zstd-ruby 1.4.5.0 → 1.5.1.1

data/ext/zstdruby/libzstd/decompress/zstd_decompress.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -55,13 +55,13 @@
 /*-*******************************************************
 *  Dependencies
 *********************************************************/
-#include <string.h>      /* memcpy, memmove, memset */
-#include "../common/cpu.h"         /* bmi2 */
+#include "../common/zstd_deps.h"   /* ZSTD_memcpy, ZSTD_memmove, ZSTD_memset */
 #include "../common/mem.h"         /* low level memory routines */
 #define FSE_STATIC_LINKING_ONLY
 #include "../common/fse.h"
 #define HUF_STATIC_LINKING_ONLY
 #include "../common/huf.h"
+#include "../common/xxhash.h" /* XXH64_reset, XXH64_update, XXH64_digest, XXH64 */
 #include "../common/zstd_internal.h"  /* blockProperties_t */
 #include "zstd_decompress_internal.h"   /* ZSTD_DCtx */
 #include "zstd_ddict.h"  /* ZSTD_DDictDictContent */
@@ -72,6 +72,147 @@
 #endif
 
 
+
+/*************************************
+ * Multiple DDicts Hashset internals *
+ *************************************/
+
+#define DDICT_HASHSET_MAX_LOAD_FACTOR_COUNT_MULT 4
+#define DDICT_HASHSET_MAX_LOAD_FACTOR_SIZE_MULT 3   /* These two constants represent SIZE_MULT/COUNT_MULT load factor without using a float.
+                                                     * Currently, that means a 0.75 load factor.
+                                                     * So, if count * COUNT_MULT / size * SIZE_MULT != 0, then we've exceeded
+                                                     * the load factor of the ddict hash set.
+                                                     */
+
+#define DDICT_HASHSET_TABLE_BASE_SIZE 64
+#define DDICT_HASHSET_RESIZE_FACTOR 2
+
+/* Hash function to determine starting position of dict insertion within the table
+ * Returns an index between [0, hashSet->ddictPtrTableSize]
+ */
+static size_t ZSTD_DDictHashSet_getIndex(const ZSTD_DDictHashSet* hashSet, U32 dictID) {
+    const U64 hash = XXH64(&dictID, sizeof(U32), 0);
+    /* DDict ptr table size is a multiple of 2, use size - 1 as mask to get index within [0, hashSet->ddictPtrTableSize) */
+    return hash & (hashSet->ddictPtrTableSize - 1);
+}
+
+/* Adds DDict to a hashset without resizing it.
+ * If inserting a DDict with a dictID that already exists in the set, replaces the one in the set.
+ * Returns 0 if successful, or a zstd error code if something went wrong.
+ */
+static size_t ZSTD_DDictHashSet_emplaceDDict(ZSTD_DDictHashSet* hashSet, const ZSTD_DDict* ddict) {
+    const U32 dictID = ZSTD_getDictID_fromDDict(ddict);
+    size_t idx = ZSTD_DDictHashSet_getIndex(hashSet, dictID);
+    const size_t idxRangeMask = hashSet->ddictPtrTableSize - 1;
+    RETURN_ERROR_IF(hashSet->ddictPtrCount == hashSet->ddictPtrTableSize, GENERIC, "Hash set is full!");
+    DEBUGLOG(4, "Hashed index: for dictID: %u is %zu", dictID, idx);
+    while (hashSet->ddictPtrTable[idx] != NULL) {
+        /* Replace existing ddict if inserting ddict with same dictID */
+        if (ZSTD_getDictID_fromDDict(hashSet->ddictPtrTable[idx]) == dictID) {
+            DEBUGLOG(4, "DictID already exists, replacing rather than adding");
+            hashSet->ddictPtrTable[idx] = ddict;
+            return 0;
+        }
+        idx &= idxRangeMask;
+        idx++;
+    }
+    DEBUGLOG(4, "Final idx after probing for dictID %u is: %zu", dictID, idx);
+    hashSet->ddictPtrTable[idx] = ddict;
+    hashSet->ddictPtrCount++;
+    return 0;
+}
+
+/* Expands hash table by factor of DDICT_HASHSET_RESIZE_FACTOR and
+ * rehashes all values, allocates new table, frees old table.
+ * Returns 0 on success, otherwise a zstd error code.
+ */
+static size_t ZSTD_DDictHashSet_expand(ZSTD_DDictHashSet* hashSet, ZSTD_customMem customMem) {
+    size_t newTableSize = hashSet->ddictPtrTableSize * DDICT_HASHSET_RESIZE_FACTOR;
+    const ZSTD_DDict** newTable = (const ZSTD_DDict**)ZSTD_customCalloc(sizeof(ZSTD_DDict*) * newTableSize, customMem);
+    const ZSTD_DDict** oldTable = hashSet->ddictPtrTable;
+    size_t oldTableSize = hashSet->ddictPtrTableSize;
+    size_t i;
+
+    DEBUGLOG(4, "Expanding DDict hash table! Old size: %zu new size: %zu", oldTableSize, newTableSize);
+    RETURN_ERROR_IF(!newTable, memory_allocation, "Expanded hashset allocation failed!");
+    hashSet->ddictPtrTable = newTable;
+    hashSet->ddictPtrTableSize = newTableSize;
+    hashSet->ddictPtrCount = 0;
+    for (i = 0; i < oldTableSize; ++i) {
+        if (oldTable[i] != NULL) {
+            FORWARD_IF_ERROR(ZSTD_DDictHashSet_emplaceDDict(hashSet, oldTable[i]), "");
+        }
+    }
+    ZSTD_customFree((void*)oldTable, customMem);
+    DEBUGLOG(4, "Finished re-hash");
+    return 0;
+}
+
+/* Fetches a DDict with the given dictID
+ * Returns the ZSTD_DDict* with the requested dictID. If it doesn't exist, then returns NULL.
+ */
+static const ZSTD_DDict* ZSTD_DDictHashSet_getDDict(ZSTD_DDictHashSet* hashSet, U32 dictID) {
+    size_t idx = ZSTD_DDictHashSet_getIndex(hashSet, dictID);
+    const size_t idxRangeMask = hashSet->ddictPtrTableSize - 1;
+    DEBUGLOG(4, "Hashed index: for dictID: %u is %zu", dictID, idx);
+    for (;;) {
+        size_t currDictID = ZSTD_getDictID_fromDDict(hashSet->ddictPtrTable[idx]);
+        if (currDictID == dictID || currDictID == 0) {
+            /* currDictID == 0 implies a NULL ddict entry */
+            break;
+        } else {
+            idx &= idxRangeMask;    /* Goes to start of table when we reach the end */
+            idx++;
+        }
+    }
+    DEBUGLOG(4, "Final idx after probing for dictID %u is: %zu", dictID, idx);
+    return hashSet->ddictPtrTable[idx];
+}
+
+/* Allocates space for and returns a ddict hash set
+ * The hash set's ZSTD_DDict* table has all values automatically set to NULL to begin with.
+ * Returns NULL if allocation failed.
+ */
+static ZSTD_DDictHashSet* ZSTD_createDDictHashSet(ZSTD_customMem customMem) {
+    ZSTD_DDictHashSet* ret = (ZSTD_DDictHashSet*)ZSTD_customMalloc(sizeof(ZSTD_DDictHashSet), customMem);
+    DEBUGLOG(4, "Allocating new hash set");
+    if (!ret)
+        return NULL;
+    ret->ddictPtrTable = (const ZSTD_DDict**)ZSTD_customCalloc(DDICT_HASHSET_TABLE_BASE_SIZE * sizeof(ZSTD_DDict*), customMem);
+    if (!ret->ddictPtrTable) {
+        ZSTD_customFree(ret, customMem);
+        return NULL;
+    }
+    ret->ddictPtrTableSize = DDICT_HASHSET_TABLE_BASE_SIZE;
+    ret->ddictPtrCount = 0;
+    return ret;
+}
+
+/* Frees the table of ZSTD_DDict* within a hashset, then frees the hashset itself.
+ * Note: The ZSTD_DDict* within the table are NOT freed.
+ */
+static void ZSTD_freeDDictHashSet(ZSTD_DDictHashSet* hashSet, ZSTD_customMem customMem) {
+    DEBUGLOG(4, "Freeing ddict hash set");
+    if (hashSet && hashSet->ddictPtrTable) {
+        ZSTD_customFree((void*)hashSet->ddictPtrTable, customMem);
+    }
+    if (hashSet) {
+        ZSTD_customFree(hashSet, customMem);
+    }
+}
+
+/* Public function: Adds a DDict into the ZSTD_DDictHashSet, possibly triggering a resize of the hash set.
+ * Returns 0 on success, or a ZSTD error.
+ */
+static size_t ZSTD_DDictHashSet_addDDict(ZSTD_DDictHashSet* hashSet, const ZSTD_DDict* ddict, ZSTD_customMem customMem) {
+    DEBUGLOG(4, "Adding dict ID: %u to hashset with - Count: %zu Tablesize: %zu", ZSTD_getDictID_fromDDict(ddict), hashSet->ddictPtrCount, hashSet->ddictPtrTableSize);
+    if (hashSet->ddictPtrCount * DDICT_HASHSET_MAX_LOAD_FACTOR_COUNT_MULT / hashSet->ddictPtrTableSize * DDICT_HASHSET_MAX_LOAD_FACTOR_SIZE_MULT != 0) {
+        FORWARD_IF_ERROR(ZSTD_DDictHashSet_expand(hashSet, customMem), "");
+    }
+    FORWARD_IF_ERROR(ZSTD_DDictHashSet_emplaceDDict(hashSet, ddict), "");
+    return 0;
+}
+
 /*-*************************************************************
 *   Context management
 ***************************************************************/
@@ -94,11 +235,19 @@ static size_t ZSTD_startingInputLength(ZSTD_format_e format)
     return startingInputLength;
 }
 
