Merge branch 'master' of https://github.com/yandex/ClickHouse into CLICKHOUSE-2720

.gitignore

@@ -199,7 +199,6 @@ vgcore*
 *.changes
 build-stamp
 configure-stamp
-debian/changelog
 debian/*.debhelper.log
 debian/*.debhelper
 debian/*.substvars

CHANGELOG.md

@@ -0,0 +1,119 @@
+# ClickHouse release 1.1.54282
+
+This is a bugfix release. The following bugs were fixed:
+* `DB::Exception: Assertion violation: !_path.empty()` error when inserting into a Distributed table.
+* Error when parsing inserted data in RowBinary format if the data begins with ';' character.
+* Errors during runtime compilation of certain aggregate functions (e.g. `groupArray()`).
+
+# ClickHouse release 1.1.54276
+
+## New features:
+
+* You can use an optional WITH clause in a SELECT query. Example query: `WITH 1+1 AS a SELECT a, a*a`
+* INSERT can be performed synchronously in a Distributed table: OK is returned only after all the data is saved on all the shards. This is activated by the setting insert_distributed_sync=1.
+* Added the UUID data type for working with 16-byte identifiers.
+* Added aliases of CHAR, FLOAT and other types for compatibility with the Tableau.
+* Added the functions toYYYYMM, toYYYYMMDD, and toYYYYMMDDhhmmss for converting time into numbers.
+* You can use IP addresses (together with the hostname) to identify servers for clustered DDL queries.
+* Added support for non-constant arguments and negative offsets in the function `substring(str, pos, len).`
+* Added the max_size parameter for the `groupArray(max_size)(column)` aggregate function, and optimized its performance.
+
+## Major changes:
+
+* Improved security: all server files are created with 0640 permissions.
+* Improved error messages for queries with invalid syntax.
+* Significantly reduced memory consumption and improved performance when merging large sections of MergeTree data.
+* Significantly increased the performance of data merges for the ReplacingMergeTree engine.
+* Improved performance for asynchronous inserts from a Distributed table by batching multiple source inserts. To enable this functionality, use the setting distributed_directory_monitor_batch_inserts=1.
+
+## Backward incompatible changes:
+
+* Changed the binary format of aggregate states of `groupArray(array_column)` functions for arrays.
+
+## Complete list of changes:
+
+* Added the `output_format_json_quote_denormals` setting, which enables outputting nan and inf values in JSON format.
+* Optimized thread allocation when reading from a Distributed table.
+* Settings can be modified in readonly mode if the value doesn't change.
+* Added the ability to read fractional granules of the MergeTree engine in order to meet restrictions on the block size specified in the preferred_block_size_bytes setting. The purpose is to reduce the consumption of RAM and increase cache locality when processing queries from tables with large columns.
+* Efficient use of indexes that contain expressions like `toStartOfHour(x)` for conditions like `toStartOfHour(x) op сonstexpr.`
+* Added new settings for MergeTree engines (the merge_tree section in config.xml):
+   - replicated_deduplication_window_seconds sets the size of deduplication window in seconds for Replicated tables.
+   - cleanup_delay_period sets how often to start cleanup to remove outdated data.
+   - replicated_can_become_leader can prevent a replica from becoming the leader (and assigning merges).
+* Accelerated cleanup to remove outdated data from ZooKeeper.
+* Multiple improvements and fixes for clustered DDL queries. Of particular interest is the new setting distributed_ddl_task_timeout, which limits the time to wait for a response from the servers in the cluster.
+* Improved display of stack traces in the server logs.
+* Added the "none" value for the compression method.
+* You can use multiple dictionaries_config sections in config.xml.
+* It is possible to connect to MySQL through a socket in the file system.
+* The `system.parts` table has a new column with information about the size of marks, in bytes.
+
+## Bug fixes:
+
+* Distributed tables using a Merge table now work correctly for a SELECT query with a condition on the  _table field.
+* Fixed a rare race condition in ReplicatedMergeTree when checking data parts.
+* Fixed possible freezing on "leader election" when starting a server.
+* The max_replica_delay_for_distributed_queries setting was ignored when using a local replica of the data source. This has been fixed.
+* Fixed incorrect behavior of `ALTER TABLE CLEAR COLUMN IN PARTITION` when attempting to clean a non-existing column.
+* Fixed an exception in the multiIf function when using empty arrays or strings.
+* Fixed excessive memory allocations when deserializing Native format.
+* Fixed incorrect auto-update of Trie dictionaries.
+* Fixed an exception when running queries with a GROUP BY clause from a Merge table when using SAMPLE.
+* Fixed a crash of GROUP BY when using distributed_aggregation_memory_efficient=1.
+* Now you can specify the database.table in the right side of IN and JOIN.
+* Too many threads were used for parallel aggregation. This has been fixed.
+* Fixed how the "if" function works with FixedString arguments.
+* SELECT worked incorrectly from a Distributed table for shards with a weight of 0. This has been fixed.
+* Crashes no longer occur when running `CREATE VIEW IF EXISTS.`
+* Fixed incorrect behavior when input_format_skip_unknown_fields=1 is set and there are negative numbers.
+* Fixed an infinite loop in the `dictGetHierarchy()` function if there is some invalid data in the dictionary.
+* Fixed `Syntax error: unexpected (...)` errors  when running distributed queries with subqueries in an IN or JOIN clause and Merge tables.
+* Fixed the incorrect interpretation of a SELECT query from Dictionary tables.
+* Fixed the "Cannot mremap" error when using arrays in IN and JOIN clauses with more than 2 billion elements.
+* Fixed the failover for dictionaries with MySQL as the source.
+
+## Improved workflow for developing and assembling ClickHouse:
+
+* Builds can be assembled in Arcadia.
+* You can use gcc 7 to compile ClickHouse.
+* Parallel builds using ccache+distcc are faster now.
+
+# ClickHouse release 1.1.54245
+
+## New features:
+
+* Distributed DDL (for example, `CREATE TABLE ON CLUSTER`).
+* The replicated request `ALTER TABLE CLEAR COLUMN IN PARTITION.`
+* The engine for Dictionary tables (access to dictionary data in the form of a table).
+* Dictionary database engine (this type of database automatically has Dictionary tables available for all the connected external dictionaries).
+* You can check for updates to the dictionary by sending a request to the source.
+* Qualified column names
+* Quoting identifiers using double quotation marks.
+* Sessions in the HTTP interface.
+* The OPTIMIZE query for a Replicated table can can run not only on the leader.
+
+## Backward incompatible changes:
+
+* Removed SET GLOBAL.
+
+## Minor changes:
+
+* If an alert is triggered, the full stack trace is printed into the log.
+* Relaxed the verification of the number of damaged or extra data parts at startup (there were too many false positives).
+
+## Bug fixes:
+
+* Fixed a bad connection "sticking" when inserting into a Distributed table.
+* GLOBAL IN now works for a query from a Merge table that looks at a Distributed table.
+* The incorrect number of cores was detected on a Google Compute Engine virtual machine. This has been fixed.
+* Changes in how an executable source of cached external dictionaries works.
+* Fixed the comparison of strings containing null characters.
+* Fixed the comparison of Float32 primary key fields with constants.
+* Previously, an incorrect estimate of the size of a field could lead to overly large allocations. This has been fixed.
+* Fixed a crash when querying a Nullable column added to a table using ALTER.
+* Fixed a crash when sorting by a Nullable column, if the number of rows is less than LIMIT.
+* Fixed an ORDER BY subquery consisting of only constant values.
+* Previously, a Replicated table could remain in the invalid state after a failed DROP TABLE.
+* Aliases for scalar subqueries with empty results are no longer lost.
+* Now a query that used compilation does not fail with an error if the .so file gets damaged.

