);
nsrc[i].rep = seed & 0xFF;
}
/* For near overflow and proper negative value handling. */
nsrc[5].matchLength = 65535 + MINMATCH;
nsrc[6].litLength = 65535;
nsrc[6].matchLength = 0;
nsrc[7].litLength = 0;
nsrc[7].matchLength = MINMATCH;
for (i = 0; i <= COUNTOF(nsrc); ++i) {
DISPLAYLEVEL(3, "test%3u : convertSequences_noRepcodes with %u inputs : ",
testNb++, (unsigned)i);
memset(ndst, 0, sizeof(ndst));
memset(rdst, 0, sizeof(rdst));
ref = convertSequences_noRepcodes_ref(rdst, nsrc, i);
ret = convertSequences_noRepcodes(ndst, nsrc, i);
CHECK_EQ(ret, ref);
CHECK_EQ(memcmp(rdst, ndst, sizeof(ndst)), 0);
DISPLAYLEVEL(3, "OK \n");
}
nsrc[7].matchLength = 65536 + MINMATCH;
for (i = 8; i <= COUNTOF(nsrc); ++i) {
DISPLAYLEVEL(3, "test%3u : convertSequences_noRepcodes with %u inputs and "
"matchLength overflow : ",
testNb++, (unsigned)i);
memset(ndst, 0, sizeof(ndst));
memset(rdst, 0, sizeof(rdst));
ref = convertSequences_noRepcodes_ref(rdst, nsrc, i);
ret = convertSequences_noRepcodes(ndst, nsrc, i);
CHECK_EQ(ret, ref);
CHECK_EQ(memcmp(rdst, ndst, sizeof(ndst)), 0);
DISPLAYLEVEL(3, "OK \n");
assert(COUNTOF(nsrc) > 8);
for (j = 4; j < 8; ++j) {
DISPLAYLEVEL(3, "test%3u : convertSequences_noRepcodes with %u inputs and "
"matchLength overflow #%u : ",
testNb++, (unsigned)i, (unsigned)(i - j));
memset(ndst, 0, sizeof(ndst));
memset(rdst, 0, sizeof(rdst));
ref = convertSequences_noRepcodes_ref(rdst, nsrc + j, i - j);
ret = convertSequences_noRepcodes(ndst, nsrc + j, i - j);
CHECK_EQ(ret, ref);
CHECK_EQ(memcmp(rdst, ndst, sizeof(ndst)), 0);
DISPLAYLEVEL(3, "OK \n");
}
}
nsrc[7].matchLength = 1;
nsrc[7].litLength = 65536;
for (i = 8; i <= COUNTOF(nsrc); ++i) {
DISPLAYLEVEL(3, "test%3u : convertSequences_noRepcodes with %u inputs and "
"litLength overflow: ",
testNb++, (unsigned)i);
memset(ndst, 0, sizeof(ndst));
memset(rdst, 0, sizeof(rdst));
ref = convertSequences_noRepcodes_ref(rdst, nsrc, i);
ret = convertSequences_noRepcodes(ndst, nsrc, i);
CHECK_EQ(ret, ref);
CHECK_EQ(memcmp(rdst, ndst, sizeof(ndst)), 0);
DISPLAYLEVEL(3, "OK \n");
assert(COUNTOF(nsrc) > 8);
for (j = 4; j < 8; ++j) {
DISPLAYLEVEL(3, "test%3u : convertSequences_noRepcodes with %u inputs and "
"litLength overflow #%u: ",
testNb++, (unsigned)i, (unsigned)(i - j));
memset(ndst, 0, sizeof(ndst));
memset(rdst, 0, sizeof(rdst));
ref = convertSequences_noRepcodes_ref(rdst, nsrc + j, i - j);
ret = convertSequences_noRepcodes(ndst, nsrc + j, i - j);
CHECK_EQ(ret, ref);
CHECK_EQ(memcmp(rdst, ndst, sizeof(ndst)), 0);
DISPLAYLEVEL(3, "OK \n");
}
}
return testNb;
}
static U32 FUZ_blockSplitHash(U32 x)
{
x ^= x >> 16;
x *= 0x7feb352dU;
x ^= x >> 15;
x *= 0x846ca68bU;
return x ^ (x >> 16);
}
/* The input is all zeroes, so all these offsets are valid matches.
* Groups of 150 sequences have distinct offset-code distributions, making
* enough dyadic splits beneficial to exhaust the block splitter's budget. */
static size_t FUZ_blockSplitSequenceProducer(void* state,
ZSTD_Sequence* out, size_t capacity,
const void* src, size_t srcSize, const void* dict, size_t dictSize,
int compressionLevel, size_t windowSize)
{
size_t i;
(void)state; (void)src; (void)dict; (void)dictSize;
(void)compressionLevel; (void)windowSize;
CHECK_EQ(srcSize, 128000);
CHECK(capacity >= 38401);
for (i = 0; i < 38400; ++i) {
U32 const code = 3 + FUZ_blockSplitHash((U32)(i / 150)) % 10;
out[i].offset = (1U << code) + (U32)(i % 4);
out[i].litLength = i == 0 ? 12800 : 0;
out[i].matchLength = 3;
out[i].rep = 0;
}
memset(out + i, 0, sizeof(*out));
return i + 1;
}
static unsigned test_blockSplitLimit(unsigned testNb)
{
size_t const srcSize = 128000;
size_t const dstCapacity = ZSTD_compressBound(srcSize);
BYTE* const src = (BYTE*)calloc(1, srcSize);
BYTE* const dst = (BYTE*)malloc(dstCapacity);
BYTE* const decoded = (BYTE*)malloc(srcSize);
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_FrameHeader header;
size_t cSize, pos, nbBlocks = 0;
DISPLAYLEVEL(3, "test%3u : external sequences exhaust the block split budget : ", testNb++);
CHECK(src != NULL && dst != NULL && decoded != NULL && cctx != NULL);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 19));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_minMatch, 3));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_splitAfterSequences, ZSTD_ps_enable));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_validateSequences, 1));
ZSTD_registerSequenceProducer(cctx, NULL, FUZ_blockSplitSequenceProducer);
cSize = ZSTD_compress2(cctx, dst, dstCapacity, src, srcSize);
CHECK_Z(cSize);
CHECK_EQ(ZSTD_getFrameHeader(&header, dst, cSize), 0);
pos = header.headerSize;
for (;;) {
U32 blockHeader;
CHECK(pos + ZSTD_blockHeaderSize <= cSize);
blockHeader = MEM_readLE24(dst + pos);
pos += ZSTD_blockHeaderSize;
pos += ((blockHeader >> 1) & 3) == bt_rle ? 1 : blockHeader >> 3;
CHECK(pos <= cSize);
++nbBlocks;
if (blockHeader & 1) break;
}
CHECK_EQ(pos, cSize);
CHECK_EQ(nbBlocks, ZSTD_MAX_NB_BLOCK_SPLITS + 1);
CHECK_EQ(cctx->blockSplitCtx.partitions[ZSTD_MAX_NB_BLOCK_SPLITS], 38400);
CHECK_EQ(ZSTD_decompress(decoded, srcSize, dst, cSize), srcSize);
CHECK_EQ(memcmp(src, decoded, srcSize), 0);
ZSTD_freeCCtx(cctx);
free(decoded);
free(dst);
free(src);
DISPLAYLEVEL(3, "OK \n");
return testNb;
}
static unsigned test_get1BlockSummary(unsigned testNb)
{
static const ZSTD_Sequence nseqs[] = {
{ 10, 2, 4, 1 },
{ 20, 3, 5, 2 },
{ 30, 6, 8, 3 },
{ 40, 7, 9, 4 },
{ 50, 10, 12, 5 },
{ 60, 11, 13, 6 },
{ 0, 14, 0, 7 },
{ 70, 15, 17, 8 },
{ 80, 16, 18, 9 },
{ 90, 19, 21, 1 },
{ 99, 20, 22, 2 },
};
static const BlockSummary blocks[] = {
{ 7, 104, 53 },
{ 6, 98, 51 },
{ 5, 90, 48 },
{ 4, 76, 42 },
{ 3, 60, 35 },
{ 2, 38, 25 },
{ 1, 14, 14 },
};
size_t i;
DISPLAYLEVEL(3, "test%3u : ZSTD_get1BlockSummary with empty array : ", testNb++);
{
BlockSummary bs = ZSTD_get1BlockSummary(nseqs, 0);
CHECK_EQ(bs.nbSequences, ERROR(externalSequences_invalid));
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3u : ZSTD_get1BlockSummary with 1 literal only : ", testNb++);
{
static const ZSTD_Sequence seqs[] = { { 0, 5, 0, 0 } };
BlockSummary bs = ZSTD_get1BlockSummary(seqs, 1);
CHECK_EQ(bs.nbSequences, 1);
CHECK_EQ(bs.litSize, 5);
CHECK_EQ(bs.blockSize, 5);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3u : ZSTD_get1BlockSummary with no terminator : ", testNb++);
{
static const ZSTD_Sequence seqs[] = { { 10, 2, 4, 0 }, { 20, 3, 5, 0 } };
BlockSummary bs = ZSTD_get1BlockSummary(seqs, 2);
CHECK_EQ(bs.nbSequences, ERROR(externalSequences_invalid));
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3u : ZSTD_get1BlockSummary with rep ignored : ", testNb++);
{
static const ZSTD_Sequence seqs[] = {
{ 10, 2, 4, 2 },
{ 10, 3, 5, 2 },
{ 0, 7, 0, 3 },
};
BlockSummary bs = ZSTD_get1BlockSummary(seqs, 3);
CHECK_EQ(bs.nbSequences, 3);
CHECK_EQ(bs.litSize, 2 + 3 + 7);
CHECK_EQ(bs.blockSize, (4 + 5) + (2 + 3 + 7));
}
DISPLAYLEVEL(3, "OK \n");
assert(COUNTOF(nseqs) > COUNTOF(blocks));
for (i = 0; i < COUNTOF(blocks); ++i) {
BlockSummary bs;
DISPLAYLEVEL(3, "test%3u : ZSTD_get1BlockSummary with %u inputs : ",
testNb++, (unsigned)(COUNTOF(nseqs) - i));
bs = ZSTD_get1BlockSummary(nseqs + i, COUNTOF(nseqs) - i);
CHECK_EQ(bs.nbSequences, blocks[i].nbSequences);
CHECK_EQ(bs.litSize, blocks[i].litSize);
CHECK_EQ(bs.blockSize, blocks[i].blockSize);
DISPLAYLEVEL(3, "OK \n");
}
return testNb;
}
/* ============================================================= */
typedef struct {
int fail;
size_t outstanding;
} FUZ_dictAllocState;
static void* FUZ_dictMalloc(void* opaque, size_t size)
{
FUZ_dictAllocState* const state = (FUZ_dictAllocState*)opaque;
void* allocation;
if (state->fail) return NULL;
allocation = malloc(size);
if (allocation != NULL) ++state->outstanding;
return allocation;
}
static void FUZ_dictFree(void* opaque, void* allocation)
{
FUZ_dictAllocState* const state = (FUZ_dictAllocState*)opaque;
if (allocation != NULL) {
CHECK(state->outstanding != 0);
--state->outstanding;
}
free(allocation);
}
/* Writes an FSE table description (NCount) for symbols [0, maxSymbol] where
* symbol 1 is absent (zero count), symbols 2..maxSymbol have probability
* 1/2^tableLog, and symbol 0 takes the rest. Returns the bytes written. */
static size_t FUZ_writeNCount(BYTE* dst, size_t capacity,
unsigned maxSymbol, unsigned tableLog)
{
short norm[MaxSeq + 1] = { 0 };
size_t n;
unsigned s;
assert(maxSymbol >= 2);
for (s = 2; s <= maxSymbol; ++s)
norm[s] = 1;
norm[0] = (short)((1U << tableLog) - (maxSymbol - 1));
CHECK_VAR(n, FSE_writeNCount(dst, capacity, norm, maxSymbol, tableLog));
return n;
}
/* Guards against the optimal parser (btopt/btultra/btultra2) reading
* uninitialized or out-of-range costs when it seeds its statistics from a
* dictionary's entropy tables.
*
* The dictionary format allows the match length and literal length FSE tables
* to omit the high symbols (they are implicitly absent), and to contain absent
* symbols inside the listed range. The compressor used to build its tables
* only up to the last listed symbol, leaving the costs of the omitted symbols
* uninitialized. An absent symbol at the maximum table log costs tableLog+1,
* which is not below the frequency scale, and a Huffman table deeper than 11
* bits is likewise too deep for the scale. Both tripped an assert and, in
* release builds, made an invalid shift (#4681).
*
* The dictionary is hand-built to follow the format, rather than trained, so
* the test does not depend on dictionary training. The only observable
* contract is that compression with this valid dictionary succeeds and
* round-trips, so that is what is checked. The bug itself is detected by
* asserts (and sanitizers), so this test is most effective in debug builds. */
static void test_optimalParserDictEntropy(unsigned tnb)
{
BYTE dict[1024];
BYTE src[1024];
BYTE compressed[ZSTD_COMPRESSBOUND(sizeof(src))];
BYTE decoded[sizeof(src)];
size_t pos = 8;
unsigned i;
ZSTD_strategy strategy;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
DISPLAYLEVEL(3, "test%3u : optimal parser with sparse dictionary entropy tables : ", tnb);
CHECK(cctx != NULL && dctx != NULL);
/* Dictionary: magic, ID, skewed 12-bit Huffman table with every literal
* present, FSE tables at the maximum table log with symbol 1 absent
* (offset: symbols 0-MaxOff, ML & LL: symbols 0-2 only), repcodes, and
* content. */
MEM_writeLE32(dict, ZSTD_MAGIC_DICTIONARY);
MEM_writeLE32(dict + 4, 1);
{ unsigned counts[MaxLit + 1];
HUF_CREATE_STATIC_CTABLE(hufTable, MaxLit);
U64 workspace[HUF_WORKSPACE_SIZE_U64];
size_t n;
for (i = 0; i <= MaxLit; ++i)
counts[i] = i < 16 ? 1U << (16 - i) : 1;
CHECK_Z(HUF_buildCTable_wksp(hufTable, counts, MaxLit, 12, workspace, sizeof(workspace)));
CHECK_VAR(n, HUF_writeCTable_wksp(dict + pos, sizeof(dict) - pos,
hufTable, MaxLit, 12, workspace, sizeof(workspace)));
pos += n;
}
pos += FUZ_writeNCount(dict + pos, sizeof(dict) - pos, MaxOff, OffFSELog);
pos += FUZ_writeNCount(dict + pos, sizeof(dict) - pos, 2, MLFSELog);
pos += FUZ_writeNCount(dict + pos, sizeof(dict) - pos, 2, LLFSELog);
for (i = 0; i < ZSTD_REP_NUM; ++i, pos += 4)
MEM_writeLE32(dict + pos, 1 + i);
for (i = 0; i < 256; ++i)
dict[pos++] = (BYTE)i;
CHECK(pos <= sizeof(dict));
for (i = 0; i < sizeof(src); ++i)
src[i] = (BYTE)(i % 251);
for (strategy = ZSTD_btopt; strategy <= ZSTD_btultra2; ++strategy) {
size_t cSize, dSize;
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_strategy, (int)strategy));
CHECK_Z(ZSTD_CCtx_loadDictionary(cctx, dict, pos));
CHECK_VAR(cSize, ZSTD_compress2(cctx, compressed, sizeof(compressed), src, sizeof(src)));
CHECK_VAR(dSize, ZSTD_decompress_usingDict(dctx, decoded, sizeof(decoded),
compressed, cSize, dict, pos));
CHECK(dSize == sizeof(src) && !memcmp(decoded, src, sizeof(src)));
}
ZSTD_freeCCtx(cctx);
ZSTD_freeDCtx(dctx);
DISPLAYLEVEL(3, "OK \n");
}
static void test_dictionaryLoadErrors(const void* dict, size_t dictSize)
{
FUZ_dictAllocState state = { 0, 0 };
ZSTD_customMem const customMem = { FUZ_dictMalloc, FUZ_dictFree, &state };
ZSTD_CCtx* const cctx = ZSTD_createCCtx_advanced(customMem);
ZSTD_CCtx_params cdictParams;
BYTE const src[32] = { 0 };
BYTE compressed[ZSTD_COMPRESSBOUND(sizeof(src))];
size_t const baseline = state.outstanding;
size_t cSize;
CHECK(cctx != NULL);
CHECK_Z(ZSTD_CCtxParams_init(&cdictParams, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 1));
CHECK_Z(ZSTD_CCtx_loadDictionary_byReference(cctx, dict, dictSize));
/* Actual allocation failures must keep their distinct error code. */
state.fail = 1;
CHECK(ZSTD_createCDict_advanced2(dict, dictSize, ZSTD_dlm_byRef,
ZSTD_dct_fullDict, &cdictParams, customMem) == NULL);
CHECK_EQ(ZSTD_getErrorCode(ZSTD_compress2(cctx, compressed, sizeof(compressed),
src, sizeof(src))), ZSTD_error_memory_allocation);
CHECK_EQ(state.outstanding, baseline);
state.fail = 0;
/* A failed CDict initialization must free its partial allocation and
* leave the CCtx able to retry. The first eight bytes retain the magic
* number and dictionary ID, but omit the required entropy tables. */
CHECK(dictSize > 8);
CHECK_Z(ZSTD_CCtx_loadDictionary_byReference(cctx, dict, 8));
CHECK(ZSTD_createCDict_advanced2(dict, 8, ZSTD_dlm_byRef,
ZSTD_dct_fullDict, &cdictParams, customMem) == NULL);
CHECK_EQ(state.outstanding, baseline);
CHECK_EQ(ZSTD_getErrorCode(ZSTD_compress2(cctx, compressed, sizeof(compressed),
src, sizeof(src))), ZSTD_error_dictionary_corrupted);
CHECK_EQ(state.outstanding, baseline);
CHECK_EQ(ZSTD_getErrorCode(ZSTD_compress2(cctx, compressed, sizeof(compressed),
src, sizeof(src))), ZSTD_error_dictionary_corrupted);
CHECK_EQ(state.outstanding, baseline);
CHECK_Z(ZSTD_CCtx_loadDictionary_byReference(cctx, dict, dictSize));
CHECK_VAR(cSize, ZSTD_compress2(cctx, compressed, sizeof(compressed), src, sizeof(src)));
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx();
BYTE decoded[sizeof(src)];
CHECK(dctx != NULL);
{ CHECK_NEWV(dSize, ZSTD_decompress_usingDict(dctx, decoded, sizeof(decoded),
compressed, cSize, dict, dictSize));
CHECK(dSize == sizeof(src) && memcmp(src, decoded, sizeof(src)) == 0);
}
ZSTD_freeDCtx(dctx);
}
ZSTD_freeCCtx(cctx);
CHECK_EQ(state.outstanding, 0);
}
#if ZSTD_WINDOW_OVERFLOW_CORRECT_FREQUENTLY
/* dict has a valid full-dictionary header followed by at least 3 MB + 64 bytes of content. */
static void test_dictionaryRepcodesAfterOverflow(BYTE* dict, size_t headerSize)
{
/* Reaching the normal overflow threshold requires compressing a few GB.
* Only run this regression in frequent-correction mode, where a smaller
* dictionary and forceMaxWindow reproduce the same state cheaply. Loading
* the dictionary into the reused CCtx retains only 3 MB of its content. */
size_t const contentSize = 3 MB + 64;
size_t const preloadSize = 256 KB - 64;
size_t const fullDictSize = headerSize + contentSize;
U32 const rep = (U32)contentSize - 16;
size_t const compressedCapacity = ZSTD_compressBound(128 KB);
void* const compressed = malloc(compressedCapacity);
ZSTD_CCtx* const fresh = ZSTD_createCCtx();
ZSTD_CCtx* const reused = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
ZSTD_parameters params = ZSTD_getParams(1, ZSTD_CONTENTSIZE_UNKNOWN, 0);
ZSTD_CCtx_params cctxParams;
BYTE src[4 KB];
BYTE decoded[sizeof(src)];
size_t remaining = preloadSize;
U32 u;
CHECK(compressed != NULL && fresh != NULL && reused != NULL && dctx != NULL);
for (u = 0; u < ZSTD_REP_NUM; ++u)
MEM_writeLE32(dict + headerSize - 12 + 4 * u, repStartValue[u]);
MEM_writeLE32(dict + headerSize - 12, rep);
memcpy(src, dict + fullDictSize - rep, sizeof(src));
params.cParams.windowLog = 20;
params.cParams.hashLog = 18;
params.cParams.chainLog = 17;
CHECK_Z(ZSTD_CCtxParams_init_advanced(&cctxParams, params));
CHECK_Z(ZSTD_CCtxParams_setParameter(&cctxParams, ZSTD_c_forceMaxWindow, 1));
/* First prove that exactly this dictionary and these parameters work. */
CHECK_Z(ZSTD_compressBegin_advanced_internal(fresh, dict, fullDictSize,
ZSTD_dct_fullDict, ZSTD_dtlm_fast, NULL, &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN));
CHECK_EQ(fresh->blockState.matchState.window.nbOverflowCorrections, 0);
{ CHECK_NEWV(cSize, ZSTD_compressEnd(fresh, compressed, compressedCapacity, src, sizeof(src)));
{ CHECK_NEWV(dSize, ZSTD_decompress_usingDict(dctx, decoded, sizeof(decoded),
compressed, cSize, dict, fullDictSize));
CHECK(dSize == sizeof(src) && memcmp(src, decoded, sizeof(src)) == 0);
}
}
/* Advance the indices using actual compression, without replacing the
* workspace or manufacturing internal match-state indices. */
CHECK_Z(ZSTD_compressBegin_advanced_internal(reused, NULL, 0,
ZSTD_dct_auto, ZSTD_dtlm_fast, NULL, &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN));
while (remaining != 0) {
size_t const chunkSize = MIN(remaining, 128 KB);
CHECK_Z(ZSTD_compressContinue(reused, compressed, compressedCapacity,
dict + headerSize, chunkSize));
remaining -= chunkSize;
}
CHECK_Z(ZSTD_compressEnd(reused, compressed, compressedCapacity, NULL, 0));
CHECK_EQ(reused->blockState.matchState.window.nbOverflowCorrections, 0);
CHECK_EQ((size_t)(reused->blockState.matchState.window.nextSrc
- reused->blockState.matchState.window.base),
preloadSize + ZSTD_WINDOW_START_INDEX);
{ size_t const result = ZSTD_compressBegin_advanced_internal(reused, dict, fullDictSize,
ZSTD_dct_fullDict, ZSTD_dtlm_fast, NULL, &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN);
ZSTD_window_t const* const window = &reused->blockState.matchState.window;
U32 const loadedDictSize = (U32)(window->nextSrc - window->base) - window->lowLimit;
CHECK_EQ(window->nbOverflowCorrections, 1);
CHECK_EQ(loadedDictSize, 3 MB);
CHECK_LT(loadedDictSize, rep);
CHECK_Z(result);
}
{ CHECK_NEWV(cSize, ZSTD_compressEnd(reused, compressed, compressedCapacity, src, sizeof(src)));
{ CHECK_NEWV(dSize, ZSTD_decompress_usingDict(dctx, decoded, sizeof(decoded),
compressed, cSize, dict, fullDictSize));
CHECK(dSize == sizeof(src) && memcmp(src, decoded, sizeof(src)) == 0);
}
}
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtx(reused);
ZSTD_freeCCtx(fresh);
free(compressed);
}
#endif
static void test_largeDictionaryRepcodes(const void* dict, size_t dictSize)
{
U32 const shortCacheMaxDictSize = (1U << (32 - ZSTD_SHORT_CACHE_TAG_BITS)) - ZSTD_WINDOW_START_INDEX;
struct DictCase { size_t contentSize; U32 rep; };
struct DictCase const cases[] = {
{ 20 MB, shortCacheMaxDictSize + 1 },
{ 20 MB, 20 MB - 16 },
{ 20 MB, shortCacheMaxDictSize },
{ 10 MB, 10 MB - 16 },
};
int const levels[] = { -5, 1, 2, 3, 4, 5 };
size_t const headerSize = ZDICT_getDictHeaderSize(dict, dictSize);
BYTE* largeDict;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
BYTE src[4 KB];
BYTE compressed[8 KB];
BYTE decoded[sizeof(src)];
size_t ci, li;
U32 r;
CHECK_Z(headerSize);
CHECK(headerSize >= 12);
largeDict = (BYTE*)malloc(headerSize + 20 MB);
CHECK(largeDict != NULL && cctx != NULL && dctx != NULL);
memcpy(largeDict, dict, headerSize);
RDG_genBuffer(largeDict + headerSize, 20 MB, 0.5, 0., 0);
#if ZSTD_WINDOW_OVERFLOW_CORRECT_FREQUENTLY
test_dictionaryRepcodesAfterOverflow(largeDict, headerSize);
#endif
for (ci = 0; ci < COUNTOF(cases); ++ci) {
size_t const largeDictSize = headerSize + cases[ci].contentSize;
/* Make the large repcode a possible match in the full dictionary. */
memcpy(src, largeDict + largeDictSize - cases[ci].rep, sizeof(src));
for (li = 0; li < COUNTOF(levels); ++li) {
int const level = levels[li];
ZSTD_CCtx_params cdictParams;
ZSTD_compressionParameters cParams;
int reject;
CHECK_Z(ZSTD_CCtxParams_init(&cdictParams, level));
cParams = ZSTD_getCParamsFromCCtxParams(&cdictParams,
ZSTD_CONTENTSIZE_UNKNOWN, largeDictSize, ZSTD_cpm_createCDict);
reject = (cParams.strategy == ZSTD_fast || cParams.strategy == ZSTD_dfast)
&& cases[ci].rep > shortCacheMaxDictSize;
for (r = 0; r < ZSTD_REP_NUM; ++r) {
ZSTD_CDict* cdict;
U32 u;
size_t cSize;
ZSTD_inBuffer in = { src, sizeof(src), 0 };
ZSTD_outBuffer out = { compressed, sizeof(compressed), 0 };
for (u = 0; u < ZSTD_REP_NUM; ++u)
MEM_writeLE32(largeDict + headerSize - 12 + 4 * u, repStartValue[u]);
MEM_writeLE32(largeDict + headerSize - 12 + 4 * r, cases[ci].rep);
cdict = ZSTD_createCDict_byReference(largeDict, largeDictSize, level);
if (reject) {
CHECK(cdict == NULL);
/* The copy constructor and advanced constructor share the check. */
CHECK(ZSTD_createCDict(largeDict, largeDictSize, level) == NULL);
CHECK(ZSTD_createCDict_advanced(largeDict, largeDictSize,
ZSTD_dlm_byRef, ZSTD_dct_fullDict,
cParams, ZSTD_defaultCMem) == NULL);
/* Local CDict creation is deferred until compression starts. */
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, level));
CHECK_Z(ZSTD_CCtx_loadDictionary(cctx, largeDict, largeDictSize));
CHECK_EQ(ZSTD_getErrorCode(ZSTD_compress2(cctx, compressed,
sizeof(compressed), src, sizeof(src))), ZSTD_error_dictionary_corrupted);
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, level));
CHECK_Z(ZSTD_CCtx_loadDictionary_byReference(cctx, largeDict, largeDictSize));
CHECK_EQ(ZSTD_getErrorCode(ZSTD_compressStream2(cctx, &out, &in,
ZSTD_e_continue)), ZSTD_error_dictionary_corrupted);
/* The dictionary is format-valid. The non-CDict path retains
* all of its content, so it must still round-trip. */
CHECK_VAR(cSize, ZSTD_compress_usingDict(cctx, compressed, sizeof(compressed),
src, sizeof(src), largeDict, largeDictSize, level));
} else {
CHECK(cdict != NULL);
CHECK_VAR(cSize, ZSTD_compress_usingCDict(cctx, compressed, sizeof(compressed),
src, sizeof(src), cdict));
{ CHECK_NEWV(dSize, ZSTD_decompress_usingDict(dctx, decoded, sizeof(decoded),
compressed, cSize, largeDict, largeDictSize));
CHECK(dSize == sizeof(src) && memcmp(src, decoded, sizeof(src)) == 0);
}
/* Start with unknown input size to exercise attached streaming. */
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
CHECK_Z(ZSTD_CCtx_refCDict(cctx, cdict));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue));
CHECK(in.pos == in.size);
CHECK(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end) == 0);
cSize = out.pos;
ZSTD_freeCDict(cdict);
}
{ CHECK_NEWV(dSize, ZSTD_decompress_usingDict(dctx, decoded, sizeof(decoded),
compressed, cSize, largeDict, largeDictSize));
CHECK(dSize == sizeof(src) && memcmp(src, decoded, sizeof(src)) == 0);
}
}
}
}
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtx(cctx);
free(largeDict);
}
/* Multi-DDict fixtures deliberately share entropy tables but have different
* content. The handcrafted frame always copies eight dictionary bytes: unlike
* compressor-generated data, it cannot turn into a literals-only block and
* accidentally pass when the decoder selects the wrong dictionary. */
enum {
FUZ_mdict_knownSize = 1,
FUZ_mdict_checksum = 2
};
static void FUZ_mdict_makeDict(BYTE dst[44], U32 id, int fill)
{
static const BYTE header[28] = {
0x37,0xa4,0x30,0xec, 0,0,0,0, /* magic and dictionary ID */
128,0x10, /* Huffman table */
0xf0,0x03, 0xf0,0x03, 0xf0,0x03, /* FSE tables */
4,0,0,0, 4,0,0,0, 4,0,0,0 /* repeat offsets */
};
memcpy(dst, header, sizeof(header));
MEM_writeLE32(dst + 4, id);
memset(dst + sizeof(header), fill, 16);
}
/* ID 0 creates a raw-content DDict; other IDs create formatted DDicts. */
static ZSTD_DDict* FUZ_mdict_createDict(U32 id, int fill)
{
BYTE bytes[44];
ZSTD_DDict* ddict;
FUZ_mdict_makeDict(bytes, id, fill);
ddict = id ? ZSTD_createDDict(bytes, sizeof(bytes))
: ZSTD_createDDict(bytes + 28, 16);
CHECK(ddict != NULL && ZSTD_getDictID_fromDDict(ddict) == id);
return ddict;
}
/* One compressed block, no literals, one match of length 8 at offset 4.
* ID 0 omits the frame's dictID. A known content size permits the streaming
* shortcut; the optional checksum is for the expected output "AAAAAAAA". */
static size_t FUZ_mdict_makeFrame(BYTE* dst, U32 id, unsigned flags)
{
static const BYTE block[] = {
0x3d,0,0, /* last compressed block, seven-byte payload */
0,1,0x54, /* no literals; one sequence; RLE sequence tables */
0,0,5,1 /* LL=0, repeat offset, ML=8, bitstream */
};
int const knownSize = (flags & FUZ_mdict_knownSize) != 0;
size_t n = 4;
MEM_writeLE32(dst, ZSTD_MAGICNUMBER);
dst[n++] = (BYTE)((id ? 3 : 0) | (knownSize ? 32 : 0)
| ((flags & FUZ_mdict_checksum) ? 4 : 0));
if (!knownSize) dst[n++] = 0; /* 1 KiB window */
if (id) { MEM_writeLE32(dst + n, id); n += 4; }
if (knownSize) dst[n++] = 8;
memcpy(dst + n, block, sizeof(block)); n += sizeof(block);
if (flags & FUZ_mdict_checksum) {
MEM_writeLE32(dst + n, (U32)XXH64("AAAAAAAA", 8, 0)); n += 4;
}
return n;
}
static size_t FUZ_mdict_makeSkip(BYTE* dst, unsigned variant)
{
MEM_writeLE32(dst, ZSTD_MAGIC_SKIPPABLE_START + variant);
MEM_writeLE32(dst + 4, 0); /* empty skippable frame */
return 8;
}
/* Register one or two DDicts in a new multiple-DDict context. */
static ZSTD_DCtx* FUZ_mdict_createContext(const ZSTD_DDict* first,
const ZSTD_DDict* second)
{
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
CHECK(dctx != NULL);
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 1));
if (first) CHECK_Z(ZSTD_DCtx_refDDict(dctx, first));
if (second) CHECK_Z(ZSTD_DCtx_refDDict(dctx, second));
return dctx;
}
/* Feed complete frames in chunks, returning the first error or total output.