+static void ZSTD_DCtx_resetParameters(ZSTD_DCtx* dctx)
+{
+    assert(dctx->streamStage == zdss_init);
+    dctx->format = ZSTD_f_zstd1;
+    dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
+    dctx->outBufferMode = ZSTD_bm_buffered;
+    dctx->forceIgnoreChecksum = ZSTD_d_validateChecksum;
+    dctx->refMultipleDDicts = ZSTD_rmd_refSingleDDict;
+}
+
 static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx)
 {
-    dctx->format = ZSTD_f_zstd1;  /* ZSTD_decompressBegin() invokes ZSTD_startingInputLength() with argument dctx->format */
     dctx->staticSize  = 0;
-    dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
     dctx->ddict       = NULL;
     dctx->ddictLocal  = NULL;
     dctx->dictEnd     = NULL;
@@ -108,12 +257,17 @@ static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx)
     dctx->inBuffSize  = 0;
     dctx->outBuffSize = 0;
     dctx->streamStage = zdss_init;
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
     dctx->legacyContext = NULL;
     dctx->previousLegacyVersion = 0;
+#endif
     dctx->noForwardProgress = 0;
     dctx->oversizedDuration = 0;
-    dctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
-    dctx->outBufferMode = ZSTD_obm_buffered;
+#if DYNAMIC_BMI2
+    dctx->bmi2 = ZSTD_cpuSupportsBmi2();
+#endif
+    dctx->ddictSet = NULL;
+    ZSTD_DCtx_resetParameters(dctx);
 #ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
     dctx->dictContentEndForFuzzing = NULL;
 #endif
@@ -132,11 +286,10 @@ ZSTD_DCtx* ZSTD_initStaticDCtx(void *workspace, size_t workspaceSize)
     return dctx;
 }
 
-ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem)
-{
-    if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
+static ZSTD_DCtx* ZSTD_createDCtx_internal(ZSTD_customMem customMem) {
+    if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
 
-    {   ZSTD_DCtx* const dctx = (ZSTD_DCtx*)ZSTD_malloc(sizeof(*dctx), customMem);
+    {   ZSTD_DCtx* const dctx = (ZSTD_DCtx*)ZSTD_customMalloc(sizeof(*dctx), customMem);
         if (!dctx) return NULL;
         dctx->customMem = customMem;
         ZSTD_initDCtx_internal(dctx);
@@ -144,10 +297,15 @@ ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem)
     }
 }
 
+ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem)
+{
+    return ZSTD_createDCtx_internal(customMem);
+}
+
 ZSTD_DCtx* ZSTD_createDCtx(void)
 {
     DEBUGLOG(3, "ZSTD_createDCtx");
-    return ZSTD_createDCtx_advanced(ZSTD_defaultCMem);
+    return ZSTD_createDCtx_internal(ZSTD_defaultCMem);
 }
 
 static void ZSTD_clearDict(ZSTD_DCtx* dctx)
@@ -164,13 +322,17 @@ size_t ZSTD_freeDCtx(ZSTD_DCtx* dctx)
     RETURN_ERROR_IF(dctx->staticSize, memory_allocation, "not compatible with static DCtx");
     {   ZSTD_customMem const cMem = dctx->customMem;
         ZSTD_clearDict(dctx);
-        ZSTD_free(dctx->inBuff, cMem);
+        ZSTD_customFree(dctx->inBuff, cMem);
         dctx->inBuff = NULL;
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
         if (dctx->legacyContext)
             ZSTD_freeLegacyStreamContext(dctx->legacyContext, dctx->previousLegacyVersion);
 #endif
-        ZSTD_free(dctx, cMem);
+        if (dctx->ddictSet) {
+            ZSTD_freeDDictHashSet(dctx->ddictSet, cMem);
+            dctx->ddictSet = NULL;
+        }
+        ZSTD_customFree(dctx, cMem);
         return 0;
     }
 }
@@ -179,7 +341,30 @@ size_t ZSTD_freeDCtx(ZSTD_DCtx* dctx)
 void ZSTD_copyDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx)
 {
     size_t const toCopy = (size_t)((char*)(&dstDCtx->inBuff) - (char*)dstDCtx);
-    memcpy(dstDCtx, srcDCtx, toCopy);  /* no need to copy workspace */
+    ZSTD_memcpy(dstDCtx, srcDCtx, toCopy);  /* no need to copy workspace */
+}
+
+/* Given a dctx with a digested frame params, re-selects the correct ZSTD_DDict based on
+ * the requested dict ID from the frame. If there exists a reference to the correct ZSTD_DDict, then
+ * accordingly sets the ddict to be used to decompress the frame.
+ *
+ * If no DDict is found, then no action is taken, and the ZSTD_DCtx::ddict remains as-is.
+ *
+ * ZSTD_d_refMultipleDDicts must be enabled for this function to be called.
+ */
+static void ZSTD_DCtx_selectFrameDDict(ZSTD_DCtx* dctx) {
+    assert(dctx->refMultipleDDicts && dctx->ddictSet);
+    DEBUGLOG(4, "Adjusting DDict based on requested dict ID from frame");
+    if (dctx->ddict) {
+        const ZSTD_DDict* frameDDict = ZSTD_DDictHashSet_getDDict(dctx->ddictSet, dctx->fParams.dictID);
+        if (frameDDict) {
+            DEBUGLOG(4, "DDict found!");
+            ZSTD_clearDict(dctx);
+            dctx->dictID = dctx->fParams.dictID;
+            dctx->ddict = frameDDict;
+            dctx->dictUses = ZSTD_use_indefinitely;
+        }
+    }
 }
 
