zstd-ruby 1.4.0.0 → 1.4.1.0

data/ext/zstdruby/libzstd/dictBuilder/zdict.c

@@ -741,7 +741,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     /* analyze, build stats, starting with literals */
     {   size_t maxNbBits = HUF_buildCTable (hufTable, countLit, 255, huffLog);
         if (HUF_isError(maxNbBits)) {
-            eSize = ERROR(GENERIC);
+            eSize = maxNbBits;
             DISPLAYLEVEL(1, " HUF_buildCTable error \n");
             goto _cleanup;
         }
@@ -764,7 +764,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     total=0; for (u=0; u<=offcodeMax; u++) total+=offcodeCount[u];
     errorCode = FSE_normalizeCount(offcodeNCount, Offlog, offcodeCount, total, offcodeMax);
     if (FSE_isError(errorCode)) {
-        eSize = ERROR(GENERIC);
+        eSize = errorCode;
         DISPLAYLEVEL(1, "FSE_normalizeCount error with offcodeCount \n");
         goto _cleanup;
     }
@@ -773,7 +773,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     total=0; for (u=0; u<=MaxML; u++) total+=matchLengthCount[u];
     errorCode = FSE_normalizeCount(matchLengthNCount, mlLog, matchLengthCount, total, MaxML);
     if (FSE_isError(errorCode)) {
-        eSize = ERROR(GENERIC);
+        eSize = errorCode;
         DISPLAYLEVEL(1, "FSE_normalizeCount error with matchLengthCount \n");
         goto _cleanup;
     }
@@ -782,7 +782,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     total=0; for (u=0; u<=MaxLL; u++) total+=litLengthCount[u];
     errorCode = FSE_normalizeCount(litLengthNCount, llLog, litLengthCount, total, MaxLL);
     if (FSE_isError(errorCode)) {
-        eSize = ERROR(GENERIC);
+        eSize = errorCode;
         DISPLAYLEVEL(1, "FSE_normalizeCount error with litLengthCount \n");
         goto _cleanup;
     }
@@ -791,7 +791,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     /* write result to buffer */
     {   size_t const hhSize = HUF_writeCTable(dstPtr, maxDstSize, hufTable, 255, huffLog);
         if (HUF_isError(hhSize)) {
-            eSize = ERROR(GENERIC);
+            eSize = hhSize;
             DISPLAYLEVEL(1, "HUF_writeCTable error \n");
             goto _cleanup;
         }
@@ -802,7 +802,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
 
     {   size_t const ohSize = FSE_writeNCount(dstPtr, maxDstSize, offcodeNCount, OFFCODE_MAX, Offlog);
         if (FSE_isError(ohSize)) {
-            eSize = ERROR(GENERIC);
+            eSize = ohSize;
             DISPLAYLEVEL(1, "FSE_writeNCount error with offcodeNCount \n");
             goto _cleanup;
         }
@@ -813,7 +813,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
 
     {   size_t const mhSize = FSE_writeNCount(dstPtr, maxDstSize, matchLengthNCount, MaxML, mlLog);
         if (FSE_isError(mhSize)) {
-            eSize = ERROR(GENERIC);
+            eSize = mhSize;
             DISPLAYLEVEL(1, "FSE_writeNCount error with matchLengthNCount \n");
             goto _cleanup;
         }
@@ -824,7 +824,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
 
     {   size_t const lhSize = FSE_writeNCount(dstPtr, maxDstSize, litLengthNCount, MaxLL, llLog);
         if (FSE_isError(lhSize)) {
-            eSize = ERROR(GENERIC);
+            eSize = lhSize;
             DISPLAYLEVEL(1, "FSE_writeNCount error with litlengthNCount \n");
             goto _cleanup;
         }
@@ -834,7 +834,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     }
 
     if (maxDstSize<12) {
-        eSize = ERROR(GENERIC);
+        eSize = ERROR(dstSize_tooSmall);
         DISPLAYLEVEL(1, "not enough space to write RepOffsets \n");
         goto _cleanup;
     }