* These tests provide enough output space, so every call must make progress. */
static size_t FUZ_mdict_decompressStream(ZSTD_DCtx* dctx, void* dst, size_t capacity,
const BYTE* src, size_t size, size_t chunk)
{
ZSTD_outBuffer out = {dst, capacity, 0};
size_t pos = 0, ret = 1;
CHECK(chunk != 0);
while (pos < size) {
size_t const n = MIN(size - pos, chunk);
ZSTD_inBuffer in = {src + pos, n, 0};
while (in.pos < in.size) {
size_t const oldIn = in.pos, oldOut = out.pos;
ret = ZSTD_decompressStream(dctx, &out, &in);
if (ZSTD_isError(ret)) return ret;
CHECK(in.pos > oldIn || out.pos > oldOut);
}
pos += n;
}
CHECK_EQ(ret, 0);
return out.pos;
}
/* The bufferless API requires exactly the input size requested by the decoder. */
static size_t FUZ_mdict_decompressContinue(ZSTD_DCtx* dctx, BYTE* dst, size_t capacity,
const BYTE* src, size_t size)
{
size_t inPos = 0, outPos = 0;
for (;;) {
size_t const n = ZSTD_nextSrcSizeToDecompress(dctx);
size_t r;
if (n == 0) break;
CHECK(inPos <= size && n <= size - inPos);
CHECK(outPos <= capacity);
r = ZSTD_decompressContinue(dctx, outPos != 0 ? dst + outPos : dst, capacity - outPos,
src + inPos, n);
if (ZSTD_isError(r)) return r;
inPos += n;
outPos += r;
}
CHECK_EQ(inPos, size);
return outPos;
}
static void FUZ_mdict_checkOutput(size_t r, const void* out, const char* expected)
{
CHECK_Z(r);
CHECK_EQ(r, strlen(expected));
CHECK_EQ(memcmp(out, expected, r), 0);
}
/* Layout of FUZ_mdict_makeDict(): 28-byte header, then 16 bytes of content. */
#define FUZ_DH_HEADER 28
#define FUZ_DH_CONTENT 16
#define FUZ_DH_DICT (FUZ_DH_HEADER + FUZ_DH_CONTENT)
/* A frame whose only sequence is a four-byte match at `offset`. */
static size_t FUZ_ddictHeader_makeFrame(BYTE* frame, U32 dictID, unsigned offset)
{
BYTE* op = frame + 17;
unsigned const ofCode = ZSTD_highbit32(offset + ZSTD_REP_NUM);
MEM_writeLE32(frame, ZSTD_MAGICNUMBER);
frame[4] = 0x83; frame[5] = 0x38; /* 4-byte dictID and size, 128 KiB window */
MEM_writeLE32(frame + 6, dictID);
MEM_writeLE32(frame + 10, 4);
*op++ = 0; /* no literals */
*op++ = 1; *op++ = 0x54; /* one sequence, three RLE tables */
*op++ = 0; *op++ = (BYTE)ofCode; *op++ = 1; /* LL=0, ML=4 */
*op++ = (BYTE)((1U << ofCode) | (offset + 3 - (1U << ofCode)));
MEM_writeLE24(frame + 14, (U32)((op - (frame + 17)) << 3) | (bt_compressed << 1) | 1);
return (size_t)(op - frame);
}
/* Matches into the header must be rejected; all others must decode. */
static void FUZ_ddictHeader_check(size_t r, const BYTE* out, const BYTE* dict,
unsigned offset, int raw)
{
if (!raw && offset > FUZ_DH_CONTENT) {
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_corruption_detected);
} else {
CHECK_EQ(r, 4);
CHECK_EQ(memcmp(out, dict + FUZ_DH_DICT - offset, 4), 0);
}
}
static void FUZ_ddictHeader_testOffset(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict,
const BYTE* dict, unsigned offset, int raw)
{
ZSTD_dictContentType_e const type = raw ? ZSTD_dct_rawContent : ZSTD_dct_fullDict;
BYTE frame[32], out[4];
size_t const size = FUZ_ddictHeader_makeFrame(frame, ZSTD_getDictID_fromDDict(ddict), offset);
int mode;
/* Every way of installing a dictionary reaches the same decoder state. */
for (mode = 0; mode < (raw ? 5 : 6); ++mode) {
size_t r;
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters));
switch (mode) {
case 0: r = ZSTD_decompress_usingDDict(dctx, out, sizeof(out), frame, size, ddict); break;
case 1: case 2:
CHECK_Z(ZSTD_DCtx_refDDict(dctx, ddict));
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, size, mode == 1 ? 1 : size);
break;
case 3:
CHECK_Z(ZSTD_DCtx_loadDictionary_advanced(dctx, dict, FUZ_DH_DICT, ZSTD_dlm_byCopy, type));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, size);
break;
case 4:
CHECK_Z(ZSTD_DCtx_refPrefix_advanced(dctx, dict, FUZ_DH_DICT, type));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, size);
break;
default: /* Control: the non-DDict API, which already rejected these. */
r = ZSTD_decompress_usingDict(dctx, out, sizeof(out), frame, size, dict, FUZ_DH_DICT);
break;
}
FUZ_ddictHeader_check(r, out, dict, offset, raw);
}
}
static void test_decompressDDictHeader(unsigned tnb)
{
/* The last valid offset, the first header byte, and the dictionary magic. */
static const unsigned offsets[] = { FUZ_DH_CONTENT, FUZ_DH_CONTENT + 1, FUZ_DH_DICT };
BYTE dict[FUZ_DH_DICT];
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
ZSTD_DDict* ddicts[2];
int raw;
size_t i;
DISPLAYLEVEL(3, "test%3u : formatted dictionary headers are not match history : ", tnb);
FUZ_mdict_makeDict(dict, 1, 0);
for (i = 0; i < FUZ_DH_CONTENT; ++i) dict[FUZ_DH_HEADER + i] = (BYTE)(0x80 + i);
ddicts[0] = ZSTD_createDDict(dict, sizeof(dict));
ddicts[1] = ZSTD_createDDict_advanced(dict, sizeof(dict),
ZSTD_dlm_byCopy, ZSTD_dct_rawContent, ZSTD_defaultCMem);
CHECK(dctx != NULL && ddicts[0] != NULL && ddicts[1] != NULL);
for (raw = 0; raw < 2; ++raw) {
/* Legacy decoders need the complete dictionary, header included. */
CHECK_EQ(ZSTD_DDict_dictSize(ddicts[raw]), sizeof(dict));
CHECK_EQ(memcmp(ZSTD_DDict_dictContent(ddicts[raw]), dict, sizeof(dict)), 0);
for (i = 0; i < COUNTOF(offsets); ++i)
FUZ_ddictHeader_testOffset(dctx, ddicts[raw], dict, offsets[i], raw);
}
ZSTD_freeDDict(ddicts[0]);
ZSTD_freeDDict(ddicts[1]);
ZSTD_freeDCtx(dctx);
DISPLAYLEVEL(3, "OK \n");
}
static void FUZ_mdict_testOneShot(void)
{
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DDict* const b = FUZ_mdict_createDict(2222, 'B');
ZSTD_DDict* const raw = FUZ_mdict_createDict(0, 'R');
BYTE frame[96], out[32], tmp[32];
size_t n, r;
int magicless, invalid;
/* B is registered last. Every frame must nevertheless install its own
* exact match, including after reset and within concatenated input. */
for (magicless = 0; magicless <= 1; ++magicless) {
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(raw, a);
size_t const skip = magicless ? 4 : 0;
static const U32 ids[] = {1111, 2222, 0};
size_t i;
CHECK_Z(ZSTD_DCtx_refDDict(dctx, b));
if (magicless) CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless));
n = FUZ_mdict_makeFrame(frame, 1111, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame + skip, n - skip);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
n = FUZ_mdict_makeFrame(frame, 2222, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame + skip, n - skip);
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
n = magicless ? 0 : FUZ_mdict_makeSkip(frame, 7);
for (i = 0; i < COUNTOF(ids); ++i) {
size_t const fSize = FUZ_mdict_makeFrame(tmp, ids[i], i == 1 ? FUZ_mdict_knownSize : 0);
memcpy(frame + n, tmp + skip, fSize - skip);
n += fSize - skip;
}
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAABBBBBBBBRRRRRRRR");
CHECK_Z(ZSTD_freeDCtx(dctx));
}
/* Failed headers must not prevent the next frame from finding its match. */
for (invalid = 0; invalid < 2; ++invalid) {
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(raw, a);
FUZ_mdict_makeFrame(frame, 1111, 0);
if (invalid) memset(frame, 0xff, 5);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, 5);
CHECK(ZSTD_isError(r));
n = FUZ_mdict_makeFrame(frame, 1111, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_freeDCtx(dctx));
}
CHECK_Z(ZSTD_freeDDict(a)); CHECK_Z(ZSTD_freeDDict(b));
CHECK_Z(ZSTD_freeDDict(raw));
}
static void FUZ_mdict_testExplicitDictionary(void)
{
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DDict* const b = FUZ_mdict_createDict(2222, 'B');
BYTE frame[32], out[32], dict[44];
size_t n, r;
int local;
for (local = 0; local < 2; ++local) {
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(b, NULL);
FUZ_mdict_makeDict(dict, 2222, 'B');
if (local) CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, dict, sizeof(dict)));
n = FUZ_mdict_makeFrame(frame, 1111, 0);
/* The set must not override the explicit argument, even though A is
* absent from the set and strict lookup would reject this frame. */
r = ZSTD_decompress_usingDDict(dctx, out, sizeof(out), frame, n, b);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
r = ZSTD_decompress_usingDDict(dctx, out, sizeof(out), frame, n, a);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
r = ZSTD_decompress_usingDict(dctx, out, sizeof(out), frame, n, dict, sizeof(dict));
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
FUZ_mdict_makeDict(dict, 1111, 'A');
r = ZSTD_decompress_usingDict(dctx, out, sizeof(out), frame, n, dict, sizeof(dict));
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
/* Header validation must likewise honor the bufferless API's DDict. */
CHECK_Z(ZSTD_decompressBegin_usingDDict(dctx, b));
r = FUZ_mdict_decompressContinue(dctx, out, sizeof(out), frame, n);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
CHECK_Z(ZSTD_decompressBegin_usingDDict(dctx, a));
r = FUZ_mdict_decompressContinue(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
/* Explicit dictionaries must not replace the configured set or the
* context-owned B. Disabling the set must not retain borrowed B. */
n = FUZ_mdict_makeFrame(frame, 2222, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
if (!local) { CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 0)); }
n = FUZ_mdict_makeFrame(frame, 0, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
if (local) { FUZ_mdict_checkOutput(r, out, "BBBBBBBB"); }
else { CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_corruption_detected); }
CHECK_Z(ZSTD_freeDCtx(dctx));
}
CHECK_Z(ZSTD_freeDDict(a)); CHECK_Z(ZSTD_freeDDict(b));
}
static void FUZ_mdict_testPrefixCompatibility(void)
{
ZSTD_DDict* const b = FUZ_mdict_createDict(2222, 'B');
BYTE prefix[16], frame[64], out[32];
size_t n, first, r;
int multiple;
memset(prefix, 'A', sizeof(prefix));
for (multiple = 0; multiple < 2; ++multiple) {
ZSTD_DCtx* const dctx = multiple ? FUZ_mdict_createContext(b, NULL) : ZSTD_createDCtx();
CHECK(dctx != NULL);
CHECK_Z(ZSTD_DCtx_refPrefix(dctx, prefix, sizeof(prefix)));
/* Preserve the existing single-DDict one-shot behavior: a prefix is
* resolved once per call, even across concatenated frames. This also
* applies when the multi-DDict flag is enabled but the set is empty. */
first = FUZ_mdict_makeFrame(frame, 0, 0);
n = first + FUZ_mdict_makeFrame(frame + first, 0, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAAAAAAAAAA");
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, first);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_corruption_detected);
/* Streaming still consumes the prefix on its first frame. */
CHECK_Z(ZSTD_DCtx_refPrefix(dctx, prefix, sizeof(prefix)));
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, first, 1);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, first, 1);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_corruption_detected);
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
/* Do not change the established skippable-frame prefix behavior in
* this multi-DDict fix. One-shot keeps its dictionary for the call;
* streaming consumes the prefix even on a skippable frame. */
CHECK_Z(ZSTD_DCtx_refPrefix(dctx, prefix, sizeof(prefix)));
first = FUZ_mdict_makeSkip(frame, 7);
n = first + FUZ_mdict_makeFrame(frame + first, 0, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_DCtx_refPrefix(dctx, prefix, sizeof(prefix)));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, first);
CHECK_EQ(r, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame + first, n - first);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_corruption_detected);
CHECK_Z(ZSTD_DCtx_refPrefix(dctx, prefix, sizeof(prefix)));
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, n, 1);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_corruption_detected);
CHECK_Z(ZSTD_freeDCtx(dctx));
}
CHECK_Z(ZSTD_freeDDict(b));
}
/* ---------------------------------------------------------------------------
* Prefix lifetime and session reset of the one-shot decompression functions.
* These test the contracts documented in zstd.h for ZSTD_DCtx_refPrefix(),
* ZSTD_DCtx_reset() and ZSTD_decompressDCtx() / _usingDict() / _usingDDict().
*
* Prefix tests use the fixture below: the frame copies 8 bytes out of the
* prefix, so it decodes to "AAAAAAAA" if and only if the prefix is in use.
* Unlike compressor-generated data it cannot degrade to a literals-only block
* and pass by accident when the prefix has been lost.
* ------------------------------------------------------------------------- */
typedef struct {
ZSTD_DCtx* dctx;
BYTE dict[44]; /* formatted dictionary "A"; its last 16 bytes are the raw content */
BYTE frame[32]; /* decodes to "AAAAAAAA" only when dictionary A is available */
size_t frameSize;
int asFullDict; /* reference the formatted dictionary instead of its raw content */
} FUZ_pfx;
static void FUZ_pfx_init(FUZ_pfx* p, unsigned asFullDict)
{
p->dctx = ZSTD_createDCtx();
CHECK(p->dctx != NULL);
FUZ_mdict_makeDict(p->dict, 1111, 'A');
p->frameSize = FUZ_mdict_makeFrame(p->frame, 0, 0);
p->asFullDict = (int)asFullDict;
}
static void FUZ_pfx_ref(FUZ_pfx* p)
{
if (p->asFullDict) {
CHECK_Z(ZSTD_DCtx_refPrefix_advanced(p->dctx, p->dict, 44, ZSTD_dct_fullDict));
} else {
CHECK_Z(ZSTD_DCtx_refPrefix(p->dctx, p->dict + 28, 16));
}
}
/* Is the prefix still referenced? Probes with ZSTD_decompressDCtx(),
* so it consumes the prefix: call it last, and reference again before reuse.
* A frame that needs a missing dictionary must not decode to "AAAAAAAA". */
static int FUZ_pfx_isSet(FUZ_pfx* p)
{
BYTE out[8];
size_t const r = ZSTD_decompressDCtx(p->dctx, out, sizeof(out), p->frame, p->frameSize);
return r == 8 && memcmp(out, "AAAAAAAA", 8) == 0;
}
/* Begin streaming a frame, but stop after `size` bytes of input. */
static void FUZ_pfx_startStream(ZSTD_DCtx* dctx, const BYTE* src, size_t size)
{
BYTE out[32];
ZSTD_inBuffer in = {src, size, 0};
ZSTD_outBuffer ob = {out, sizeof(out), 0};
size_t const r = ZSTD_decompressStream(dctx, &ob, &in);
CHECK_Z(r);
CHECK(r != 0); /* frame not finished */
}
/* ZSTD_decompressDCtx() uses the prefix for exactly one call, whatever the
* input is. The explicit dictionary functions use their own dictionary and
* leave the prefix alone. */
static void FUZ_pfx_testOneShot(unsigned mode)
{
FUZ_pfx p;
ZSTD_DDict* const b = FUZ_mdict_createDict(0, 'B');
BYTE rawB[16], in[64], out[32];
size_t n, r;
unsigned i;
FUZ_pfx_init(&p, mode);
memset(rawB, 'B', sizeof(rawB));
/* Every frame of a single call, including after a skippable frame, uses the prefix. */
n = FUZ_mdict_makeSkip(in, 0);
memcpy(in + n, p.frame, p.frameSize); n += p.frameSize;
memcpy(in + n, p.frame, p.frameSize); n += p.frameSize;
FUZ_pfx_ref(&p);
r = ZSTD_decompressDCtx(p.dctx, out, sizeof(out), in, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAAAAAAAAAA");
CHECK(!FUZ_pfx_isSet(&p));
/* Consumed even when there is no frame to decode, or the call fails. */
for (i = 0; i < 3; ++i) {
if (i == 0) n = 0;
else if (i == 1) n = FUZ_mdict_makeSkip(in, 0);
else { memset(in, 0xff, 8); n = 8; }
FUZ_pfx_ref(&p);
r = ZSTD_decompressDCtx(p.dctx, out, sizeof(out), in, n);
CHECK((int)ZSTD_isError(r) == (int)(i == 2));
CHECK(!FUZ_pfx_isSet(&p));
}
/* Explicit dictionaries take precedence and do not consume the prefix. */
FUZ_pfx_ref(&p);
r = ZSTD_decompress_usingDict(p.dctx, out, sizeof(out), p.frame, p.frameSize, rawB, sizeof(rawB));
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
r = ZSTD_decompress_usingDDict(p.dctx, out, sizeof(out), p.frame, p.frameSize, b);
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
CHECK(FUZ_pfx_isSet(&p));
CHECK_Z(ZSTD_freeDDict(b));
CHECK_Z(ZSTD_freeDCtx(p.dctx));
}
/* ZSTD_decompressStream() uses the prefix until the first frame is complete
* (returns 0), including frames with no output. Input is fed in chunks of every
* size, so the header and the payload are split at every possible position. */
static void FUZ_pfx_testStreaming(unsigned mode)
{
static const BYTE skipEmpty[] = {0x50,0x2a,0x4d,0x18, 0,0,0,0};
static const BYTE skipData[] = {0x5f,0x2a,0x4d,0x18, 5,0,0,0, 1,2,3,4,5};
static const BYTE emptyFrame[] = {0x28,0xb5,0x2f,0xfd,0x20,0,1,0,0};
static const struct { const BYTE* src; size_t size; } noOutput[] = {
{skipEmpty, sizeof(skipEmpty)},
{skipData, sizeof(skipData)},
{emptyFrame, sizeof(emptyFrame)} /* zstd frame with no content */
};
FUZ_pfx p;
BYTE frame[32], out[32];
size_t size, chunk, r;
unsigned i;
FUZ_pfx_init(&p, mode);
/* Frames with and without a known size, which can be decoded in one pass. */
for (i = 0; i < 2; ++i) {
size = FUZ_mdict_makeFrame(frame, 0, i ? FUZ_mdict_knownSize : 0);
for (chunk = 1; chunk <= size; ++chunk) {
FUZ_pfx_ref(&p);
r = FUZ_mdict_decompressStream(p.dctx, out, sizeof(out), frame, size, chunk);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK(!FUZ_pfx_isSet(&p));
}
}
/* Frames that produce no output also consume the prefix. */
for (i = 0; i < COUNTOF(noOutput); ++i) {
for (chunk = 1; chunk <= noOutput[i].size; ++chunk) {
FUZ_pfx_ref(&p);
CHECK_EQ(FUZ_mdict_decompressStream(p.dctx, out, sizeof(out),
noOutput[i].src, noOutput[i].size, chunk), 0);
CHECK(!FUZ_pfx_isSet(&p));
}
}
CHECK_Z(ZSTD_freeDCtx(p.dctx));
}
/* Interaction with ZSTD_DCtx_reset() and with one-shot calls after a stream is
* started and abandoned. This guards the bug where a prefix belonging to an
* abandoned stream was silently applied to the next, unrelated frame. */
static void FUZ_pfx_testReset(unsigned mode)
{
FUZ_pfx p;
ZSTD_DDict* const b = FUZ_mdict_createDict(0, 'B');
BYTE rawB[16], out[32];
size_t r;
unsigned how;
FUZ_pfx_init(&p, mode);
memset(rawB, 'B', sizeof(rawB));
/* A session reset keeps a prefix that streaming has not touched yet. */
FUZ_pfx_ref(&p);
CHECK_Z(ZSTD_DCtx_reset(p.dctx, ZSTD_reset_session_only));
CHECK(FUZ_pfx_isSet(&p));
/* Once streaming has started, even with an incomplete header, abandoning the
* frame discards the prefix: either explicitly, or implicitly by calling any
* one-shot function (which must not leak the prefix to the next frame). */
for (how = 0; how < 5; ++how) {
size_t const sizes[] = {1, p.frameSize - 1};
size_t i;
for (i = 0; i < COUNTOF(sizes); ++i) {
FUZ_pfx_ref(&p);
FUZ_pfx_startStream(p.dctx, p.frame, sizes[i]);
switch (how) {
case 0: CHECK_Z(ZSTD_DCtx_reset(p.dctx, ZSTD_reset_session_only)); break;
case 1: CHECK_Z(ZSTD_DCtx_reset(p.dctx, ZSTD_reset_session_and_parameters)); break;
case 2: break; /* FUZ_pfx_isSet() is itself a one-shot call */
case 3:
r = ZSTD_decompress_usingDict(p.dctx, out, sizeof(out), p.frame, p.frameSize, rawB, sizeof(rawB));
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
break;
default:
r = ZSTD_decompress_usingDDict(p.dctx, out, sizeof(out), p.frame, p.frameSize, b);
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
break;
}
CHECK(!FUZ_pfx_isSet(&p));
}
}
/* Resetting the parameters discards the prefix when no frame is active ... */
FUZ_pfx_ref(&p);
CHECK_Z(ZSTD_DCtx_reset(p.dctx, ZSTD_reset_parameters));
CHECK(!FUZ_pfx_isSet(&p));
/* ... and is rejected, with no effect, while a frame is being decompressed. */
FUZ_pfx_ref(&p);
FUZ_pfx_startStream(p.dctx, p.frame, 1);
CHECK_EQ(ZSTD_getErrorCode(ZSTD_DCtx_reset(p.dctx, ZSTD_reset_parameters)), ZSTD_error_stage_wrong);
r = FUZ_mdict_decompressStream(p.dctx, out, sizeof(out), p.frame + 1, p.frameSize - 1, p.frameSize);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_freeDDict(b));
CHECK_Z(ZSTD_freeDCtx(p.dctx));
}
/* Referencing any prefix or dictionary (including NULL or empty) replaces the
* previous prefix, and so does ZSTD_initDStream(). */
static void FUZ_pfx_testReplacement(unsigned mode)
{
FUZ_pfx p;
ZSTD_DDict* const b = FUZ_mdict_createDict(0, 'B');
BYTE rawB[16], out[32];
size_t r;
unsigned i;
FUZ_pfx_init(&p, mode);
memset(rawB, 'B', sizeof(rawB));
/* Replaced by something else: only the new dictionary is used. */
for (i = 0; i < 3; ++i) {
FUZ_pfx_ref(&p);
switch (i) {
case 0: CHECK_Z(ZSTD_DCtx_refPrefix(p.dctx, rawB, sizeof(rawB))); break;
case 1: CHECK_Z(ZSTD_DCtx_loadDictionary(p.dctx, rawB, sizeof(rawB))); break;
default: CHECK_Z(ZSTD_DCtx_refDDict(p.dctx, b)); break;
}
r = ZSTD_decompressDCtx(p.dctx, out, sizeof(out), p.frame, p.frameSize);
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
}
/* Replaced by nothing. */
for (i = 0; i < 4; ++i) {
FUZ_pfx_ref(&p);
switch (i) {
case 0: CHECK_Z(ZSTD_DCtx_refPrefix(p.dctx, NULL, 0)); break;
case 1: CHECK_Z(ZSTD_DCtx_loadDictionary(p.dctx, NULL, 0)); break;
case 2: CHECK_Z(ZSTD_DCtx_refDDict(p.dctx, NULL)); break;
default: CHECK_Z(ZSTD_initDStream(p.dctx)); break;
}
CHECK(!FUZ_pfx_isSet(&p));
}
CHECK_Z(ZSTD_freeDDict(b));
CHECK_Z(ZSTD_freeDCtx(p.dctx));
}
/* One-shot functions reset the streaming session first: they work on a DCtx
* left mid-stream or after a streaming error, and leave it ready to stream a
* new frame without any explicit ZSTD_DCtx_reset(). Before the fix they failed
* (or decoded garbage) on a dirty DCtx, and the single-pass streaming shortcut
* was entangled with this.
* `api` selects the one-shot function:
* 0: decompressDCtx, 1: usingDict, 2: usingDDict, 3: decompressDCtx with
* ZSTD_d_refMultipleDDicts (a separate code path). The DCtx always
* references dictionary A, so streaming can decode the frame as well. */
static void FUZ_reset_testDirtyStream(unsigned api)
{
ZSTD_DDict* const ddict = FUZ_mdict_createDict(1111, 'A');
ZSTD_DCtx* const dctx = api == 3 ? FUZ_mdict_createContext(ddict, NULL) : ZSTD_createDCtx();
BYTE dict[44], frame[32], bad[32], out[32];
size_t const frameSize = FUZ_mdict_makeFrame(frame, 1111, 0);
unsigned dirty;
CHECK(dctx != NULL);
if (api != 3) CHECK_Z(ZSTD_DCtx_refDDict(dctx, ddict));
FUZ_mdict_makeDict(dict, 1111, 'A');
memcpy(bad, frame, frameSize);
bad[frameSize - 10] = 0x3f; /* block header (last 10 bytes): reserved block type */
for (dirty = 0; dirty < 3; ++dirty) {
/* 0: incomplete frame header, 1: inside the block, 2: streaming error */
const BYTE* const src = dirty == 2 ? bad : frame;
size_t const size = dirty == 0 ? 3 : dirty == 1 ? frameSize - 1 : frameSize;
ZSTD_inBuffer in = {src, size, 0};
ZSTD_outBuffer ob = {out, sizeof(out), 0};
size_t r = ZSTD_decompressStream(dctx, &ob, &in);
CHECK((int)ZSTD_isError(r) == (int)(dirty == 2));
switch (api) {
case 1: r = ZSTD_decompress_usingDict(dctx, out, sizeof(out), frame, frameSize, dict, sizeof(dict)); break;
case 2: r = ZSTD_decompress_usingDDict(dctx, out, sizeof(out), frame, frameSize, ddict); break;
default: r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, frameSize); break;
}
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, frameSize, frameSize);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
}
CHECK_Z(ZSTD_freeDCtx(dctx));
CHECK_Z(ZSTD_freeDDict(ddict));
}
static unsigned test_sessionReset(unsigned testNb)
{
static const struct { const char* name; void (*run)(unsigned); unsigned nbParams; } tests[] = {
{"prefix lifetime: one-shot calls", FUZ_pfx_testOneShot, 2},
{"prefix lifetime: streaming", FUZ_pfx_testStreaming, 2},
{"prefix lifetime: reset and abandoned streams", FUZ_pfx_testReset, 2},
{"prefix lifetime: replacement", FUZ_pfx_testReplacement, 2},
{"one-shot decompression resets the streaming session", FUZ_reset_testDirtyStream, 4},
};
size_t i;
unsigned param;
for (i = 0; i < COUNTOF(tests); ++i) {
DISPLAYLEVEL(3, "test%3u : %s : ", testNb++, tests[i].name);
for (param = 0; param < tests[i].nbParams; ++param) tests[i].run(param);
DISPLAYLEVEL(3, "OK \n");
}
return testNb;
}
static void FUZ_mdict_testZeroID(void)
{
U32 const zero = 0;
U32 id = 1;
ZSTD_DDict* const raw = FUZ_mdict_createDict(0, 'R');
ZSTD_DDict* full;
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(raw, NULL);
BYTE frame[32], out[32];
size_t n, r;
/* Find a nonzero ID in raw dictionary 0's initial 64-slot hash bucket. */
while ((XXH64(&id, sizeof(id), 0) & 63) != (XXH64(&zero, sizeof(zero), 0) & 63))
++id;
full = FUZ_mdict_createDict(id, 'A');
n = FUZ_mdict_makeFrame(frame, id, 0);
/* An occupied ID-0 slot is not a match for the missing nonzero ID. */
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, n, 1);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
/* After insertion, lookup must probe past raw dictionary 0 to find A.
* An actual ID-less frame must still find the raw dictionary itself. */
CHECK_Z(ZSTD_DCtx_refDDict(dctx, full));
CHECK_Z(ZSTD_DCtx_refDDict(dctx, raw));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, n, 1);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
n = FUZ_mdict_makeFrame(frame, 0, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "RRRRRRRR");
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, n, 1);
FUZ_mdict_checkOutput(r, out, "RRRRRRRR");
CHECK_Z(ZSTD_freeDCtx(dctx));
CHECK_Z(ZSTD_freeDDict(raw)); CHECK_Z(ZSTD_freeDDict(full));
}
static void FUZ_mdict_testClearReferences(void)
{
int mode;
/* Register A, clear its references, free A, then register B and request A.
* The result must be dictionary_wrong, without consulting freed A. */
for (mode = 0; mode < 5; ++mode) {
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DDict* const b = FUZ_mdict_createDict(2222, 'B');
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(a, NULL);
BYTE frame[32], out[32], dummy = 0;
size_t const n = FUZ_mdict_makeFrame(frame, 1111, 0);
size_t r;
switch (mode) {
case 0: CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_parameters)); break;
case 1: CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, NULL, 0)); break;
case 2: CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, &dummy, 0)); break;
case 3: CHECK_Z(ZSTD_DCtx_refDDict(dctx, NULL)); break;
default:
/* Even a dictionary already selected for decoding is released. */
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 0));
break;
}
CHECK_Z(ZSTD_freeDDict(a));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 1));
CHECK_Z(ZSTD_DCtx_refDDict(dctx, b));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
CHECK_Z(ZSTD_freeDCtx(dctx)); CHECK_Z(ZSTD_freeDDict(b));
}
}
static void FUZ_mdict_testDisableSet(void)
{
int mode;
for (mode = 0; mode < 4; ++mode) {
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DDict* const b = FUZ_mdict_createDict(2222, 'B');
ZSTD_DDict* const raw = FUZ_mdict_createDict(0, 'R');
ZSTD_DDict* const d = FUZ_mdict_createDict(3333, 'D');
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(a, b);
BYTE frame[32], out[32];
size_t n, r;
CHECK_Z(ZSTD_DCtx_refDDict(dctx, raw));
n = FUZ_mdict_makeFrame(frame, 1111, mode == 3 ? FUZ_mdict_knownSize : 0);
if (mode != 0) {
r = mode == 1 ? ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n)
: FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, n,
mode == 2 ? 1 : sizeof(frame));
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
}
/* All registered DDicts can be freed immediately, including the most
* recently registered one and any DDict already used for decoding. */
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 0));
CHECK_Z(ZSTD_freeDDict(a)); CHECK_Z(ZSTD_freeDDict(b));
CHECK_Z(ZSTD_freeDDict(raw));
n = FUZ_mdict_makeFrame(frame, 0, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_corruption_detected);
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 1));
n = FUZ_mdict_makeFrame(frame, 1111, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
CHECK_Z(ZSTD_DCtx_refDDict(dctx, d));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
n = FUZ_mdict_makeFrame(frame, 3333, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "DDDDDDDD");
CHECK_Z(ZSTD_freeDCtx(dctx)); CHECK_Z(ZSTD_freeDDict(d));
}
/* With no set, toggling the opt-in parameter does not discard a separately
* configured local dictionary, prefix, or single borrowed DDict. */
for (mode = 0; mode < 3; ++mode) {
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
BYTE dict[44], frame[32], out[32];
size_t n, r;
CHECK(dctx != NULL);
FUZ_mdict_makeDict(dict, 1111, 'A');
if (mode == 0) { CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, dict, sizeof(dict))); }
if (mode == 1) { CHECK_Z(ZSTD_DCtx_refPrefix_advanced(dctx, dict, sizeof(dict), ZSTD_dct_fullDict)); }
if (mode == 2) { CHECK_Z(ZSTD_DCtx_refDDict(dctx, a)); }
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 1));
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 0));
n = FUZ_mdict_makeFrame(frame, 1111, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_freeDCtx(dctx)); CHECK_Z(ZSTD_freeDDict(a));
}
}
static void FUZ_mdict_testDictionaryReplacement(void)
{
int mode, used;
for (mode = 0; mode < 5; ++mode) {
for (used = 0; used < 2; ++used) {
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DDict* const b = FUZ_mdict_createDict(2222, 'B');
ZSTD_DDict* const d = FUZ_mdict_createDict(4444, 'D');
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(a, b);
BYTE* const dict = (BYTE*)malloc(44);
BYTE frame[32], out[32];
size_t n, r;
CHECK(dict != NULL);
FUZ_mdict_makeDict(dict, 3333, 'C');
switch (mode) {
case 0: CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, dict, 44)); break;
case 1: CHECK_Z(ZSTD_DCtx_loadDictionary_byReference(dctx, dict, 44)); break;
case 2: CHECK_Z(ZSTD_DCtx_loadDictionary_advanced(dctx, dict + 28, 16,
ZSTD_dlm_byCopy, ZSTD_dct_rawContent)); break;
case 3: CHECK_Z(ZSTD_DCtx_refPrefix(dctx, dict + 28, 16)); break;
default: CHECK_Z(ZSTD_DCtx_refPrefix_advanced(dctx, dict, 44, ZSTD_dct_fullDict)); break;
}
/* Every local dictionary/prefix replacement releases all borrowed
* DDicts, even though multiple-DDict mode remains enabled. */
CHECK_Z(ZSTD_freeDDict(a)); CHECK_Z(ZSTD_freeDDict(b));
if (used) {
n = FUZ_mdict_makeFrame(frame, 0, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "CCCCCCCC");
}
/* Going back to the set must discard both an unused and an already
* consumed prefix, as well as a sticky local dictionary. */
CHECK_Z(ZSTD_DCtx_refDDict(dctx, d));
free(dict);
n = FUZ_mdict_makeFrame(frame, 1111, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_dictionary_wrong);
n = FUZ_mdict_makeFrame(frame, 4444, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "DDDDDDDD");
CHECK_Z(ZSTD_freeDCtx(dctx)); CHECK_Z(ZSTD_freeDDict(d));
}
}
}
static void FUZ_mdict_testStreaming(void)
{
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DDict* const b = FUZ_mdict_createDict(7, 'B');
BYTE data[64], out[32];
static const struct { unsigned flags; size_t chunk; } cases[] = {
{0, 1}, /* fragmented/buffered */
{FUZ_mdict_knownSize, sizeof(data)}, /* single-pass shortcut */
{FUZ_mdict_knownSize | FUZ_mdict_checksum, sizeof(data)}
};
size_t i, n, r;
/* Skippable frames do not require a dictionary, including an ID-0 entry.
* The following ordinary frame must still select its exact match. */
for (i = 0; i < COUNTOF(cases); ++i) {
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(b, a);
n = FUZ_mdict_makeSkip(data, 7);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), data, n, cases[i].chunk);
CHECK_EQ(r, 0);
n = FUZ_mdict_makeFrame(data, 1111, cases[i].flags);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), data, n, cases[i].chunk);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_freeDCtx(dctx));
}
/* Ordinary frames select an exact match on both buffered and shortcut paths. */
for (i = 0; i < COUNTOF(cases); ++i) {
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(a, b);
n = FUZ_mdict_makeFrame(data, 1111, cases[i].flags);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), data, n, cases[i].chunk);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_freeDCtx(dctx));
}
/* Incomplete and rejected headers must not trigger an ID-0 lookup.
* This set deliberately has no ID-0 entry. */
for (i = 0; i < 2; ++i) {
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(a, b);
ZSTD_inBuffer in = {data, 5, 0};
ZSTD_outBuffer ob = {out, sizeof(out), 0};
FUZ_mdict_makeFrame(data, 1111, 0);
if (i) memset(data, 0xff, 5);
r = ZSTD_decompressStream(dctx, &ob, &in);
if (i) { CHECK_EQ(ZSTD_getErrorCode(r), ZSTD_error_prefix_unknown); }
else { CHECK_Z(r); CHECK(r != 0); CHECK_EQ(in.pos, in.size); }
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
n = FUZ_mdict_makeFrame(data, 1111, 0);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), data, n, 1);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
CHECK_Z(ZSTD_freeDCtx(dctx));
}
CHECK_Z(ZSTD_freeDDict(a)); CHECK_Z(ZSTD_freeDDict(b));
}
static void FUZ_mdict_testStrictMatching(void)
{
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DDict* const b = FUZ_mdict_createDict(2222, 'B');
ZSTD_DDict* const raw = FUZ_mdict_createDict(0, 'R');
static const U32 missingIDs[] = {0, 3333};
BYTE frame[32], out[32];
size_t i, n, r;
int mode;
for (mode = 0; mode < 4; ++mode) {
int const magicless = mode >> 1;
int const shortcut = mode & 1;
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(a, b);
unsigned const flags = shortcut ? FUZ_mdict_knownSize : 0;
size_t const chunk = shortcut ? sizeof(frame) : 1;
size_t const skip = magicless ? 4 : 0;
if (magicless) { CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless)); }
/* Neither a missing nonzero ID nor an omitted ID may fall back to B. */
for (i = 0; i < COUNTOF(missingIDs); ++i) {
ZSTD_ErrorCode const expected = missingIDs[i] == 0
? ZSTD_error_corruption_detected : ZSTD_error_dictionary_wrong;
n = FUZ_mdict_makeFrame(frame, 2222, flags);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame + skip, n - skip);
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
n = FUZ_mdict_makeFrame(frame, missingIDs[i], flags);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame + skip, n - skip);
CHECK_EQ(ZSTD_getErrorCode(r), expected);
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
n = FUZ_mdict_makeFrame(frame, 2222, flags);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame + skip, n - skip, chunk);
FUZ_mdict_checkOutput(r, out, "BBBBBBBB");
n = FUZ_mdict_makeFrame(frame, missingIDs[i], flags);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame + skip, n - skip, chunk);
CHECK_EQ(ZSTD_getErrorCode(r), expected);
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
}
/* ID 0 is a real key: registering raw R makes an ID-less frame use R,
* even when a different DDict was most recently referenced or used. */
CHECK_Z(ZSTD_DCtx_refDDict(dctx, raw));
CHECK_Z(ZSTD_DCtx_refDDict(dctx, a));
n = FUZ_mdict_makeFrame(frame, 0, flags);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame + skip, n - skip);
FUZ_mdict_checkOutput(r, out, "RRRRRRRR");
CHECK_Z(ZSTD_DCtx_refDDict(dctx, a));
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame + skip, n - skip, chunk);
FUZ_mdict_checkOutput(r, out, "RRRRRRRR");
n = FUZ_mdict_makeFrame(frame, 1111, flags);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame + skip, n - skip, chunk);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
n = FUZ_mdict_makeFrame(frame, 0, flags);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame + skip, n - skip, chunk);
FUZ_mdict_checkOutput(r, out, "RRRRRRRR");
CHECK_Z(ZSTD_freeDCtx(dctx));
}
/* Ordinary frames compressed without a dictionary work with or without
* an explicitly registered empty ID-0 DDict. */
{ ZSTD_DDict* const empty = ZSTD_createDDict(NULL, 0);
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(a, NULL);
BYTE plain[128];
CHECK(empty != NULL && ZSTD_getDictID_fromDDict(empty) == 0);
n = ZSTD_compress(plain, sizeof(plain), "no dictionary", 13, 1);
CHECK_Z(n);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), plain, n);
FUZ_mdict_checkOutput(r, out, "no dictionary");
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), plain, n, 1);
FUZ_mdict_checkOutput(r, out, "no dictionary");
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
CHECK_Z(ZSTD_DCtx_refDDict(dctx, empty));
CHECK_Z(ZSTD_DCtx_refDDict(dctx, a));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), plain, n);
FUZ_mdict_checkOutput(r, out, "no dictionary");
CHECK_Z(ZSTD_DCtx_refDDict(dctx, a));
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), plain, n, 1);
FUZ_mdict_checkOutput(r, out, "no dictionary");
CHECK_Z(ZSTD_freeDCtx(dctx)); CHECK_Z(ZSTD_freeDDict(empty));
}
CHECK_Z(ZSTD_freeDDict(a)); CHECK_Z(ZSTD_freeDDict(b));
CHECK_Z(ZSTD_freeDDict(raw));
}
static void FUZ_mdict_testEmptyDDict(void)
{
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
BYTE dummy = 0, plain[128], emptyFrame[32], frame[32], out[32];
ZSTD_DDict* const empty[] = {
ZSTD_createDDict(NULL, 0),
ZSTD_createDDict_byReference(NULL, 0),
ZSTD_createDDict(&dummy, 0),
ZSTD_createDDict_byReference(&dummy, 0)
};
size_t const plainSize = ZSTD_compress(plain, sizeof(plain), "no dictionary", 13, 1);
size_t const emptySize = ZSTD_compress(emptyFrame, sizeof(emptyFrame), NULL, 0, 1);
size_t i;
CHECK_Z(plainSize);
CHECK_Z(emptySize);
for (i = 0; i < COUNTOF(empty); ++i) {
ZSTD_DCtx* dctx;
size_t n, r;
CHECK(empty[i] != NULL && ZSTD_getDictID_fromDDict(empty[i]) == 0);
dctx = FUZ_mdict_createContext(empty[i], a);
/* Installing an empty DDict must replace the previous frame's history,
* without doing arithmetic on a possibly NULL dictionary pointer. */
n = FUZ_mdict_makeFrame(frame, 1111, 0);
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), frame, n);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA");
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), plain, plainSize);
FUZ_mdict_checkOutput(r, out, "no dictionary");
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), plain, plainSize, 1);
FUZ_mdict_checkOutput(r, out, "no dictionary");
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), plain, plainSize, plainSize);
FUZ_mdict_checkOutput(r, out, "no dictionary");
/* The explicit and bufferless APIs must accept the same DDict. */
r = ZSTD_decompress_usingDDict(dctx, out, sizeof(out), plain, plainSize, empty[i]);
FUZ_mdict_checkOutput(r, out, "no dictionary");
CHECK_Z(ZSTD_decompressBegin_usingDDict(dctx, empty[i]));
r = FUZ_mdict_decompressContinue(dctx, out, sizeof(out), plain, plainSize);
FUZ_mdict_checkOutput(r, out, "no dictionary");
/* Empty frames also permit a NULL, zero-capacity output buffer. */
CHECK_EQ(ZSTD_decompressDCtx(dctx, NULL, 0, emptyFrame, emptySize), 0);
CHECK_EQ(FUZ_mdict_decompressStream(dctx, NULL, 0, emptyFrame, emptySize, 1), 0);
CHECK_EQ(FUZ_mdict_decompressStream(dctx, NULL, 0, emptyFrame, emptySize, emptySize), 0);
CHECK_EQ(ZSTD_decompress_usingDDict(dctx, NULL, 0, emptyFrame, emptySize, empty[i]), 0);
CHECK_Z(ZSTD_decompressBegin_usingDDict(dctx, empty[i]));
CHECK_EQ(FUZ_mdict_decompressContinue(dctx, NULL, 0, emptyFrame, emptySize), 0);
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 0));
CHECK_Z(ZSTD_DCtx_refDDict(dctx, empty[i]));
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), plain, plainSize);
FUZ_mdict_checkOutput(r, out, "no dictionary");
CHECK_Z(ZSTD_freeDCtx(dctx));
CHECK_Z(ZSTD_freeDDict(empty[i]));
}
CHECK_Z(ZSTD_freeDDict(a));
}
static void FUZ_mdict_testNoDictionary(void)
{
ZSTD_DDict* const a = FUZ_mdict_createDict(1111, 'A');
ZSTD_DDict* const b = FUZ_mdict_createDict(2222, 'B');
BYTE plain[128], frame[32], data[256], out[128];
size_t const plainSize = ZSTD_compress(plain, sizeof(plain), "no dictionary", 13, 1);
int mode;
CHECK_Z(plainSize);
CHECK_EQ(ZSTD_getDictID_fromFrame(plain, plainSize), 0);
CHECK_EQ(ZSTD_getFrameContentSize(plain, plainSize), 13);
for (mode = 0; mode < 4; ++mode) {
int const magicless = mode >> 1;
int const shortcut = mode & 1;
ZSTD_DCtx* const dctx = FUZ_mdict_createContext(a, b);
unsigned const flags = shortcut ? FUZ_mdict_knownSize : 0;
size_t const chunk = shortcut ? sizeof(data) : 1;
size_t const skip = magicless ? 4 : 0;
static const U32 ids[] = {2222, 1111};
size_t i, n = 0, r;
if (magicless) { CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless)); }
/* Dictionary-free frames neither inherit the preceding dictionary nor
* prevent the following frame from selecting its own exact match. */
for (i = 0; i < COUNTOF(ids); ++i) {
size_t const frameSize = FUZ_mdict_makeFrame(frame, ids[i], flags);
memcpy(data + n, frame + skip, frameSize - skip); n += frameSize - skip;
memcpy(data + n, plain + skip, plainSize - skip); n += plainSize - skip;
}
r = ZSTD_decompressDCtx(dctx, out, sizeof(out), data, n);
FUZ_mdict_checkOutput(r, out, "BBBBBBBBno dictionaryAAAAAAAAno dictionary");
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), data, n, chunk);
FUZ_mdict_checkOutput(r, out, "BBBBBBBBno dictionaryAAAAAAAAno dictionary");
CHECK_Z(ZSTD_freeDCtx(dctx));
}
CHECK_Z(ZSTD_freeDDict(a)); CHECK_Z(ZSTD_freeDDict(b));
}
/* The DDicts use the default allocator. Counting the DCtx's first 128-slot
* table lets the test observe growth without inspecting private structs. */
static void* FUZ_mdict_alloc(void* opaque, size_t size)
{
if (size == 128 * sizeof(ZSTD_DDict*)) ++*(unsigned*)opaque;
return malloc(size);
}
static void FUZ_mdict_free(void* opaque, void* ptr)
{
(void)opaque;
free(ptr);
}
static void FUZ_mdict_testGrowth(void)
{
unsigned expansions = 0;
ZSTD_customMem const mem = {FUZ_mdict_alloc, FUZ_mdict_free, &expansions};
ZSTD_DCtx* const dctx = ZSTD_createDCtx_advanced(mem);
ZSTD_DDict* ddicts[49];
BYTE frame[32], out[32];
size_t i, n, r;
CHECK(dctx != NULL);
expansions = 0; /* Do not count allocation of the DCtx itself. */
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, 1));
for (i = 0; i < COUNTOF(ddicts); ++i) {
ddicts[i] = FUZ_mdict_createDict(10000 + (U32)i, 'A' + (int)(i % 26));
CHECK_Z(ZSTD_DCtx_refDDict(dctx, ddicts[i]));
/* A 64-slot table grows when inserting entry 49, after reaching 75%.
