Update zstd to 1.4.5.

54 files changed, 1613 insertions, 1230 deletions

diff --git a/zstd/bitstream.h b/zstd/bitstream.h
index fe3148ee..37b99c01 100644
--- a/zstd/bitstream.h
+++ b/zstd/bitstream.h
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   bitstream
-   Part of FSE library
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-   You can contact the author at :
-   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * bitstream
+ * Part of FSE library
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *
+ * 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.
 ****************************************************************** */
 #ifndef BITSTREAM_H_MODULE
 #define BITSTREAM_H_MODULE
@@ -162,8 +142,7 @@ MEM_STATIC unsigned BIT_highbit32 (U32 val)
     {
 #   if defined(_MSC_VER)   /* Visual */
         unsigned long r=0;
-        _BitScanReverse ( &r, val );
-        return (unsigned) r;
+        return _BitScanReverse ( &r, val ) ? (unsigned)r : 0;
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* Use GCC Intrinsic */
         return __builtin_clz (val) ^ 31;
 #   elif defined(__ICCARM__)    /* IAR Intrinsic */
diff --git a/zstd/compiler.h b/zstd/compiler.h
index 228845aa..95e94835 100644
--- a/zstd/compiler.h
+++ b/zstd/compiler.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -17,7 +17,7 @@
 /* force inlining */
 
 #if !defined(ZSTD_NO_INLINE)
-#if defined (__GNUC__) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#if (defined(__GNUC__) && !defined(__STRICT_ANSI__)) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
 #  define INLINE_KEYWORD inline
 #else
 #  define INLINE_KEYWORD
@@ -114,6 +114,9 @@
 #    include <mmintrin.h>   /* https://msdn.microsoft.com/fr-fr/library/84szxsww(v=vs.90).aspx */
 #    define PREFETCH_L1(ptr)  _mm_prefetch((const char*)(ptr), _MM_HINT_T0)
 #    define PREFETCH_L2(ptr)  _mm_prefetch((const char*)(ptr), _MM_HINT_T1)
+#    elif defined(__aarch64__)
+#     define PREFETCH_L1(ptr)  __asm__ __volatile__("prfm pldl1keep, %0" ::"Q"(*(ptr)))
+#     define PREFETCH_L2(ptr)  __asm__ __volatile__("prfm pldl2keep, %0" ::"Q"(*(ptr)))
 #  elif defined(__GNUC__) && ( (__GNUC__ >= 4) || ( (__GNUC__ == 3) && (__GNUC_MINOR__ >= 1) ) )
 #    define PREFETCH_L1(ptr)  __builtin_prefetch((ptr), 0 /* rw==read */, 3 /* locality */)
 #    define PREFETCH_L2(ptr)  __builtin_prefetch((ptr), 0 /* rw==read */, 2 /* locality */)
@@ -136,7 +139,7 @@
 
 /* vectorization
  * older GCC (pre gcc-4.3 picked as the cutoff) uses a different syntax */
-#if !defined(__clang__) && defined(__GNUC__)
+#if !defined(__INTEL_COMPILER) && !defined(__clang__) && defined(__GNUC__)
 #  if (__GNUC__ == 4 && __GNUC_MINOR__ > 3) || (__GNUC__ >= 5)
 #    define DONT_VECTORIZE __attribute__((optimize("no-tree-vectorize")))
 #  else
diff --git a/zstd/cpu.h b/zstd/cpu.h
index 5f0923fc..6e8a974f 100644
--- a/zstd/cpu.h
+++ b/zstd/cpu.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2018-present, Facebook, Inc.
+ * Copyright (c) 2018-2020, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/debug.c b/zstd/debug.c
index 3ebdd1cb..f303f4a2 100644
--- a/zstd/debug.c
+++ b/zstd/debug.c
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   debug
-   Part of FSE library
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-   You can contact the author at :
-   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * debug
+ * Part of FSE library
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *
+ * 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.
 ****************************************************************** */
 
 
diff --git a/zstd/debug.h b/zstd/debug.h
index b4fc89d4..ac622488 100644
--- a/zstd/debug.h
+++ b/zstd/debug.h
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   debug
-   Part of FSE library
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-   You can contact the author at :
-   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * debug
+ * Part of FSE library
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *
+ * 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.
 ****************************************************************** */
 
 
diff --git a/zstd/entropy_common.c b/zstd/entropy_common.c
index b12944e1..9d3e4e8e 100644
--- a/zstd/entropy_common.c
+++ b/zstd/entropy_common.c
@@ -1,36 +1,16 @@
-/*
-   Common functions of New Generation Entropy library
-   Copyright (C) 2016, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
-    - Public forum : https://groups.google.com/forum/#!forum/lz4c
-*************************************************************************** */
+/* ******************************************************************
+ * Common functions of New Generation Entropy library
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *  - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * 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.
+****************************************************************** */
 
 /* *************************************
 *  Dependencies
diff --git a/zstd/error_private.c b/zstd/error_private.c
index 7c1bb67a..cd437529 100644
--- a/zstd/error_private.c
+++ b/zstd/error_private.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -47,6 +47,7 @@ const char* ERR_getErrorString(ERR_enum code)
         /* following error codes are not stable and may be removed or changed in a future version */
     case PREFIX(frameIndex_tooLarge): return "Frame index is too large";
     case PREFIX(seekableIO): return "An I/O error occurred when reading/seeking";
+    case PREFIX(dstBuffer_wrong): return "Destination buffer is wrong";
     case PREFIX(maxCode):
     default: return notErrorCode;
     }
diff --git a/zstd/error_private.h b/zstd/error_private.h
index 0d2fa7e3..982cf8e9 100644
--- a/zstd/error_private.h
+++ b/zstd/error_private.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -49,7 +49,7 @@ typedef ZSTD_ErrorCode ERR_enum;
 /*-****************************************
 *  Error codes handling
 ******************************************/
-#undef ERROR   /* reported already defined on VS 2015 (Rich Geldreich) */
+#undef ERROR   /* already defined on Visual Studio */
 #define ERROR(name) ZSTD_ERROR(name)
 #define ZSTD_ERROR(name) ((size_t)-PREFIX(name))
 
@@ -57,6 +57,10 @@ ERR_STATIC unsigned ERR_isError(size_t code) { return (code > ERROR(maxCode)); }
 
 ERR_STATIC ERR_enum ERR_getErrorCode(size_t code) { if (!ERR_isError(code)) return (ERR_enum)0; return (ERR_enum) (0-code); }
 
+/* check and forward error code */
+#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return e
+#define CHECK_F(f)   { CHECK_V_F(_var_err__, f); }
+
 
 /*-****************************************
 *  Error Strings
diff --git a/zstd/fse.h b/zstd/fse.h
index a7553e37..ff54e70e 100644
--- a/zstd/fse.h
+++ b/zstd/fse.h
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   FSE : Finite State Entropy codec
-   Public Prototypes declaration
-   Copyright (C) 2013-2016, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-   You can contact the author at :
-   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * FSE : Finite State Entropy codec
+ * Public Prototypes declaration
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *
+ * 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.
 ****************************************************************** */
 
 #if defined (__cplusplus)
diff --git a/zstd/fse_compress.c b/zstd/fse_compress.c
index 68b47e10..e09b646d 100644
--- a/zstd/fse_compress.c
+++ b/zstd/fse_compress.c
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   FSE : Finite State Entropy encoder
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
-    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ * FSE : Finite State Entropy encoder
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *  - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * 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.
 ****************************************************************** */
 
 /* **************************************************************
@@ -645,9 +625,6 @@ size_t FSE_compress_usingCTable (void* dst, size_t dstSize,
 
 size_t FSE_compressBound(size_t size) { return FSE_COMPRESSBOUND(size); }
 
-#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return e
-#define CHECK_F(f)   { CHECK_V_F(_var_err__, f); }
-
 /* FSE_compress_wksp() :
  * Same as FSE_compress2(), but using an externally allocated scratch buffer (`workSpace`).
  * `wkspSize` size must be `(1<<tableLog)`.
diff --git a/zstd/fse_decompress.c b/zstd/fse_decompress.c
index 4f073789..bcc2223c 100644
--- a/zstd/fse_decompress.c
+++ b/zstd/fse_decompress.c
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   FSE : Finite State Entropy decoder
-   Copyright (C) 2013-2015, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
-    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ * FSE : Finite State Entropy decoder
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *  - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * 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.
 ****************************************************************** */
 
 
@@ -51,11 +31,6 @@
 #define FSE_isError ERR_isError
 #define FSE_STATIC_ASSERT(c) DEBUG_STATIC_ASSERT(c)   /* use only *after* variable declarations */
 
-/* check and forward error code */
-#ifndef CHECK_F
-#define CHECK_F(f) { size_t const e = f; if (FSE_isError(e)) return e; }
-#endif
-
 
 /* **************************************************************
 *  Templates
@@ -287,7 +262,7 @@ size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size
     /* normal FSE decoding mode */
     size_t const NCountLength = FSE_readNCount (counting, &maxSymbolValue, &tableLog, istart, cSrcSize);
     if (FSE_isError(NCountLength)) return NCountLength;
-    //if (NCountLength >= cSrcSize) return ERROR(srcSize_wrong);   /* too small input size; supposed to be already checked in NCountLength, only remaining case : NCountLength==cSrcSize */
+    /* if (NCountLength >= cSrcSize) return ERROR(srcSize_wrong); */  /* too small input size; supposed to be already checked in NCountLength, only remaining case : NCountLength==cSrcSize */
     if (tableLog > maxLog) return ERROR(tableLog_tooLarge);
     ip += NCountLength;
     cSrcSize -= NCountLength;
diff --git a/zstd/hist.c b/zstd/hist.c
index 45b7babc..c17b9725 100644
--- a/zstd/hist.c
+++ b/zstd/hist.c
@@ -1,36 +1,16 @@
 /* ******************************************************************
-   hist : Histogram functions
-   part of Finite State Entropy project
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
-    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ * hist : Histogram functions
+ * part of Finite State Entropy project
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *  - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * 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.
 ****************************************************************** */
 
 /* --- dependencies --- */
diff --git a/zstd/hist.h b/zstd/hist.h
index 8b389358..77e3ec4f 100644
--- a/zstd/hist.h
+++ b/zstd/hist.h
@@ -1,36 +1,16 @@
 /* ******************************************************************
-   hist : Histogram functions
-   part of Finite State Entropy project
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
-    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ * hist : Histogram functions
+ * part of Finite State Entropy project
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *  - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * 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.
 ****************************************************************** */
 
 /* --- dependencies --- */
diff --git a/zstd/huf.h b/zstd/huf.h
index 4a87db5c..ef432685 100644
--- a/zstd/huf.h
+++ b/zstd/huf.h
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   huff0 huffman codec,
-   part of Finite State Entropy library
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-   You can contact the author at :
-   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * huff0 huffman codec,
+ * part of Finite State Entropy library
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *
+ * 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.
 ****************************************************************** */
 
 #if defined (__cplusplus)
@@ -110,7 +90,7 @@ HUF_PUBLIC_API size_t HUF_compress2 (void* dst, size_t dstCapacity,
 /** HUF_compress4X_wksp() :
  *  Same as HUF_compress2(), but uses externally allocated `workSpace`.
  * `workspace` must have minimum alignment of 4, and be at least as large as HUF_WORKSPACE_SIZE */
-#define HUF_WORKSPACE_SIZE (6 << 10)
+#define HUF_WORKSPACE_SIZE ((6 << 10) + 256)
 #define HUF_WORKSPACE_SIZE_U32 (HUF_WORKSPACE_SIZE / sizeof(U32))
 HUF_PUBLIC_API size_t HUF_compress4X_wksp (void* dst, size_t dstCapacity,
                                      const void* src, size_t srcSize,
@@ -209,6 +189,7 @@ size_t HUF_buildCTable (HUF_CElt* CTable, const unsigned* count, unsigned maxSym
 size_t HUF_writeCTable (void* dst, size_t maxDstSize, const HUF_CElt* CTable, unsigned maxSymbolValue, unsigned huffLog);
 size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable);
 size_t HUF_estimateCompressedSize(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue);
+int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue);
 
 typedef enum {
    HUF_repeat_none,  /**< Cannot use the previous table */
diff --git a/zstd/huf_compress.c b/zstd/huf_compress.c
index b8e6fb38..4efffbb5 100644
--- a/zstd/huf_compress.c
+++ b/zstd/huf_compress.c
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   Huffman encoder, part of New Generation Entropy library
-   Copyright (C) 2013-2016, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
-    - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ * Huffman encoder, part of New Generation Entropy library
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *  - Public forum : https://groups.google.com/forum/#!forum/lz4c
+ *
+ * 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.
 ****************************************************************** */
 
 /* **************************************************************
@@ -60,8 +40,6 @@
 ****************************************************************/
 #define HUF_isError ERR_isError
 #define HUF_STATIC_ASSERT(c) DEBUG_STATIC_ASSERT(c)   /* use only *after* variable declarations */
-#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return e
-#define CHECK_F(f)   { CHECK_V_F(_var_err__, f); }
 
 
 /* **************************************************************
@@ -110,18 +88,18 @@ static size_t HUF_compressWeights (void* dst, size_t dstSize, const void* weight
     CHECK_F( FSE_normalizeCount(norm, tableLog, count, wtSize, maxSymbolValue) );
 
     /* Write table description header */
-    {   CHECK_V_F(hSize, FSE_writeNCount(op, oend-op, norm, maxSymbolValue, tableLog) );
+    {   CHECK_V_F(hSize, FSE_writeNCount(op, (size_t)(oend-op), norm, maxSymbolValue, tableLog) );
         op += hSize;
     }
 
     /* Compress */
     CHECK_F( FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, scratchBuffer, sizeof(scratchBuffer)) );
-    {   CHECK_V_F(cSize, FSE_compress_usingCTable(op, oend - op, weightTable, wtSize, CTable) );
+    {   CHECK_V_F(cSize, FSE_compress_usingCTable(op, (size_t)(oend - op), weightTable, wtSize, CTable) );
         if (cSize == 0) return 0;   /* not enough space for compressed data */
         op += cSize;
     }
 
-    return op-ostart;
+    return (size_t)(op-ostart);
 }
 
 
@@ -242,7 +220,7 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
     /* there are several too large elements (at least >= 2) */
     {   int totalCost = 0;
         const U32 baseCost = 1 << (largestBits - maxNbBits);
-        U32 n = lastNonNull;
+        int n = (int)lastNonNull;
 
         while (huffNode[n].nbBits > maxNbBits) {
             totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits));
@@ -257,22 +235,22 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
         /* repay normalized cost */
         {   U32 const noSymbol = 0xF0F0F0F0;
             U32 rankLast[HUF_TABLELOG_MAX+2];
-            int pos;
 
             /* Get pos of last (smallest) symbol per rank */
             memset(rankLast, 0xF0, sizeof(rankLast));
             {   U32 currentNbBits = maxNbBits;
+                int pos;
                 for (pos=n ; pos >= 0; pos--) {
                     if (huffNode[pos].nbBits >= currentNbBits) continue;
                     currentNbBits = huffNode[pos].nbBits;   /* < maxNbBits */
-                    rankLast[maxNbBits-currentNbBits] = pos;
+                    rankLast[maxNbBits-currentNbBits] = (U32)pos;
             }   }
 
             while (totalCost > 0) {
-                U32 nBitsToDecrease = BIT_highbit32(totalCost) + 1;
+                U32 nBitsToDecrease = BIT_highbit32((U32)totalCost) + 1;
                 for ( ; nBitsToDecrease > 1; nBitsToDecrease--) {
-                    U32 highPos = rankLast[nBitsToDecrease];
-                    U32 lowPos = rankLast[nBitsToDecrease-1];
+                    U32 const highPos = rankLast[nBitsToDecrease];
+                    U32 const lowPos = rankLast[nBitsToDecrease-1];
                     if (highPos == noSymbol) continue;
                     if (lowPos == noSymbol) break;
                     {   U32 const highTotal = huffNode[highPos].count;
@@ -299,7 +277,8 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
                 if (rankLast[1] == noSymbol) {  /* special case : no rank 1 symbol (using maxNbBits-1); let's create one from largest rank 0 (using maxNbBits) */
                     while (huffNode[n].nbBits == maxNbBits) n--;
                     huffNode[n+1].nbBits--;
-                    rankLast[1] = n+1;
+                    assert(n >= 0);
+                    rankLast[1] = (U32)(n+1);
                     totalCost++;
                     continue;
                 }
@@ -311,29 +290,36 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
     return maxNbBits;
 }
 
-
 typedef struct {
     U32 base;
     U32 current;
 } rankPos;
 
-static void HUF_sort(nodeElt* huffNode, const unsigned* count, U32 maxSymbolValue)
+typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32];
+
+#define RANK_POSITION_TABLE_SIZE 32
+
+typedef struct {
+  huffNodeTable huffNodeTbl;
+  rankPos rankPosition[RANK_POSITION_TABLE_SIZE];
+} HUF_buildCTable_wksp_tables;
+
+static void HUF_sort(nodeElt* huffNode, const unsigned* count, U32 maxSymbolValue, rankPos* rankPosition)
 {
-    rankPos rank[32];
     U32 n;
 
-    memset(rank, 0, sizeof(rank));
+    memset(rankPosition, 0, sizeof(*rankPosition) * RANK_POSITION_TABLE_SIZE);
     for (n=0; n<=maxSymbolValue; n++) {
         U32 r = BIT_highbit32(count[n] + 1);
-        rank[r].base ++;
+        rankPosition[r].base ++;
     }
-    for (n=30; n>0; n--) rank[n-1].base += rank[n].base;
-    for (n=0; n<32; n++) rank[n].current = rank[n].base;
+    for (n=30; n>0; n--) rankPosition[n-1].base += rankPosition[n].base;
+    for (n=0; n<32; n++) rankPosition[n].current = rankPosition[n].base;
     for (n=0; n<=maxSymbolValue; n++) {
         U32 const c = count[n];
         U32 const r = BIT_highbit32(c+1) + 1;
-        U32 pos = rank[r].current++;
-        while ((pos > rank[r].base) && (c > huffNode[pos-1].count)) {
+        U32 pos = rankPosition[r].current++;
+        while ((pos > rankPosition[r].base) && (c > huffNode[pos-1].count)) {
             huffNode[pos] = huffNode[pos-1];
             pos--;
         }
@@ -345,45 +331,48 @@ static void HUF_sort(nodeElt* huffNode, const unsigned* count, U32 maxSymbolValu
 
 /** HUF_buildCTable_wksp() :
  *  Same as HUF_buildCTable(), but using externally allocated scratch buffer.
- *  `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as a table of HUF_CTABLE_WORKSPACE_SIZE_U32 unsigned.
+ *  `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as sizeof(HUF_buildCTable_wksp_tables).
  */
 #define STARTNODE (HUF_SYMBOLVALUE_MAX+1)
-typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32];
+
 size_t HUF_buildCTable_wksp (HUF_CElt* tree, const unsigned* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
 {
-    nodeElt* const huffNode0 = (nodeElt*)workSpace;
+    HUF_buildCTable_wksp_tables* const wksp_tables = (HUF_buildCTable_wksp_tables*)workSpace;
+    nodeElt* const huffNode0 = wksp_tables->huffNodeTbl;
     nodeElt* const huffNode = huffNode0+1;
-    U32 n, nonNullRank;
+    int nonNullRank;
     int lowS, lowN;
-    U16 nodeNb = STARTNODE;
-    U32 nodeRoot;
+    int nodeNb = STARTNODE;
+    int n, nodeRoot;
 
     /* safety checks */
     if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC);  /* must be aligned on 4-bytes boundaries */
-    if (wkspSize < sizeof(huffNodeTable)) return ERROR(workSpace_tooSmall);
+    if (wkspSize < sizeof(HUF_buildCTable_wksp_tables))
+      return ERROR(workSpace_tooSmall);
     if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
-    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
+    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX)
+      return ERROR(maxSymbolValue_tooLarge);
     memset(huffNode0, 0, sizeof(huffNodeTable));
 
     /* sort, decreasing order */
-    HUF_sort(huffNode, count, maxSymbolValue);
+    HUF_sort(huffNode, count, maxSymbolValue, wksp_tables->rankPosition);
 
     /* init for parents */
-    nonNullRank = maxSymbolValue;
+    nonNullRank = (int)maxSymbolValue;
     while(huffNode[nonNullRank].count == 0) nonNullRank--;
     lowS = nonNullRank; nodeRoot = nodeNb + lowS - 1; lowN = nodeNb;
     huffNode[nodeNb].count = huffNode[lowS].count + huffNode[lowS-1].count;
-    huffNode[lowS].parent = huffNode[lowS-1].parent = nodeNb;
+    huffNode[lowS].parent = huffNode[lowS-1].parent = (U16)nodeNb;
     nodeNb++; lowS-=2;
     for (n=nodeNb; n<=nodeRoot; n++) huffNode[n].count = (U32)(1U<<30);
     huffNode0[0].count = (U32)(1U<<31);  /* fake entry, strong barrier */
 
     /* create parents */
     while (nodeNb <= nodeRoot) {
-        U32 n1 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
-        U32 n2 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
+        int const n1 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
+        int const n2 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
         huffNode[nodeNb].count = huffNode[n1].count + huffNode[n2].count;
-        huffNode[n1].parent = huffNode[n2].parent = nodeNb;
+        huffNode[n1].parent = huffNode[n2].parent = (U16)nodeNb;
         nodeNb++;
     }
 
@@ -395,24 +384,25 @@ size_t HUF_buildCTable_wksp (HUF_CElt* tree, const unsigned* count, U32 maxSymbo
         huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
 
     /* enforce maxTableLog */
-    maxNbBits = HUF_setMaxHeight(huffNode, nonNullRank, maxNbBits);
+    maxNbBits = HUF_setMaxHeight(huffNode, (U32)nonNullRank, maxNbBits);
 
     /* fill result into tree (val, nbBits) */
     {   U16 nbPerRank[HUF_TABLELOG_MAX+1] = {0};
         U16 valPerRank[HUF_TABLELOG_MAX+1] = {0};
+        int const alphabetSize = (int)(maxSymbolValue + 1);
         if (maxNbBits > HUF_TABLELOG_MAX) return ERROR(GENERIC);   /* check fit into table */
         for (n=0; n<=nonNullRank; n++)
             nbPerRank[huffNode[n].nbBits]++;
         /* determine stating value per rank */
         {   U16 min = 0;
-            for (n=maxNbBits; n>0; n--) {
+            for (n=(int)maxNbBits; n>0; n--) {
                 valPerRank[n] = min;      /* get starting value within each rank */
                 min += nbPerRank[n];
                 min >>= 1;
         }   }
-        for (n=0; n<=maxSymbolValue; n++)
+        for (n=0; n<alphabetSize; n++)
             tree[huffNode[n].byte].nbBits = huffNode[n].nbBits;   /* push nbBits per symbol, symbol order */
-        for (n=0; n<=maxSymbolValue; n++)
+        for (n=0; n<alphabetSize; n++)
             tree[n].val = valPerRank[tree[n].nbBits]++;   /* assign value within rank, symbol order */
     }
 
@@ -425,8 +415,8 @@ size_t HUF_buildCTable_wksp (HUF_CElt* tree, const unsigned* count, U32 maxSymbo
  */
 size_t HUF_buildCTable (HUF_CElt* tree, const unsigned* count, unsigned maxSymbolValue, unsigned maxNbBits)
 {
-    huffNodeTable nodeTable;
-    return HUF_buildCTable_wksp(tree, count, maxSymbolValue, maxNbBits, nodeTable, sizeof(nodeTable));
+    HUF_buildCTable_wksp_tables workspace;
+    return HUF_buildCTable_wksp(tree, count, maxSymbolValue, maxNbBits, &workspace, sizeof(workspace));
 }
 
 size_t HUF_estimateCompressedSize(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue)
@@ -439,7 +429,7 @@ size_t HUF_estimateCompressedSize(const HUF_CElt* CTable, const unsigned* count,
     return nbBits >> 3;
 }
 
-static int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue) {
+int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue) {
   int bad = 0;
   int s;
   for (s = 0; s <= (int)maxSymbolValue; ++s) {
@@ -478,7 +468,7 @@ HUF_compress1X_usingCTable_internal_body(void* dst, size_t dstSize,
 
     /* init */
     if (dstSize < 8) return 0;   /* not enough space to compress */
-    { size_t const initErr = BIT_initCStream(&bitC, op, oend-op);
+    { size_t const initErr = BIT_initCStream(&bitC, op, (size_t)(oend-op));
       if (HUF_isError(initErr)) return 0; }
 
     n = srcSize & ~3;  /* join to mod 4 */
@@ -575,7 +565,8 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
     if (srcSize < 12) return 0;   /* no saving possible : too small input */
     op += 6;   /* jumpTable */
 
-    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
+    assert(op <= oend);
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
         if (cSize==0) return 0;
         assert(cSize <= 65535);
         MEM_writeLE16(ostart, (U16)cSize);
@@ -583,7 +574,8 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
     }
 
     ip += segmentSize;
-    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
+    assert(op <= oend);
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
         if (cSize==0) return 0;
         assert(cSize <= 65535);
         MEM_writeLE16(ostart+2, (U16)cSize);
@@ -591,7 +583,8 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
     }
 
     ip += segmentSize;
-    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
+    assert(op <= oend);
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
         if (cSize==0) return 0;
         assert(cSize <= 65535);
         MEM_writeLE16(ostart+4, (U16)cSize);
@@ -599,12 +592,14 @@ HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
     }
 
     ip += segmentSize;
-    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, iend-ip, CTable, bmi2) );
+    assert(op <= oend);
+    assert(ip <= iend);
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, (size_t)(iend-ip), CTable, bmi2) );
         if (cSize==0) return 0;
         op += cSize;
     }
 
-    return op-ostart;
+    return (size_t)(op-ostart);
 }
 
 size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
@@ -620,20 +615,21 @@ static size_t HUF_compressCTable_internal(
                 HUF_nbStreams_e nbStreams, const HUF_CElt* CTable, const int bmi2)
 {
     size_t const cSize = (nbStreams==HUF_singleStream) ?
-                         HUF_compress1X_usingCTable_internal(op, oend - op, src, srcSize, CTable, bmi2) :
-                         HUF_compress4X_usingCTable_internal(op, oend - op, src, srcSize, CTable, bmi2);
+                         HUF_compress1X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, bmi2) :
+                         HUF_compress4X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, bmi2);
     if (HUF_isError(cSize)) { return cSize; }
     if (cSize==0) { return 0; }   /* uncompressible */
     op += cSize;
     /* check compressibility */
+    assert(op >= ostart);
     if ((size_t)(op-ostart) >= srcSize-1) { return 0; }
-    return op-ostart;
+    return (size_t)(op-ostart);
 }
 
 typedef struct {
     unsigned count[HUF_SYMBOLVALUE_MAX + 1];
     HUF_CElt CTable[HUF_SYMBOLVALUE_MAX + 1];
-    huffNodeTable nodeTable;
+    HUF_buildCTable_wksp_tables buildCTable_wksp;
 } HUF_compress_tables_t;
 
 /* HUF_compress_internal() :
@@ -652,6 +648,8 @@ HUF_compress_internal (void* dst, size_t dstSize,
     BYTE* const oend = ostart + dstSize;
     BYTE* op = ostart;
 
+    HUF_STATIC_ASSERT(sizeof(*table) <= HUF_WORKSPACE_SIZE);
+
     /* checks & inits */
     if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC);  /* must be aligned on 4-bytes boundaries */
     if (wkspSize < HUF_WORKSPACE_SIZE) return ERROR(workSpace_tooSmall);
@@ -693,7 +691,7 @@ HUF_compress_internal (void* dst, size_t dstSize,
     huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
     {   size_t const maxBits = HUF_buildCTable_wksp(table->CTable, table->count,
                                             maxSymbolValue, huffLog,
-                                            table->nodeTable, sizeof(table->nodeTable));
+                                            &table->buildCTable_wksp, sizeof(table->buildCTable_wksp));
         CHECK_F(maxBits);
         huffLog = (U32)maxBits;
         /* Zero unused symbols in CTable, so we can check it for validity */
diff --git a/zstd/huf_decompress.c b/zstd/huf_decompress.c
index e599070a..ab5db481 100644
--- a/zstd/huf_decompress.c
+++ b/zstd/huf_decompress.c
@@ -1,35 +1,15 @@
 /* ******************************************************************
-   huff0 huffman decoder,
-   part of Finite State Entropy library
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * huff0 huffman decoder,
+ * part of Finite State Entropy library
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *
+ * 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.
 ****************************************************************** */
 
 /* **************************************************************
@@ -61,9 +41,6 @@
 *  Error Management
 ****************************************************************/
 #define HUF_isError ERR_isError
-#ifndef CHECK_F
-#define CHECK_F(f) { size_t const err_ = (f); if (HUF_isError(err_)) return err_; }
-#endif
 
 
 /* **************************************************************
@@ -855,7 +832,7 @@ HUF_decompress4X2_usingDTable_internal_body(
             HUF_DECODE_SYMBOLX2_0(op2, &bitD2);
             HUF_DECODE_SYMBOLX2_0(op3, &bitD3);
             HUF_DECODE_SYMBOLX2_0(op4, &bitD4);
-            endSignal = LIKELY(
+            endSignal = (U32)LIKELY(
                         (BIT_reloadDStreamFast(&bitD1) == BIT_DStream_unfinished)
                       & (BIT_reloadDStreamFast(&bitD2) == BIT_DStream_unfinished)
                       & (BIT_reloadDStreamFast(&bitD3) == BIT_DStream_unfinished)
diff --git a/zstd/mem.h b/zstd/mem.h
index 530d30c8..89c8aea7 100644
--- a/zstd/mem.h
+++ b/zstd/mem.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/pool.c b/zstd/pool.c
index f5759350..aa4b4de0 100644
--- a/zstd/pool.c
+++ b/zstd/pool.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/pool.h b/zstd/pool.h
index 458d37f1..8503e2d2 100644
--- a/zstd/pool.h
+++ b/zstd/pool.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/threading.c b/zstd/threading.c
index 482664bd..e2edb313 100644
--- a/zstd/threading.c
+++ b/zstd/threading.c
@@ -2,12 +2,13 @@
  * Copyright (c) 2016 Tino Reichardt
  * All rights reserved.
  *
+ * You can contact the author at:
+ * - zstdmt source repository: https://github.com/mcmilk/zstdmt
+ *
  * 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 can contact the author at:
- * - zstdmt source repository: https://github.com/mcmilk/zstdmt
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 /**
diff --git a/zstd/threading.h b/zstd/threading.h
index 3193ca7d..fd0060d5 100644
--- a/zstd/threading.h
+++ b/zstd/threading.h
@@ -2,12 +2,13 @@
  * Copyright (c) 2016 Tino Reichardt
  * All rights reserved.
  *
+ * You can contact the author at:
+ * - zstdmt source repository: https://github.com/mcmilk/zstdmt
+ *
  * 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 can contact the author at:
- * - zstdmt source repository: https://github.com/mcmilk/zstdmt
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 #ifndef THREADING_H_938743
diff --git a/zstd/xxhash.c b/zstd/xxhash.c
index 99d24596..597de18f 100644
--- a/zstd/xxhash.c
+++ b/zstd/xxhash.c
@@ -1,35 +1,15 @@
 /*
-*  xxHash - Fast Hash algorithm
-*  Copyright (C) 2012-2016, Yann Collet
-*
-*  BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-*
-*  Redistribution and use in source and binary forms, with or without
-*  modification, are permitted provided that the following conditions are
-*  met:
-*
-*  * Redistributions of source code must retain the above copyright
-*  notice, this list of conditions and the following disclaimer.
-*  * Redistributions in binary form must reproduce the above
-*  copyright notice, this list of conditions and the following disclaimer
-*  in the documentation and/or other materials provided with the
-*  distribution.
-*
-*  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-*  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-*  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-*  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-*  OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-*  SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-*  LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-*  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-*  THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-*  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-*  OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-*
-*  You can contact the author at :
-*  - xxHash homepage: http://www.xxhash.com
-*  - xxHash source repository : https://github.com/Cyan4973/xxHash
+ *  xxHash - Fast Hash algorithm
+ *  Copyright (c) 2012-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - xxHash homepage: http://www.xxhash.com
+ *  - xxHash source repository : https://github.com/Cyan4973/xxHash
+ * 
+ * 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.
 */
 