data/ext/zstdruby/libzstd/dictBuilder/zdict.h

@@ -94,6 +94,8 @@ typedef struct {
     unsigned steps;              /* Number of steps : Only used for optimization : 0 means default (40) : Higher means more parameters checked */
     unsigned nbThreads;          /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
     double splitPoint;           /* Percentage of samples used for training: Only used for optimization : the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (1.0), 1.0 when all samples are used for both training and testing */
+    unsigned shrinkDict;         /* Train dictionaries to shrink in size starting from the minimum size and selects the smallest dictionary that is shrinkDictMaxRegression% worse than the largest dictionary. 0 means no shrinking and 1 means shrinking  */
+    unsigned shrinkDictMaxRegression; /* Sets shrinkDictMaxRegression so that a smaller dictionary can be at worse shrinkDictMaxRegression% worse than the max dict size dictionary. */
     ZDICT_params_t zParams;
 } ZDICT_cover_params_t;
 
@@ -105,6 +107,9 @@ typedef struct {
     unsigned nbThreads;          /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
     double splitPoint;           /* Percentage of samples used for training: Only used for optimization : the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (0.75), 1.0 when all samples are used for both training and testing */
     unsigned accel;              /* Acceleration level: constraint: 0 < accel <= 10, higher means faster and less accurate, 0 means default(1) */
+    unsigned shrinkDict;         /* Train dictionaries to shrink in size starting from the minimum size and selects the smallest dictionary that is shrinkDictMaxRegression% worse than the largest dictionary. 0 means no shrinking and 1 means shrinking  */
+    unsigned shrinkDictMaxRegression; /* Sets shrinkDictMaxRegression so that a smaller dictionary can be at worse shrinkDictMaxRegression% worse than the max dict size dictionary. */
+
     ZDICT_params_t zParams;
 } ZDICT_fastCover_params_t;
 

data/ext/zstdruby/libzstd/legacy/zstd_legacy.h

@@ -238,6 +238,10 @@ MEM_STATIC ZSTD_frameSizeInfo ZSTD_findFrameSizeInfoLegacy(const void *src, size
             frameSizeInfo.decompressedBound = ZSTD_CONTENTSIZE_ERROR;
             break;
     }
+    if (!ZSTD_isError(frameSizeInfo.compressedSize) && frameSizeInfo.compressedSize > srcSize) {
+        frameSizeInfo.compressedSize = ERROR(srcSize_wrong);
+        frameSizeInfo.decompressedBound = ZSTD_CONTENTSIZE_ERROR;
+    }
     return frameSizeInfo;
 }
 