 
@@ -205,6 +390,19 @@ unsigned ZSTD_isFrame(const void* buffer, size_t size)
     return 0;
 }
 
+/*! ZSTD_isSkippableFrame() :
+ *  Tells if the content of `buffer` starts with a valid Frame Identifier for a skippable frame.
+ *  Note : Frame Identifier is 4 bytes. If `size < 4`, @return will always be 0.
+ */
+unsigned ZSTD_isSkippableFrame(const void* buffer, size_t size)
+{
+    if (size < ZSTD_FRAMEIDSIZE) return 0;
+    {   U32 const magic = MEM_readLE32(buffer);
+        if ((magic & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
+    }
+    return 0;
+}
+
 /** ZSTD_frameHeaderSize_internal() :
  *  srcSize must be large enough to reach header size fields.
  *  note : only works for formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless.
@@ -246,7 +444,7 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
     const BYTE* ip = (const BYTE*)src;
     size_t const minInputSize = ZSTD_startingInputLength(format);
 
-    memset(zfhPtr, 0, sizeof(*zfhPtr));   /* not strictly necessary, but static analyzer do not understand that zfhPtr is only going to be read only if return value is zero, since they are 2 different signals */
+    ZSTD_memset(zfhPtr, 0, sizeof(*zfhPtr));   /* not strictly necessary, but static analyzer do not understand that zfhPtr is only going to be read only if return value is zero, since they are 2 different signals */
     if (srcSize < minInputSize) return minInputSize;
     RETURN_ERROR_IF(src==NULL, GENERIC, "invalid parameter");
 
@@ -256,7 +454,7 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
             /* skippable frame */
             if (srcSize < ZSTD_SKIPPABLEHEADERSIZE)
                 return ZSTD_SKIPPABLEHEADERSIZE; /* magic number + frame length */
-            memset(zfhPtr, 0, sizeof(*zfhPtr));
+            ZSTD_memset(zfhPtr, 0, sizeof(*zfhPtr));
             zfhPtr->frameContentSize = MEM_readLE32((const char *)src + ZSTD_FRAMEIDSIZE);
             zfhPtr->frameType = ZSTD_skippableFrame;
             return 0;
@@ -291,7 +489,9 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
         }
         switch(dictIDSizeCode)
         {
-            default: assert(0);  /* impossible */
+            default:
+                assert(0);  /* impossible */
+                ZSTD_FALLTHROUGH;
             case 0 : break;
             case 1 : dictID = ip[pos]; pos++; break;
             case 2 : dictID = MEM_readLE16(ip+pos); pos+=2; break;
@@ -299,7 +499,9 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
         }
         switch(fcsID)
         {
-            default: assert(0);  /* impossible */
+            default:
+                assert(0);  /* impossible */
+                ZSTD_FALLTHROUGH;
             case 0 : if (singleSegment) frameContentSize = ip[pos]; break;
             case 1 : frameContentSize = MEM_readLE16(ip+pos)+256; break;
             case 2 : frameContentSize = MEM_readLE32(ip+pos); break;
@@ -328,7 +530,6 @@ size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t src
     return ZSTD_getFrameHeader_advanced(zfhPtr, src, srcSize, ZSTD_f_zstd1);
 }
 
-
 /** ZSTD_getFrameContentSize() :
  *  compatible with legacy mode
  * @return : decompressed size of the single frame pointed to be `src` if known, otherwise
@@ -369,6 +570,37 @@ static size_t readSkippableFrameSize(void const* src, size_t srcSize)
     }
 }
 
+/*! ZSTD_readSkippableFrame() :
+ * Retrieves a zstd skippable frame containing data given by src, and writes it to dst buffer.
+ *
+ * The parameter magicVariant will receive the magicVariant that was supplied when the frame was written,
+ * i.e. magicNumber - ZSTD_MAGIC_SKIPPABLE_START.  This can be NULL if the caller is not interested
+ * in the magicVariant.
+ *
+ * Returns an error if destination buffer is not large enough, or if the frame is not skippable.
+ *
+ * @return : number of bytes written or a ZSTD error.
+ */
+ZSTDLIB_API size_t ZSTD_readSkippableFrame(void* dst, size_t dstCapacity, unsigned* magicVariant,
+                                            const void* src, size_t srcSize)
+{
+    U32 const magicNumber = MEM_readLE32(src);
+    size_t skippableFrameSize = readSkippableFrameSize(src, srcSize);
+    size_t skippableContentSize = skippableFrameSize - ZSTD_SKIPPABLEHEADERSIZE;
+
+    /* check input validity */
+    RETURN_ERROR_IF(!ZSTD_isSkippableFrame(src, srcSize), frameParameter_unsupported, "");
+    RETURN_ERROR_IF(skippableFrameSize < ZSTD_SKIPPABLEHEADERSIZE || skippableFrameSize > srcSize, srcSize_wrong, "");
+    RETURN_ERROR_IF(skippableContentSize > dstCapacity, dstSize_tooSmall, "");
+
+    /* deliver payload */
+    if (skippableContentSize > 0  && dst != NULL)
+        ZSTD_memcpy(dst, (const BYTE *)src + ZSTD_SKIPPABLEHEADERSIZE, skippableContentSize);
+    if (magicVariant != NULL)
+        *magicVariant = magicNumber - ZSTD_MAGIC_SKIPPABLE_START;
+    return skippableContentSize;
+}
+
 /** ZSTD_findDecompressedSize() :
  *  compatible with legacy mode
  *  `srcSize` must be the exact length of some number of ZSTD compressed and/or
@@ -433,12 +665,19 @@ unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize)
 
 /** ZSTD_decodeFrameHeader() :
  * `headerSize` must be the size provided by ZSTD_frameHeaderSize().
+ * If multiple DDict references are enabled, also will choose the correct DDict to use.
  * @return : 0 if success, or an error code, which can be tested using ZSTD_isError() */
 static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t headerSize)
 {
     size_t const result = ZSTD_getFrameHeader_advanced(&(dctx->fParams), src, headerSize, dctx->format);
     if (ZSTD_isError(result)) return result;    /* invalid header */
     RETURN_ERROR_IF(result>0, srcSize_wrong, "headerSize too small");
+
+    /* Reference DDict requested by frame if dctx references multiple ddicts */
+    if (dctx->refMultipleDDicts == ZSTD_rmd_refMultipleDDicts && dctx->ddictSet) {
+        ZSTD_DCtx_selectFrameDDict(dctx);
+    }
+
 #ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
     /* Skip the dictID check in fuzzing mode, because it makes the search
      * harder.
@@ -446,7 +685,9 @@ static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t he
     RETURN_ERROR_IF(dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID),
                     dictionary_wrong, "");
 #endif
-    if (dctx->fParams.checksumFlag) XXH64_reset(&dctx->xxhState, 0);
+    dctx->validateChecksum = (dctx->fParams.checksumFlag && !dctx->forceIgnoreChecksum) ? 1 : 0;
+    if (dctx->validateChecksum) XXH64_reset(&dctx->xxhState, 0);
+    dctx->processedCSize += headerSize;
     return 0;
 }
 