 
@@ -115,7 +95,7 @@ static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcp
 /* *************************************
 *  Compiler Specific Options
 ***************************************/
-#if defined (__GNUC__) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#if (defined(__GNUC__) && !defined(__STRICT_ANSI__)) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
 #  define INLINE_KEYWORD inline
 #else
 #  define INLINE_KEYWORD
@@ -729,7 +709,9 @@ FORCE_INLINE_TEMPLATE XXH_errorcode XXH64_update_endian (XXH64_state_t* state, c
     state->total_len += len;
 
     if (state->memsize + len < 32) {  /* fill in tmp buffer */
-        XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len);
+        if (input != NULL) {
+            XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len);
+        }
         state->memsize += (U32)len;
         return XXH_OK;
     }
diff --git a/zstd/xxhash.h b/zstd/xxhash.h
index 9bad1f59..4207eba8 100644
--- a/zstd/xxhash.h
+++ b/zstd/xxhash.h
@@ -1,35 +1,15 @@
 /*
-   xxHash - Extremely Fast Hash algorithm
-   Header File
-   Copyright (C) 2012-2016, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-   You can contact the author at :
-   - xxHash source repository : https://github.com/Cyan4973/xxHash
+ * xxHash - Extremely Fast Hash algorithm
+ * Header File
+ * Copyright (c) 2012-2020, Yann Collet, Facebook, Inc.
+ *
+ * You can contact the author at :
+ * - xxHash source repository : https://github.com/Cyan4973/xxHash
+ * 
+ * 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.
 */
 
 /* Notice extracted from xxHash homepage :
diff --git a/zstd/zstd.h b/zstd/zstd.h
index 73cb3d31..8c6fc6ae 100644
--- a/zstd/zstd.h
+++ b/zstd/zstd.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -274,7 +274,10 @@ typedef enum {
                               * Default level is ZSTD_CLEVEL_DEFAULT==3.
                               * Special: value 0 means default, which is controlled by ZSTD_CLEVEL_DEFAULT.
                               * Note 1 : it's possible to pass a negative compression level.
-                              * Note 2 : setting a level resets all other compression parameters to default */
+                              * Note 2 : setting a level does not automatically set all other compression parameters
+                              *   to default. Setting this will however eventually dynamically impact the compression
+                              *   parameters which have not been manually set. The manually set
+                              *   ones will 'stick'. */
     /* Advanced compression parameters :
      * It's possible to pin down compression parameters to some specific values.
      * In which case, these values are no longer dynamically selected by the compressor */
@@ -519,11 +522,13 @@ typedef enum {
     /* note : additional experimental parameters are also available
      * within the experimental section of the API.
      * At the time of this writing, they include :
-     * ZSTD_c_format
+     * ZSTD_d_format
+     * ZSTD_d_stableOutBuffer
      * Because they are not stable, it's necessary to define ZSTD_STATIC_LINKING_ONLY to access them.
      * note : never ever use experimentalParam? names directly
      */
-     ZSTD_d_experimentalParam1=1000
+     ZSTD_d_experimentalParam1=1000,
+     ZSTD_d_experimentalParam2=1001
 
 } ZSTD_dParameter;
 
@@ -1263,23 +1268,28 @@ ZSTDLIB_API size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,
 ***************************************/
 
 /*! ZSTD_estimate*() :
- *  These functions make it possible to estimate memory usage of a future
- *  {D,C}Ctx, before its creation.
+ *  These functions make it possible to estimate memory usage
+ *  of a future {D,C}Ctx, before its creation.
+ *
+ *  ZSTD_estimateCCtxSize() will provide a memory budget large enough
+ *  for any compression level up to selected one.
+ *  Note : Unlike ZSTD_estimateCStreamSize*(), this estimate
+ *         does not include space for a window buffer.
+ *         Therefore, the estimation is only guaranteed for single-shot compressions, not streaming.
+ *  The estimate will assume the input may be arbitrarily large,
+ *  which is the worst case.
  *
- *  ZSTD_estimateCCtxSize() will provide a budget large enough for any
- *  compression level up to selected one. Unlike ZSTD_estimateCStreamSize*(),
- *  this estimate does not include space for a window buffer, so this estimate
- *  is guaranteed to be enough for single-shot compressions, but not streaming
- *  compressions. It will however assume the input may be arbitrarily large,
- *  which is the worst case. If srcSize is known to always be small,
- *  ZSTD_estimateCCtxSize_usingCParams() can provide a tighter estimation.
- *  ZSTD_estimateCCtxSize_usingCParams() can be used in tandem with
- *  ZSTD_getCParams() to create cParams from compressionLevel.
- *  ZSTD_estimateCCtxSize_usingCCtxParams() can be used in tandem with
- *  ZSTD_CCtxParams_setParameter().
+ *  When srcSize can be bound by a known and rather "small" value,
+ *  this fact can be used to provide a tighter estimation
+ *  because the CCtx compression context will need less memory.
+ *  This tighter estimation can be provided by more advanced functions
+ *  ZSTD_estimateCCtxSize_usingCParams(), which can be used in tandem with ZSTD_getCParams(),
+ *  and ZSTD_estimateCCtxSize_usingCCtxParams(), which can be used in tandem with ZSTD_CCtxParams_setParameter().
+ *  Both can be used to estimate memory using custom compression parameters and arbitrary srcSize limits.
  *
- *  Note: only single-threaded compression is supported. This function will
- *  return an error code if ZSTD_c_nbWorkers is >= 1. */
+ *  Note 2 : only single-threaded compression is supported.
+ *  ZSTD_estimateCCtxSize_usingCCtxParams() will return an error code if ZSTD_c_nbWorkers is >= 1.
+ */
 ZSTDLIB_API size_t ZSTD_estimateCCtxSize(int compressionLevel);
 ZSTDLIB_API size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams);
 ZSTDLIB_API size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params);
@@ -1642,6 +1652,37 @@ ZSTDLIB_API size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowS
  * allowing selection between ZSTD_format_e input compression formats
  */
 #define ZSTD_d_format ZSTD_d_experimentalParam1
+/* ZSTD_d_stableOutBuffer
+ * Experimental parameter.
+ * Default is 0 == disabled. Set to 1 to enable.
+ *
+ * Tells the decompressor that the ZSTD_outBuffer will ALWAYS be the same
+ * between calls, except for the modifications that zstd makes to pos (the
+ * caller must not modify pos). This is checked by the decompressor, and
+ * decompression will fail if it ever changes. Therefore the ZSTD_outBuffer
+ * MUST be large enough to fit the entire decompressed frame. This will be
+ * checked when the frame content size is known. The data in the ZSTD_outBuffer
+ * in the range [dst, dst + pos) MUST not be modified during decompression
+ * or you will get data corruption.
+ *
+ * When this flags is enabled zstd won't allocate an output buffer, because
+ * it can write directly to the ZSTD_outBuffer, but it will still allocate
+ * an input buffer large enough to fit any compressed block. This will also
+ * avoid the memcpy() from the internal output buffer to the ZSTD_outBuffer.
+ * If you need to avoid the input buffer allocation use the buffer-less
+ * streaming API.
+ *
+ * NOTE: So long as the ZSTD_outBuffer always points to valid memory, using
+ * this flag is ALWAYS memory safe, and will never access out-of-bounds
+ * memory. However, decompression WILL fail if you violate the preconditions.
+ *
+ * WARNING: The data in the ZSTD_outBuffer in the range [dst, dst + pos) MUST
+ * not be modified during decompression or you will get data corruption. This
+ * is because zstd needs to reference data in the ZSTD_outBuffer to regenerate
+ * matches. Normally zstd maintains its own buffer for this purpose, but passing
+ * this flag tells zstd to use the user provided buffer.
+ */
+#define ZSTD_d_stableOutBuffer ZSTD_d_experimentalParam2
 
 /*! ZSTD_DCtx_setFormat() :
  *  Instruct the decoder context about what kind of data to decode next.
diff --git a/zstd/zstd_common.c b/zstd/zstd_common.c
index 667f4a27..91fe3323 100644
--- a/zstd/zstd_common.c
+++ b/zstd/zstd_common.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstd_compress.c b/zstd/zstd_compress.c
index 69ee7275..526cb21f 100644
--- a/zstd/zstd_compress.c
+++ b/zstd/zstd_compress.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -90,7 +90,7 @@ ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem)
     }
 }
 
-ZSTD_CCtx* ZSTD_initStaticCCtx(void *workspace, size_t workspaceSize)
+ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize)
 {
     ZSTD_cwksp ws;
     ZSTD_CCtx* cctx;
@@ -99,9 +99,8 @@ ZSTD_CCtx* ZSTD_initStaticCCtx(void *workspace, size_t workspaceSize)
     ZSTD_cwksp_init(&ws, workspace, workspaceSize);
 
     cctx = (ZSTD_CCtx*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CCtx));
-    if (cctx == NULL) {
-        return NULL;
-    }
+    if (cctx == NULL) return NULL;
+
     memset(cctx, 0, sizeof(ZSTD_CCtx));
     ZSTD_cwksp_move(&cctx->workspace, &ws);
     cctx->staticSize = workspaceSize;
@@ -110,8 +109,7 @@ ZSTD_CCtx* ZSTD_initStaticCCtx(void *workspace, size_t workspaceSize)
     if (!ZSTD_cwksp_check_available(&cctx->workspace, HUF_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t))) return NULL;
     cctx->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
     cctx->blockState.nextCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
-    cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(
-        &cctx->workspace, HUF_WORKSPACE_SIZE);
+    cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cctx->workspace, HUF_WORKSPACE_SIZE);
     cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
     return cctx;
 }
@@ -235,7 +233,7 @@ size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params)
 }
 
 size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel) {
-    RETURN_ERROR_IF(!cctxParams, GENERIC);
+    RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!");
     memset(cctxParams, 0, sizeof(*cctxParams));
     cctxParams->compressionLevel = compressionLevel;
     cctxParams->fParams.contentSizeFlag = 1;
@@ -244,8 +242,8 @@ size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel)
 
 size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params)
 {
-    RETURN_ERROR_IF(!cctxParams, GENERIC);
-    FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) );
+    RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!");
+    FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "");
     memset(cctxParams, 0, sizeof(*cctxParams));
     assert(!ZSTD_checkCParams(params.cParams));
     cctxParams->cParams = params.cParams;
@@ -421,9 +419,8 @@ ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param)
         return bounds;
 
     default:
-        {   ZSTD_bounds const boundError = { ERROR(parameter_unsupported), 0, 0 };
-            return boundError;
-        }
+        bounds.error = ERROR(parameter_unsupported);
+        return bounds;
     }
 }
 
@@ -441,7 +438,7 @@ static size_t ZSTD_cParam_clampBounds(ZSTD_cParameter cParam, int* value)
 
 #define BOUNDCHECK(cParam, val) { \
     RETURN_ERROR_IF(!ZSTD_cParam_withinBounds(cParam,val), \
-                    parameter_outOfBound); \
+                    parameter_outOfBound, "Param out of bounds"); \
 }
 
 
@@ -489,7 +486,7 @@ size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value)
         if (ZSTD_isUpdateAuthorized(param)) {
             cctx->cParamsChanged = 1;
         } else {
-            RETURN_ERROR(stage_wrong);
+            RETURN_ERROR(stage_wrong, "can only set params in ctx init stage");
     }   }
 
     switch(param)
@@ -526,7 +523,7 @@ size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value)
     case ZSTD_c_srcSizeHint:
         break;
 
-    default: RETURN_ERROR(parameter_unsupported);
+    default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
     }
     return ZSTD_CCtxParams_setParameter(&cctx->requestedParams, param, value);
 }
@@ -543,7 +540,7 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
         return (size_t)CCtxParams->format;
 
     case ZSTD_c_compressionLevel : {
-        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value));
+        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
         if (value) {  /* 0 : does not change current level */
             CCtxParams->compressionLevel = value;
         }
@@ -631,7 +628,7 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
         RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
         return 0;
 #else
-        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value));
+        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
         CCtxParams->nbWorkers = value;
         return CCtxParams->nbWorkers;
 #endif
@@ -644,7 +641,7 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
         /* Adjust to the minimum non-default value. */
         if (value != 0 && value < ZSTDMT_JOBSIZE_MIN)
             value = ZSTDMT_JOBSIZE_MIN;
-        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value));
+        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
         assert(value >= 0);
         CCtxParams->jobSize = value;
         return CCtxParams->jobSize;
@@ -655,7 +652,7 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
         RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
         return 0;
 #else
-        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value));
+        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "");
         CCtxParams->overlapLog = value;
         return CCtxParams->overlapLog;
 #endif
@@ -665,7 +662,7 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
         RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
         return 0;
 #else
-        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value));
+        FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "");
         CCtxParams->rsyncable = value;
         return CCtxParams->rsyncable;
 #endif
@@ -694,7 +691,7 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
 
     case ZSTD_c_ldmHashRateLog :
         RETURN_ERROR_IF(value > ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN,
-                        parameter_outOfBound);
+                        parameter_outOfBound, "Param out of bounds!");
         CCtxParams->ldmParams.hashRateLog = value;
         return CCtxParams->ldmParams.hashRateLog;
 
@@ -834,8 +831,11 @@ size_t ZSTD_CCtx_setParametersUsingCCtxParams(
         ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params)
 {
     DEBUGLOG(4, "ZSTD_CCtx_setParametersUsingCCtxParams");
-    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong);
-    RETURN_ERROR_IF(cctx->cdict, stage_wrong);
+    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+                    "The context is in the wrong stage!");
+    RETURN_ERROR_IF(cctx->cdict, stage_wrong,
+                    "Can't override parameters with cdict attached (some must "
+                    "be inherited from the cdict).");
 
     cctx->requestedParams = *params;
     return 0;
@@ -844,7 +844,8 @@ size_t ZSTD_CCtx_setParametersUsingCCtxParams(
 ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize)
 {
     DEBUGLOG(4, "ZSTD_CCtx_setPledgedSrcSize to %u bytes", (U32)pledgedSrcSize);
-    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong);
+    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+                    "Can't set pledgedSrcSize when not in init stage.");
     cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1;
     return 0;
 }
@@ -882,7 +883,7 @@ static size_t ZSTD_initLocalDict(ZSTD_CCtx* cctx)
             dl->dictContentType,
             cParams,
             cctx->customMem);
-    RETURN_ERROR_IF(!dl->cdict, memory_allocation);
+    RETURN_ERROR_IF(!dl->cdict, memory_allocation, "ZSTD_createCDict_advanced failed");
     cctx->cdict = dl->cdict;
     return 0;
 }
@@ -891,7 +892,8 @@ size_t ZSTD_CCtx_loadDictionary_advanced(
         ZSTD_CCtx* cctx, const void* dict, size_t dictSize,
         ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType)
 {
-    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong);
+    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+                    "Can't load a dictionary when ctx is not in init stage.");
     RETURN_ERROR_IF(cctx->staticSize, memory_allocation,
                     "no malloc for static CCtx");
     DEBUGLOG(4, "ZSTD_CCtx_loadDictionary_advanced (size: %u)", (U32)dictSize);
@@ -902,7 +904,7 @@ size_t ZSTD_CCtx_loadDictionary_advanced(
         cctx->localDict.dict = dict;
     } else {
         void* dictBuffer = ZSTD_malloc(dictSize, cctx->customMem);
-        RETURN_ERROR_IF(!dictBuffer, memory_allocation);
+        RETURN_ERROR_IF(!dictBuffer, memory_allocation, "NULL pointer!");
         memcpy(dictBuffer, dict, dictSize);
         cctx->localDict.dictBuffer = dictBuffer;
         cctx->localDict.dict = dictBuffer;
@@ -928,7 +930,8 @@ ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, s
 
 size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)
 {
-    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong);
+    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+                    "Can't ref a dict when ctx not in init stage.");
     /* Free the existing local cdict (if any) to save memory. */
     ZSTD_clearAllDicts(cctx);
     cctx->cdict = cdict;
@@ -943,11 +946,14 @@ size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSiz
 size_t ZSTD_CCtx_refPrefix_advanced(
         ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType)
 {
-    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong);
+    RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+                    "Can't ref a prefix when ctx not in init stage.");
     ZSTD_clearAllDicts(cctx);
-    cctx->prefixDict.dict = prefix;
-    cctx->prefixDict.dictSize = prefixSize;
-    cctx->prefixDict.dictContentType = dictContentType;
+    if (prefix != NULL && prefixSize > 0) {
+        cctx->prefixDict.dict = prefix;
+        cctx->prefixDict.dictSize = prefixSize;
+        cctx->prefixDict.dictContentType = dictContentType;
+    }
     return 0;
 }
 
@@ -962,7 +968,8 @@ size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset)
     }
     if ( (reset == ZSTD_reset_parameters)
       || (reset == ZSTD_reset_session_and_parameters) ) {
-        RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong);
+        RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
+                        "Can't reset parameters only when not in init stage.");
         ZSTD_clearAllDicts(cctx);
         return ZSTD_CCtxParams_reset(&cctx->requestedParams);
     }
@@ -1009,7 +1016,7 @@ ZSTD_clampCParams(ZSTD_compressionParameters cParams)
 
 /** ZSTD_cycleLog() :
  *  condition for correct operation : hashLog > 1 */
-static U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
+U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
 {
     U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2);
     return hashLog - btScale;
@@ -1134,13 +1141,26 @@ size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params)
         size_t const ldmSpace = ZSTD_ldm_getTableSize(params->ldmParams);
         size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize) * sizeof(rawSeq));
 
-        size_t const neededSpace = entropySpace + blockStateSpace + tokenSpace +
-                                   matchStateSize + ldmSpace + ldmSeqSpace;
+        /* estimateCCtxSize is for one-shot compression. So no buffers should
+         * be needed. However, we still allocate two 0-sized buffers, which can
+         * take space under ASAN. */
+        size_t const bufferSpace = ZSTD_cwksp_alloc_size(0)
+                                 + ZSTD_cwksp_alloc_size(0);
+
         size_t const cctxSpace = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx));
 
-        DEBUGLOG(5, "sizeof(ZSTD_CCtx) : %u", (U32)cctxSpace);
+        size_t const neededSpace =
+            cctxSpace +
+            entropySpace +
+            blockStateSpace +
+            ldmSpace +
+            ldmSeqSpace +
+            matchStateSize +
+            tokenSpace +
+            bufferSpace;
+
         DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace);
-        return cctxSpace + neededSpace;
+        return neededSpace;
     }
 }
 
@@ -1435,7 +1455,7 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
             needsIndexReset = ZSTDirp_reset;
         }
 
-        ZSTD_cwksp_bump_oversized_duration(ws, 0);
+        if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0);
 
         /* Check if workspace is large enough, alloc a new one if needed */
         {   size_t const cctxSpace = zc->staticSize ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0;
@@ -1472,7 +1492,7 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
                 needsIndexReset = ZSTDirp_reset;
 
                 ZSTD_cwksp_free(ws, zc->customMem);
-                FORWARD_IF_ERROR(ZSTD_cwksp_create(ws, neededSpace, zc->customMem));
+                FORWARD_IF_ERROR(ZSTD_cwksp_create(ws, neededSpace, zc->customMem), "");
 
                 DEBUGLOG(5, "reserving object space");
                 /* Statically sized space.
@@ -1543,7 +1563,7 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
             &params.cParams,
             crp,
             needsIndexReset,
-            ZSTD_resetTarget_CCtx));
+            ZSTD_resetTarget_CCtx), "");
 
         /* ldm hash table */
         if (params.ldmParams.enableLdm) {
@@ -1556,6 +1576,7 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
 
             ZSTD_window_init(&zc->ldmState.window);
             ZSTD_window_clear(&zc->ldmState.window);
+            zc->ldmState.loadedDictEnd = 0;
         }
 
         DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws));
@@ -1621,7 +1642,7 @@ ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx,
         params.cParams = ZSTD_adjustCParams_internal(*cdict_cParams, pledgedSrcSize, 0);
         params.cParams.windowLog = windowLog;
         FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
-                                                 ZSTDcrp_makeClean, zbuff));
+                                                 ZSTDcrp_makeClean, zbuff), "");
         assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy);
     }
 
@@ -1670,7 +1691,7 @@ static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx,
         params.cParams = *cdict_cParams;
         params.cParams.windowLog = windowLog;
         FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
-                                                 ZSTDcrp_leaveDirty, zbuff));
+                                                 ZSTDcrp_leaveDirty, zbuff), "");
         assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy);
         assert(cctx->appliedParams.cParams.hashLog == cdict_cParams->hashLog);
         assert(cctx->appliedParams.cParams.chainLog == cdict_cParams->chainLog);
@@ -1751,7 +1772,8 @@ static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx,
                             ZSTD_buffered_policy_e zbuff)
 {
     DEBUGLOG(5, "ZSTD_copyCCtx_internal");
-    RETURN_ERROR_IF(srcCCtx->stage!=ZSTDcs_init, stage_wrong);
+    RETURN_ERROR_IF(srcCCtx->stage!=ZSTDcs_init, stage_wrong,
+                    "Can't copy a ctx that's not in init stage.");
 
     memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem));
     {   ZSTD_CCtx_params params = dstCCtx->requestedParams;
@@ -1926,21 +1948,6 @@ void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)
         mlCodeTable[seqStorePtr->longLengthPos] = MaxML;
 }
 
-static int ZSTD_disableLiteralsCompression(const ZSTD_CCtx_params* cctxParams)
-{
-    switch (cctxParams->literalCompressionMode) {
-    case ZSTD_lcm_huffman:
-        return 0;
-    case ZSTD_lcm_uncompressed:
-        return 1;
-    default:
-        assert(0 /* impossible: pre-validated */);
-        /* fall-through */
-    case ZSTD_lcm_auto:
-        return (cctxParams->cParams.strategy == ZSTD_fast) && (cctxParams->cParams.targetLength > 0);
-    }
-}
-
 /* ZSTD_useTargetCBlockSize():
  * Returns if target compressed block size param is being used.
  * If used, compression will do best effort to make a compressed block size to be around targetCBlockSize.
@@ -1994,14 +2001,14 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                     literals, litSize,
                                     entropyWorkspace, entropyWkspSize,
                                     bmi2);
-        FORWARD_IF_ERROR(cSize);
+        FORWARD_IF_ERROR(cSize, "ZSTD_compressLiterals failed");
         assert(cSize <= dstCapacity);
         op += cSize;
     }
 
     /* Sequences Header */
     RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,
-                    dstSize_tooSmall);
+                    dstSize_tooSmall, "Can't fit seq hdr in output buf!");
     if (nbSeq < 128) {
         *op++ = (BYTE)nbSeq;
     } else if (nbSeq < LONGNBSEQ) {
@@ -2046,7 +2053,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                 prevEntropy->fse.litlengthCTable,
                 sizeof(prevEntropy->fse.litlengthCTable),
                 entropyWorkspace, entropyWkspSize);
-            FORWARD_IF_ERROR(countSize);
+            FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for LitLens failed");
             if (LLtype == set_compressed)
                 lastNCount = op;
             op += countSize;
@@ -2074,7 +2081,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                 prevEntropy->fse.offcodeCTable,
                 sizeof(prevEntropy->fse.offcodeCTable),
                 entropyWorkspace, entropyWkspSize);
-            FORWARD_IF_ERROR(countSize);
+            FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for Offsets failed");
             if (Offtype == set_compressed)
                 lastNCount = op;
             op += countSize;
@@ -2100,7 +2107,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                 prevEntropy->fse.matchlengthCTable,
                 sizeof(prevEntropy->fse.matchlengthCTable),
                 entropyWorkspace, entropyWkspSize);
-            FORWARD_IF_ERROR(countSize);
+            FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for MatchLengths failed");
             if (MLtype == set_compressed)
                 lastNCount = op;
             op += countSize;
@@ -2116,7 +2123,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                         CTable_LitLength, llCodeTable,
                                         sequences, nbSeq,
                                         longOffsets, bmi2);
-        FORWARD_IF_ERROR(bitstreamSize);
+        FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed");
         op += bitstreamSize;
         assert(op <= oend);
         /* zstd versions <= 1.3.4 mistakenly report corruption when
@@ -2160,7 +2167,7 @@ ZSTD_compressSequences(seqStore_t* seqStorePtr,
      */
     if ((cSize == ERROR(dstSize_tooSmall)) & (srcSize <= dstCapacity))
         return 0;  /* block not compressed */
-    FORWARD_IF_ERROR(cSize);
+    FORWARD_IF_ERROR(cSize, "ZSTD_compressSequences_internal failed");
 
     /* Check compressibility */
     {   size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, cctxParams->cParams.strategy);
@@ -2286,7 +2293,7 @@ static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize)
             /* Updates ldmSeqStore.size */
             FORWARD_IF_ERROR(ZSTD_ldm_generateSequences(&zc->ldmState, &ldmSeqStore,
                                                &zc->appliedParams.ldmParams,
-                                               src, srcSize));
+                                               src, srcSize), "");
             /* Updates ldmSeqStore.pos */
             lastLLSize =
                 ZSTD_ldm_blockCompress(&ldmSeqStore,
@@ -2362,7 +2369,7 @@ size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,
     void* dst = ZSTD_malloc(dstCapacity, ZSTD_defaultCMem);
     SeqCollector seqCollector;
 
-    RETURN_ERROR_IF(dst == NULL, memory_allocation);
+    RETURN_ERROR_IF(dst == NULL, memory_allocation, "NULL pointer!");
 
     seqCollector.collectSequences = 1;
     seqCollector.seqStart = outSeqs;
@@ -2385,6 +2392,18 @@ static int ZSTD_isRLE(const BYTE *ip, size_t length) {
     return 1;
 }
 
+/* Returns true if the given block may be RLE.
+ * This is just a heuristic based on the compressibility.
+ * It may return both false positives and false negatives.
+ */
+static int ZSTD_maybeRLE(seqStore_t const* seqStore)
+{
+    size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart);
+    size_t const nbLits = (size_t)(seqStore->lit - seqStore->litStart);
+
+    return nbSeqs < 4 && nbLits < 10;
+}
+
 static void ZSTD_confirmRepcodesAndEntropyTables(ZSTD_CCtx* zc)
 {
     ZSTD_compressedBlockState_t* const tmp = zc->blockState.prevCBlock;
@@ -2409,7 +2428,7 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
                 (unsigned)zc->blockState.matchState.nextToUpdate);
 
     {   const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
-        FORWARD_IF_ERROR(bss);
+        FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");
         if (bss == ZSTDbss_noCompress) { cSize = 0; goto out; }
     }
 
@@ -2461,6 +2480,16 @@ static size_t ZSTD_compressBlock_targetCBlockSize_body(ZSTD_CCtx* zc,
 {
     DEBUGLOG(6, "Attempting ZSTD_compressSuperBlock()");
     if (bss == ZSTDbss_compress) {
+        if (/* We don't want to emit our first block as a RLE even if it qualifies because
+            * doing so will cause the decoder (cli only) to throw a "should consume all input error."
+            * This is only an issue for zstd <= v1.4.3
+            */
+            !zc->isFirstBlock &&
+            ZSTD_maybeRLE(&zc->seqStore) &&
+            ZSTD_isRLE((BYTE const*)src, srcSize))
+        {
+            return ZSTD_rleCompressBlock(dst, dstCapacity, *(BYTE const*)src, srcSize, lastBlock);
+        }
         /* Attempt superblock compression.
          *
          * Note that compressed size of ZSTD_compressSuperBlock() is not bound by the
@@ -2479,12 +2508,15 @@ static size_t ZSTD_compressBlock_targetCBlockSize_body(ZSTD_CCtx* zc,
          *   * cSize >= blockBound(srcSize): We have expanded the block too much so
          *     emit an uncompressed block.
          */
-        size_t const cSize = ZSTD_compressSuperBlock(zc, dst, dstCapacity, lastBlock);
-        if (cSize != ERROR(dstSize_tooSmall)) {
-            FORWARD_IF_ERROR(cSize);
-            if (cSize != 0 && cSize < srcSize + ZSTD_blockHeaderSize) {
-                ZSTD_confirmRepcodesAndEntropyTables(zc);
-                return cSize;
+        {
+            size_t const cSize = ZSTD_compressSuperBlock(zc, dst, dstCapacity, src, srcSize, lastBlock);
+            if (cSize != ERROR(dstSize_tooSmall)) {
+                size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, zc->appliedParams.cParams.strategy);
+                FORWARD_IF_ERROR(cSize, "ZSTD_compressSuperBlock failed");
+                if (cSize != 0 && cSize < maxCSize + ZSTD_blockHeaderSize) {
+                    ZSTD_confirmRepcodesAndEntropyTables(zc);
+                    return cSize;
+                }
             }
         }
     }
@@ -2505,10 +2537,10 @@ static size_t ZSTD_compressBlock_targetCBlockSize(ZSTD_CCtx* zc,
     const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
     DEBUGLOG(5, "ZSTD_compressBlock_targetCBlockSize (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u, srcSize=%zu)",
                 (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, (unsigned)zc->blockState.matchState.nextToUpdate, srcSize);
-    FORWARD_IF_ERROR(bss);
+    FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");
 
     cSize = ZSTD_compressBlock_targetCBlockSize_body(zc, dst, dstCapacity, src, srcSize, bss, lastBlock);
-    FORWARD_IF_ERROR(cSize);
+    FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize_body failed");
 
     if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
         zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
@@ -2584,18 +2616,18 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
         {   size_t cSize;
             if (ZSTD_useTargetCBlockSize(&cctx->appliedParams)) {
                 cSize = ZSTD_compressBlock_targetCBlockSize(cctx, op, dstCapacity, ip, blockSize, lastBlock);
-                FORWARD_IF_ERROR(cSize);
+                FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize failed");
                 assert(cSize > 0);
                 assert(cSize <= blockSize + ZSTD_blockHeaderSize);
             } else {
                 cSize = ZSTD_compressBlock_internal(cctx,
                                         op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize,
                                         ip, blockSize, 1 /* frame */);
-                FORWARD_IF_ERROR(cSize);
+                FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_internal failed");
 
                 if (cSize == 0) {  /* block is not compressible */
                     cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
-                    FORWARD_IF_ERROR(cSize);
+                    FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
                 } else {
                     U32 const cBlockHeader = cSize == 1 ?
                         lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) :
@@ -2637,7 +2669,8 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
     size_t pos=0;
 
     assert(!(params->fParams.contentSizeFlag && pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN));
-    RETURN_ERROR_IF(dstCapacity < ZSTD_FRAMEHEADERSIZE_MAX, dstSize_tooSmall);
+    RETURN_ERROR_IF(dstCapacity < ZSTD_FRAMEHEADERSIZE_MAX, dstSize_tooSmall,
+                    "dst buf is too small to fit worst-case frame header size.");
     DEBUGLOG(4, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u",
                 !params->fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode);
 