CHANGELOG_RU.md

@@ -0,0 +1,110 @@
+# Релиз ClickHouse 1.1.54282
+
+Релиз содержит исправления к предыдущему релизу 1.1.54276:
+* Исправлена ошибка `DB::Exception: Assertion violation: !_path.empty()` при вставке в Distributed таблицу.
+* Исправлен парсинг при вставке в формате RowBinary, если входные данные начинаются с ';'.
+* Исправлена ошибка при рантайм-компиляции некоторых агрегатных функций (например, `groupArray()`).
+
+# Релиз ClickHouse 1.1.54276
+
+## Новые возможности:
+* Добавлена опциональная секция WITH запроса SELECT. Пример запроса: `WITH 1+1 AS a SELECT a, a*a`
+* Добавлена возможность синхронной вставки в Distributed таблицу: выдается Ok только после того как все данные записались на все шарды. Активируется настройкой insert_distributed_sync=1
+* Добавлен тип данных UUID для работы с 16-байтовыми идентификаторами
+* Добавлены алиасы типов CHAR, FLOAT и т.д. для совместимости с Tableau
+* Добавлены функции toYYYYMM, toYYYYMMDD, toYYYYMMDDhhmmss для перевода времени в числа
+* Добавлена возможность использовать IP адреса (совместно с hostname) для идентификации сервера при работе с кластерными DDL запросами
+* Добавлена поддержка неконстантных аргументов и отрицательных смещений в функции `substring(str, pos, len)`
+* Добавлен параметр max_size для агрегатной функции `groupArray(max_size)(column)`, и оптимизирована её производительность
+
+## Основные изменения:
+* Улучшение безопасности: все файлы сервера создаются с правами 0640
+* Улучшены сообщения об ошибках в случае синтаксически неверных запросов
+* Значительно уменьшен расход оперативной памяти и улучшена производительность слияний больших MergeTree-кусков данных
+* Значительно увеличена производительность слияний данных для движка ReplacingMergeTree
+* Улучшена производительность асинхронных вставок из Distributed таблицы за счет объединения нескольких исходных вставок. Функциональность включается настройкой distributed_directory_monitor_batch_inserts=1.
+
+## Обратно несовместимые изменения:
+* Изменился бинарный формат агрегатных состояний функции `groupArray(array_column)` для массивов
+
+## Полный список изменений:
+* Добавлена настройка `output_format_json_quote_denormals`, включающая вывод nan и inf значений в формате JSON
+* Более оптимальное выделение потоков при чтении из Distributed таблиц
+* Разрешено задавать настройки в режиме readonly, если их значение не изменяется
+* Добавлена возможность считывать нецелые гранулы движка MergeTree для выполнения ограничений на размер блока, задаваемый настройкой preferred_block_size_bytes - для уменьшения потребления оперативной памяти и увеличения кэш-локальности при обработке запросов из таблиц со столбцами большого размера
+* Эффективное использование индекса, содержащего выражения типа `toStartOfHour(x)`, для условий вида `toStartOfHour(x) op сonstexpr`
+* Добавлены новые настройки для MergeTree движков (секция merge_tree в config.xml):
+  - replicated_deduplication_window_seconds позволяет задать интервал дедупликации вставок в Replicated-таблицы в секундах
+  - cleanup_delay_period - периодичность запуска очистки неактуальных данных
+  - replicated_can_become_leader - запретить реплике становиться лидером (и назначать мержи)
+* Ускорена очистка неактуальных данных из ZooKeeper
+* Множественные улучшения и исправления работы кластерных DDL запросов. В частности, добавлена настройка distributed_ddl_task_timeout, ограничивающая время ожидания ответов серверов кластера.
+* Улучшено отображение стэктрейсов в логах сервера
+* Добавлен метод сжатия none
+* Возможность использования нескольких секций dictionaries_config в config.xml
+* Возможность подключения к MySQL через сокет на файловой системе
+* В таблицу system.parts добавлен столбец с информацией о размере marks в байтах
+
+## Исправления багов:
+* Исправлена некорректная работа Distributed таблиц, использующих Merge таблицы, при SELECT с условием на поле _table
+* Исправлен редкий race condition в ReplicatedMergeTree при проверке кусков данных
+* Исправлено возможное зависание процедуры leader election при старте сервера
+* Исправлено игнорирование настройки max_replica_delay_for_distributed_queries при использовании локальной реплики в качестве источника данных
+* Исправлено некорректное поведение `ALTER TABLE CLEAR COLUMN IN PARTITION` при попытке очистить несуществующую колонку
+* Исправлено исключение в функции multiIf при использовании пустых массивов или строк
+* Исправлено чрезмерное выделение памяти при десериализации формата Native
+* Исправлено некорректное автообновление Trie словарей
+* Исправлено исключение при выполнении запросов с GROUP BY из Merge-таблицы при использовании SAMPLE
+* Исправлено падение GROUP BY при использовании настройки distributed_aggregation_memory_efficient=1
+* Добавлена возможность указывать database.table в правой части IN и JOIN
+* Исправлено использование слишком большого количества потоков при параллельной агрегации
+* Исправлена работа функции if с аргументами FixedString
+* Исправлена некорректная работа SELECT из Distributed-таблицы для шардов с весом 0
+* Исправлено падение запроса `CREATE VIEW IF EXISTS`
+* Исправлено некорректное поведение при input_format_skip_unknown_fields=1 в случае отрицательных чисел
+* Исправлен бесконечный цикл в функции `dictGetHierarchy()` в случае некоторых некорректных данных словаря
+* Исправлены ошибки типа `Syntax error: unexpected (...)` при выполнении распределенных запросов с подзапросами в секции IN или JOIN, в случае использования совместно с Merge таблицами
+* Исправлена неправильная интерпретация SELECT запроса из таблиц типа Dictionary
+* Исправлена ошибка "Cannot mremap" при использовании множеств в секциях IN, JOIN, содержащих более 2 млрд. элементов
+* Исправлен failover для словарей с источником MySQL
+
+## Улучшения процесса разработки и сборки ClickHouse:
+* Добавлена возмозможность сборки в Arcadia
+* Добавлена возможность сборки с помощью gcc 7
+* Ускорена параллельная сборка с помощью ccache+distcc
+
+
+# Релиз ClickHouse 1.1.54245
+
+## Новые возможности:
+* Распределённые DDL (например, `CREATE TABLE ON CLUSTER`)
+* Реплицируемый запрос `ALTER TABLE CLEAR COLUMN IN PARTITION`
+* Движок таблиц Dictionary (доступ к данным словаря в виде таблицы)
+* Движок баз данных Dictionary (в такой базе автоматически доступны Dictionary-таблицы для всех подключённых внешних словарей)
+* Возможность проверки необходимости обновления словаря путём отправки запроса в источник
+* Qualified имена столбцов
+* Квотирование идентификаторов двойными кавычками
+* Сессии в HTTP интерфейсе
+* Запрос OPTIMIZE для Replicated таблицы теперь можно выполнять не только на лидере
+
+## Обратно несовместимые изменения:
+* Убрана команда SET GLOBAL
+
+## Мелкие изменения:
+* Теперь после получения сигнала в лог печатается полный стектрейс
+* Ослаблена проверка на количество повреждённых/лишних кусков при старте (было слишком много ложных срабатываний)
+
+## Исправления багов:
+* Исправлено залипание плохого соединения при вставке в Distributed таблицу
+* GLOBAL IN теперь работает при запросе из таблицы Merge, смотрящей в Distributed
+* Теперь правильно определяется количество ядер на виртуалках Google Compute Engine
+* Исправления в работе executable источника кэшируемых внешних словарей
+* Исправлены сравнения строк, содержащих нулевые символы
+* Исправлено сравнение полей первичного ключа типа Float32 с константами
+* Раньше неправильная оценка размера поля могла приводить к слишком большим аллокациям
+* Исправлено падение при запросе Nullable столбца, добавленного в таблицу ALTER-ом
+* Исправлено падение при сортировке по Nullable столбцу, если количество строк меньше LIMIT
+* Исправлен ORDER BY подзапроса, состоящего только из константных значений
+* Раньше Replicated таблица могла остаться в невалидном состоянии после неудавшегося DROP TABLE
+* Алиасы для скалярных подзапросов с пустым результатом теперь не теряются
+* Теперь запрос, в котором использовалась компиляция, не завершается ошибкой, если .so файл повреждается