@@ -461,7 +702,7 @@ static ZSTD_frameSizeInfo ZSTD_errorFrameSizeInfo(size_t ret)
 static ZSTD_frameSizeInfo ZSTD_findFrameSizeInfo(const void* src, size_t srcSize)
 {
     ZSTD_frameSizeInfo frameSizeInfo;
-    memset(&frameSizeInfo, 0, sizeof(ZSTD_frameSizeInfo));
+    ZSTD_memset(&frameSizeInfo, 0, sizeof(ZSTD_frameSizeInfo));
 
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
     if (ZSTD_isLegacy(src, srcSize))
@@ -516,7 +757,7 @@ static ZSTD_frameSizeInfo ZSTD_findFrameSizeInfo(const void* src, size_t srcSize
             ip += 4;
         }
 
-        frameSizeInfo.compressedSize = ip - ipstart;
+        frameSizeInfo.compressedSize = (size_t)(ip - ipstart);
         frameSizeInfo.decompressedBound = (zfh.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN)
                                         ? zfh.frameContentSize
                                         : nbBlocks * zfh.blockSizeMax;
@@ -569,7 +810,7 @@ unsigned long long ZSTD_decompressBound(const void* src, size_t srcSize)
 size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize)
 {
     DEBUGLOG(5, "ZSTD_insertBlock: %u bytes", (unsigned)blockSize);
-    ZSTD_checkContinuity(dctx, blockStart);
+    ZSTD_checkContinuity(dctx, blockStart, blockSize);
     dctx->previousDstEnd = (const char*)blockStart + blockSize;
     return blockSize;
 }
@@ -579,12 +820,12 @@ static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity,
                           const void* src, size_t srcSize)
 {
     DEBUGLOG(5, "ZSTD_copyRawBlock");
+    RETURN_ERROR_IF(srcSize > dstCapacity, dstSize_tooSmall, "");
     if (dst == NULL) {
         if (srcSize == 0) return 0;
         RETURN_ERROR(dstBuffer_null, "");
     }
-    RETURN_ERROR_IF(srcSize > dstCapacity, dstSize_tooSmall, "");
-    memcpy(dst, src, srcSize);
+    ZSTD_memcpy(dst, src, srcSize);
     return srcSize;
 }
 
@@ -592,15 +833,41 @@ static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity,
                                BYTE b,
                                size_t regenSize)
 {
+    RETURN_ERROR_IF(regenSize > dstCapacity, dstSize_tooSmall, "");
     if (dst == NULL) {
         if (regenSize == 0) return 0;
         RETURN_ERROR(dstBuffer_null, "");
     }
-    RETURN_ERROR_IF(regenSize > dstCapacity, dstSize_tooSmall, "");
-    memset(dst, b, regenSize);
+    ZSTD_memset(dst, b, regenSize);
     return regenSize;
 }
 
+static void ZSTD_DCtx_trace_end(ZSTD_DCtx const* dctx, U64 uncompressedSize, U64 compressedSize, unsigned streaming)
+{
+#if ZSTD_TRACE
+    if (dctx->traceCtx && ZSTD_trace_decompress_end != NULL) {
+        ZSTD_Trace trace;
+        ZSTD_memset(&trace, 0, sizeof(trace));
+        trace.version = ZSTD_VERSION_NUMBER;
+        trace.streaming = streaming;
+        if (dctx->ddict) {
+            trace.dictionaryID = ZSTD_getDictID_fromDDict(dctx->ddict);
+            trace.dictionarySize = ZSTD_DDict_dictSize(dctx->ddict);
+            trace.dictionaryIsCold = dctx->ddictIsCold;
+        }
+        trace.uncompressedSize = (size_t)uncompressedSize;
+        trace.compressedSize = (size_t)compressedSize;
+        trace.dctx = dctx;
+        ZSTD_trace_decompress_end(dctx->traceCtx, &trace);
+    }
+#else
+    (void)dctx;
+    (void)uncompressedSize;
+    (void)compressedSize;
+    (void)streaming;
+#endif
+}
+
 
 /*! ZSTD_decompressFrame() :
  * @dctx must be properly initialized
@@ -610,8 +877,9 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
                                    void* dst, size_t dstCapacity,
                              const void** srcPtr, size_t *srcSizePtr)
 {
-    const BYTE* ip = (const BYTE*)(*srcPtr);
-    BYTE* const ostart = (BYTE* const)dst;
+    const BYTE* const istart = (const BYTE*)(*srcPtr);
+    const BYTE* ip = istart;
+    BYTE* const ostart = (BYTE*)dst;
     BYTE* const oend = dstCapacity != 0 ? ostart + dstCapacity : ostart;
     BYTE* op = ostart;
     size_t remainingSrcSize = *srcSizePtr;
@@ -647,13 +915,13 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
         switch(blockProperties.blockType)
         {
         case bt_compressed:
-            decodedSize = ZSTD_decompressBlock_internal(dctx, op, oend-op, ip, cBlockSize, /* frame */ 1);
+            decodedSize = ZSTD_decompressBlock_internal(dctx, op, (size_t)(oend-op), ip, cBlockSize, /* frame */ 1, not_streaming);
             break;
         case bt_raw :
-            decodedSize = ZSTD_copyRawBlock(op, oend-op, ip, cBlockSize);
+            decodedSize = ZSTD_copyRawBlock(op, (size_t)(oend-op), ip, cBlockSize);
             break;
         case bt_rle :
-            decodedSize = ZSTD_setRleBlock(op, oend-op, *ip, blockProperties.origSize);
+            decodedSize = ZSTD_setRleBlock(op, (size_t)(oend-op), *ip, blockProperties.origSize);
             break;
         case bt_reserved :
         default:
@@ -661,7 +929,7 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
         }
 
         if (ZSTD_isError(decodedSize)) return decodedSize;
-        if (dctx->fParams.checksumFlag)
+        if (dctx->validateChecksum)
             XXH64_update(&dctx->xxhState, op, decodedSize);
         if (decodedSize != 0)
             op += decodedSize;
@@ -676,19 +944,21 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
                         corruption_detected, "");
     }
     if (dctx->fParams.checksumFlag) { /* Frame content checksum verification */
-        U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState);
-        U32 checkRead;
         RETURN_ERROR_IF(remainingSrcSize<4, checksum_wrong, "");
-        checkRead = MEM_readLE32(ip);
-        RETURN_ERROR_IF(checkRead != checkCalc, checksum_wrong, "");
+        if (!dctx->forceIgnoreChecksum) {
+            U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState);
+            U32 checkRead;
+            checkRead = MEM_readLE32(ip);
+            RETURN_ERROR_IF(checkRead != checkCalc, checksum_wrong, "");
+        }
         ip += 4;
         remainingSrcSize -= 4;
     }
-
+    ZSTD_DCtx_trace_end(dctx, (U64)(op-ostart), (U64)(ip-istart), /* streaming */ 0);
     /* Allow caller to get size read */
     *srcPtr = ip;
     *srcSizePtr = remainingSrcSize;
-    return op-ostart;
+    return (size_t)(op-ostart);
 }
 
 static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
@@ -721,7 +991,7 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
             decodedSize = ZSTD_decompressLegacy(dst, dstCapacity, src, frameSize, dict, dictSize);
             if (ZSTD_isError(decodedSize)) return decodedSize;
 
-            assert(decodedSize <=- dstCapacity);
+            assert(decodedSize <= dstCapacity);
             dst = (BYTE*)dst + decodedSize;
             dstCapacity -= decodedSize;
 