@@ -2673,7 +2706,8 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
  */
 size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity)
 {
-    RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall);
+    RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall,
+                    "dst buf is too small to write frame trailer empty block.");
     {   U32 const cBlockHeader24 = 1 /*lastBlock*/ + (((U32)bt_raw)<<1);  /* 0 size */
         MEM_writeLE24(dst, cBlockHeader24);
         return ZSTD_blockHeaderSize;
@@ -2682,9 +2716,11 @@ size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity)
 
 size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq)
 {
-    RETURN_ERROR_IF(cctx->stage != ZSTDcs_init, stage_wrong);
+    RETURN_ERROR_IF(cctx->stage != ZSTDcs_init, stage_wrong,
+                    "wrong cctx stage");
     RETURN_ERROR_IF(cctx->appliedParams.ldmParams.enableLdm,
-                    parameter_unsupported);
+                    parameter_unsupported,
+                    "incompatible with ldm");
     cctx->externSeqStore.seq = seq;
     cctx->externSeqStore.size = nbSeq;
     cctx->externSeqStore.capacity = nbSeq;
@@ -2709,7 +2745,7 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
     if (frame && (cctx->stage==ZSTDcs_init)) {
         fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams,
                                        cctx->pledgedSrcSizePlusOne-1, cctx->dictID);
-        FORWARD_IF_ERROR(fhSize);
+        FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed");
         assert(fhSize <= dstCapacity);
         dstCapacity -= fhSize;
         dst = (char*)dst + fhSize;
@@ -2736,7 +2772,7 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
     {   size_t const cSize = frame ?
                              ZSTD_compress_frameChunk (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :
                              ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize, 0 /* frame */);
-        FORWARD_IF_ERROR(cSize);
+        FORWARD_IF_ERROR(cSize, "%s", frame ? "ZSTD_compress_frameChunk failed" : "ZSTD_compressBlock_internal failed");
         cctx->consumedSrcSize += srcSize;
         cctx->producedCSize += (cSize + fhSize);
         assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0));
@@ -2773,7 +2809,7 @@ size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const
 {
     DEBUGLOG(5, "ZSTD_compressBlock: srcSize = %u", (unsigned)srcSize);
     { size_t const blockSizeMax = ZSTD_getBlockSize(cctx);
-      RETURN_ERROR_IF(srcSize > blockSizeMax, srcSize_wrong); }
+      RETURN_ERROR_IF(srcSize > blockSizeMax, srcSize_wrong, "input is larger than a block"); }
 
     return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 0 /* frame mode */, 0 /* last chunk */);
 }
@@ -2782,6 +2818,7 @@ size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const
  *  @return : 0, or an error code
  */
 static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
+                                         ldmState_t* ls,
                                          ZSTD_cwksp* ws,
                                          ZSTD_CCtx_params const* params,
                                          const void* src, size_t srcSize,
@@ -2793,6 +2830,11 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
     ZSTD_window_update(&ms->window, src, srcSize);
     ms->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ms->window.base);
 
+    if (params->ldmParams.enableLdm && ls != NULL) {
+        ZSTD_window_update(&ls->window, src, srcSize);
+        ls->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ls->window.base);
+    }
+
     /* Assert that we the ms params match the params we're being given */
     ZSTD_assertEqualCParams(params->cParams, ms->cParams);
 
@@ -2805,6 +2847,9 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
 
         ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, ichunk);
 
+        if (params->ldmParams.enableLdm && ls != NULL)
+            ZSTD_ldm_fillHashTable(ls, (const BYTE*)src, (const BYTE*)src + srcSize, &params->ldmParams);
+
         switch(params->cParams.strategy)
         {
         case ZSTD_fast:
@@ -2847,9 +2892,9 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
    NOTE: This behavior is not standard and could be improved in the future. */
 static size_t ZSTD_checkDictNCount(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue) {
     U32 s;
-    RETURN_ERROR_IF(dictMaxSymbolValue < maxSymbolValue, dictionary_corrupted);
+    RETURN_ERROR_IF(dictMaxSymbolValue < maxSymbolValue, dictionary_corrupted, "dict fse tables don't have all symbols");
     for (s = 0; s <= maxSymbolValue; ++s) {
-        RETURN_ERROR_IF(normalizedCounter[s] == 0, dictionary_corrupted);
+        RETURN_ERROR_IF(normalizedCounter[s] == 0, dictionary_corrupted, "dict fse tables don't have all symbols");
     }
     return 0;
 }
@@ -2873,56 +2918,56 @@ size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace,
         if (!hasZeroWeights)
             bs->entropy.huf.repeatMode = HUF_repeat_valid;
 
-        RETURN_ERROR_IF(HUF_isError(hufHeaderSize), dictionary_corrupted);
-        RETURN_ERROR_IF(maxSymbolValue < 255, dictionary_corrupted);
+        RETURN_ERROR_IF(HUF_isError(hufHeaderSize), dictionary_corrupted, "");
+        RETURN_ERROR_IF(maxSymbolValue < 255, dictionary_corrupted, "");
         dictPtr += hufHeaderSize;
     }
 
     {   unsigned offcodeLog;
         size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
-        RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted);
-        RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted);
+        RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "");
+        RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "");
         /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */
         /* fill all offset symbols to avoid garbage at end of table */
         RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
                 bs->entropy.fse.offcodeCTable,
                 offcodeNCount, MaxOff, offcodeLog,
                 workspace, HUF_WORKSPACE_SIZE)),
-            dictionary_corrupted);
+            dictionary_corrupted, "");
         dictPtr += offcodeHeaderSize;
     }
 
     {   short matchlengthNCount[MaxML+1];
         unsigned matchlengthMaxValue = MaxML, matchlengthLog;
         size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);
-        RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted);
-        RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted);
+        RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "");
+        RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "");
         /* Every match length code must have non-zero probability */
-        FORWARD_IF_ERROR( ZSTD_checkDictNCount(matchlengthNCount, matchlengthMaxValue, MaxML));
+        FORWARD_IF_ERROR( ZSTD_checkDictNCount(matchlengthNCount, matchlengthMaxValue, MaxML), "");
         RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
                 bs->entropy.fse.matchlengthCTable,
                 matchlengthNCount, matchlengthMaxValue, matchlengthLog,
                 workspace, HUF_WORKSPACE_SIZE)),
-            dictionary_corrupted);
+            dictionary_corrupted, "");
         dictPtr += matchlengthHeaderSize;
     }
 
     {   short litlengthNCount[MaxLL+1];
         unsigned litlengthMaxValue = MaxLL, litlengthLog;
         size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);
-        RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted);
-        RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted);
+        RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "");
+        RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "");
         /* Every literal length code must have non-zero probability */
-        FORWARD_IF_ERROR( ZSTD_checkDictNCount(litlengthNCount, litlengthMaxValue, MaxLL));
+        FORWARD_IF_ERROR( ZSTD_checkDictNCount(litlengthNCount, litlengthMaxValue, MaxLL), "");
         RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
                 bs->entropy.fse.litlengthCTable,
                 litlengthNCount, litlengthMaxValue, litlengthLog,
                 workspace, HUF_WORKSPACE_SIZE)),
-            dictionary_corrupted);
+            dictionary_corrupted, "");
         dictPtr += litlengthHeaderSize;
     }
 
-    RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted);
+    RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "");
     bs->rep[0] = MEM_readLE32(dictPtr+0);
     bs->rep[1] = MEM_readLE32(dictPtr+4);
     bs->rep[2] = MEM_readLE32(dictPtr+8);
@@ -2961,7 +3006,7 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
 
     dictID = params->fParams.noDictIDFlag ? 0 :  MEM_readLE32(dictPtr + 4 /* skip magic number */ );
     eSize = ZSTD_loadCEntropy(bs, workspace, offcodeNCount, &offcodeMaxValue, dict, dictSize);
-    FORWARD_IF_ERROR(eSize);
+    FORWARD_IF_ERROR(eSize, "ZSTD_loadCEntropy failed");
     dictPtr += eSize;
 
     {   size_t const dictContentSize = (size_t)(dictEnd - dictPtr);
@@ -2971,19 +3016,19 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
             offcodeMax = ZSTD_highbit32(maxOffset); /* Calculate minimum offset code required to represent maxOffset */
         }
         /* All offset values <= dictContentSize + 128 KB must be representable */
-        FORWARD_IF_ERROR(ZSTD_checkDictNCount(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff)));
+        FORWARD_IF_ERROR(ZSTD_checkDictNCount(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff)), "");
         /* All repCodes must be <= dictContentSize and != 0*/
         {   U32 u;
             for (u=0; u<3; u++) {
-                RETURN_ERROR_IF(bs->rep[u] == 0, dictionary_corrupted);
-                RETURN_ERROR_IF(bs->rep[u] > dictContentSize, dictionary_corrupted);
+                RETURN_ERROR_IF(bs->rep[u] == 0, dictionary_corrupted, "");
+                RETURN_ERROR_IF(bs->rep[u] > dictContentSize, dictionary_corrupted, "");
         }   }
 
         bs->entropy.fse.offcode_repeatMode = FSE_repeat_valid;
         bs->entropy.fse.matchlength_repeatMode = FSE_repeat_valid;
         bs->entropy.fse.litlength_repeatMode = FSE_repeat_valid;
         FORWARD_IF_ERROR(ZSTD_loadDictionaryContent(
-            ms, ws, params, dictPtr, dictContentSize, dtlm));
+            ms, NULL, ws, params, dictPtr, dictContentSize, dtlm), "");
         return dictID;
     }
 }
@@ -2993,6 +3038,7 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
 static size_t
 ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs,
                                ZSTD_matchState_t* ms,
+                               ldmState_t* ls,
                                ZSTD_cwksp* ws,
                          const ZSTD_CCtx_params* params,
                          const void* dict, size_t dictSize,
@@ -3002,7 +3048,7 @@ ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs,
 {
     DEBUGLOG(4, "ZSTD_compress_insertDictionary (dictSize=%u)", (U32)dictSize);
     if ((dict==NULL) || (dictSize<8)) {
-        RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong);
+        RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "");
         return 0;
     }
 
@@ -3010,15 +3056,15 @@ ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs,
 
     /* dict restricted modes */
     if (dictContentType == ZSTD_dct_rawContent)
-        return ZSTD_loadDictionaryContent(ms, ws, params, dict, dictSize, dtlm);
+        return ZSTD_loadDictionaryContent(ms, ls, ws, params, dict, dictSize, dtlm);
 
     if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) {
         if (dictContentType == ZSTD_dct_auto) {
             DEBUGLOG(4, "raw content dictionary detected");
             return ZSTD_loadDictionaryContent(
-                ms, ws, params, dict, dictSize, dtlm);
+                ms, ls, ws, params, dict, dictSize, dtlm);
         }
-        RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong);
+        RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "");
         assert(0);   /* impossible */
     }
 
@@ -3055,18 +3101,18 @@ static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx,
     }
 
     FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, *params, pledgedSrcSize,
-                                     ZSTDcrp_makeClean, zbuff) );
+                                     ZSTDcrp_makeClean, zbuff) , "");
     {   size_t const dictID = cdict ?
                 ZSTD_compress_insertDictionary(
                         cctx->blockState.prevCBlock, &cctx->blockState.matchState,
-                        &cctx->workspace, &cctx->appliedParams, cdict->dictContent,
+                        &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, cdict->dictContent,
                         cdict->dictContentSize, dictContentType, dtlm,
                         cctx->entropyWorkspace)
               : ZSTD_compress_insertDictionary(
                         cctx->blockState.prevCBlock, &cctx->blockState.matchState,
-                        &cctx->workspace, &cctx->appliedParams, dict, dictSize,
+                        &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, dict, dictSize,
                         dictContentType, dtlm, cctx->entropyWorkspace);
-        FORWARD_IF_ERROR(dictID);
+        FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed");
         assert(dictID <= UINT_MAX);
         cctx->dictID = (U32)dictID;
     }
@@ -3083,7 +3129,7 @@ size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
 {
     DEBUGLOG(4, "ZSTD_compressBegin_advanced_internal: wlog=%u", params->cParams.windowLog);
     /* compression parameters verification and optimization */
-    FORWARD_IF_ERROR( ZSTD_checkCParams(params->cParams) );
+    FORWARD_IF_ERROR( ZSTD_checkCParams(params->cParams) , "");
     return ZSTD_compressBegin_internal(cctx,
                                        dict, dictSize, dictContentType, dtlm,
                                        cdict,
@@ -3136,7 +3182,7 @@ static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
     /* special case : empty frame */
     if (cctx->stage == ZSTDcs_init) {
         fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams, 0, 0);
-        FORWARD_IF_ERROR(fhSize);
+        FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed");
         dstCapacity -= fhSize;
         op += fhSize;
         cctx->stage = ZSTDcs_ongoing;
@@ -3145,7 +3191,7 @@ static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
     if (cctx->stage != ZSTDcs_ending) {
         /* write one last empty block, make it the "last" block */
         U32 const cBlockHeader24 = 1 /* last block */ + (((U32)bt_raw)<<1) + 0;
-        RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall);
+        RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for epilogue");
         MEM_writeLE32(op, cBlockHeader24);
         op += ZSTD_blockHeaderSize;
         dstCapacity -= ZSTD_blockHeaderSize;
@@ -3153,7 +3199,7 @@ static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
 
     if (cctx->appliedParams.fParams.checksumFlag) {
         U32 const checksum = (U32) XXH64_digest(&cctx->xxhState);
-        RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall);
+        RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum");
         DEBUGLOG(4, "ZSTD_writeEpilogue: write checksum : %08X", (unsigned)checksum);
         MEM_writeLE32(op, checksum);
         op += 4;
@@ -3171,9 +3217,9 @@ size_t ZSTD_compressEnd (ZSTD_CCtx* cctx,
     size_t const cSize = ZSTD_compressContinue_internal(cctx,
                                 dst, dstCapacity, src, srcSize,
                                 1 /* frame mode */, 1 /* last chunk */);
-    FORWARD_IF_ERROR(cSize);
+    FORWARD_IF_ERROR(cSize, "ZSTD_compressContinue_internal failed");
     endResult = ZSTD_writeEpilogue(cctx, (char*)dst + cSize, dstCapacity-cSize);
-    FORWARD_IF_ERROR(endResult);
+    FORWARD_IF_ERROR(endResult, "ZSTD_writeEpilogue failed");
     assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0));
     if (cctx->pledgedSrcSizePlusOne != 0) {  /* control src size */
         ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1);
@@ -3212,7 +3258,7 @@ size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx,
                                ZSTD_parameters params)
 {
     DEBUGLOG(4, "ZSTD_compress_advanced");
-    FORWARD_IF_ERROR(ZSTD_checkCParams(params.cParams));
+    FORWARD_IF_ERROR(ZSTD_checkCParams(params.cParams), "");
     return ZSTD_compress_internal(cctx,
                                   dst, dstCapacity,
                                   src, srcSize,
@@ -3231,7 +3277,7 @@ size_t ZSTD_compress_advanced_internal(
     DEBUGLOG(4, "ZSTD_compress_advanced_internal (srcSize:%u)", (unsigned)srcSize);
     FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,
                          dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,
-                         params, srcSize, ZSTDb_not_buffered) );
+                         params, srcSize, ZSTDb_not_buffered) , "");
     return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
 }
 
@@ -3243,6 +3289,7 @@ size_t ZSTD_compress_usingDict(ZSTD_CCtx* cctx,
 {
     ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, srcSize, dict ? dictSize : 0);
     ZSTD_CCtx_params cctxParams = ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, &params);
+    DEBUGLOG(4, "ZSTD_compress_usingDict (srcSize=%u)", (unsigned)srcSize);
     assert(params.fParams.contentSizeFlag == 1);
     return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctxParams);
 }
@@ -3315,7 +3362,7 @@ static size_t ZSTD_initCDict_internal(
         cdict->dictContent = dictBuffer;
     } else {
          void *internalBuffer = ZSTD_cwksp_reserve_object(&cdict->workspace, ZSTD_cwksp_align(dictSize, sizeof(void*)));
-        RETURN_ERROR_IF(!internalBuffer, memory_allocation);
+        RETURN_ERROR_IF(!internalBuffer, memory_allocation, "NULL pointer!");
         cdict->dictContent = internalBuffer;
         memcpy(internalBuffer, dictBuffer, dictSize);
     }
@@ -3332,7 +3379,7 @@ static size_t ZSTD_initCDict_internal(
         &cParams,
         ZSTDcrp_makeClean,
         ZSTDirp_reset,
-        ZSTD_resetTarget_CDict));
+        ZSTD_resetTarget_CDict), "");
     /* (Maybe) load the dictionary
      * Skips loading the dictionary if it is < 8 bytes.
      */
@@ -3342,10 +3389,10 @@ static size_t ZSTD_initCDict_internal(
         params.fParams.contentSizeFlag = 1;
         params.cParams = cParams;
         {   size_t const dictID = ZSTD_compress_insertDictionary(
-                    &cdict->cBlockState, &cdict->matchState, &cdict->workspace,
+                    &cdict->cBlockState, &cdict->matchState, NULL, &cdict->workspace,
                     &params, cdict->dictContent, cdict->dictContentSize,
                     dictContentType, ZSTD_dtlm_full, cdict->entropyWorkspace);
-            FORWARD_IF_ERROR(dictID);
+            FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed");
             assert(dictID <= (size_t)(U32)-1);
             cdict->dictID = (U32)dictID;
         }
@@ -3493,7 +3540,7 @@ size_t ZSTD_compressBegin_usingCDict_advanced(
     ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize)
 {
     DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_advanced");
-    RETURN_ERROR_IF(cdict==NULL, dictionary_wrong);
+    RETURN_ERROR_IF(cdict==NULL, dictionary_wrong, "NULL pointer!");
     {   ZSTD_CCtx_params params = cctx->requestedParams;
         params.cParams = ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF
                         || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER
@@ -3537,7 +3584,7 @@ size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx,
                                 const void* src, size_t srcSize,
                                 const ZSTD_CDict* cdict, ZSTD_frameParameters fParams)
 {
-    FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, srcSize));   /* will check if cdict != NULL */
+    FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, srcSize), "");   /* will check if cdict != NULL */
     return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
 }
 
@@ -3609,7 +3656,7 @@ static size_t ZSTD_resetCStream_internal(ZSTD_CStream* cctx,
                                          dict, dictSize, dictContentType, ZSTD_dtlm_fast,
                                          cdict,
                                          &params, pledgedSrcSize,
-                                         ZSTDb_buffered) );
+                                         ZSTDb_buffered) , "");
 
     cctx->inToCompress = 0;
     cctx->inBuffPos = 0;
@@ -3631,8 +3678,8 @@ size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pss)
      */
     U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
     DEBUGLOG(4, "ZSTD_resetCStream: pledgedSrcSize = %u", (unsigned)pledgedSrcSize);
-    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
     return 0;
 }
 
@@ -3646,16 +3693,16 @@ size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
                     unsigned long long pledgedSrcSize)
 {
     DEBUGLOG(4, "ZSTD_initCStream_internal");
-    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
     assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
     zcs->requestedParams = *params;
     assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
     if (dict) {
-        FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) );
+        FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
     } else {
         /* Dictionary is cleared if !cdict */
-        FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) );
+        FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
     }
     return 0;
 }
@@ -3668,10 +3715,10 @@ size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs,
                                             unsigned long long pledgedSrcSize)
 {
     DEBUGLOG(4, "ZSTD_initCStream_usingCDict_advanced");
-    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
     zcs->requestedParams.fParams = fParams;
-    FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
     return 0;
 }
 
@@ -3679,8 +3726,8 @@ size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs,
 size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict)
 {
     DEBUGLOG(4, "ZSTD_initCStream_usingCDict");
-    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
     return 0;
 }
 
@@ -3699,20 +3746,20 @@ size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
      */
     U64 const pledgedSrcSize = (pss==0 && params.fParams.contentSizeFlag==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
     DEBUGLOG(4, "ZSTD_initCStream_advanced");
-    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) );
-    FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
+    FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "");
     zcs->requestedParams = ZSTD_assignParamsToCCtxParams(&zcs->requestedParams, &params);
-    FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
     return 0;
 }
 
 size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel)
 {
     DEBUGLOG(4, "ZSTD_initCStream_usingDict");
-    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
     return 0;
 }
 
@@ -3724,19 +3771,19 @@ size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigne
      */
     U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
     DEBUGLOG(4, "ZSTD_initCStream_srcSize");
-    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
     return 0;
 }
 
 size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel)
 {
     DEBUGLOG(4, "ZSTD_initCStream");
-    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) );
-    FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) );
+    FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "");
+    FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
     return 0;
 }
 
@@ -3749,14 +3796,6 @@ static size_t ZSTD_nextInputSizeHint(const ZSTD_CCtx* cctx)
     return hintInSize;
 }
 
-static size_t ZSTD_limitCopy(void* dst, size_t dstCapacity,
-                       const void* src, size_t srcSize)
-{
-    size_t const length = MIN(dstCapacity, srcSize);
-    if (length) memcpy(dst, src, length);
-    return length;
-}
-
 /** ZSTD_compressStream_generic():
  *  internal function for all *compressStream*() variants
  *  non-static, because can be called from zstdmt_compress.c
@@ -3797,7 +3836,7 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
                 size_t const cSize = ZSTD_compressEnd(zcs,
                                                 op, oend-op, ip, iend-ip);
                 DEBUGLOG(4, "ZSTD_compressEnd : cSize=%u", (unsigned)cSize);
-                FORWARD_IF_ERROR(cSize);
+                FORWARD_IF_ERROR(cSize, "ZSTD_compressEnd failed");
                 ip = iend;
                 op += cSize;
                 zcs->frameEnded = 1;
@@ -3839,7 +3878,7 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
                                     zcs->inBuff + zcs->inToCompress, iSize) :
                         ZSTD_compressContinue(zcs, cDst, oSize,
                                     zcs->inBuff + zcs->inToCompress, iSize);
-                FORWARD_IF_ERROR(cSize);
+                FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed");
                 zcs->frameEnded = lastBlock;
                 /* prepare next block */
                 zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize;
@@ -3916,7 +3955,7 @@ static size_t ZSTD_nextInputSizeHint_MTorST(const ZSTD_CCtx* cctx)
 
 size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
 {
-    FORWARD_IF_ERROR( ZSTD_compressStream2(zcs, output, input, ZSTD_e_continue) );
+    FORWARD_IF_ERROR( ZSTD_compressStream2(zcs, output, input, ZSTD_e_continue) , "");
     return ZSTD_nextInputSizeHint_MTorST(zcs);
 }
 
@@ -3928,15 +3967,15 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
 {
     DEBUGLOG(5, "ZSTD_compressStream2, endOp=%u ", (unsigned)endOp);
     /* check conditions */
-    RETURN_ERROR_IF(output->pos > output->size, GENERIC);
-    RETURN_ERROR_IF(input->pos  > input->size, GENERIC);
+    RETURN_ERROR_IF(output->pos > output->size, GENERIC, "invalid buffer");
+    RETURN_ERROR_IF(input->pos  > input->size, GENERIC, "invalid buffer");
     assert(cctx!=NULL);
 
     /* transparent initialization stage */
     if (cctx->streamStage == zcss_init) {
         ZSTD_CCtx_params params = cctx->requestedParams;
         ZSTD_prefixDict const prefixDict = cctx->prefixDict;
-        FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) ); /* Init the local dict if present. */
+        FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , ""); /* Init the local dict if present. */
         memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict));   /* single usage */
         assert(prefixDict.dict==NULL || cctx->cdict==NULL);    /* only one can be set */
         DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage");
@@ -3955,14 +3994,14 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
                 DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u",
                             params.nbWorkers);
                 cctx->mtctx = ZSTDMT_createCCtx_advanced((U32)params.nbWorkers, cctx->customMem);
-                RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation);
+                RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation, "NULL pointer!");
             }
             /* mt compression */
             DEBUGLOG(4, "call ZSTDMT_initCStream_internal as nbWorkers=%u", params.nbWorkers);
             FORWARD_IF_ERROR( ZSTDMT_initCStream_internal(
                         cctx->mtctx,
-                        prefixDict.dict, prefixDict.dictSize, ZSTD_dct_rawContent,
-                        cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) );
+                        prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType,
+                        cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) , "");
             cctx->streamStage = zcss_load;
             cctx->appliedParams.nbWorkers = params.nbWorkers;
         } else
@@ -3970,7 +4009,7 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
         {   FORWARD_IF_ERROR( ZSTD_resetCStream_internal(cctx,
                             prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType,
                             cctx->cdict,
-                            params, cctx->pledgedSrcSizePlusOne-1) );
+                            params, cctx->pledgedSrcSizePlusOne-1) , "");
             assert(cctx->streamStage == zcss_load);
             assert(cctx->appliedParams.nbWorkers == 0);
     }   }
@@ -3992,7 +4031,7 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
               || (endOp == ZSTD_e_end && flushMin == 0) ) { /* compression completed */
                 ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
             }
-            FORWARD_IF_ERROR(flushMin);
+            FORWARD_IF_ERROR(flushMin, "ZSTDMT_compressStream_generic failed");
         } while (forceMaxProgress && flushMin != 0 && output->pos < output->size);
         DEBUGLOG(5, "completed ZSTD_compressStream2 delegating to ZSTDMT_compressStream_generic");
         /* Either we don't require maximum forward progress, we've finished the
@@ -4002,7 +4041,7 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
         return flushMin;
     }
 #endif
-    FORWARD_IF_ERROR( ZSTD_compressStream_generic(cctx, output, input, endOp) );
+    FORWARD_IF_ERROR( ZSTD_compressStream_generic(cctx, output, input, endOp) , "");
     DEBUGLOG(5, "completed ZSTD_compressStream2");
     return cctx->outBuffContentSize - cctx->outBuffFlushedSize; /* remaining to flush */
 }
@@ -4026,6 +4065,7 @@ size_t ZSTD_compress2(ZSTD_CCtx* cctx,
                       void* dst, size_t dstCapacity,
                       const void* src, size_t srcSize)
 {
+    DEBUGLOG(4, "ZSTD_compress2 (srcSize=%u)", (unsigned)srcSize);
     ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
     {   size_t oPos = 0;
         size_t iPos = 0;
@@ -4033,10 +4073,10 @@ size_t ZSTD_compress2(ZSTD_CCtx* cctx,
                                         dst, dstCapacity, &oPos,
                                         src, srcSize, &iPos,
                                         ZSTD_e_end);
-        FORWARD_IF_ERROR(result);
+        FORWARD_IF_ERROR(result, "ZSTD_compressStream2_simpleArgs failed");
         if (result != 0) {  /* compression not completed, due to lack of output space */
             assert(oPos == dstCapacity);
-            RETURN_ERROR(dstSize_tooSmall);
+            RETURN_ERROR(dstSize_tooSmall, "");
         }
         assert(iPos == srcSize);   /* all input is expected consumed */
         return oPos;
@@ -4058,7 +4098,7 @@ size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
 {
     ZSTD_inBuffer input = { NULL, 0, 0 };
     size_t const remainingToFlush = ZSTD_compressStream2(zcs, output, &input, ZSTD_e_end);
-    FORWARD_IF_ERROR( remainingToFlush );
+    FORWARD_IF_ERROR( remainingToFlush , "ZSTD_compressStream2 failed");
     if (zcs->appliedParams.nbWorkers > 0) return remainingToFlush;   /* minimal estimation */
     /* single thread mode : attempt to calculate remaining to flush more precisely */
     {   size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE;
diff --git a/zstd/zstd_compress_internal.h b/zstd/zstd_compress_internal.h
index 243f0e07..c500687c 100644
--- a/zstd/zstd_compress_internal.h
+++ b/zstd/zstd_compress_internal.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -166,6 +166,7 @@ typedef struct {
 typedef struct {
     ZSTD_window_t window;   /* State for the window round buffer management */
     ldmEntry_t* hashTable;
+    U32 loadedDictEnd;
     BYTE* bucketOffsets;    /* Next position in bucket to insert entry */
     U64 hashPower;          /* Used to compute the rolling hash.
                              * Depends on ldmParams.minMatchLength */
@@ -325,6 +326,31 @@ MEM_STATIC U32 ZSTD_MLcode(U32 mlBase)
     return (mlBase > 127) ? ZSTD_highbit32(mlBase) + ML_deltaCode : ML_Code[mlBase];
 }
 
+typedef struct repcodes_s {
+    U32 rep[3];
+} repcodes_t;
+
+MEM_STATIC repcodes_t ZSTD_updateRep(U32 const rep[3], U32 const offset, U32 const ll0)
+{
+    repcodes_t newReps;
+    if (offset >= ZSTD_REP_NUM) {  /* full offset */
+        newReps.rep[2] = rep[1];
+        newReps.rep[1] = rep[0];
+        newReps.rep[0] = offset - ZSTD_REP_MOVE;
+    } else {   /* repcode */
+        U32 const repCode = offset + ll0;
+        if (repCode > 0) {  /* note : if repCode==0, no change */
+            U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
+            newReps.rep[2] = (repCode >= 2) ? rep[1] : rep[2];
+            newReps.rep[1] = rep[0];
+            newReps.rep[0] = currentOffset;
+        } else {   /* repCode == 0 */
+            memcpy(&newReps, rep, sizeof(newReps));
+        }
+    }
+    return newReps;
+}
+
 /* ZSTD_cParam_withinBounds:
  * @return 1 if value is within cParam bounds,
  * 0 otherwise */
@@ -344,12 +370,22 @@ MEM_STATIC size_t ZSTD_noCompressBlock (void* dst, size_t dstCapacity, const voi
 {
     U32 const cBlockHeader24 = lastBlock + (((U32)bt_raw)<<1) + (U32)(srcSize << 3);
     RETURN_ERROR_IF(srcSize + ZSTD_blockHeaderSize > dstCapacity,
-                    dstSize_tooSmall);
+                    dstSize_tooSmall, "dst buf too small for uncompressed block");
     MEM_writeLE24(dst, cBlockHeader24);
     memcpy((BYTE*)dst + ZSTD_blockHeaderSize, src, srcSize);
     return ZSTD_blockHeaderSize + srcSize;
 }
 
+MEM_STATIC size_t ZSTD_rleCompressBlock (void* dst, size_t dstCapacity, BYTE src, size_t srcSize, U32 lastBlock)
+{
+    BYTE* const op = (BYTE*)dst;
+    U32 const cBlockHeader = lastBlock + (((U32)bt_rle)<<1) + (U32)(srcSize << 3);
+    RETURN_ERROR_IF(dstCapacity < 4, dstSize_tooSmall, "");
+    MEM_writeLE24(op, cBlockHeader);
+    op[3] = src;
+    return 4;
+}
+
 
 /* ZSTD_minGain() :
  * minimum compression required
@@ -363,6 +399,21 @@ MEM_STATIC size_t ZSTD_minGain(size_t srcSize, ZSTD_strategy strat)
     return (srcSize >> minlog) + 2;
 }
 