CMakeLists.txt

@@ -27,11 +27,11 @@ endif ()
 cmake_policy (SET CMP0014 OLD) # Ignore warning about CMakeLists.txt in each directory
 cmake_policy (SET CMP0012 NEW) # Don't dereference TRUE and FALSE
 
-find_program(CCACHE_FOUND ccache)
-if(CCACHE_FOUND AND NOT CMAKE_CXX_COMPILER_LAUNCHER MATCHES "ccache")
-  set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE "ccache")
-  set_property(GLOBAL PROPERTY RULE_LAUNCH_LINK "ccache")
-endif()
+find_program (CCACHE_FOUND ccache)
+if (CCACHE_FOUND AND NOT CMAKE_CXX_COMPILER_LAUNCHER MATCHES "ccache" AND NOT CMAKE_CXX_COMPILER MATCHES "ccache")
+  set_property (GLOBAL PROPERTY RULE_LAUNCH_COMPILE "ccache")
+  set_property (GLOBAL PROPERTY RULE_LAUNCH_LINK "ccache")
+endif ()
 
 if (NOT CMAKE_BUILD_TYPE OR CMAKE_BUILD_TYPE STREQUAL "None")
     message (STATUS "CMAKE_BUILD_TYPE is not set, set to default = RELWITHDEBINFO")
@@ -45,10 +45,10 @@ message (STATUS "CMAKE_BUILD_TYPE: " ${CMAKE_BUILD_TYPE} )
 set (CMAKE_CONFIGURATION_TYPES "RelWithDebInfo;Debug;Release;MinSizeRel;ASan;UBSan" CACHE STRING "" FORCE)
 
 if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64.*|AARCH64.*)")
-    set (AARCH64 1)
+    set (ARCH_AARCH64 1)
 endif ()
-if (AARCH64 OR CMAKE_SYSTEM_PROCESSOR MATCHES "arm")
-    set (ARM 1)
+if (ARCH_AARCH64 OR CMAKE_SYSTEM_PROCESSOR MATCHES "arm")
+    set (ARCH_ARM 1)
 endif ()
 
 set (COMMON_WARNING_FLAGS "-Wall")    # -Werror is also added inside directories with our own code.
@@ -175,6 +175,7 @@ if (NOT OPENSSL_FOUND)
     message (FATAL_ERROR "Need openssl for build. debian tip: sudo apt install libssl-dev")
 endif ()
 
+include (cmake/lib_name.cmake)
 include (cmake/find_icu4c.cmake)
 include (cmake/find_boost.cmake)
 # openssl, zlib before poco
@@ -183,19 +184,24 @@ include (cmake/find_zstd.cmake)
 include (cmake/find_poco.cmake)
 include (cmake/find_lz4.cmake)
 include (cmake/find_sparsehash.cmake)
-include (cmake/find_libtool.cmake)
 include (cmake/find_rt.cmake)
 include (cmake/find_readline_edit.cmake)
 include (cmake/find_zookeeper.cmake)
-include (cmake/find_double-conversion.cmake)
 include (cmake/find_re2.cmake)
+
+include (cmake/find_contrib_lib.cmake)
+find_contrib_lib(cityhash)
+find_contrib_lib(farmhash)
+find_contrib_lib(metrohash)
+find_contrib_lib(btrie)
+find_contrib_lib(double-conversion)
+
 # Need to process before "contrib" dir:
 include (libs/libcommon/cmake/find_gperftools.cmake)
 include (libs/libcommon/cmake/find_jemalloc.cmake)
 include (libs/libcommon/cmake/find_cctz.cmake)
 include (libs/libmysqlxx/cmake/find_mysqlclient.cmake)
 include (libs/libdaemon/cmake/find_unwind.cmake)
-include (cmake/lib_name.cmake)
 
 
 set (FULL_C_FLAGS "${CMAKE_C_FLAGS} ${CMAKE_C_FLAGS_${CMAKE_BUILD_TYPE}}")