@@ -753,7 +1023,7 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
              * use this in all cases but ddict */
             FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDict(dctx, dict, dictSize), "");
         }
-        ZSTD_checkContinuity(dctx, dst);
+        ZSTD_checkContinuity(dctx, dst, dstCapacity);
 
         {   const size_t res = ZSTD_decompressFrame(dctx, dst, dstCapacity,
                                                     &src, &srcSize);
@@ -761,15 +1031,13 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
                 (ZSTD_getErrorCode(res) == ZSTD_error_prefix_unknown)
              && (moreThan1Frame==1),
                 srcSize_wrong,
-                "at least one frame successfully completed, but following "
-                "bytes are garbage: it's more likely to be a srcSize error, "
-                "specifying more bytes than compressed size of frame(s). This "
-                "error message replaces ERROR(prefix_unknown), which would be "
-                "confusing, as the first header is actually correct. Note that "
-                "one could be unlucky, it might be a corruption error instead, "
-                "happening right at the place where we expect zstd magic "
-                "bytes. But this is _much_ less likely than a srcSize field "
-                "error.");
+                "At least one frame successfully completed, "
+                "but following bytes are garbage: "
+                "it's more likely to be a srcSize error, "
+                "specifying more input bytes than size of frame(s). "
+                "Note: one could be unlucky, it might be a corruption error instead, "
+                "happening right at the place where we expect zstd magic bytes. "
+                "But this is _much_ less likely than a srcSize field error.");
             if (ZSTD_isError(res)) return res;
             assert(res <= dstCapacity);
             if (res != 0)
@@ -781,7 +1049,7 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
 
     RETURN_ERROR_IF(srcSize, srcSize_wrong, "input not entirely consumed");
 
-    return (BYTE*)dst - (BYTE*)dststart;
+    return (size_t)((BYTE*)dst - (BYTE*)dststart);
 }
 
 size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
@@ -798,7 +1066,7 @@ static ZSTD_DDict const* ZSTD_getDDict(ZSTD_DCtx* dctx)
     switch (dctx->dictUses) {
     default:
         assert(0 /* Impossible */);
-        /* fall-through */
+        ZSTD_FALLTHROUGH;
     case ZSTD_dont_use:
         ZSTD_clearDict(dctx);
         return NULL;
@@ -820,7 +1088,7 @@ size_t ZSTD_decompress(void* dst, size_t dstCapacity, const void* src, size_t sr
 {
 #if defined(ZSTD_HEAPMODE) && (ZSTD_HEAPMODE>=1)
     size_t regenSize;
-    ZSTD_DCtx* const dctx = ZSTD_createDCtx();
+    ZSTD_DCtx* const dctx =  ZSTD_createDCtx_internal(ZSTD_defaultCMem);
     RETURN_ERROR_IF(dctx==NULL, memory_allocation, "NULL pointer!");
     regenSize = ZSTD_decompressDCtx(dctx, dst, dstCapacity, src, srcSize);
     ZSTD_freeDCtx(dctx);
@@ -854,7 +1122,7 @@ static size_t ZSTD_nextSrcSizeToDecompressWithInputSize(ZSTD_DCtx* dctx, size_t
         return dctx->expected;
     if (dctx->bType != bt_raw)
         return dctx->expected;
-    return MIN(MAX(inputSize, 1), dctx->expected);
+    return BOUNDED(1, inputSize, dctx->expected);
 }
 
 ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
@@ -862,7 +1130,9 @@ ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
     {
     default:   /* should not happen */
         assert(0);
+        ZSTD_FALLTHROUGH;
     case ZSTDds_getFrameHeaderSize:
+        ZSTD_FALLTHROUGH;
     case ZSTDds_decodeFrameHeader:
         return ZSTDnit_frameHeader;
     case ZSTDds_decodeBlockHeader:
@@ -874,6 +1144,7 @@ ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
     case ZSTDds_checkChecksum:
         return ZSTDnit_checksum;
     case ZSTDds_decodeSkippableHeader:
+        ZSTD_FALLTHROUGH;
     case ZSTDds_skipFrame:
         return ZSTDnit_skippableFrame;
     }
@@ -890,7 +1161,9 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
     DEBUGLOG(5, "ZSTD_decompressContinue (srcSize:%u)", (unsigned)srcSize);
     /* Sanity check */
     RETURN_ERROR_IF(srcSize != ZSTD_nextSrcSizeToDecompressWithInputSize(dctx, srcSize), srcSize_wrong, "not allowed");
-    if (dstCapacity) ZSTD_checkContinuity(dctx, dst);
+    ZSTD_checkContinuity(dctx, dst, dstCapacity);
+
+    dctx->processedCSize += srcSize;
 
     switch (dctx->stage)
     {
@@ -899,21 +1172,21 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
         if (dctx->format == ZSTD_f_zstd1) {  /* allows header */
             assert(srcSize >= ZSTD_FRAMEIDSIZE);  /* to read skippable magic number */
             if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {        /* skippable frame */
-                memcpy(dctx->headerBuffer, src, srcSize);
+                ZSTD_memcpy(dctx->headerBuffer, src, srcSize);
                 dctx->expected = ZSTD_SKIPPABLEHEADERSIZE - srcSize;  /* remaining to load to get full skippable frame header */
                 dctx->stage = ZSTDds_decodeSkippableHeader;
                 return 0;
         }   }
         dctx->headerSize = ZSTD_frameHeaderSize_internal(src, srcSize, dctx->format);
         if (ZSTD_isError(dctx->headerSize)) return dctx->headerSize;
-        memcpy(dctx->headerBuffer, src, srcSize);
+        ZSTD_memcpy(dctx->headerBuffer, src, srcSize);
         dctx->expected = dctx->headerSize - srcSize;
         dctx->stage = ZSTDds_decodeFrameHeader;
         return 0;
 
     case ZSTDds_decodeFrameHeader:
         assert(src != NULL);
-        memcpy(dctx->headerBuffer + (dctx->headerSize - srcSize), src, srcSize);
+        ZSTD_memcpy(dctx->headerBuffer + (dctx->headerSize - srcSize), src, srcSize);
         FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize), "");
         dctx->expected = ZSTD_blockHeaderSize;
         dctx->stage = ZSTDds_decodeBlockHeader;
@@ -955,7 +1228,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
             {
             case bt_compressed:
                 DEBUGLOG(5, "ZSTD_decompressContinue: case bt_compressed");
-                rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 1);
+                rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 1, is_streaming);
                 dctx->expected = 0;  /* Streaming not supported */
                 break;
             case bt_raw :
@@ -977,7 +1250,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
             RETURN_ERROR_IF(rSize > dctx->fParams.blockSizeMax, corruption_detected, "Decompressed Block Size Exceeds Maximum");
             DEBUGLOG(5, "ZSTD_decompressContinue: decoded size from block : %u", (unsigned)rSize);
             dctx->decodedSize += rSize;
-            if (dctx->fParams.checksumFlag) XXH64_update(&dctx->xxhState, dst, rSize);
+            if (dctx->validateChecksum) XXH64_update(&dctx->xxhState, dst, rSize);
             dctx->previousDstEnd = (char*)dst + rSize;
 