+MEM_STATIC int ZSTD_disableLiteralsCompression(const ZSTD_CCtx_params* cctxParams)
+{
+    switch (cctxParams->literalCompressionMode) {
+    case ZSTD_lcm_huffman:
+        return 0;
+    case ZSTD_lcm_uncompressed:
+        return 1;
+    default:
+        assert(0 /* impossible: pre-validated */);
+        /* fall-through */
+    case ZSTD_lcm_auto:
+        return (cctxParams->cParams.strategy == ZSTD_fast) && (cctxParams->cParams.targetLength > 0);
+    }
+}
+
 /*! ZSTD_safecopyLiterals() :
  *  memcpy() function that won't read beyond more than WILDCOPY_OVERLENGTH bytes past ilimit_w.
  *  Only called when the sequence ends past ilimit_w, so it only needs to be optimized for single
@@ -448,8 +499,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
         if (MEM_64bits()) {
 #       if defined(_MSC_VER) && defined(_WIN64)
             unsigned long r = 0;
-            _BitScanForward64( &r, (U64)val );
-            return (unsigned)(r>>3);
+            return _BitScanForward64( &r, (U64)val ) ? (unsigned)(r >> 3) : 0;
 #       elif defined(__GNUC__) && (__GNUC__ >= 4)
             return (__builtin_ctzll((U64)val) >> 3);
 #       else
@@ -466,8 +516,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
         } else { /* 32 bits */
 #       if defined(_MSC_VER)
             unsigned long r=0;
-            _BitScanForward( &r, (U32)val );
-            return (unsigned)(r>>3);
+            return _BitScanForward( &r, (U32)val ) ? (unsigned)(r >> 3) : 0;
 #       elif defined(__GNUC__) && (__GNUC__ >= 3)
             return (__builtin_ctz((U32)val) >> 3);
 #       else
@@ -482,8 +531,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
         if (MEM_64bits()) {
 #       if defined(_MSC_VER) && defined(_WIN64)
             unsigned long r = 0;
-            _BitScanReverse64( &r, val );
-            return (unsigned)(r>>3);
+            return _BitScanReverse64( &r, val ) ? (unsigned)(r >> 3) : 0;
 #       elif defined(__GNUC__) && (__GNUC__ >= 4)
             return (__builtin_clzll(val) >> 3);
 #       else
@@ -497,8 +545,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
         } else { /* 32 bits */
 #       if defined(_MSC_VER)
             unsigned long r = 0;
-            _BitScanReverse( &r, (unsigned long)val );
-            return (unsigned)(r>>3);
+            return _BitScanReverse( &r, (unsigned long)val ) ? (unsigned)(r >> 3) : 0;
 #       elif defined(__GNUC__) && (__GNUC__ >= 3)
             return (__builtin_clz((U32)val) >> 3);
 #       else
@@ -907,7 +954,7 @@ MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
         window->dictLimit = (U32)distanceFromBase;
         window->dictBase = window->base;
         window->base = ip - distanceFromBase;
-        // ms->nextToUpdate = window->dictLimit;
+        /* ms->nextToUpdate = window->dictLimit; */
         if (window->dictLimit - window->lowLimit < HASH_READ_SIZE) window->lowLimit = window->dictLimit;   /* too small extDict */
         contiguous = 0;
     }
@@ -923,6 +970,9 @@ MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
     return contiguous;
 }
 
+/**
+ * Returns the lowest allowed match index. It may either be in the ext-dict or the prefix.
+ */
 MEM_STATIC U32 ZSTD_getLowestMatchIndex(const ZSTD_matchState_t* ms, U32 current, unsigned windowLog)
 {
     U32    const maxDistance = 1U << windowLog;
@@ -933,6 +983,19 @@ MEM_STATIC U32 ZSTD_getLowestMatchIndex(const ZSTD_matchState_t* ms, U32 current
     return matchLowest;
 }
 
+/**
+ * Returns the lowest allowed match index in the prefix.
+ */
+MEM_STATIC U32 ZSTD_getLowestPrefixIndex(const ZSTD_matchState_t* ms, U32 current, unsigned windowLog)
+{
+    U32    const maxDistance = 1U << windowLog;
+    U32    const lowestValid = ms->window.dictLimit;
+    U32    const withinWindow = (current - lowestValid > maxDistance) ? current - maxDistance : lowestValid;
+    U32    const isDictionary = (ms->loadedDictEnd != 0);
+    U32    const matchLowest = isDictionary ? lowestValid : withinWindow;
+    return matchLowest;
+}
+
 
 
 /* debug functions */
@@ -1055,5 +1118,8 @@ size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity);
  */
 size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq);
 
+/** ZSTD_cycleLog() :
+ *  condition for correct operation : hashLog > 1 */
+U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat);
 
 #endif /* ZSTD_COMPRESS_H */
diff --git a/zstd/zstd_compress_literals.c b/zstd/zstd_compress_literals.c
index df3f269d..17e7168d 100644
--- a/zstd/zstd_compress_literals.c
+++ b/zstd/zstd_compress_literals.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -18,7 +18,7 @@ size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src,
     BYTE* const ostart = (BYTE* const)dst;
     U32   const flSize = 1 + (srcSize>31) + (srcSize>4095);
 
-    RETURN_ERROR_IF(srcSize + flSize > dstCapacity, dstSize_tooSmall);
+    RETURN_ERROR_IF(srcSize + flSize > dstCapacity, dstSize_tooSmall, "");
 
     switch(flSize)
     {
@@ -36,6 +36,7 @@ size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src,
     }
 
     memcpy(ostart + flSize, src, srcSize);
+    DEBUGLOG(5, "Raw literals: %u -> %u", (U32)srcSize, (U32)(srcSize + flSize));
     return srcSize + flSize;
 }
 
@@ -62,6 +63,7 @@ size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, const void*
     }
 
     ostart[flSize] = *(const BYTE*)src;
+    DEBUGLOG(5, "RLE literals: %u -> %u", (U32)srcSize, (U32)flSize + 1);
     return flSize+1;
 }
 
@@ -80,8 +82,8 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
     symbolEncodingType_e hType = set_compressed;
     size_t cLitSize;
 
-    DEBUGLOG(5,"ZSTD_compressLiterals (disableLiteralCompression=%i)",
-                disableLiteralCompression);
+    DEBUGLOG(5,"ZSTD_compressLiterals (disableLiteralCompression=%i srcSize=%u)",
+                disableLiteralCompression, (U32)srcSize);
 
     /* Prepare nextEntropy assuming reusing the existing table */
     memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
@@ -110,6 +112,7 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
                 (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2);
         if (repeat != HUF_repeat_none) {
             /* reused the existing table */
+            DEBUGLOG(5, "Reusing previous huffman table");
             hType = set_repeat;
         }
     }
@@ -150,5 +153,6 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
     default:  /* not possible : lhSize is {3,4,5} */
         assert(0);
     }
+    DEBUGLOG(5, "Compressed literals: %u -> %u", (U32)srcSize, (U32)(lhSize+cLitSize));
     return lhSize+cLitSize;
 }
diff --git a/zstd/zstd_compress_literals.h b/zstd/zstd_compress_literals.h
index 97273d7c..8b087057 100644
--- a/zstd/zstd_compress_literals.h
+++ b/zstd/zstd_compress_literals.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstd_compress_sequences.c b/zstd/zstd_compress_sequences.c
index 950471c4..f9f8097c 100644
--- a/zstd/zstd_compress_sequences.c
+++ b/zstd/zstd_compress_sequences.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -60,7 +60,7 @@ static size_t ZSTD_NCountCost(unsigned const* count, unsigned const max,
     BYTE wksp[FSE_NCOUNTBOUND];
     S16 norm[MaxSeq + 1];
     const U32 tableLog = FSE_optimalTableLog(FSELog, nbSeq, max);
-    FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq, max));
+    FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq, max), "");
     return FSE_writeNCount(wksp, sizeof(wksp), norm, max, tableLog);
 }
 
@@ -96,18 +96,22 @@ size_t ZSTD_fseBitCost(
     unsigned s;
     FSE_CState_t cstate;
     FSE_initCState(&cstate, ctable);
-    RETURN_ERROR_IF(ZSTD_getFSEMaxSymbolValue(ctable) < max, GENERIC,
-                    "Repeat FSE_CTable has maxSymbolValue %u < %u",
+    if (ZSTD_getFSEMaxSymbolValue(ctable) < max) {
+        DEBUGLOG(5, "Repeat FSE_CTable has maxSymbolValue %u < %u",
                     ZSTD_getFSEMaxSymbolValue(ctable), max);
+        return ERROR(GENERIC);
+    }
     for (s = 0; s <= max; ++s) {
         unsigned const tableLog = cstate.stateLog;
         unsigned const badCost = (tableLog + 1) << kAccuracyLog;
         unsigned const bitCost = FSE_bitCost(cstate.symbolTT, tableLog, s, kAccuracyLog);
         if (count[s] == 0)
             continue;
-        RETURN_ERROR_IF(bitCost >= badCost, GENERIC,
-                        "Repeat FSE_CTable has Prob[%u] == 0", s);
-        cost += count[s] * bitCost;
+        if (bitCost >= badCost) {
+            DEBUGLOG(5, "Repeat FSE_CTable has Prob[%u] == 0", s);
+            return ERROR(GENERIC);
+        }
+        cost += (size_t)count[s] * bitCost;
     }
     return cost >> kAccuracyLog;
 }
@@ -125,7 +129,7 @@ size_t ZSTD_crossEntropyCost(short const* norm, unsigned accuracyLog,
     unsigned s;
     assert(accuracyLog <= 8);
     for (s = 0; s <= max; ++s) {
-        unsigned const normAcc = norm[s] != -1 ? norm[s] : 1;
+        unsigned const normAcc = (norm[s] != -1) ? (unsigned)norm[s] : 1;
         unsigned const norm256 = normAcc << shift;
         assert(norm256 > 0);
         assert(norm256 < 256);
@@ -230,15 +234,15 @@ ZSTD_buildCTable(void* dst, size_t dstCapacity,
 
     switch (type) {
     case set_rle:
-        FORWARD_IF_ERROR(FSE_buildCTable_rle(nextCTable, (BYTE)max));
-        RETURN_ERROR_IF(dstCapacity==0, dstSize_tooSmall);
+        FORWARD_IF_ERROR(FSE_buildCTable_rle(nextCTable, (BYTE)max), "");
+        RETURN_ERROR_IF(dstCapacity==0, dstSize_tooSmall, "not enough space");
         *op = codeTable[0];
         return 1;
     case set_repeat:
         memcpy(nextCTable, prevCTable, prevCTableSize);
         return 0;
     case set_basic:
-        FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, defaultNorm, defaultMax, defaultNormLog, entropyWorkspace, entropyWorkspaceSize));  /* note : could be pre-calculated */
+        FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, defaultNorm, defaultMax, defaultNormLog, entropyWorkspace, entropyWorkspaceSize), "");  /* note : could be pre-calculated */
         return 0;
     case set_compressed: {
         S16 norm[MaxSeq + 1];
@@ -249,14 +253,14 @@ ZSTD_buildCTable(void* dst, size_t dstCapacity,
             nbSeq_1--;
         }
         assert(nbSeq_1 > 1);
-        FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max));
+        FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max), "");
         {   size_t const NCountSize = FSE_writeNCount(op, oend - op, norm, max, tableLog);   /* overflow protected */
-            FORWARD_IF_ERROR(NCountSize);
-            FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, norm, max, tableLog, entropyWorkspace, entropyWorkspaceSize));
+            FORWARD_IF_ERROR(NCountSize, "FSE_writeNCount failed");
+            FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, norm, max, tableLog, entropyWorkspace, entropyWorkspaceSize), "");
             return NCountSize;
         }
     }
-    default: assert(0); RETURN_ERROR(GENERIC);
+    default: assert(0); RETURN_ERROR(GENERIC, "impossible to reach");
     }
 }
 
@@ -290,7 +294,7 @@ ZSTD_encodeSequences_body(
     if (MEM_32bits()) BIT_flushBits(&blockStream);
     if (longOffsets) {
         U32 const ofBits = ofCodeTable[nbSeq-1];
-        int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
+        unsigned const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
         if (extraBits) {
             BIT_addBits(&blockStream, sequences[nbSeq-1].offset, extraBits);
             BIT_flushBits(&blockStream);
@@ -327,7 +331,7 @@ ZSTD_encodeSequences_body(
             BIT_addBits(&blockStream, sequences[n].matchLength, mlBits);
             if (MEM_32bits() || (ofBits+mlBits+llBits > 56)) BIT_flushBits(&blockStream);
             if (longOffsets) {
-                int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
+                unsigned const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
                 if (extraBits) {
                     BIT_addBits(&blockStream, sequences[n].offset, extraBits);
                     BIT_flushBits(&blockStream);                            /* (7)*/
diff --git a/zstd/zstd_compress_sequences.h b/zstd/zstd_compress_sequences.h
index 10cf6781..b884e45b 100644
--- a/zstd/zstd_compress_sequences.h
+++ b/zstd/zstd_compress_sequences.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstd_compress_superblock.c b/zstd/zstd_compress_superblock.c
index 39af1aba..7fa0f222 100644
--- a/zstd/zstd_compress_superblock.c
+++ b/zstd/zstd_compress_superblock.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -11,11 +11,13 @@
  /*-*************************************
  *  Dependencies
  ***************************************/
-#include "hist.h"           /* HIST_countFast_wksp */
+#include "zstd_compress_superblock.h"
+
+#include "zstd_internal.h"  /* ZSTD_getSequenceLength */
+#include "hist.h"                     /* HIST_countFast_wksp */
 #include "zstd_compress_internal.h"
 #include "zstd_compress_sequences.h"
 #include "zstd_compress_literals.h"
-#include "zstd_compress_superblock.h"
 
 /*-*************************************
 *  Superblock entropy buffer structs
@@ -27,7 +29,7 @@
  *  This metadata is populated in ZSTD_buildSuperBlockEntropy_literal() */
 typedef struct {
     symbolEncodingType_e hType;
-    BYTE hufDesBuffer[500]; // TODO give name to this value
+    BYTE hufDesBuffer[500]; /* TODO give name to this value */
     size_t hufDesSize;
 } ZSTD_hufCTablesMetadata_t;
 
@@ -40,9 +42,9 @@ typedef struct {
     symbolEncodingType_e llType;
     symbolEncodingType_e ofType;
     symbolEncodingType_e mlType;
-    BYTE fseTablesBuffer[500]; // TODO give name to this value
+    BYTE fseTablesBuffer[500]; /* TODO give name to this value */
     size_t fseTablesSize;
-    size_t lastCountSize; // This is to account for bug in 1.3.4. More detail in ZSTD_compressSubBlock_sequences()
+    size_t lastCountSize; /* This is to account for bug in 1.3.4. More detail in ZSTD_compressSubBlock_sequences() */
 } ZSTD_fseCTablesMetadata_t;
 
 typedef struct {
@@ -53,15 +55,14 @@ typedef struct {
 
 /** ZSTD_buildSuperBlockEntropy_literal() :
  *  Builds entropy for the super-block literals.
- *  Stores literals block type (raw, rle, compressed) and
+ *  Stores literals block type (raw, rle, compressed, repeat) and
  *  huffman description table to hufMetadata.
- *  Currently, this does not consider the option of reusing huffman table from
- *  previous super-block. I think it would be a good improvement to add that option.
  *  @return : size of huffman description table or error code */
 static size_t ZSTD_buildSuperBlockEntropy_literal(void* const src, size_t srcSize,
                                             const ZSTD_hufCTables_t* prevHuf,
                                                   ZSTD_hufCTables_t* nextHuf,
                                                   ZSTD_hufCTablesMetadata_t* hufMetadata,
+                                                  const int disableLiteralsCompression,
                                                   void* workspace, size_t wkspSize)
 {
     BYTE* const wkspStart = (BYTE*)workspace;
@@ -72,26 +73,49 @@ static size_t ZSTD_buildSuperBlockEntropy_literal(void* const src, size_t srcSiz
     BYTE* const nodeWksp = countWkspStart + countWkspSize;
     const size_t nodeWkspSize = wkspEnd-nodeWksp;
     unsigned maxSymbolValue = 255;
-    unsigned huffLog = 11;
+    unsigned huffLog = HUF_TABLELOG_DEFAULT;
+    HUF_repeat repeat = prevHuf->repeatMode;
 
     DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy_literal (srcSize=%zu)", srcSize);
 
     /* Prepare nextEntropy assuming reusing the existing table */
     memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
 
+    if (disableLiteralsCompression) {
+        DEBUGLOG(5, "set_basic - disabled");
+        hufMetadata->hType = set_basic;
+        return 0;
+    }
+
     /* small ? don't even attempt compression (speed opt) */
 #   define COMPRESS_LITERALS_SIZE_MIN 63
-    {   size_t const minLitSize = COMPRESS_LITERALS_SIZE_MIN;
-        if (srcSize <= minLitSize) { hufMetadata->hType = set_basic; return 0; }
+    {   size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN;
+        if (srcSize <= minLitSize) {
+            DEBUGLOG(5, "set_basic - too small");
+            hufMetadata->hType = set_basic;
+            return 0;
+        }
     }
 
     /* Scan input and build symbol stats */
     {   size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)src, srcSize, workspace, wkspSize);
-        FORWARD_IF_ERROR(largest);
-        if (largest == srcSize) { hufMetadata->hType = set_rle; return 0; }
-        if (largest <= (srcSize >> 7)+4) { hufMetadata->hType = set_basic; return 0; }
+        FORWARD_IF_ERROR(largest, "HIST_count_wksp failed");
+        if (largest == srcSize) {
+            DEBUGLOG(5, "set_rle");
+            hufMetadata->hType = set_rle;
+            return 0;
+        }
+        if (largest <= (srcSize >> 7)+4) {
+            DEBUGLOG(5, "set_basic - no gain");
+            hufMetadata->hType = set_basic;
+            return 0;
+        }
     }
 
+    /* Validate the previous Huffman table */
+    if (repeat == HUF_repeat_check && !HUF_validateCTable((HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue)) {
+        repeat = HUF_repeat_none;
+    }
 
     /* Build Huffman Tree */
     memset(nextHuf->CTable, 0, sizeof(nextHuf->CTable));
@@ -99,15 +123,34 @@ static size_t ZSTD_buildSuperBlockEntropy_literal(void* const src, size_t srcSiz
     {   size_t const maxBits = HUF_buildCTable_wksp((HUF_CElt*)nextHuf->CTable, countWksp,
                                                     maxSymbolValue, huffLog,
                                                     nodeWksp, nodeWkspSize);
-        FORWARD_IF_ERROR(maxBits);
+        FORWARD_IF_ERROR(maxBits, "HUF_buildCTable_wksp");
         huffLog = (U32)maxBits;
-        {   size_t cSize = HUF_estimateCompressedSize(
-                              (HUF_CElt*)nextHuf->CTable, countWksp, maxSymbolValue);
-            size_t hSize = HUF_writeCTable(
-                              hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer),
-                              (HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog);
-            if (cSize + hSize >= srcSize) { hufMetadata->hType = set_basic; return 0; }
+        {   /* Build and write the CTable */
+            size_t const newCSize = HUF_estimateCompressedSize(
+                    (HUF_CElt*)nextHuf->CTable, countWksp, maxSymbolValue);
+            size_t const hSize = HUF_writeCTable(
+                    hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer),
+                    (HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog);
+            /* Check against repeating the previous CTable */
+            if (repeat != HUF_repeat_none) {
+                size_t const oldCSize = HUF_estimateCompressedSize(
+                        (HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue);
+                if (oldCSize < srcSize && (oldCSize <= hSize + newCSize || hSize + 12 >= srcSize)) {
+                    DEBUGLOG(5, "set_repeat - smaller");
+                    memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+                    hufMetadata->hType = set_repeat;
+                    return 0;
+                }
+            }
+            if (newCSize + hSize >= srcSize) {
+                DEBUGLOG(5, "set_basic - no gains");
+                memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
+                hufMetadata->hType = set_basic;
+                return 0;
+            }
+            DEBUGLOG(5, "set_compressed (hSize=%u)", (U32)hSize);
             hufMetadata->hType = set_compressed;
+            nextHuf->repeatMode = HUF_repeat_check;
             return hSize;
         }
     }
@@ -167,7 +210,7 @@ static size_t ZSTD_buildSuperBlockEntropy_sequences(seqStore_t* seqStorePtr,
                                                     countWksp, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL,
                                                     prevEntropy->litlengthCTable, sizeof(prevEntropy->litlengthCTable),
                                                     cTableWksp, cTableWkspSize);
-            FORWARD_IF_ERROR(countSize);
+            FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for LitLens failed");
             if (LLtype == set_compressed)
                 fseMetadata->lastCountSize = countSize;
             op += countSize;
@@ -191,7 +234,7 @@ static size_t ZSTD_buildSuperBlockEntropy_sequences(seqStore_t* seqStorePtr,
                                                     countWksp, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
                                                     prevEntropy->offcodeCTable, sizeof(prevEntropy->offcodeCTable),
                                                     cTableWksp, cTableWkspSize);
-            FORWARD_IF_ERROR(countSize);
+            FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for Offsets failed");
             if (Offtype == set_compressed)
                 fseMetadata->lastCountSize = countSize;
             op += countSize;
@@ -213,7 +256,7 @@ static size_t ZSTD_buildSuperBlockEntropy_sequences(seqStore_t* seqStorePtr,
                                                     countWksp, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML,
                                                     prevEntropy->matchlengthCTable, sizeof(prevEntropy->matchlengthCTable),
                                                     cTableWksp, cTableWkspSize);
-            FORWARD_IF_ERROR(countSize);
+            FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for MatchLengths failed");
             if (MLtype == set_compressed)
                 fseMetadata->lastCountSize = countSize;
             op += countSize;
@@ -241,35 +284,34 @@ ZSTD_buildSuperBlockEntropy(seqStore_t* seqStorePtr,
         ZSTD_buildSuperBlockEntropy_literal(seqStorePtr->litStart, litSize,
                                             &prevEntropy->huf, &nextEntropy->huf,
                                             &entropyMetadata->hufMetadata,
+                                            ZSTD_disableLiteralsCompression(cctxParams),
                                             workspace, wkspSize);
-    FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize);
+    FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize, "ZSTD_buildSuperBlockEntropy_literal failed");
     entropyMetadata->fseMetadata.fseTablesSize =
         ZSTD_buildSuperBlockEntropy_sequences(seqStorePtr,
                                               &prevEntropy->fse, &nextEntropy->fse,
                                               cctxParams,
                                               &entropyMetadata->fseMetadata,
                                               workspace, wkspSize);
-    FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize);
+    FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize, "ZSTD_buildSuperBlockEntropy_sequences failed");
     return 0;
 }
 
 /** ZSTD_compressSubBlock_literal() :
  *  Compresses literals section for a sub-block.
- *  Compressed literal size needs to be less than uncompressed literal size.
- *      ZSTD spec doesn't have this constaint. I will explain why I have this constraint here.
- *      Literals section header size ranges from 1 to 5 bytes,
- *      which is dictated by regenerated size and compressed size.
- *      In order to figure out the memory address to start writing compressed literal,
- *      it is necessary to figure out the literals section header size.
- *      The challenge is that compressed size is only known after compression.
- *      This is a chicken and egg problem.
- *      I am simplifying the problem by assuming that
- *      compressed size will always be less than or equal to regenerated size,
- *      and using regenerated size to calculate literals section header size.
+ *  When we have to write the Huffman table we will sometimes choose a header
+ *  size larger than necessary. This is because we have to pick the header size
+ *  before we know the table size + compressed size, so we have a bound on the
+ *  table size. If we guessed incorrectly, we fall back to uncompressed literals.
+ *
+ *  We write the header when writeEntropy=1 and set entropyWrriten=1 when we succeeded
+ *  in writing the header, otherwise it is set to 0.
+ *
  *  hufMetadata->hType has literals block type info.
  *      If it is set_basic, all sub-blocks literals section will be Raw_Literals_Block.
  *      If it is set_rle, all sub-blocks literals section will be RLE_Literals_Block.
  *      If it is set_compressed, first sub-block's literals section will be Compressed_Literals_Block
+ *      If it is set_compressed, first sub-block's literals section will be Treeless_Literals_Block
  *      and the following sub-blocks' literals sections will be Treeless_Literals_Block.
  *  @return : compressed size of literals section of a sub-block
  *            Or 0 if it unable to compress.
@@ -278,28 +320,22 @@ static size_t ZSTD_compressSubBlock_literal(const HUF_CElt* hufTable,
                                     const ZSTD_hufCTablesMetadata_t* hufMetadata,
                                     const BYTE* literals, size_t litSize,
                                     void* dst, size_t dstSize,
-                                    const int bmi2, int writeEntropy)
+                                    const int bmi2, int writeEntropy, int* entropyWritten)
 {
-    size_t const lhSize = 3 + (litSize >= 1 KB) + (litSize >= 16 KB);
+    size_t const header = writeEntropy ? 200 : 0;
+    size_t const lhSize = 3 + (litSize >= (1 KB - header)) + (litSize >= (16 KB - header));
     BYTE* const ostart = (BYTE*)dst;
     BYTE* const oend = ostart + dstSize;
     BYTE* op = ostart + lhSize;
-    U32 singleStream = litSize < 256;
-    symbolEncodingType_e hType = writeEntropy ? set_compressed : set_repeat;
+    U32 const singleStream = lhSize == 3;
+    symbolEncodingType_e hType = writeEntropy ? hufMetadata->hType : set_repeat;
     size_t cLitSize = 0;
 
-    (void)bmi2; // TODO bmi2...
+    (void)bmi2; /* TODO bmi2... */
 
     DEBUGLOG(5, "ZSTD_compressSubBlock_literal (litSize=%zu, lhSize=%zu, writeEntropy=%d)", litSize, lhSize, writeEntropy);
 
-    if (writeEntropy && litSize == 0) {
-      /* Literals section cannot be compressed mode when litSize == 0.
-       * (This seems to be decoder constraint.)
-       * Entropy cannot be written if literals section is not compressed mode.
-       */
-      return 0;
-    }
-
+    *entropyWritten = 0;
     if (litSize == 0 || hufMetadata->hType == set_basic) {
       DEBUGLOG(5, "ZSTD_compressSubBlock_literal using raw literal");
       return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
@@ -308,25 +344,35 @@ static size_t ZSTD_compressSubBlock_literal(const HUF_CElt* hufTable,
       return ZSTD_compressRleLiteralsBlock(dst, dstSize, literals, litSize);
     }
 
-    if (lhSize == 3) singleStream = 1;
-    if (writeEntropy) {
+    assert(litSize > 0);
+    assert(hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat);
+
+    if (writeEntropy && hufMetadata->hType == set_compressed) {
         memcpy(op, hufMetadata->hufDesBuffer, hufMetadata->hufDesSize);
         op += hufMetadata->hufDesSize;
         cLitSize += hufMetadata->hufDesSize;
         DEBUGLOG(5, "ZSTD_compressSubBlock_literal (hSize=%zu)", hufMetadata->hufDesSize);
     }
 
-    // TODO bmi2
+    /* TODO bmi2 */
     {   const size_t cSize = singleStream ? HUF_compress1X_usingCTable(op, oend-op, literals, litSize, hufTable)
                                           : HUF_compress4X_usingCTable(op, oend-op, literals, litSize, hufTable);
         op += cSize;
         cLitSize += cSize;
         if (cSize == 0 || ERR_isError(cSize)) {
-          return 0;
+            DEBUGLOG(5, "Failed to write entropy tables %s", ZSTD_getErrorName(cSize));
+            return 0;
         }
-        if (cLitSize > litSize) {
-            if (writeEntropy) return 0;
-            else return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
+        /* If we expand and we aren't writing a header then emit uncompressed */
+        if (!writeEntropy && cLitSize >= litSize) {
+            DEBUGLOG(5, "ZSTD_compressSubBlock_literal using raw literal because uncompressible");
+            return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
+        }
+        /* If we are writing headers then allow expansion that doesn't change our header size. */
+        if (lhSize < (size_t)(3 + (cLitSize >= 1 KB) + (cLitSize >= 16 KB))) {
+            assert(cLitSize > litSize);
+            DEBUGLOG(5, "Literals expanded beyond allowed header size");
+            return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
         }
         DEBUGLOG(5, "ZSTD_compressSubBlock_literal (cSize=%zu)", cSize);
     }
@@ -353,17 +399,26 @@ static size_t ZSTD_compressSubBlock_literal(const HUF_CElt* hufTable,
     default:  /* not possible : lhSize is {3,4,5} */
         assert(0);
     }
+    *entropyWritten = 1;
+    DEBUGLOG(5, "Compressed literals: %u -> %u", (U32)litSize, (U32)(op-ostart));
     return op-ostart;
 }
 
-static size_t ZSTD_seqDecompressedSize(const seqDef* sequences, size_t nbSeq, size_t litSize) {
+static size_t ZSTD_seqDecompressedSize(seqStore_t const* seqStore, const seqDef* sequences, size_t nbSeq, size_t litSize, int lastSequence) {
     const seqDef* const sstart = sequences;
     const seqDef* const send = sequences + nbSeq;
     const seqDef* sp = sstart;
     size_t matchLengthSum = 0;
+    size_t litLengthSum = 0;
     while (send-sp > 0) {
-      matchLengthSum += sp->matchLength + MINMATCH;
-      sp++;
+        ZSTD_sequenceLength const seqLen = ZSTD_getSequenceLength(seqStore, sp);
+        litLengthSum += seqLen.litLength;
+        matchLengthSum += seqLen.matchLength;
+        sp++;
+    }
+    assert(litLengthSum <= litSize);
+    if (!lastSequence) {
+        assert(litLengthSum == litSize);
     }
     return matchLengthSum + litSize;
 }
@@ -372,8 +427,9 @@ static size_t ZSTD_seqDecompressedSize(const seqDef* sequences, size_t nbSeq, si
  *  Compresses sequences section for a sub-block.
  *  fseMetadata->llType, fseMetadata->ofType, and fseMetadata->mlType have
  *  symbol compression modes for the super-block.
- *  First sub-block will have these in its header. The following sub-blocks
- *  will always have repeat mode.
+ *  The first successfully compressed block will have these in its header.
+ *  We set entropyWritten=1 when we succeed in compressing the sequences.
+ *  The following sub-blocks will always have repeat mode.
  *  @return : compressed size of sequences section of a sub-block
  *            Or 0 if it is unable to compress
  *            Or error code. */
@@ -383,7 +439,7 @@ static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables
                                               const BYTE* llCode, const BYTE* mlCode, const BYTE* ofCode,
                                               const ZSTD_CCtx_params* cctxParams,
                                               void* dst, size_t dstCapacity,
-                                              const int bmi2, int writeEntropy)
+                                              const int bmi2, int writeEntropy, int* entropyWritten)
 {
     const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;
     BYTE* const ostart = (BYTE*)dst;
@@ -393,18 +449,16 @@ static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables
 
     DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (nbSeq=%zu, writeEntropy=%d, longOffsets=%d)", nbSeq, writeEntropy, longOffsets);
 
+    *entropyWritten = 0;
     /* Sequences Header */
     RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,
-                    dstSize_tooSmall);
+                    dstSize_tooSmall, "");
     if (nbSeq < 0x7F)
         *op++ = (BYTE)nbSeq;
     else if (nbSeq < LONGNBSEQ)
         op[0] = (BYTE)((nbSeq>>8) + 0x80), op[1] = (BYTE)nbSeq, op+=2;
     else
         op[0]=0xFF, MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)), op+=3;
-    if (writeEntropy && nbSeq == 0) {
-        return 0;
-    }
     if (nbSeq==0) {
         return op - ostart;
     }
@@ -434,7 +488,7 @@ static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables
                                         fseTables->litlengthCTable, llCode,
                                         sequences, nbSeq,
                                         longOffsets, bmi2);
-        FORWARD_IF_ERROR(bitstreamSize);
+        FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed");
         op += bitstreamSize;
         /* zstd versions <= 1.3.4 mistakenly report corruption when
          * FSE_readNCount() receives a buffer < 4 bytes.
@@ -444,6 +498,7 @@ static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables
          * In this exceedingly rare case, we will simply emit an uncompressed
          * block, since it isn't worth optimizing.
          */
+#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
         if (writeEntropy && fseMetadata->lastCountSize && fseMetadata->lastCountSize + bitstreamSize < 4) {
             /* NCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */
             assert(fseMetadata->lastCountSize + bitstreamSize == 3);
@@ -451,6 +506,7 @@ static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables
                         "emitting an uncompressed block.");
             return 0;
         }
+#endif
         DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (bitstreamSize=%zu)", bitstreamSize);
     }
 