MacOS.md

@@ -1,6 +1,6 @@
-## How to increase maxfiles on Mac OS X
+## How to increase maxfiles on macOS
 
-To increase maxfiles on MacOS, create the following file:
+To increase maxfiles on macOS, create the following file:
 
 (Note: you'll need to use sudo)
 

cmake/Modules/Findbtrie.cmake

@@ -0,0 +1,44 @@
+# - Try to find btrie headers and libraries.
+#
+# Usage of this module as follows:
+#
+#     find_package(btrie)
+#
+# Variables used by this module, they can change the default behaviour and need
+# to be set before calling find_package:
+#
+#  BTRIE_ROOT_DIR Set this variable to the root installation of
+#                    btrie if the module has problems finding
+#                    the proper installation path.
+#
+# Variables defined by this module:
+#
+#  BTRIE_FOUND             System has btrie libs/headers
+#  BTRIE_LIBRARIES         The btrie library/libraries
+#  BTRIE_INCLUDE_DIR       The location of btrie headers
+
+find_path(BTRIE_ROOT_DIR
+    NAMES include/btrie.h
+)
+
+find_library(BTRIE_LIBRARIES
+    NAMES btrie
+    PATHS ${BTRIE_ROOT_DIR}/lib ${BTRIE_LIBRARIES_PATHS}
+)
+
+find_path(BTRIE_INCLUDE_DIR
+    NAMES btrie.h
+    PATHS ${BTRIE_ROOT_DIR}/include ${BTRIE_INCLUDE_PATHS}
+)
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(btrie DEFAULT_MSG
+    BTRIE_LIBRARIES
+    BTRIE_INCLUDE_DIR
+)
+
+mark_as_advanced(
+    BTRIE_ROOT_DIR
+    BTRIE_LIBRARIES
+    BTRIE_INCLUDE_DIR
+)

cmake/Modules/Findcityhash.cmake

@@ -0,0 +1,44 @@
+# - Try to find cityhash headers and libraries.
+#
+# Usage of this module as follows:
+#
+#     find_package(cityhash)
+#
+# Variables used by this module, they can change the default behaviour and need
+# to be set before calling find_package:
+#
+#  CITYHASH_ROOT_DIR Set this variable to the root installation of
+#                    cityhash if the module has problems finding
+#                    the proper installation path.
+#
+# Variables defined by this module:
+#
+#  CITYHASH_FOUND             System has cityhash libs/headers
+#  CITYHASH_LIBRARIES         The cityhash library/libraries
+#  CITYHASH_INCLUDE_DIR       The location of cityhash headers
+
+find_path(CITYHASH_ROOT_DIR
+    NAMES include/city.h
+)
+
+find_library(CITYHASH_LIBRARIES
+    NAMES cityhash
+    PATHS ${CITYHASH_ROOT_DIR}/lib ${CITYHASH_LIBRARIES_PATHS}
+)
+
+find_path(CITYHASH_INCLUDE_DIR
+    NAMES city.h
+    PATHS ${CITYHASH_ROOT_DIR}/include ${CITYHASH_INCLUDE_PATHS}
+)
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(cityhash DEFAULT_MSG
+    CITYHASH_LIBRARIES
+    CITYHASH_INCLUDE_DIR
+)
+
+mark_as_advanced(
+    CITYHASH_ROOT_DIR
+    CITYHASH_LIBRARIES
+    CITYHASH_INCLUDE_DIR
+)

cmake/Modules/Finddouble-conversion.cmake

@@ -0,0 +1,44 @@
+# - Try to find double-conversion headers and libraries.
+#
+# Usage of this module as follows:
+#
+#     find_package(double-conversion)
+#
+# Variables used by this module, they can change the default behaviour and need
+# to be set before calling find_package:
+#
+#  DOUBLE_CONVERSION_ROOT_DIR Set this variable to the root installation of
+#                    double-conversion if the module has problems finding
+#                    the proper installation path.
+#
+# Variables defined by this module:
+#
+#  DOUBLE_CONVERSION_FOUND             System has double-conversion libs/headers
+#  DOUBLE_CONVERSION_LIBRARIES         The double-conversion library/libraries
+#  DOUBLE_CONVERSION_INCLUDE_DIR       The location of double-conversion headers
+
+find_path(DOUBLE_CONVERSION_ROOT_DIR
+    NAMES include/double-conversion/double-conversion.h
+)
+
+find_library(DOUBLE_CONVERSION_LIBRARIES
+    NAMES double-conversion
+    PATHS ${DOUBLE_CONVERSION_ROOT_DIR}/lib ${BTRIE_CITYHASH_PATHS}
+)
+
+find_path(DOUBLE_CONVERSION_INCLUDE_DIR
+    NAMES double-conversion/double-conversion.h
+    PATHS ${DOUBLE_CONVERSION_ROOT_DIR}/include ${DOUBLE_CONVERSION_INCLUDE_PATHS}
+)
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(double_conversion DEFAULT_MSG
+    DOUBLE_CONVERSION_LIBRARIES
+    DOUBLE_CONVERSION_INCLUDE_DIR
+)
+
+mark_as_advanced(
+    DOUBLE_CONVERSION_ROOT_DIR
+    DOUBLE_CONVERSION_LIBRARIES
+    DOUBLE_CONVERSION_INCLUDE_DIR
+)