             /* Stay on the same stage until we are finished streaming the block. */
@@ -995,6 +1268,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
                     dctx->expected = 4;
                     dctx->stage = ZSTDds_checkChecksum;
                 } else {
+                    ZSTD_DCtx_trace_end(dctx, dctx->decodedSize, dctx->processedCSize, /* streaming */ 1);
                     dctx->expected = 0;   /* ends here */
                     dctx->stage = ZSTDds_getFrameHeaderSize;
                 }
@@ -1007,10 +1281,14 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
 
     case ZSTDds_checkChecksum:
         assert(srcSize == 4);  /* guaranteed by dctx->expected */
-        {   U32 const h32 = (U32)XXH64_digest(&dctx->xxhState);
-            U32 const check32 = MEM_readLE32(src);
-            DEBUGLOG(4, "ZSTD_decompressContinue: checksum : calculated %08X :: %08X read", (unsigned)h32, (unsigned)check32);
-            RETURN_ERROR_IF(check32 != h32, checksum_wrong, "");
+        {
+            if (dctx->validateChecksum) {
+                U32 const h32 = (U32)XXH64_digest(&dctx->xxhState);
+                U32 const check32 = MEM_readLE32(src);
+                DEBUGLOG(4, "ZSTD_decompressContinue: checksum : calculated %08X :: %08X read", (unsigned)h32, (unsigned)check32);
+                RETURN_ERROR_IF(check32 != h32, checksum_wrong, "");
+            }
+            ZSTD_DCtx_trace_end(dctx, dctx->decodedSize, dctx->processedCSize, /* streaming */ 1);
             dctx->expected = 0;
             dctx->stage = ZSTDds_getFrameHeaderSize;
             return 0;
@@ -1019,7 +1297,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
     case ZSTDds_decodeSkippableHeader:
         assert(src != NULL);
         assert(srcSize <= ZSTD_SKIPPABLEHEADERSIZE);
-        memcpy(dctx->headerBuffer + (ZSTD_SKIPPABLEHEADERSIZE - srcSize), src, srcSize);   /* complete skippable header */
+        ZSTD_memcpy(dctx->headerBuffer + (ZSTD_SKIPPABLEHEADERSIZE - srcSize), src, srcSize);   /* complete skippable header */
         dctx->expected = MEM_readLE32(dctx->headerBuffer + ZSTD_FRAMEIDSIZE);   /* note : dctx->expected can grow seriously large, beyond local buffer size */
         dctx->stage = ZSTDds_skipFrame;
         return 0;
@@ -1075,7 +1353,7 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
                                                 workspace, workspaceSize);
 #else
         size_t const hSize = HUF_readDTableX2_wksp(entropy->hufTable,
-                                                dictPtr, dictEnd - dictPtr,
+                                                dictPtr, (size_t)(dictEnd - dictPtr),
                                                 workspace, workspaceSize);
 #endif
         RETURN_ERROR_IF(HUF_isError(hSize), dictionary_corrupted, "");
@@ -1084,40 +1362,46 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
 
     {   short offcodeNCount[MaxOff+1];
         unsigned offcodeMaxValue = MaxOff, offcodeLog;
-        size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
+        size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, (size_t)(dictEnd-dictPtr));
         RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "");
         RETURN_ERROR_IF(offcodeMaxValue > MaxOff, dictionary_corrupted, "");
         RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "");
         ZSTD_buildFSETable( entropy->OFTable,
                             offcodeNCount, offcodeMaxValue,
                             OF_base, OF_bits,
-                            offcodeLog);
+                            offcodeLog,
+                            entropy->workspace, sizeof(entropy->workspace),
+                            /* bmi2 */0);
         dictPtr += offcodeHeaderSize;
     }
 
     {   short matchlengthNCount[MaxML+1];
         unsigned matchlengthMaxValue = MaxML, matchlengthLog;
-        size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);
+        size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, (size_t)(dictEnd-dictPtr));
         RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "");
         RETURN_ERROR_IF(matchlengthMaxValue > MaxML, dictionary_corrupted, "");
         RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "");
         ZSTD_buildFSETable( entropy->MLTable,
                             matchlengthNCount, matchlengthMaxValue,
                             ML_base, ML_bits,
-                            matchlengthLog);
+                            matchlengthLog,
+                            entropy->workspace, sizeof(entropy->workspace),
+                            /* bmi2 */ 0);
         dictPtr += matchlengthHeaderSize;
     }
 
     {   short litlengthNCount[MaxLL+1];
         unsigned litlengthMaxValue = MaxLL, litlengthLog;
-        size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);
+        size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, (size_t)(dictEnd-dictPtr));
         RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "");
         RETURN_ERROR_IF(litlengthMaxValue > MaxLL, dictionary_corrupted, "");
         RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "");
         ZSTD_buildFSETable( entropy->LLTable,
                             litlengthNCount, litlengthMaxValue,
                             LL_base, LL_bits,
-                            litlengthLog);
+                            litlengthLog,
+                            entropy->workspace, sizeof(entropy->workspace),
+                            /* bmi2 */ 0);
         dictPtr += litlengthHeaderSize;
     }
 
@@ -1131,7 +1415,7 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
             entropy->rep[i] = rep;
     }   }
 
-    return dictPtr - (const BYTE*)dict;
+    return (size_t)(dictPtr - (const BYTE*)dict);
 }
 
 static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
@@ -1158,8 +1442,12 @@ static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict
 size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
 {
     assert(dctx != NULL);
+#if ZSTD_TRACE
+    dctx->traceCtx = (ZSTD_trace_decompress_begin != NULL) ? ZSTD_trace_decompress_begin(dctx) : 0;
+#endif
     dctx->expected = ZSTD_startingInputLength(dctx->format);  /* dctx->format must be properly set */
     dctx->stage = ZSTDds_getFrameHeaderSize;
+    dctx->processedCSize = 0;
     dctx->decodedSize = 0;
     dctx->previousDstEnd = NULL;
     dctx->prefixStart = NULL;
@@ -1170,7 +1458,7 @@ size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
     dctx->dictID = 0;
     dctx->bType = bt_reserved;
     ZSTD_STATIC_ASSERT(sizeof(dctx->entropy.rep) == sizeof(repStartValue));
-    memcpy(dctx->entropy.rep, repStartValue, sizeof(repStartValue));  /* initial repcodes */
+    ZSTD_memcpy(dctx->entropy.rep, repStartValue, sizeof(repStartValue));  /* initial repcodes */
     dctx->LLTptr = dctx->entropy.LLTable;
     dctx->MLTptr = dctx->entropy.MLTable;
     dctx->OFTptr = dctx->entropy.OFTable;
@@ -1265,7 +1553,7 @@ size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
 ZSTD_DStream* ZSTD_createDStream(void)
 {
     DEBUGLOG(3, "ZSTD_createDStream");
-    return ZSTD_createDStream_advanced(ZSTD_defaultCMem);
+    return ZSTD_createDCtx_internal(ZSTD_defaultCMem);
 }
 
 ZSTD_DStream* ZSTD_initStaticDStream(void *workspace, size_t workspaceSize)
@@ -1275,7 +1563,7 @@ ZSTD_DStream* ZSTD_initStaticDStream(void *workspace, size_t workspaceSize)
 
 ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem)
 {
-    return ZSTD_createDCtx_advanced(customMem);
+    return ZSTD_createDCtx_internal(customMem);
 }
 
 size_t ZSTD_freeDStream(ZSTD_DStream* zds)
@@ -1373,6 +1661,16 @@ size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
     if (ddict) {
         dctx->ddict = ddict;
         dctx->dictUses = ZSTD_use_indefinitely;
+        if (dctx->refMultipleDDicts == ZSTD_rmd_refMultipleDDicts) {
+            if (dctx->ddictSet == NULL) {
+                dctx->ddictSet = ZSTD_createDDictHashSet(dctx->customMem);
+                if (!dctx->ddictSet) {
+                    RETURN_ERROR(memory_allocation, "Failed to allocate memory for hash set!");
+                }
+            }
+            assert(!dctx->staticSize);  /* Impossible: ddictSet cannot have been allocated if static dctx */
+            FORWARD_IF_ERROR(ZSTD_DDictHashSet_addDDict(dctx->ddictSet, ddict, dctx->customMem), "");
+        }
     }
     return 0;
 }
@@ -1394,7 +1692,7 @@ size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize)
 
 size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format)
 {
-    return ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, format);
+    return ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, (int)format);
 }
 
 ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam)
@@ -1411,8 +1709,16 @@ ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam)
             ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);
             return bounds;
         case ZSTD_d_stableOutBuffer:
-            bounds.lowerBound = (int)ZSTD_obm_buffered;
-            bounds.upperBound = (int)ZSTD_obm_stable;
+            bounds.lowerBound = (int)ZSTD_bm_buffered;
+            bounds.upperBound = (int)ZSTD_bm_stable;
+            return bounds;
+        case ZSTD_d_forceIgnoreChecksum:
+            bounds.lowerBound = (int)ZSTD_d_validateChecksum;
+            bounds.upperBound = (int)ZSTD_d_ignoreChecksum;
+            return bounds;
+        case ZSTD_d_refMultipleDDicts:
+            bounds.lowerBound = (int)ZSTD_rmd_refSingleDDict;
+            bounds.upperBound = (int)ZSTD_rmd_refMultipleDDicts;
             return bounds;
         default:;
     }
@@ -1436,6 +1742,29 @@ static int ZSTD_dParam_withinBounds(ZSTD_dParameter dParam, int value)
     RETURN_ERROR_IF(!ZSTD_dParam_withinBounds(p, v), parameter_outOfBound, ""); \
 }
 
+size_t ZSTD_DCtx_getParameter(ZSTD_DCtx* dctx, ZSTD_dParameter param, int* value)
+{
+    switch (param) {
+        case ZSTD_d_windowLogMax:
+            *value = (int)ZSTD_highbit32((U32)dctx->maxWindowSize);
+            return 0;
+        case ZSTD_d_format:
+            *value = (int)dctx->format;
+            return 0;
+        case ZSTD_d_stableOutBuffer:
+            *value = (int)dctx->outBufferMode;
+            return 0;
+        case ZSTD_d_forceIgnoreChecksum:
+            *value = (int)dctx->forceIgnoreChecksum;
+            return 0;
+        case ZSTD_d_refMultipleDDicts:
+            *value = (int)dctx->refMultipleDDicts;
+            return 0;
+        default:;
+    }
+    RETURN_ERROR(parameter_unsupported, "");
+}
+
 size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter dParam, int value)
 {
     RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
@@ -1451,7 +1780,18 @@ size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter dParam, int value
             return 0;
         case ZSTD_d_stableOutBuffer:
             CHECK_DBOUNDS(ZSTD_d_stableOutBuffer, value);
-            dctx->outBufferMode = (ZSTD_outBufferMode_e)value;
+            dctx->outBufferMode = (ZSTD_bufferMode_e)value;
+            return 0;
+        case ZSTD_d_forceIgnoreChecksum:
+            CHECK_DBOUNDS(ZSTD_d_forceIgnoreChecksum, value);
+            dctx->forceIgnoreChecksum = (ZSTD_forceIgnoreChecksum_e)value;
+            return 0;
+        case ZSTD_d_refMultipleDDicts:
+            CHECK_DBOUNDS(ZSTD_d_refMultipleDDicts, value);
+            if (dctx->staticSize != 0) {
+                RETURN_ERROR(parameter_unsupported, "Static dctx does not support multiple DDicts!");
+            }
+            dctx->refMultipleDDicts = (ZSTD_refMultipleDDicts_e)value;
             return 0;
         default:;
     }
@@ -1469,8 +1809,7 @@ size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset)
       || (reset == ZSTD_reset_session_and_parameters) ) {
         RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
         ZSTD_clearDict(dctx);
-        dctx->format = ZSTD_f_zstd1;
-        dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
+        ZSTD_DCtx_resetParameters(dctx);
     }
     return 0;
 }
@@ -1484,7 +1823,8 @@ size_t ZSTD_sizeof_DStream(const ZSTD_DStream* dctx)
 size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize)
 {
     size_t const blockSize = (size_t) MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
-    unsigned long long const neededRBSize = windowSize + blockSize + (WILDCOPY_OVERLENGTH * 2);
+    /* space is needed to store the litbuffer after the output of a given block without stomping the extDict of a previous run, as well as to cover both windows against wildcopy*/
+    unsigned long long const neededRBSize = windowSize + blockSize + ZSTD_BLOCKSIZE_MAX + (WILDCOPY_OVERLENGTH * 2);
     unsigned long long const neededSize = MIN(frameContentSize, neededRBSize);
     size_t const minRBSize = (size_t) neededSize;
     RETURN_ERROR_IF((unsigned long long)minRBSize != neededSize,
@@ -1524,7 +1864,7 @@ static void ZSTD_DCtx_updateOversizedDuration(ZSTD_DStream* zds, size_t const ne
 {
     if (ZSTD_DCtx_isOverflow(zds, neededInBuffSize, neededOutBuffSize))
         zds->oversizedDuration++;
-    else 
+    else
         zds->oversizedDuration = 0;
 }
 
@@ -1538,7 +1878,7 @@ static size_t ZSTD_checkOutBuffer(ZSTD_DStream const* zds, ZSTD_outBuffer const*
 {
     ZSTD_outBuffer const expect = zds->expectedOutBuffer;
     /* No requirement when ZSTD_obm_stable is not enabled. */
-    if (zds->outBufferMode != ZSTD_obm_stable)
+    if (zds->outBufferMode != ZSTD_bm_stable)
         return 0;
     /* Any buffer is allowed in zdss_init, this must be the same for every other call until
      * the context is reset.
@@ -1548,7 +1888,7 @@ static size_t ZSTD_checkOutBuffer(ZSTD_DStream const* zds, ZSTD_outBuffer const*
     /* The buffer must match our expectation exactly. */
     if (expect.dst == output->dst && expect.pos == output->pos && expect.size == output->size)
         return 0;
-    RETURN_ERROR(dstBuffer_wrong, "ZSTD_obm_stable enabled but output differs!");
+    RETURN_ERROR(dstBuffer_wrong, "ZSTD_d_stableOutBuffer enabled but output differs!");
 }
 