data/ext/zstdruby/libzstd/legacy/zstd_v01.c

@@ -1073,99 +1073,102 @@ static size_t HUF_decompress_usingDTable(   /* -3% slower when non static */
     const void* cSrc, size_t cSrcSize,
     const U16* DTable)
 {
-    BYTE* const ostart = (BYTE*) dst;
-    BYTE* op = ostart;
-    BYTE* const omax = op + maxDstSize;
-    BYTE* const olimit = omax-15;
-
-    const void* ptr = DTable;
-    const HUF_DElt* const dt = (const HUF_DElt*)(ptr)+1;
-    const U32 dtLog = DTable[0];
-    size_t errorCode;
-    U32 reloadStatus;
-
-    /* Init */
-
-    const U16* jumpTable = (const U16*)cSrc;
-    const size_t length1 = FSE_readLE16(jumpTable);
-    const size_t length2 = FSE_readLE16(jumpTable+1);
-    const size_t length3 = FSE_readLE16(jumpTable+2);
-    const size_t length4 = cSrcSize - 6 - length1 - length2 - length3;   // check coherency !!
-    const char* const start1 = (const char*)(cSrc) + 6;
-    const char* const start2 = start1 + length1;
-    const char* const start3 = start2 + length2;
-    const char* const start4 = start3 + length3;
-    FSE_DStream_t bitD1, bitD2, bitD3, bitD4;
-
-    if (length1+length2+length3+6 >= cSrcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
-
-    errorCode = FSE_initDStream(&bitD1, start1, length1);
-    if (FSE_isError(errorCode)) return errorCode;
-    errorCode = FSE_initDStream(&bitD2, start2, length2);
-    if (FSE_isError(errorCode)) return errorCode;
-    errorCode = FSE_initDStream(&bitD3, start3, length3);
-    if (FSE_isError(errorCode)) return errorCode;
-    errorCode = FSE_initDStream(&bitD4, start4, length4);
-    if (FSE_isError(errorCode)) return errorCode;
-
-    reloadStatus=FSE_reloadDStream(&bitD2);
-
-    /* 16 symbols per loop */
-    for ( ; (reloadStatus<FSE_DStream_completed) && (op<olimit);  /* D2-3-4 are supposed to be synchronized and finish together */
-        op+=16, reloadStatus = FSE_reloadDStream(&bitD2) | FSE_reloadDStream(&bitD3) | FSE_reloadDStream(&bitD4), FSE_reloadDStream(&bitD1))
+    if (cSrcSize < 6) return (size_t)-FSE_ERROR_srcSize_wrong;
     {
-#define HUF_DECODE_SYMBOL_0(n, Dstream) \
-        op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog);
-
-#define HUF_DECODE_SYMBOL_1(n, Dstream) \
-        op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog); \
-        if (FSE_32bits() && (HUF_MAX_TABLELOG>12)) FSE_reloadDStream(&Dstream)
-
-#define HUF_DECODE_SYMBOL_2(n, Dstream) \
-        op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog); \
-        if (FSE_32bits()) FSE_reloadDStream(&Dstream)
-
-        HUF_DECODE_SYMBOL_1( 0, bitD1);
-        HUF_DECODE_SYMBOL_1( 1, bitD2);
-        HUF_DECODE_SYMBOL_1( 2, bitD3);
-        HUF_DECODE_SYMBOL_1( 3, bitD4);
-        HUF_DECODE_SYMBOL_2( 4, bitD1);
-        HUF_DECODE_SYMBOL_2( 5, bitD2);
-        HUF_DECODE_SYMBOL_2( 6, bitD3);
-        HUF_DECODE_SYMBOL_2( 7, bitD4);
-        HUF_DECODE_SYMBOL_1( 8, bitD1);
-        HUF_DECODE_SYMBOL_1( 9, bitD2);
-        HUF_DECODE_SYMBOL_1(10, bitD3);
-        HUF_DECODE_SYMBOL_1(11, bitD4);
-        HUF_DECODE_SYMBOL_0(12, bitD1);
-        HUF_DECODE_SYMBOL_0(13, bitD2);
-        HUF_DECODE_SYMBOL_0(14, bitD3);
-        HUF_DECODE_SYMBOL_0(15, bitD4);
-    }
+        BYTE* const ostart = (BYTE*) dst;
+        BYTE* op = ostart;
+        BYTE* const omax = op + maxDstSize;
+        BYTE* const olimit = omax-15;
+
+        const void* ptr = DTable;
+        const HUF_DElt* const dt = (const HUF_DElt*)(ptr)+1;
+        const U32 dtLog = DTable[0];
+        size_t errorCode;
+        U32 reloadStatus;
+
+        /* Init */
+
+        const U16* jumpTable = (const U16*)cSrc;
+        const size_t length1 = FSE_readLE16(jumpTable);
+        const size_t length2 = FSE_readLE16(jumpTable+1);
+        const size_t length3 = FSE_readLE16(jumpTable+2);
+        const size_t length4 = cSrcSize - 6 - length1 - length2 - length3;   // check coherency !!
+        const char* const start1 = (const char*)(cSrc) + 6;
+        const char* const start2 = start1 + length1;
+        const char* const start3 = start2 + length2;
+        const char* const start4 = start3 + length3;
+        FSE_DStream_t bitD1, bitD2, bitD3, bitD4;
+
+        if (length1+length2+length3+6 >= cSrcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
+
+        errorCode = FSE_initDStream(&bitD1, start1, length1);
+        if (FSE_isError(errorCode)) return errorCode;
+        errorCode = FSE_initDStream(&bitD2, start2, length2);
+        if (FSE_isError(errorCode)) return errorCode;
+        errorCode = FSE_initDStream(&bitD3, start3, length3);
+        if (FSE_isError(errorCode)) return errorCode;
+        errorCode = FSE_initDStream(&bitD4, start4, length4);
+        if (FSE_isError(errorCode)) return errorCode;
+
+        reloadStatus=FSE_reloadDStream(&bitD2);
+
+        /* 16 symbols per loop */
+        for ( ; (reloadStatus<FSE_DStream_completed) && (op<olimit);  /* D2-3-4 are supposed to be synchronized and finish together */
+            op+=16, reloadStatus = FSE_reloadDStream(&bitD2) | FSE_reloadDStream(&bitD3) | FSE_reloadDStream(&bitD4), FSE_reloadDStream(&bitD1))
+        {
+    #define HUF_DECODE_SYMBOL_0(n, Dstream) \
+            op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog);
+
+    #define HUF_DECODE_SYMBOL_1(n, Dstream) \
+            op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog); \
+            if (FSE_32bits() && (HUF_MAX_TABLELOG>12)) FSE_reloadDStream(&Dstream)
+
+    #define HUF_DECODE_SYMBOL_2(n, Dstream) \
+            op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog); \
+            if (FSE_32bits()) FSE_reloadDStream(&Dstream)
+
+            HUF_DECODE_SYMBOL_1( 0, bitD1);
+            HUF_DECODE_SYMBOL_1( 1, bitD2);
+            HUF_DECODE_SYMBOL_1( 2, bitD3);
+            HUF_DECODE_SYMBOL_1( 3, bitD4);
+            HUF_DECODE_SYMBOL_2( 4, bitD1);
+            HUF_DECODE_SYMBOL_2( 5, bitD2);
+            HUF_DECODE_SYMBOL_2( 6, bitD3);
+            HUF_DECODE_SYMBOL_2( 7, bitD4);
+            HUF_DECODE_SYMBOL_1( 8, bitD1);
+            HUF_DECODE_SYMBOL_1( 9, bitD2);
+            HUF_DECODE_SYMBOL_1(10, bitD3);
+            HUF_DECODE_SYMBOL_1(11, bitD4);
+            HUF_DECODE_SYMBOL_0(12, bitD1);
+            HUF_DECODE_SYMBOL_0(13, bitD2);
+            HUF_DECODE_SYMBOL_0(14, bitD3);
+            HUF_DECODE_SYMBOL_0(15, bitD4);
+        }
 