@@ -461,10 +517,15 @@ static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables
      * with rle mode and the current block's sequences section is compressed
      * with repeat mode where sequences section body size can be 1 byte.
      */
+#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
     if (op-seqHead < 4) {
+        DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.4.0 by emitting "
+                    "an uncompressed block when sequences are < 4 bytes");
         return 0;
     }
+#endif
 
+    *entropyWritten = 1;
     return op - ostart;
 }
 
@@ -479,17 +540,20 @@ static size_t ZSTD_compressSubBlock(const ZSTD_entropyCTables_t* entropy,
                                     const BYTE* llCode, const BYTE* mlCode, const BYTE* ofCode,
                                     const ZSTD_CCtx_params* cctxParams,
                                     void* dst, size_t dstCapacity,
-                                    const int bmi2, int writeEntropy, U32 lastBlock)
+                                    const int bmi2,
+                                    int writeLitEntropy, int writeSeqEntropy,
+                                    int* litEntropyWritten, int* seqEntropyWritten,
+                                    U32 lastBlock)
 {
     BYTE* const ostart = (BYTE*)dst;
     BYTE* const oend = ostart + dstCapacity;
     BYTE* op = ostart + ZSTD_blockHeaderSize;
-    DEBUGLOG(5, "ZSTD_compressSubBlock (litSize=%zu, nbSeq=%zu, writeEntropy=%d, lastBlock=%d)",
-                litSize, nbSeq, writeEntropy, lastBlock);
+    DEBUGLOG(5, "ZSTD_compressSubBlock (litSize=%zu, nbSeq=%zu, writeLitEntropy=%d, writeSeqEntropy=%d, lastBlock=%d)",
+                litSize, nbSeq, writeLitEntropy, writeSeqEntropy, lastBlock);
     {   size_t cLitSize = ZSTD_compressSubBlock_literal((const HUF_CElt*)entropy->huf.CTable,
                                                         &entropyMetadata->hufMetadata, literals, litSize,
-                                                        op, oend-op, bmi2, writeEntropy);
-        FORWARD_IF_ERROR(cLitSize);
+                                                        op, oend-op, bmi2, writeLitEntropy, litEntropyWritten);
+        FORWARD_IF_ERROR(cLitSize, "ZSTD_compressSubBlock_literal failed");
         if (cLitSize == 0) return 0;
         op += cLitSize;
     }
@@ -499,8 +563,8 @@ static size_t ZSTD_compressSubBlock(const ZSTD_entropyCTables_t* entropy,
                                                   llCode, mlCode, ofCode,
                                                   cctxParams,
                                                   op, oend-op,
-                                                  bmi2, writeEntropy);
-        FORWARD_IF_ERROR(cSeqSize);
+                                                  bmi2, writeSeqEntropy, seqEntropyWritten);
+        FORWARD_IF_ERROR(cSeqSize, "ZSTD_compressSubBlock_sequences failed");
         if (cSeqSize == 0) return 0;
         op += cSeqSize;
     }
@@ -524,7 +588,7 @@ static size_t ZSTD_estimateSubBlockSize_literal(const BYTE* literals, size_t lit
 
     if (hufMetadata->hType == set_basic) return litSize;
     else if (hufMetadata->hType == set_rle) return 1;
-    else if (hufMetadata->hType == set_compressed) {
+    else if (hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat) {
         size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)literals, litSize, workspace, wkspSize);
         if (ZSTD_isError(largest)) return litSize;
         {   size_t cLitSizeEstimate = HUF_estimateCompressedSize((const HUF_CElt*)huf->CTable, countWksp, maxSymbolValue);
@@ -601,17 +665,28 @@ static size_t ZSTD_estimateSubBlockSize(const BYTE* literals, size_t litSize,
                                         const ZSTD_entropyCTables_t* entropy,
                                         const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
                                         void* workspace, size_t wkspSize,
-                                        int writeEntropy) {
+                                        int writeLitEntropy, int writeSeqEntropy) {
     size_t cSizeEstimate = 0;
     cSizeEstimate += ZSTD_estimateSubBlockSize_literal(literals, litSize,
                                                          &entropy->huf, &entropyMetadata->hufMetadata,
-                                                         workspace, wkspSize, writeEntropy);
+                                                         workspace, wkspSize, writeLitEntropy);
     cSizeEstimate += ZSTD_estimateSubBlockSize_sequences(ofCodeTable, llCodeTable, mlCodeTable,
                                                          nbSeq, &entropy->fse, &entropyMetadata->fseMetadata,
-                                                         workspace, wkspSize, writeEntropy);
+                                                         workspace, wkspSize, writeSeqEntropy);
     return cSizeEstimate + ZSTD_blockHeaderSize;
 }
 
+static int ZSTD_needSequenceEntropyTables(ZSTD_fseCTablesMetadata_t const* fseMetadata)
+{
+    if (fseMetadata->llType == set_compressed || fseMetadata->llType == set_rle)
+        return 1;
+    if (fseMetadata->mlType == set_compressed || fseMetadata->mlType == set_rle)
+        return 1;
+    if (fseMetadata->ofType == set_compressed || fseMetadata->ofType == set_rle)
+        return 1;
+    return 0;
+}
+
 /** ZSTD_compressSubBlock_multi() :
  *  Breaks super-block into multiple sub-blocks and compresses them.
  *  Entropy will be written to the first block.
@@ -620,10 +695,12 @@ static size_t ZSTD_estimateSubBlockSize(const BYTE* literals, size_t litSize,
  *  @return : compressed size of the super block (which is multiple ZSTD blocks)
  *            Or 0 if it failed to compress. */
 static size_t ZSTD_compressSubBlock_multi(const seqStore_t* seqStorePtr,
-                            const ZSTD_entropyCTables_t* entropy,
+                            const ZSTD_compressedBlockState_t* prevCBlock,
+                            ZSTD_compressedBlockState_t* nextCBlock,
                             const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
                             const ZSTD_CCtx_params* cctxParams,
                                   void* dst, size_t dstCapacity,
+                            const void* src, size_t srcSize,
                             const int bmi2, U32 lastBlock,
                             void* workspace, size_t wkspSize)
 {
@@ -633,6 +710,8 @@ static size_t ZSTD_compressSubBlock_multi(const seqStore_t* seqStorePtr,
     const BYTE* const lstart = seqStorePtr->litStart;
     const BYTE* const lend = seqStorePtr->lit;
     const BYTE* lp = lstart;
+    BYTE const* ip = (BYTE const*)src;
+    BYTE const* const iend = ip + srcSize;
     BYTE* const ostart = (BYTE*)dst;
     BYTE* const oend = ostart + dstCapacity;
     BYTE* op = ostart;
@@ -641,41 +720,56 @@ static size_t ZSTD_compressSubBlock_multi(const seqStore_t* seqStorePtr,
     const BYTE* ofCodePtr = seqStorePtr->ofCode;
     size_t targetCBlockSize = cctxParams->targetCBlockSize;
     size_t litSize, seqCount;
-    int writeEntropy = 1;
-    size_t remaining = ZSTD_seqDecompressedSize(sstart, send-sstart, lend-lstart);
-    size_t cBlockSizeEstimate = 0;
+    int writeLitEntropy = entropyMetadata->hufMetadata.hType == set_compressed;
+    int writeSeqEntropy = 1;
+    int lastSequence = 0;
 
     DEBUGLOG(5, "ZSTD_compressSubBlock_multi (litSize=%u, nbSeq=%u)",
                 (unsigned)(lend-lp), (unsigned)(send-sstart));
 
     litSize = 0;
     seqCount = 0;
-    while (sp + seqCount < send) {
-        const seqDef* const sequence = sp + seqCount;
-        const U32 lastSequence = sequence+1 == send;
-        litSize = (sequence == send) ? (size_t)(lend-lp) : litSize + sequence->litLength;
-        seqCount++;
+    do {
+        size_t cBlockSizeEstimate = 0;
+        if (sstart == send) {
+            lastSequence = 1;
+        } else {
+            const seqDef* const sequence = sp + seqCount;
+            lastSequence = sequence == send - 1;
+            litSize += ZSTD_getSequenceLength(seqStorePtr, sequence).litLength;
+            seqCount++;
+        }
+        if (lastSequence) {
+            assert(lp <= lend);
+            assert(litSize <= (size_t)(lend - lp));
+            litSize = (size_t)(lend - lp);
+        }
         /* I think there is an optimization opportunity here.
          * Calling ZSTD_estimateSubBlockSize for every sequence can be wasteful
          * since it recalculates estimate from scratch.
          * For example, it would recount literal distribution and symbol codes everytime.
          */
         cBlockSizeEstimate = ZSTD_estimateSubBlockSize(lp, litSize, ofCodePtr, llCodePtr, mlCodePtr, seqCount,
-                                                       entropy, entropyMetadata,
-                                                       workspace, wkspSize, writeEntropy);
+                                                       &nextCBlock->entropy, entropyMetadata,
+                                                       workspace, wkspSize, writeLitEntropy, writeSeqEntropy);
         if (cBlockSizeEstimate > targetCBlockSize || lastSequence) {
-            const size_t decompressedSize = ZSTD_seqDecompressedSize(sp, seqCount, litSize);
-            const size_t cSize = ZSTD_compressSubBlock(entropy, entropyMetadata,
+            int litEntropyWritten = 0;
+            int seqEntropyWritten = 0;
+            const size_t decompressedSize = ZSTD_seqDecompressedSize(seqStorePtr, sp, seqCount, litSize, lastSequence);
+            const size_t cSize = ZSTD_compressSubBlock(&nextCBlock->entropy, entropyMetadata,
                                                        sp, seqCount,
                                                        lp, litSize,
                                                        llCodePtr, mlCodePtr, ofCodePtr,
                                                        cctxParams,
                                                        op, oend-op,
-                                                       bmi2, writeEntropy, lastBlock && lastSequence);
-            FORWARD_IF_ERROR(cSize);
+                                                       bmi2, writeLitEntropy, writeSeqEntropy,
+                                                       &litEntropyWritten, &seqEntropyWritten,
+                                                       lastBlock && lastSequence);
+            FORWARD_IF_ERROR(cSize, "ZSTD_compressSubBlock failed");
             if (cSize > 0 && cSize < decompressedSize) {
-                assert(remaining >= decompressedSize);
-                remaining -= decompressedSize;
+                DEBUGLOG(5, "Committed the sub-block");
+                assert(ip + decompressedSize <= iend);
+                ip += decompressedSize;
                 sp += seqCount;
                 lp += litSize;
                 op += cSize;
@@ -684,20 +778,51 @@ static size_t ZSTD_compressSubBlock_multi(const seqStore_t* seqStorePtr,
                 ofCodePtr += seqCount;
                 litSize = 0;
                 seqCount = 0;
-                writeEntropy = 0; // Entropy only needs to be written once
+                /* Entropy only needs to be written once */
+                if (litEntropyWritten) {
+                    writeLitEntropy = 0;
+                }
+                if (seqEntropyWritten) {
+                    writeSeqEntropy = 0;
+                }
             }
         }
+    } while (!lastSequence);
+    if (writeLitEntropy) {
+        DEBUGLOG(5, "ZSTD_compressSubBlock_multi has literal entropy tables unwritten");
+        memcpy(&nextCBlock->entropy.huf, &prevCBlock->entropy.huf, sizeof(prevCBlock->entropy.huf));
     }
-    if (remaining) {
-        DEBUGLOG(5, "ZSTD_compressSubBlock_multi failed to compress");
+    if (writeSeqEntropy && ZSTD_needSequenceEntropyTables(&entropyMetadata->fseMetadata)) {
+        /* If we haven't written our entropy tables, then we've violated our contract and
+         * must emit an uncompressed block.
+         */
+        DEBUGLOG(5, "ZSTD_compressSubBlock_multi has sequence entropy tables unwritten");
         return 0;
     }
+    if (ip < iend) {
+        size_t const cSize = ZSTD_noCompressBlock(op, oend - op, ip, iend - ip, lastBlock);
+        DEBUGLOG(5, "ZSTD_compressSubBlock_multi last sub-block uncompressed, %zu bytes", (size_t)(iend - ip));
+        FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
+        assert(cSize != 0);
+        op += cSize;
+        /* We have to regenerate the repcodes because we've skipped some sequences */
+        if (sp < send) {
+            seqDef const* seq;
+            repcodes_t rep;
+            memcpy(&rep, prevCBlock->rep, sizeof(rep)); 
+            for (seq = sstart; seq < sp; ++seq) {
+                rep = ZSTD_updateRep(rep.rep, seq->offset - 1, ZSTD_getSequenceLength(seqStorePtr, seq).litLength == 0);
+            }
+            memcpy(nextCBlock->rep, &rep, sizeof(rep));
+        }
+    }
     DEBUGLOG(5, "ZSTD_compressSubBlock_multi compressed");
     return op-ostart;
 }
 
 size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
                                void* dst, size_t dstCapacity,
+                               void const* src, size_t srcSize,
                                unsigned lastBlock) {
     ZSTD_entropyCTablesMetadata_t entropyMetadata;
 
@@ -706,13 +831,15 @@ size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
           &zc->blockState.nextCBlock->entropy,
           &zc->appliedParams,
           &entropyMetadata,
-          zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */));
+          zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */), "");
 
     return ZSTD_compressSubBlock_multi(&zc->seqStore,
-            &zc->blockState.nextCBlock->entropy,
+            zc->blockState.prevCBlock,
+            zc->blockState.nextCBlock,
             &entropyMetadata,
             &zc->appliedParams,
             dst, dstCapacity,
+            src, srcSize,
             zc->bmi2, lastBlock,
             zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */);
 }
diff --git a/zstd/zstd_compress_superblock.h b/zstd/zstd_compress_superblock.h
index 099ad3bb..35d20729 100644
--- a/zstd/zstd_compress_superblock.h
+++ b/zstd/zstd_compress_superblock.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -26,6 +26,7 @@
  * The given block will be compressed into multiple sub blocks that are around targetCBlockSize. */
 size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
                                void* dst, size_t dstCapacity,
+                               void const* src, size_t srcSize,
                                unsigned lastBlock);
 
 #endif /* ZSTD_COMPRESS_ADVANCED_H */
diff --git a/zstd/zstd_cwksp.h b/zstd/zstd_cwksp.h
index fc9765bd..91f812fa 100644
--- a/zstd/zstd_cwksp.h
+++ b/zstd/zstd_cwksp.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -24,16 +24,6 @@ extern "C" {
 *  Constants
 ***************************************/
 
-/* define "workspace is too large" as this number of times larger than needed */
-#define ZSTD_WORKSPACETOOLARGE_FACTOR 3
-
-/* when workspace is continuously too large
- * during at least this number of times,
- * context's memory usage is considered wasteful,
- * because it's sized to handle a worst case scenario which rarely happens.
- * In which case, resize it down to free some memory */
-#define ZSTD_WORKSPACETOOLARGE_MAXDURATION 128
-
 /* Since the workspace is effectively its own little malloc implementation /
  * arena, when we run under ASAN, we should similarly insert redzones between
  * each internal element of the workspace, so ASAN will catch overruns that
@@ -468,7 +458,7 @@ MEM_STATIC void ZSTD_cwksp_init(ZSTD_cwksp* ws, void* start, size_t size) {
 MEM_STATIC size_t ZSTD_cwksp_create(ZSTD_cwksp* ws, size_t size, ZSTD_customMem customMem) {
     void* workspace = ZSTD_malloc(size, customMem);
     DEBUGLOG(4, "cwksp: creating new workspace with %zd bytes", size);
-    RETURN_ERROR_IF(workspace == NULL, memory_allocation);
+    RETURN_ERROR_IF(workspace == NULL, memory_allocation, "NULL pointer!");
     ZSTD_cwksp_init(ws, workspace, size);
     return 0;
 }
diff --git a/zstd/zstd_ddict.c b/zstd/zstd_ddict.c
index 0af3d23b..47e985fc 100644
--- a/zstd/zstd_ddict.c
+++ b/zstd/zstd_ddict.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -25,7 +25,7 @@
 #include "zstd_ddict.h"
 
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
-#  include "zstd_legacy.h"
+#  include "../legacy/zstd_legacy.h"
 #endif
 
 
@@ -65,6 +65,10 @@ void ZSTD_copyDDictParameters(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
     dctx->virtualStart = ddict->dictContent;
     dctx->dictEnd = (const BYTE*)ddict->dictContent + ddict->dictSize;
     dctx->previousDstEnd = dctx->dictEnd;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+    dctx->dictContentBeginForFuzzing = dctx->prefixStart;
+    dctx->dictContentEndForFuzzing = dctx->previousDstEnd;
+#endif
     if (ddict->entropyPresent) {
         dctx->litEntropy = 1;
         dctx->fseEntropy = 1;
@@ -107,7 +111,7 @@ ZSTD_loadEntropy_intoDDict(ZSTD_DDict* ddict,
     /* load entropy tables */
     RETURN_ERROR_IF(ZSTD_isError(ZSTD_loadDEntropy(
             &ddict->entropy, ddict->dictContent, ddict->dictSize)),
-        dictionary_corrupted);
+        dictionary_corrupted, "");
     ddict->entropyPresent = 1;
     return 0;
 }
@@ -133,7 +137,7 @@ static size_t ZSTD_initDDict_internal(ZSTD_DDict* ddict,
     ddict->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001);  /* cover both little and big endian */
 
     /* parse dictionary content */
-    FORWARD_IF_ERROR( ZSTD_loadEntropy_intoDDict(ddict, dictContentType) );
+    FORWARD_IF_ERROR( ZSTD_loadEntropy_intoDDict(ddict, dictContentType) , "");
 
     return 0;
 }
diff --git a/zstd/zstd_ddict.h b/zstd/zstd_ddict.h
index 0479d11b..3ab7532e 100644
--- a/zstd/zstd_ddict.h
+++ b/zstd/zstd_ddict.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstd_decompress.c b/zstd/zstd_decompress.c
index f87e2618..5b06e7d3 100644
--- a/zstd/zstd_decompress.c
+++ b/zstd/zstd_decompress.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -68,7 +68,7 @@
 #include "zstd_decompress_block.h"   /* ZSTD_decompressBlock_internal */
 
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
-#  include "zstd_legacy.h"
+#  include "../legacy/zstd_legacy.h"
 #endif
 
 
@@ -111,7 +111,12 @@ static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx)
     dctx->legacyContext = NULL;
     dctx->previousLegacyVersion = 0;
     dctx->noForwardProgress = 0;
+    dctx->oversizedDuration = 0;
     dctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
+    dctx->outBufferMode = ZSTD_obm_buffered;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+    dctx->dictContentEndForFuzzing = NULL;
+#endif
 }
 
 ZSTD_DCtx* ZSTD_initStaticDCtx(void *workspace, size_t workspaceSize)
@@ -208,7 +213,7 @@ unsigned ZSTD_isFrame(const void* buffer, size_t size)
 static size_t ZSTD_frameHeaderSize_internal(const void* src, size_t srcSize, ZSTD_format_e format)
 {
     size_t const minInputSize = ZSTD_startingInputLength(format);
-    RETURN_ERROR_IF(srcSize < minInputSize, srcSize_wrong);
+    RETURN_ERROR_IF(srcSize < minInputSize, srcSize_wrong, "");
 
     {   BYTE const fhd = ((const BYTE*)src)[minInputSize-1];
         U32 const dictID= fhd & 3;
@@ -256,7 +261,7 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
             zfhPtr->frameType = ZSTD_skippableFrame;
             return 0;
         }
-        RETURN_ERROR(prefix_unknown);
+        RETURN_ERROR(prefix_unknown, "");
     }
 
     /* ensure there is enough `srcSize` to fully read/decode frame header */
@@ -280,7 +285,7 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
         if (!singleSegment) {
             BYTE const wlByte = ip[pos++];
             U32 const windowLog = (wlByte >> 3) + ZSTD_WINDOWLOG_ABSOLUTEMIN;
-            RETURN_ERROR_IF(windowLog > ZSTD_WINDOWLOG_MAX, frameParameter_windowTooLarge);
+            RETURN_ERROR_IF(windowLog > ZSTD_WINDOWLOG_MAX, frameParameter_windowTooLarge, "");
             windowSize = (1ULL << windowLog);
             windowSize += (windowSize >> 3) * (wlByte&7);
         }
@@ -352,14 +357,14 @@ static size_t readSkippableFrameSize(void const* src, size_t srcSize)
     size_t const skippableHeaderSize = ZSTD_SKIPPABLEHEADERSIZE;
     U32 sizeU32;
 
-    RETURN_ERROR_IF(srcSize < ZSTD_SKIPPABLEHEADERSIZE, srcSize_wrong);
+    RETURN_ERROR_IF(srcSize < ZSTD_SKIPPABLEHEADERSIZE, srcSize_wrong, "");
 
     sizeU32 = MEM_readLE32((BYTE const*)src + ZSTD_FRAMEIDSIZE);
     RETURN_ERROR_IF((U32)(sizeU32 + ZSTD_SKIPPABLEHEADERSIZE) < sizeU32,
-                    frameParameter_unsupported);
+                    frameParameter_unsupported, "");
     {
         size_t const skippableSize = skippableHeaderSize + sizeU32;
-        RETURN_ERROR_IF(skippableSize > srcSize, srcSize_wrong);
+        RETURN_ERROR_IF(skippableSize > srcSize, srcSize_wrong, "");
         return skippableSize;
     }
 }
@@ -439,7 +444,7 @@ static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t he
      * harder.
      */
     RETURN_ERROR_IF(dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID),
-                    dictionary_wrong);
+                    dictionary_wrong, "");
 #endif
     if (dctx->fParams.checksumFlag) XXH64_reset(&dctx->xxhState, 0);
     return 0;
@@ -576,9 +581,9 @@ static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity,
     DEBUGLOG(5, "ZSTD_copyRawBlock");
     if (dst == NULL) {
         if (srcSize == 0) return 0;
-        RETURN_ERROR(dstBuffer_null);
+        RETURN_ERROR(dstBuffer_null, "");
     }
-    RETURN_ERROR_IF(srcSize > dstCapacity, dstSize_tooSmall);
+    RETURN_ERROR_IF(srcSize > dstCapacity, dstSize_tooSmall, "");
     memcpy(dst, src, srcSize);
     return srcSize;
 }
@@ -589,9 +594,9 @@ static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity,
 {
     if (dst == NULL) {
         if (regenSize == 0) return 0;
-        RETURN_ERROR(dstBuffer_null);
+        RETURN_ERROR(dstBuffer_null, "");
     }
-    RETURN_ERROR_IF(regenSize > dstCapacity, dstSize_tooSmall);
+    RETURN_ERROR_IF(regenSize > dstCapacity, dstSize_tooSmall, "");
     memset(dst, b, regenSize);
     return regenSize;
 }
@@ -616,15 +621,15 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
     /* check */
     RETURN_ERROR_IF(
         remainingSrcSize < ZSTD_FRAMEHEADERSIZE_MIN(dctx->format)+ZSTD_blockHeaderSize,
-        srcSize_wrong);
+        srcSize_wrong, "");
 
     /* Frame Header */
     {   size_t const frameHeaderSize = ZSTD_frameHeaderSize_internal(
                 ip, ZSTD_FRAMEHEADERSIZE_PREFIX(dctx->format), dctx->format);
         if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize;
         RETURN_ERROR_IF(remainingSrcSize < frameHeaderSize+ZSTD_blockHeaderSize,
-                        srcSize_wrong);
-        FORWARD_IF_ERROR( ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize) );
+                        srcSize_wrong, "");
+        FORWARD_IF_ERROR( ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize) , "");
         ip += frameHeaderSize; remainingSrcSize -= frameHeaderSize;
     }
 
@@ -637,7 +642,7 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
 
         ip += ZSTD_blockHeaderSize;
         remainingSrcSize -= ZSTD_blockHeaderSize;
-        RETURN_ERROR_IF(cBlockSize > remainingSrcSize, srcSize_wrong);
+        RETURN_ERROR_IF(cBlockSize > remainingSrcSize, srcSize_wrong, "");
 
         switch(blockProperties.blockType)
         {
@@ -652,7 +657,7 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
             break;
         case bt_reserved :
         default:
-            RETURN_ERROR(corruption_detected);
+            RETURN_ERROR(corruption_detected, "invalid block type");
         }
 
         if (ZSTD_isError(decodedSize)) return decodedSize;
@@ -668,14 +673,14 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
 
     if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) {
         RETURN_ERROR_IF((U64)(op-ostart) != dctx->fParams.frameContentSize,
-                        corruption_detected);
+                        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);
+        RETURN_ERROR_IF(remainingSrcSize<4, checksum_wrong, "");
         checkRead = MEM_readLE32(ip);
-        RETURN_ERROR_IF(checkRead != checkCalc, checksum_wrong);
+        RETURN_ERROR_IF(checkRead != checkCalc, checksum_wrong, "");
         ip += 4;
         remainingSrcSize -= 4;
     }
@@ -732,7 +737,7 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
                         (unsigned)magicNumber, ZSTD_MAGICNUMBER);
             if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
                 size_t const skippableSize = readSkippableFrameSize(src, srcSize);
-                FORWARD_IF_ERROR(skippableSize);
+                FORWARD_IF_ERROR(skippableSize, "readSkippableFrameSize failed");
                 assert(skippableSize <= srcSize);
 
                 src = (const BYTE *)src + skippableSize;
@@ -742,11 +747,11 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
 
         if (ddict) {
             /* we were called from ZSTD_decompress_usingDDict */
-            FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(dctx, ddict));
+            FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(dctx, ddict), "");
         } else {
             /* this will initialize correctly with no dict if dict == NULL, so
              * use this in all cases but ddict */
-            FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDict(dctx, dict, dictSize));
+            FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDict(dctx, dict, dictSize), "");
         }
         ZSTD_checkContinuity(dctx, dst);
 
@@ -816,7 +821,7 @@ 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();
-    RETURN_ERROR_IF(dctx==NULL, memory_allocation);
+    RETURN_ERROR_IF(dctx==NULL, memory_allocation, "NULL pointer!");
     regenSize = ZSTD_decompressDCtx(dctx, dst, dstCapacity, src, srcSize);
     ZSTD_freeDCtx(dctx);
     return regenSize;
@@ -909,7 +914,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
     case ZSTDds_decodeFrameHeader:
         assert(src != NULL);
         memcpy(dctx->headerBuffer + (dctx->headerSize - srcSize), src, srcSize);
-        FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize));
+        FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize), "");
         dctx->expected = ZSTD_blockHeaderSize;
         dctx->stage = ZSTDds_decodeBlockHeader;
         return 0;
@@ -956,7 +961,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
             case bt_raw :
                 assert(srcSize <= dctx->expected);
                 rSize = ZSTD_copyRawBlock(dst, dstCapacity, src, srcSize);
-                FORWARD_IF_ERROR(rSize);
+                FORWARD_IF_ERROR(rSize, "ZSTD_copyRawBlock failed");
                 assert(rSize == srcSize);
                 dctx->expected -= rSize;
                 break;
@@ -966,9 +971,9 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
                 break;
             case bt_reserved :   /* should never happen */
             default:
-                RETURN_ERROR(corruption_detected);
+                RETURN_ERROR(corruption_detected, "invalid block type");
             }
-            FORWARD_IF_ERROR(rSize);
+            FORWARD_IF_ERROR(rSize, "");
             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;
@@ -985,7 +990,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
                 RETURN_ERROR_IF(
                     dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
                  && dctx->decodedSize != dctx->fParams.frameContentSize,
-                    corruption_detected);
+                    corruption_detected, "");
                 if (dctx->fParams.checksumFlag) {  /* another round for frame checksum */
                     dctx->expected = 4;
                     dctx->stage = ZSTDds_checkChecksum;
@@ -1005,7 +1010,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
         {   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);
+            RETURN_ERROR_IF(check32 != h32, checksum_wrong, "");
             dctx->expected = 0;
             dctx->stage = ZSTDds_getFrameHeaderSize;
             return 0;
@@ -1026,7 +1031,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
 
     default:
         assert(0);   /* impossible */
-        RETURN_ERROR(GENERIC);   /* some compiler require default to do something */
+        RETURN_ERROR(GENERIC, "impossible to reach");   /* some compiler require default to do something */
     }
 }
 
@@ -1037,6 +1042,10 @@ static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dict
     dctx->virtualStart = (const char*)dict - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart));
     dctx->prefixStart = dict;
     dctx->previousDstEnd = (const char*)dict + dictSize;
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+    dctx->dictContentBeginForFuzzing = dctx->prefixStart;
+    dctx->dictContentEndForFuzzing = dctx->previousDstEnd;
+#endif
     return 0;
 }
 
@@ -1050,7 +1059,7 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
     const BYTE* dictPtr = (const BYTE*)dict;
     const BYTE* const dictEnd = dictPtr + dictSize;
 
-    RETURN_ERROR_IF(dictSize <= 8, dictionary_corrupted);
+    RETURN_ERROR_IF(dictSize <= 8, dictionary_corrupted, "dict is too small");
     assert(MEM_readLE32(dict) == ZSTD_MAGIC_DICTIONARY);   /* dict must be valid */
     dictPtr += 8;   /* skip header = magic + dictID */
 
@@ -1069,16 +1078,16 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
                                                 dictPtr, dictEnd - dictPtr,
                                                 workspace, workspaceSize);
 #endif
-        RETURN_ERROR_IF(HUF_isError(hSize), dictionary_corrupted);
+        RETURN_ERROR_IF(HUF_isError(hSize), dictionary_corrupted, "");
         dictPtr += hSize;
     }
 
     {   short offcodeNCount[MaxOff+1];
         unsigned offcodeMaxValue = MaxOff, offcodeLog;
         size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
-        RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted);
-        RETURN_ERROR_IF(offcodeMaxValue > MaxOff, dictionary_corrupted);
-        RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted);
+        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,
@@ -1089,9 +1098,9 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
     {   short matchlengthNCount[MaxML+1];
         unsigned matchlengthMaxValue = MaxML, matchlengthLog;
         size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);
-        RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted);
-        RETURN_ERROR_IF(matchlengthMaxValue > MaxML, dictionary_corrupted);
-        RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted);
+        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,
@@ -1102,9 +1111,9 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
     {   short litlengthNCount[MaxLL+1];
         unsigned litlengthMaxValue = MaxLL, litlengthLog;
         size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);
-        RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted);
-        RETURN_ERROR_IF(litlengthMaxValue > MaxLL, dictionary_corrupted);
-        RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted);
+        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,
@@ -1112,13 +1121,13 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
         dictPtr += litlengthHeaderSize;
     }
 
-    RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted);
+    RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "");
     {   int i;
         size_t const dictContentSize = (size_t)(dictEnd - (dictPtr+12));
         for (i=0; i<3; i++) {
             U32 const rep = MEM_readLE32(dictPtr); dictPtr += 4;
             RETURN_ERROR_IF(rep==0 || rep > dictContentSize,
-                            dictionary_corrupted);
+                            dictionary_corrupted, "");
             entropy->rep[i] = rep;
     }   }
 
@@ -1136,7 +1145,7 @@ static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict
 
     /* load entropy tables */
     {   size_t const eSize = ZSTD_loadDEntropy(&dctx->entropy, dict, dictSize);
-        RETURN_ERROR_IF(ZSTD_isError(eSize), dictionary_corrupted);
+        RETURN_ERROR_IF(ZSTD_isError(eSize), dictionary_corrupted, "");
         dict = (const char*)dict + eSize;
         dictSize -= eSize;
     }
@@ -1171,11 +1180,11 @@ size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
 
 size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
 {
-    FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) );
+    FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) , "");
     if (dict && dictSize)
         RETURN_ERROR_IF(
             ZSTD_isError(ZSTD_decompress_insertDictionary(dctx, dict, dictSize)),
-            dictionary_corrupted);
+            dictionary_corrupted, "");
     return 0;
 }
 