cmake/Modules/Findfarmhash.cmake

@@ -0,0 +1,44 @@
+# - Try to find farmhash headers and libraries.
+#
+# Usage of this module as follows:
+#
+#     find_package(farmhash)
+#
+# Variables used by this module, they can change the default behaviour and need
+# to be set before calling find_package:
+#
+#  FARMHASH_ROOT_DIR Set this variable to the root installation of
+#                    farmhash if the module has problems finding
+#                    the proper installation path.
+#
+# Variables defined by this module:
+#
+#  FARMHASH_FOUND             System has farmhash libs/headers
+#  FARMHASH_LIBRARIES         The farmhash library/libraries
+#  FARMHASH_INCLUDE_DIR       The location of farmhash headers
+
+find_path(FARMHASH_ROOT_DIR
+    NAMES include/farmhash.h
+)
+
+find_library(FARMHASH_LIBRARIES
+    NAMES farmhash
+    PATHS ${FARMHASH_ROOT_DIR}/lib ${FARMHASH_LIBRARIES_PATHS}
+)
+
+find_path(FARMHASH_INCLUDE_DIR
+    NAMES farmhash.h
+    PATHS ${FARMHASH_ROOT_DIR}/include ${FARMHASH_INCLUDE_PATHS}
+)
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(farmhash DEFAULT_MSG
+    FARMHASH_LIBRARIES
+    FARMHASH_INCLUDE_DIR
+)
+
+mark_as_advanced(
+    FARMHASH_ROOT_DIR
+    FARMHASH_LIBRARIES
+    FARMHASH_INCLUDE_DIR
+)

cmake/Modules/Findmetrohash.cmake

@@ -0,0 +1,44 @@
+# - Try to find metrohash headers and libraries.
+#
+# Usage of this module as follows:
+#
+#     find_package(metrohash)
+#
+# Variables used by this module, they can change the default behaviour and need
+# to be set before calling find_package:
+#
+#  METROHASH_ROOT_DIR Set this variable to the root installation of
+#                    metrohash if the module has problems finding
+#                    the proper installation path.
+#
+# Variables defined by this module:
+#
+#  METROHASH_FOUND             System has metrohash libs/headers
+#  METROHASH_LIBRARIES         The metrohash library/libraries
+#  METROHASH_INCLUDE_DIR       The location of metrohash headers
+
+find_path(METROHASH_ROOT_DIR
+    NAMES include/metrohash.h
+)
+
+find_library(METROHASH_LIBRARIES
+    NAMES metrohash
+    PATHS ${METROHASH_ROOT_DIR}/lib ${METROHASH_LIBRARIES_PATHS}
+)
+
+find_path(METROHASH_INCLUDE_DIR
+    NAMES metrohash.h
+    PATHS ${METROHASH_ROOT_DIR}/include ${METROHASH_INCLUDE_PATHS}
+)
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(metrohash DEFAULT_MSG
+    METROHASH_LIBRARIES
+    METROHASH_INCLUDE_DIR
+)
+
+mark_as_advanced(
+    METROHASH_ROOT_DIR
+    METROHASH_LIBRARIES
+    METROHASH_INCLUDE_DIR
+)

cmake/add_check.cmake

@@ -1,14 +1,16 @@
 # Adding test output on failure
 enable_testing ()
 
-if (CMAKE_CONFIGURATION_TYPES)
+if (NOT TARGET check)
+    if (CMAKE_CONFIGURATION_TYPES)
         add_custom_target (check COMMAND ${CMAKE_CTEST_COMMAND}
             --force-new-ctest-process --output-on-failure --build-config "$<CONFIGURATION>"
             WORKING_DIRECTORY ${CMAKE_BINARY_DIR})
-else ()
+    else ()
         add_custom_target (check COMMAND ${CMAKE_CTEST_COMMAND}
             --force-new-ctest-process --output-on-failure
             WORKING_DIRECTORY ${CMAKE_BINARY_DIR})
+    endif ()
 endif ()
 
 macro (add_check target)

cmake/find_boost.cmake

@@ -16,9 +16,6 @@ if (NOT USE_INTERNAL_BOOST_LIBRARY)
         set (Boost_SYSTEM_LIBRARY "")
     endif ()
 
-    if (Boost_INCLUDE_DIRS)
-        include_directories (${Boost_INCLUDE_DIRS})
-    endif ()
 endif ()
 
 if (NOT Boost_SYSTEM_LIBRARY)
@@ -28,7 +25,6 @@ if (NOT Boost_SYSTEM_LIBRARY)
     set (Boost_SYSTEM_LIBRARY boost_system_internal)
     set (Boost_FILESYSTEM_LIBRARY boost_filesystem_internal)
     set (Boost_INCLUDE_DIRS "${ClickHouse_SOURCE_DIR}/contrib/libboost/boost_1_62_0/")
-    include_directories (BEFORE ${Boost_INCLUDE_DIRS})
 endif ()
 
 message (STATUS "Using Boost: ${Boost_INCLUDE_DIRS} : ${Boost_PROGRAM_OPTIONS_LIBRARY},${Boost_SYSTEM_LIBRARY},${Boost_FILESYSTEM_LIBRARY}")

cmake/find_contrib_lib.cmake

@@ -0,0 +1,21 @@
+macro(find_contrib_lib LIB_NAME)
+
+    string(TOLOWER ${LIB_NAME} LIB_NAME_LC)
+    string(TOUPPER ${LIB_NAME} LIB_NAME_UC)
+    string(REPLACE "-" "_" LIB_NAME_UC ${LIB_NAME_UC})
+
+    option (USE_INTERNAL_${LIB_NAME_UC}_LIBRARY "Use bundled library ${LIB_NAME} instead of system" ${NOT_UNBUNDLED})
+
+    if (NOT USE_INTERNAL_${LIB_NAME_UC}_LIBRARY)
+        find_package ("${LIB_NAME}")
+    endif ()
+
+    if (NOT ${LIB_NAME_UC}_FOUND)
+        set (USE_INTERNAL_${LIB_NAME_UC}_LIBRARY 1)
+        set (${LIB_NAME_UC}_LIBRARIES ${LIB_NAME_LC})
+        set (${LIB_NAME_UC}_INCLUDE_DIR ${${LIB_NAME_UC}_CONTRIB_INCLUDE_DIR})
+    endif ()
+
+    message (STATUS "Using ${LIB_NAME}: ${${LIB_NAME_UC}_INCLUDE_DIR} : ${${LIB_NAME_UC}_LIBRARIES}")
+
+endmacro()