* The old expression incorrectly grew it when inserting entry 17. */
CHECK_EQ(expansions, (unsigned)(i == 48));
}
n = FUZ_mdict_makeFrame(frame, 10000, 0);
r = FUZ_mdict_decompressStream(dctx, out, sizeof(out), frame, n, 1);
FUZ_mdict_checkOutput(r, out, "AAAAAAAA"); /* Rehashing retained entry 1. */
CHECK_Z(ZSTD_freeDCtx(dctx));
for (i = 0; i < COUNTOF(ddicts); ++i) CHECK_Z(ZSTD_freeDDict(ddicts[i]));
}
static unsigned test_multipleDDicts(unsigned testNb)
{
static const struct { const char* name; void (*run)(void); } tests[] = {
{"one-shot frame selection", FUZ_mdict_testOneShot},
{"explicit dictionary APIs", FUZ_mdict_testExplicitDictionary},
{"single-DDict prefix compatibility", FUZ_mdict_testPrefixCompatibility},
{"raw dictionary ID zero", FUZ_mdict_testZeroID},
{"clearing dictionary references", FUZ_mdict_testClearReferences},
{"disabling the dictionary set", FUZ_mdict_testDisableSet},
{"local dictionary replacement", FUZ_mdict_testDictionaryReplacement},
{"streaming frame selection", FUZ_mdict_testStreaming},
{"strict dictionary-ID matching", FUZ_mdict_testStrictMatching},
{"empty dictionaries", FUZ_mdict_testEmptyDDict},
{"dictionary-free frames", FUZ_mdict_testNoDictionary},
{"hash-table growth", FUZ_mdict_testGrowth},
};
size_t i;
for (i = 0; i < COUNTOF(tests); ++i) {
DISPLAYLEVEL(3, "test%3u : multi-DDict %s : ", testNb++, tests[i].name);
tests[i].run();
DISPLAYLEVEL(3, "OK \n");
}
return testNb;
}
static void FUZ_checkEmptyDDict(const ZSTD_DDict* ddict)
{
static const BYTE frame[] = {
0x28, 0xB5, 0x2F, 0xFD, 0x20, 4, 0x21, 0, 0, 'T', 'E', 'S', 'T'
};
BYTE dst[4];
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
CHECK(ddict != NULL && dctx != NULL);
CHECK(ZSTD_DDict_dictContent(ddict) == NULL);
CHECK_EQ(ZSTD_DDict_dictSize(ddict), 0);
CHECK_Z(ZSTD_decompressBegin_usingDDict(dctx, ddict));
CHECK(dctx->prefixStart == NULL && dctx->virtualStart == NULL);
CHECK(dctx->dictEnd == NULL && dctx->previousDstEnd == NULL);
ZSTD_copyDDictParameters(dctx, ddict);
CHECK(dctx->dictEnd == NULL && dctx->previousDstEnd == NULL);
CHECK_EQ(ZSTD_decompress_usingDDict(dctx, dst, sizeof(dst), frame, sizeof(frame), ddict), 4);
CHECK(!memcmp(dst, "TEST", 4));
CHECK_Z(ZSTD_DCtx_refDDict(dctx, ddict));
{ ZSTD_inBuffer in = { frame, sizeof(frame), 0 };
ZSTD_outBuffer out = { dst, sizeof(dst), 0 };
CHECK_EQ(ZSTD_decompressStream(dctx, &out, &in), 0);
CHECK_EQ(in.pos, sizeof(frame));
CHECK_EQ(out.pos, sizeof(dst));
CHECK(!memcmp(dst, "TEST", 4));
}
ZSTD_freeDCtx(dctx);
}
static void FUZ_checkHufBoundaries(HUF_DTable* dtable, HUF_CElt const* ctable)
{
BYTE src[512];
BYTE compressed[1024];
BYTE decoded[sizeof(src)];
size_t i;
int const bmi2 = ZSTD_cpuSupportsBmi2() ? HUF_flags_bmi2 : 0;
int const flags[] = { 0, HUF_flags_disableAsm, HUF_flags_disableFast };
unsigned flagNb;
for (i = 0; i < sizeof(src); ++i) src[i] = (BYTE)(i & 1);
for (i = 1; i <= 17; ++i) {
size_t const cSize = HUF_compress1X_usingCTable(compressed, sizeof(compressed), src, i, ctable, 0);
CHECK_Z(cSize);
CHECK(cSize != 0);
CHECK_EQ(HUF_decompress1X_usingDTable(decoded, i, compressed, cSize, dtable, bmi2), i);
CHECK(!memcmp(src, decoded, i));
}
/* The two-symbol table is expanded to the fast decoder's table log.
* Each of these four streams is at least eight compressed bytes long. */
{ size_t const cSize = HUF_compress4X_usingCTable(compressed, sizeof(compressed), src, sizeof(src), ctable, 0);
CHECK_Z(cSize);
CHECK(cSize != 0);
for (flagNb = 0; flagNb < COUNTOF(flags); ++flagNb) {
CHECK_EQ(HUF_decompress4X_usingDTable(decoded, sizeof(decoded), compressed, cSize, dtable, bmi2 | flags[flagNb]), sizeof(decoded));
CHECK(!memcmp(src, decoded, sizeof(src)));
}
}
}
static void test_decoderPointerBoundaries(unsigned tnb)
{
static const BYTE emptyFrame[] = {
0x28, 0xB5, 0x2F, 0xFD, 0, 0, 3, 0, 0, 'A'
};
ZSTD_dictLoadMethod_e const methods[] = { ZSTD_dlm_byCopy, ZSTD_dlm_byRef };
unsigned methodNb;
unsigned nullInput;
DISPLAYLEVEL(3, "test%3u : empty dictionaries and decoder pointer boundaries : ", tnb);
for (methodNb = 0; methodNb < COUNTOF(methods); ++methodNb) {
for (nullInput = 0; nullInput < 2; ++nullInput) {
BYTE* const input = (BYTE*)malloc(16);
ZSTD_DDict* ddict;
const ZSTD_DDict* staticDDict;
size_t const workspaceSize = ZSTD_estimateDDictSize(0, methods[methodNb]);
void* const workspace = malloc(workspaceSize);
CHECK(input != NULL && workspace != NULL);
memset(input, 0xA5, 16);
ddict = ZSTD_createDDict_advanced(nullInput ? NULL : input, 0,
methods[methodNb], ZSTD_dct_auto, ZSTD_defaultCMem);
staticDDict = ZSTD_initStaticDDict(workspace, workspaceSize, nullInput ? NULL : input, 0,
methods[methodNb], ZSTD_dct_auto);
free(input);
FUZ_checkEmptyDDict(ddict);
FUZ_checkEmptyDDict(staticDDict);
ZSTD_freeDDict(ddict);
free(workspace);
}
/* Do not change nonempty raw-dictionary copying/reference semantics. */
{ BYTE input[16];
ZSTD_DDict* ddict;
memset(input, 0x5A, sizeof(input));
ddict = ZSTD_createDDict_advanced(input, sizeof(input), methods[methodNb],
ZSTD_dct_rawContent, ZSTD_defaultCMem);
CHECK(ddict != NULL);
CHECK_EQ(ZSTD_DDict_dictSize(ddict), sizeof(input));
CHECK(!memcmp(ZSTD_DDict_dictContent(ddict), input, sizeof(input)));
CHECK((ZSTD_DDict_dictContent(ddict) == input) == (methods[methodNb] == ZSTD_dlm_byRef));
ZSTD_freeDDict(ddict);
}
}
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx();
size_t pos = 0;
CHECK(dctx != NULL);
CHECK_Z(ZSTD_decompressBegin(dctx));
while (ZSTD_nextSrcSizeToDecompress(dctx) != 0) {
size_t const next = ZSTD_nextSrcSizeToDecompress(dctx);
CHECK(next <= sizeof(emptyFrame) - pos);
CHECK_EQ(ZSTD_decompressContinue(dctx, NULL, 0, emptyFrame + pos, next), 0);
pos += next;
}
CHECK_EQ(pos, sizeof(emptyFrame));
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters));
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_stableOutBuffer, 1));
{ ZSTD_inBuffer in1 = { emptyFrame, 6, 0 };
ZSTD_inBuffer in2 = { emptyFrame + 6, sizeof(emptyFrame) - 6, 0 };
ZSTD_outBuffer out = { NULL, 0, 0 };
CHECK_Z(ZSTD_decompressStream(dctx, &out, &in1));
CHECK_EQ(ZSTD_decompressStream(dctx, &out, &in2), 0);
CHECK_EQ(out.pos, 0);
}
ZSTD_freeDCtx(dctx);
}
{ BYTE input[2 * sizeof(BitContainerType)];
size_t distance;
U32 nbBytes;
memset(input, 0xA5, sizeof(input));
for (distance = 1; distance < sizeof(BitContainerType); ++distance) {
for (nbBytes = 0; nbBytes <= sizeof(BitContainerType); ++nbBytes) {
BIT_DStream_t bit;
size_t const consumed = MIN(distance, (size_t)nbBytes);
CHECK_Z(BIT_initDStream(&bit, input, sizeof(input)));
bit.ptr = bit.start + distance;
bit.bitsConsumed = nbBytes * 8;
CHECK_EQ(BIT_reloadDStream(&bit), nbBytes > distance ? BIT_DStream_endOfBuffer : BIT_DStream_unfinished);
CHECK(bit.ptr == bit.start + distance - consumed);
CHECK_EQ(bit.bitsConsumed, (nbBytes - consumed) * 8);
}
}
}
{ static const BYTE header[] = { 128, 0x10 };
HUF_CREATE_STATIC_CTABLE(ctable, HUF_SYMBOLVALUE_MAX);
U32 workspace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX;
unsigned hasZeroWeights = 0;
CHECK_EQ(HUF_readCTable(ctable, &maxSymbolValue, header, sizeof(header), &hasZeroWeights), sizeof(header));
#ifndef HUF_FORCE_DECOMPRESS_X2
{ HUF_CREATE_STATIC_DTABLEX1(dtable, HUF_TABLELOG_MAX);
CHECK_EQ(HUF_readDTableX1_wksp(dtable, header, sizeof(header), workspace, sizeof(workspace), 0), sizeof(header));
FUZ_checkHufBoundaries(dtable, ctable);
}
#endif
#ifndef HUF_FORCE_DECOMPRESS_X1
{ HUF_CREATE_STATIC_DTABLEX2(dtable, HUF_TABLELOG_MAX);
CHECK_EQ(HUF_readDTableX2_wksp(dtable, header, sizeof(header), workspace, sizeof(workspace), 0), sizeof(header));
FUZ_checkHufBoundaries(dtable, ctable);
}
#endif
}
DISPLAYLEVEL(3, "OK \n");
}
static int basicUnitTests(U32 const seed, double compressibility)
{
size_t const CNBuffSize = 5 MB;
void* const CNBuffer = malloc(CNBuffSize);
size_t const compressedBufferSize = ZSTD_compressBound(CNBuffSize);
void* const compressedBuffer = malloc(compressedBufferSize);
void* const decodedBuffer = malloc(CNBuffSize);
int testResult = 0;
unsigned testNb=0;
size_t cSize;
/* Create compressible noise */
if (!CNBuffer || !compressedBuffer || !decodedBuffer) {
DISPLAY("Not enough memory, aborting\n");
testResult = 1;
goto _end;
}
RDG_genBuffer(CNBuffer, CNBuffSize, compressibility, 0., seed);
/* Basic tests */
DISPLAYLEVEL(3, "test%3u : ZSTD_getErrorName : ", testNb++);
{ const char* errorString = ZSTD_getErrorName(0);
DISPLAYLEVEL(3, "OK : %s \n", errorString);
}
DISPLAYLEVEL(3, "test%3u : ZSTD_getErrorName with wrong value : ", testNb++);
{ const char* errorString = ZSTD_getErrorName(499);
DISPLAYLEVEL(3, "OK : %s \n", errorString);
}
DISPLAYLEVEL(3, "test%3u : min compression level : ", testNb++);
{ int const mcl = ZSTD_minCLevel();
DISPLAYLEVEL(3, "%i (OK) \n", mcl);
}
DISPLAYLEVEL(3, "test%3u : default compression level : ", testNb++);
{ int const defaultCLevel = ZSTD_defaultCLevel();
if (defaultCLevel != ZSTD_CLEVEL_DEFAULT) goto _output_error;
DISPLAYLEVEL(3, "%i (OK) \n", defaultCLevel);
}
DISPLAYLEVEL(3, "test%3u : ZSTD_versionNumber : ", testNb++);
{ unsigned const vn = ZSTD_versionNumber();
DISPLAYLEVEL(3, "%u (OK) \n", vn);
}
test_compressBound(testNb++);
test_decompressBound(testNb++);
test_decompressBoundOverflow(testNb++);
test_decompressHuffmanTable(testNb++);
test_decompressDDictHeader(testNb++);
test_decoderPointerBoundaries(testNb++);
test_decompressSplitLitBuffer(testNb++);
test_decompressContinueWithoutBegin(testNb++);
test_decompressZeroByteHistory(testNb++);
test_decompressStreamSplitHeader(testNb++);
test_setCParams(testNb++);
test_optimalParserDictEntropy(testNb++);
DISPLAYLEVEL(3, "test%3u : ZSTD_adjustCParams : ", testNb++);
{
ZSTD_compressionParameters params;
memset(&params, 0, sizeof(params));
params.windowLog = 10;
params.hashLog = 19;
params.chainLog = 19;
params = ZSTD_adjustCParams(params, 1000, 100000);
if (params.hashLog != 18) goto _output_error;
if (params.chainLog != 17) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
testNb = test_histCountWksp(seed, testNb);
test_histCountOverflow(testNb++);
DISPLAYLEVEL(3, "test%3u : compress %u bytes : ", testNb++, (unsigned)CNBuffSize);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
if (cctx==NULL) goto _output_error;
CHECK_VAR(cSize, ZSTD_compressCCtx(cctx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, 1) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : size of cctx for level 1 : ", testNb++);
{ size_t const cctxSize = ZSTD_sizeof_CCtx(cctx);
DISPLAYLEVEL(3, "%u bytes \n", (unsigned)cctxSize);
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "test%3i : decompress skippable frame -8 size : ", testNb++);
{
char const skippable8[] = "\x50\x2a\x4d\x18\xf8\xff\xff\xff";
size_t const size = ZSTD_decompress(NULL, 0, skippable8, 8);
if (!ZSTD_isError(size)) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_getFrameContentSize test : ", testNb++);
{ unsigned long long const rSize = ZSTD_getFrameContentSize(compressedBuffer, cSize);
if (rSize != CNBuffSize) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_getDecompressedSize test : ", testNb++);
{ unsigned long long const rSize = ZSTD_getDecompressedSize(compressedBuffer, cSize);
if (rSize != CNBuffSize) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_findDecompressedSize test : ", testNb++);
{ unsigned long long const rSize = ZSTD_findDecompressedSize(compressedBuffer, cSize);
if (rSize != CNBuffSize) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : tight ZSTD_decompressBound test : ", testNb++);
{
unsigned long long bound = ZSTD_decompressBound(compressedBuffer, cSize);
if (bound != CNBuffSize) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_decompressBound test with invalid srcSize : ", testNb++);
{
unsigned long long bound = ZSTD_decompressBound(compressedBuffer, cSize - 1);
if (bound != ZSTD_CONTENTSIZE_ERROR) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress %u bytes : ", testNb++, (unsigned)CNBuffSize);
{ size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (r != CNBuffSize) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress %u bytes with Huffman assembly disabled : ", testNb++, (unsigned)CNBuffSize);
{
ZSTD_DCtx* dctx = ZSTD_createDCtx();
size_t r;
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_disableHuffmanAssembly, 1));
r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (r != CNBuffSize || memcmp(decodedBuffer, CNBuffer, CNBuffSize)) goto _output_error;
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : check decompressed result : ", testNb++);
{ size_t u;
for (u=0; u<CNBuffSize; u++) {
if (((BYTE*)decodedBuffer)[u] != ((BYTE*)CNBuffer)[u]) goto _output_error;
} }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3u : invalid endDirective : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_inBuffer inb = { CNBuffer, CNBuffSize, 0 };
ZSTD_outBuffer outb = { compressedBuffer, compressedBufferSize, 0 };
if (cctx==NULL) goto _output_error;
CHECK( ZSTD_isError( ZSTD_compressStream2(cctx, &outb, &inb, (ZSTD_EndDirective) 3) ) ); /* must fail */
CHECK( ZSTD_isError( ZSTD_compressStream2(cctx, &outb, &inb, (ZSTD_EndDirective)-1) ) ); /* must fail */
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_checkCParams : ", testNb++);
{
ZSTD_parameters params = ZSTD_getParams(3, 0, 0);
assert(!ZSTD_checkCParams(params.cParams));
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_createDCtx_advanced and ZSTD_sizeof_DCtx: ", testNb++);
{
ZSTD_DCtx* const dctx = ZSTD_createDCtx_advanced(ZSTD_defaultCMem);
assert(dctx != NULL);
assert(ZSTD_sizeof_DCtx(dctx) != 0);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : misc unaccounted for zstd symbols : ", testNb++);
{
/* %p takes a void*. In ISO C, it's illegal to cast a function pointer
* to a data pointer. (Although in POSIX you're required to be allowed
* to do it...) So we have to fall back to our trusty friend memcpy. */
unsigned (* const funcptr_getDictID)(const ZSTD_DDict* ddict) =
ZSTD_getDictID_fromDDict;
ZSTD_DStream* (* const funcptr_createDStream)(
ZSTD_customMem customMem) = ZSTD_createDStream_advanced;
ZSTD_nextInputType_e (* const funcptr_nextInputType)(ZSTD_DCtx* dctx) =
ZSTD_nextInputType;
const void *voidptr_getDictID;
const void *voidptr_createDStream;
const void *voidptr_nextInputType;
DEBUG_STATIC_ASSERT(sizeof(funcptr_getDictID) == sizeof(voidptr_getDictID));
memcpy(
(void*)&voidptr_getDictID,
(const void*)&funcptr_getDictID,
sizeof(void*));
memcpy(
(void*)&voidptr_createDStream,
(const void*)&funcptr_createDStream,
sizeof(void*));
memcpy(
(void*)&voidptr_nextInputType,
(const void*)&funcptr_nextInputType,
sizeof(void*));
DISPLAYLEVEL(3, "%p ", voidptr_getDictID);
DISPLAYLEVEL(3, "%p ", voidptr_createDStream);
DISPLAYLEVEL(3, "%p ", voidptr_nextInputType);
}
DISPLAYLEVEL(3, ": OK \n");
DISPLAYLEVEL(3, "test%3i : decompress with null dict : ", testNb++);
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx(); assert(dctx != NULL);
{ size_t const r = ZSTD_decompress_usingDict(dctx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize,
NULL, 0);
if (r != CNBuffSize) goto _output_error;
}
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress with null DDict : ", testNb++);
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx(); assert(dctx != NULL);
{ size_t const r = ZSTD_decompress_usingDDict(dctx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize,
NULL);
if (r != CNBuffSize) goto _output_error;
}
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress with 1 missing byte : ", testNb++);
{ size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize-1);
if (!ZSTD_isError(r)) goto _output_error;
if (ZSTD_getErrorCode((size_t)r) != ZSTD_error_srcSize_wrong) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress with 1 too much byte : ", testNb++);
{ size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize+1);
if (!ZSTD_isError(r)) goto _output_error;
if (ZSTD_getErrorCode(r) != ZSTD_error_srcSize_wrong) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress too large input : ", testNb++);
{ size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, compressedBufferSize);
if (!ZSTD_isError(r)) goto _output_error;
if (ZSTD_getErrorCode(r) != ZSTD_error_srcSize_wrong) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress into NULL buffer : ", testNb++);
{ size_t const r = ZSTD_decompress(NULL, 0, compressedBuffer, compressedBufferSize);
if (!ZSTD_isError(r)) goto _output_error;
if (ZSTD_getErrorCode(r) != ZSTD_error_dstSize_tooSmall) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress with corrupted checksum : ", testNb++);
{ /* create compressed buffer with checksumming enabled */
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
if (!cctx) {
DISPLAY("Not enough memory, aborting\n");
testResult = 1;
goto _end;
}
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1) );
CHECK_VAR(cSize, ZSTD_compress2(cctx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize) );
ZSTD_freeCCtx(cctx);
}
{ /* copy the compressed buffer and corrupt the checksum */
size_t r;
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
if (!dctx) {
DISPLAY("Not enough memory, aborting\n");
testResult = 1;
goto _end;
}
((char*)compressedBuffer)[cSize-1] += 1;
r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (!ZSTD_isError(r)) goto _output_error;
if (ZSTD_getErrorCode(r) != ZSTD_error_checksum_wrong) goto _output_error;
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_forceIgnoreChecksum, ZSTD_d_ignoreChecksum));
r = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize-1);
if (!ZSTD_isError(r)) goto _output_error; /* wrong checksum size should still throw error */
r = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (ZSTD_isError(r)) goto _output_error;
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_decompressBound test with content size missing : ", testNb++);
{ /* create compressed buffer with content size missing */
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_contentSizeFlag, 0) );
CHECK_VAR(cSize, ZSTD_compress2(cctx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize) );
ZSTD_freeCCtx(cctx);
}
{ /* ensure frame content size is missing */
ZSTD_FrameHeader zfh;
size_t const ret = ZSTD_getFrameHeader(&zfh, compressedBuffer, compressedBufferSize);
if (ret != 0 || zfh.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) goto _output_error;
}
{ /* ensure CNBuffSize <= decompressBound */
unsigned long long const bound = ZSTD_decompressBound(compressedBuffer, compressedBufferSize);
if (CNBuffSize > bound) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d: check DCtx size is reduced after many oversized calls : ", testNb++);
{
size_t const largeFrameSrcSize = 200;
size_t const smallFrameSrcSize = 10;
size_t const nbFrames = 256;
size_t i = 0, consumed = 0, produced = 0, prevDCtxSize = 0;
int sizeReduced = 0;
BYTE* const dst = (BYTE*)compressedBuffer;
ZSTD_DCtx* dctx = ZSTD_createDCtx();
/* create a large frame and then a bunch of small frames */
size_t srcSize = ZSTD_compress((void*)dst,
compressedBufferSize, CNBuffer, largeFrameSrcSize, 3);
for (i = 0; i < nbFrames; i++)
srcSize += ZSTD_compress((void*)(dst + srcSize),
compressedBufferSize - srcSize, CNBuffer,
smallFrameSrcSize, 3);
/* decompressStream and make sure that dctx size was reduced at least once */
while (consumed < srcSize) {
ZSTD_inBuffer in = {(void*)(dst + consumed), MIN(1, srcSize - consumed), 0};
ZSTD_outBuffer out = {(BYTE*)CNBuffer + produced, CNBuffSize - produced, 0};
ZSTD_decompressStream(dctx, &out, &in);
consumed += in.pos;
produced += out.pos;
/* success! size was reduced from the previous frame */
if (prevDCtxSize > ZSTD_sizeof_DCtx(dctx))
sizeReduced = 1;
prevDCtxSize = ZSTD_sizeof_DCtx(dctx);
}
assert(sizeReduced);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
{
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_CDict* const cdict = ZSTD_createCDict(CNBuffer, 100, 1);
ZSTD_parameters const params = ZSTD_getParams(1, 0, 0);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_format, ZSTD_f_zstd1_magicless) );
DISPLAYLEVEL(3, "test%3i : ZSTD_compressCCtx() doesn't use advanced parameters", testNb++);
CHECK_Z(ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, NULL, 0, 1));
if (MEM_readLE32(compressedBuffer) != ZSTD_MAGICNUMBER) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_compress_usingDict() doesn't use advanced parameters: ", testNb++);
CHECK_Z(ZSTD_compress_usingDict(cctx, compressedBuffer, compressedBufferSize, NULL, 0, NULL, 0, 1));
if (MEM_readLE32(compressedBuffer) != ZSTD_MAGICNUMBER) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_compress_usingCDict() doesn't use advanced parameters: ", testNb++);
CHECK_Z(ZSTD_compress_usingCDict(cctx, compressedBuffer, compressedBufferSize, NULL, 0, cdict));
if (MEM_readLE32(compressedBuffer) != ZSTD_MAGICNUMBER) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_compress_advanced() doesn't use advanced parameters: ", testNb++);
CHECK_Z(ZSTD_compress_advanced(cctx, compressedBuffer, compressedBufferSize, NULL, 0, NULL, 0, params));
if (MEM_readLE32(compressedBuffer) != ZSTD_MAGICNUMBER) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_compress_usingCDict_advanced() doesn't use advanced parameters: ", testNb++);
CHECK_Z(ZSTD_compress_usingCDict_advanced(cctx, compressedBuffer, compressedBufferSize, NULL, 0, cdict, params.fParams));
if (MEM_readLE32(compressedBuffer) != ZSTD_MAGICNUMBER) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
ZSTD_freeCDict(cdict);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "test%3i : maxBlockSize = 2K : ", testNb++);
{
ZSTD_CCtx* cctx = ZSTD_createCCtx();
ZSTD_DCtx* dctx = ZSTD_createDCtx();
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_maxBlockSize, 2048));
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_maxBlockSize, 2048));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize);
CHECK_Z(ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize));
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_maxBlockSize, 1024));
CHECK(ZSTD_isError(ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize)));
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : maxBlockSize rejects raw & RLE blocks : ", testNb++);
{
ZSTD_CCtx* cctx = ZSTD_createCCtx();
ZSTD_DCtx* dctx = ZSTD_createDCtx();
size_t const srcSize = 8 KB;
BYTE* const src = (BYTE*)malloc(srcSize);
int i;
CHECK(src != NULL);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_maxBlockSize, 2048));
CHECK_Z(ZSTD_DCtx_setParameter(dctx, ZSTD_d_maxBlockSize, 1024));
/* i = 0 : incompressible data, which is stored in raw blocks
* i = 1 : a single repeated byte, which is stored in RLE blocks */
for (i = 0; i < 2; i++) {
if (i == 0) {
RDG_genBuffer(src, srcSize, 0.0, 0.0, seed);
} else {
memset(src, 'a', srcSize);
}
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, src, srcSize);
CHECK_Z(cSize);
CHECK(ZSTD_isError(ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize)));
}
free(src);
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ldm fill dict out-of-bounds check", testNb++);
{
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const size = (1U << 10);
size_t const dstCapacity = ZSTD_compressBound(size);
void* dict = (void*)malloc(size);
void* src = (void*)malloc(size);
void* dst = (void*)malloc(dstCapacity);
RDG_genBuffer(dict, size, 0.5, 0.5, seed);
RDG_genBuffer(src, size, 0.5, 0.5, seed);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable));
assert(!ZSTD_isError(ZSTD_compress_usingDict(cctx, dst, dstCapacity, src, size, dict, size, 3)));
ZSTD_freeCCtx(cctx);
free(dict);
free(src);
free(dst);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ldm hashRateLog > windowLog underflow check : ", testNb++);
{
/* Test that when windowLog < hashRateLog, we don't get excessive memory usage
* due to underflow in hashLog calculation (windowLog - hashRateLog). */
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const size = (1U << 10); // 1 KB
size_t const dstCapacity = ZSTD_compressBound(size);
void* src = (void*)malloc(size);
void* dst = (void*)malloc(dstCapacity);
RDG_genBuffer(src, size, 0.5, 0.5, seed);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_ldmHashLog, 0));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, 12));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_ldmHashRateLog, 13));
CHECK_Z(ZSTD_compress2(cctx, dst, dstCapacity, src, size));
{ size_t const cctxSize = ZSTD_sizeof_CCtx(cctx);
DISPLAYLEVEL(3, "CCtx size: %u bytes ", (unsigned)cctxSize);
CHECK_LT(cctxSize, 50 MB);
}
ZSTD_freeCCtx(cctx);
free(src);
free(dst);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : testing dict compression with enableLdm and forceMaxWindow : ", testNb++);
{
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
void* dict = (void*)malloc(CNBuffSize);
int nbWorkers;
for (nbWorkers = 0; nbWorkers < 3; ++nbWorkers) {
RDG_genBuffer(dict, CNBuffSize, 0.5, 0.5, seed);
RDG_genBuffer(CNBuffer, CNBuffSize, 0.6, 0.6, seed);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, nbWorkers));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_forceMaxWindow, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable));
CHECK_Z(ZSTD_CCtx_refPrefix(cctx, dict, CNBuffSize));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize);
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, dict, CNBuffSize));
}
ZSTD_freeCCtx(cctx);
ZSTD_freeDCtx(dctx);
free(dict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : testing dict compression for determinism : ", testNb++);
{
size_t const testSize = 1024;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
char* dict = (char*)malloc(2 * testSize);
int ldmEnabled, level;
RDG_genBuffer(dict, testSize, 0.5, 0.5, seed);
RDG_genBuffer(CNBuffer, testSize, 0.6, 0.6, seed);
memcpy(dict + testSize, CNBuffer, testSize);
for (level = 1; level <= 5; ++level) {
for (ldmEnabled = ZSTD_ps_enable; ldmEnabled <= ZSTD_ps_disable; ++ldmEnabled) {
size_t cSize0;
XXH64_hash_t compressedChecksum0;
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, level));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ldmEnabled));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_deterministicRefPrefix, 1));
CHECK_Z(ZSTD_CCtx_refPrefix(cctx, dict, testSize));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, testSize);
CHECK_Z(cSize);
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, testSize, compressedBuffer, cSize, dict, testSize));
cSize0 = cSize;
compressedChecksum0 = XXH64(compressedBuffer, cSize, 0);
CHECK_Z(ZSTD_CCtx_refPrefix(cctx, dict, testSize));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, dict + testSize, testSize);
CHECK_Z(cSize);
if (cSize != cSize0) goto _output_error;
if (XXH64(compressedBuffer, cSize, 0) != compressedChecksum0) goto _output_error;
}
}
ZSTD_freeCCtx(cctx);
ZSTD_freeDCtx(dctx);
free(dict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : LDM + opt parser with small uncompressible block ", testNb++);
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
ZSTD_DCtx* dctx = ZSTD_createDCtx();
size_t const srcSize = 300 KB;
size_t const flushSize = 128 KB + 5;
size_t const dstSize = ZSTD_compressBound(srcSize);
char* src = (char*)CNBuffer;
char* dst = (char*)compressedBuffer;
ZSTD_outBuffer out = { dst, dstSize, 0 };
ZSTD_inBuffer in = { src, flushSize, 0 };
if (!cctx || !dctx) {
DISPLAY("Not enough memory, aborting\n");
testResult = 1;
goto _end;
}
RDG_genBuffer(src, srcSize, 0.5, 0.5, seed);
/* Force an LDM to exist that crosses block boundary into uncompressible block */
memcpy(src + 125 KB, src, 3 KB + 5);
/* Enable MT, LDM, and opt parser */
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 19));
/* Flushes a block of 128 KB and block of 5 bytes */
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush));
/* Compress the rest */
in.size = 300 KB;
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
CHECK_Z(ZSTD_decompress(decodedBuffer, CNBuffSize, dst, out.pos));
ZSTD_freeCCtx(cctx);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : testing ldm dictionary gets invalidated : ", testNb++);
{
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
void* dict = (void*)malloc(CNBuffSize);
size_t const kWindowLog = 10;
size_t const kWindowSize = (size_t)1 << kWindowLog;
size_t const dictSize = kWindowSize * 10;
size_t const srcSize1 = kWindowSize / 2;
size_t const srcSize2 = kWindowSize * 10;
CHECK(cctx!=NULL);
CHECK(dctx!=NULL);
CHECK(dict!=NULL);
if (CNBuffSize < dictSize) goto _output_error;
RDG_genBuffer(dict, dictSize, 0.5, 0.5, seed);
RDG_genBuffer(CNBuffer, srcSize1 + srcSize2, 0.5, 0.5, seed);
/* Enable checksum to verify round trip. */
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
/* Disable content size to skip single-pass decompression. */
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_contentSizeFlag, 0));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, (int)kWindowLog));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_ldmMinMatch, 32));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_ldmHashRateLog, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_ldmHashLog, 16));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_ldmBucketSizeLog, 3));
/* Round trip once with a dictionary. */
CHECK_Z(ZSTD_CCtx_refPrefix(cctx, dict, dictSize));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, srcSize1);
CHECK_Z(cSize);
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, dict, dictSize));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, srcSize2);
/* Streaming decompression to catch out of bounds offsets. */
{
ZSTD_inBuffer in = {compressedBuffer, cSize, 0};
ZSTD_outBuffer out = {decodedBuffer, CNBuffSize, 0};
size_t const dSize = ZSTD_decompressStream(dctx, &out, &in);
CHECK_Z(dSize);
if (dSize != 0) goto _output_error;
}
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, 2));
/* Round trip once with a dictionary. */
CHECK_Z(ZSTD_CCtx_refPrefix(cctx, dict, dictSize));
{ ZSTD_inBuffer in = {CNBuffer, srcSize1, 0};
ZSTD_outBuffer out = {compressedBuffer, compressedBufferSize, 0};
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
cSize = out.pos;
}
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, dict, dictSize));
{ ZSTD_inBuffer in = {CNBuffer, srcSize2, 0};
ZSTD_outBuffer out = {compressedBuffer, compressedBufferSize, 0};
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
cSize = out.pos;
}
/* Streaming decompression to catch out of bounds offsets. */
{ ZSTD_inBuffer in = {compressedBuffer, cSize, 0};
ZSTD_outBuffer out = {decodedBuffer, CNBuffSize, 0};
size_t const dSize = ZSTD_decompressStream(dctx, &out, &in);
CHECK_Z(dSize);
if (dSize != 0) goto _output_error;
}
ZSTD_freeCCtx(cctx);
ZSTD_freeDCtx(dctx);
free(dict);
}
DISPLAYLEVEL(3, "OK \n");
/* Note: this test takes 0.5 seconds to run */
DISPLAYLEVEL(3, "test%3i : testing refPrefx vs refPrefx + ldm (size comparison) : ", testNb++);
{
/* test a big buffer so that ldm can take effect */
size_t const size = 100 MB;
int const windowLog = 27;
size_t const dstSize = ZSTD_compressBound(size);
void* dict = (void*)malloc(size);
void* src = (void*)malloc(size);
void* dst = (void*)malloc(dstSize);
void* recon = (void*)malloc(size);
size_t refPrefixCompressedSize = 0;
size_t refPrefixLdmCompressedSize = 0;
size_t reconSize = 0;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
/* make dict and src the same uncompressible data */
RDG_genBuffer(src, size, 0, 0, seed);
memcpy(dict, src, size);
assert(!memcmp(dict, src, size));
/* set level 1 and windowLog to cover src */
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, windowLog));
/* compress on level 1 using just refPrefix and no ldm */
ZSTD_CCtx_refPrefix(cctx, dict, size);
refPrefixCompressedSize = ZSTD_compress2(cctx, dst, dstSize, src, size);
assert(!ZSTD_isError(refPrefixCompressedSize));
/* test round trip just refPrefix */
ZSTD_DCtx_refPrefix(dctx, dict, size);
reconSize = ZSTD_decompressDCtx(dctx, recon, size, dst, refPrefixCompressedSize);
assert(!ZSTD_isError(reconSize));
assert(reconSize == size);
assert(!memcmp(recon, src, size));
/* compress on level 1 using refPrefix and ldm */
ZSTD_CCtx_refPrefix(cctx, dict, size);;
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable))
refPrefixLdmCompressedSize = ZSTD_compress2(cctx, dst, dstSize, src, size);
assert(!ZSTD_isError(refPrefixLdmCompressedSize));
/* test round trip refPrefix + ldm*/
ZSTD_DCtx_refPrefix(dctx, dict, size);
reconSize = ZSTD_decompressDCtx(dctx, recon, size, dst, refPrefixLdmCompressedSize);
assert(!ZSTD_isError(reconSize));
assert(reconSize == size);
assert(!memcmp(recon, src, size));
/* make sure that refPrefixCompressedSize is significantly greater */
assert(refPrefixCompressedSize > 10 * refPrefixLdmCompressedSize);
/* make sure the ldm compressed size is less than 1% of original */
assert((double)refPrefixLdmCompressedSize / (double)size < 0.01);
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtx(cctx);
free(recon);
free(dict);
free(src);
free(dst);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : in-place decompression : ", testNb++);
cSize = ZSTD_compress(compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize, -ZSTD_BLOCKSIZE_MAX);
CHECK_Z(cSize);
CHECK_LT(CNBuffSize, cSize);
{
size_t const margin = ZSTD_decompressionMargin(compressedBuffer, cSize);
size_t const outputSize = (CNBuffSize + margin);
char* output = malloc(outputSize);
char* input = output + outputSize - cSize;
CHECK_LT(cSize, CNBuffSize + margin);
CHECK(output != NULL);
CHECK_Z(margin);
CHECK(margin <= ZSTD_DECOMPRESSION_MARGIN(CNBuffSize, ZSTD_BLOCKSIZE_MAX));
memcpy(input, compressedBuffer, cSize);
{
size_t const dSize = ZSTD_decompress(output, outputSize, input, cSize);
CHECK_Z(dSize);
CHECK_EQ(dSize, CNBuffSize);
}
CHECK(!memcmp(output, CNBuffer, CNBuffSize));
free(output);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : in-place decompression with 2 frames : ", testNb++);
cSize = ZSTD_compress(compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize / 3, -ZSTD_BLOCKSIZE_MAX);
CHECK_Z(cSize);
{
size_t const cSize2 = ZSTD_compress((char*)compressedBuffer + cSize, compressedBufferSize - cSize, (char const*)CNBuffer + (CNBuffSize / 3), CNBuffSize / 3, -ZSTD_BLOCKSIZE_MAX);
CHECK_Z(cSize2);
cSize += cSize2;
}
{
size_t const srcSize = (CNBuffSize / 3) * 2;
size_t const margin = ZSTD_decompressionMargin(compressedBuffer, cSize);
size_t const outputSize = (CNBuffSize + margin);
char* output = malloc(outputSize);
char* input = output + outputSize - cSize;
CHECK_LT(cSize, CNBuffSize + margin);
CHECK(output != NULL);
CHECK_Z(margin);
memcpy(input, compressedBuffer, cSize);
{
size_t const dSize = ZSTD_decompress(output, outputSize, input, cSize);
CHECK_Z(dSize);
CHECK_EQ(dSize, srcSize);
}
CHECK(!memcmp(output, CNBuffer, srcSize));
free(output);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Check block splitter with 64K literal length : ", testNb++);
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
size_t const srcSize = 256 * 1024;
U32 const compressibleLenU32 = 32 * 1024 / 4;
U32 const blockSizeU32 = 128 * 1024 / 4;
U32 const litLenU32 = 64 * 1024 / 4;
U32* data = (U32*)malloc(srcSize);
size_t dSize;
if (data == NULL || cctx == NULL) goto _output_error;
/* Generate data without any matches */
RDG_genBuffer(data, srcSize, 0.0, 0.01, 2654435761U);
/* Generate 32K of compressible data */
RDG_genBuffer(data, compressibleLenU32 * 4, 0.5, 0.5, 0xcafebabe);
/* Add a match of offset=12, length=8 at idx=16, 32, 48, 64 */
data[compressibleLenU32 + 0] = 0xFFFFFFFF;
data[compressibleLenU32 + 1] = 0xEEEEEEEE;
data[compressibleLenU32 + 4] = 0xFFFFFFFF;
data[compressibleLenU32 + 5] = 0xEEEEEEEE;
/* Add a match of offset=16, length=8 at idx=64K + 64.