@@ -1194,7 +1203,7 @@ size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
         DEBUGLOG(4, "DDict is %s",
                     dctx->ddictIsCold ? "~cold~" : "hot!");
     }
-    FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) );
+    FORWARD_IF_ERROR( ZSTD_decompressBegin(dctx) , "");
     if (ddict) {   /* NULL ddict is equivalent to no dictionary */
         ZSTD_copyDDictParameters(dctx, ddict);
     }
@@ -1285,11 +1294,11 @@ size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx,
                                          ZSTD_dictLoadMethod_e dictLoadMethod,
                                          ZSTD_dictContentType_e dictContentType)
 {
-    RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong);
+    RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
     ZSTD_clearDict(dctx);
     if (dict && dictSize != 0) {
         dctx->ddictLocal = ZSTD_createDDict_advanced(dict, dictSize, dictLoadMethod, dictContentType, dctx->customMem);
-        RETURN_ERROR_IF(dctx->ddictLocal == NULL, memory_allocation);
+        RETURN_ERROR_IF(dctx->ddictLocal == NULL, memory_allocation, "NULL pointer!");
         dctx->ddict = dctx->ddictLocal;
         dctx->dictUses = ZSTD_use_indefinitely;
     }
@@ -1308,7 +1317,7 @@ size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSi
 
 size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType)
 {
-    FORWARD_IF_ERROR(ZSTD_DCtx_loadDictionary_advanced(dctx, prefix, prefixSize, ZSTD_dlm_byRef, dictContentType));
+    FORWARD_IF_ERROR(ZSTD_DCtx_loadDictionary_advanced(dctx, prefix, prefixSize, ZSTD_dlm_byRef, dictContentType), "");
     dctx->dictUses = ZSTD_use_once;
     return 0;
 }
@@ -1325,8 +1334,8 @@ size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSiz
 size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize)
 {
     DEBUGLOG(4, "ZSTD_initDStream_usingDict");
-    FORWARD_IF_ERROR( ZSTD_DCtx_reset(zds, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_DCtx_loadDictionary(zds, dict, dictSize) );
+    FORWARD_IF_ERROR( ZSTD_DCtx_reset(zds, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_DCtx_loadDictionary(zds, dict, dictSize) , "");
     return ZSTD_startingInputLength(zds->format);
 }
 
@@ -1342,8 +1351,8 @@ size_t ZSTD_initDStream(ZSTD_DStream* zds)
  * this function cannot fail */
 size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* dctx, const ZSTD_DDict* ddict)
 {
-    FORWARD_IF_ERROR( ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only) );
-    FORWARD_IF_ERROR( ZSTD_DCtx_refDDict(dctx, ddict) );
+    FORWARD_IF_ERROR( ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only) , "");
+    FORWARD_IF_ERROR( ZSTD_DCtx_refDDict(dctx, ddict) , "");
     return ZSTD_startingInputLength(dctx->format);
 }
 
@@ -1352,14 +1361,14 @@ size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* dctx, const ZSTD_DDict* ddict)
  * this function cannot fail */
 size_t ZSTD_resetDStream(ZSTD_DStream* dctx)
 {
-    FORWARD_IF_ERROR(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
+    FORWARD_IF_ERROR(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only), "");
     return ZSTD_startingInputLength(dctx->format);
 }
 
 
 size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
 {
-    RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong);
+    RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
     ZSTD_clearDict(dctx);
     if (ddict) {
         dctx->ddict = ddict;
@@ -1376,9 +1385,9 @@ size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize)
     ZSTD_bounds const bounds = ZSTD_dParam_getBounds(ZSTD_d_windowLogMax);
     size_t const min = (size_t)1 << bounds.lowerBound;
     size_t const max = (size_t)1 << bounds.upperBound;
-    RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong);
-    RETURN_ERROR_IF(maxWindowSize < min, parameter_outOfBound);
-    RETURN_ERROR_IF(maxWindowSize > max, parameter_outOfBound);
+    RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
+    RETURN_ERROR_IF(maxWindowSize < min, parameter_outOfBound, "");
+    RETURN_ERROR_IF(maxWindowSize > max, parameter_outOfBound, "");
     dctx->maxWindowSize = maxWindowSize;
     return 0;
 }
@@ -1401,6 +1410,10 @@ ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam)
             bounds.upperBound = (int)ZSTD_f_zstd1_magicless;
             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;
+            return bounds;
         default:;
     }
     bounds.error = ERROR(parameter_unsupported);
@@ -1420,12 +1433,12 @@ static int ZSTD_dParam_withinBounds(ZSTD_dParameter dParam, int value)
 }
 
 #define CHECK_DBOUNDS(p,v) {                \
-    RETURN_ERROR_IF(!ZSTD_dParam_withinBounds(p, v), parameter_outOfBound); \
+    RETURN_ERROR_IF(!ZSTD_dParam_withinBounds(p, v), parameter_outOfBound, ""); \
 }
 
 size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter dParam, int value)
 {
-    RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong);
+    RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
     switch(dParam) {
         case ZSTD_d_windowLogMax:
             if (value == 0) value = ZSTD_WINDOWLOG_LIMIT_DEFAULT;
@@ -1436,9 +1449,13 @@ size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter dParam, int value
             CHECK_DBOUNDS(ZSTD_d_format, value);
             dctx->format = (ZSTD_format_e)value;
             return 0;
+        case ZSTD_d_stableOutBuffer:
+            CHECK_DBOUNDS(ZSTD_d_stableOutBuffer, value);
+            dctx->outBufferMode = (ZSTD_outBufferMode_e)value;
+            return 0;
         default:;
     }
-    RETURN_ERROR(parameter_unsupported);
+    RETURN_ERROR(parameter_unsupported, "");
 }
 
 size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset)
@@ -1450,7 +1467,7 @@ size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset)
     }
     if ( (reset == ZSTD_reset_parameters)
       || (reset == ZSTD_reset_session_and_parameters) ) {
-        RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong);
+        RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong, "");
         ZSTD_clearDict(dctx);
         dctx->format = ZSTD_f_zstd1;
         dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
@@ -1471,7 +1488,7 @@ size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long
     unsigned long long const neededSize = MIN(frameContentSize, neededRBSize);
     size_t const minRBSize = (size_t) neededSize;
     RETURN_ERROR_IF((unsigned long long)minRBSize != neededSize,
-                    frameParameter_windowTooLarge);
+                    frameParameter_windowTooLarge, "");
     return minRBSize;
 }
 
@@ -1489,24 +1506,84 @@ size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize)
     ZSTD_frameHeader zfh;
     size_t const err = ZSTD_getFrameHeader(&zfh, src, srcSize);
     if (ZSTD_isError(err)) return err;
-    RETURN_ERROR_IF(err>0, srcSize_wrong);
+    RETURN_ERROR_IF(err>0, srcSize_wrong, "");
     RETURN_ERROR_IF(zfh.windowSize > windowSizeMax,
-                    frameParameter_windowTooLarge);
+                    frameParameter_windowTooLarge, "");
     return ZSTD_estimateDStreamSize((size_t)zfh.windowSize);
 }
 
 
 /* *****   Decompression   ***** */
 
-MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+static int ZSTD_DCtx_isOverflow(ZSTD_DStream* zds, size_t const neededInBuffSize, size_t const neededOutBuffSize)
 {
-    size_t const length = MIN(dstCapacity, srcSize);
-    if (length > 0) {
-        memcpy(dst, src, length);
-    }
-    return length;
+    return (zds->inBuffSize + zds->outBuffSize) >= (neededInBuffSize + neededOutBuffSize) * ZSTD_WORKSPACETOOLARGE_FACTOR;
+}
+
+static void ZSTD_DCtx_updateOversizedDuration(ZSTD_DStream* zds, size_t const neededInBuffSize, size_t const neededOutBuffSize)
+{
+    if (ZSTD_DCtx_isOverflow(zds, neededInBuffSize, neededOutBuffSize))
+        zds->oversizedDuration++;
+    else 
+        zds->oversizedDuration = 0;
+}
+
+static int ZSTD_DCtx_isOversizedTooLong(ZSTD_DStream* zds)
+{
+    return zds->oversizedDuration >= ZSTD_WORKSPACETOOLARGE_MAXDURATION;
+}
+
+/* Checks that the output buffer hasn't changed if ZSTD_obm_stable is used. */
+static size_t ZSTD_checkOutBuffer(ZSTD_DStream const* zds, ZSTD_outBuffer const* output)
+{
+    ZSTD_outBuffer const expect = zds->expectedOutBuffer;
+    /* No requirement when ZSTD_obm_stable is not enabled. */
+    if (zds->outBufferMode != ZSTD_obm_stable)
+        return 0;
+    /* Any buffer is allowed in zdss_init, this must be the same for every other call until
+     * the context is reset.
+     */
+    if (zds->streamStage == zdss_init)
+        return 0;
+    /* 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!");
 }
 
+/* Calls ZSTD_decompressContinue() with the right parameters for ZSTD_decompressStream()
+ * and updates the stage and the output buffer state. This call is extracted so it can be
+ * used both when reading directly from the ZSTD_inBuffer, and in buffered input mode.
+ * NOTE: You must break after calling this function since the streamStage is modified.
+ */
+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) {
+        size_t const dstSize = isSkipFrame ? 0 : zds->outBuffSize - zds->outStart;
+        size_t const decodedSize = ZSTD_decompressContinue(zds,
+                zds->outBuff + zds->outStart, dstSize, src, srcSize);
+        FORWARD_IF_ERROR(decodedSize, "");
+        if (!decodedSize && !isSkipFrame) {
+            zds->streamStage = zdss_read;
+        } else {
+            zds->outEnd = zds->outStart + decodedSize;
+            zds->streamStage = zdss_flush;
+        }
+    } else {
+        /* Write directly into the output buffer */
+        size_t const dstSize = isSkipFrame ? 0 : 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);
+    }
+    return 0;
+}
 
 size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
 {
@@ -1532,6 +1609,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
         "forbidden. out: pos: %u   vs size: %u",
         (U32)output->pos, (U32)output->size);
     DEBUGLOG(5, "input size : %u", (U32)(input->size - input->pos));
+    FORWARD_IF_ERROR(ZSTD_checkOutBuffer(zds, output), "");
 
     while (someMoreWork) {
         switch(zds->streamStage)
@@ -1542,6 +1620,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
             zds->legacyVersion = 0;
             zds->hostageByte = 0;
+            zds->expectedOutBuffer = *output;
             /* fall-through */
 
         case zdss_loadHeader :
@@ -1569,7 +1648,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                             "legacy support is incompatible with static dctx");
                         FORWARD_IF_ERROR(ZSTD_initLegacyStream(&zds->legacyContext,
                                     zds->previousLegacyVersion, legacyVersion,
-                                    dict, dictSize));
+                                    dict, dictSize), "");
                         zds->legacyVersion = zds->previousLegacyVersion = legacyVersion;
                         {   size_t const hint = ZSTD_decompressLegacyStream(zds->legacyContext, legacyVersion, output, input);
                             if (hint==0) zds->streamStage = zdss_init;   /* or stay in stage zdss_loadHeader */
@@ -1596,7 +1675,8 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             }   }
 
             /* check for single-pass mode opportunity */
-            if (zds->fParams.frameContentSize && zds->fParams.windowSize /* skippable frame if == 0 */
+            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);
                 if (cSize <= (size_t)(iend-istart)) {
@@ -1612,15 +1692,23 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                     break;
             }   }
 
+            /* Check output buffer is large enough for ZSTD_odm_stable. */
+            if (zds->outBufferMode == ZSTD_obm_stable
+                && zds->fParams.frameType != ZSTD_skippableFrame
+                && zds->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN
+                && (U64)(size_t)(oend-op) < zds->fParams.frameContentSize) {
+                RETURN_ERROR(dstSize_tooSmall, "ZSTD_obm_stable passed but ZSTD_outBuffer is too small");
+            }
+
             /* Consume header (see ZSTDds_decodeFrameHeader) */
             DEBUGLOG(4, "Consume header");
-            FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(zds, ZSTD_getDDict(zds)));
+            FORWARD_IF_ERROR(ZSTD_decompressBegin_usingDDict(zds, ZSTD_getDDict(zds)), "");
 
             if ((MEM_readLE32(zds->headerBuffer) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {  /* skippable frame */
                 zds->expected = MEM_readLE32(zds->headerBuffer + ZSTD_FRAMEIDSIZE);
                 zds->stage = ZSTDds_skipFrame;
             } else {
-                FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(zds, zds->headerBuffer, zds->lhSize));
+                FORWARD_IF_ERROR(ZSTD_decodeFrameHeader(zds, zds->headerBuffer, zds->lhSize), "");
                 zds->expected = ZSTD_blockHeaderSize;
                 zds->stage = ZSTDds_decodeBlockHeader;
             }
@@ -1631,34 +1719,42 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                         (U32)(zds->maxWindowSize >> 10) );
             zds->fParams.windowSize = MAX(zds->fParams.windowSize, 1U << ZSTD_WINDOWLOG_ABSOLUTEMIN);
             RETURN_ERROR_IF(zds->fParams.windowSize > zds->maxWindowSize,
-                            frameParameter_windowTooLarge);
+                            frameParameter_windowTooLarge, "");
 
             /* Adapt buffer sizes to frame header instructions */
             {   size_t const neededInBuffSize = MAX(zds->fParams.blockSizeMax, 4 /* frame checksum */);
-                size_t const neededOutBuffSize = ZSTD_decodingBufferSize_min(zds->fParams.windowSize, zds->fParams.frameContentSize);
-                if ((zds->inBuffSize < neededInBuffSize) || (zds->outBuffSize < neededOutBuffSize)) {
-                    size_t const bufferSize = neededInBuffSize + neededOutBuffSize;
-                    DEBUGLOG(4, "inBuff  : from %u to %u",
-                                (U32)zds->inBuffSize, (U32)neededInBuffSize);
-                    DEBUGLOG(4, "outBuff : from %u to %u",
-                                (U32)zds->outBuffSize, (U32)neededOutBuffSize);
-                    if (zds->staticSize) {  /* static DCtx */
-                        DEBUGLOG(4, "staticSize : %u", (U32)zds->staticSize);
-                        assert(zds->staticSize >= sizeof(ZSTD_DCtx));  /* controlled at init */
-                        RETURN_ERROR_IF(
-                            bufferSize > zds->staticSize - sizeof(ZSTD_DCtx),
-                            memory_allocation);
-                    } else {
-                        ZSTD_free(zds->inBuff, zds->customMem);
-                        zds->inBuffSize = 0;
-                        zds->outBuffSize = 0;
-                        zds->inBuff = (char*)ZSTD_malloc(bufferSize, zds->customMem);
-                        RETURN_ERROR_IF(zds->inBuff == NULL, memory_allocation);
-                    }
-                    zds->inBuffSize = neededInBuffSize;
-                    zds->outBuff = zds->inBuff + zds->inBuffSize;
-                    zds->outBuffSize = neededOutBuffSize;
-            }   }
+                size_t const neededOutBuffSize = zds->outBufferMode == ZSTD_obm_buffered
+                        ? ZSTD_decodingBufferSize_min(zds->fParams.windowSize, zds->fParams.frameContentSize)
+                        : 0;
+
+                ZSTD_DCtx_updateOversizedDuration(zds, neededInBuffSize, neededOutBuffSize);
+
+                {   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",
+                                    (U32)zds->inBuffSize, (U32)neededInBuffSize);
+                        DEBUGLOG(4, "outBuff : from %u to %u",
+                                    (U32)zds->outBuffSize, (U32)neededOutBuffSize);
+                        if (zds->staticSize) {  /* static DCtx */
+                            DEBUGLOG(4, "staticSize : %u", (U32)zds->staticSize);
+                            assert(zds->staticSize >= sizeof(ZSTD_DCtx));  /* controlled at init */
+                            RETURN_ERROR_IF(
+                                bufferSize > zds->staticSize - sizeof(ZSTD_DCtx),
+                                memory_allocation, "");
+                        } else {
+                            ZSTD_free(zds->inBuff, zds->customMem);
+                            zds->inBuffSize = 0;
+                            zds->outBuffSize = 0;
+                            zds->inBuff = (char*)ZSTD_malloc(bufferSize, zds->customMem);
+                            RETURN_ERROR_IF(zds->inBuff == NULL, memory_allocation, "");
+                        }
+                        zds->inBuffSize = neededInBuffSize;
+                        zds->outBuff = zds->inBuff + zds->inBuffSize;
+                        zds->outBuffSize = neededOutBuffSize;
+            }   }   }
             zds->streamStage = zdss_read;
             /* fall-through */
 
@@ -1672,15 +1768,9 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                     break;
                 }
                 if ((size_t)(iend-ip) >= neededInSize) {  /* decode directly from src */
-                    int const isSkipFrame = ZSTD_isSkipFrame(zds);
-                    size_t const decodedSize = ZSTD_decompressContinue(zds,
-                        zds->outBuff + zds->outStart, (isSkipFrame ? 0 : zds->outBuffSize - zds->outStart),
-                        ip, neededInSize);
-                    if (ZSTD_isError(decodedSize)) return decodedSize;
+                    FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, ip, neededInSize), "");
                     ip += neededInSize;
-                    if (!decodedSize && !isSkipFrame) break;   /* this was just a header */
-                    zds->outEnd = zds->outStart + decodedSize;
-                    zds->streamStage = zdss_flush;
+                    /* Function modifies the stage so we must break */
                     break;
             }   }
             if (ip==iend) { someMoreWork = 0; break; }   /* no more input */
@@ -1707,17 +1797,11 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                 if (loadedSize < toLoad) { someMoreWork = 0; break; }   /* not enough input, wait for more */
 
                 /* decode loaded input */
-                {   size_t const decodedSize = ZSTD_decompressContinue(zds,
-                        zds->outBuff + zds->outStart, zds->outBuffSize - zds->outStart,
-                        zds->inBuff, neededInSize);
-                    if (ZSTD_isError(decodedSize)) return decodedSize;
-                    zds->inPos = 0;   /* input is consumed */
-                    if (!decodedSize && !isSkipFrame) { zds->streamStage = zdss_read; break; }   /* this was just a header */
-                    zds->outEnd = zds->outStart +  decodedSize;
-            }   }
-            zds->streamStage = zdss_flush;
-            /* fall-through */
-
+                zds->inPos = 0;   /* input is consumed */
+                FORWARD_IF_ERROR(ZSTD_decompressContinueStream(zds, &op, oend, zds->inBuff, neededInSize), "");
+                /* Function modifies the stage so we must break */
+                break;
+            }
         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);
@@ -1740,17 +1824,21 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
 
         default:
             assert(0);    /* impossible */
-            RETURN_ERROR(GENERIC);   /* some compiler require default to do something */
+            RETURN_ERROR(GENERIC, "impossible to reach");   /* some compiler require default to do something */
     }   }
 
     /* result */
     input->pos = (size_t)(ip - (const char*)(input->src));
     output->pos = (size_t)(op - (char*)(output->dst));
+
+    /* Update the expected output buffer for ZSTD_obm_stable. */
+    zds->expectedOutBuffer = *output;
+
     if ((ip==istart) && (op==ostart)) {  /* no forward progress */
         zds->noForwardProgress ++;
         if (zds->noForwardProgress >= ZSTD_NO_FORWARD_PROGRESS_MAX) {
-            RETURN_ERROR_IF(op==oend, dstSize_tooSmall);
-            RETURN_ERROR_IF(ip==iend, srcSize_wrong);
+            RETURN_ERROR_IF(op==oend, dstSize_tooSmall, "");
+            RETURN_ERROR_IF(ip==iend, srcSize_wrong, "");
             assert(0);
         }
     } else {
diff --git a/zstd/zstd_decompress_block.c b/zstd/zstd_decompress_block.c
index d4d23357..fcb381b9 100644
--- a/zstd/zstd_decompress_block.c
+++ b/zstd/zstd_decompress_block.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -56,7 +56,7 @@ static void ZSTD_copy4(void* dst, const void* src) { memcpy(dst, src, 4); }
 size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
                           blockProperties_t* bpPtr)
 {
-    RETURN_ERROR_IF(srcSize < ZSTD_blockHeaderSize, srcSize_wrong);
+    RETURN_ERROR_IF(srcSize < ZSTD_blockHeaderSize, srcSize_wrong, "");
 
     {   U32 const cBlockHeader = MEM_readLE24(src);
         U32 const cSize = cBlockHeader >> 3;
@@ -64,7 +64,7 @@ size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
         bpPtr->blockType = (blockType_e)((cBlockHeader >> 1) & 3);
         bpPtr->origSize = cSize;   /* only useful for RLE */
         if (bpPtr->blockType == bt_rle) return 1;
-        RETURN_ERROR_IF(bpPtr->blockType == bt_reserved, corruption_detected);
+        RETURN_ERROR_IF(bpPtr->blockType == bt_reserved, corruption_detected, "");
         return cSize;
     }
 }
@@ -80,7 +80,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
                           const void* src, size_t srcSize)   /* note : srcSize < BLOCKSIZE */
 {
     DEBUGLOG(5, "ZSTD_decodeLiteralsBlock");
-    RETURN_ERROR_IF(srcSize < MIN_CBLOCK_SIZE, corruption_detected);
+    RETURN_ERROR_IF(srcSize < MIN_CBLOCK_SIZE, corruption_detected, "");
 
     {   const BYTE* const istart = (const BYTE*) src;
         symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3);
@@ -89,7 +89,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
         {
         case set_repeat:
             DEBUGLOG(5, "set_repeat flag : re-using stats from previous compressed literals block");
-            RETURN_ERROR_IF(dctx->litEntropy==0, dictionary_corrupted);
+            RETURN_ERROR_IF(dctx->litEntropy==0, dictionary_corrupted, "");
             /* fall-through */
 
         case set_compressed:
@@ -121,8 +121,8 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
                     litCSize = (lhc >> 22) + ((size_t)istart[4] << 10);
                     break;
                 }
-                RETURN_ERROR_IF(litSize > ZSTD_BLOCKSIZE_MAX, corruption_detected);
-                RETURN_ERROR_IF(litCSize + lhSize > srcSize, corruption_detected);
+                RETURN_ERROR_IF(litSize > ZSTD_BLOCKSIZE_MAX, corruption_detected, "");
+                RETURN_ERROR_IF(litCSize + lhSize > srcSize, corruption_detected, "");
 
                 /* prefetch huffman table if cold */
                 if (dctx->ddictIsCold && (litSize > 768 /* heuristic */)) {
@@ -160,7 +160,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
                     }
                 }
 
-                RETURN_ERROR_IF(HUF_isError(hufSuccess), corruption_detected);
+                RETURN_ERROR_IF(HUF_isError(hufSuccess), corruption_detected, "");
 
                 dctx->litPtr = dctx->litBuffer;
                 dctx->litSize = litSize;
@@ -190,7 +190,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
                 }
 
                 if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) {  /* risk reading beyond src buffer with wildcopy */
-                    RETURN_ERROR_IF(litSize+lhSize > srcSize, corruption_detected);
+                    RETURN_ERROR_IF(litSize+lhSize > srcSize, corruption_detected, "");
                     memcpy(dctx->litBuffer, istart+lhSize, litSize);
                     dctx->litPtr = dctx->litBuffer;
                     dctx->litSize = litSize;
@@ -222,7 +222,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
                     RETURN_ERROR_IF(srcSize<4, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4");
                     break;
                 }
-                RETURN_ERROR_IF(litSize > ZSTD_BLOCKSIZE_MAX, corruption_detected);
+                RETURN_ERROR_IF(litSize > ZSTD_BLOCKSIZE_MAX, corruption_detected, "");
                 memset(dctx->litBuffer, istart[lhSize], litSize + WILDCOPY_OVERLENGTH);
                 dctx->litPtr = dctx->litBuffer;
                 dctx->litSize = litSize;
@@ -440,8 +440,8 @@ static size_t ZSTD_buildSeqTable(ZSTD_seqSymbol* DTableSpace, const ZSTD_seqSymb
     switch(type)
     {
     case set_rle :
-        RETURN_ERROR_IF(!srcSize, srcSize_wrong);
-        RETURN_ERROR_IF((*(const BYTE*)src) > max, corruption_detected);
+        RETURN_ERROR_IF(!srcSize, srcSize_wrong, "");
+        RETURN_ERROR_IF((*(const BYTE*)src) > max, corruption_detected, "");
         {   U32 const symbol = *(const BYTE*)src;
             U32 const baseline = baseValue[symbol];
             U32 const nbBits = nbAdditionalBits[symbol];
@@ -453,7 +453,7 @@ static size_t ZSTD_buildSeqTable(ZSTD_seqSymbol* DTableSpace, const ZSTD_seqSymb
         *DTablePtr = defaultTable;
         return 0;
     case set_repeat:
-        RETURN_ERROR_IF(!flagRepeatTable, corruption_detected);
+        RETURN_ERROR_IF(!flagRepeatTable, corruption_detected, "");
         /* prefetch FSE table if used */
         if (ddictIsCold && (nbSeq > 24 /* heuristic */)) {
             const void* const pStart = *DTablePtr;
@@ -465,8 +465,8 @@ static size_t ZSTD_buildSeqTable(ZSTD_seqSymbol* DTableSpace, const ZSTD_seqSymb
         {   unsigned tableLog;
             S16 norm[MaxSeq+1];
             size_t const headerSize = FSE_readNCount(norm, &max, &tableLog, src, srcSize);
-            RETURN_ERROR_IF(FSE_isError(headerSize), corruption_detected);
-            RETURN_ERROR_IF(tableLog > maxLog, corruption_detected);
+            RETURN_ERROR_IF(FSE_isError(headerSize), corruption_detected, "");
+            RETURN_ERROR_IF(tableLog > maxLog, corruption_detected, "");
             ZSTD_buildFSETable(DTableSpace, norm, max, baseValue, nbAdditionalBits, tableLog);
             *DTablePtr = DTableSpace;
             return headerSize;
@@ -487,28 +487,28 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
     DEBUGLOG(5, "ZSTD_decodeSeqHeaders");
 
     /* check */
-    RETURN_ERROR_IF(srcSize < MIN_SEQUENCES_SIZE, srcSize_wrong);
+    RETURN_ERROR_IF(srcSize < MIN_SEQUENCES_SIZE, srcSize_wrong, "");
 
     /* SeqHead */
     nbSeq = *ip++;
     if (!nbSeq) {
         *nbSeqPtr=0;
-        RETURN_ERROR_IF(srcSize != 1, srcSize_wrong);
+        RETURN_ERROR_IF(srcSize != 1, srcSize_wrong, "");
         return 1;
     }
     if (nbSeq > 0x7F) {
         if (nbSeq == 0xFF) {
-            RETURN_ERROR_IF(ip+2 > iend, srcSize_wrong);
+            RETURN_ERROR_IF(ip+2 > iend, srcSize_wrong, "");
             nbSeq = MEM_readLE16(ip) + LONGNBSEQ, ip+=2;
         } else {
-            RETURN_ERROR_IF(ip >= iend, srcSize_wrong);
+            RETURN_ERROR_IF(ip >= iend, srcSize_wrong, "");
             nbSeq = ((nbSeq-0x80)<<8) + *ip++;
         }
     }
     *nbSeqPtr = nbSeq;
 
     /* FSE table descriptors */
-    RETURN_ERROR_IF(ip+1 > iend, srcSize_wrong); /* minimum possible size: 1 byte for symbol encoding types */
+    RETURN_ERROR_IF(ip+1 > iend, srcSize_wrong, ""); /* minimum possible size: 1 byte for symbol encoding types */
     {   symbolEncodingType_e const LLtype = (symbolEncodingType_e)(*ip >> 6);
         symbolEncodingType_e const OFtype = (symbolEncodingType_e)((*ip >> 4) & 3);
         symbolEncodingType_e const MLtype = (symbolEncodingType_e)((*ip >> 2) & 3);
@@ -521,7 +521,7 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
                                                       LL_base, LL_bits,
                                                       LL_defaultDTable, dctx->fseEntropy,
                                                       dctx->ddictIsCold, nbSeq);
-            RETURN_ERROR_IF(ZSTD_isError(llhSize), corruption_detected);
+            RETURN_ERROR_IF(ZSTD_isError(llhSize), corruption_detected, "ZSTD_buildSeqTable failed");
             ip += llhSize;
         }
 
@@ -531,7 +531,7 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
                                                       OF_base, OF_bits,
                                                       OF_defaultDTable, dctx->fseEntropy,
                                                       dctx->ddictIsCold, nbSeq);
-            RETURN_ERROR_IF(ZSTD_isError(ofhSize), corruption_detected);
+            RETURN_ERROR_IF(ZSTD_isError(ofhSize), corruption_detected, "ZSTD_buildSeqTable failed");
             ip += ofhSize;
         }
 
@@ -541,7 +541,7 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
                                                       ML_base, ML_bits,
                                                       ML_defaultDTable, dctx->fseEntropy,
                                                       dctx->ddictIsCold, nbSeq);
-            RETURN_ERROR_IF(ZSTD_isError(mlhSize), corruption_detected);
+            RETURN_ERROR_IF(ZSTD_isError(mlhSize), corruption_detected, "ZSTD_buildSeqTable failed");
             ip += mlhSize;
         }
     }
@@ -665,15 +665,15 @@ size_t ZSTD_execSequenceEnd(BYTE* op,
 {
     BYTE* const oLitEnd = op + sequence.litLength;
     size_t const sequenceLength = sequence.litLength + sequence.matchLength;
-    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
     const BYTE* const iLitEnd = *litPtr + sequence.litLength;
     const BYTE* match = oLitEnd - sequence.offset;
     BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
 
-    /* bounds checks */
-    assert(oLitEnd < oMatchEnd);
-    RETURN_ERROR_IF(oMatchEnd > oend, dstSize_tooSmall, "last match must fit within dstBuffer");
-    RETURN_ERROR_IF(iLitEnd > litLimit, corruption_detected, "try to read beyond literal buffer");
+    /* bounds checks : careful of address space overflow in 32-bit mode */
+    RETURN_ERROR_IF(sequenceLength > (size_t)(oend - op), dstSize_tooSmall, "last match must fit within dstBuffer");
+    RETURN_ERROR_IF(sequence.litLength > (size_t)(litLimit - *litPtr), corruption_detected, "try to read beyond literal buffer");
+    assert(op < op + sequenceLength);
+    assert(oLitEnd < op + sequenceLength);
 
     /* copy literals */
     ZSTD_safecopy(op, oend_w, *litPtr, sequence.litLength, ZSTD_no_overlap);
@@ -683,7 +683,7 @@ size_t ZSTD_execSequenceEnd(BYTE* op,
     /* copy Match */
     if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
         /* offset beyond prefix */
-        RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - virtualStart), corruption_detected);
+        RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - virtualStart), corruption_detected, "");
         match = dictEnd - (prefixStart-match);
         if (match + sequence.matchLength <= dictEnd) {
             memmove(oLitEnd, match, sequence.matchLength);
@@ -709,16 +709,27 @@ size_t ZSTD_execSequence(BYTE* op,
     BYTE* const oLitEnd = op + sequence.litLength;
     size_t const sequenceLength = sequence.litLength + sequence.matchLength;
     BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
-    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
+    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;   /* risk : address space underflow on oend=NULL */
     const BYTE* const iLitEnd = *litPtr + sequence.litLength;
     const BYTE* match = oLitEnd - sequence.offset;
 