cmake/find_double-conversion.cmake

@@ -1,17 +0,0 @@
-option (USE_INTERNAL_DOUBLE_CONVERSION_LIBRARY "Set to FALSE to use system double-conversion library instead of bundled" ${NOT_UNBUNDLED})
-
-if (NOT USE_INTERNAL_DOUBLE_CONVERSION_LIBRARY)
-    find_library (DOUBLE_CONVERSION_LIBRARY double-conversion)
-    find_path (DOUBLE_CONVERSION_INCLUDE_DIR NAMES double-conversion/double-conversion.h PATHS ${DOUBLE_CONVERSION_INCLUDE_PATHS})
-endif ()
-
-if (DOUBLE_CONVERSION_LIBRARY AND DOUBLE_CONVERSION_INCLUDE_DIR)
-    include_directories (${DOUBLE_CONVERSION_INCLUDE_DIR})
-else ()
-    set (USE_INTERNAL_DOUBLE_CONVERSION_LIBRARY 1)
-    set (DOUBLE_CONVERSION_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/libdouble-conversion")
-    include_directories (BEFORE ${DOUBLE_CONVERSION_INCLUDE_DIR})
-    set (DOUBLE_CONVERSION_LIBRARY double-conversion)
-endif ()
-
-message (STATUS "Using double-conversion: ${DOUBLE_CONVERSION_INCLUDE_DIR} : ${DOUBLE_CONVERSION_LIBRARY}")

cmake/find_lz4.cmake

@@ -9,8 +9,6 @@ if (LZ4_LIBRARY AND LZ4_INCLUDE_DIR)
     include_directories (${LZ4_INCLUDE_DIR})
 else ()
     set (USE_INTERNAL_LZ4_LIBRARY 1)
-    set (LZ4_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/liblz4/include/lz4")
-    include_directories (BEFORE ${LZ4_INCLUDE_DIR})
     set (LZ4_LIBRARY lz4)
 endif ()
 

cmake/find_poco.cmake

@@ -10,6 +10,7 @@ else ()
 
     set (USE_INTERNAL_POCO_LIBRARY 1)
 
+    include (${ClickHouse_SOURCE_DIR}/cmake/find_ltdl.cmake)
     include (${ClickHouse_SOURCE_DIR}/contrib/libpoco/cmake/FindODBC.cmake)
 
     list (APPEND Poco_INCLUDE_DIRS
@@ -29,6 +30,7 @@ else ()
     if (ODBC_FOUND)
         set (Poco_DataODBC_FOUND 1)
         set (Poco_DataODBC_LIBRARY PocoDataODBC)
+        list (APPEND Poco_DataODBC_LIBRARY ${LTDL_LIB})
         list (APPEND Poco_INCLUDE_DIRS "${ClickHouse_SOURCE_DIR}/contrib/libpoco/Data/ODBC/include/")
     endif ()
 

cmake/find_re2.cmake

@@ -10,10 +10,6 @@ if (RE2_LIBRARY AND RE2_INCLUDE_DIR)
     set (RE2_ST_LIBRARY ${RE2_LIBRARY})
 else ()
     set (USE_INTERNAL_RE2_LIBRARY 1)
-    set (RE2_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/libre2")
-    set (RE2_ST_INCLUDE_DIR "${ClickHouse_BINARY_DIR}/contrib/libre2")
-    include_directories (BEFORE ${RE2_INCLUDE_DIR})
-    include_directories (BEFORE ${RE2_ST_INCLUDE_DIR})
     set (RE2_LIBRARY re2)
     set (RE2_ST_LIBRARY re2_st)
     set (USE_RE2_ST 1)

cmake/find_sparsehash.cmake

@@ -9,7 +9,6 @@ if (SPARCEHASH_INCLUDE_DIR)
 else ()
     set (USE_INTERNAL_SPARCEHASH_LIBRARY 1)
     set (SPARCEHASH_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/libsparsehash")
-    include_directories (BEFORE ${SPARCEHASH_INCLUDE_DIR})
 endif ()
 
 message (STATUS "Using sparsehash: ${SPARCEHASH_INCLUDE_DIR}")

cmake/find_zlib.cmake

@@ -2,15 +2,11 @@ option (USE_INTERNAL_ZLIB_LIBRARY "Set to FALSE to use system zlib library inste
 
 if (NOT USE_INTERNAL_ZLIB_LIBRARY)
     find_package (ZLIB)
-    if (ZLIB_FOUND)
-        include_directories (${ZLIB_INCLUDE_DIRS})
-    endif ()
 endif ()
 
 if (NOT ZLIB_FOUND)
     set (USE_INTERNAL_ZLIB_LIBRARY 1)
     set (ZLIB_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/libzlib-ng")
-    include_directories (BEFORE ${ZLIB_INCLUDE_DIR})
     if (USE_STATIC_LIBRARIES)
         set (ZLIB_LIBRARIES zlibstatic)
     else ()

cmake/find_zookeeper.cmake

@@ -9,8 +9,6 @@ if (ZOOKEEPER_LIBRARY AND ZOOKEEPER_INCLUDE_DIR)
     include_directories (${ZOOKEEPER_INCLUDE_DIR})
 else ()
     set (USE_INTERNAL_ZOOKEEPER_LIBRARY 1)
-    set (ZOOKEEPER_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/libzookeeper/include")
-    include_directories (BEFORE ${ZOOKEEPER_INCLUDE_DIR})
     set (ZOOKEEPER_LIBRARY zookeeper_mt)
 endif ()
 

cmake/find_zstd.cmake

@@ -9,8 +9,6 @@ if (ZSTD_LIBRARY AND ZSTD_INCLUDE_DIR)
     include_directories (${ZSTD_INCLUDE_DIR})
 else ()
     set (USE_INTERNAL_ZSTD_LIBRARY 1)
-    set (ZSTD_INCLUDE_DIR "${ClickHouse_SOURCE_DIR}/contrib/libzstd/include/zstd")
-    include_directories (BEFORE ${ZSTD_INCLUDE_DIR})
     set (ZSTD_LIBRARY zstd)
 endif ()
 

cmake/lib_name.cmake

@@ -1,10 +1,6 @@
 
-set(CITYHASH_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/contrib/libcityhash/include)
-set(CPUID_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/contrib/libcpuid/include)
 set(DIVIDE_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/contrib/libdivide)
-set(BTRIE_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/contrib/libbtrie/include)
-set(CITYHASH_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/contrib/libcityhash/include)
-set(MYSQLXX_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/libs/libmysqlxx/include)
-set(POCOEXT_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/libs/libpocoext/include)
+set(CITYHASH_CONTRIB_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/contrib/libcityhash/include)
 set(COMMON_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/libs/libcommon/include ${ClickHouse_BINARY_DIR}/libs/libcommon/include)
 set(DBMS_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/dbms/src ${ClickHouse_BINARY_DIR}/dbms/src)
+set(DOUBLE_CONVERSION_CONTRIB_INCLUDE_DIR ${ClickHouse_SOURCE_DIR}/contrib/libdouble-conversion)