-    if (reloadStatus!=FSE_DStream_completed)   /* not complete : some bitStream might be FSE_DStream_unfinished */
-        return (size_t)-FSE_ERROR_corruptionDetected;
+        if (reloadStatus!=FSE_DStream_completed)   /* not complete : some bitStream might be FSE_DStream_unfinished */
+            return (size_t)-FSE_ERROR_corruptionDetected;
 
-    /* tail */
-    {
-        // bitTail = bitD1;   // *much* slower : -20% !??!
-        FSE_DStream_t bitTail;
-        bitTail.ptr = bitD1.ptr;
-        bitTail.bitsConsumed = bitD1.bitsConsumed;
-        bitTail.bitContainer = bitD1.bitContainer;   // required in case of FSE_DStream_endOfBuffer
-        bitTail.start = start1;
-        for ( ; (FSE_reloadDStream(&bitTail) < FSE_DStream_completed) && (op<omax) ; op++)
+        /* tail */
         {
-            HUF_DECODE_SYMBOL_0(0, bitTail);
-        }
+            // bitTail = bitD1;   // *much* slower : -20% !??!
+            FSE_DStream_t bitTail;
+            bitTail.ptr = bitD1.ptr;
+            bitTail.bitsConsumed = bitD1.bitsConsumed;
+            bitTail.bitContainer = bitD1.bitContainer;   // required in case of FSE_DStream_endOfBuffer
+            bitTail.start = start1;
+            for ( ; (FSE_reloadDStream(&bitTail) < FSE_DStream_completed) && (op<omax) ; op++)
+            {
+                HUF_DECODE_SYMBOL_0(0, bitTail);
+            }
 
-        if (FSE_endOfDStream(&bitTail))
-            return op-ostart;
-    }
+            if (FSE_endOfDStream(&bitTail))
+                return op-ostart;
+        }
 
-    if (op==omax) return (size_t)-FSE_ERROR_dstSize_tooSmall;   /* dst buffer is full, but cSrc unfinished */
+        if (op==omax) return (size_t)-FSE_ERROR_dstSize_tooSmall;   /* dst buffer is full, but cSrc unfinished */
 
-    return (size_t)-FSE_ERROR_corruptionDetected;
+        return (size_t)-FSE_ERROR_corruptionDetected;
+    }
 }
 
 
@@ -1355,8 +1358,6 @@ static unsigned ZSTD_isLittleEndian(void)
 
 static U16    ZSTD_read16(const void* p) { U16 r; memcpy(&r, p, sizeof(r)); return r; }
 