-    /* Errors and uncommon cases handled here. */
-    assert(oLitEnd < oMatchEnd);
-    if (UNLIKELY(iLitEnd > litLimit || oMatchEnd > oend_w))
+    assert(op != NULL /* Precondition */);
+    assert(oend_w < oend /* No underflow */);
+    /* Handle edge cases in a slow path:
+     *   - Read beyond end of literals
+     *   - Match end is within WILDCOPY_OVERLIMIT of oend
+     *   - 32-bit mode and the match length overflows
+     */
+    if (UNLIKELY(
+            iLitEnd > litLimit ||
+            oMatchEnd > oend_w ||
+            (MEM_32bits() && (size_t)(oend - op) < sequenceLength + WILDCOPY_OVERLENGTH)))
         return ZSTD_execSequenceEnd(op, oend, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd);
 
     /* Assumptions (everything else goes into ZSTD_execSequenceEnd()) */
+    assert(op <= oLitEnd /* No overflow */);
+    assert(oLitEnd < oMatchEnd /* Non-zero match & no overflow */);
+    assert(oMatchEnd <= oend /* No underflow */);
     assert(iLitEnd <= litLimit /* Literal length is in bounds */);
     assert(oLitEnd <= oend_w /* Can wildcopy literals */);
     assert(oMatchEnd <= oend_w /* Can wildcopy matches */);
@@ -738,7 +749,7 @@ size_t ZSTD_execSequence(BYTE* op,
     /* Copy Match */
     if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
         /* offset beyond prefix -> go into extDict */
-        RETURN_ERROR_IF(UNLIKELY(sequence.offset > (size_t)(oLitEnd - virtualStart)), corruption_detected);
+        RETURN_ERROR_IF(UNLIKELY(sequence.offset > (size_t)(oLitEnd - virtualStart)), corruption_detected, "");
         match = dictEnd + (match - prefixStart);
         if (match + sequence.matchLength <= dictEnd) {
             memmove(oLitEnd, match, sequence.matchLength);
@@ -936,13 +947,56 @@ ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets, c
     return seq;
 }
 
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+static int ZSTD_dictionaryIsActive(ZSTD_DCtx const* dctx, BYTE const* prefixStart, BYTE const* oLitEnd)
+{
+    size_t const windowSize = dctx->fParams.windowSize;
+    /* No dictionary used. */
+    if (dctx->dictContentEndForFuzzing == NULL) return 0;
+    /* Dictionary is our prefix. */
+    if (prefixStart == dctx->dictContentBeginForFuzzing) return 1;
+    /* Dictionary is not our ext-dict. */
+    if (dctx->dictEnd != dctx->dictContentEndForFuzzing) return 0;
+    /* Dictionary is not within our window size. */
+    if ((size_t)(oLitEnd - prefixStart) >= windowSize) return 0;
+    /* Dictionary is active. */
+    return 1;
+}
+
+MEM_STATIC void ZSTD_assertValidSequence(
+        ZSTD_DCtx const* dctx,
+        BYTE const* op, BYTE const* oend,
+        seq_t const seq,
+        BYTE const* prefixStart, BYTE const* virtualStart)
+{
+    size_t const windowSize = dctx->fParams.windowSize;
+    size_t const sequenceSize = seq.litLength + seq.matchLength;
+    BYTE const* const oLitEnd = op + seq.litLength;
+    DEBUGLOG(6, "Checking sequence: litL=%u matchL=%u offset=%u",
+            (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset);
+    assert(op <= oend);
+    assert((size_t)(oend - op) >= sequenceSize);
+    assert(sequenceSize <= ZSTD_BLOCKSIZE_MAX);
+    if (ZSTD_dictionaryIsActive(dctx, prefixStart, oLitEnd)) {
+        size_t const dictSize = (size_t)((char const*)dctx->dictContentEndForFuzzing - (char const*)dctx->dictContentBeginForFuzzing);
+        /* Offset must be within the dictionary. */
+        assert(seq.offset <= (size_t)(oLitEnd - virtualStart));
+        assert(seq.offset <= windowSize + dictSize);
+    } else {
+        /* Offset must be within our window. */
+        assert(seq.offset <= windowSize);
+    }
+}
+#endif
+
 #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
 FORCE_INLINE_TEMPLATE size_t
 DONT_VECTORIZE
 ZSTD_decompressSequences_body( ZSTD_DCtx* dctx,
                                void* dst, size_t maxDstSize,
                          const void* seqStart, size_t seqSize, int nbSeq,
-                         const ZSTD_longOffset_e isLongOffset)
+                         const ZSTD_longOffset_e isLongOffset,
+                         const int frame)
 {
     const BYTE* ip = (const BYTE*)seqStart;
     const BYTE* const iend = ip + seqSize;
@@ -955,6 +1009,7 @@ ZSTD_decompressSequences_body( ZSTD_DCtx* dctx,
     const BYTE* const vBase = (const BYTE*) (dctx->virtualStart);
     const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
     DEBUGLOG(5, "ZSTD_decompressSequences_body");
+    (void)frame;
 
     /* Regen sequences */
     if (nbSeq) {
@@ -964,19 +1019,65 @@ ZSTD_decompressSequences_body( ZSTD_DCtx* dctx,
         { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) seqState.prevOffset[i] = dctx->entropy.rep[i]; }
         RETURN_ERROR_IF(
             ERR_isError(BIT_initDStream(&seqState.DStream, ip, iend-ip)),
-            corruption_detected);
+            corruption_detected, "");
         ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
         ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
         ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
+        assert(dst != NULL);
 
         ZSTD_STATIC_ASSERT(
                 BIT_DStream_unfinished < BIT_DStream_completed &&
                 BIT_DStream_endOfBuffer < BIT_DStream_completed &&
                 BIT_DStream_completed < BIT_DStream_overflow);
 
+#if defined(__GNUC__) && defined(__x86_64__)
+        /* Align the decompression loop to 32 + 16 bytes.
+         *
+         * zstd compiled with gcc-9 on an Intel i9-9900k shows 10% decompression
+         * speed swings based on the alignment of the decompression loop. This
+         * performance swing is caused by parts of the decompression loop falling
+         * out of the DSB. The entire decompression loop should fit in the DSB,
+         * when it can't we get much worse performance. You can measure if you've
+         * hit the good case or the bad case with this perf command for some
+         * compressed file test.zst:
+         *
+         *   perf stat -e cycles -e instructions -e idq.all_dsb_cycles_any_uops \
+         *             -e idq.all_mite_cycles_any_uops -- ./zstd -tq test.zst
+         *
+         * If you see most cycles served out of the MITE you've hit the bad case.
+         * If you see most cycles served out of the DSB you've hit the good case.
+         * If it is pretty even then you may be in an okay case.
+         *
+         * I've been able to reproduce this issue on the following CPUs:
+         *   - Kabylake: Macbook Pro (15-inch, 2019) 2.4 GHz Intel Core i9
+         *               Use Instruments->Counters to get DSB/MITE cycles.
+         *               I never got performance swings, but I was able to
+         *               go from the good case of mostly DSB to half of the
+         *               cycles served from MITE.
+         *   - Coffeelake: Intel i9-9900k
+         *
+         * I haven't been able to reproduce the instability or DSB misses on any
+         * of the following CPUS:
+         *   - Haswell
+         *   - Broadwell: Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GH
+         *   - Skylake
+         *
+         * If you are seeing performance stability this script can help test.
+         * It tests on 4 commits in zstd where I saw performance change.
+         *
+         *   https://gist.github.com/terrelln/9889fc06a423fd5ca6e99351564473f4
+         */
+        __asm__(".p2align 5");
+        __asm__("nop");
+        __asm__(".p2align 4");
+#endif
         for ( ; ; ) {
             seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset, ZSTD_p_noPrefetch);
             size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, prefixStart, vBase, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+            assert(!ZSTD_isError(oneSeqSize));
+            if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequence, prefixStart, vBase);
+#endif
             DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
             BIT_reloadDStream(&(seqState.DStream));
             /* gcc and clang both don't like early returns in this loop.
@@ -993,17 +1094,19 @@ ZSTD_decompressSequences_body( ZSTD_DCtx* dctx,
         /* check if reached exact end */
         DEBUGLOG(5, "ZSTD_decompressSequences_body: after decode loop, remaining nbSeq : %i", nbSeq);
         if (ZSTD_isError(error)) return error;
-        RETURN_ERROR_IF(nbSeq, corruption_detected);
-        RETURN_ERROR_IF(BIT_reloadDStream(&seqState.DStream) < BIT_DStream_completed, corruption_detected);
+        RETURN_ERROR_IF(nbSeq, corruption_detected, "");
+        RETURN_ERROR_IF(BIT_reloadDStream(&seqState.DStream) < BIT_DStream_completed, corruption_detected, "");
         /* save reps for next block */
         { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); }
     }
 
     /* last literal segment */
     {   size_t const lastLLSize = litEnd - litPtr;
-        RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall);
-        memcpy(op, litPtr, lastLLSize);
-        op += lastLLSize;
+        RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, "");
+        if (op != NULL) {
+            memcpy(op, litPtr, lastLLSize);
+            op += lastLLSize;
+        }
     }
 
     return op-ostart;
@@ -1013,9 +1116,10 @@ static size_t
 ZSTD_decompressSequences_default(ZSTD_DCtx* dctx,
                                  void* dst, size_t maxDstSize,
                            const void* seqStart, size_t seqSize, int nbSeq,
-                           const ZSTD_longOffset_e isLongOffset)
+                           const ZSTD_longOffset_e isLongOffset,
+                           const int frame)
 {
-    return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+    return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
 }
 #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
 
@@ -1025,7 +1129,8 @@ ZSTD_decompressSequencesLong_body(
                                ZSTD_DCtx* dctx,
                                void* dst, size_t maxDstSize,
                          const void* seqStart, size_t seqSize, int nbSeq,
-                         const ZSTD_longOffset_e isLongOffset)
+                         const ZSTD_longOffset_e isLongOffset,
+                         const int frame)
 {
     const BYTE* ip = (const BYTE*)seqStart;
     const BYTE* const iend = ip + seqSize;
@@ -1037,6 +1142,7 @@ ZSTD_decompressSequencesLong_body(
     const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
     const BYTE* const dictStart = (const BYTE*) (dctx->virtualStart);
     const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
+    (void)frame;
 
     /* Regen sequences */
     if (nbSeq) {
@@ -1052,10 +1158,11 @@ ZSTD_decompressSequencesLong_body(
         seqState.prefixStart = prefixStart;
         seqState.pos = (size_t)(op-prefixStart);
         seqState.dictEnd = dictEnd;
+        assert(dst != NULL);
         assert(iend >= ip);
         RETURN_ERROR_IF(
             ERR_isError(BIT_initDStream(&seqState.DStream, ip, iend-ip)),
-            corruption_detected);
+            corruption_detected, "");
         ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
         ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
         ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
@@ -1065,23 +1172,31 @@ ZSTD_decompressSequencesLong_body(
             sequences[seqNb] = ZSTD_decodeSequence(&seqState, isLongOffset, ZSTD_p_prefetch);
             PREFETCH_L1(sequences[seqNb].match); PREFETCH_L1(sequences[seqNb].match + sequences[seqNb].matchLength - 1); /* note : it's safe to invoke PREFETCH() on any memory address, including invalid ones */
         }
-        RETURN_ERROR_IF(seqNb<seqAdvance, corruption_detected);
+        RETURN_ERROR_IF(seqNb<seqAdvance, corruption_detected, "");
 
         /* decode and decompress */
         for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && (seqNb<nbSeq) ; seqNb++) {
             seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset, ZSTD_p_prefetch);
             size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequences[(seqNb-ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litEnd, prefixStart, dictStart, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+            assert(!ZSTD_isError(oneSeqSize));
+            if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequences[(seqNb-ADVANCED_SEQS) & STORED_SEQS_MASK], prefixStart, dictStart);
+#endif
             if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
             PREFETCH_L1(sequence.match); PREFETCH_L1(sequence.match + sequence.matchLength - 1); /* note : it's safe to invoke PREFETCH() on any memory address, including invalid ones */
             sequences[seqNb & STORED_SEQS_MASK] = sequence;
             op += oneSeqSize;
         }
-        RETURN_ERROR_IF(seqNb<nbSeq, corruption_detected);
+        RETURN_ERROR_IF(seqNb<nbSeq, corruption_detected, "");
 
         /* finish queue */
         seqNb -= seqAdvance;
         for ( ; seqNb<nbSeq ; seqNb++) {
             size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequences[seqNb&STORED_SEQS_MASK], &litPtr, litEnd, prefixStart, dictStart, dictEnd);
+#if defined(FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION) && defined(FUZZING_ASSERT_VALID_SEQUENCE)
+            assert(!ZSTD_isError(oneSeqSize));
+            if (frame) ZSTD_assertValidSequence(dctx, op, oend, sequences[seqNb&STORED_SEQS_MASK], prefixStart, dictStart);
+#endif
             if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
             op += oneSeqSize;
         }
@@ -1092,9 +1207,11 @@ ZSTD_decompressSequencesLong_body(
 
     /* last literal segment */
     {   size_t const lastLLSize = litEnd - litPtr;
-        RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall);
-        memcpy(op, litPtr, lastLLSize);
-        op += lastLLSize;
+        RETURN_ERROR_IF(lastLLSize > (size_t)(oend-op), dstSize_tooSmall, "");
+        if (op != NULL) {
+            memcpy(op, litPtr, lastLLSize);
+            op += lastLLSize;
+        }
     }
 
     return op-ostart;
@@ -1104,9 +1221,10 @@ static size_t
 ZSTD_decompressSequencesLong_default(ZSTD_DCtx* dctx,
                                  void* dst, size_t maxDstSize,
                            const void* seqStart, size_t seqSize, int nbSeq,
-                           const ZSTD_longOffset_e isLongOffset)
+                           const ZSTD_longOffset_e isLongOffset,
+                           const int frame)
 {
-    return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+    return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
 }
 #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
 
@@ -1120,9 +1238,10 @@ DONT_VECTORIZE
 ZSTD_decompressSequences_bmi2(ZSTD_DCtx* dctx,
                                  void* dst, size_t maxDstSize,
                            const void* seqStart, size_t seqSize, int nbSeq,
-                           const ZSTD_longOffset_e isLongOffset)
+                           const ZSTD_longOffset_e isLongOffset,
+                           const int frame)
 {
-    return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+    return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
 }
 #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
 
@@ -1131,9 +1250,10 @@ static TARGET_ATTRIBUTE("bmi2") size_t
 ZSTD_decompressSequencesLong_bmi2(ZSTD_DCtx* dctx,
                                  void* dst, size_t maxDstSize,
                            const void* seqStart, size_t seqSize, int nbSeq,
-                           const ZSTD_longOffset_e isLongOffset)
+                           const ZSTD_longOffset_e isLongOffset,
+                           const int frame)
 {
-    return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+    return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
 }
 #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
 
@@ -1143,21 +1263,23 @@ typedef size_t (*ZSTD_decompressSequences_t)(
                             ZSTD_DCtx* dctx,
                             void* dst, size_t maxDstSize,
                             const void* seqStart, size_t seqSize, int nbSeq,
-                            const ZSTD_longOffset_e isLongOffset);
+                            const ZSTD_longOffset_e isLongOffset,
+                            const int frame);
 
 #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
 static size_t
 ZSTD_decompressSequences(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize,
                    const void* seqStart, size_t seqSize, int nbSeq,
-                   const ZSTD_longOffset_e isLongOffset)
+                   const ZSTD_longOffset_e isLongOffset,
+                   const int frame)
 {
     DEBUGLOG(5, "ZSTD_decompressSequences");
 #if DYNAMIC_BMI2
     if (dctx->bmi2) {
-        return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+        return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
     }
 #endif
-  return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+  return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
 }
 #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
 
@@ -1172,15 +1294,16 @@ static size_t
 ZSTD_decompressSequencesLong(ZSTD_DCtx* dctx,
                              void* dst, size_t maxDstSize,
                              const void* seqStart, size_t seqSize, int nbSeq,
-                             const ZSTD_longOffset_e isLongOffset)
+                             const ZSTD_longOffset_e isLongOffset,
+                             const int frame)
 {
     DEBUGLOG(5, "ZSTD_decompressSequencesLong");
 #if DYNAMIC_BMI2
     if (dctx->bmi2) {
-        return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+        return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
     }
 #endif
-  return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+  return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
 }
 #endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
 
@@ -1214,7 +1337,6 @@ ZSTD_getLongOffsetsShare(const ZSTD_seqSymbol* offTable)
 }
 #endif
 
-
 size_t
 ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
                               void* dst, size_t dstCapacity,
@@ -1230,7 +1352,7 @@ ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
     ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || (dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN))));
     DEBUGLOG(5, "ZSTD_decompressBlock_internal (size : %u)", (U32)srcSize);
 
-    RETURN_ERROR_IF(srcSize >= ZSTD_BLOCKSIZE_MAX, srcSize_wrong);
+    RETURN_ERROR_IF(srcSize >= ZSTD_BLOCKSIZE_MAX, srcSize_wrong, "");
 
     /* Decode literals section */
     {   size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize);
@@ -1256,6 +1378,8 @@ ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
         ip += seqHSize;
         srcSize -= seqHSize;
 
+        RETURN_ERROR_IF(dst == NULL && nbSeq > 0, dstSize_tooSmall, "NULL not handled");
+
 #if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
     !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
         if ( !usePrefetchDecoder
@@ -1274,12 +1398,12 @@ ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
         if (usePrefetchDecoder)
 #endif
 #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
-            return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
+            return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset, frame);
 #endif
 
 #ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
         /* else */
-        return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
+        return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset, frame);
 #endif
     }
 }
diff --git a/zstd/zstd_decompress_block.h b/zstd/zstd_decompress_block.h
index 7e929604..a065f8d9 100644
--- a/zstd/zstd_decompress_block.h
+++ b/zstd/zstd_decompress_block.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstd_decompress_internal.h b/zstd/zstd_decompress_internal.h
index 99eab854..a5cafc04 100644
--- a/zstd/zstd_decompress_internal.h
+++ b/zstd/zstd_decompress_internal.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -95,6 +95,11 @@ typedef enum {
     ZSTD_use_once = 1            /* Use the dictionary once and set to ZSTD_dont_use */
 } ZSTD_dictUses_e;
 
+typedef enum {
+    ZSTD_obm_buffered = 0,  /* Buffer the output */
+    ZSTD_obm_stable = 1     /* ZSTD_outBuffer is stable */
+} ZSTD_outBufferMode_e;
+
 struct ZSTD_DCtx_s
 {
     const ZSTD_seqSymbol* LLTptr;
@@ -147,10 +152,19 @@ struct ZSTD_DCtx_s
     U32 legacyVersion;
     U32 hostageByte;
     int noForwardProgress;
+    ZSTD_outBufferMode_e outBufferMode;
+    ZSTD_outBuffer expectedOutBuffer;
 
     /* workspace */
     BYTE litBuffer[ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH];
     BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
+
+    size_t oversizedDuration;
+
+#ifdef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
+    void const* dictContentBeginForFuzzing;
+    void const* dictContentEndForFuzzing;
+#endif
 };  /* typedef'd to ZSTD_DCtx within "zstd.h" */
 
 
diff --git a/zstd/zstd_double_fast.c b/zstd/zstd_double_fast.c
index 2e657f7c..27eed66c 100644
--- a/zstd/zstd_double_fast.c
+++ b/zstd/zstd_double_fast.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -63,10 +63,8 @@ size_t ZSTD_compressBlock_doubleFast_generic(
     const BYTE* ip = istart;
     const BYTE* anchor = istart;
     const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
-    const U32 lowestValid = ms->window.dictLimit;
-    const U32 maxDistance = 1U << cParams->windowLog;
     /* presumes that, if there is a dictionary, it must be using Attach mode */
-    const U32 prefixLowestIndex = (endIndex - lowestValid > maxDistance) ? endIndex - maxDistance : lowestValid;
+    const U32 prefixLowestIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
     const BYTE* const prefixLowest = base + prefixLowestIndex;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - HASH_READ_SIZE;
@@ -104,13 +102,15 @@ size_t ZSTD_compressBlock_doubleFast_generic(
 
     /* if a dictionary is attached, it must be within window range */
     if (dictMode == ZSTD_dictMatchState) {
-        assert(lowestValid + maxDistance >= endIndex);
+        assert(ms->window.dictLimit + (1U << cParams->windowLog) >= endIndex);
     }
 
     /* init */
     ip += (dictAndPrefixLength == 0);
     if (dictMode == ZSTD_noDict) {
-        U32 const maxRep = (U32)(ip - prefixLowest);
+        U32 const current = (U32)(ip - base);
+        U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, cParams->windowLog);
+        U32 const maxRep = current - windowLow;
         if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
         if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
     }
@@ -198,6 +198,9 @@ size_t ZSTD_compressBlock_doubleFast_generic(
         }   }
 
         ip += ((ip-anchor) >> kSearchStrength) + 1;
+#if defined(__aarch64__)
+        PREFETCH_L1(ip+256);
+#endif
         continue;
 
 _search_next_long:
diff --git a/zstd/zstd_double_fast.h b/zstd/zstd_double_fast.h
index 4fa31acf..73a2002f 100644
--- a/zstd/zstd_double_fast.h
+++ b/zstd/zstd_double_fast.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstd_errors.h b/zstd/zstd_errors.h
index 92a34338..998398e7 100644
--- a/zstd/zstd_errors.h
+++ b/zstd/zstd_errors.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -76,6 +76,7 @@ typedef enum {
   /* following error codes are __NOT STABLE__, they can be removed or changed in future versions */
   ZSTD_error_frameIndex_tooLarge = 100,
   ZSTD_error_seekableIO          = 102,
+  ZSTD_error_dstBuffer_wrong     = 104,
   ZSTD_error_maxCode = 120  /* never EVER use this value directly, it can change in future versions! Use ZSTD_isError() instead */
 } ZSTD_ErrorCode;
 
diff --git a/zstd/zstd_fast.c b/zstd/zstd_fast.c
index 6dbefee6..85a3a7a9 100644
--- a/zstd/zstd_fast.c
+++ b/zstd/zstd_fast.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -61,9 +61,7 @@ ZSTD_compressBlock_fast_generic(
     const BYTE* ip1;
     const BYTE* anchor = istart;
     const U32   endIndex = (U32)((size_t)(istart - base) + srcSize);
-    const U32   maxDistance = 1U << cParams->windowLog;
-    const U32   validStartIndex = ms->window.dictLimit;
-    const U32   prefixStartIndex = (endIndex - validStartIndex > maxDistance) ? endIndex - maxDistance : validStartIndex;
+    const U32   prefixStartIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
     const BYTE* const prefixStart = base + prefixStartIndex;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - HASH_READ_SIZE;
@@ -74,12 +72,21 @@ ZSTD_compressBlock_fast_generic(
     DEBUGLOG(5, "ZSTD_compressBlock_fast_generic");
     ip0 += (ip0 == prefixStart);
     ip1 = ip0 + 1;
-    {   U32 const maxRep = (U32)(ip0 - prefixStart);
+    {   U32 const current = (U32)(ip0 - base);
+        U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, cParams->windowLog);
+        U32 const maxRep = current - windowLow;
         if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
         if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
     }
 
     /* Main Search Loop */
+#ifdef __INTEL_COMPILER
+    /* From intel 'The vector pragma indicates that the loop should be
+     * vectorized if it is legal to do so'. Can be used together with
+     * #pragma ivdep (but have opted to exclude that because intel
+     * warns against using it).*/
+    #pragma vector always
+#endif
     while (ip1 < ilimit) {   /* < instead of <=, because check at ip0+2 */
         size_t mLength;
         BYTE const* ip2 = ip0 + 2;
@@ -91,19 +98,25 @@ ZSTD_compressBlock_fast_generic(
         U32 const current1 = (U32)(ip1-base);
         U32 const matchIndex0 = hashTable[h0];
         U32 const matchIndex1 = hashTable[h1];
-        BYTE const* repMatch = ip2-offset_1;
+        BYTE const* repMatch = ip2 - offset_1;
         const BYTE* match0 = base + matchIndex0;
         const BYTE* match1 = base + matchIndex1;
         U32 offcode;
+
+#if defined(__aarch64__)
+        PREFETCH_L1(ip0+256);
+#endif
+
         hashTable[h0] = current0;   /* update hash table */
         hashTable[h1] = current1;   /* update hash table */
 
         assert(ip0 + 1 == ip1);
 
         if ((offset_1 > 0) & (MEM_read32(repMatch) == MEM_read32(ip2))) {
-            mLength = ip2[-1] == repMatch[-1] ? 1 : 0;
+            mLength = (ip2[-1] == repMatch[-1]) ? 1 : 0;
             ip0 = ip2 - mLength;
             match0 = repMatch - mLength;
+            mLength += 4;
             offcode = 0;
             goto _match;
         }
@@ -128,19 +141,18 @@ _offset: /* Requires: ip0, match0 */
         offset_2 = offset_1;
         offset_1 = (U32)(ip0-match0);
         offcode = offset_1 + ZSTD_REP_MOVE;
-        mLength = 0;
+        mLength = 4;
         /* Count the backwards match length */
         while (((ip0>anchor) & (match0>prefixStart))
              && (ip0[-1] == match0[-1])) { ip0--; match0--; mLength++; } /* catch up */
 
 _match: /* Requires: ip0, match0, offcode */
         /* Count the forward length */
-        mLength += ZSTD_count(ip0+mLength+4, match0+mLength+4, iend) + 4;
+        mLength += ZSTD_count(ip0+mLength, match0+mLength, iend);
         ZSTD_storeSeq(seqStore, (size_t)(ip0-anchor), anchor, iend, offcode, mLength-MINMATCH);
         /* match found */
         ip0 += mLength;
         anchor = ip0;
-        ip1 = ip0 + 1;
 
         if (ip0 <= ilimit) {
             /* Fill Table */
@@ -148,19 +160,18 @@ _match: /* Requires: ip0, match0, offcode */
             hashTable[ZSTD_hashPtr(base+current0+2, hlog, mls)] = current0+2;  /* here because current+2 could be > iend-8 */
             hashTable[ZSTD_hashPtr(ip0-2, hlog, mls)] = (U32)(ip0-2-base);
 
-            while ( ((ip0 <= ilimit) & (offset_2>0))  /* offset_2==0 means offset_2 is invalidated */
-                 && (MEM_read32(ip0) == MEM_read32(ip0 - offset_2)) ) {
-                /* store sequence */
-                size_t const rLength = ZSTD_count(ip0+4, ip0+4-offset_2, iend) + 4;
-                { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
-                hashTable[ZSTD_hashPtr(ip0, hlog, mls)] = (U32)(ip0-base);
-                ip0 += rLength;
-                ip1 = ip0 + 1;
-                ZSTD_storeSeq(seqStore, 0 /*litLen*/, anchor, iend, 0 /*offCode*/, rLength-MINMATCH);
-                anchor = ip0;
-                continue;   /* faster when present (confirmed on gcc-8) ... (?) */
-            }
-        }
+            if (offset_2 > 0) { /* offset_2==0 means offset_2 is invalidated */
+                while ( (ip0 <= ilimit) && (MEM_read32(ip0) == MEM_read32(ip0 - offset_2)) ) {
+                    /* store sequence */
+                    size_t const rLength = ZSTD_count(ip0+4, ip0+4-offset_2, iend) + 4;
+                    { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
+                    hashTable[ZSTD_hashPtr(ip0, hlog, mls)] = (U32)(ip0-base);
+                    ip0 += rLength;
+                    ZSTD_storeSeq(seqStore, 0 /*litLen*/, anchor, iend, 0 /*offCode*/, rLength-MINMATCH);
+                    anchor = ip0;
+                    continue;   /* faster when present (confirmed on gcc-8) ... (?) */
+        }   }   }
+        ip1 = ip0 + 1;
     }
 
     /* save reps for next block */
@@ -387,7 +398,7 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
     const BYTE* const ilimit = iend - 8;
     U32 offset_1=rep[0], offset_2=rep[1];
 
-    DEBUGLOG(5, "ZSTD_compressBlock_fast_extDict_generic");
+    DEBUGLOG(5, "ZSTD_compressBlock_fast_extDict_generic (offset_1=%u)", offset_1);
 
     /* switch to "regular" variant if extDict is invalidated due to maxDistance */
     if (prefixStartIndex == dictStartIndex)
@@ -404,6 +415,7 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
         const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
         const BYTE* const repMatch = repBase + repIndex;
         hashTable[h] = current;   /* update hash table */
+        DEBUGLOG(7, "offset_1 = %u , current = %u", offset_1, current);
         assert(offset_1 <= current +1);   /* check repIndex */
 
         if ( (((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > dictStartIndex))
diff --git a/zstd/zstd_fast.h b/zstd/zstd_fast.h
index b74a88c5..3e15cc85 100644
--- a/zstd/zstd_fast.h
+++ b/zstd/zstd_fast.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstd_internal.h b/zstd/zstd_internal.h
index d0b01435..8e4201e1 100644
--- a/zstd/zstd_internal.h
+++ b/zstd/zstd_internal.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -19,6 +19,9 @@
 /*-*************************************
 *  Dependencies
 ***************************************/
+#ifdef __aarch64__
+#include <arm_neon.h>
+#endif
 #include "compiler.h"
 #include "mem.h"
 #include "debug.h"                 /* assert, DEBUGLOG, RAWLOG, g_debuglevel */
@@ -54,6 +57,31 @@ extern "C" {
 #define MAX(a,b) ((a)>(b) ? (a) : (b))
 
 /**
+ * Ignore: this is an internal helper.
+ *
+ * This is a helper function to help force C99-correctness during compilation.
+ * Under strict compilation modes, variadic macro arguments can't be empty.
+ * However, variadic function arguments can be. Using a function therefore lets
+ * us statically check that at least one (string) argument was passed,
+ * independent of the compilation flags.
+ */
+static INLINE_KEYWORD UNUSED_ATTR
+void _force_has_format_string(const char *format, ...) {
+  (void)format;
+}
+
+/**
+ * Ignore: this is an internal helper.
+ *
+ * We want to force this function invocation to be syntactically correct, but
+ * we don't want to force runtime evaluation of its arguments.
+ */
+#define _FORCE_HAS_FORMAT_STRING(...) \
+  if (0) { \
+    _force_has_format_string(__VA_ARGS__); \
+  }
+
+/**
  * Return the specified error if the condition evaluates to true.
  *
  * In debug modes, prints additional information.
@@ -62,7 +90,9 @@ extern "C" {
  */
 #define RETURN_ERROR_IF(cond, err, ...) \
   if (cond) { \
-    RAWLOG(3, "%s:%d: ERROR!: check %s failed, returning %s", __FILE__, __LINE__, ZSTD_QUOTE(cond), ZSTD_QUOTE(ERROR(err))); \
+    RAWLOG(3, "%s:%d: ERROR!: check %s failed, returning %s", \
+           __FILE__, __LINE__, ZSTD_QUOTE(cond), ZSTD_QUOTE(ERROR(err))); \
+    _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
     RAWLOG(3, ": " __VA_ARGS__); \
     RAWLOG(3, "\n"); \
     return ERROR(err); \
@@ -75,7 +105,9 @@ extern "C" {
  */
 #define RETURN_ERROR(err, ...) \
   do { \
-    RAWLOG(3, "%s:%d: ERROR!: unconditional check failed, returning %s", __FILE__, __LINE__, ZSTD_QUOTE(ERROR(err))); \
+    RAWLOG(3, "%s:%d: ERROR!: unconditional check failed, returning %s", \
+           __FILE__, __LINE__, ZSTD_QUOTE(ERROR(err))); \
+    _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
     RAWLOG(3, ": " __VA_ARGS__); \
     RAWLOG(3, "\n"); \
     return ERROR(err); \
@@ -90,7 +122,9 @@ extern "C" {
   do { \
     size_t const err_code = (err); \
     if (ERR_isError(err_code)) { \
-      RAWLOG(3, "%s:%d: ERROR!: forwarding error in %s: %s", __FILE__, __LINE__, ZSTD_QUOTE(err), ERR_getErrorName(err_code)); \
+      RAWLOG(3, "%s:%d: ERROR!: forwarding error in %s: %s", \
+             __FILE__, __LINE__, ZSTD_QUOTE(err), ERR_getErrorName(err_code)); \
+      _FORCE_HAS_FORMAT_STRING(__VA_ARGS__); \
       RAWLOG(3, ": " __VA_ARGS__); \
       RAWLOG(3, "\n"); \
       return err_code; \
@@ -193,10 +227,22 @@ static const U32 OF_defaultNormLog = OF_DEFAULTNORMLOG;
 /*-*******************************************
 *  Shared functions to include for inlining
 *********************************************/
-static void ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
+static void ZSTD_copy8(void* dst, const void* src) {
+#ifdef __aarch64__
+    vst1_u8((uint8_t*)dst, vld1_u8((const uint8_t*)src));
+#else
+    memcpy(dst, src, 8);
+#endif
+}
 