* This generates a sequence with llen=64K, and repeat code 1.
* The block splitter thought this was ll0, and corrupted the
* repeat offset history.
*/
data[compressibleLenU32 + litLenU32 + 2 + 0] = 0xDDDDDDDD;
data[compressibleLenU32 + litLenU32 + 2 + 1] = 0xCCCCCCCC;
data[compressibleLenU32 + litLenU32 + 2 + 4] = 0xDDDDDDDD;
data[compressibleLenU32 + litLenU32 + 2 + 5] = 0xCCCCCCCC;
/* Add a match of offset=16, length=8 at idx=128K + 16.
* This should generate a sequence with repeat code = 1.
* But the block splitters mistake caused zstd to generate
* repeat code = 2, corrupting the data.
*/
data[blockSizeU32] = 0xBBBBBBBB;
data[blockSizeU32 + 1] = 0xAAAAAAAA;
data[blockSizeU32 + 4] = 0xBBBBBBBB;
data[blockSizeU32 + 5] = 0xAAAAAAAA;
/* Generate a golden file from this data in case datagen changes and
* doesn't generate the exact same data. We will also test this golden file.
*/
if (0) {
FILE* f = fopen("golden-compression/PR-3517-block-splitter-corruption-test", "wb");
fwrite(data, 1, srcSize, f);
fclose(f);
}
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 19));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_minMatch, 7));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_splitAfterSequences, ZSTD_ps_enable));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, data, srcSize);
CHECK_Z(cSize);
dSize = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize);
CHECK_Z(dSize);
CHECK_EQ(dSize, srcSize);
CHECK(!memcmp(decodedBuffer, data, srcSize));
free(data);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
test_blockSplitter_incompressibleExpansionProtection(testNb++, seed);
test_targetCBlockSize_dstCapacity(testNb++);
DISPLAYLEVEL(3, "test%3d : superblock uncompressible data: too many nocompress superblocks : ", testNb++);
{
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
const BYTE* src = (BYTE*)CNBuffer; BYTE* dst = (BYTE*)compressedBuffer;
size_t srcSize = 321656; size_t dstCapacity = ZSTD_compressBound(srcSize);
/* This is the number of bytes to stream before ending. This value
* was obtained by trial and error :/. */
const size_t streamCompressThreshold = 161792;
const size_t streamCompressDelta = 1024;
/* The first 1/5 of the buffer is compressible and the last 4/5 is
* uncompressible. This is an approximation of the type of data
* the fuzzer generated to catch this bug. Streams like this were making
* zstd generate noCompress superblocks (which are larger than the src
* they come from). Do this enough times, and we'll run out of room
* and throw a dstSize_tooSmall error. */
const size_t compressiblePartSize = srcSize/5;
const size_t uncompressiblePartSize = srcSize-compressiblePartSize;
RDG_genBuffer(CNBuffer, compressiblePartSize, 0.5, 0.5, seed);
RDG_genBuffer((BYTE*)CNBuffer+compressiblePartSize, uncompressiblePartSize, 0, 0, seed);
/* Setting target block size so that superblock is used */
assert(cctx != NULL);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_targetCBlockSize, 81);
{ size_t read;
for (read = 0; read < streamCompressThreshold; read += streamCompressDelta) {
ZSTD_inBuffer in = {src, streamCompressDelta, 0};
ZSTD_outBuffer out = {dst, dstCapacity, 0};
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
src += streamCompressDelta; srcSize -= streamCompressDelta;
dst += out.pos; dstCapacity -= out.pos;
} }
/* This is trying to catch a dstSize_tooSmall error */
{ ZSTD_inBuffer in = {src, srcSize, 0};
ZSTD_outBuffer out = {dst, dstCapacity, 0};
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d: superblock with no literals : ", testNb++);
/* Generate the same data 20 times over */
{ size_t const avgChunkSize = CNBuffSize / 20;
size_t b;
for (b = 0; b < CNBuffSize; b += avgChunkSize) {
size_t const chunkSize = MIN(CNBuffSize - b, avgChunkSize);
RDG_genBuffer((char*)CNBuffer + b, chunkSize, compressibility, 0. /* auto */, seed);
} }
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const normalCSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
size_t const allowedExpansion = (CNBuffSize * 3 / 1000);
size_t superCSize;
CHECK_Z(normalCSize);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 19);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_targetCBlockSize, 1000);
superCSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(superCSize);
if (superCSize > normalCSize + allowedExpansion) {
DISPLAYLEVEL(1, "Superblock too big: %u > %u + %u \n", (U32)superCSize, (U32)normalCSize, (U32)allowedExpansion);
goto _output_error;
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
RDG_genBuffer(CNBuffer, CNBuffSize, compressibility, 0. /*auto*/, seed);
DISPLAYLEVEL(3, "test%3d: superblock enough room for checksum : ", testNb++)
/* This tests whether or not we leave enough room for the checksum at the end
* of the dst buffer. The bug that motivated this test was found by the
* stream_round_trip fuzzer but this crashes for the same reason and is
* far more compact than re-creating the stream_round_trip fuzzer's code path */
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_CCtx_setParameter(cctx, ZSTD_c_targetCBlockSize, 64);
assert(!ZSTD_isError(ZSTD_compress2(cctx, compressedBuffer, 1339, CNBuffer, 1278)));
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress a NULL input with each level : ", testNb++);
{ int level = -1;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
if (!cctx) goto _output_error;
for (level = -1; level <= ZSTD_maxCLevel(); ++level) {
CHECK_Z( ZSTD_compress(compressedBuffer, compressedBufferSize, NULL, 0, level) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, level) );
CHECK_Z( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, NULL, 0) );
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : check CCtx size after compressing empty input : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const r = ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, NULL, 0, 19);
if (ZSTD_isError(r)) goto _output_error;
if (ZSTD_sizeof_CCtx(cctx) > (1U << 20)) goto _output_error;
ZSTD_freeCCtx(cctx);
cSize = r;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : decompress empty frame into NULL : ", testNb++);
{ size_t const r = ZSTD_decompress(NULL, 0, compressedBuffer, cSize);
if (ZSTD_isError(r)) goto _output_error;
if (r != 0) goto _output_error;
}
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_outBuffer output;
if (cctx==NULL) goto _output_error;
output.dst = compressedBuffer;
output.size = compressedBufferSize;
output.pos = 0;
CHECK_Z( ZSTD_initCStream(cctx, 1) ); /* content size unknown */
CHECK_Z( ZSTD_flushStream(cctx, &output) ); /* ensure no possibility to "concatenate" and determine the content size */
CHECK_Z( ZSTD_endStream(cctx, &output) );
ZSTD_freeCCtx(cctx);
/* single scan decompression */
{ size_t const r = ZSTD_decompress(NULL, 0, compressedBuffer, output.pos);
if (ZSTD_isError(r)) goto _output_error;
if (r != 0) goto _output_error;
}
/* streaming decompression */
{ ZSTD_DCtx* const dstream = ZSTD_createDStream();
ZSTD_inBuffer dinput;
ZSTD_outBuffer doutput;
size_t ipos;
if (dstream==NULL) goto _output_error;
dinput.src = compressedBuffer;
dinput.size = 0;
dinput.pos = 0;
doutput.dst = NULL;
doutput.size = 0;
doutput.pos = 0;
CHECK_Z ( ZSTD_initDStream(dstream) );
for (ipos=1; ipos<=output.pos; ipos++) {
dinput.size = ipos;
CHECK_Z ( ZSTD_decompressStream(dstream, &doutput, &dinput) );
}
if (doutput.pos != 0) goto _output_error;
ZSTD_freeDStream(dstream);
}
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : reuse CCtx with expanding block size : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_parameters const params = ZSTD_getParams(1, ZSTD_CONTENTSIZE_UNKNOWN, 0);
assert(params.fParams.contentSizeFlag == 1); /* block size will be adapted if pledgedSrcSize is enabled */
CHECK_Z( ZSTD_compressBegin_advanced(cctx, NULL, 0, params, 1 /*pledgedSrcSize*/) );
CHECK_Z( ZSTD_compressEnd(cctx, compressedBuffer, compressedBufferSize, CNBuffer, 1) ); /* creates a block size of 1 */
CHECK_Z( ZSTD_compressBegin_advanced(cctx, NULL, 0, params, ZSTD_CONTENTSIZE_UNKNOWN) ); /* reuse same parameters */
{ size_t const inSize = 2* 128 KB;
size_t const outSize = ZSTD_compressBound(inSize);
CHECK_Z( ZSTD_compressEnd(cctx, compressedBuffer, outSize, CNBuffer, inSize) );
/* will fail if blockSize is not resized */
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : re-using a CCtx should compress the same : ", testNb++);
{ size_t const sampleSize = 30;
int i;
for (i=0; i<20; i++)
((char*)CNBuffer)[i] = (char)i; /* ensure no match during initial section */
memcpy((char*)CNBuffer + 20, CNBuffer, 10); /* create one match, starting from beginning of sample, which is the difficult case (see #1241) */
for (i=1; i<=19; i++) {
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t size1, size2;
DISPLAYLEVEL(5, "l%i ", i);
size1 = ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, CNBuffer, sampleSize, i);
CHECK_Z(size1);
size2 = ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, CNBuffer, sampleSize, i);
CHECK_Z(size2);
CHECK_EQ(size1, size2);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, i) );
size2 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, sampleSize);
CHECK_Z(size2);
CHECK_EQ(size1, size2);
size2 = ZSTD_compress2(cctx, compressedBuffer, ZSTD_compressBound(sampleSize) - 1, CNBuffer, sampleSize); /* force streaming, as output buffer is not large enough to guarantee success */
CHECK_Z(size2);
CHECK_EQ(size1, size2);
{ ZSTD_inBuffer inb;
ZSTD_outBuffer outb;
inb.src = CNBuffer;
inb.pos = 0;
inb.size = sampleSize;
outb.dst = compressedBuffer;
outb.pos = 0;
outb.size = ZSTD_compressBound(sampleSize) - 1; /* force streaming, as output buffer is not large enough to guarantee success */
CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_end) );
assert(inb.pos == inb.size);
CHECK_EQ(size1, outb.pos);
}
ZSTD_freeCCtx(cctx);
}
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : btultra2 & 1st block : ", testNb++);
{ size_t const sampleSize = 1024;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_inBuffer inb;
ZSTD_outBuffer outb;
inb.src = CNBuffer;
inb.pos = 0;
inb.size = 0;
outb.dst = compressedBuffer;
outb.pos = 0;
outb.size = compressedBufferSize;
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, ZSTD_maxCLevel()) );
inb.size = sampleSize; /* start with something, so that context is already used */
CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_end) ); /* will break internal assert if stats_init is not disabled */
assert(inb.pos == inb.size);
outb.pos = 0; /* cancel output */
CHECK_Z( ZSTD_CCtx_setPledgedSrcSize(cctx, sampleSize) );
inb.size = 4; /* too small size : compression will be skipped */
inb.pos = 0;
CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_flush) );
assert(inb.pos == inb.size);
inb.size += 5; /* too small size : compression will be skipped */
CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_flush) );
assert(inb.pos == inb.size);
inb.size += 11; /* small enough to attempt compression */
CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_flush) );
assert(inb.pos == inb.size);
assert(inb.pos < sampleSize);
inb.size = sampleSize; /* large enough to trigger stats_init, but no longer at beginning */
CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_end) ); /* will break internal assert if stats_init is not disabled */
assert(inb.pos == inb.size);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : ZSTD_CCtx_getParameter() : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_outBuffer out = {NULL, 0, 0};
ZSTD_inBuffer in = {NULL, 0, 0};
int value;
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
CHECK_EQ(value, 3);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_hashLog, ZSTD_HASHLOG_MIN));
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
CHECK_EQ(value, 3);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, ZSTD_HASHLOG_MIN);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 7));
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
CHECK_EQ(value, 7);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, ZSTD_HASHLOG_MIN);
/* Start a compression job */
ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
CHECK_EQ(value, 7);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, ZSTD_HASHLOG_MIN);
/* Reset the CCtx */
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
CHECK_EQ(value, 7);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, ZSTD_HASHLOG_MIN);
/* Reset the parameters */
ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
CHECK_EQ(value, 3);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, 0);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : ZSTD_CCtx_setCParams() : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
int value;
ZSTD_compressionParameters cparams = ZSTD_getCParams(1, 0, 0);
cparams.strategy = (ZSTD_strategy)-1; /* set invalid value, on purpose */
/* Set invalid cParams == error out, and no change. */
CHECK(ZSTD_isError(ZSTD_CCtx_setCParams(cctx, cparams)));
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_windowLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_chainLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_searchLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_minMatch, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_targetLength, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_strategy, &value));
CHECK_EQ(value, 0);
cparams = ZSTD_getCParams(12, 0, 0);
CHECK_Z(ZSTD_CCtx_setCParams(cctx, cparams));
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_windowLog, &value));
CHECK_EQ(value, (int)cparams.windowLog);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_chainLog, &value));
CHECK_EQ(value, (int)cparams.chainLog);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, (int)cparams.hashLog);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_searchLog, &value));
CHECK_EQ(value, (int)cparams.searchLog);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_minMatch, &value));
CHECK_EQ(value, (int)cparams.minMatch);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_targetLength, &value));
CHECK_EQ(value, (int)cparams.targetLength);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_strategy, &value));
CHECK_EQ(value, (int)cparams.strategy);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "test%3d : ZSTD_CCtx_setFParams() : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
int value;
ZSTD_frameParameters fparams = {0, 1, 1};
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_contentSizeFlag, &value));
CHECK_EQ(value, 1);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_checksumFlag, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_dictIDFlag, &value));
CHECK_EQ(value, 1);
CHECK_Z(ZSTD_CCtx_setFParams(cctx, fparams));
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_contentSizeFlag, &value));
CHECK_EQ(value, fparams.contentSizeFlag);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_checksumFlag, &value));
CHECK_EQ(value, fparams.checksumFlag);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_dictIDFlag, &value));
CHECK_EQ(value, !fparams.noDictIDFlag);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "test%3d : ZSTD_CCtx_setParams() : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
int value;
ZSTD_parameters params = ZSTD_getParams(1, 0, 0);
params.cParams.strategy = (ZSTD_strategy)-1; /* set invalid value, on purpose */
/* Set invalid params == error out, and no change. */
CHECK(ZSTD_isError(ZSTD_CCtx_setParams(cctx, params)));
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_windowLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_chainLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_searchLog, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_minMatch, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_targetLength, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_strategy, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_contentSizeFlag, &value));
CHECK_EQ(value, 1);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_checksumFlag, &value));
CHECK_EQ(value, 0);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_dictIDFlag, &value));
CHECK_EQ(value, 1);
params = ZSTD_getParams(12, 0, 0);
params.fParams.contentSizeFlag = 0;
params.fParams.checksumFlag = 1;
params.fParams.noDictIDFlag = 1;
CHECK_Z(ZSTD_CCtx_setParams(cctx, params));
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_windowLog, &value));
CHECK_EQ(value, (int)params.cParams.windowLog);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_chainLog, &value));
CHECK_EQ(value, (int)params.cParams.chainLog);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
CHECK_EQ(value, (int)params.cParams.hashLog);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_searchLog, &value));
CHECK_EQ(value, (int)params.cParams.searchLog);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_minMatch, &value));
CHECK_EQ(value, (int)params.cParams.minMatch);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_targetLength, &value));
CHECK_EQ(value, (int)params.cParams.targetLength);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_strategy, &value));
CHECK_EQ(value, (int)params.cParams.strategy);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_contentSizeFlag, &value));
CHECK_EQ(value, params.fParams.contentSizeFlag);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_checksumFlag, &value));
CHECK_EQ(value, params.fParams.checksumFlag);
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_dictIDFlag, &value));
CHECK_EQ(value, !params.fParams.noDictIDFlag);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "test%3d : ldm conditionally enabled by default doesn't change cctx params: ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_outBuffer out = {NULL, 0, 0};
ZSTD_inBuffer in = {NULL, 0, 0};
int value;
/* Even if LDM will be enabled by default in the applied params (since wlog >= 27 and strategy >= btopt),
* we should not modify the actual parameter specified by the user within the CCtx
*/
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, 27));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_strategy, ZSTD_btopt));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue));
CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_enableLongDistanceMatching, &value));
CHECK_EQ(value, 0);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
/* this test is really too long, and should be made faster */
DISPLAYLEVEL(3, "test%3d : overflow protection with large windowLog : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_parameters params = ZSTD_getParams(-999, ZSTD_CONTENTSIZE_UNKNOWN, 0);
size_t const nbCompressions = ((1U << 31) / CNBuffSize) + 2; /* ensure U32 overflow protection is triggered */
size_t cnb;
assert(cctx != NULL);
params.fParams.contentSizeFlag = 0;
params.cParams.windowLog = ZSTD_WINDOWLOG_MAX;
for (cnb = 0; cnb < nbCompressions; ++cnb) {
DISPLAYLEVEL(6, "run %u / %u \n", (unsigned)cnb, (unsigned)nbCompressions);
CHECK_Z( ZSTD_compressBegin_advanced(cctx, NULL, 0, params, ZSTD_CONTENTSIZE_UNKNOWN) ); /* reuse same parameters */
CHECK_Z( ZSTD_compressEnd(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize) );
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : size down context : ", testNb++);
{ ZSTD_CCtx* const largeCCtx = ZSTD_createCCtx();
assert(largeCCtx != NULL);
CHECK_Z( ZSTD_compressBegin(largeCCtx, 19) ); /* streaming implies ZSTD_CONTENTSIZE_UNKNOWN, which maximizes memory usage */
CHECK_Z( ZSTD_compressEnd(largeCCtx, compressedBuffer, compressedBufferSize, CNBuffer, 1) );
{ size_t const largeCCtxSize = ZSTD_sizeof_CCtx(largeCCtx); /* size of context must be measured after compression */
{ ZSTD_CCtx* const smallCCtx = ZSTD_createCCtx();
assert(smallCCtx != NULL);
CHECK_Z(ZSTD_compressCCtx(smallCCtx, compressedBuffer, compressedBufferSize, CNBuffer, 1, 1));
{ size_t const smallCCtxSize = ZSTD_sizeof_CCtx(smallCCtx);
DISPLAYLEVEL(5, "(large) %uKB > 32*%uKB (small) : ",
(unsigned)(largeCCtxSize>>10), (unsigned)(smallCCtxSize>>10));
assert(largeCCtxSize > 32* smallCCtxSize); /* note : "too large" definition is handled within zstd_compress.c .
* make this test case extreme, so that it doesn't depend on a possibly fluctuating definition */
}
ZSTD_freeCCtx(smallCCtx);
}
{ U32 const maxNbAttempts = 1100; /* nb of usages before triggering size down is handled within zstd_compress.c.
* currently defined as 128x, but could be adjusted in the future.
* make this test long enough so that it's not too much tied to the current definition within zstd_compress.c */
unsigned u;
for (u=0; u<maxNbAttempts; u++) {
CHECK_Z(ZSTD_compressCCtx(largeCCtx, compressedBuffer, compressedBufferSize, CNBuffer, 1, 1));
if (ZSTD_sizeof_CCtx(largeCCtx) < largeCCtxSize) break; /* sized down */
}
DISPLAYLEVEL(5, "size down after %u attempts : ", u);
if (u==maxNbAttempts) goto _output_error; /* no sizedown happened */
}
}
ZSTD_freeCCtx(largeCCtx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : static DCtx dictionary allocation : ", testNb++);
{
static const char dict[] = "static DCtx dictionary regression test";
size_t const dictSize = sizeof(dict);
size_t const dctxSize = ZSTD_estimateDCtxSize();
size_t const dstreamSize = ZSTD_estimateDStreamSize(1 KB);
size_t const ddictSize = ZSTD_estimateDDictSize(dictSize, ZSTD_dlm_byRef);
void* const workspace = malloc(dstreamSize);
void* const ddictWorkspace = malloc(ddictSize);
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
const ZSTD_DDict* ddict;
char compressed[ZSTD_COMPRESSBOUND(sizeof(dict))];
char decoded[sizeof(dict)];
size_t compressedSize;
int fill;
int stream;
CHECK(workspace != NULL && ddictWorkspace != NULL && cctx != NULL);
ddict = ZSTD_initStaticDDict(ddictWorkspace, ddictSize, dict, dictSize,
ZSTD_dlm_byRef, ZSTD_dct_rawContent);
CHECK(ddict != NULL);
CHECK_VAR(compressedSize, ZSTD_compress_usingDict(cctx, compressed, sizeof(compressed),
dict, dictSize, dict, dictSize, 1));
CHECK_Z(ZSTD_freeCCtx(cctx));
/* Both zeroed and reused workspaces must reject internal DDict
* creation without invoking an allocator. */
for (fill = 0; fill < 2; ++fill) {
for (stream = 0; stream < 2; ++stream) {
ZSTD_DCtx* dctx;
memset(workspace, fill ? 0xA5 : 0, dstreamSize);
dctx = stream ? ZSTD_initStaticDStream(workspace, dstreamSize)
: ZSTD_initStaticDCtx(workspace, dctxSize);
CHECK(dctx != NULL);
CHECK_EQ(ZSTD_DCtx_loadDictionary(dctx, dict, dictSize), ERROR(memory_allocation));
CHECK_EQ(ZSTD_DCtx_loadDictionary_byReference(dctx, dict, dictSize), ERROR(memory_allocation));
CHECK_EQ(ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_rawContent), ERROR(memory_allocation));
CHECK_EQ(ZSTD_DCtx_loadDictionary_advanced(dctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_rawContent), ERROR(memory_allocation));
CHECK_EQ(ZSTD_DCtx_refPrefix(dctx, dict, dictSize), ERROR(memory_allocation));
CHECK_EQ(ZSTD_DCtx_refPrefix_advanced(dctx, dict, dictSize, ZSTD_dct_auto), ERROR(memory_allocation));
CHECK_EQ(ZSTD_initDStream_usingDict(dctx, dict, dictSize), ERROR(memory_allocation));
CHECK_EQ(dctx->customMem.customAlloc, NULL);
CHECK_EQ(dctx->customMem.customFree, NULL);
CHECK_EQ(dctx->customMem.opaque, NULL);
/* Clearing a dictionary does not allocate. */
CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, NULL, dictSize));
CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, dict, 0));
CHECK_Z(ZSTD_DCtx_loadDictionary_byReference(dctx, NULL, dictSize));
CHECK_Z(ZSTD_DCtx_loadDictionary_byReference(dctx, dict, 0));
CHECK_Z(ZSTD_DCtx_refPrefix(dctx, NULL, dictSize));
CHECK_Z(ZSTD_DCtx_refPrefix(dctx, dict, 0));
CHECK_Z(ZSTD_initDStream_usingDict(dctx, NULL, dictSize));
CHECK_Z(ZSTD_initDStream_usingDict(dctx, dict, 0));
/* A caller-owned DDict remains usable without allocation. */
CHECK_Z(ZSTD_DCtx_refDDict(dctx, ddict));
CHECK_EQ(ZSTD_decompressDCtx(dctx, decoded, sizeof(decoded), compressed, compressedSize), dictSize);
CHECK_EQ(memcmp(decoded, dict, dictSize), 0);
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters));
if (stream) {
ZSTD_inBuffer input = { compressed, compressedSize, 0 };
ZSTD_outBuffer output = { decoded, sizeof(decoded), 0 };
CHECK_Z(ZSTD_initDStream_usingDDict(dctx, ddict));
CHECK_EQ(ZSTD_decompressStream(dctx, &output, &input), 0);
CHECK_EQ(input.pos, compressedSize);
CHECK_EQ(output.pos, dictSize);
CHECK_EQ(memcmp(decoded, dict, dictSize), 0);
}
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters));
}
}
free(ddictWorkspace);
free(workspace);
}
DISPLAYLEVEL(3, "OK \n");
/* Static CCtx tests */
#define STATIC_CCTX_LEVEL 4
DISPLAYLEVEL(3, "test%3i : create static CCtx for level %u : ", testNb++, STATIC_CCTX_LEVEL);
{ size_t const staticCStreamSize = ZSTD_estimateCStreamSize(STATIC_CCTX_LEVEL);
void* const staticCCtxBuffer = malloc(staticCStreamSize);
size_t const staticDCtxSize = ZSTD_estimateDCtxSize();
void* const staticDCtxBuffer = malloc(staticDCtxSize);
DISPLAYLEVEL(4, "CStream size = %u, ", (U32)staticCStreamSize);
if (staticCCtxBuffer==NULL || staticDCtxBuffer==NULL) {
free(staticCCtxBuffer);
free(staticDCtxBuffer);
DISPLAY("Not enough memory, aborting\n");
testResult = 1;
goto _end;
}
{ size_t const smallInSize = 32 KB;
ZSTD_compressionParameters const cparams_small = ZSTD_getCParams(STATIC_CCTX_LEVEL, smallInSize, 0);
size_t const smallCCtxSize = ZSTD_estimateCCtxSize_usingCParams(cparams_small);
size_t const staticCCtxSize = ZSTD_estimateCCtxSize(STATIC_CCTX_LEVEL);
ZSTD_CCtx* staticCCtx = ZSTD_initStaticCCtx(staticCCtxBuffer, smallCCtxSize);
ZSTD_DCtx* const staticDCtx = ZSTD_initStaticDCtx(staticDCtxBuffer, staticDCtxSize);
DISPLAYLEVEL(4, "Full CCtx size = %u, ", (U32)staticCCtxSize);
DISPLAYLEVEL(4, "CCtx for 32 KB = %u, ", (U32)smallCCtxSize);
if ((staticCCtx==NULL) || (staticDCtx==NULL)) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress small input with small static CCtx : ", testNb++);
CHECK_VAR(cSize, ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
CNBuffer, smallInSize, STATIC_CCTX_LEVEL) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n",
(unsigned)cSize, (double)cSize/smallInSize*100);
DISPLAYLEVEL(3, "test%3i : compress large input with small static CCtx (must fail) : ", testNb++);
{ size_t const r = ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, STATIC_CCTX_LEVEL);
if (ZSTD_getErrorCode((size_t)r) != ZSTD_error_memory_allocation) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : resize context to full CCtx size : ", testNb++);
staticCCtx = ZSTD_initStaticCStream(staticCCtxBuffer, staticCCtxSize);
DISPLAYLEVEL(4, "staticCCtxBuffer = %p, staticCCtx = %p , ", staticCCtxBuffer, (void*)staticCCtx);
if (staticCCtx == NULL) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress large input with static CCtx : ", testNb++);
CHECK_VAR(cSize, ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, STATIC_CCTX_LEVEL) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n",
(unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : compress small input often enough to trigger context reduce : ", testNb++);
{ int nbc;
assert(staticCCtxSize > smallCCtxSize * ZSTD_WORKSPACETOOLARGE_FACTOR); /* ensure size down scenario */
assert(CNBuffSize > smallInSize + ZSTD_WORKSPACETOOLARGE_MAXDURATION + 3);
for (nbc=0; nbc<ZSTD_WORKSPACETOOLARGE_MAXDURATION+2; nbc++) {
CHECK_Z(ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
(char*)CNBuffer + nbc, smallInSize,
STATIC_CCTX_LEVEL) );
} }
DISPLAYLEVEL(3, "OK \n")
DISPLAYLEVEL(3, "test%3i : init CCtx for level %u : ", testNb++, STATIC_CCTX_LEVEL);
CHECK_Z( ZSTD_compressBegin(staticCCtx, STATIC_CCTX_LEVEL) );
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compression again with static CCtx : ", testNb++);
CHECK_VAR(cSize, ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, STATIC_CCTX_LEVEL) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n",
(unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : simple decompression test with static DCtx : ", testNb++);
{ size_t const r = ZSTD_decompressDCtx(staticDCtx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize);
if (r != CNBuffSize) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : check decompressed result : ", testNb++);
if (memcmp(decodedBuffer, CNBuffer, CNBuffSize)) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : init CCtx for too large level (must fail) : ", testNb++);
{ size_t const r = ZSTD_compressBegin(staticCCtx, ZSTD_maxCLevel());
if (!ZSTD_isError(r)) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : init CCtx for small level %u (should work again) : ", testNb++, 1);
CHECK_Z( ZSTD_compressBegin(staticCCtx, 1) );
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : use CStream on CCtx-sized static context (should fail) : ", testNb++);
CHECK_Z( ZSTD_initCStream(staticCCtx, STATIC_CCTX_LEVEL) ); /* note : doesn't allocate */
{ ZSTD_outBuffer output = { compressedBuffer, compressedBufferSize, 0 };
ZSTD_inBuffer input = { CNBuffer, CNBuffSize, 0 };
size_t const r = ZSTD_compressStream(staticCCtx, &output, &input); /* now allocates, should fail */
if (!ZSTD_isError(r)) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : resize context to CStream size, then stream compress : ", testNb++);
staticCCtx = ZSTD_initStaticCStream(staticCCtxBuffer, staticCStreamSize);
assert(staticCCtx != NULL);
CHECK_Z( ZSTD_initCStream(staticCCtx, STATIC_CCTX_LEVEL) ); /* note : doesn't allocate */
{ ZSTD_outBuffer output = { compressedBuffer, compressedBufferSize, 0 };
ZSTD_inBuffer input = { CNBuffer, CNBuffSize, 0 };
CHECK_Z( ZSTD_compressStream(staticCCtx, &output, &input) );
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : CStream for small level %u : ", testNb++, 1);
CHECK_Z( ZSTD_initCStream(staticCCtx, 1) ); /* note : doesn't allocate */
{ ZSTD_outBuffer output = { compressedBuffer, compressedBufferSize, 0 };
ZSTD_inBuffer input = { CNBuffer, CNBuffSize, 0 };
CHECK_Z( ZSTD_compressStream(staticCCtx, &output, &input) );
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : init static CStream with dictionary (should fail) : ", testNb++);
{ size_t const r = ZSTD_initCStream_usingDict(staticCCtx, CNBuffer, 64 KB, 1);
if (!ZSTD_isError(r)) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : use DStream on DCtx-sized static context (should fail) : ", testNb++);
CHECK_Z( ZSTD_initDStream(staticDCtx) );
{ ZSTD_outBuffer output = { decodedBuffer, CNBuffSize, 0 };
ZSTD_inBuffer input = { compressedBuffer, ZSTD_FRAMEHEADERSIZE_MAX+1, 0 };
size_t const r = ZSTD_decompressStream(staticDCtx, &output, &input);
if (!ZSTD_isError(r)) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : test estimation functions with default cctx params : ", testNb++);
{
// Test ZSTD_estimateCCtxSize_usingCCtxParams
{
ZSTD_CCtx_params* params = ZSTD_createCCtxParams();
size_t const cctxSizeDefault = ZSTD_estimateCCtxSize_usingCCtxParams(params);
staticCCtx = ZSTD_initStaticCCtx(staticCCtxBuffer, cctxSizeDefault);
CHECK_VAR(cSize, ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, 3));
{
size_t const r = ZSTD_decompressDCtx(staticDCtx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize);
if (r != CNBuffSize) goto _output_error;
if (memcmp(decodedBuffer, CNBuffer, CNBuffSize)) goto _output_error;
}
ZSTD_freeCCtxParams(params);
}
// Test ZSTD_estimateCStreamSize_usingCCtxParams
{
ZSTD_CCtx_params* params = ZSTD_createCCtxParams();
size_t const cctxSizeDefault = ZSTD_estimateCStreamSize_usingCCtxParams(params);
staticCCtx = ZSTD_initStaticCCtx(staticCCtxBuffer, cctxSizeDefault);
CHECK_VAR(cSize, ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, 3) );
{
size_t const r = ZSTD_decompressDCtx(staticDCtx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize);
if (r != CNBuffSize) goto _output_error;
if (memcmp(decodedBuffer, CNBuffer, CNBuffSize)) goto _output_error;
}
ZSTD_freeCCtxParams(params);
}
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : test estimation functions with maxBlockSize = 0 : ", testNb++);
{
// Test ZSTD_estimateCCtxSize_usingCCtxParams
{
ZSTD_CCtx_params* params = ZSTD_createCCtxParams();
size_t cctxSizeDefault;
CHECK_Z(ZSTD_CCtxParams_setParameter(params, ZSTD_c_maxBlockSize, 0));
cctxSizeDefault = ZSTD_estimateCCtxSize_usingCCtxParams(params);
staticCCtx = ZSTD_initStaticCCtx(staticCCtxBuffer, cctxSizeDefault);
CHECK_VAR(cSize, ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, 3) );
{
size_t const r = ZSTD_decompressDCtx(staticDCtx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize);
if (r != CNBuffSize) goto _output_error;
if (memcmp(decodedBuffer, CNBuffer, CNBuffSize)) goto _output_error;
}
ZSTD_freeCCtxParams(params);
}
// Test ZSTD_estimateCStreamSize_usingCCtxParams
{
ZSTD_CCtx_params* params = ZSTD_createCCtxParams();
size_t cctxSizeDefault;
CHECK_Z(ZSTD_CCtxParams_setParameter(params, ZSTD_c_maxBlockSize, 0));
cctxSizeDefault = ZSTD_estimateCStreamSize_usingCCtxParams(params);
staticCCtx = ZSTD_initStaticCCtx(staticCCtxBuffer, cctxSizeDefault);
CHECK_VAR(cSize, ZSTD_compressCCtx(staticCCtx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, 3) );
{
size_t const r = ZSTD_decompressDCtx(staticDCtx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize);
if (r != CNBuffSize) goto _output_error;
if (memcmp(decodedBuffer, CNBuffer, CNBuffSize)) goto _output_error;
}
ZSTD_freeCCtxParams(params);
}
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : estimation functions with LDM enabled (issue #4590) : ", testNb++);
{
/* ZSTD_estimateCCtxSize_usingCCtxParams must adjust zeroed-out
* LDM parameters when LDM is enabled, to avoid division by zero
* in ZSTD_ldm_getMaxNbSeq. */
ZSTD_CCtx_params* params = ZSTD_createCCtxParams();
size_t cctxSize;
CHECK_Z(ZSTD_CCtxParams_setParameter(params, ZSTD_c_compressionLevel, 22));
CHECK_Z(ZSTD_CCtxParams_setParameter(params, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable));
cctxSize = ZSTD_estimateCCtxSize_usingCCtxParams(params);
if (ZSTD_isError(cctxSize)) goto _output_error;
if (cctxSize == 0) goto _output_error;
cctxSize = ZSTD_estimateCStreamSize_usingCCtxParams(params);
if (ZSTD_isError(cctxSize)) goto _output_error;
if (cctxSize == 0) goto _output_error;
ZSTD_freeCCtxParams(params);
}
DISPLAYLEVEL(3, "OK \n");
}
free(staticCCtxBuffer);
free(staticDCtxBuffer);
}
DISPLAYLEVEL(3, "test%3i : Static context sizes for negative levels : ", testNb++);
{ size_t const cctxSizeN1 = ZSTD_estimateCCtxSize(-1);
size_t const cctxSizeP1 = ZSTD_estimateCCtxSize(1);
size_t const cstreamSizeN1 = ZSTD_estimateCStreamSize(-1);
size_t const cstreamSizeP1 = ZSTD_estimateCStreamSize(1);
if (!(0 < cctxSizeN1 && cctxSizeN1 <= cctxSizeP1)) goto _output_error;
if (!(0 < cstreamSizeN1 && cstreamSizeN1 <= cstreamSizeP1)) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
/* ZSTDMT simple MT compression test */
DISPLAYLEVEL(3, "test%3i : create ZSTDMT CCtx : ", testNb++);
{ ZSTD_CCtx* const mtctx = ZSTD_createCCtx();
if (mtctx==NULL) {
DISPLAY("mtctx : not enough memory, aborting \n");
testResult = 1;
goto _end;
}
CHECK_Z( ZSTD_CCtx_setParameter(mtctx, ZSTD_c_nbWorkers, 2) );
CHECK_Z( ZSTD_CCtx_setParameter(mtctx, ZSTD_c_compressionLevel, 1) );
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3u : compress %u bytes with 2 threads : ", testNb++, (unsigned)CNBuffSize);
CHECK_VAR(cSize, ZSTD_compress2(mtctx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : decompressed size test : ", testNb++);
{ unsigned long long const rSize = ZSTD_getFrameContentSize(compressedBuffer, cSize);
if (rSize != CNBuffSize) {
DISPLAY("ZSTD_getFrameContentSize incorrect : %u != %u \n", (unsigned)rSize, (unsigned)CNBuffSize);
goto _output_error;
} }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress %u bytes : ", testNb++, (unsigned)CNBuffSize);
{ size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (r != CNBuffSize) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : check decompressed result : ", testNb++);
{ size_t u;
for (u=0; u<CNBuffSize; u++) {
if (((BYTE*)decodedBuffer)[u] != ((BYTE*)CNBuffer)[u]) goto _output_error;
} }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress -T2 with checksum : ", testNb++);
CHECK_Z( ZSTD_CCtx_setParameter(mtctx, ZSTD_c_checksumFlag, 1) );
CHECK_Z( ZSTD_CCtx_setParameter(mtctx, ZSTD_c_contentSizeFlag, 1) );
CHECK_Z( ZSTD_CCtx_setParameter(mtctx, ZSTD_c_overlapLog, 3) );
CHECK_VAR(cSize, ZSTD_compress2(mtctx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : decompress %u bytes : ", testNb++, (unsigned)CNBuffSize);
{ size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (r != CNBuffSize) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
ZSTD_freeCCtx(mtctx);
}
DISPLAYLEVEL(3, "test%3u : compress empty string and decompress with small window log : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
char out[32];
if (cctx == NULL || dctx == NULL) goto _output_error;
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_contentSizeFlag, 0) );
CHECK_VAR(cSize, ZSTD_compress2(cctx, out, sizeof(out), NULL, 0) );
DISPLAYLEVEL(3, "OK (%u bytes)\n", (unsigned)cSize);
CHECK_Z( ZSTD_DCtx_setParameter(dctx, ZSTD_d_windowLogMax, 10) );
{ char const* outPtr = out;
ZSTD_inBuffer inBuffer = { outPtr, cSize, 0 };
ZSTD_outBuffer outBuffer = { NULL, 0, 0 };
size_t dSize;
CHECK_VAR(dSize, ZSTD_decompressStream(dctx, &outBuffer, &inBuffer) );
if (dSize != 0) goto _output_error;
}
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "test%3i : compress with block splitting : ", testNb++)
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_splitAfterSequences, ZSTD_ps_enable) );
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress -T2 with/without literals compression : ", testNb++)
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
size_t cSize1, cSize2;
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 1) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, 2) );
cSize1 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize1);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_literalCompressionMode, ZSTD_ps_disable) );
cSize2 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize2);
CHECK_LT(cSize1, cSize2);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Multithreaded ZSTD_compress2() with rsyncable : ", testNb++)
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
/* Set rsyncable and don't give the ZSTD_compressBound(CNBuffSize) so
* ZSTDMT is forced to not take the shortcut.