cmake/print_include_directories.cmake

@@ -1,4 +1,17 @@
-get_property (dirs TARGET dbms PROPERTY INCLUDE_DIRECTORIES)
+
+# TODO? Maybe recursive collect on all deps
+
+get_property (dirs1 TARGET dbms PROPERTY INCLUDE_DIRECTORIES)
+list(APPEND dirs ${dirs1})
+
+get_property (dirs1 TARGET common PROPERTY INCLUDE_DIRECTORIES)
+list(APPEND dirs ${dirs1})
+
+if (USE_INTERNAL_BOOST_LIBRARY)
+    get_property (dirs1 TARGET ${Boost_PROGRAM_OPTIONS_LIBRARY} PROPERTY INCLUDE_DIRECTORIES)
+    list(APPEND dirs ${dirs1})
+endif ()
+
 list(REMOVE_DUPLICATES dirs)
 file (WRITE ${CMAKE_CURRENT_BINARY_DIR}/include_directories.txt "")
 foreach (dir ${dirs})

cmake/test_compiler.cmake

@@ -1,21 +1,26 @@
 include (CheckCXXSourceCompiles)
 include (CMakePushCheckState)
 
-cmake_push_check_state ()
+if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+# clang4 : -no-pie cause error
+# clang6 : -no-pie cause warning
+else ()
 
-set (TEST_FLAG "-no-pie")
-set (CMAKE_REQUIRED_FLAGS "${TEST_FLAG}")
+    cmake_push_check_state ()
 
-check_cxx_source_compiles("
+    set (TEST_FLAG "-no-pie")
+    set (CMAKE_REQUIRED_FLAGS "${TEST_FLAG}")
+
+    check_cxx_source_compiles("
         int main() {
             return 0;
         }
-" HAVE_NO_PIE)
+        " HAVE_NO_PIE)
 
-set (CMAKE_REQUIRED_FLAGS "")
-
-if (HAVE_NO_PIE)
+    if (HAVE_NO_PIE)
         set (FLAG_NO_PIE ${TEST_FLAG})
-endif ()
+    endif ()
 
-cmake_pop_check_state ()
+    cmake_pop_check_state ()
+
+endif ()

cmake/test_cpu.cmake

@@ -57,7 +57,7 @@ check_cxx_source_compiles("
     }
 " HAVE_POPCNT)
 
-if (HAVE_POPCNT AND NOT AARCH64)
+if (HAVE_POPCNT AND NOT ARCH_AARCH64)
     set (COMPILER_FLAGS "${COMPILER_FLAGS} ${TEST_FLAG}")
 endif ()
 

contrib/CMakeLists.txt

@@ -28,10 +28,21 @@ if (USE_INTERNAL_ZOOKEEPER_LIBRARY)
 	add_subdirectory (libzookeeper)
 endif ()
 
-add_subdirectory (libcityhash)
-add_subdirectory (libfarmhash)
-add_subdirectory (libmetrohash)
-add_subdirectory (libbtrie)
+if (USE_INTERNAL_CITYHASH_LIBRARY)
+    add_subdirectory (libcityhash)
+endif ()
+
+if (USE_INTERNAL_FARMHASH_LIBRARY)
+    add_subdirectory (libfarmhash)
+endif ()
+
+if (USE_INTERNAL_METROHASH_LIBRARY)
+    add_subdirectory (libmetrohash)
+endif ()
+
+if (USE_INTERNAL_BTRIE_LIBRARY)
+    add_subdirectory (libbtrie)
+endif ()
 
 if (USE_INTERNAL_UNWIND_LIBRARY)
 	add_subdirectory (libunwind)
@@ -49,6 +60,6 @@ if (ENABLE_LIBTCMALLOC AND USE_INTERNAL_GPERFTOOLS_LIBRARY)
 	add_subdirectory (libtcmalloc)
 endif ()
 
-if (NOT ARM)
+if (NOT ARCH_ARM)
 	add_subdirectory (libcpuid)
 endif ()

contrib/libboost/CMakeLists.txt

@@ -48,3 +48,8 @@ boost_1_62_0/libs/test/src/unit_test_monitor.cpp
 boost_1_62_0/libs/test/src/unit_test_parameters.cpp
 boost_1_62_0/libs/test/src/xml_log_formatter.cpp
 boost_1_62_0/libs/test/src/xml_report_formatter.cpp)
+
+target_include_directories (boost_program_options_internal BEFORE PUBLIC ${Boost_INCLUDE_DIRS})
+target_include_directories (boost_filesystem_internal BEFORE PUBLIC ${Boost_INCLUDE_DIRS})
+target_include_directories (boost_system_internal BEFORE PUBLIC ${Boost_INCLUDE_DIRS})
+target_include_directories (boost_test_internal BEFORE PUBLIC ${Boost_INCLUDE_DIRS})

contrib/libbtrie/CMakeLists.txt

@@ -1,6 +1,6 @@
-include_directories (BEFORE include)
-
 add_library (btrie
     src/btrie.c
     include/btrie.h
 )
+
+target_include_directories (btrie PUBLIC include)

contrib/libcctz/CMakeLists.txt

@@ -1,9 +1,3 @@
-include_directories (include)
-
-if (CMAKE_SYSTEM MATCHES "FreeBSD")
-	# yes, need linux, because bsd check inside linux in time_zone_libc.cc:24
-	add_definitions (-D__USE_BSD -Dlinux -D_XOPEN_SOURCE=600)
-endif ()
 
 add_library(cctz
     src/time_zone_libc.cc
@@ -24,3 +18,10 @@ add_library(cctz
     include/time_zone.h
     include/civil_time_detail.h
     include/civil_time.h)
+
+if (CMAKE_SYSTEM MATCHES "FreeBSD")
+    # yes, need linux, because bsd check inside linux in time_zone_libc.cc:24
+    target_compile_definitions (cctz PRIVATE __USE_BSD linux _XOPEN_SOURCE=600)
+endif ()
+
+target_include_directories (cctz PUBLIC include)

contrib/libcityhash/CMakeLists.txt

@@ -1,8 +1,9 @@
-include_directories (include src)
-
 add_library(cityhash
 	src/city.cc
 
 	include/citycrc.h
 	include/city.h
 	src/config.h)
+
+target_include_directories (cityhash PUBLIC include)
+target_include_directories (cityhash PRIVATE src)

contrib/libcpuid/CMakeLists.txt

@@ -16,3 +16,5 @@ include/cpuid/rdtsc.h
 include/cpuid/recog_amd.h
 include/cpuid/recog_intel.h
 )
+
+target_include_directories (cpuid PUBLIC include)

contrib/libdouble-conversion/CMakeLists.txt

@@ -18,3 +18,5 @@ double-conversion/strtod.cc
 double-conversion/strtod.h
 double-conversion/utils.h
 )
+
+target_include_directories (double-conversion PUBLIC .)