 /* Calls ZSTD_decompressContinue() with the right parameters for ZSTD_decompressStream()
@@ -1560,7 +1900,7 @@ static size_t ZSTD_decompressContinueStream(
             ZSTD_DStream* zds, char** op, char* oend,
             void const* src, size_t srcSize) {
     int const isSkipFrame = ZSTD_isSkipFrame(zds);
-    if (zds->outBufferMode == ZSTD_obm_buffered) {
+    if (zds->outBufferMode == ZSTD_bm_buffered) {
         size_t const dstSize = isSkipFrame ? 0 : zds->outBuffSize - zds->outStart;
         size_t const decodedSize = ZSTD_decompressContinue(zds,
                 zds->outBuff + zds->outStart, dstSize, src, srcSize);
@@ -1573,14 +1913,14 @@ static size_t ZSTD_decompressContinueStream(
         }
     } else {
         /* Write directly into the output buffer */
-        size_t const dstSize = isSkipFrame ? 0 : oend - *op;
+        size_t const dstSize = isSkipFrame ? 0 : (size_t)(oend - *op);
         size_t const decodedSize = ZSTD_decompressContinue(zds, *op, dstSize, src, srcSize);
         FORWARD_IF_ERROR(decodedSize, "");
         *op += decodedSize;
         /* Flushing is not needed. */
         zds->streamStage = zdss_read;
         assert(*op <= oend);
-        assert(zds->outBufferMode == ZSTD_obm_stable);
+        assert(zds->outBufferMode == ZSTD_bm_stable);
     }
     return 0;
 }
@@ -1618,10 +1958,12 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             DEBUGLOG(5, "stage zdss_init => transparent reset ");
             zds->streamStage = zdss_loadHeader;
             zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
             zds->legacyVersion = 0;
+#endif
             zds->hostageByte = 0;
             zds->expectedOutBuffer = *output;
-            /* fall-through */
+            ZSTD_FALLTHROUGH;
 
         case zdss_loadHeader :
             DEBUGLOG(5, "stage zdss_loadHeader (srcSize : %u)", (U32)(iend - ip));
@@ -1635,6 +1977,9 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             }   }
 #endif
             {   size_t const hSize = ZSTD_getFrameHeader_advanced(&zds->fParams, zds->headerBuffer, zds->lhSize, zds->format);
+                if (zds->refMultipleDDicts && zds->ddictSet) {
+                    ZSTD_DCtx_selectFrameDDict(zds);
+                }
                 DEBUGLOG(5, "header size : %u", (U32)hSize);
                 if (ZSTD_isError(hSize)) {
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
@@ -1663,14 +2008,14 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                     assert(iend >= ip);
                     if (toLoad > remainingInput) {   /* not enough input to load full header */
                         if (remainingInput > 0) {
-                            memcpy(zds->headerBuffer + zds->lhSize, ip, remainingInput);
+                            ZSTD_memcpy(zds->headerBuffer + zds->lhSize, ip, remainingInput);
                             zds->lhSize += remainingInput;
                         }
                         input->pos = input->size;
                         return (MAX((size_t)ZSTD_FRAMEHEADERSIZE_MIN(zds->format), hSize) - zds->lhSize) + ZSTD_blockHeaderSize;   /* remaining header bytes + next block header */
                     }
                     assert(ip != NULL);
-                    memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
+                    ZSTD_memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
                     break;
             }   }
 
@@ -1678,10 +2023,10 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             if (zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
                 && zds->fParams.frameType != ZSTD_skippableFrame
                 && (U64)(size_t)(oend-op) >= zds->fParams.frameContentSize) {
-                size_t const cSize = ZSTD_findFrameCompressedSize(istart, iend-istart);
+                size_t const cSize = ZSTD_findFrameCompressedSize(istart, (size_t)(iend-istart));
                 if (cSize <= (size_t)(iend-istart)) {
                     /* shortcut : using single-pass mode */
-                    size_t const decompressedSize = ZSTD_decompress_usingDDict(zds, op, oend-op, istart, cSize, ZSTD_getDDict(zds));
+                    size_t const decompressedSize = ZSTD_decompress_usingDDict(zds, op, (size_t)(oend-op), istart, cSize, ZSTD_getDDict(zds));
                     if (ZSTD_isError(decompressedSize)) return decompressedSize;
                     DEBUGLOG(4, "shortcut to single-pass ZSTD_decompress_usingDDict()")
                     ip = istart + cSize;
@@ -1693,7 +2038,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             }   }
 
             /* Check output buffer is large enough for ZSTD_odm_stable. */
-            if (zds->outBufferMode == ZSTD_obm_stable
+            if (zds->outBufferMode == ZSTD_bm_stable
                 && zds->fParams.frameType != ZSTD_skippableFrame
                 && zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
                 && (U64)(size_t)(oend-op) < zds->fParams.frameContentSize) {
@@ -1723,7 +2068,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
 
             /* Adapt buffer sizes to frame header instructions */
             {   size_t const neededInBuffSize = MAX(zds->fParams.blockSizeMax, 4 /* frame checksum */);
-                size_t const neededOutBuffSize = zds->outBufferMode == ZSTD_obm_buffered
+                size_t const neededOutBuffSize = zds->outBufferMode == ZSTD_bm_buffered
                         ? ZSTD_decodingBufferSize_min(zds->fParams.windowSize, zds->fParams.frameContentSize)
                         : 0;
 
@@ -1731,7 +2076,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
 
                 {   int const tooSmall = (zds->inBuffSize < neededInBuffSize) || (zds->outBuffSize < neededOutBuffSize);
                     int const tooLarge = ZSTD_DCtx_isOversizedTooLong(zds);
-                    
+
                     if (tooSmall || tooLarge) {
                         size_t const bufferSize = neededInBuffSize + neededOutBuffSize;
                         DEBUGLOG(4, "inBuff  : from %u to %u",
@@ -1745,10 +2090,10 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                                 bufferSize > zds->staticSize - sizeof(ZSTD_DCtx),
                                 memory_allocation, "");
                         } else {
-                            ZSTD_free(zds->inBuff, zds->customMem);
+                            ZSTD_customFree(zds->inBuff, zds->customMem);
                             zds->inBuffSize = 0;
                             zds->outBuffSize = 0;
-                            zds->inBuff = (char*)ZSTD_malloc(bufferSize, zds->customMem);
+                            zds->inBuff = (char*)ZSTD_customMalloc(bufferSize, zds->customMem);
                             RETURN_ERROR_IF(zds->inBuff == NULL, memory_allocation, "");
                         }
                         zds->inBuffSize = neededInBuffSize;
@@ -1756,11 +2101,11 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                         zds->outBuffSize = neededOutBuffSize;
             }   }   }
             zds->streamStage = zdss_read;
-            /* fall-through */
+            ZSTD_FALLTHROUGH;
 
         case zdss_read:
             DEBUGLOG(5, "stage zdss_read");
-            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompressWithInputSize(zds, iend - ip);
+            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompressWithInputSize(zds, (size_t)(iend - ip));
                 DEBUGLOG(5, "neededInSize = %u", (U32)neededInSize);
                 if (neededInSize==0) {  /* end of frame */
                     zds->streamStage = zdss_init;
@@ -1775,7 +2120,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             }   }
             if (ip==iend) { someMoreWork = 0; break; }   /* no more input */
             zds->streamStage = zdss_load;
-            /* fall-through */
+            ZSTD_FALLTHROUGH;
 
         case zdss_load:
             {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds);
@@ -1790,7 +2135,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                     RETURN_ERROR_IF(toLoad > zds->inBuffSize - zds->inPos,
                                     corruption_detected,
                                     "should never happen");
-                    loadedSize = ZSTD_limitCopy(zds->inBuff + zds->inPos, toLoad, ip, iend-ip);
+                    loadedSize = ZSTD_limitCopy(zds->inBuff + zds->inPos, toLoad, ip, (size_t)(iend-ip));
                 }
                 ip += loadedSize;
                 zds->inPos += loadedSize;
@@ -1804,7 +2149,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             }
         case zdss_flush:
             {   size_t const toFlushSize = zds->outEnd - zds->outStart;
-                size_t const flushedSize = ZSTD_limitCopy(op, oend-op, zds->outBuff + zds->outStart, toFlushSize);
+                size_t const flushedSize = ZSTD_limitCopy(op, (size_t)(oend-op), zds->outBuff + zds->outStart, toFlushSize);
                 op += flushedSize;
                 zds->outStart += flushedSize;
                 if (flushedSize == toFlushSize) {  /* flush completed */