*/
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 1) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, 1) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_rsyncable, 1) );
CHECK_Z( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize - 1, CNBuffer, CNBuffSize) );
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : setting multithreaded parameters : ", testNb++)
{ ZSTD_CCtx_params* params = ZSTD_createCCtxParams();
int const jobSize = 512 KB;
int value;
/* Check that the overlap log and job size are unset. */
CHECK_Z( ZSTD_CCtxParams_getParameter(params, ZSTD_c_overlapLog, &value) );
CHECK_EQ(value, 0);
CHECK_Z( ZSTD_CCtxParams_getParameter(params, ZSTD_c_jobSize, &value) );
CHECK_EQ(value, 0);
/* Set and check the overlap log and job size. */
CHECK_Z( ZSTD_CCtxParams_setParameter(params, ZSTD_c_overlapLog, 5) );
CHECK_Z( ZSTD_CCtxParams_setParameter(params, ZSTD_c_jobSize, jobSize) );
CHECK_Z( ZSTD_CCtxParams_getParameter(params, ZSTD_c_overlapLog, &value) );
CHECK_EQ(value, 5);
CHECK_Z( ZSTD_CCtxParams_getParameter(params, ZSTD_c_jobSize, &value) );
CHECK_EQ(value, jobSize);
/* Set the number of workers and check the overlap log and job size. */
CHECK_Z( ZSTD_CCtxParams_setParameter(params, ZSTD_c_nbWorkers, 2) );
CHECK_Z( ZSTD_CCtxParams_getParameter(params, ZSTD_c_overlapLog, &value) );
CHECK_EQ(value, 5);
CHECK_Z( ZSTD_CCtxParams_getParameter(params, ZSTD_c_jobSize, &value) );
CHECK_EQ(value, jobSize);
ZSTD_freeCCtxParams(params);
}
DISPLAYLEVEL(3, "OK \n");
/* Simple API multiframe test */
DISPLAYLEVEL(3, "test%3i : compress multiple frames : ", testNb++);
{ size_t off = 0;
int i;
int const segs = 4;
/* only use the first half so we don't push against size limit of compressedBuffer */
size_t const segSize = (CNBuffSize / 2) / segs;
const U32 skipLen = 129 KB;
char* const skipBuff = (char*)malloc(skipLen);
assert(skipBuff != NULL);
memset(skipBuff, 0, skipLen);
for (i = 0; i < segs; i++) {
CHECK_NEWV(r, ZSTD_compress(
(BYTE*)compressedBuffer + off, CNBuffSize - off,
(BYTE*)CNBuffer + segSize * (size_t)i, segSize,
5) );
off += r;
if (i == segs/2) {
/* insert skippable frame */
size_t const skippableSize =
ZSTD_writeSkippableFrame((BYTE*)compressedBuffer + off, compressedBufferSize,
skipBuff, skipLen, seed % 15);
CHECK_Z(skippableSize);
off += skippableSize;
}
}
cSize = off;
free(skipBuff);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : get decompressed size of multiple frames : ", testNb++);
{ unsigned long long const r = ZSTD_findDecompressedSize(compressedBuffer, cSize);
if (r != CNBuffSize / 2) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : get tight decompressed bound of multiple frames : ", testNb++);
{ unsigned long long const bound = ZSTD_decompressBound(compressedBuffer, cSize);
if (bound != CNBuffSize / 2) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress multiple frames : ", testNb++);
{ CHECK_NEWV(r, ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize));
if (r != CNBuffSize / 2) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : check decompressed result : ", testNb++);
if (memcmp(decodedBuffer, CNBuffer, CNBuffSize / 2) != 0) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
/* Simple API skippable frame test */
DISPLAYLEVEL(3, "test%3i : read/write a skippable frame : ", testNb++);
{ U32 i;
unsigned readMagic;
unsigned long long receivedSize;
size_t skippableSize;
const U32 skipLen = 129 KB;
char* const skipBuff = (char*)malloc(skipLen);
assert(skipBuff != NULL);
for (i = 0; i < skipLen; i++)
skipBuff[i] = (char) ((seed + i) % 256);
skippableSize = ZSTD_writeSkippableFrame(
compressedBuffer, compressedBufferSize,
skipBuff, skipLen, seed % 15);
CHECK_Z(skippableSize);
CHECK_EQ(1, ZSTD_isSkippableFrame(compressedBuffer, skippableSize));
receivedSize = ZSTD_readSkippableFrame(decodedBuffer, CNBuffSize, &readMagic, compressedBuffer, skippableSize);
CHECK_EQ(skippableSize, receivedSize + ZSTD_SKIPPABLEHEADERSIZE);
CHECK_EQ(seed % 15, readMagic);
if (memcmp(decodedBuffer, skipBuff, skipLen) != 0) goto _output_error;
free(skipBuff);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : read/write an empty skippable frame : ", testNb++);
{
unsigned readMagic;
unsigned long long receivedSize;
size_t skippableSize;
skippableSize = ZSTD_writeSkippableFrame(
compressedBuffer, compressedBufferSize,
CNBuffer, 0, seed % 15);
CHECK_EQ(ZSTD_SKIPPABLEHEADERSIZE, skippableSize);
CHECK_EQ(1, ZSTD_isSkippableFrame(compressedBuffer, skippableSize));
receivedSize = ZSTD_readSkippableFrame(NULL, 0, &readMagic, compressedBuffer, skippableSize);
CHECK_EQ(skippableSize, receivedSize + ZSTD_SKIPPABLEHEADERSIZE);
CHECK_EQ(seed % 15, readMagic);
}
DISPLAYLEVEL(3, "OK \n");
/* Dictionary and CCtx Duplication tests */
{ ZSTD_CCtx* const ctxOrig = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
static const size_t dictSize = 551;
assert(dctx != NULL); assert(ctxOrig != NULL);
DISPLAYLEVEL(3, "test%3i : load dictionary into context : ", testNb++);
CHECK_Z( ZSTD_compressBegin_usingDict(ctxOrig, CNBuffer, dictSize, 2) );
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress with flat dictionary : ", testNb++);
cSize = 0;
CHECKPLUS(r, ZSTD_compressEnd(ctxOrig,
compressedBuffer, compressedBufferSize,
(const char*)CNBuffer + dictSize, CNBuffSize - dictSize),
cSize += r);
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : frame built with flat dictionary should be decompressible : ", testNb++);
CHECKPLUS(r, ZSTD_decompress_usingDict(dctx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize,
CNBuffer, dictSize),
if (r != CNBuffSize - dictSize) goto _output_error);
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : decompress with DDict : ", testNb++);
{ ZSTD_DDict* const ddict = ZSTD_createDDict(CNBuffer, dictSize);
size_t const r = ZSTD_decompress_usingDDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, ddict);
if (r != CNBuffSize - dictSize) goto _output_error;
DISPLAYLEVEL(3, "OK (size of DDict : %u) \n", (unsigned)ZSTD_sizeof_DDict(ddict));
ZSTD_freeDDict(ddict);
}
DISPLAYLEVEL(3, "test%3i : decompress with static DDict : ", testNb++);
{ size_t const ddictBufferSize = ZSTD_estimateDDictSize(dictSize, ZSTD_dlm_byCopy);
void* const ddictBuffer = malloc(ddictBufferSize);
if (ddictBuffer == NULL) goto _output_error;
{ const ZSTD_DDict* const ddict = ZSTD_initStaticDDict(ddictBuffer, ddictBufferSize, CNBuffer, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto);
size_t const r = ZSTD_decompress_usingDDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, ddict);
if (r != CNBuffSize - dictSize) goto _output_error;
}
free(ddictBuffer);
DISPLAYLEVEL(3, "OK (size of static DDict : %u) \n", (unsigned)ddictBufferSize);
}
#if !defined(ZSTD_EXCLUDE_DFAST_BLOCK_COMPRESSOR) \
&& !defined(ZSTD_EXCLUDE_GREEDY_BLOCK_COMPRESSOR) \
&& !defined(ZSTD_EXCLUDE_DFAST_BLOCK_COMPRESSOR) \
&& !defined(ZSTD_EXCLUDE_LAZY_BLOCK_COMPRESSOR) \
&& !defined(ZSTD_EXCLUDE_LAZY2_BLOCK_COMPRESSOR) \
&& !defined(ZSTD_EXCLUDE_BTLAZY2_BLOCK_COMPRESSOR) \
&& !defined(ZSTD_EXCLUDE_BTOPT_BLOCK_COMPRESSOR) \
&& !defined(ZSTD_EXCLUDE_BTULTRA_BLOCK_COMPRESSOR)
/* Note : these tests should be replaced by proper regression tests,
* but existing ones do not focus on small data + dictionary + all levels.
*/
if ((int)(compressibility * 100 + 0.1) == FUZ_compressibility_default) { /* test only valid with known input */
size_t const flatdictSize = 22 KB;
size_t const contentSize = 9 KB;
const void* const dict = (const char*)CNBuffer;
const void* const contentStart = (const char*)dict + flatdictSize;
/* These upper bounds are generally within a few bytes of the compressed size */
size_t target_nodict_cSize[22+1] = { 3840, 3770, 3870, 3830, 3770,
3770, 3770, 3770, 3750, 3750,
3742, 3675, 3674, 3665, 3664,
3663, 3662, 3661, 3660, 3660,
3660, 3660, 3660 };
size_t const target_wdict_cSize[22+1] = { 2830, 2896, 2893, 2840, 2950,
2950, 2950, 2925, 2900, 2892,
2910, 2910, 2910, 2780, 2775,
2765, 2760, 2755, 2754, 2753,
2753, 2753, 2753 };
int l = 1;
int const maxLevel = ZSTD_maxCLevel();
/* clevels with strategies that support rowhash on small inputs */
int rowLevel = 4;
int const rowLevelEnd = 8;
DISPLAYLEVEL(3, "test%3i : flat-dictionary efficiency test : \n", testNb++);
assert(maxLevel == 22);
RDG_genBuffer(CNBuffer, flatdictSize + contentSize, compressibility, 0., seed);
DISPLAYLEVEL(4, "content hash : %016llx; dict hash : %016llx \n",
(unsigned long long)XXH64(contentStart, contentSize, 0),
(unsigned long long)XXH64(dict, flatdictSize, 0));
for ( ; l <= maxLevel; l++) {
size_t const nodict_cSize = ZSTD_compress(compressedBuffer, compressedBufferSize,
contentStart, contentSize, l);
if (nodict_cSize > target_nodict_cSize[l]) {
DISPLAYLEVEL(1, "error : compression at level %i worse than expected (%u > %u) \n",
l, (unsigned)nodict_cSize, (unsigned)target_nodict_cSize[l]);
goto _output_error;
}
DISPLAYLEVEL(4, "level %i : max expected %u >= reached %u \n",
l, (unsigned)target_nodict_cSize[l], (unsigned)nodict_cSize);
}
for ( l=1 ; l <= maxLevel; l++) {
size_t const wdict_cSize = ZSTD_compress_usingDict(ctxOrig,
compressedBuffer, compressedBufferSize,
contentStart, contentSize,
dict, flatdictSize,
l);
if (wdict_cSize > target_wdict_cSize[l]) {
DISPLAYLEVEL(1, "error : compression with dictionary at level %i worse than expected (%u > %u) \n",
l, (unsigned)wdict_cSize, (unsigned)target_wdict_cSize[l]);
goto _output_error;
}
DISPLAYLEVEL(4, "level %i with dictionary : max expected %u >= reached %u \n",
l, (unsigned)target_wdict_cSize[l], (unsigned)wdict_cSize);
}
/* Compression with ZSTD_compress2 and row match finder force enabled.
* Give some slack for force-enabled row matchfinder since we're on a small input (9KB)
*/
for ( ; rowLevel <= rowLevelEnd; ++rowLevel) target_nodict_cSize[rowLevel] += 5;
for (l=1 ; l <= maxLevel; l++) {
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t nodict_cSize;
ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, l);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_useRowMatchFinder, ZSTD_ps_enable);
nodict_cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize,
contentStart, contentSize);
if (nodict_cSize > target_nodict_cSize[l]) {
DISPLAYLEVEL(1, "error : compression with compress2 at level %i worse than expected (%u > %u) \n",
l, (unsigned)nodict_cSize, (unsigned)target_nodict_cSize[l]);
ZSTD_freeCCtx(cctx);
goto _output_error;
}
DISPLAYLEVEL(4, "level %i with compress2 : max expected %u >= reached %u \n",
l, (unsigned)target_nodict_cSize[l], (unsigned)nodict_cSize);
ZSTD_freeCCtx(cctx);
}
/* Dict compression with DMS */
for ( l=1 ; l <= maxLevel; l++) {
size_t wdict_cSize;
CHECK_Z( ZSTD_CCtx_loadDictionary(ctxOrig, dict, flatdictSize) );
CHECK_Z( ZSTD_CCtx_setParameter(ctxOrig, ZSTD_c_compressionLevel, l) );
CHECK_Z( ZSTD_CCtx_setParameter(ctxOrig, ZSTD_c_enableDedicatedDictSearch, 0) );
CHECK_Z( ZSTD_CCtx_setParameter(ctxOrig, ZSTD_c_forceAttachDict, ZSTD_dictForceAttach) );
CHECK_Z( ZSTD_CCtx_setParameter(ctxOrig, ZSTD_c_prefetchCDictTables, seed % 3) );
wdict_cSize = ZSTD_compress2(ctxOrig, compressedBuffer, compressedBufferSize, contentStart, contentSize);
if (wdict_cSize > target_wdict_cSize[l]) {
DISPLAYLEVEL(1, "error : compression with dictionary and compress2 at level %i worse than expected (%u > %u) \n",
l, (unsigned)wdict_cSize, (unsigned)target_wdict_cSize[l]);
goto _output_error;
}
DISPLAYLEVEL(4, "level %i with dictionary and compress2 : max expected %u >= reached %u \n",
l, (unsigned)target_wdict_cSize[l], (unsigned)wdict_cSize);
}
DISPLAYLEVEL(4, "compression efficiency tests OK \n");
}
#endif
ZSTD_freeCCtx(ctxOrig);
ZSTD_freeDCtx(dctx);
}
/* Dictionary and dictBuilder tests */
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const dictBufferCapacity = 16 KB;
void* const dictBuffer = malloc(dictBufferCapacity);
size_t const totalSampleSize = 1 MB;
size_t const sampleUnitSize = 8 KB;
U32 const nbSamples = (U32)(totalSampleSize / sampleUnitSize);
size_t* const samplesSizes = (size_t*) malloc(nbSamples * sizeof(size_t));
size_t dictSize;
U32 dictID;
size_t dictHeaderSize;
size_t dictBufferFixedSize = 144;
unsigned char const dictBufferFixed[144] = {0x37, 0xa4, 0x30, 0xec, 0x63, 0x00, 0x00, 0x00, 0x08, 0x10, 0x00, 0x1f,
0x0f, 0x00, 0x28, 0xe5, 0x03, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x80, 0x0f, 0x9e, 0x0f, 0x00, 0x00, 0x24, 0x40, 0x80, 0x00, 0x01,
0x02, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0xde, 0x08,
0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08,
0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08,
0x08, 0x08, 0x08, 0x08, 0xbc, 0xe1, 0x4b, 0x92, 0x0e, 0xb4, 0x7b, 0x18,
0x86, 0x61, 0x18, 0xc6, 0x18, 0x63, 0x8c, 0x31, 0xc6, 0x18, 0x63, 0x8c,
0x31, 0x66, 0x66, 0x66, 0x66, 0xb6, 0x6d, 0x01, 0x00, 0x00, 0x00, 0x04,
0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00, 0x20, 0x73, 0x6f, 0x64, 0x61,
0x6c, 0x65, 0x73, 0x20, 0x74, 0x6f, 0x72, 0x74, 0x6f, 0x72, 0x20, 0x65,
0x6c, 0x65, 0x69, 0x66, 0x65, 0x6e, 0x64, 0x2e, 0x20, 0x41, 0x6c, 0x69};
if (dictBuffer==NULL || samplesSizes==NULL) {
free(dictBuffer);
free(samplesSizes);
goto _output_error;
}
DISPLAYLEVEL(3, "test%3i : dictBuilder on cyclic data : ", testNb++);
assert(compressedBufferSize >= totalSampleSize);
{ U32 u; for (u=0; u<totalSampleSize; u++) ((BYTE*)decodedBuffer)[u] = (BYTE)u; }
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
{ size_t const sDictSize = ZDICT_trainFromBuffer(dictBuffer, dictBufferCapacity,
decodedBuffer, samplesSizes, nbSamples);
if (ZDICT_isError(sDictSize)) goto _output_error;
DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)sDictSize);
}
DISPLAYLEVEL(3, "test%3i : dictBuilder : ", testNb++);
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
dictSize = ZDICT_trainFromBuffer(dictBuffer, dictBufferCapacity,
CNBuffer, samplesSizes, nbSamples);
if (ZDICT_isError(dictSize)) goto _output_error;
DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)dictSize);
DISPLAYLEVEL(3, "test%3i : FASTCOVER dictBuilder samples smaller than a dmer (issue #4499) : ", testNb++);
{ const char smallData[8] = { 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H' };
const size_t smallSizes[8] = { 1, 1, 1, 1, 1, 1, 1, 1 };
size_t const r = ZDICT_trainFromBuffer(dictBuffer, dictBufferCapacity,
smallData, smallSizes, 8);
if (!ZDICT_isError(r)) goto _output_error;
if (ZSTD_getErrorCode(r) != ZSTD_error_srcSize_wrong) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Multithreaded COVER dictBuilder : ", testNb++);
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
{ ZDICT_cover_params_t coverParams;
memset(&coverParams, 0, sizeof(coverParams));
coverParams.steps = 8;
coverParams.nbThreads = 4;
dictSize = ZDICT_optimizeTrainFromBuffer_cover(
dictBuffer, dictBufferCapacity,
CNBuffer, samplesSizes, nbSamples/8, /* less samples for faster tests */
&coverParams);
if (ZDICT_isError(dictSize)) goto _output_error;
}
DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)dictSize);
DISPLAYLEVEL(3, "test%3i : COVER dictBuilder with shrinkDict: ", testNb++);
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
{ ZDICT_cover_params_t coverParams;
memset(&coverParams, 0, sizeof(coverParams));
coverParams.steps = 8;
coverParams.nbThreads = 4;
coverParams.shrinkDict = 1;
coverParams.shrinkDictMaxRegression = 1;
dictSize = ZDICT_optimizeTrainFromBuffer_cover(
dictBuffer, dictBufferCapacity,
CNBuffer, samplesSizes, nbSamples/8, /* less samples for faster tests */
&coverParams);
if (ZDICT_isError(dictSize)) goto _output_error;
}
DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)dictSize);
DISPLAYLEVEL(3, "test%3i : Multithreaded FASTCOVER dictBuilder : ", testNb++);
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
{ ZDICT_fastCover_params_t fastCoverParams;
memset(&fastCoverParams, 0, sizeof(fastCoverParams));
fastCoverParams.steps = 8;
fastCoverParams.nbThreads = 4;
dictSize = ZDICT_optimizeTrainFromBuffer_fastCover(
dictBuffer, dictBufferCapacity,
CNBuffer, samplesSizes, nbSamples,
&fastCoverParams);
if (ZDICT_isError(dictSize)) goto _output_error;
}
DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)dictSize);
DISPLAYLEVEL(3, "test%3i : FASTCOVER dictBuilder with shrinkDict: ", testNb++);
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
{ ZDICT_fastCover_params_t fastCoverParams;
memset(&fastCoverParams, 0, sizeof(fastCoverParams));
fastCoverParams.steps = 8;
fastCoverParams.nbThreads = 4;
fastCoverParams.shrinkDict = 1;
fastCoverParams.shrinkDictMaxRegression = 1;
dictSize = ZDICT_optimizeTrainFromBuffer_fastCover(
dictBuffer, dictBufferCapacity,
CNBuffer, samplesSizes, nbSamples,
&fastCoverParams);
if (ZDICT_isError(dictSize)) goto _output_error;
}
DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)dictSize);
DISPLAYLEVEL(3, "test%3i : COVER dictBuilder orders equal dmers by position : ", testNb++);
/* COVER sorts its suffix array with qsort(), breaking ties between
* equal dmers by sample position; COVER_group() relies on that order
* to count how many samples contain a dmer. A tie-break that depends on
* the sort's internal state instead (the former address comparison)
* still produces a dictionary, just a different one. Low-entropy
* samples make ties frequent, so train on them and compare against the
* dictionary the position order produces, for both comparator paths. */
{ size_t const dictCapacity = 16 KB;
void* const coverDict = malloc(dictCapacity);
unsigned const dValues[2] = { 8, 16 }; /* COVER_cmp8() and COVER_cmp() */
XXH64_hash_t const expectedHash[2] = { 0x77EB0C1D595FF7A3ULL, 0xE02EC0FD3A5BBF6DULL };
unsigned dIdx;
if (coverDict==NULL) goto _output_error;
RDG_genBuffer(decodedBuffer, totalSampleSize, 0.5, 0.05, 1); /* low-entropy: many repeated dmers */
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
for (dIdx = 0; dIdx < 2; dIdx++) {
ZDICT_cover_params_t coverParams;
size_t coverDictSize;
XXH64_hash_t hash;
memset(&coverParams, 0, sizeof(coverParams));
coverParams.k = 256;
coverParams.d = dValues[dIdx];
coverDictSize = ZDICT_trainFromBuffer_cover(coverDict, dictCapacity,
decodedBuffer, samplesSizes, nbSamples, coverParams);
if (ZDICT_isError(coverDictSize)) goto _output_error;
hash = XXH64(coverDict, coverDictSize, 0);
if (hash != expectedHash[dIdx]) {
DISPLAYLEVEL(1, "error : COVER dictionary for d=%u hashes to %08X%08X, expected %08X%08X \n",
coverParams.d,
(unsigned)(hash >> 32), (unsigned)hash,
(unsigned)(expectedHash[dIdx] >> 32), (unsigned)expectedHash[dIdx]);
goto _output_error;
}
}
free(coverDict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : COVER_selectDict shrinkDict with content < ZDICT_DICTSIZE_MIN : ", testNb++);
/* The shrink loop tries suffixes of the content, starting at ZDICT_DICTSIZE_MIN.
* With smaller content no suffix fits, so the full dictionary must be returned.