-static U32    ZSTD_read32(const void* p) { U32 r; memcpy(&r, p, sizeof(r)); return r; }
-
 static void   ZSTD_copy4(void* dst, const void* src) { memcpy(dst, src, 4); }
 
 static void   ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
@@ -1381,16 +1382,9 @@ static U16 ZSTD_readLE16(const void* memPtr)
     }
 }
 
-
-static U32 ZSTD_readLE32(const void* memPtr)
+static U32 ZSTD_readLE24(const void* memPtr)
 {
-    if (ZSTD_isLittleEndian())
-        return ZSTD_read32(memPtr);
-    else
-    {
-        const BYTE* p = (const BYTE*)memPtr;
-        return (U32)((U32)p[0] + ((U32)p[1]<<8) + ((U32)p[2]<<16) + ((U32)p[3]<<24));
-    }
+    return ZSTD_readLE16(memPtr) + (((const BYTE*)memPtr)[2] << 16);
 }
 
 static U32 ZSTD_readBE32(const void* memPtr)
@@ -1704,13 +1698,13 @@ static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
     seqState->prevOffset = seq->offset;
     if (litLength == MaxLL)
     {
-        U32 add = dumps<de ? *dumps++ : 0;
+        const U32 add = dumps<de ? *dumps++ : 0;
         if (add < 255) litLength += add;
         else
         {
             if (dumps<=(de-3))
             {
-                litLength = ZSTD_readLE32(dumps) & 0xFFFFFF;  /* no pb : dumps is always followed by seq tables > 1 byte */
+                litLength = ZSTD_readLE24(dumps);
                 dumps += 3;
             }
         }
@@ -1732,13 +1726,13 @@ static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
     matchLength = FSE_decodeSymbol(&(seqState->stateML), &(seqState->DStream));
     if (matchLength == MaxML)
     {
-        U32 add = dumps<de ? *dumps++ : 0;
+        const U32 add = dumps<de ? *dumps++ : 0;
         if (add < 255) matchLength += add;
         else
         {
             if (dumps<=(de-3))
             {
-                matchLength = ZSTD_readLE32(dumps) & 0xFFFFFF;  /* no pb : dumps is always followed by seq tables > 1 byte */
+                matchLength = ZSTD_readLE24(dumps);
                 dumps += 3;
             }
         }

data/ext/zstdruby/libzstd/legacy/zstd_v02.c

@@ -217,6 +217,11 @@ MEM_STATIC void MEM_writeLE16(void* memPtr, U16 val)
     }
 }
 
+MEM_STATIC U32 MEM_readLE24(const void* memPtr)
+{
+    return MEM_readLE16(memPtr) + (((const BYTE*)memPtr)[2] << 16);
+}
+
 MEM_STATIC U32 MEM_readLE32(const void* memPtr)
 {
     if (MEM_isLittleEndian())
@@ -3043,11 +3048,11 @@ static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
     seqState->prevOffset = seq->offset;
     if (litLength == MaxLL)
     {
-        U32 add = *dumps++;
+        const U32 add = dumps<de ? *dumps++ : 0;
         if (add < 255) litLength += add;
-        else
+        else if (dumps + 3 <= de)
         {
-            litLength = MEM_readLE32(dumps) & 0xFFFFFF;  /* no pb : dumps is always followed by seq tables > 1 byte */
+            litLength = MEM_readLE24(dumps);
             dumps += 3;
         }
         if (dumps >= de) dumps = de-1;   /* late correction, to avoid read overflow (data is now corrupted anyway) */
@@ -3073,11 +3078,11 @@ static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
     matchLength = FSE_decodeSymbol(&(seqState->stateML), &(seqState->DStream));
     if (matchLength == MaxML)
     {
-        U32 add = *dumps++;
+        const U32 add = dumps<de ? *dumps++ : 0;
         if (add < 255) matchLength += add;
-        else
+        else if (dumps + 3 <= de)
         {
-            matchLength = MEM_readLE32(dumps) & 0xFFFFFF;  /* no pb : dumps is always followed by seq tables > 1 byte */
+            matchLength = MEM_readLE24(dumps);
             dumps += 3;
         }
         if (dumps >= de) dumps = de-1;   /* late correction, to avoid read overflow (data is now corrupted anyway) */

data/ext/zstdruby/libzstd/legacy/zstd_v03.c

@@ -219,6 +219,11 @@ MEM_STATIC void MEM_writeLE16(void* memPtr, U16 val)
     }
 }
 