 #define COPY8(d,s) { ZSTD_copy8(d,s); d+=8; s+=8; }
-static void ZSTD_copy16(void* dst, const void* src) { memcpy(dst, src, 16); }
+static void ZSTD_copy16(void* dst, const void* src) {
+#ifdef __aarch64__
+    vst1q_u8((uint8_t*)dst, vld1q_u8((const uint8_t*)src));
+#else
+    memcpy(dst, src, 16);
+#endif
+}
 #define COPY16(d,s) { ZSTD_copy16(d,s); d+=16; s+=16; }
 
 #define WILDCOPY_OVERLENGTH 32
@@ -235,9 +281,15 @@ void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length, ZSTD_overlap_e
         /* Separate out the first COPY16() call because the copy length is
          * almost certain to be short, so the branches have different
          * probabilities. Since it is almost certain to be short, only do
-	 * one COPY16() in the first call. Then, do two calls per loop since
-	 * at that point it is more likely to have a high trip count.
+         * one COPY16() in the first call. Then, do two calls per loop since
+         * at that point it is more likely to have a high trip count.
          */
+#ifndef __aarch64__
+        do {
+            COPY16(op, ip);
+        }
+        while (op < oend);
+#else
         COPY16(op, ip);
         if (op >= oend) return;
         do {
@@ -245,9 +297,29 @@ void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length, ZSTD_overlap_e
             COPY16(op, ip);
         }
         while (op < oend);
+#endif
     }
 }
 
+MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    size_t const length = MIN(dstCapacity, srcSize);
+    if (length > 0) {
+        memcpy(dst, src, length);
+    }
+    return length;
+}
+
+/* define "workspace is too large" as this number of times larger than needed */
+#define ZSTD_WORKSPACETOOLARGE_FACTOR 3
+
+/* when workspace is continuously too large
+ * during at least this number of times,
+ * context's memory usage is considered wasteful,
+ * because it's sized to handle a worst case scenario which rarely happens.
+ * In which case, resize it down to free some memory */
+#define ZSTD_WORKSPACETOOLARGE_MAXDURATION 128
+
 
 /*-*******************************************
 *  Private declarations
@@ -272,6 +344,31 @@ typedef struct {
     U32   longLengthPos;
 } seqStore_t;
 
+typedef struct {
+    U32 litLength;
+    U32 matchLength;
+} ZSTD_sequenceLength;
+
+/**
+ * Returns the ZSTD_sequenceLength for the given sequences. It handles the decoding of long sequences
+ * indicated by longLengthPos and longLengthID, and adds MINMATCH back to matchLength.
+ */
+MEM_STATIC ZSTD_sequenceLength ZSTD_getSequenceLength(seqStore_t const* seqStore, seqDef const* seq)
+{
+    ZSTD_sequenceLength seqLen;
+    seqLen.litLength = seq->litLength;
+    seqLen.matchLength = seq->matchLength + MINMATCH;
+    if (seqStore->longLengthPos == (U32)(seq - seqStore->sequencesStart)) {
+        if (seqStore->longLengthID == 1) {
+            seqLen.litLength += 0xFFFF;
+        }
+        if (seqStore->longLengthID == 2) {
+            seqLen.matchLength += 0xFFFF;
+        }
+    }
+    return seqLen;
+}
+
 /**
  * Contains the compressed frame size and an upper-bound for the decompressed frame size.
  * Note: before using `compressedSize`, check for errors using ZSTD_isError().
@@ -298,8 +395,7 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val)   /* compress, dictBuilder, decodeCorpus
     {
 #   if defined(_MSC_VER)   /* Visual */
         unsigned long r=0;
-        _BitScanReverse(&r, val);
-        return (unsigned)r;
+        return _BitScanReverse(&r, val) ? (unsigned)r : 0;
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* GCC Intrinsic */
         return __builtin_clz (val) ^ 31;
 #   elif defined(__ICCARM__)    /* IAR Intrinsic */
diff --git a/zstd/zstd_lazy.c b/zstd/zstd_lazy.c
index dd82830e..4cf5c88b 100644
--- a/zstd/zstd_lazy.c
+++ b/zstd/zstd_lazy.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -662,10 +662,14 @@ ZSTD_compressBlock_lazy_generic(
                                      0;
     const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictLowest));
 
+    DEBUGLOG(5, "ZSTD_compressBlock_lazy_generic (dictMode=%u)", (U32)dictMode);
+
     /* init */
     ip += (dictAndPrefixLength == 0);
     if (dictMode == ZSTD_noDict) {
-        U32 const maxRep = (U32)(ip - prefixLowest);
+        U32 const current = (U32)(ip - base);
+        U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, ms->cParams.windowLog);
+        U32 const maxRep = current - windowLow;
         if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
         if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
     }
@@ -677,6 +681,12 @@ ZSTD_compressBlock_lazy_generic(
     }
 
     /* Match Loop */
+#if defined(__GNUC__) && defined(__x86_64__)
+    /* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
+     * code alignment is perturbed. To fix the instability align the loop on 32-bytes.
+     */
+    __asm__(".p2align 5");
+#endif
     while (ip < ilimit) {
         size_t matchLength=0;
         size_t offset=0;
@@ -929,11 +939,11 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(
     const BYTE* const ilimit = iend - 8;
     const BYTE* const base = ms->window.base;
     const U32 dictLimit = ms->window.dictLimit;
-    const U32 lowestIndex = ms->window.lowLimit;
     const BYTE* const prefixStart = base + dictLimit;
     const BYTE* const dictBase = ms->window.dictBase;
     const BYTE* const dictEnd  = dictBase + dictLimit;
-    const BYTE* const dictStart  = dictBase + lowestIndex;
+    const BYTE* const dictStart  = dictBase + ms->window.lowLimit;
+    const U32 windowLog = ms->cParams.windowLog;
 
     typedef size_t (*searchMax_f)(
                         ZSTD_matchState_t* ms,
@@ -942,10 +952,18 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(
 
     U32 offset_1 = rep[0], offset_2 = rep[1];
 
+    DEBUGLOG(5, "ZSTD_compressBlock_lazy_extDict_generic");
+
     /* init */
     ip += (ip == prefixStart);
 
     /* Match Loop */
+#if defined(__GNUC__) && defined(__x86_64__)
+    /* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
+     * code alignment is perturbed. To fix the instability align the loop on 32-bytes.
+     */
+    __asm__(".p2align 5");
+#endif
     while (ip < ilimit) {
         size_t matchLength=0;
         size_t offset=0;
@@ -953,10 +971,11 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(
         U32 current = (U32)(ip-base);
 
         /* check repCode */
-        {   const U32 repIndex = (U32)(current+1 - offset_1);
+        {   const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current+1, windowLog);
+            const U32 repIndex = (U32)(current+1 - offset_1);
             const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
             const BYTE* const repMatch = repBase + repIndex;
-            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))   /* intentional overflow */
+            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow))   /* intentional overflow */
             if (MEM_read32(ip+1) == MEM_read32(repMatch)) {
                 /* repcode detected we should take it */
                 const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
@@ -983,10 +1002,11 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(
             current++;
             /* check repCode */
             if (offset) {
+                const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current, windowLog);
                 const U32 repIndex = (U32)(current - offset_1);
                 const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
                 const BYTE* const repMatch = repBase + repIndex;
-                if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+                if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow))  /* intentional overflow */
                 if (MEM_read32(ip) == MEM_read32(repMatch)) {
                     /* repcode detected */
                     const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
@@ -1013,10 +1033,11 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(
                 current++;
                 /* check repCode */
                 if (offset) {
+                    const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current, windowLog);
                     const U32 repIndex = (U32)(current - offset_1);
                     const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
                     const BYTE* const repMatch = repBase + repIndex;
-                    if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+                    if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow))  /* intentional overflow */
                     if (MEM_read32(ip) == MEM_read32(repMatch)) {
                         /* repcode detected */
                         const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
@@ -1057,10 +1078,12 @@ _storeSequence:
 
         /* check immediate repcode */
         while (ip <= ilimit) {
-            const U32 repIndex = (U32)((ip-base) - offset_2);
+            const U32 repCurrent = (U32)(ip-base);
+            const U32 windowLow = ZSTD_getLowestMatchIndex(ms, repCurrent, windowLog);
+            const U32 repIndex = repCurrent - offset_2;
             const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
             const BYTE* const repMatch = repBase + repIndex;
-            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow))  /* intentional overflow */
             if (MEM_read32(ip) == MEM_read32(repMatch)) {
                 /* repcode detected we should take it */
                 const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
diff --git a/zstd/zstd_lazy.h b/zstd/zstd_lazy.h
index bb176306..581936f0 100644
--- a/zstd/zstd_lazy.h
+++ b/zstd/zstd_lazy.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstd_ldm.c b/zstd/zstd_ldm.c
index c3312ad3..3916da49 100644
--- a/zstd/zstd_ldm.c
+++ b/zstd/zstd_ldm.c
@@ -1,10 +1,11 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * 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.
  */
 
 #include "zstd_ldm.h"
@@ -223,6 +224,20 @@ static U64 ZSTD_ldm_fillLdmHashTable(ldmState_t* state,
     return rollingHash;
 }
 
+void ZSTD_ldm_fillHashTable(
+            ldmState_t* state, const BYTE* ip,
+            const BYTE* iend, ldmParams_t const* params)
+{
+    DEBUGLOG(5, "ZSTD_ldm_fillHashTable");
+    if ((size_t)(iend - ip) >= params->minMatchLength) {
+        U64 startingHash = ZSTD_rollingHash_compute(ip, params->minMatchLength);
+        ZSTD_ldm_fillLdmHashTable(
+            state, startingHash, ip, iend - params->minMatchLength, state->window.base,
+            params->hashLog - params->bucketSizeLog,
+            *params);
+    }
+}
+
 
 /** ZSTD_ldm_limitTableUpdate() :
  *
@@ -449,6 +464,8 @@ size_t ZSTD_ldm_generateSequences(
             U32 const correction = ZSTD_window_correctOverflow(
                 &ldmState->window, /* cycleLog */ 0, maxDist, chunkStart);
             ZSTD_ldm_reduceTable(ldmState->hashTable, ldmHSize, correction);
+            /* invalidate dictionaries on overflow correction */
+            ldmState->loadedDictEnd = 0;
         }
         /* 2. We enforce the maximum offset allowed.
          *
@@ -457,8 +474,14 @@ size_t ZSTD_ldm_generateSequences(
          * TODO: * Test the chunk size.
          *       * Try invalidation after the sequence generation and test the
          *         the offset against maxDist directly.
+         *
+         * NOTE: Because of dictionaries + sequence splitting we MUST make sure
+         * that any offset used is valid at the END of the sequence, since it may
+         * be split into two sequences. This condition holds when using
+         * ZSTD_window_enforceMaxDist(), but if we move to checking offsets
+         * against maxDist directly, we'll have to carefully handle that case.
          */
-        ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, NULL, NULL);
+        ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, &ldmState->loadedDictEnd, NULL);
         /* 3. Generate the sequences for the chunk, and get newLeftoverSize. */
         newLeftoverSize = ZSTD_ldm_generateSequences_internal(
             ldmState, sequences, params, chunkStart, chunkSize);
@@ -566,14 +589,13 @@ size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
         if (sequence.offset == 0)
             break;
 
-        assert(sequence.offset <= (1U << cParams->windowLog));
         assert(ip + sequence.litLength + sequence.matchLength <= iend);
 
         /* Fill tables for block compressor */
         ZSTD_ldm_limitTableUpdate(ms, ip);
         ZSTD_ldm_fillFastTables(ms, ip);
         /* Run the block compressor */
-        DEBUGLOG(5, "calling block compressor on segment of size %u", sequence.litLength);
+        DEBUGLOG(5, "pos %u : calling block compressor on segment of size %u", (unsigned)(ip-istart), sequence.litLength);
         {
             size_t const newLitLength =
                 blockCompressor(ms, seqStore, rep, ip, sequence.litLength);
diff --git a/zstd/zstd_ldm.h b/zstd/zstd_ldm.h
index a4784612..aa2b6c52 100644
--- a/zstd/zstd_ldm.h
+++ b/zstd/zstd_ldm.h
@@ -1,10 +1,11 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * 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.
  */
 
 #ifndef ZSTD_LDM_H
@@ -23,6 +24,10 @@ extern "C" {
 
 #define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_LIMIT_DEFAULT
 
+void ZSTD_ldm_fillHashTable(
+            ldmState_t* state, const BYTE* ip,
+            const BYTE* iend, ldmParams_t const* params);
+
 /**
  * ZSTD_ldm_generateSequences():
  *
diff --git a/zstd/zstd_opt.c b/zstd/zstd_opt.c
index 2e50fca6..36fff050 100644
--- a/zstd/zstd_opt.c
+++ b/zstd/zstd_opt.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -249,40 +249,6 @@ static U32 ZSTD_litLengthPrice(U32 const litLength, const optState_t* const optP
     }
 }
 
-/* ZSTD_litLengthContribution() :
- * @return ( cost(litlength) - cost(0) )
- * this value can then be added to rawLiteralsCost()
- * to provide a cost which is directly comparable to a match ending at same position */
-static int ZSTD_litLengthContribution(U32 const litLength, const optState_t* const optPtr, int optLevel)
-{
-    if (optPtr->priceType >= zop_predef) return (int)WEIGHT(litLength, optLevel);
-
-    /* dynamic statistics */
-    {   U32 const llCode = ZSTD_LLcode(litLength);
-        int const contribution = (int)(LL_bits[llCode] * BITCOST_MULTIPLIER)
-                               + (int)WEIGHT(optPtr->litLengthFreq[0], optLevel)   /* note: log2litLengthSum cancel out */
-                               - (int)WEIGHT(optPtr->litLengthFreq[llCode], optLevel);
-#if 1
-        return contribution;
-#else
-        return MAX(0, contribution); /* sometimes better, sometimes not ... */
-#endif
-    }
-}
-
-/* ZSTD_literalsContribution() :
- * creates a fake cost for the literals part of a sequence
- * which can be compared to the ending cost of a match
- * should a new match start at this position */
-static int ZSTD_literalsContribution(const BYTE* const literals, U32 const litLength,
-                                     const optState_t* const optPtr,
-                                     int optLevel)
-{
-    int const contribution = (int)ZSTD_rawLiteralsCost(literals, litLength, optPtr, optLevel)
-                           + ZSTD_litLengthContribution(litLength, optPtr, optLevel);
-    return contribution;
-}
-
 /* ZSTD_getMatchPrice() :
  * Provides the cost of the match part (offset + matchLength) of a sequence
  * Must be combined with ZSTD_fullLiteralsCost() to get the full cost of a sequence.
@@ -603,7 +569,10 @@ U32 ZSTD_insertBtAndGetAllMatches (
             U32 repLen = 0;
             assert(current >= dictLimit);
             if (repOffset-1 /* intentional overflow, discards 0 and -1 */ < current-dictLimit) {  /* equivalent to `current > repIndex >= dictLimit` */
-                if (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repOffset, minMatch)) {
+                /* We must validate the repcode offset because when we're using a dictionary the
+                 * valid offset range shrinks when the dictionary goes out of bounds.
+                 */
+                if ((repIndex >= windowLow) & (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repOffset, minMatch))) {
                     repLen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repOffset, iLimit) + minMatch;
                 }
             } else {  /* repIndex < dictLimit || repIndex >= current */
@@ -799,30 +768,6 @@ FORCE_INLINE_TEMPLATE U32 ZSTD_BtGetAllMatches (
 /*-*******************************
 *  Optimal parser
 *********************************/
-typedef struct repcodes_s {
-    U32 rep[3];
-} repcodes_t;
-
-static repcodes_t ZSTD_updateRep(U32 const rep[3], U32 const offset, U32 const ll0)
-{
-    repcodes_t newReps;
-    if (offset >= ZSTD_REP_NUM) {  /* full offset */
-        newReps.rep[2] = rep[1];
-        newReps.rep[1] = rep[0];
-        newReps.rep[0] = offset - ZSTD_REP_MOVE;
-    } else {   /* repcode */
-        U32 const repCode = offset + ll0;
-        if (repCode > 0) {  /* note : if repCode==0, no change */
-            U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
-            newReps.rep[2] = (repCode >= 2) ? rep[1] : rep[2];
-            newReps.rep[1] = rep[0];
-            newReps.rep[0] = currentOffset;
-        } else {   /* repCode == 0 */
-            memcpy(&newReps, rep, sizeof(newReps));
-        }
-    }
-    return newReps;
-}
 
 
 static U32 ZSTD_totalLen(ZSTD_optimal_t sol)
@@ -839,7 +784,7 @@ listStats(const U32* table, int lastEltID)
     int enb;
     for (enb=0; enb < nbElts; enb++) {
         (void)table;
-        //RAWLOG(2, "%3i:%3i,  ", enb, table[enb]);
+        /* RAWLOG(2, "%3i:%3i,  ", enb, table[enb]); */
         RAWLOG(2, "%4i,", table[enb]);
     }
     RAWLOG(2, " \n");
@@ -894,7 +839,12 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
             { U32 i ; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
             opt[0].mlen = 0;  /* means is_a_literal */
             opt[0].litlen = litlen;
-            opt[0].price = ZSTD_literalsContribution(anchor, litlen, optStatePtr, optLevel);
+            /* We don't need to include the actual price of the literals because
+             * it is static for the duration of the forward pass, and is included
+             * in every price. We include the literal length to avoid negative
+             * prices when we subtract the previous literal length.
+             */
+            opt[0].price = ZSTD_litLengthPrice(litlen, optStatePtr, optLevel);
 
             /* large match -> immediate encoding */
             {   U32 const maxML = matches[nbMatches-1].len;
@@ -923,7 +873,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                 for (matchNb = 0; matchNb < nbMatches; matchNb++) {
                     U32 const offset = matches[matchNb].off;
                     U32 const end = matches[matchNb].len;
-                    repcodes_t const repHistory = ZSTD_updateRep(rep, offset, ll0);
                     for ( ; pos <= end ; pos++ ) {
                         U32 const matchPrice = ZSTD_getMatchPrice(offset, pos, optStatePtr, optLevel);
                         U32 const sequencePrice = literalsPrice + matchPrice;
@@ -933,8 +882,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                         opt[pos].off = offset;
                         opt[pos].litlen = litlen;
                         opt[pos].price = sequencePrice;
-                        ZSTD_STATIC_ASSERT(sizeof(opt[pos].rep) == sizeof(repHistory));
-                        memcpy(opt[pos].rep, &repHistory, sizeof(repHistory));
                 }   }
                 last_pos = pos-1;
             }
@@ -961,7 +908,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                     opt[cur].off = 0;
                     opt[cur].litlen = litlen;
                     opt[cur].price = price;
-                    memcpy(opt[cur].rep, opt[cur-1].rep, sizeof(opt[cur].rep));
                 } else {
                     DEBUGLOG(7, "cPos:%zi==rPos:%u : literal would cost more (%.2f>%.2f) (hist:%u,%u,%u)",
                                 inr-istart, cur, ZSTD_fCost(price), ZSTD_fCost(opt[cur].price),
@@ -969,6 +915,21 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                 }
             }
 
+            /* Set the repcodes of the current position. We must do it here
+             * because we rely on the repcodes of the 2nd to last sequence being
+             * correct to set the next chunks repcodes during the backward
+             * traversal.
+             */
+            ZSTD_STATIC_ASSERT(sizeof(opt[cur].rep) == sizeof(repcodes_t));
+            assert(cur >= opt[cur].mlen);
+            if (opt[cur].mlen != 0) {
+                U32 const prev = cur - opt[cur].mlen;
+                repcodes_t newReps = ZSTD_updateRep(opt[prev].rep, opt[cur].off, opt[cur].litlen==0);
+                memcpy(opt[cur].rep, &newReps, sizeof(repcodes_t));
+            } else {
+                memcpy(opt[cur].rep, opt[cur - 1].rep, sizeof(repcodes_t));
+            }
+
             /* last match must start at a minimum distance of 8 from oend */
             if (inr > ilimit) continue;
 
@@ -1009,7 +970,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                 /* set prices using matches found at position == cur */
                 for (matchNb = 0; matchNb < nbMatches; matchNb++) {
                     U32 const offset = matches[matchNb].off;
-                    repcodes_t const repHistory = ZSTD_updateRep(opt[cur].rep, offset, ll0);
                     U32 const lastML = matches[matchNb].len;
                     U32 const startML = (matchNb>0) ? matches[matchNb-1].len+1 : minMatch;
                     U32 mlen;
@@ -1029,8 +989,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                             opt[pos].off = offset;
                             opt[pos].litlen = litlen;
                             opt[pos].price = price;
-                            ZSTD_STATIC_ASSERT(sizeof(opt[pos].rep) == sizeof(repHistory));
-                            memcpy(opt[pos].rep, &repHistory, sizeof(repHistory));
                         } else {
                             DEBUGLOG(7, "rPos:%u (ml=%2u) => new price is worse (%.2f>=%.2f)",
                                         pos, mlen, ZSTD_fCost(price), ZSTD_fCost(opt[pos].price));
@@ -1046,6 +1004,17 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
 _shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
         assert(opt[0].mlen == 0);
 
+        /* Set the next chunk's repcodes based on the repcodes of the beginning
+         * of the last match, and the last sequence. This avoids us having to
+         * update them while traversing the sequences.
+         */
+        if (lastSequence.mlen != 0) {
+            repcodes_t reps = ZSTD_updateRep(opt[cur].rep, lastSequence.off, lastSequence.litlen==0);
+            memcpy(rep, &reps, sizeof(reps));
+        } else {
+            memcpy(rep, opt[cur].rep, sizeof(repcodes_t));
+        }
+
         {   U32 const storeEnd = cur + 1;
             U32 storeStart = storeEnd;
             U32 seqPos = cur;
@@ -1082,20 +1051,6 @@ _shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
                         continue;   /* will finish */
                     }
 
-                    /* repcodes update : like ZSTD_updateRep(), but update in place */
-                    if (offCode >= ZSTD_REP_NUM) {  /* full offset */
-                        rep[2] = rep[1];
-                        rep[1] = rep[0];
-                        rep[0] = offCode - ZSTD_REP_MOVE;
-                    } else {   /* repcode */
-                        U32 const repCode = offCode + (llen==0);
-                        if (repCode) {  /* note : if repCode==0, no change */
-                            U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
-                            if (repCode >= 2) rep[2] = rep[1];
-                            rep[1] = rep[0];
-                            rep[0] = currentOffset;
-                    }   }
-
                     assert(anchor + llen <= iend);
                     ZSTD_updateStats(optStatePtr, llen, anchor, offCode, mlen);
                     ZSTD_storeSeq(seqStore, llen, anchor, iend, offCode, mlen-MINMATCH);
@@ -1104,7 +1059,6 @@ _shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
             }   }
             ZSTD_setBasePrices(optStatePtr, optLevel);
         }
-
     }   /* while (ip < ilimit) */
 
     /* Return the last literals size */
diff --git a/zstd/zstd_opt.h b/zstd/zstd_opt.h
index 094f7476..9aba8a90 100644
--- a/zstd/zstd_opt.h
+++ b/zstd/zstd_opt.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/zstd/zstdmt_compress.c b/zstd/zstdmt_compress.c
index 4eaa3fcd..101f6bf0 100644
--- a/zstd/zstdmt_compress.c
+++ b/zstd/zstdmt_compress.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -461,7 +461,13 @@ typedef struct {
     ZSTD_window_t ldmWindow;  /* A thread-safe copy of ldmState.window */
 } serialState_t;
 
-static int ZSTDMT_serialState_reset(serialState_t* serialState, ZSTDMT_seqPool* seqPool, ZSTD_CCtx_params params, size_t jobSize)
+static int
+ZSTDMT_serialState_reset(serialState_t* serialState,
+                         ZSTDMT_seqPool* seqPool,
+                         ZSTD_CCtx_params params,
+                         size_t jobSize,
+                         const void* dict, size_t const dictSize,
+                         ZSTD_dictContentType_e dictContentType)
 {
     /* Adjust parameters */
     if (params.ldmParams.enableLdm) {
@@ -491,7 +497,6 @@ static int ZSTDMT_serialState_reset(serialState_t* serialState, ZSTDMT_seqPool*
         ZSTDMT_setNbSeq(seqPool, ZSTD_ldm_getMaxNbSeq(params.ldmParams, jobSize));
         /* Reset the window */
         ZSTD_window_init(&serialState->ldmState.window);
-        serialState->ldmWindow = serialState->ldmState.window;
         /* Resize tables and output space if necessary. */
         if (serialState->ldmState.hashTable == NULL || serialState->params.ldmParams.hashLog < hashLog) {
             ZSTD_free(serialState->ldmState.hashTable, cMem);
@@ -506,7 +511,24 @@ static int ZSTDMT_serialState_reset(serialState_t* serialState, ZSTDMT_seqPool*
         /* Zero the tables */
         memset(serialState->ldmState.hashTable, 0, hashSize);
         memset(serialState->ldmState.bucketOffsets, 0, bucketSize);
+
+        /* Update window state and fill hash table with dict */
+        serialState->ldmState.loadedDictEnd = 0;
+        if (dictSize > 0) {
+            if (dictContentType == ZSTD_dct_rawContent) {
+                BYTE const* const dictEnd = (const BYTE*)dict + dictSize;
+                ZSTD_window_update(&serialState->ldmState.window, dict, dictSize);
+                ZSTD_ldm_fillHashTable(&serialState->ldmState, (const BYTE*)dict, dictEnd, &params.ldmParams);
+                serialState->ldmState.loadedDictEnd = params.forceWindow ? 0 : (U32)(dictEnd - serialState->ldmState.window.base);
+            } else {
+                /* don't even load anything */
+            }
+        }
+
+        /* Initialize serialState's copy of ldmWindow. */
+        serialState->ldmWindow = serialState->ldmState.window;
     }
+
     serialState->params = params;
     serialState->params.jobSize = (U32)jobSize;
     return 0;
@@ -1054,7 +1076,7 @@ static ZSTD_CCtx_params ZSTDMT_initJobCCtxParams(const ZSTD_CCtx_params* params)
 static size_t ZSTDMT_resize(ZSTDMT_CCtx* mtctx, unsigned nbWorkers)
 {
     if (POOL_resize(mtctx->factory, nbWorkers)) return ERROR(memory_allocation);
-    FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbWorkers) );
+    FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbWorkers) , "");
     mtctx->bufPool = ZSTDMT_expandBufferPool(mtctx->bufPool, nbWorkers);
     if (mtctx->bufPool == NULL) return ERROR(memory_allocation);
     mtctx->cctxPool = ZSTDMT_expandCCtxPool(mtctx->cctxPool, nbWorkers);
@@ -1235,7 +1257,8 @@ ZSTDMT_computeNbJobs(const ZSTD_CCtx_params* params, size_t srcSize, unsigned nb
 /* ZSTDMT_compress_advanced_internal() :
  * This is a blocking function : it will only give back control to caller after finishing its compression job.
  */
-static size_t ZSTDMT_compress_advanced_internal(
+static size_t
+ZSTDMT_compress_advanced_internal(
                 ZSTDMT_CCtx* mtctx,
                 void* dst, size_t dstCapacity,
           const void* src, size_t srcSize,
@@ -1267,10 +1290,11 @@ static size_t ZSTDMT_compress_advanced_internal(
 
     assert(avgJobSize >= 256 KB);  /* condition for ZSTD_compressBound(A) + ZSTD_compressBound(B) <= ZSTD_compressBound(A+B), required to compress directly into Dst (no additional buffer) */
     ZSTDMT_setBufferSize(mtctx->bufPool, ZSTD_compressBound(avgJobSize) );
-    if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, avgJobSize))
+    /* LDM doesn't even try to load the dictionary in single-ingestion mode */
+    if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, avgJobSize, NULL, 0, ZSTD_dct_auto))
         return ERROR(memory_allocation);
 
-    FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbJobs) );  /* only expands if necessary */
+    FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbJobs) , "");  /* only expands if necessary */
 
     {   unsigned u;
         for (u=0; u<nbJobs; u++) {
@@ -1403,7 +1427,7 @@ size_t ZSTDMT_initCStream_internal(
 
     /* init */
     if (params.nbWorkers != mtctx->params.nbWorkers)
-        FORWARD_IF_ERROR( ZSTDMT_resize(mtctx, params.nbWorkers) );
+        FORWARD_IF_ERROR( ZSTDMT_resize(mtctx, params.nbWorkers) , "");
 
     if (params.jobSize != 0 && params.jobSize < ZSTDMT_JOBSIZE_MIN) params.jobSize = ZSTDMT_JOBSIZE_MIN;
     if (params.jobSize > (size_t)ZSTDMT_JOBSIZE_MAX) params.jobSize = (size_t)ZSTDMT_JOBSIZE_MAX;
@@ -1500,7 +1524,8 @@ size_t ZSTDMT_initCStream_internal(
     mtctx->allJobsCompleted = 0;
     mtctx->consumed = 0;
     mtctx->produced = 0;
-    if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, mtctx->targetSectionSize))
+    if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, mtctx->targetSectionSize,
+                                 dict, dictSize, dictContentType))
         return ERROR(memory_allocation);
     return 0;
 }
@@ -2062,7 +2087,7 @@ size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
       || ((endOp == ZSTD_e_end) && (!mtctx->frameEnded)) ) {   /* must finish the frame with a zero-size block */
         size_t const jobSize = mtctx->inBuff.filled;
         assert(mtctx->inBuff.filled <= mtctx->targetSectionSize);
-        FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, jobSize, endOp) );
+        FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, jobSize, endOp) , "");
     }
 
     /* check for potential compressed data ready to be flushed */
@@ -2076,7 +2101,7 @@ size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
 
 size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
 {
-    FORWARD_IF_ERROR( ZSTDMT_compressStream_generic(mtctx, output, input, ZSTD_e_continue) );
+    FORWARD_IF_ERROR( ZSTDMT_compressStream_generic(mtctx, output, input, ZSTD_e_continue) , "");
 
     /* recommended next input size : fill current input buffer */
     return mtctx->targetSectionSize - mtctx->inBuff.filled;   /* note : could be zero when input buffer is fully filled and no more availability to create new job */
@@ -2093,7 +2118,7 @@ static size_t ZSTDMT_flushStream_internal(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* ou
       || ((endFrame==ZSTD_e_end) && !mtctx->frameEnded)) {  /* need a last 0-size block to end frame */
            DEBUGLOG(5, "ZSTDMT_flushStream_internal : create a new job (%u bytes, end:%u)",
                         (U32)srcSize, (U32)endFrame);
-        FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, srcSize, endFrame) );
+        FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, srcSize, endFrame) , "");
     }
 
     /* check if there is any data available to flush */
diff --git a/zstd/zstdmt_compress.h b/zstd/zstdmt_compress.h
index 12a52608..b90fd389 100644
--- a/zstd/zstdmt_compress.h
+++ b/zstd/zstdmt_compress.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the