* The content is allocated at its exact size, so under ASan any read before it
* is reported as a heap-buffer-overflow. */
{ size_t offsets[8];
unsigned const nbSelectSamples = (unsigned)(sizeof(offsets) / sizeof(offsets[0]));
size_t const contentSize = ZDICT_DICTSIZE_MIN - 16;
size_t const selectCapacity = 4 KB;
BYTE* const content = (BYTE*)malloc(contentSize);
BYTE* const expected = (BYTE*)malloc(selectCapacity);
ZDICT_cover_params_t coverParams;
COVER_dictSelection_t selection;
size_t expectedSize;
unsigned u;
if (content == NULL || expected == NULL) {
free(content);
free(expected);
goto _output_error;
}
memcpy(content, CNBuffer, contentSize);
for (u = 0; u < nbSelectSamples; u++) offsets[u] = u * sampleUnitSize;
memset(&coverParams, 0, sizeof(coverParams));
coverParams.splitPoint = 1.0;
coverParams.shrinkDict = 1;
coverParams.shrinkDictMaxRegression = 100000; /* accept any candidate */
expectedSize = ZDICT_finalizeDictionary(expected, selectCapacity,
content, contentSize,
CNBuffer, samplesSizes, nbSelectSamples,
coverParams.zParams);
selection = COVER_selectDict(content, selectCapacity, contentSize,
(const BYTE*)CNBuffer, samplesSizes, nbSelectSamples,
nbSelectSamples, nbSelectSamples, coverParams, offsets, 0);
{ int const ok = !ZDICT_isError(expectedSize)
&& !COVER_dictSelectionIsError(selection)
&& selection.dictSize == expectedSize
&& !memcmp(selection.dictContent, expected, expectedSize);
COVER_dictSelectionFree(selection);
free(content);
free(expected);
if (!ok) goto _output_error;
}
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : check dictID : ", testNb++);
dictID = ZDICT_getDictID(dictBuffer, dictSize);
if (dictID==0) goto _output_error;
DISPLAYLEVEL(3, "OK : %u \n", (unsigned)dictID);
DISPLAYLEVEL(3, "test%3i : check dict header size no error : ", testNb++);
dictHeaderSize = ZDICT_getDictHeaderSize(dictBuffer, dictSize);
if (dictHeaderSize==0) goto _output_error;
DISPLAYLEVEL(3, "OK : %u \n", (unsigned)dictHeaderSize);
DISPLAYLEVEL(3, "test%3i : check dict header size correctness : ", testNb++);
{ dictHeaderSize = ZDICT_getDictHeaderSize(dictBufferFixed, dictBufferFixedSize);
if (dictHeaderSize != 115) goto _output_error;
}
DISPLAYLEVEL(3, "OK : %u \n", (unsigned)dictHeaderSize);
DISPLAYLEVEL(3, "test%3i : Dictionary creation errors : ", testNb++);
test_dictionaryLoadErrors(dictBuffer, dictSize);
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Large dictionary repcodes : ", testNb++);
test_largeDictionaryRepcodes(dictBuffer, dictSize);
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress with dictionary : ", testNb++);
cSize = ZSTD_compress_usingDict(cctx, compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize,
dictBuffer, dictSize, 4);
if (ZSTD_isError(cSize)) goto _output_error;
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : retrieve dictID from dictionary : ", testNb++);
{ U32 const did = ZSTD_getDictID_fromDict(dictBuffer, dictSize);
if (did != dictID) goto _output_error; /* non-conformant (content-only) dictionary */
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : retrieve dictID from frame : ", testNb++);
{ U32 const did = ZSTD_getDictID_fromFrame(compressedBuffer, cSize);
if (did != dictID) goto _output_error; /* non-conformant (content-only) dictionary */
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : frame built with dictionary should be decompressible : ", testNb++);
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx(); assert(dctx != NULL);
CHECKPLUS(r, ZSTD_decompress_usingDict(dctx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize,
dictBuffer, dictSize),
if (r != CNBuffSize) goto _output_error);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : dictionary IDs in the encoded error range (32-bit size_t) : ", testNb++);
/* On 32-bit, (size_t)-120 == 4294967176 :
* dictIDs above must not be confused with an error code while loading the dictionary. */
{ U32 const dictIDs[] = { 1U, 0x7FFFFFFFU, 0x80000000U,
0xFFFFFF88U /* last non-error value */,
0xFFFFFF89U /* first value in the error range */,
0xFFFFFFFFU };
size_t const nbDictIDs = sizeof(dictIDs) / sizeof(dictIDs[0]);
size_t const srcSize = MIN(CNBuffSize, (size_t)(16 KB));
void* const dictCopy = malloc(dictSize);
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
size_t n;
CHECK(dictCopy != NULL);
CHECK(dctx != NULL);
memcpy(dictCopy, dictBuffer, dictSize);
for (n = 0; n < nbDictIDs; n++) {
int mode;
MEM_writeLE32((char*)dictCopy + 4, dictIDs[n]); /* after the magic number */
CHECK_EQ(ZSTD_getDictID_fromDict(dictCopy, dictSize), dictIDs[n]);
for (mode = 0; mode < 3; mode++) {
size_t r;
if (mode == 0) { /* ZSTD_compress_usingDict() */
cSize = ZSTD_compress_usingDict(cctx, compressedBuffer, compressedBufferSize,
CNBuffer, srcSize, dictCopy, dictSize, 3);
} else if (mode == 1) { /* ZSTD_CDict */
ZSTD_CDict* const cdict = ZSTD_createCDict(dictCopy, dictSize, 3);
CHECK(cdict != NULL);
CHECK_EQ(ZSTD_getDictID_fromCDict(cdict), dictIDs[n]);
cSize = ZSTD_compress_usingCDict(cctx, compressedBuffer, compressedBufferSize,
CNBuffer, srcSize, cdict);
ZSTD_freeCDict(cdict);
} else { /* ZSTD_CCtx_loadDictionary() */
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
CHECK_Z(ZSTD_CCtx_loadDictionary(cctx, dictCopy, dictSize));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize,
CNBuffer, srcSize);
}
CHECK_Z(cSize);
CHECK_EQ(ZSTD_getDictID_fromFrame(compressedBuffer, cSize), dictIDs[n]);
r = ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize,
compressedBuffer, cSize, dictCopy, dictSize);
CHECK_Z(r);
CHECK_EQ(r, srcSize);
CHECK(!memcmp(decodedBuffer, CNBuffer, srcSize));
}
}
ZSTD_freeDCtx(dctx);
free(dictCopy);
}
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : estimate CDict size : ", testNb++);
{ ZSTD_compressionParameters const cParams = ZSTD_getCParams(1, CNBuffSize, dictSize);
size_t const estimatedSize = ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byRef);
DISPLAYLEVEL(3, "OK : %u \n", (unsigned)estimatedSize);
}
DISPLAYLEVEL(3, "test%3i : compress with CDict ", testNb++);
{ ZSTD_compressionParameters const cParams = ZSTD_getCParams(1, CNBuffSize, dictSize);
ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dictBuffer, dictSize,
ZSTD_dlm_byRef, ZSTD_dct_auto,
cParams, ZSTD_defaultCMem);
assert(cdict != NULL);
DISPLAYLEVEL(3, "(size : %u) : ", (unsigned)ZSTD_sizeof_CDict(cdict));
assert(ZSTD_getDictID_fromDict(dictBuffer, dictSize) == ZSTD_getDictID_fromCDict(cdict));
cSize = ZSTD_compress_usingCDict(cctx, compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize, cdict);
ZSTD_freeCDict(cdict);
if (ZSTD_isError(cSize)) goto _output_error;
}
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : retrieve dictID from frame : ", testNb++);
{ U32 const did = ZSTD_getDictID_fromFrame(compressedBuffer, cSize);
if (did != dictID) goto _output_error; /* non-conformant (content-only) dictionary */
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : frame built with dictionary should be decompressible : ", testNb++);
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx(); assert(dctx != NULL);
CHECKPLUS(r, ZSTD_decompress_usingDict(dctx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize,
dictBuffer, dictSize),
if (r != CNBuffSize) goto _output_error);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress with static CDict : ", testNb++);
{ int const maxLevel = ZSTD_maxCLevel();
int level;
for (level = 1; level <= maxLevel; ++level) {
ZSTD_compressionParameters const cParams = ZSTD_getCParams(level, CNBuffSize, dictSize);
size_t const cdictSize = ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byCopy);
void* const cdictBuffer = malloc(cdictSize);
if (cdictBuffer==NULL) goto _output_error;
{ const ZSTD_CDict* const cdict = ZSTD_initStaticCDict(
cdictBuffer, cdictSize,
dictBuffer, dictSize,
ZSTD_dlm_byCopy, ZSTD_dct_auto,
cParams);
if (cdict == NULL) {
DISPLAY("ZSTD_initStaticCDict failed ");
goto _output_error;
}
cSize = ZSTD_compress_usingCDict(cctx,
compressedBuffer, compressedBufferSize,
CNBuffer, MIN(10 KB, CNBuffSize), cdict);
if (ZSTD_isError(cSize)) {
DISPLAY("ZSTD_compress_usingCDict failed ");
goto _output_error;
} }
free(cdictBuffer);
} }
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : ZSTD_compress_usingCDict_advanced, no contentSize, no dictID : ", testNb++);
{ ZSTD_frameParameters const fParams = { 0 /* frameSize */, 1 /* checksum */, 1 /* noDictID*/ };
ZSTD_compressionParameters const cParams = ZSTD_getCParams(1, CNBuffSize, dictSize);
ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, cParams, ZSTD_defaultCMem);
assert(cdict != NULL);
cSize = ZSTD_compress_usingCDict_advanced(cctx,
compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize,
cdict, fParams);
ZSTD_freeCDict(cdict);
if (ZSTD_isError(cSize)) goto _output_error;
}
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : try retrieving contentSize from frame : ", testNb++);
{ U64 const contentSize = ZSTD_getFrameContentSize(compressedBuffer, cSize);
if (contentSize != ZSTD_CONTENTSIZE_UNKNOWN) goto _output_error;
}
DISPLAYLEVEL(3, "OK (unknown)\n");
DISPLAYLEVEL(3, "test%3i : frame built without dictID should be decompressible : ", testNb++);
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx();
assert(dctx != NULL);
CHECKPLUS(r, ZSTD_decompress_usingDict(dctx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize,
dictBuffer, dictSize),
if (r != CNBuffSize) goto _output_error);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_compress_advanced, no dictID : ", testNb++);
{ ZSTD_parameters p = ZSTD_getParams(3, CNBuffSize, dictSize);
p.fParams.noDictIDFlag = 1;
cSize = ZSTD_compress_advanced(cctx, compressedBuffer, compressedBufferSize,
CNBuffer, CNBuffSize,
dictBuffer, dictSize, p);
if (ZSTD_isError(cSize)) goto _output_error;
}
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
DISPLAYLEVEL(3, "test%3i : frame built without dictID should be decompressible : ", testNb++);
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx(); assert(dctx != NULL);
CHECKPLUS(r, ZSTD_decompress_usingDict(dctx,
decodedBuffer, CNBuffSize,
compressedBuffer, cSize,
dictBuffer, dictSize),
if (r != CNBuffSize) goto _output_error);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : dictionary containing only header should return error : ", testNb++);
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx();
assert(dctx != NULL);
{ const size_t ret = ZSTD_decompress_usingDict(
dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize,
"\x37\xa4\x30\xec\x11\x22\x33\x44", 8);
if (ZSTD_getErrorCode(ret) != ZSTD_error_dictionary_corrupted)
goto _output_error;
}
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : loading a dictionary containing only header should return dictionary_corrupted : ", testNb++);
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx();
const char* const headerOnly = "\x37\xa4\x30\xec\x11\x22\x33\x44";
assert(dctx != NULL);
CHECK_EQ(ZSTD_getErrorCode(ZSTD_DCtx_loadDictionary(dctx, headerOnly, 8)), ZSTD_error_dictionary_corrupted);
CHECK_EQ(ZSTD_getErrorCode(ZSTD_DCtx_loadDictionary_byReference(dctx, headerOnly, 8)), ZSTD_error_dictionary_corrupted);
/* the context remains usable */
CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, dictBuffer, dictSize));
CHECK_EQ(ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize), CNBuffSize);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3d : bufferless api with cdict : ", testNb++);
{ ZSTD_CDict* const cdict = ZSTD_createCDict(dictBuffer, dictSize, 1);
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
ZSTD_frameParameters const fParams = { 0, 1, 0 };
size_t cBlockSize;
cSize = 0;
CHECK_Z(ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, ZSTD_CONTENTSIZE_UNKNOWN));
cBlockSize = ZSTD_compressContinue(cctx, (char*)compressedBuffer + cSize, compressedBufferSize - cSize, CNBuffer, 1000);
CHECK_Z(cBlockSize);
cSize += cBlockSize;
cBlockSize = ZSTD_compressEnd(cctx, (char*)compressedBuffer + cSize, compressedBufferSize - cSize, (char const*)CNBuffer + 2000, 1000);
CHECK_Z(cBlockSize);
cSize += cBlockSize;
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, dictBuffer, dictSize));
ZSTD_freeCDict(cdict);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Building cdict w/ ZSTD_dct_fullDict on a good dictionary : ", testNb++);
{ ZSTD_compressionParameters const cParams = ZSTD_getCParams(1, CNBuffSize, dictSize);
ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_fullDict, cParams, ZSTD_defaultCMem);
if (cdict==NULL) goto _output_error;
ZSTD_freeCDict(cdict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Building cdict w/ ZSTD_dct_fullDict on a rawContent (must fail) : ", testNb++);
{ ZSTD_compressionParameters const cParams = ZSTD_getCParams(1, CNBuffSize, dictSize);
ZSTD_CDict* const cdict = ZSTD_createCDict_advanced((const char*)dictBuffer+1, dictSize-1, ZSTD_dlm_byRef, ZSTD_dct_fullDict, cParams, ZSTD_defaultCMem);
if (cdict!=NULL) goto _output_error;
ZSTD_freeCDict(cdict);
}
DISPLAYLEVEL(3, "OK \n");
{ char* rawDictBuffer = (char*)malloc(dictSize);
assert(rawDictBuffer);
memcpy(rawDictBuffer, (char*)dictBuffer + 2, dictSize - 2);
memset(rawDictBuffer + dictSize - 2, 0, 2);
MEM_writeLE32((char*)rawDictBuffer, ZSTD_MAGIC_DICTIONARY);
DISPLAYLEVEL(3, "test%3i : Loading rawContent starting with dict header w/ ZSTD_dct_auto should fail : ", testNb++);
{
size_t ret;
/* Either operation is allowed to fail, but one must fail. */
ret = ZSTD_CCtx_loadDictionary_advanced(
cctx, (const char*)rawDictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto);
if (!ZSTD_isError(ret)) {
ret = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100));
if (!ZSTD_isError(ret)) goto _output_error;
}
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loading rawContent starting with dict header w/ ZSTD_dct_rawContent should pass : ", testNb++);
{
size_t ret;
ret = ZSTD_CCtx_loadDictionary_advanced(
cctx, (const char*)rawDictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_rawContent);
if (ZSTD_isError(ret)) goto _output_error;
ret = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100));
if (ZSTD_isError(ret)) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Testing non-attached CDict with ZSTD_dct_rawContent : ", testNb++);
{ size_t const srcSize = MIN(CNBuffSize, 100);
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
/* Force the dictionary to be reloaded in raw content mode */
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_forceAttachDict, ZSTD_dictForceLoad));
CHECK_Z(ZSTD_CCtx_loadDictionary_advanced(cctx, rawDictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_rawContent));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, srcSize);
CHECK_Z(cSize);
}
DISPLAYLEVEL(3, "OK \n");
free(rawDictBuffer);
}
DISPLAYLEVEL(3, "test%3i : ZSTD_CCtx_refCDict() then set parameters : ", testNb++);
{ ZSTD_CDict* const cdict = ZSTD_createCDict(CNBuffer, dictSize, 1);
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 1) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_hashLog, 12 ));
CHECK_Z( ZSTD_CCtx_refCDict(cctx, cdict) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 1) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_hashLog, 12 ));
ZSTD_freeCDict(cdict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loading dictionary before setting parameters is the same as loading after : ", testNb++);
{
size_t size1, size2;
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 7) );
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, CNBuffer, MIN(CNBuffSize, 10 KB)) );
size1 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100 KB));
if (ZSTD_isError(size1)) goto _output_error;
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, CNBuffer, MIN(CNBuffSize, 10 KB)) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 7) );
size2 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100 KB));
if (ZSTD_isError(size2)) goto _output_error;
if (size1 != size2) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loading a dictionary clears the prefix : ", testNb++);
{
CHECK_Z( ZSTD_CCtx_refPrefix(cctx, (const char*)dictBuffer, dictSize) );
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, (const char*)dictBuffer, dictSize) );
CHECK_Z( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100)) );
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loading a dictionary clears the cdict : ", testNb++);
{
ZSTD_CDict* const cdict = ZSTD_createCDict(dictBuffer, dictSize, 1);
CHECK_Z( ZSTD_CCtx_refCDict(cctx, cdict) );
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, (const char*)dictBuffer, dictSize) );
CHECK_Z( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100)) );
ZSTD_freeCDict(cdict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loading a cdict clears the prefix : ", testNb++);
{
ZSTD_CDict* const cdict = ZSTD_createCDict(dictBuffer, dictSize, 1);
CHECK_Z( ZSTD_CCtx_refPrefix(cctx, (const char*)dictBuffer, dictSize) );
CHECK_Z( ZSTD_CCtx_refCDict(cctx, cdict) );
CHECK_Z( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100)) );
ZSTD_freeCDict(cdict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loading a cdict clears the dictionary : ", testNb++);
{
ZSTD_CDict* const cdict = ZSTD_createCDict(dictBuffer, dictSize, 1);
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, (const char*)dictBuffer, dictSize) );
CHECK_Z( ZSTD_CCtx_refCDict(cctx, cdict) );
CHECK_Z( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100)) );
ZSTD_freeCDict(cdict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loading a prefix clears the dictionary : ", testNb++);
{
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, (const char*)dictBuffer, dictSize) );
CHECK_Z( ZSTD_CCtx_refPrefix(cctx, (const char*)dictBuffer, dictSize) );
CHECK_Z( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100)) );
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loading a prefix clears the cdict : ", testNb++);
{
ZSTD_CDict* const cdict = ZSTD_createCDict(dictBuffer, dictSize, 1);
CHECK_Z( ZSTD_CCtx_refCDict(cctx, cdict) );
CHECK_Z( ZSTD_CCtx_refPrefix(cctx, (const char*)dictBuffer, dictSize) );
CHECK_Z( ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100)) );
ZSTD_freeCDict(cdict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loaded dictionary persists across reset session : ", testNb++);
{
size_t size1, size2;
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, CNBuffer, MIN(CNBuffSize, 10 KB)) );
size1 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100 KB));
if (ZSTD_isError(size1)) goto _output_error;
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
size2 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100 KB));
if (ZSTD_isError(size2)) goto _output_error;
if (size1 != size2) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Loaded dictionary is cleared after resetting parameters : ", testNb++);
{
size_t size1, size2;
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, CNBuffer, MIN(CNBuffSize, 10 KB)) );
size1 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100 KB));
if (ZSTD_isError(size1)) goto _output_error;
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
size2 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100 KB));
if (ZSTD_isError(size2)) goto _output_error;
if (size1 == size2) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_CCtx_loadDictionary(cctx, dictBuffer, dictSize) );
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, MIN(CNBuffSize, 100 KB));
CHECK_Z(cSize);
DISPLAYLEVEL(3, "test%3i : ZSTD_decompressDCtx() with dictionary : ", testNb++);
{
ZSTD_DCtx* dctx = ZSTD_createDCtx();
size_t ret;
/* We should fail to decompress without a dictionary. */
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
ret = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (!ZSTD_isError(ret)) goto _output_error;
/* We should succeed to decompress with the dictionary. */
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_DCtx_loadDictionary(dctx, dictBuffer, dictSize) );
CHECK_Z( ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize) );
/* The dictionary should persist across calls. */
CHECK_Z( ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize) );
/* When we reset the context the dictionary is cleared. */
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
ret = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (!ZSTD_isError(ret)) goto _output_error;
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_decompressDCtx() with ddict : ", testNb++);
{
ZSTD_DCtx* dctx = ZSTD_createDCtx();
ZSTD_DDict* ddict = ZSTD_createDDict(dictBuffer, dictSize);
size_t ret;
/* We should succeed to decompress with the ddict. */
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_DCtx_refDDict(dctx, ddict) );
CHECK_Z( ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize) );
/* The ddict should persist across calls. */
CHECK_Z( ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize) );
/* When we reset the context the ddict is cleared. */
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
ret = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (!ZSTD_isError(ret)) goto _output_error;
ZSTD_freeDCtx(dctx);
ZSTD_freeDDict(ddict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_decompressDCtx() with prefix : ", testNb++);
{
ZSTD_DCtx* dctx = ZSTD_createDCtx();
size_t ret;
/* We should succeed to decompress with the prefix. */
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_DCtx_refPrefix_advanced(dctx, dictBuffer, dictSize, ZSTD_dct_auto) );
CHECK_Z( ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize) );
/* The prefix should be cleared after the first compression. */
ret = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (!ZSTD_isError(ret)) goto _output_error;
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_fast attach dictionary with hashLog = 25 and chainLog = 25 : ", testNb++);
{
ZSTD_CCtx_params* cctxParams = ZSTD_createCCtxParams();
ZSTD_customMem customMem = {NULL, NULL, NULL};
ZSTD_DCtx* dctx = ZSTD_createDCtx();
ZSTD_CDict* cdict;
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_strategy, ZSTD_fast));
/* Set windowLog to 25 so hash/chain logs don't get sized down */
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_windowLog, 25));
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_hashLog, 25));
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_chainLog, 25));
/* Set srcSizeHint to 2^25 so hash/chain logs don't get sized down */
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_srcSizeHint, 1u << 25));
cdict = ZSTD_createCDict_advanced2(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, cctxParams, customMem);
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_forceAttachDict, ZSTD_dictForceAttach));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_refCDict(cctx, cdict));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize);
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, dictBuffer, dictSize));
ZSTD_freeCDict(cdict);
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtxParams(cctxParams);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_dfast attach dictionary with hashLog = 25 and chainLog = 25 : ", testNb++);
{
ZSTD_CCtx_params* cctxParams = ZSTD_createCCtxParams();
ZSTD_customMem customMem = {NULL, NULL, NULL};
ZSTD_DCtx* dctx = ZSTD_createDCtx();
ZSTD_CDict* cdict;
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_strategy, ZSTD_dfast));
/* Set windowLog to 25 so hash/chain logs don't get sized down */
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_windowLog, 25));
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_hashLog, 25));
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_chainLog, 25));
/* Set srcSizeHint to 2^25 so hash/chain logs don't get sized down */
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_srcSizeHint, 1u << 25));
cdict = ZSTD_createCDict_advanced2(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, cctxParams, customMem);
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_forceAttachDict, ZSTD_dictForceAttach));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_refCDict(cctx, cdict));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize);
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, dictBuffer, dictSize));
ZSTD_freeCDict(cdict);
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtxParams(cctxParams);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_lazy attach dictionary with hashLog = 29 and searchLog = 4 : ", testNb++);
if (MEM_64bits()) {
ZSTD_CCtx_params* cctxParams = ZSTD_createCCtxParams();
ZSTD_customMem customMem = {NULL, NULL, NULL};
ZSTD_DCtx* dctx = ZSTD_createDCtx();
ZSTD_CDict* cdict;
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_strategy, ZSTD_lazy));
/* Force enable row based match finder, and disable dedicated dict search. */
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_useRowMatchFinder, ZSTD_ps_enable));
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_enableDedicatedDictSearch, 0));
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_searchLog, 4));
/* Set windowLog to 29 so hash/chain logs don't get sized down */
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_windowLog, 29));
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_hashLog, 29));
/* Set srcSizeHint to 2^29 so hash/chain logs don't get sized down */
CHECK_Z(ZSTD_CCtxParams_setParameter(cctxParams, ZSTD_c_srcSizeHint, 1u << 29));
cdict = ZSTD_createCDict_advanced2(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, cctxParams, customMem);
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_forceAttachDict, ZSTD_dictForceAttach));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_refCDict(cctx, cdict));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize);
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, dictBuffer, dictSize));
ZSTD_freeCDict(cdict);
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtxParams(cctxParams);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Dictionary with non-default repcodes : ", testNb++);
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
dictSize = ZDICT_trainFromBuffer(dictBuffer, dictSize,
CNBuffer, samplesSizes, nbSamples);
if (ZDICT_isError(dictSize)) goto _output_error;
/* Set all the repcodes to non-default */
{
BYTE* dictPtr = (BYTE*)dictBuffer;
BYTE* dictLimit = dictPtr + dictSize - 12;
/* Find the repcodes */
while (dictPtr < dictLimit &&
(MEM_readLE32(dictPtr) != 1 || MEM_readLE32(dictPtr + 4) != 4 ||
MEM_readLE32(dictPtr + 8) != 8)) {
++dictPtr;
}
if (dictPtr >= dictLimit) goto _output_error;
MEM_writeLE32(dictPtr + 0, 10);
MEM_writeLE32(dictPtr + 4, 10);
MEM_writeLE32(dictPtr + 8, 10);
/* Set the last 8 bytes to 'x' */
memset((BYTE*)dictBuffer + dictSize - 8, 'x', 8);
}
/* The optimal parser checks all the repcodes.
* Make sure at least one is a match >= targetLength so that it is
* immediately chosen. This will make sure that the compressor and
* decompressor agree on at least one of the repcodes.
*/
{ size_t dSize;
BYTE data[1024];
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
ZSTD_compressionParameters const cParams = ZSTD_getCParams(19, CNBuffSize, dictSize);
ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dictBuffer, dictSize,
ZSTD_dlm_byRef, ZSTD_dct_auto,
cParams, ZSTD_defaultCMem);
assert(dctx != NULL); assert(cdict != NULL);
memset(data, 'x', sizeof(data));
cSize = ZSTD_compress_usingCDict(cctx, compressedBuffer, compressedBufferSize,
data, sizeof(data), cdict);
ZSTD_freeCDict(cdict);
if (ZSTD_isError(cSize)) { DISPLAYLEVEL(5, "Compression error %s : ", ZSTD_getErrorName(cSize)); goto _output_error; }
dSize = ZSTD_decompress_usingDict(dctx, decodedBuffer, sizeof(data), compressedBuffer, cSize, dictBuffer, dictSize);
if (ZSTD_isError(dSize)) { DISPLAYLEVEL(5, "Decompression error %s : ", ZSTD_getErrorName(dSize)); goto _output_error; }
if (memcmp(data, decodedBuffer, sizeof(data))) { DISPLAYLEVEL(5, "Data corruption : "); goto _output_error; }
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
testNb = test_sessionReset(testNb);
testNb = test_multipleDDicts(testNb);
DISPLAYLEVEL(3, "test%3i : ZSTD_decompressDCtx() with multiple ddicts : ", testNb++);
{
const size_t numDicts = 128;
const size_t numFrames = 4;
size_t i;
ZSTD_DCtx* dctx = ZSTD_createDCtx();
ZSTD_DDict** ddictTable = (ZSTD_DDict**)malloc(sizeof(ZSTD_DDict*)*numDicts);
ZSTD_CDict** cdictTable = (ZSTD_CDict**)malloc(sizeof(ZSTD_CDict*)*numDicts);
U32 dictIDSeed = seed;
/* Create new compressed buffer that will hold frames with differing dictIDs */
char* dictBufferMulti = (char*)malloc(sizeof(char) * dictBufferFixedSize); /* Modifiable copy of fixed full dict buffer */
ZSTD_memcpy(dictBufferMulti, dictBufferFixed, dictBufferFixedSize);
/* Create a bunch of DDicts with random dict IDs */
for (i = 0; i < numDicts; ++i) {
U32 currDictID = FUZ_rand(&dictIDSeed);
MEM_writeLE32(dictBufferMulti+ZSTD_FRAMEIDSIZE, currDictID);
ddictTable[i] = ZSTD_createDDict(dictBufferMulti, dictBufferFixedSize);
cdictTable[i] = ZSTD_createCDict(dictBufferMulti, dictBufferFixedSize, 3);
if (!ddictTable[i] || !cdictTable[i] || ZSTD_getDictID_fromCDict(cdictTable[i]) != ZSTD_getDictID_fromDDict(ddictTable[i])) {
goto _output_error;
}
}
/* Compress a few frames using random CDicts */
{
size_t off = 0;
/* only use the first half so we don't push against size limit of compressedBuffer */
size_t const segSize = (CNBuffSize / 2) / numFrames;
for (i = 0; i < numFrames; i++) {
size_t dictIdx = FUZ_rand(&dictIDSeed) % numDicts;
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
{ CHECK_NEWV(r, ZSTD_compress_usingCDict(cctx,
(BYTE*)compressedBuffer + off, CNBuffSize - off,
(BYTE*)CNBuffer + segSize * (size_t)i, segSize,
cdictTable[dictIdx]));
off += r;
}
}
cSize = off;
}
/* We should succeed to decompression even though different dicts were used on different frames */
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, ZSTD_rmd_refMultipleDDicts);
/* Reference every single ddict we made */
for (i = 0; i < numDicts; ++i) {
CHECK_Z( ZSTD_DCtx_refDDict(dctx, ddictTable[i]));
}
CHECK_Z( ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize) );
/* Streaming decompression should also work */
{
ZSTD_inBuffer in = {compressedBuffer, cSize, 0};
ZSTD_outBuffer out = {decodedBuffer, CNBuffSize, 0};
while (in.pos < in.size) {
CHECK_Z(ZSTD_decompressStream(dctx, &out, &in));
}
}
ZSTD_freeDCtx(dctx);
for (i = 0; i < numDicts; ++i) {
ZSTD_freeCDict(cdictTable[i]);
ZSTD_freeDDict(ddictTable[i]);
}
free(dictBufferMulti);
free(ddictTable);
free(cdictTable);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : DDict hashset OOB T283341343 (ASAN) : ", testNb++);
{
/* Simplified repro for T283341343 - relies on ASAN to catch OOB read.
* Original bug: probe does idx &= mask; idx++ -> idx can become 64 OOB.
* IDs 3 and 47 both hash XXH64(id) & 63 == 63 (last slot).
* Setup: register dict 3 (occupies slot 63), then decompress frame
* whose header dictID is patched to 47 (collides to same slot).
* Buggy getDDict reads slot 64 OOB. With ASAN=1 this is reported as
* heap-buffer-overflow even though glibc would see 0 and return
* "dictionary mismatch". No custom allocator needed.
* See P2447503582 mode 0.
*/
U32 victimID;
U32 attackerID;
size_t victimIdx;
size_t attackerIdx;
ZSTD_DCtx* dctx2;
ZSTD_CCtx* cctx2;
char* dictBufVictim;
ZSTD_DDict* victimDDict;
size_t srcSize;
BYTE* frame;
BYTE fhd;
int dictIDSizeCode;
int nb;
int singleSegment;
int pos;
int i;
victimID = 3;
attackerID = 47;
victimIdx = XXH64(&victimID, sizeof(victimID), 0) & 63;
attackerIdx = XXH64(&attackerID, sizeof(attackerID), 0) & 63;
if (victimIdx != 63 || attackerIdx != 63) {
DISPLAY("hash assumption broken\n");
goto _output_error;
}
dctx2 = ZSTD_createDCtx();
cctx2 = ZSTD_createCCtx();
dictBufVictim = (char*)malloc(dictBufferFixedSize);
victimDDict = NULL;
if (!dctx2 || !cctx2 || !dictBufVictim) {
DISPLAY("alloc failed\n");
goto _oob_error2;
}
ZSTD_memcpy(dictBufVictim, dictBufferFixed, dictBufferFixedSize);
MEM_writeLE32(dictBufVictim + ZSTD_FRAMEIDSIZE, victimID);
victimDDict = ZSTD_createDDict(dictBufVictim, dictBufferFixedSize);
if (!victimDDict) {
DISPLAY("DDict creation failed\n");
goto _oob_error2;
}
CHECK_Z( ZSTD_DCtx_setParameter(dctx2, ZSTD_d_refMultipleDDicts, ZSTD_rmd_refMultipleDDicts) );
CHECK_Z( ZSTD_DCtx_refDDict(dctx2, victimDDict) );
ZSTD_CCtx_reset(cctx2, ZSTD_reset_session_and_parameters);
srcSize = MIN(CNBuffSize, 1 KB);
cSize = ZSTD_compress_usingDict(cctx2, compressedBuffer, compressedBufferSize,
CNBuffer, srcSize,
dictBufVictim, dictBufferFixedSize, 3);
if (ZSTD_isError(cSize)) {
DISPLAY("compress_usingDict failed %s\n", ZSTD_getErrorName(cSize));
goto _oob_error2;
}
/* Patch dictID in frame header to attackerID (47) */
frame = (BYTE*)compressedBuffer;
fhd = frame[4];
dictIDSizeCode = fhd & 3;
nb = (dictIDSizeCode == 3) ? 4 : dictIDSizeCode;
singleSegment = (fhd >> 5) & 1;
pos = 4 + 1 + (singleSegment ? 0 : 1);
if (nb == 0) {
DISPLAY("frame has no dictID field\n");
goto _oob_error2;
}
for (i = 0; i < nb; i++) frame[pos + i] = (BYTE)(attackerID >> (8 * i));
/* With ASAN=1 buggy code triggers heap-buffer-overflow here.
* Without ASAN it just returns dictionary_wrong, which is OK for fixed code.
* So this test always passes without ASAN, but fails under ASAN if bug present.
*/
{
size_t const ret = ZSTD_decompressDCtx(dctx2, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (!ZSTD_isError(ret)) {
DISPLAY("unexpected success %zu (should be dict mismatch)\n", ret);
goto _oob_error2;
}
/* Expected: dictionary_wrong / mismatch -> OK means OOB was not taken as success,
* but ASAN would have already aborted if OOB read happened. */
}
ZSTD_freeDDict(victimDDict);
ZSTD_freeDCtx(dctx2);
ZSTD_freeCCtx(cctx2);
free(dictBufVictim);
goto _oob_skip2;
_oob_error2:
if (victimDDict) ZSTD_freeDDict(victimDDict);
if (dctx2) ZSTD_freeDCtx(dctx2);
if (cctx2) ZSTD_freeCCtx(cctx2);
if (dictBufVictim) free(dictBufVictim);
goto _output_error;
_oob_skip2: ;
}
DISPLAYLEVEL(3, "OK (ASAN would report OOB read if vulnerable)\n");
DISPLAYLEVEL(3, "test%3i : DDict hashset OOB write T283341343 (ASAN) : ", testNb++);
{
/* Write path version of above: same colliding IDs 3 and 47 → slot 63.
* Insertion does while(table[idx]!=NULL) { idx&=mask; idx++; }
* Second insertion with colliding ID should wrap to slot 0, but buggy code
* writes to slot 64 OOB (8-byte write). With ASAN this is heap-buffer-overflow
* on write. Without ASAN glibc slack hides it.
*/
U32 victimID2;
U32 attackerID2;
size_t vIdx2;
size_t aIdx2;
ZSTD_DCtx* dctx3;
char* dictBufV2;
char* dictBufA2;
ZSTD_DDict* victimDDict2;
ZSTD_DDict* attackerDDict2;
victimID2 = 3;
attackerID2 = 47;
vIdx2 = XXH64(&victimID2, sizeof(victimID2), 0) & 63;
aIdx2 = XXH64(&attackerID2, sizeof(attackerID2), 0) & 63;
if (vIdx2 != 63 || aIdx2 != 63) {
DISPLAY("hash assumption broken for write test\n");
goto _output_error;
}
dctx3 = ZSTD_createDCtx();
dictBufV2 = (char*)malloc(dictBufferFixedSize);
dictBufA2 = (char*)malloc(dictBufferFixedSize);
victimDDict2 = NULL;
attackerDDict2 = NULL;
if (!dctx3 || !dictBufV2 || !dictBufA2) {
DISPLAY("alloc failed for OOB write test\n");
goto _oob_write_error2;
}
ZSTD_memcpy(dictBufV2, dictBufferFixed, dictBufferFixedSize);
ZSTD_memcpy(dictBufA2, dictBufferFixed, dictBufferFixedSize);
MEM_writeLE32(dictBufV2 + ZSTD_FRAMEIDSIZE, victimID2);
MEM_writeLE32(dictBufA2 + ZSTD_FRAMEIDSIZE, attackerID2);
victimDDict2 = ZSTD_createDDict(dictBufV2, dictBufferFixedSize);
attackerDDict2 = ZSTD_createDDict(dictBufA2, dictBufferFixedSize);
if (!victimDDict2 || !attackerDDict2) {
DISPLAY("DDict creation failed for write test\n");
goto _oob_write_error2;
}
CHECK_Z( ZSTD_DCtx_setParameter(dctx3, ZSTD_d_refMultipleDDicts, ZSTD_rmd_refMultipleDDicts) );
CHECK_Z( ZSTD_DCtx_refDDict(dctx3, victimDDict2) );
/* Second colliding insertion - triggers OOB write in buggy emplaceDDict */
CHECK_Z( ZSTD_DCtx_refDDict(dctx3, attackerDDict2) );
ZSTD_freeDDict(victimDDict2);
ZSTD_freeDDict(attackerDDict2);
ZSTD_freeDCtx(dctx3);
free(dictBufV2);
free(dictBufA2);
goto _oob_write_skip2;
_oob_write_error2:
if (victimDDict2) ZSTD_freeDDict(victimDDict2);
if (attackerDDict2) ZSTD_freeDDict(attackerDDict2);
if (dctx3) ZSTD_freeDCtx(dctx3);
if (dictBufV2) free(dictBufV2);
if (dictBufA2) free(dictBufA2);
goto _output_error;
_oob_write_skip2: ;
}
DISPLAYLEVEL(3, "OK (ASAN would report OOB write if vulnerable)\n");
DISPLAYLEVEL(3, "test%3i : DDict hashset is dropped by reset and refDDict(NULL) : ", testNb++);
{ /* After a parameter reset or refDDict(NULL), a frame must not select
* a DDict referenced before, which the caller may have freed since */
char* const dictBufA = (char*)malloc(dictBufferFixedSize);
ZSTD_DDict* const ddictB = ZSTD_createDDict(dictBufferFixed, dictBufferFixedSize);
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
size_t frameSize;
int n;
if (!dictBufA || !ddictB || !dctx) goto _output_error;
ZSTD_memcpy(dictBufA, dictBufferFixed, dictBufferFixedSize);
MEM_writeLE32(dictBufA + ZSTD_FRAMEIDSIZE, ZSTD_getDictID_fromDDict(ddictB) + 1);
frameSize = ZSTD_compress_usingDict(cctx, compressedBuffer, compressedBufferSize,
CNBuffer, 100, dictBufA, dictBufferFixedSize, 1);
CHECK_Z(frameSize);
for (n = 0; n < 2; n++) {
ZSTD_DDict* const ddictA = ZSTD_createDDict(dictBufA, dictBufferFixedSize);
if (!ddictA) goto _output_error;
CHECK_Z( ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, ZSTD_rmd_refMultipleDDicts) );
CHECK_Z( ZSTD_DCtx_refDDict(dctx, ddictA) );
if (n == 0) {
CHECK_Z( ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters) );
} else {
CHECK_Z( ZSTD_DCtx_refDDict(dctx, NULL) );
}
ZSTD_freeDDict(ddictA);
/* only ddictB is referenced now : the frame needs ddictA, so it must fail */
CHECK_Z( ZSTD_DCtx_setParameter(dctx, ZSTD_d_refMultipleDDicts, ZSTD_rmd_refMultipleDDicts) );
CHECK_Z( ZSTD_DCtx_refDDict(dctx, ddictB) );
if (!ZSTD_isError(ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, frameSize)))
goto _output_error;
}
ZSTD_freeDCtx(dctx);
ZSTD_freeDDict(ddictB);
free(dictBufA);
}
DISPLAYLEVEL(3, "OK \n");
ZSTD_freeCCtx(cctx);
free(dictBuffer);
free(samplesSizes);
}
/* COVER dictionary builder tests */
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t dictSize = 16 KB;
size_t optDictSize = dictSize;
void* dictBuffer = malloc(dictSize);
size_t const totalSampleSize = 1 MB;
size_t const sampleUnitSize = 8 KB;
U32 const nbSamples = (U32)(totalSampleSize / sampleUnitSize);
size_t* const samplesSizes = (size_t*) malloc(nbSamples * sizeof(size_t));
U32 seed32 = seed;
ZDICT_cover_params_t params;
U32 dictID;
if (dictBuffer==NULL || samplesSizes==NULL) {
free(dictBuffer);
free(samplesSizes);
goto _output_error;
}
DISPLAYLEVEL(3, "test%3i : ZDICT_trainFromBuffer_cover : ", testNb++);
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
memset(&params, 0, sizeof(params));
params.d = 1 + (FUZ_rand(&seed32) % 16);
params.k = params.d + (FUZ_rand(&seed32) % 256);
dictSize = ZDICT_trainFromBuffer_cover(dictBuffer, dictSize,
CNBuffer, samplesSizes, nbSamples,
params);
if (ZDICT_isError(dictSize)) goto _output_error;
DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)dictSize);
DISPLAYLEVEL(3, "test%3i : check dictID : ", testNb++);
dictID = ZDICT_getDictID(dictBuffer, dictSize);
if (dictID==0) goto _output_error;
DISPLAYLEVEL(3, "OK : %u \n", (unsigned)dictID);
DISPLAYLEVEL(3, "test%3i : ZDICT_optimizeTrainFromBuffer_cover : ", testNb++);
memset(&params, 0, sizeof(params));
params.steps = 4;
optDictSize = ZDICT_optimizeTrainFromBuffer_cover(dictBuffer, optDictSize,
CNBuffer, samplesSizes,
nbSamples / 4, &params);
if (ZDICT_isError(optDictSize)) goto _output_error;
DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)optDictSize);
DISPLAYLEVEL(3, "test%3i : check dictID : ", testNb++);
dictID = ZDICT_getDictID(dictBuffer, optDictSize);
if (dictID==0) goto _output_error;
DISPLAYLEVEL(3, "OK : %u \n", (unsigned)dictID);
DISPLAYLEVEL(3, "test%3i : ZDICT_optimizeTrainFromBuffer_cover with tiny training set : ", testNb++);
{ /* With splitPoint < 1.0 the training set is a subset of the corpus.
* If that subset is smaller than MAX(d, sizeof(U64)) the partial
* suffix array size used to underflow size_t (issue #4682). The
* builder must reject it cleanly instead. The first few samples are
* tiny so the 0.75 training split stays below the threshold, while
* the total corpus is comfortably large. */
size_t tinySizes[8] = { 1, 1, 1, 1, 1, 1, 100, 100 };
ZDICT_cover_params_t tinyParams;
size_t tinyResult;
memset(&tinyParams, 0, sizeof(tinyParams));
tinyParams.steps = 4;
tinyParams.splitPoint = 0.75;
tinyResult = ZDICT_optimizeTrainFromBuffer_cover(
dictBuffer, optDictSize, CNBuffer, tinySizes, 8, &tinyParams);
if (!ZDICT_isError(tinyResult)) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
ZSTD_freeCCtx(cctx);
free(dictBuffer);
free(samplesSizes);
}
/* Decompression defense tests */
DISPLAYLEVEL(3, "test%3i : Check input length for magic number : ", testNb++);
{ size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, CNBuffer, 3); /* too small input */
if (!ZSTD_isError(r)) goto _output_error;
if (ZSTD_getErrorCode(r) != ZSTD_error_srcSize_wrong) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Check magic Number : ", testNb++);
((char*)(CNBuffer))[0] = 1;
{ size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, CNBuffer, 4);
if (!ZSTD_isError(r)) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
/* content size verification test */
DISPLAYLEVEL(3, "test%3i : Content size verification : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const srcSize = 5000;
size_t const wrongSrcSize = (srcSize + 1000);
ZSTD_parameters params = ZSTD_getParams(1, wrongSrcSize, 0);
params.fParams.contentSizeFlag = 1;
CHECK_Z( ZSTD_compressBegin_advanced(cctx, NULL, 0, params, wrongSrcSize) );
{ size_t const result = ZSTD_compressEnd(cctx, decodedBuffer, CNBuffSize, CNBuffer, srcSize);
if (!ZSTD_isError(result)) goto _output_error;
if (ZSTD_getErrorCode(result) != ZSTD_error_srcSize_wrong) goto _output_error;
DISPLAYLEVEL(3, "OK : %s \n", ZSTD_getErrorName(result));
}
ZSTD_freeCCtx(cctx);
}
/* negative compression level test : ensure simple API and advanced API produce same result */
DISPLAYLEVEL(3, "test%3i : negative compression level : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const srcSize = CNBuffSize / 5;
int const compressionLevel = -1;
assert(cctx != NULL);
{ size_t const cSize_1pass = ZSTD_compress(compressedBuffer, compressedBufferSize,
CNBuffer, srcSize, compressionLevel);
if (ZSTD_isError(cSize_1pass)) goto _output_error;
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, compressionLevel) );
{ size_t const compressionResult = ZSTD_compress2(cctx,
compressedBuffer, compressedBufferSize,
CNBuffer, srcSize);
DISPLAYLEVEL(5, "simple=%u vs %u=advanced : ", (unsigned)cSize_1pass, (unsigned)compressionResult);
if (ZSTD_isError(compressionResult)) goto _output_error;
if (compressionResult != cSize_1pass) goto _output_error;
} }
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
/* parameters order test */
{ size_t const inputSize = CNBuffSize / 2;
U64 xxh64;
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
DISPLAYLEVEL(3, "test%3i : parameters in order : ", testNb++);
assert(cctx != NULL);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 2) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, 18) );
{ size_t const compressedSize = ZSTD_compress2(cctx,
compressedBuffer, ZSTD_compressBound(inputSize),
CNBuffer, inputSize);
CHECK_Z(compressedSize);
cSize = compressedSize;
xxh64 = XXH64(compressedBuffer, compressedSize, 0);
}
DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (unsigned)inputSize, (unsigned)cSize);
ZSTD_freeCCtx(cctx);
}
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
DISPLAYLEVEL(3, "test%3i : parameters disordered : ", testNb++);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, 18) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 2) );
{ size_t const result = ZSTD_compress2(cctx,
compressedBuffer, ZSTD_compressBound(inputSize),
CNBuffer, inputSize);
CHECK_Z(result);
if (result != cSize) goto _output_error; /* must result in same compressed result, hence same size */
if (XXH64(compressedBuffer, result, 0) != xxh64) goto _output_error; /* must result in exactly same content, hence same hash */
DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (unsigned)inputSize, (unsigned)result);
}
ZSTD_freeCCtx(cctx);
}
}
DISPLAYLEVEL(3, "test%3i : oversized round buffer must fail cleanly on 32-bit : ", testNb++);
/* On 32-bit builds the round buffer size can exceed 4 GiB, which
* wraps in 32-bit size_t. Without the overflow check in
* ZSTDMT_initCStream_internal(), init succeeds with a buffer that
* is too small for the requested configuration (and if it wraps
* to 0, the first streaming copy overruns it). With the check,
* the first ZSTD_compressStream2() call fails cleanly with
* memory_allocation and nothing is copied. Both cases are
* unreachable on 64-bit builds, where they get a real buffer.
* - case 0: windowLog 30 with btultra2 (level 22) and LDM disabled
* makes the overlap prefix 1 GiB, which bumps the job size to
* 1 GiB; 1 worker needs 4 GiB, which wraps to 0.