+MEM_STATIC U32 MEM_readLE24(const void* memPtr)
+{
+    return MEM_readLE16(memPtr) + (((const BYTE*)memPtr)[2] << 16);
+}
+
 MEM_STATIC U32 MEM_readLE32(const void* memPtr)
 {
     if (MEM_isLittleEndian())
@@ -2684,11 +2689,11 @@ static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
     seqState->prevOffset = seq->offset;
     if (litLength == MaxLL)
     {
-        U32 add = *dumps++;
+        const U32 add = dumps<de ? *dumps++ : 0;
         if (add < 255) litLength += add;
-        else
+        else if (dumps + 3 <= de)
         {
-            litLength = MEM_readLE32(dumps) & 0xFFFFFF;  /* no pb : dumps is always followed by seq tables > 1 byte */
+            litLength = MEM_readLE24(dumps);
             dumps += 3;
         }
         if (dumps >= de) dumps = de-1;   /* late correction, to avoid read overflow (data is now corrupted anyway) */
@@ -2714,11 +2719,11 @@ static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
     matchLength = FSE_decodeSymbol(&(seqState->stateML), &(seqState->DStream));
     if (matchLength == MaxML)
     {
-        U32 add = *dumps++;
+        const U32 add = dumps<de ? *dumps++ : 0;
         if (add < 255) matchLength += add;
-        else
+        else if (dumps + 3 <= de)
         {
-            matchLength = MEM_readLE32(dumps) & 0xFFFFFF;  /* no pb : dumps is always followed by seq tables > 1 byte */
+            matchLength = MEM_readLE24(dumps);
             dumps += 3;
         }
         if (dumps >= de) dumps = de-1;   /* late correction, to avoid read overflow (data is now corrupted anyway) */

data/ext/zstdruby/libzstd/legacy/zstd_v04.c

@@ -189,6 +189,11 @@ MEM_STATIC void MEM_writeLE16(void* memPtr, U16 val)
     }
 }
 
+MEM_STATIC U32 MEM_readLE24(const void* memPtr)
+{
+    return MEM_readLE16(memPtr) + (((const BYTE*)memPtr)[2] << 16);
+}
+
 MEM_STATIC U32 MEM_readLE32(const void* memPtr)
 {
     if (MEM_isLittleEndian())
@@ -2808,13 +2813,12 @@ static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
     litLength = FSE_decodeSymbol(&(seqState->stateLL), &(seqState->DStream));
     prevOffset = litLength ? seq->offset : seqState->prevOffset;
     if (litLength == MaxLL) {
-        U32 add = *dumps++;
+        const U32 add = dumps<de ? *dumps++ : 0;
         if (add < 255) litLength += add;
-        else {
-            litLength = dumps[0] + (dumps[1]<<8) + (dumps[2]<<16);
+        else if (dumps + 3 <= de) {
+            litLength = MEM_readLE24(dumps);
             dumps += 3;
         }
-        if (dumps > de) { litLength = MaxLL+255; }  /* late correction, to avoid using uninitialized memory */
         if (dumps >= de) { dumps = de-1; }  /* late correction, to avoid read overflow (data is now corrupted anyway) */
     }
 
@@ -2837,13 +2841,12 @@ static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
     /* MatchLength */
     matchLength = FSE_decodeSymbol(&(seqState->stateML), &(seqState->DStream));
     if (matchLength == MaxML) {
-        U32 add = *dumps++;
+        const U32 add = dumps<de ? *dumps++ : 0;
         if (add < 255) matchLength += add;
-        else {
-            matchLength = dumps[0] + (dumps[1]<<8) + (dumps[2]<<16);
+        else if (dumps + 3 <= de){
+            matchLength = MEM_readLE24(dumps);
             dumps += 3;
         }
-        if (dumps > de) { matchLength = MaxML+255; }  /* late correction, to avoid using uninitialized memory */
         if (dumps >= de) { dumps = de-1; }  /* late correction, to avoid read overflow (data is now corrupted anyway) */
     }
     matchLength += MINMATCH;