* - case 1: 16 workers with 256 MB jobs need 4.75 GiB, which
* wraps to 768 MB. */
if (MEM_32bits()) {
#ifdef ZSTD_MULTITHREAD
struct { int nbWorkers; int windowLog; int jobSize; } const cases[2] = {
{ 1, 30, 0 },
{ 16, 27, 256 MB },
};
size_t caseNb;
for (caseNb = 0; caseNb < 2; caseNb++) {
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_outBuffer out = { compressedBuffer, compressedBufferSize, 0 };
ZSTD_inBuffer in = { CNBuffer, CNBuffSize, 0 };
size_t r;
CHECK(cctx != NULL);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, cases[caseNb].nbWorkers) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, ZSTD_maxCLevel()) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, cases[caseNb].windowLog) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_jobSize, cases[caseNb].jobSize) );
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_disable) );
r = ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue);
ZSTD_freeCCtx(cctx);
if (ZSTD_getErrorCode(r) != ZSTD_error_memory_allocation) goto _output_error;
}
#endif
}
DISPLAYLEVEL(3, "OK \n");
/* advanced parameters for decompression */
{ ZSTD_DCtx* const dctx = ZSTD_createDCtx();
assert(dctx != NULL);
DISPLAYLEVEL(3, "test%3i : get dParameter bounds ", testNb++);
{ ZSTD_bounds const bounds = ZSTD_dParam_getBounds(ZSTD_d_windowLogMax);
CHECK_Z(bounds.error);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : wrong dParameter : ", testNb++);
{ size_t const sr = ZSTD_DCtx_setParameter(dctx, (ZSTD_dParameter)999999, 0);
if (!ZSTD_isError(sr)) goto _output_error;
}
{ ZSTD_bounds const bounds = ZSTD_dParam_getBounds((ZSTD_dParameter)999998);
if (!ZSTD_isError(bounds.error)) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : out of bound dParameter : ", testNb++);
{ size_t const sr = ZSTD_DCtx_setParameter(dctx, ZSTD_d_windowLogMax, 9999);
if (!ZSTD_isError(sr)) goto _output_error;
}
{ size_t const sr = ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, (ZSTD_format_e)888);
if (!ZSTD_isError(sr)) goto _output_error;
}
DISPLAYLEVEL(3, "OK \n");
ZSTD_freeDCtx(dctx);
}
/* custom formats tests */
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
size_t const inputSize = CNBuffSize / 2; /* won't cause pb with small dict size */
assert(dctx != NULL); assert(cctx != NULL);
/* basic block compression */
DISPLAYLEVEL(3, "test%3i : magic-less format test : ", testNb++);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_format, ZSTD_f_zstd1_magicless) );
{ ZSTD_inBuffer in = { CNBuffer, inputSize, 0 };
ZSTD_outBuffer out = { compressedBuffer, ZSTD_compressBound(inputSize), 0 };
size_t const result = ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end);
if (result != 0) goto _output_error;
if (in.pos != in.size) goto _output_error;
cSize = out.pos;
}
DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (unsigned)inputSize, (unsigned)cSize);
DISPLAYLEVEL(3, "test%3i : decompress normally (should fail) : ", testNb++);
{ size_t const decodeResult = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (ZSTD_getErrorCode(decodeResult) != ZSTD_error_prefix_unknown) goto _output_error;
DISPLAYLEVEL(3, "OK : %s \n", ZSTD_getErrorName(decodeResult));
}
DISPLAYLEVEL(3, "test%3i : decompress of magic-less frame : ", testNb++);
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless) );
{ ZSTD_FrameHeader zfh;
size_t const zfhrt = ZSTD_getFrameHeader_advanced(&zfh, compressedBuffer, cSize, ZSTD_f_zstd1_magicless);
if (zfhrt != 0) goto _output_error;
}
/* one shot */
{ size_t const result = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (result != inputSize) goto _output_error;
DISPLAYLEVEL(3, "one-shot OK, ");
}
/* streaming */
{ ZSTD_inBuffer in = { compressedBuffer, cSize, 0 };
ZSTD_outBuffer out = { decodedBuffer, CNBuffSize, 0 };
size_t const result = ZSTD_decompressStream(dctx, &out, &in);
if (result != 0) goto _output_error;
if (in.pos != in.size) goto _output_error;
if (out.pos != inputSize) goto _output_error;
DISPLAYLEVEL(3, "streaming OK : regenerated %u bytes \n", (unsigned)out.pos);
}
/* basic block compression */
DISPLAYLEVEL(3, "test%3i : empty magic-less format test : ", testNb++);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_format, ZSTD_f_zstd1_magicless) );
{ ZSTD_inBuffer in = { CNBuffer, 0, 0 };
ZSTD_outBuffer out = { compressedBuffer, ZSTD_compressBound(0), 0 };
size_t const result = ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end);
if (result != 0) goto _output_error;
if (in.pos != in.size) goto _output_error;
cSize = out.pos;
}
DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (unsigned)0, (unsigned)cSize);
DISPLAYLEVEL(3, "test%3i : decompress of empty magic-less frame : ", testNb++);
ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
CHECK_Z( ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless) );
/* one shot */
{ size_t const result = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
if (result != 0) goto _output_error;
DISPLAYLEVEL(3, "one-shot OK, ");
}
/* streaming */
{ ZSTD_inBuffer in = { compressedBuffer, cSize, 0 };
ZSTD_outBuffer out = { decodedBuffer, CNBuffSize, 0 };
size_t const result = ZSTD_decompressStream(dctx, &out, &in);
if (result != 0) goto _output_error;
if (in.pos != in.size) goto _output_error;
if (out.pos != 0) goto _output_error;
DISPLAYLEVEL(3, "streaming OK : regenerated %u bytes \n", (unsigned)out.pos);
}
ZSTD_freeCCtx(cctx);
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "test%3i : Decompression parameter reset test : ", testNb++);
{
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
/* Attempt to future proof this to new parameters. */
int const maxParam = 2000;
int param;
if (ZSTD_d_experimentalParam3 > maxParam) goto _output_error;
for (param = 0; param < maxParam; ++param) {
ZSTD_dParameter dParam = (ZSTD_dParameter)param;
ZSTD_bounds bounds = ZSTD_dParam_getBounds(dParam);
int value1;
int value2;
int check;
if (ZSTD_isError(bounds.error))
continue;
CHECK_Z(ZSTD_DCtx_getParameter(dctx, dParam, &value1));
value2 = (value1 != bounds.lowerBound) ? bounds.lowerBound : bounds.upperBound;
CHECK_Z(ZSTD_DCtx_setParameter(dctx, dParam, value2));
CHECK_Z(ZSTD_DCtx_getParameter(dctx, dParam, &check));
if (check != value2) goto _output_error;
CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_parameters));
CHECK_Z(ZSTD_DCtx_getParameter(dctx, dParam, &check));
if (check != value1) goto _output_error;
}
ZSTD_freeDCtx(dctx);
}
DISPLAYLEVEL(3, "OK \n");
/* block API tests */
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
static const size_t dictSize = 65 KB;
static const size_t blockSize = 100 KB; /* won't cause pb with small dict size */
size_t cSize2;
assert(cctx != NULL); assert(dctx != NULL);
/* basic block compression */
DISPLAYLEVEL(3, "test%3i : Block compression test : ", testNb++);
CHECK_Z( ZSTD_compressBegin(cctx, 5) );
CHECK_Z( ZSTD_getBlockSize(cctx) >= blockSize);
CHECK_VAR(cSize, ZSTD_compressBlock(cctx, compressedBuffer, ZSTD_compressBound(blockSize), CNBuffer, blockSize) );
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Block decompression test : ", testNb++);
CHECK_Z( ZSTD_decompressBegin(dctx) );
{ CHECK_NEWV(r, ZSTD_decompressBlock(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize) );
if (r != blockSize) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
/* very long stream of block compression */
DISPLAYLEVEL(3, "test%3i : Huge block streaming compression test : ", testNb++);
CHECK_Z( ZSTD_compressBegin(cctx, -199) ); /* we just want to quickly overflow internal U32 index */
CHECK_Z( ZSTD_getBlockSize(cctx) >= blockSize);
{ U64 const toCompress = 5000000000ULL; /* > 4 GB */
U64 compressed = 0;
while (compressed < toCompress) {
size_t const blockCSize = ZSTD_compressBlock(cctx, compressedBuffer, ZSTD_compressBound(blockSize), CNBuffer, blockSize);
assert(blockCSize != 0);
if (ZSTD_isError(blockCSize)) goto _output_error;
compressed += blockCSize;
} }
DISPLAYLEVEL(3, "OK \n");
/* dictionary block compression */
DISPLAYLEVEL(3, "test%3i : Dictionary Block compression test : ", testNb++);
CHECK_Z( ZSTD_compressBegin_usingDict(cctx, CNBuffer, dictSize, 5) );
CHECK_VAR(cSize, ZSTD_compressBlock(cctx, compressedBuffer, ZSTD_compressBound(blockSize), (char*)CNBuffer+dictSize, blockSize));
RDG_genBuffer((char*)CNBuffer+dictSize+blockSize, blockSize, 0.0, 0.0, seed); /* create a non-compressible second block */
{ CHECK_NEWV(r, ZSTD_compressBlock(cctx, (char*)compressedBuffer+cSize, ZSTD_compressBound(blockSize), (char*)CNBuffer+dictSize+blockSize, blockSize) ); /* for cctx history consistency */
assert(r == 0); /* non-compressible block */ }
memcpy((char*)compressedBuffer+cSize, (char*)CNBuffer+dictSize+blockSize, blockSize); /* send non-compressed block (without header) */
CHECK_VAR(cSize2, ZSTD_compressBlock(cctx, (char*)compressedBuffer+cSize+blockSize, ZSTD_compressBound(blockSize),
(char*)CNBuffer+dictSize+2*blockSize, blockSize));
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Dictionary Block decompression test : ", testNb++);
CHECK_Z( ZSTD_decompressBegin_usingDict(dctx, CNBuffer, dictSize) );
{ CHECK_NEWV( r, ZSTD_decompressBlock(dctx, decodedBuffer, blockSize, compressedBuffer, cSize) );
if (r != blockSize) {
DISPLAYLEVEL(1, "ZSTD_decompressBlock() with _usingDict() fails : %u, instead of %u expected \n", (unsigned)r, (unsigned)blockSize);
goto _output_error;
} }
memcpy((char*)decodedBuffer+blockSize, (char*)compressedBuffer+cSize, blockSize);
ZSTD_insertBlock(dctx, (char*)decodedBuffer+blockSize, blockSize); /* insert non-compressed block into dctx history */
{ CHECK_NEWV( r, ZSTD_decompressBlock(dctx, (char*)decodedBuffer+2*blockSize, blockSize, (char*)compressedBuffer+cSize+blockSize, cSize2) );
if (r != blockSize) {
DISPLAYLEVEL(1, "ZSTD_decompressBlock() with _usingDict() and after insertBlock() fails : %u, instead of %u expected \n", (unsigned)r, (unsigned)blockSize);
goto _output_error;
} }
assert(memcpy((char*)CNBuffer+dictSize, decodedBuffer, blockSize*3)); /* ensure regenerated content is identical to origin */
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : Block compression with CDict : ", testNb++);
{ ZSTD_CDict* const cdict = ZSTD_createCDict(CNBuffer, dictSize, 3);
if (cdict==NULL) goto _output_error;
CHECK_Z( ZSTD_compressBegin_usingCDict(cctx, cdict) );
CHECK_Z( ZSTD_compressBlock(cctx, compressedBuffer, ZSTD_compressBound(blockSize), (char*)CNBuffer+dictSize, blockSize) );
ZSTD_freeCDict(cdict);
}
DISPLAYLEVEL(3, "OK \n");
ZSTD_freeCCtx(cctx);
ZSTD_freeDCtx(dctx);
}
/* rle detection test: must compress better blocks with a single identical byte repeated */
{ size_t sampleSize = 0;
size_t maxCompressedSize = 46; /* block 1, 2: compressed, block 3: RLE, zstd 1.4.4 */
DISPLAYLEVEL(3, "test%3i : RLE detection test : ", testNb++);
memset((char*)CNBuffer+sampleSize, 'B', 256 KB - 2);
sampleSize += 256 KB - 2;
memset((char*)CNBuffer+sampleSize, 'A', 100 KB);
sampleSize += 100 KB;
cSize = ZSTD_compress(compressedBuffer, ZSTD_compressBound(sampleSize), CNBuffer, sampleSize, 1);
if (ZSTD_isError(cSize) || cSize > maxCompressedSize) {
DISPLAYLEVEL(4, "error: cSize %u > %u expected ! \n", (unsigned)cSize, (unsigned)maxCompressedSize);
goto _output_error;
}
{ CHECK_NEWV(regenSize, ZSTD_decompress(decodedBuffer, sampleSize, compressedBuffer, cSize));
if (regenSize!=sampleSize) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
}
DISPLAYLEVEL(3, "test%3i : ZSTD_generateSequences decode from sequences test : ", testNb++);
{
size_t srcSize = 150 KB;
BYTE* src = (BYTE*)CNBuffer;
BYTE* decoded = (BYTE*)compressedBuffer;
ZSTD_CCtx* cctx = ZSTD_createCCtx();
ZSTD_Sequence* seqs = (ZSTD_Sequence*)malloc(srcSize * sizeof(ZSTD_Sequence));
size_t seqsSize;
if (seqs == NULL) goto _output_error;
assert(cctx != NULL);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 19);
/* Populate src with random data */
RDG_genBuffer(CNBuffer, srcSize, compressibility, 0.5, seed);
/* Test with block delimiters roundtrip */
seqsSize = ZSTD_generateSequences(cctx, seqs, srcSize, src, srcSize);
CHECK_Z(seqsSize);
FUZ_decodeSequences(decoded, seqs, seqsSize, src, srcSize, ZSTD_sf_explicitBlockDelimiters);
assert(!memcmp(CNBuffer, compressedBuffer, srcSize));
/* Test no block delimiters roundtrip */
seqsSize = ZSTD_mergeBlockDelimiters(seqs, seqsSize);
CHECK_Z(seqsSize);
FUZ_decodeSequences(decoded, seqs, seqsSize, src, srcSize, ZSTD_sf_noBlockDelimiters);
assert(!memcmp(CNBuffer, compressedBuffer, srcSize));
ZSTD_freeCCtx(cctx);
free(seqs);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_generateSequences too small output buffer : ", testNb++);
{
const size_t seqsCapacity = 10;
const size_t srcSize = 150 KB;
const BYTE* src = (BYTE*)CNBuffer;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_Sequence* const seqs = (ZSTD_Sequence*)malloc(seqsCapacity * sizeof(ZSTD_Sequence));
if (seqs == NULL) goto _output_error;
if (cctx == NULL) goto _output_error;
/* Populate src with random data */
RDG_genBuffer(CNBuffer, srcSize, compressibility, 0.5, seed);
/* Test with block delimiters roundtrip */
{
size_t const seqsSize = ZSTD_generateSequences(cctx, seqs, seqsCapacity, src, srcSize);
if (!ZSTD_isError(seqsSize)) goto _output_error;
}
ZSTD_freeCCtx(cctx);
free(seqs);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_getSequences followed by ZSTD_compressSequences : ", testNb++);
{
const size_t srcSize = 500 KB;
const BYTE* const src = (BYTE*)CNBuffer;
BYTE* const dst = (BYTE*)compressedBuffer;
const size_t dstCapacity = ZSTD_compressBound(srcSize);
const size_t decompressSize = srcSize;
char* const decompressBuffer = (char*)malloc(decompressSize);
size_t compressedSize;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_Sequence* const seqs = (ZSTD_Sequence*)malloc(srcSize * sizeof(ZSTD_Sequence));
size_t nbSeqs;
if (seqs == NULL) goto _output_error;
assert(cctx != NULL);
/* Populate src with compressible random data */
RDG_genBuffer(CNBuffer, srcSize, compressibility, 0., seed);
/* Roundtrip Test with block delimiters generated by ZSTD_generateSequences() */
nbSeqs = ZSTD_generateSequences(cctx, seqs, srcSize, src, srcSize);
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_blockDelimiters, ZSTD_sf_explicitBlockDelimiters);
compressedSize = ZSTD_compressSequences(cctx, dst, dstCapacity, seqs, nbSeqs, src, srcSize);
if (ZSTD_isError(compressedSize)) {
DISPLAY("Error in sequence compression with block delims\n");
goto _output_error;
}
{ size_t const dSize = ZSTD_decompress(decompressBuffer, decompressSize, dst, compressedSize);
if (ZSTD_isError(dSize)) {
DISPLAY("Error in sequence compression roundtrip with block delims\n");
goto _output_error;
} }
assert(!memcmp(decompressBuffer, src, srcSize));
/* Roundtrip Test with no block delimiters */
{ size_t const nbSeqsAfterMerge = ZSTD_mergeBlockDelimiters(seqs, nbSeqs);
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_blockDelimiters, ZSTD_sf_noBlockDelimiters);
compressedSize = ZSTD_compressSequences(cctx, dst, dstCapacity, seqs, nbSeqsAfterMerge, src, srcSize);
}
if (ZSTD_isError(compressedSize)) {
DISPLAY("Error in sequence compression with no block delims\n");
goto _output_error;
}
{ size_t const dSize = ZSTD_decompress(decompressBuffer, decompressSize, dst, compressedSize);
if (ZSTD_isError(dSize)) {
DISPLAY("Error in sequence compression roundtrip with no block delims\n");
goto _output_error;
} }
assert(!memcmp(decompressBuffer, src, srcSize));
ZSTD_freeCCtx(cctx);
free(decompressBuffer);
free(seqs);
}
DISPLAYLEVEL(3, "OK \n");
testNb = test_convertSequences_noRepcodes(seed, testNb);
testNb = test_get1BlockSummary(testNb);
testNb = test_blockSplitLimit(testNb);
DISPLAYLEVEL(3, "test%3i : ZSTD_compressSequencesAndLiterals : ", testNb++);
{
const size_t srcSize = 497000;
const BYTE* const src = (BYTE*)CNBuffer;
BYTE* const dst = (BYTE*)compressedBuffer;
const size_t dstCapacity = ZSTD_compressBound(srcSize);
const size_t decompressSize = srcSize;
char* const decompressBuffer = (char*)malloc(decompressSize);
char* const litBuffer = (char*)malloc(decompressSize);
size_t compressedSize;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_Sequence* const seqs = (ZSTD_Sequence*)malloc(srcSize * sizeof(ZSTD_Sequence));
size_t nbSeqs;
if (litBuffer == NULL) goto _output_error;
if (decompressBuffer == NULL) goto _output_error;
if (seqs == NULL) goto _output_error;
assert(cctx != NULL);
/* Populate src with compressible random data */
RDG_genBuffer(CNBuffer, srcSize, compressibility, 0., seed);
/* Roundtrip Test using the AndLiterals() variant */
nbSeqs = ZSTD_generateSequences(cctx, seqs, srcSize, src, srcSize);
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
ZSTD_CCtx_setParameter(cctx, ZSTD_c_blockDelimiters, ZSTD_sf_explicitBlockDelimiters);
{ size_t const litSize = FUZ_getLitSize(seqs, nbSeqs);
FUZ_transferLiterals(litBuffer, decompressSize, CNBuffer, srcSize, seqs, nbSeqs);
/* literal buffer capacity without 8 bytes of margin: must fail */
compressedSize = ZSTD_compressSequencesAndLiterals(cctx, dst, dstCapacity, seqs, nbSeqs, litBuffer, litSize, litSize+7, srcSize);
if (ZSTD_getErrorCode(compressedSize) != ZSTD_error_workSpace_tooSmall) {
DISPLAY("ZSTD_compressSequencesAndLiterals() should have failed: literals buffer capacity is too small\n");
goto _output_error;
}
/* not enough literals: must fail */
compressedSize = ZSTD_compressSequencesAndLiterals(cctx, dst, dstCapacity, seqs, nbSeqs, src, litSize-1, decompressSize, srcSize);
if (!ZSTD_isError(compressedSize)) {
DISPLAY("ZSTD_compressSequencesAndLiterals() should have failed: not enough literals provided\n");
goto _output_error;
}
/* too many literals: must fail */
compressedSize = ZSTD_compressSequencesAndLiterals(cctx, dst, dstCapacity, seqs, nbSeqs, src, litSize+1, decompressSize, srcSize);
if (!ZSTD_isError(compressedSize)) {
DISPLAY("ZSTD_compressSequencesAndLiterals() should have failed: too many literals provided\n");
goto _output_error;
}
/* srcSize too large: must fail */
compressedSize = ZSTD_compressSequencesAndLiterals(cctx, dst, dstCapacity, seqs, nbSeqs, litBuffer, litSize, decompressSize, srcSize+1);
if (!ZSTD_isError(compressedSize)) {
DISPLAY("ZSTD_compressSequencesAndLiterals() should have failed: srcSize is too large\n");
goto _output_error;
}
/* srcSize too small: must fail */
compressedSize = ZSTD_compressSequencesAndLiterals(cctx, dst, dstCapacity, seqs, nbSeqs, litBuffer, litSize, decompressSize, srcSize-1);
if (!ZSTD_isError(compressedSize)) {
DISPLAY("ZSTD_compressSequencesAndLiterals() should have failed: srcSize is too small\n");
goto _output_error;
}
/* correct amount of literals: should compress successfully */
compressedSize = ZSTD_compressSequencesAndLiterals(cctx, dst, dstCapacity, seqs, nbSeqs, litBuffer, litSize, litSize+8, srcSize);
if (ZSTD_isError(compressedSize)) {
DISPLAY("Error in ZSTD_compressSequencesAndLiterals()\n");
goto _output_error;
}
}
{ ZSTD_FrameHeader zfh;
size_t const zfhStatus = ZSTD_getFrameHeader(&zfh, dst, compressedSize);
if (zfhStatus != 0) {
DISPLAY("Error reading frame header\n");
goto _output_error;
}
if (zfh.frameContentSize != srcSize) {
DISPLAY("Error: ZSTD_compressSequencesAndLiterals() did not report srcSize in the frame header\n");
goto _output_error;
}
if (zfh.windowSize > srcSize) {
DISPLAY("Error: ZSTD_compressSequencesAndLiterals() did not resized window size to smaller contentSize\n");
goto _output_error;
}
}
{ size_t const dSize = ZSTD_decompress(decompressBuffer, decompressSize, dst, compressedSize);
if (ZSTD_isError(dSize)) {
DISPLAY("Error during decompression of frame produced by ZSTD_compressSequencesAndLiterals()\n");
goto _output_error;
}
if (dSize != srcSize) {
DISPLAY("Error: decompression of frame produced by ZSTD_compressSequencesAndLiterals() has different size\n");
goto _output_error;
}
if (memcmp(decompressBuffer, src, srcSize)) {
DISPLAY("Error: decompression of frame produced by ZSTD_compressSequencesAndLiterals() produces a different content (of same size)\n");
goto _output_error;
}
}
ZSTD_freeCCtx(cctx);
free(litBuffer);
free(decompressBuffer);
free(seqs);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_compressSequencesAndLiterals : reject a too large block : ", testNb++);
{ /* one block of ZSTD_BLOCKSIZE_MAX + 1 bytes : 16 literals and a match, then the block delimiter */
ZSTD_Sequence const seqs[2] = { { 1, 16, ZSTD_BLOCKSIZE_MAX + 1 - 16, 0 }, { 0, 0, 0, 0 } };
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
CHECK(cctx != NULL);
CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_blockDelimiters, ZSTD_sf_explicitBlockDelimiters) );
CHECK( ZSTD_isError(ZSTD_compressSequencesAndLiterals(cctx, compressedBuffer, compressedBufferSize,
seqs, 2, CNBuffer, 16, CNBuffSize, ZSTD_BLOCKSIZE_MAX + 1)) );
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
/* Multiple blocks of zeros test */
#define LONGZEROSLENGTH 1000000 /* 1MB of zeros */
DISPLAYLEVEL(3, "test%3i : compress %u zeroes : ", testNb++, LONGZEROSLENGTH);
memset(CNBuffer, 0, LONGZEROSLENGTH);
CHECK_VAR(cSize, ZSTD_compress(compressedBuffer, ZSTD_compressBound(LONGZEROSLENGTH), CNBuffer, LONGZEROSLENGTH, 1) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/LONGZEROSLENGTH*100);
DISPLAYLEVEL(3, "test%3i : decompress %u zeroes : ", testNb++, LONGZEROSLENGTH);
{ CHECK_NEWV(r, ZSTD_decompress(decodedBuffer, LONGZEROSLENGTH, compressedBuffer, cSize) );
if (r != LONGZEROSLENGTH) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
/* All zeroes test (test bug #137) */
#define ZEROESLENGTH 100
DISPLAYLEVEL(3, "test%3i : compress %u zeroes : ", testNb++, ZEROESLENGTH);
memset(CNBuffer, 0, ZEROESLENGTH);
CHECK_VAR(cSize, ZSTD_compress(compressedBuffer, ZSTD_compressBound(ZEROESLENGTH), CNBuffer, ZEROESLENGTH, 1) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/ZEROESLENGTH*100);
DISPLAYLEVEL(3, "test%3i : decompress %u zeroes : ", testNb++, ZEROESLENGTH);
{ CHECK_NEWV(r, ZSTD_decompress(decodedBuffer, ZEROESLENGTH, compressedBuffer, cSize) );
if (r != ZEROESLENGTH) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
/* nbSeq limit test */
#define _3BYTESTESTLENGTH 131000
#define NB3BYTESSEQLOG 9
#define NB3BYTESSEQ (1 << NB3BYTESSEQLOG)
#define NB3BYTESSEQMASK (NB3BYTESSEQ-1)
/* creates a buffer full of 3-bytes sequences */
{ BYTE _3BytesSeqs[NB3BYTESSEQ][3];
U32 rSeed = 1;
/* create batch of 3-bytes sequences */
{ int i;
for (i=0; i < NB3BYTESSEQ; i++) {
_3BytesSeqs[i][0] = (BYTE)(FUZ_rand(&rSeed) & 255);
_3BytesSeqs[i][1] = (BYTE)(FUZ_rand(&rSeed) & 255);
_3BytesSeqs[i][2] = (BYTE)(FUZ_rand(&rSeed) & 255);
} }
/* randomly fills CNBuffer with prepared 3-bytes sequences */
{ int i;
for (i=0; i < _3BYTESTESTLENGTH; i += 3) { /* note : CNBuffer size > _3BYTESTESTLENGTH+3 */
U32 const id = FUZ_rand(&rSeed) & NB3BYTESSEQMASK;
((BYTE*)CNBuffer)[i+0] = _3BytesSeqs[id][0];
((BYTE*)CNBuffer)[i+1] = _3BytesSeqs[id][1];
((BYTE*)CNBuffer)[i+2] = _3BytesSeqs[id][2];
} } }
DISPLAYLEVEL(3, "test%3i : growing nbSeq : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const maxNbSeq = _3BYTESTESTLENGTH / 3;
size_t const bound = ZSTD_compressBound(_3BYTESTESTLENGTH);
size_t nbSeq = 1;
while (nbSeq <= maxNbSeq) {
CHECK_Z(ZSTD_compressCCtx(cctx, compressedBuffer, bound, CNBuffer, nbSeq * 3, 19));
/* Check every sequence for the first 100, then skip more rapidly. */
if (nbSeq < 100) {
++nbSeq;
} else {
nbSeq += (nbSeq >> 2);
}
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : compress lots 3-bytes sequences : ", testNb++);
CHECK_VAR(cSize, ZSTD_compress(compressedBuffer, ZSTD_compressBound(_3BYTESTESTLENGTH),
CNBuffer, _3BYTESTESTLENGTH, 19) );
DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/_3BYTESTESTLENGTH*100);
DISPLAYLEVEL(3, "test%3i : decompress lots 3-bytes sequence : ", testNb++);
{ CHECK_NEWV(r, ZSTD_decompress(decodedBuffer, _3BYTESTESTLENGTH, compressedBuffer, cSize) );
if (r != _3BYTESTESTLENGTH) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : growing literals buffer : ", testNb++);
RDG_genBuffer(CNBuffer, CNBuffSize, 0.0, 0.1, seed);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
size_t const bound = ZSTD_compressBound(CNBuffSize);
size_t size = 1;
while (size <= CNBuffSize) {
CHECK_Z(ZSTD_compressCCtx(cctx, compressedBuffer, bound, CNBuffer, size, 3));
/* Check every size for the first 100, then skip more rapidly. */
if (size < 100) {
++size;
} else {
size += (size >> 2);
}
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : incompressible data and ill suited dictionary : ", testNb++);
{ /* Train a dictionary on low characters */
size_t dictSize = 16 KB;
void* const dictBuffer = malloc(dictSize);
size_t const totalSampleSize = 1 MB;
size_t const sampleUnitSize = 8 KB;
U32 const nbSamples = (U32)(totalSampleSize / sampleUnitSize);
size_t* const samplesSizes = (size_t*) malloc(nbSamples * sizeof(size_t));
if (!dictBuffer || !samplesSizes) goto _output_error;
{ U32 u; for (u=0; u<nbSamples; u++) samplesSizes[u] = sampleUnitSize; }
dictSize = ZDICT_trainFromBuffer(dictBuffer, dictSize, CNBuffer, samplesSizes, nbSamples);
if (ZDICT_isError(dictSize)) goto _output_error;
/* Reverse the characters to make the dictionary ill suited */
{ U32 u;
for (u = 0; u < CNBuffSize; ++u) {
((BYTE*)CNBuffer)[u] = 255 - ((BYTE*)CNBuffer)[u];
} }
{ /* Compress the data */
size_t const inputSize = 500;
size_t const outputSize = ZSTD_compressBound(inputSize);
void* const outputBuffer = malloc(outputSize);
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
if (!outputBuffer || !cctx) goto _output_error;
CHECK_Z(ZSTD_compress_usingDict(cctx, outputBuffer, outputSize, CNBuffer, inputSize, dictBuffer, dictSize, 1));
free(outputBuffer);
ZSTD_freeCCtx(cctx);
}
free(dictBuffer);
free(samplesSizes);
}
DISPLAYLEVEL(3, "OK \n");
/* frame operations on skippable frames */
{ const char skippableFrame[] = "\x52\x2a\x4d\x18\x05\x0\x0\0abcde";
size_t const skippableFrameSize = sizeof(skippableFrame) - 1 /* remove the terminating /0 */;
DISPLAYLEVEL(3, "test%3i : ZSTD_findFrameCompressedSize on skippable frame : ", testNb++);
CHECK(ZSTD_findFrameCompressedSize(skippableFrame, skippableFrameSize) == skippableFrameSize);
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_getFrameContentSize on skippable frame : ", testNb++);
CHECK(ZSTD_getFrameContentSize(skippableFrame, skippableFrameSize) == 0);
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_getFrameHeader on skippable frame : ", testNb++);
{ ZSTD_FrameHeader zfh;
size_t const s = ZSTD_getFrameHeader(&zfh, skippableFrame, skippableFrameSize);
CHECK_Z(s);
CHECK(s == 0); /* success */
CHECK(zfh.frameType == ZSTD_skippableFrame);
CHECK(zfh.headerSize == ZSTD_SKIPPABLEHEADERSIZE);
CHECK(zfh.dictID == 2); /* magic variant */
assert(skippableFrameSize >= ZSTD_SKIPPABLEHEADERSIZE);
CHECK(zfh.frameContentSize == skippableFrameSize - ZSTD_SKIPPABLEHEADERSIZE);
}
DISPLAYLEVEL(3, "OK \n");
}
/* error string tests */
DISPLAYLEVEL(3, "test%3i : testing ZSTD error code strings : ", testNb++);
if (strcmp("No error detected", ZSTD_getErrorName((ZSTD_ErrorCode)(0-ZSTD_error_no_error))) != 0) goto _output_error;
if (strcmp("No error detected", ZSTD_getErrorString(ZSTD_error_no_error)) != 0) goto _output_error;
if (strcmp("Unspecified error code", ZSTD_getErrorString((ZSTD_ErrorCode)(0-ZSTD_error_GENERIC))) != 0) goto _output_error;
if (strcmp("Error (generic)", ZSTD_getErrorName((size_t)0-ZSTD_error_GENERIC)) != 0) goto _output_error;
if (strcmp("Error (generic)", ZSTD_getErrorString(ZSTD_error_GENERIC)) != 0) goto _output_error;
if (strcmp("No error detected", ZSTD_getErrorName(ZSTD_error_GENERIC)) != 0) goto _output_error;
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : testing ZSTD dictionary sizes : ", testNb++);
RDG_genBuffer(CNBuffer, CNBuffSize, compressibility, 0., seed);
{
size_t const size = MIN(128 KB, CNBuffSize);
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_CDict* const lgCDict = ZSTD_createCDict(CNBuffer, size, 1);
ZSTD_CDict* const smCDict = ZSTD_createCDict(CNBuffer, 1 KB, 1);
ZSTD_FrameHeader lgHeader;
ZSTD_FrameHeader smHeader;
CHECK_Z(ZSTD_compress_usingCDict(cctx, compressedBuffer, compressedBufferSize, CNBuffer, size, lgCDict));
CHECK_Z(ZSTD_getFrameHeader(&lgHeader, compressedBuffer, compressedBufferSize));
CHECK_Z(ZSTD_compress_usingCDict(cctx, compressedBuffer, compressedBufferSize, CNBuffer, size, smCDict));
CHECK_Z(ZSTD_getFrameHeader(&smHeader, compressedBuffer, compressedBufferSize));
if (lgHeader.windowSize != smHeader.windowSize) goto _output_error;
ZSTD_freeCDict(smCDict);
ZSTD_freeCDict(lgCDict);
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : testing FSE_normalizeCount() PR#1255: ", testNb++);
{
short norm[32];
unsigned count[32];
unsigned const tableLog = 5;
size_t const nbSeq = 32;
unsigned const maxSymbolValue = 31;
size_t i;
for (i = 0; i < 32; ++i)
count[i] = 1;
/* Calling FSE_normalizeCount() on a uniform distribution should not
* cause a division by zero.
*/
FSE_normalizeCount(norm, tableLog, count, nbSeq, maxSymbolValue, /* useLowProbCount */ 1);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : testing FSE_writeNCount() PR#2779: ", testNb++);
{
size_t const outBufSize = 9;
short const count[11] = {1, 0, 1, 0, 1, 0, 1, 0, 1, 9, 18};
unsigned const tableLog = 5;
unsigned const maxSymbolValue = 10;
BYTE* outBuf = (BYTE*)malloc(outBufSize*sizeof(BYTE));
/* Ensure that this write doesn't write out of bounds, and that
* FSE_writeNCount_generic() is *not* called with writeIsSafe == 1.
*/
FSE_writeNCount(outBuf, outBufSize, count, maxSymbolValue, tableLog);
free(outBuf);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : testing bitwise intrinsics PR#3045: ", testNb++);
{
U32 seed_copy = seed; /* need non-const seed to avoid compiler warning for FUZ_rand(&seed) */
U32 rand32 = FUZ_rand(&seed_copy);
U64 rand64 = ((U64)FUZ_rand(&seed_copy) << 32) | FUZ_rand(&seed_copy);
U32 lowbit_only_32 = 1;
U64 lowbit_only_64 = 1;
U32 highbit_only_32 = (U32)1 << 31;
U64 highbit_only_64 = (U64)1 << 63;
U32 i;
if (rand32 == 0) rand32 = 1; /* CLZ and CTZ are undefined on 0 */
if (rand64 == 0) rand64 = 1; /* CLZ and CTZ are undefined on 0 */
/* Test ZSTD_countTrailingZeros32 */
CHECK_EQ(ZSTD_countTrailingZeros32(lowbit_only_32), 0u);
CHECK_EQ(ZSTD_countTrailingZeros32(highbit_only_32), 31u);
CHECK_EQ(ZSTD_countTrailingZeros32(rand32), ZSTD_countTrailingZeros32_fallback(rand32));
/* Test ZSTD_countLeadingZeros32 */
CHECK_EQ(ZSTD_countLeadingZeros32(lowbit_only_32), 31u);
CHECK_EQ(ZSTD_countLeadingZeros32(highbit_only_32), 0u);
CHECK_EQ(ZSTD_countLeadingZeros32(rand32), ZSTD_countLeadingZeros32_fallback(rand32));
/* Test ZSTD_countTrailingZeros64 */
CHECK_EQ(ZSTD_countTrailingZeros64(lowbit_only_64), 0u);
CHECK_EQ(ZSTD_countTrailingZeros64(highbit_only_64), 63u);
/* Test ZSTD_countLeadingZeros64 */
CHECK_EQ(ZSTD_countLeadingZeros64(lowbit_only_64), 63u);
CHECK_EQ(ZSTD_countLeadingZeros64(highbit_only_64), 0u);
/* Test ZSTD_highbit32 */
CHECK_EQ(ZSTD_highbit32(lowbit_only_32), 0u);
CHECK_EQ(ZSTD_highbit32(highbit_only_32), 31u);
/* Test ZSTD_NbCommonBytes */
if (MEM_isLittleEndian()) {
if (MEM_64bits()) {
CHECK_EQ(ZSTD_NbCommonBytes(lowbit_only_32), 0u);
CHECK_EQ(ZSTD_NbCommonBytes(highbit_only_32), 3u);
} else {
CHECK_EQ(ZSTD_NbCommonBytes(lowbit_only_32), 0u);
CHECK_EQ(ZSTD_NbCommonBytes(highbit_only_32), 3u);
}
} else {
if (MEM_64bits()) {
CHECK_EQ(ZSTD_NbCommonBytes(lowbit_only_32), 7u);
CHECK_EQ(ZSTD_NbCommonBytes(highbit_only_32), 4u);
} else {
CHECK_EQ(ZSTD_NbCommonBytes(lowbit_only_32), 3u);
CHECK_EQ(ZSTD_NbCommonBytes(highbit_only_32), 0u);
}
}
/* Test MEM_ intrinsics */
CHECK_EQ(MEM_swap32(rand32), MEM_swap32_fallback(rand32));
CHECK_EQ(MEM_swap64(rand64), MEM_swap64_fallback(rand64));
/* Test fallbacks vs intrinsics on a range of small integers */
for (i=1; i <= 1000; i++) {
CHECK_EQ(MEM_swap32(i), MEM_swap32_fallback(i));
CHECK_EQ(MEM_swap64((U64)i), MEM_swap64_fallback((U64)i));
CHECK_EQ(ZSTD_countTrailingZeros32(i), ZSTD_countTrailingZeros32_fallback(i));
CHECK_EQ(ZSTD_countLeadingZeros32(i), ZSTD_countLeadingZeros32_fallback(i));
}
}
DISPLAYLEVEL(3, "OK \n");
#ifdef ZSTD_MULTITHREAD
DISPLAYLEVEL(3, "test%3i : passing wrong full dict should fail on compressStream2 refPrefix ", testNb++);
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
size_t const srcSize = 1 MB + 5; /* A little more than ZSTDMT_JOBSIZE_MIN */
size_t const dstSize = ZSTD_compressBound(srcSize);
void* const src = CNBuffer;
void* const dst = compressedBuffer;
void* dict = (void*)malloc(srcSize);
RDG_genBuffer(src, srcSize, compressibility, 0.5, seed);
RDG_genBuffer(dict, srcSize, compressibility, 0., seed);
/* Make sure there is no ZSTD_MAGIC_NUMBER */
memset(dict, 0, sizeof(U32));
/* something more than 1 */
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, 2));
/* lie and claim this is a full dict */
CHECK_Z(ZSTD_CCtx_refPrefix_advanced(cctx, dict, srcSize, ZSTD_dct_fullDict));
{ ZSTD_outBuffer out = {dst, dstSize, 0};
ZSTD_inBuffer in = {src, srcSize, 0};
/* should fail because its not a full dict like we said it was */
assert(ZSTD_isError(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush)));
}
ZSTD_freeCCtx(cctx);
free(dict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : small dictionary with multithreading and LDM ", testNb++);
{ ZSTD_CCtx* cctx = ZSTD_createCCtx();
size_t const srcSize = 1 MB + 5; /* A little more than ZSTDMT_JOBSIZE_MIN */
size_t const dictSize = 10;
size_t const dstSize = ZSTD_compressBound(srcSize);
void* const src = CNBuffer;
void* const dst = compressedBuffer;
void* dict = (void*)malloc(dictSize);
RDG_genBuffer(src, srcSize, compressibility, 0.5, seed);
RDG_genBuffer(dict, dictSize, compressibility, 0., seed);
/* Make sure there is no ZSTD_MAGIC_NUMBER */
memset(dict, 0, sizeof(U32));
/* Enable MT, LDM, and use refPrefix() for a small dict */
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, 2));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable));
CHECK_Z(ZSTD_CCtx_refPrefix(cctx, dict, dictSize));
CHECK_Z(ZSTD_compress2(cctx, dst, dstSize, src, srcSize));
ZSTD_freeCCtx(cctx);
free(dict);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_getCParams() + dictionary ", testNb++);
{
ZSTD_compressionParameters const medium = ZSTD_getCParams(1, 16*1024-1, 0);
ZSTD_compressionParameters const large = ZSTD_getCParams(1, 128*1024-1, 0);
ZSTD_compressionParameters const smallDict = ZSTD_getCParams(1, 0, 400);
ZSTD_compressionParameters const mediumDict = ZSTD_getCParams(1, 0, 10000);
ZSTD_compressionParameters const largeDict = ZSTD_getCParams(1, 0, 100000);
assert(!memcmp(&smallDict, &mediumDict, sizeof(smallDict)));
assert(!memcmp(&medium, &mediumDict, sizeof(medium)));
assert(!memcmp(&large, &largeDict, sizeof(large)));
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : ZSTD_adjustCParams() + dictionary ", testNb++);
{
ZSTD_compressionParameters const cParams = ZSTD_getCParams(1, 0, 0);
ZSTD_compressionParameters const smallDict = ZSTD_adjustCParams(cParams, 0, 400);
ZSTD_compressionParameters const smallSrcAndDict = ZSTD_adjustCParams(cParams, 500, 400);
assert(smallSrcAndDict.windowLog == 10);
assert(!memcmp(&cParams, &smallDict, sizeof(cParams)));
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : check compression mem usage monotonicity over levels for estimateCCtxSize() : ", testNb++);
{
int level = 1;
size_t prevSize = 0;
for (; level < ZSTD_maxCLevel(); ++level) {
size_t const currSize = ZSTD_estimateCCtxSize(level);
if (prevSize > currSize) {
DISPLAYLEVEL(3, "Error! previous cctx size: %u at level: %d is larger than current cctx size: %u at level: %d",
(unsigned)prevSize, level-1, (unsigned)currSize, level);
goto _output_error;
}
prevSize = currSize;
}
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : check estimateCCtxSize() always larger or equal to ZSTD_estimateCCtxSize_usingCParams() : ", testNb++);
{
size_t const kSizeIncrement = 2 KB;
int level = -3;
for (; level <= ZSTD_maxCLevel(); ++level) {
size_t dictSize = 0;
for (; dictSize <= 256 KB; dictSize += 8 * kSizeIncrement) {
size_t srcSize = 2 KB;
for (; srcSize < 300 KB; srcSize += kSizeIncrement) {
ZSTD_compressionParameters const cParams = ZSTD_getCParams(level, srcSize, dictSize);
size_t const cctxSizeUsingCParams = ZSTD_estimateCCtxSize_usingCParams(cParams);
size_t const cctxSizeUsingLevel = ZSTD_estimateCCtxSize(level);
if (cctxSizeUsingLevel < cctxSizeUsingCParams
|| ZSTD_isError(cctxSizeUsingCParams)
|| ZSTD_isError(cctxSizeUsingLevel)) {
DISPLAYLEVEL(3, "error! l: %d dict: %u srcSize: %u cctx size cpar: %u, cctx size level: %u\n",
level, (unsigned)dictSize, (unsigned)srcSize, (unsigned)cctxSizeUsingCParams, (unsigned)cctxSizeUsingLevel);
goto _output_error;
} } } } }
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "test%3i : thread pool API tests : \n", testNb++)
{
int const threadPoolTestResult = threadPoolTests();
if (threadPoolTestResult) {
goto _output_error;
}
}
DISPLAYLEVEL(3, "thread pool tests OK \n");
#endif /* ZSTD_MULTITHREAD */
_end:
free(CNBuffer);
free(compressedBuffer);
free(decodedBuffer);
return testResult;
_output_error:
testResult = 1;
DISPLAY("Error detected in Unit tests ! \n");
goto _end;
}
static int longUnitTests(U32 const seed, double compressibility)
{
size_t const CNBuffSize = 5 MB;
void* const CNBuffer = malloc(CNBuffSize);
size_t const compressedBufferSize = ZSTD_compressBound(CNBuffSize);
void* const compressedBuffer = malloc(compressedBufferSize);
void* const decodedBuffer = malloc(CNBuffSize);
int testResult = 0;
unsigned testNb=0;
size_t cSize;
/* Create compressible noise */
if (!CNBuffer || !compressedBuffer || !decodedBuffer) {
DISPLAY("Not enough memory, aborting\n");
testResult = 1;
goto _end;
}
RDG_genBuffer(CNBuffer, CNBuffSize, compressibility, 0., seed);
/* note : this test is rather long, it would be great to find a way to speed up its execution */
DISPLAYLEVEL(3, "longtest%3i : table cleanliness through index reduction : ", testNb++);
{ int cLevel;
size_t approxIndex = 0;
size_t maxIndex = ((3U << 29) + (1U << ZSTD_WINDOWLOG_MAX)); /* ZSTD_CURRENT_MAX from zstd_compress_internal.h */
/* Provision enough space in a static context so that we can do all
* this without ever reallocating, which would reset the indices. */
size_t const staticCCtxSize = ZSTD_estimateCStreamSize(22);
void* const staticCCtxBuffer = malloc(staticCCtxSize);
ZSTD_CCtx* const cctx = ZSTD_initStaticCCtx(staticCCtxBuffer, staticCCtxSize);
/* bump the indices so the following compressions happen at high
* indices. */
{ ZSTD_outBuffer out = { compressedBuffer, compressedBufferSize, 0 };
ZSTD_inBuffer in = { CNBuffer, CNBuffSize, 0 };
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, -500));
while (approxIndex <= (maxIndex / 4) * 3) {
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush));
approxIndex += in.pos;
CHECK_Z(in.pos == in.size);
in.pos = 0;
out.pos = 0;
}
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
}
/* spew a bunch of stuff into the table area */
for (cLevel = 1; cLevel <= 22; cLevel++) {
ZSTD_outBuffer out = { compressedBuffer, compressedBufferSize / (unsigned)cLevel, 0 };
ZSTD_inBuffer in = { CNBuffer, CNBuffSize, 0 };
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, cLevel));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
approxIndex += in.pos;
}
/* now crank the indices so we overflow */
{ ZSTD_outBuffer out = { compressedBuffer, compressedBufferSize, 0 };
ZSTD_inBuffer in = { CNBuffer, CNBuffSize, 0 };
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, -500));
while (approxIndex <= maxIndex) {
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush));
approxIndex += in.pos;
CHECK_Z(in.pos == in.size);
in.pos = 0;
out.pos = 0;
}
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
}
/* do a bunch of compressions again in low indices and ensure we don't
* hit untracked invalid indices */
for (cLevel = 1; cLevel <= 22; cLevel++) {
ZSTD_outBuffer out = { compressedBuffer, compressedBufferSize / (unsigned)cLevel, 0 };
ZSTD_inBuffer in = { CNBuffer, CNBuffSize, 0 };
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, cLevel));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush));
CHECK_Z(ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end));
approxIndex += in.pos;
}
free(staticCCtxBuffer);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "longtest%3i : testing ldm no regressions in size for opt parser : ", testNb++);
{ size_t cSizeLdm;
size_t cSizeNoLdm;
ZSTD_CCtx* const cctx = ZSTD_createCCtx();
RDG_genBuffer(CNBuffer, CNBuffSize, 0.5, 0.5, seed);
/* Enable checksum to verify round trip. */
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_enable));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 19));
/* Round trip once with ldm. */
cSizeLdm = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSizeLdm);
CHECK_Z(ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSizeLdm));
ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, 1));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, ZSTD_ps_disable));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 19));
/* Round trip once without ldm. */
cSizeNoLdm = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSizeNoLdm);
CHECK_Z(ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSizeNoLdm));
if (cSizeLdm > cSizeNoLdm) {
DISPLAY("Using long mode should not cause regressions for btopt+\n");
testResult = 1;
goto _end;
}
ZSTD_freeCCtx(cctx);
}
DISPLAYLEVEL(3, "OK \n");
DISPLAYLEVEL(3, "longtest%3i : testing cdict compression with different attachment strategies : ", testNb++);
{ ZSTD_CCtx* const cctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
size_t dictSize = CNBuffSize;
void* dict = (void*)malloc(dictSize);
ZSTD_CCtx_params* cctx_params = ZSTD_createCCtxParams();
ZSTD_dictAttachPref_e const attachPrefs[] = {
ZSTD_dictDefaultAttach,
ZSTD_dictForceAttach,
ZSTD_dictForceCopy,
ZSTD_dictForceLoad,
ZSTD_dictDefaultAttach,
ZSTD_dictForceAttach,
ZSTD_dictForceCopy,
ZSTD_dictForceLoad
};
int const enableDedicatedDictSearch[] = {0, 0, 0, 0, 1, 1, 1, 1};
int cLevel;
int i;
RDG_genBuffer(dict, dictSize, 0.5, 0.5, seed);
RDG_genBuffer(CNBuffer, CNBuffSize, 0.6, 0.6, seed);
CHECK_Z(cctx_params != NULL);
for (dictSize = CNBuffSize; dictSize; dictSize = dictSize >> 3) {
DISPLAYLEVEL(3, "\n Testing with dictSize %u ", (U32)dictSize);
for (cLevel = 4; cLevel < 13; cLevel++) {
for (i = 0; i < 8; ++i) {
ZSTD_dictAttachPref_e const attachPref = attachPrefs[i];
int const enableDDS = enableDedicatedDictSearch[i];
ZSTD_CDict* cdict;
DISPLAYLEVEL(5, "\n dictSize %u cLevel %d iter %d ", (U32)dictSize, cLevel, i);
ZSTD_CCtxParams_init(cctx_params, cLevel);
CHECK_Z(ZSTD_CCtxParams_setParameter(cctx_params, ZSTD_c_enableDedicatedDictSearch, enableDDS));
cdict = ZSTD_createCDict_advanced2(dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, cctx_params, ZSTD_defaultCMem);
CHECK(cdict != NULL);
CHECK_Z(ZSTD_CCtx_refCDict(cctx, cdict));
CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_forceAttachDict, (int)attachPref));
cSize = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, CNBuffSize);
CHECK_Z(cSize);
CHECK_Z(ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, dict, dictSize));
DISPLAYLEVEL(5, "compressed to %u bytes ", (U32)cSize);
CHECK_Z(ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters));
ZSTD_freeCDict(cdict);
} } }
ZSTD_freeCCtx(cctx);
ZSTD_freeDCtx(dctx);
ZSTD_freeCCtxParams(cctx_params);
free(dict);
}
DISPLAYLEVEL(3, "OK \n");
_end:
free(CNBuffer);
free(compressedBuffer);
free(decodedBuffer);
return testResult;
}
static size_t findDiff(const void* buf1, const void* buf2, size_t max)
{
const BYTE* b1 = (const BYTE*)buf1;
const BYTE* b2 = (const BYTE*)buf2;
size_t u;
for (u=0; u<max; u++) {
if (b1[u] != b2[u]) break;
}
return u;
}
static ZSTD_parameters FUZ_makeParams(ZSTD_compressionParameters cParams, ZSTD_frameParameters fParams)
{
ZSTD_parameters params;
params.cParams = cParams;
params.fParams = fParams;
return params;
}
static size_t FUZ_rLogLength(U32* seed, U32 logLength)
{
size_t const lengthMask = ((size_t)1 << logLength) - 1;
return (lengthMask+1) + (FUZ_rand(seed) & lengthMask);
}
static size_t FUZ_randomLength(U32* seed, U32 maxLog)
{
U32 const logLength = FUZ_rand(seed) % maxLog;
return FUZ_rLogLength(seed, logLength);
}
#undef CHECK
#define CHECK(cond, ...) { \
if (cond) { \
DISPLAY("Error => "); \
DISPLAY(__VA_ARGS__); \
DISPLAY(" (seed %u, test nb %u) \n", (unsigned)seed, testNb); \
goto _output_error; \
} }
#undef CHECK_Z
#define CHECK_Z(f) { \
size_t const err = f; \
if (ZSTD_isError(err)) { \
DISPLAY("Error => %s : %s ", \
#f, ZSTD_getErrorName(err)); \
DISPLAY(" (seed %u, test nb %u) \n", (unsigned)seed, testNb); \
goto _output_error; \
} }
static int fuzzerTests(U32 seed, unsigned nbTests, unsigned startTest, U32 const maxDurationS, double compressibility, int bigTests)
{
static const U32 maxSrcLog = 23;
static const U32 maxSampleLog = 22;
size_t const srcBufferSize = (size_t)1<<maxSrcLog;
size_t const dstBufferSize = (size_t)1<<maxSampleLog;
size_t const cBufferSize = ZSTD_compressBound(dstBufferSize);
BYTE* cNoiseBuffer[5];
BYTE* const cBuffer = (BYTE*) malloc (cBufferSize);
BYTE* const dstBuffer = (BYTE*) malloc (dstBufferSize);
BYTE* const mirrorBuffer = (BYTE*) malloc (dstBufferSize);
ZSTD_CCtx* const ctx = ZSTD_createCCtx();
ZSTD_DCtx* const dctx = ZSTD_createDCtx();
U32 result = 0;
unsigned testNb = 0;
U32 coreSeed = seed;
UTIL_time_t const startClock = UTIL_getTime();
U64 const maxClockSpan = maxDurationS * SEC_TO_MICRO;
int const cLevelLimiter = bigTests ? 3 : 2;
/* allocation */
cNoiseBuffer[0] = (BYTE*)malloc (srcBufferSize);
cNoiseBuffer[1] = (BYTE*)malloc (srcBufferSize);
cNoiseBuffer[2] = (BYTE*)malloc (srcBufferSize);
cNoiseBuffer[3] = (BYTE*)malloc (srcBufferSize);
cNoiseBuffer[4] = (BYTE*)malloc (srcBufferSize);
CHECK (!cNoiseBuffer[0] || !cNoiseBuffer[1] || !cNoiseBuffer[2] || !cNoiseBuffer[3] || !cNoiseBuffer[4]
|| !dstBuffer || !mirrorBuffer || !cBuffer || !ctx || !dctx,
"Not enough memory, fuzzer tests cancelled");
/* Create initial samples */
RDG_genBuffer(cNoiseBuffer[0], srcBufferSize, 0.00, 0., coreSeed); /* pure noise */
RDG_genBuffer(cNoiseBuffer[1], srcBufferSize, 0.05, 0., coreSeed); /* barely compressible */
RDG_genBuffer(cNoiseBuffer[2], srcBufferSize, compressibility, 0., coreSeed);
RDG_genBuffer(cNoiseBuffer[3], srcBufferSize, 0.95, 0., coreSeed); /* highly compressible */
RDG_genBuffer(cNoiseBuffer[4], srcBufferSize, 1.00, 0., coreSeed); /* sparse content */
/* catch up testNb */
for (testNb=1; testNb < startTest; testNb++) FUZ_rand(&coreSeed);
/* main test loop */
for ( ; (testNb <= nbTests) || (UTIL_clockSpanMicro(startClock) < maxClockSpan); testNb++ ) {
BYTE* srcBuffer; /* jumping pointer */
U32 lseed;
size_t sampleSize, maxTestSize, totalTestSize;
size_t cSize, totalCSize, totalGenSize;
U64 crcOrig;
BYTE* sampleBuffer;
const BYTE* dict;
size_t dictSize;
/* notification */
if (nbTests >= testNb) { DISPLAYUPDATE(2, "\r%6u/%6u ", testNb, nbTests); }
else { DISPLAYUPDATE(2, "\r%6u ", testNb); }
FUZ_rand(&coreSeed);
{ U32 const prime1 = 2654435761U; lseed = coreSeed ^ prime1; }
/* srcBuffer selection [0-4] */
{ U32 buffNb = FUZ_rand(&lseed) & 0x7F;
if (buffNb & 7) buffNb=2; /* most common : compressible (P) */
else {
buffNb >>= 3;
if (buffNb & 7) {
const U32 tnb[2] = { 1, 3 }; /* barely/highly compressible */
buffNb = tnb[buffNb >> 3];
} else {
const U32 tnb[2] = { 0, 4 }; /* not compressible / sparse */
buffNb = tnb[buffNb >> 3];
} }
srcBuffer = cNoiseBuffer[buffNb];
}
/* select src segment */
sampleSize = FUZ_randomLength(&lseed, maxSampleLog);
/* create sample buffer (to catch read error with valgrind & sanitizers) */
sampleBuffer = (BYTE*)malloc(sampleSize);
CHECK(sampleBuffer==NULL, "not enough memory for sample buffer");
{ size_t const sampleStart = FUZ_rand(&lseed) % (srcBufferSize - sampleSize);
memcpy(sampleBuffer, srcBuffer + sampleStart, sampleSize); }
crcOrig = XXH64(sampleBuffer, sampleSize, 0);
/* compression tests */
{ int const cLevelPositive = (int)
( FUZ_rand(&lseed) %
((U32)ZSTD_maxCLevel() - (FUZ_highbit32((U32)sampleSize) / (U32)cLevelLimiter)) )
+ 1;
int const cLevel = ((FUZ_rand(&lseed) & 15) == 3) ?
- (int)((FUZ_rand(&lseed) & 7) + 1) : /* test negative cLevel */
cLevelPositive;
DISPLAYLEVEL(5, "fuzzer t%u: Simple compression test (level %i) \n", testNb, cLevel);
cSize = ZSTD_compressCCtx(ctx, cBuffer, cBufferSize, sampleBuffer, sampleSize, cLevel);
CHECK(ZSTD_isError(cSize), "ZSTD_compressCCtx failed : %s", ZSTD_getErrorName(cSize));
/* compression failure test : too small dest buffer */
assert(cSize > 3);
{ const size_t missing = (FUZ_rand(&lseed) % (cSize-2)) + 1;
const size_t tooSmallSize = cSize - missing;
const unsigned endMark = 0x4DC2B1A9;
memcpy(dstBuffer+tooSmallSize, &endMark, sizeof(endMark));
DISPLAYLEVEL(5, "fuzzer t%u: compress into too small buffer of size %u (missing %u bytes) \n",
testNb, (unsigned)tooSmallSize, (unsigned)missing);
{ size_t const errorCode = ZSTD_compressCCtx(ctx, dstBuffer, tooSmallSize, sampleBuffer, sampleSize, cLevel);
CHECK(ZSTD_getErrorCode(errorCode) != ZSTD_error_dstSize_tooSmall, "ZSTD_compressCCtx should have failed ! (buffer too small : %u < %u)", (unsigned)tooSmallSize, (unsigned)cSize); }
{ unsigned endCheck; memcpy(&endCheck, dstBuffer+tooSmallSize, sizeof(endCheck));
CHECK(endCheck != endMark, "ZSTD_compressCCtx : dst buffer overflow (check.%08X != %08X.mark)", endCheck, endMark); }
} }
/* frame header decompression test */
{ ZSTD_FrameHeader zfh;
CHECK_Z( ZSTD_getFrameHeader(&zfh, cBuffer, cSize) );
CHECK(zfh.frameContentSize != sampleSize, "Frame content size incorrect");
}
/* Decompressed size test */
{ unsigned long long const rSize = ZSTD_findDecompressedSize(cBuffer, cSize);
CHECK(rSize != sampleSize, "decompressed size incorrect");
}
/* successful decompression test */
DISPLAYLEVEL(5, "fuzzer t%u: simple decompression test \n", testNb);
{ size_t const margin = (FUZ_rand(&lseed) & 1) ? 0 : (FUZ_rand(&lseed) & 31) + 1;
size_t const dSize = ZSTD_decompress(dstBuffer, sampleSize + margin, cBuffer, cSize);
CHECK(dSize != sampleSize, "ZSTD_decompress failed (%s) (srcSize : %u ; cSize : %u)", ZSTD_getErrorName(dSize), (unsigned)sampleSize, (unsigned)cSize);
{ U64 const crcDest = XXH64(dstBuffer, sampleSize, 0);
CHECK(crcOrig != crcDest, "decompression result corrupted (pos %u / %u)", (unsigned)findDiff(sampleBuffer, dstBuffer, sampleSize), (unsigned)sampleSize);
} }
free(sampleBuffer); /* no longer useful after this point */
/* truncated src decompression test */
DISPLAYLEVEL(5, "fuzzer t%u: decompression of truncated source \n", testNb);
{ size_t const missing = (FUZ_rand(&lseed) % (cSize-2)) + 1; /* no problem, as cSize > 4 (frameHeaderSizer) */
size_t const tooSmallSize = cSize - missing;
void* cBufferTooSmall = malloc(tooSmallSize); /* valgrind will catch read overflows */
CHECK(cBufferTooSmall == NULL, "not enough memory !");
memcpy(cBufferTooSmall, cBuffer, tooSmallSize);
{ size_t const errorCode = ZSTD_decompress(dstBuffer, dstBufferSize, cBufferTooSmall, tooSmallSize);
CHECK(!ZSTD_isError(errorCode), "ZSTD_decompress should have failed ! (truncated src buffer)"); }
free(cBufferTooSmall);
}
/* too small dst decompression test */
DISPLAYLEVEL(5, "fuzzer t%u: decompress into too small dst buffer \n", testNb);
if (sampleSize > 3) {
size_t const missing = (FUZ_rand(&lseed) % (sampleSize-2)) + 1; /* no problem, as cSize > 4 (frameHeaderSizer) */
size_t const tooSmallSize = sampleSize - missing;
static const BYTE token = 0xA9;
dstBuffer[tooSmallSize] = token;
{ size_t const errorCode = ZSTD_decompress(dstBuffer, tooSmallSize, cBuffer, cSize);
CHECK(ZSTD_getErrorCode(errorCode) != ZSTD_error_dstSize_tooSmall, "ZSTD_decompress should have failed : %u > %u (dst buffer too small)", (unsigned)errorCode, (unsigned)tooSmallSize); }
CHECK(dstBuffer[tooSmallSize] != token, "ZSTD_decompress : dst buffer overflow");
}
/* noisy src decompression test */
if (cSize > 6) {
/* insert noise into src */
{ U32 const maxNbBits = FUZ_highbit32((U32)(cSize-4));
size_t pos = 4; /* preserve magic number (too easy to detect) */
for (;;) {
/* keep some original src */
{ U32 const nbBits = FUZ_rand(&lseed) % maxNbBits;
size_t const mask = (1<<nbBits) - 1;
size_t const skipLength = FUZ_rand(&lseed) & mask;
pos += skipLength;
}
if (pos >= cSize) break;
/* add noise */
{ U32 const nbBitsCodes = FUZ_rand(&lseed) % maxNbBits;
U32 const nbBits = nbBitsCodes ? nbBitsCodes-1 : 0;
size_t const mask = (1<<nbBits) - 1;
size_t const rNoiseLength = (FUZ_rand(&lseed) & mask) + 1;
size_t const noiseLength = MIN(rNoiseLength, cSize-pos);
size_t const noiseStart = FUZ_rand(&lseed) % (srcBufferSize - noiseLength);
memcpy(cBuffer + pos, srcBuffer + noiseStart, noiseLength);
pos += noiseLength;
} } }
/* decompress noisy source */
DISPLAYLEVEL(5, "fuzzer t%u: decompress noisy source \n", testNb);
{ U32 const endMark = 0xA9B1C3D6;
memcpy(dstBuffer+sampleSize, &endMark, 4);
{ size_t const decompressResult = ZSTD_decompress(dstBuffer, sampleSize, cBuffer, cSize);
/* result *may* be an unlikely success, but even then, it must strictly respect dst buffer boundaries */
CHECK((!ZSTD_isError(decompressResult)) && (decompressResult>sampleSize),
"ZSTD_decompress on noisy src : result is too large : %u > %u (dst buffer)", (unsigned)decompressResult, (unsigned)sampleSize);
}
{ U32 endCheck; memcpy(&endCheck, dstBuffer+sampleSize, 4);
CHECK(endMark!=endCheck, "ZSTD_decompress on noisy src : dst buffer overflow");
} } } /* noisy src decompression test */
/*===== Bufferless streaming compression test, scattered segments and dictionary =====*/
DISPLAYLEVEL(5, "fuzzer t%u: Bufferless streaming compression test \n", testNb);
{ U32 const testLog = FUZ_rand(&lseed) % maxSrcLog;
U32 const dictLog = FUZ_rand(&lseed) % maxSrcLog;
int const cLevel = (int)(FUZ_rand(&lseed) %
((U32)ZSTD_maxCLevel() -
(MAX(testLog, dictLog) / (U32)cLevelLimiter))) +
1;
maxTestSize = FUZ_rLogLength(&lseed, testLog);
if (maxTestSize >= dstBufferSize) maxTestSize = dstBufferSize-1;
dictSize = FUZ_rLogLength(&lseed, dictLog); /* needed also for decompression */
dict = srcBuffer + (FUZ_rand(&lseed) % (srcBufferSize - dictSize));
DISPLAYLEVEL(6, "fuzzer t%u: Compressing up to <=%u bytes at level %i with dictionary size %u \n",
testNb, (unsigned)maxTestSize, cLevel, (unsigned)dictSize);
if (FUZ_rand(&lseed) & 0xF) {
CHECK_Z ( ZSTD_compressBegin_usingDict(ctx, dict, dictSize, cLevel) );
} else {
ZSTD_compressionParameters const cPar = ZSTD_getCParams(cLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize);
ZSTD_frameParameters const fPar = { FUZ_rand(&lseed)&1 /* contentSizeFlag */,
!(FUZ_rand(&lseed)&3) /* contentChecksumFlag*/,
0 /*NodictID*/ }; /* note : since dictionary is fake, dictIDflag has no impact */
ZSTD_parameters const p = FUZ_makeParams(cPar, fPar);
CHECK_Z ( ZSTD_compressBegin_advanced(ctx, dict, dictSize, p, ZSTD_CONTENTSIZE_UNKNOWN) );
}
}
{ U32 const nbChunks = (FUZ_rand(&lseed) & 127) + 2;
U32 n;
XXH64_state_t xxhState;
XXH64_reset(&xxhState, 0);
for (totalTestSize=0, cSize=0, n=0 ; n<nbChunks ; n++) {
size_t const segmentSize = FUZ_randomLength(&lseed, maxSampleLog);
size_t const segmentStart = FUZ_rand(&lseed) % (srcBufferSize - segmentSize);
if (cBufferSize-cSize < ZSTD_compressBound(segmentSize)) break; /* avoid invalid dstBufferTooSmall */
if (totalTestSize+segmentSize > maxTestSize) break;
{ size_t const compressResult = ZSTD_compressContinue(ctx, cBuffer+cSize, cBufferSize-cSize, srcBuffer+segmentStart, segmentSize);
CHECK (ZSTD_isError(compressResult), "multi-segments compression error : %s", ZSTD_getErrorName(compressResult));
cSize += compressResult;
}
XXH64_update(&xxhState, srcBuffer+segmentStart, segmentSize);
memcpy(mirrorBuffer + totalTestSize, srcBuffer+segmentStart, segmentSize);
totalTestSize += segmentSize;
}
{ size_t const flushResult = ZSTD_compressEnd(ctx, cBuffer+cSize, cBufferSize-cSize, NULL, 0);
CHECK (ZSTD_isError(flushResult), "multi-segments epilogue error : %s", ZSTD_getErrorName(flushResult));
cSize += flushResult;
}
crcOrig = XXH64_digest(&xxhState);
}
/* streaming decompression test */
DISPLAYLEVEL(5, "fuzzer t%u: Bufferless streaming decompression test \n", testNb);
/* ensure memory requirement is good enough (should always be true) */
{ ZSTD_FrameHeader zfh;
CHECK( ZSTD_getFrameHeader(&zfh, cBuffer, ZSTD_FRAMEHEADERSIZE_MAX),
"ZSTD_getFrameHeader(): error retrieving frame information");
{ size_t const roundBuffSize = ZSTD_decodingBufferSize_min(zfh.windowSize, zfh.frameContentSize);
CHECK_Z(roundBuffSize);
CHECK((roundBuffSize > totalTestSize) && (zfh.frameContentSize!=ZSTD_CONTENTSIZE_UNKNOWN),
"ZSTD_decodingBufferSize_min() requires more memory (%u) than necessary (%u)",
(unsigned)roundBuffSize, (unsigned)totalTestSize );
} }
if (dictSize<8) dictSize=0, dict=NULL; /* disable dictionary */
CHECK_Z( ZSTD_decompressBegin_usingDict(dctx, dict, dictSize) );
totalCSize = 0;
totalGenSize = 0;
while (totalCSize < cSize) {
size_t const inSize = ZSTD_nextSrcSizeToDecompress(dctx);
size_t const genSize = ZSTD_decompressContinue(dctx, dstBuffer+totalGenSize, dstBufferSize-totalGenSize, cBuffer+totalCSize, inSize);
CHECK (ZSTD_isError(genSize), "ZSTD_decompressContinue error : %s", ZSTD_getErrorName(genSize));
totalGenSize += genSize;
totalCSize += inSize;
}
CHECK (ZSTD_nextSrcSizeToDecompress(dctx) != 0, "frame not fully decoded");
CHECK (totalGenSize != totalTestSize, "streaming decompressed data : wrong size")
CHECK (totalCSize != cSize, "compressed data should be fully read")
{ U64 const crcDest = XXH64(dstBuffer, totalTestSize, 0);
CHECK(crcOrig != crcDest, "streaming decompressed data corrupted (pos %u / %u)",
(unsigned)findDiff(mirrorBuffer, dstBuffer, totalTestSize), (unsigned)totalTestSize);
}
} /* for ( ; (testNb <= nbTests) */
DISPLAY("\r%u fuzzer tests completed \n", testNb-1);
_cleanup:
ZSTD_freeCCtx(ctx);
ZSTD_freeDCtx(dctx);
free(cNoiseBuffer[0]);
free(cNoiseBuffer[1]);
free(cNoiseBuffer[2]);
free(cNoiseBuffer[3]);
free(cNoiseBuffer[4]);
free(cBuffer);
free(dstBuffer);
free(mirrorBuffer);
return (int)result;
_output_error:
result = 1;
goto _cleanup;
}
/*_*******************************************************
* Command line
*********************************************************/
static int FUZ_usage(const char* programName)
{
DISPLAY( "Usage :\n");
DISPLAY( " %s [args]\n", programName);
DISPLAY( "\n");
DISPLAY( "Arguments :\n");
DISPLAY( " -i# : Number of tests (default:%i)\n", nbTestsDefault);
DISPLAY( " -T# : Max duration to run for. Overrides number of tests. (e.g. -T1m or -T60s for one minute)\n");
DISPLAY( " -s# : Select seed (default:prompt user)\n");
DISPLAY( " -t# : Select starting test number (default:0)\n");
DISPLAY( " -P# : Select compressibility in %% (default:%i%%)\n", FUZ_compressibility_default);
DISPLAY( " -v : verbose\n");
DISPLAY( " -p : pause at the end\n");
DISPLAY( " -h : display help and exit\n");
return 0;
}
/*! readU32FromChar() :
@return : unsigned integer value read from input in `char` format
allows and interprets K, KB, KiB, M, MB and MiB suffix.
Will also modify `*stringPtr`, advancing it to position where it stopped reading.
Note : function result can overflow if digit string > MAX_UINT */
static unsigned readU32FromChar(const char** stringPtr)
{
unsigned result = 0;
while ((**stringPtr >='0') && (**stringPtr <='9'))
result *= 10, result += (unsigned)(**stringPtr - '0'), (*stringPtr)++ ;
if ((**stringPtr=='K') || (**stringPtr=='M')) {
result <<= 10;
if (**stringPtr=='M') result <<= 10;
(*stringPtr)++ ;
if (**stringPtr=='i') (*stringPtr)++;
if (**stringPtr=='B') (*stringPtr)++;
}
return result;
}
/** longCommandWArg() :
* check if *stringPtr is the same as longCommand.
* If yes, @return 1 and advances *stringPtr to the position which immediately follows longCommand.
* @return 0 and doesn't modify *stringPtr otherwise.
*/
static int longCommandWArg(const char** stringPtr, const char* longCommand)
{
size_t const comSize = strlen(longCommand);
int const result = !strncmp(*stringPtr, longCommand, comSize);
if (result) *stringPtr += comSize;
return result;
}
int main(int argc, const char** argv)
{
U32 seed = 0;
int seedset = 0;
int argNb;
int nbTests = nbTestsDefault;
int testNb = 0;
int proba = FUZ_compressibility_default;
double probfloat;
int result = 0;
U32 mainPause = 0;
U32 maxDuration = 0;
int bigTests = 1;
int longTests = 0;
U32 memTestsOnly = 0;
const char* const programName = argv[0];
/* Check command line */
for (argNb=1; argNb<argc; argNb++) {
const char* argument = argv[argNb];
if(!argument) continue; /* Protection if argument empty */
/* Handle commands. Aggregated commands are allowed */
if (argument[0]=='-') {
if (longCommandWArg(&argument, "--memtest=")) { memTestsOnly = readU32FromChar(&argument); continue; }
if (!strcmp(argument, "--memtest")) { memTestsOnly=1; continue; }
if (!strcmp(argument, "--no-big-tests")) { bigTests=0; continue; }
if (!strcmp(argument, "--long-tests")) { longTests=1; continue; }
if (!strcmp(argument, "--no-long-tests")) { longTests=0; continue; }
argument++;
while (*argument!=0) {
switch(*argument)
{
case 'h':
return FUZ_usage(programName);
case 'v':
argument++;
g_displayLevel++;
break;
case 'q':
argument++;
g_displayLevel--;
break;
case 'p': /* pause at the end */
argument++;
mainPause = 1;
break;
case 'i':
argument++; maxDuration = 0;
nbTests = (int)readU32FromChar(&argument);
break;
case 'T':
argument++;
nbTests = 0;
maxDuration = readU32FromChar(&argument);
if (*argument=='s') argument++; /* seconds */
if (*argument=='m') maxDuration *= 60, argument++; /* minutes */
if (*argument=='n') argument++;
break;
case 's':
argument++;
seedset = 1;
seed = readU32FromChar(&argument);
break;
case 't':
argument++;
testNb = (int)readU32FromChar(&argument);
break;
case 'P': /* compressibility % */
argument++;
proba = (int)readU32FromChar(&argument);
if (proba>100) proba = 100;
break;
default:
return (FUZ_usage(programName), 1);
} } } } /* for (argNb=1; argNb<argc; argNb++) */
/* Get Seed */
DISPLAY("Starting zstd tester (%i-bits, %s)\n", (int)(sizeof(size_t)*8), ZSTD_VERSION_STRING);
if (!seedset) {
time_t const t = time(NULL);
U32 const h = XXH32(&t, sizeof(t), 1);
seed = h % 10000;
}
DISPLAY("Seed = %u\n", (unsigned)seed);
if (proba!=FUZ_compressibility_default) DISPLAY("Compressibility : %i%%\n", proba);
probfloat = ((double)proba) / 100;
if (memTestsOnly) {
g_displayLevel = MAX(3, g_displayLevel);
return FUZ_mallocTests(seed, probfloat, memTestsOnly);
}
if (nbTests < testNb) nbTests = testNb;
if (testNb==0) {
result = basicUnitTests(0, probfloat); /* constant seed for predictability */
if (!result && longTests) {
result = longUnitTests(0, probfloat);
}
}
if (!result)
result = fuzzerTests(seed, (unsigned)nbTests, (unsigned)testNb, maxDuration, ((double)proba) / 100, bigTests);
if (mainPause) {
int unused;
DISPLAY("Press Enter \n");
unused = getchar();
(void)unused;
}
return result;
}