contrib/libfarmhash/CMakeLists.txt

@@ -1,5 +1,5 @@
-include_directories (${CMAKE_CURRENT_BINARY_DIR})
-
 add_library(farmhash
 	farmhash.cc
 	farmhash.h)
+
+target_include_directories (farmhash PUBLIC ${CMAKE_CURRENT_BINARY_DIR})

contrib/liblz4/CMakeLists.txt

@@ -1,5 +1,3 @@
-include_directories (BEFORE include)
-
 add_library (lz4
     src/lz4.c
     src/lz4hc.c
@@ -8,3 +6,4 @@ add_library (lz4
     include/lz4/lz4hc.h
     include/lz4/lz4opt.h)
 
+target_include_directories(lz4 PUBLIC include/lz4)

contrib/libmetrohash/CMakeLists.txt

@@ -1,5 +1,3 @@
-include_directories (${CMAKE_CURRENT_BINARY_DIR})
-
 if (HAVE_SSE42) # Not used. Pretty easy to port.
     set (SOURCES_SSE42_ONLY src/metrohash128crc.cpp)
 endif ()
@@ -11,3 +9,5 @@ add_library(metrohash
     src/metrohash64.cpp
     src/metrohash128.cpp
     ${SOURCES_SSE42_ONLY})
+
+target_include_directories(metrohash PRIVATE ${CMAKE_CURRENT_BINARY_DIR})

contrib/libre2/CMakeLists.txt

@@ -34,14 +34,15 @@ set (re2_sources
 # re2 changes its state during matching of regular expression, e.g. creates temporary DFA.
 # It uses RWLock to process the same regular expression object from different threads.
 # In order to avoid redundant locks in some cases, we use not thread-safe version of the library (re2_st).
-add_definitions (-DNDEBUG)
-
-include_directories (BEFORE .)
 
 add_library (re2 ${re2_sources})
 add_library (re2_st ${re2_sources})
 
-set_target_properties (re2_st PROPERTIES COMPILE_DEFINITIONS "NO_THREADS;re2=re2_st")
+target_compile_definitions (re2 PRIVATE NDEBUG)
+target_compile_definitions (re2_st PRIVATE NDEBUG NO_THREADS re2=re2_st)
+
+target_include_directories (re2 PUBLIC .)
+target_include_directories (re2_st PRIVATE . PUBLIC ${CMAKE_CURRENT_BINARY_DIR})
 
 file (MAKE_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/re2_st)
 foreach (FILENAME filtered_re2.h re2.h set.h stringpiece.h variadic_function.h)

contrib/libtcmalloc/CMakeLists.txt

@@ -1,39 +1,5 @@
-add_definitions(
-    -DNO_TCMALLOC_SAMPLES
-    -DNDEBUG
-    -DNO_FRAME_POINTER
-    -Wwrite-strings
-    -Wno-sign-compare
-    -Wno-unused-result
-    -Wno-deprecated-declarations
-    -Wno-unused-function
-    -Wno-unused-private-field
-    -fno-builtin-malloc
-    -fno-builtin-free
-    -fno-builtin-realloc
-    -fno-builtin-calloc
-    -fno-builtin-cfree
-    -fno-builtin-memalign
-    -fno-builtin-posix_memalign
-    -fno-builtin-valloc
-    -fno-builtin-pvalloc
-)
 
-if(CMAKE_CXX_COMPILER_ID STREQUAL "Clang" AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 3.9)
-    add_definitions( -Wno-dynamic-exception-spec )
-endif()
-
-if(CMAKE_SYSTEM MATCHES "FreeBSD" AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
-    add_definitions(-Wno-unused-but-set-variable)
-endif()
-
-if(CMAKE_SYSTEM MATCHES "FreeBSD")
-    add_definitions(-D_GNU_SOURCE)
-endif()
-
-include_directories (include src)
-
-message(STATUS "Building: tcmalloc_minimal_internal")
+message (STATUS "Building: tcmalloc_minimal_internal")
 
 add_library (tcmalloc_minimal_internal
 ./src/malloc_hook.cc
@@ -71,3 +37,40 @@ add_library (tcmalloc_minimal_internal
 ./src/raw_printer.cc
 ./src/memory_region_map.cc
 )
+
+target_compile_options (tcmalloc_minimal_internal PUBLIC
+    -DNO_TCMALLOC_SAMPLES
+    -DNDEBUG
+    -DNO_FRAME_POINTER
+    -Wwrite-strings
+    -Wno-sign-compare
+    -Wno-unused-result
+    -Wno-deprecated-declarations
+    -Wno-unused-function
+    -Wno-unused-private-field
+    -fno-builtin-malloc
+    -fno-builtin-free
+    -fno-builtin-realloc
+    -fno-builtin-calloc
+    -fno-builtin-cfree
+    -fno-builtin-memalign
+    -fno-builtin-posix_memalign
+    -fno-builtin-valloc
+    -fno-builtin-pvalloc
+)
+
+if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang" AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 3.9)
+    target_compile_options(tcmalloc_minimal_internal PUBLIC -Wno-dynamic-exception-spec )
+endif ()
+
+if (CMAKE_SYSTEM MATCHES "FreeBSD" AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+    target_compile_options(tcmalloc_minimal_internal PUBLIC -Wno-unused-but-set-variable)
+endif ()
+
+if (CMAKE_SYSTEM MATCHES "FreeBSD")
+    target_compile_definitions(tcmalloc_minimal_internal PUBLIC _GNU_SOURCE)
+endif ()
+
+target_include_directories (tcmalloc_minimal_internal PUBLIC include)
+target_include_directories (tcmalloc_minimal_internal PRIVATE src)
+

contrib/libunwind/CMakeLists.txt

@@ -1,6 +1,3 @@
-add_definitions(-DHAVE_CONFIG_H=1 -D_XOPEN_SOURCE -D_GNU_SOURCE -Wno-visibility -Wno-header-guard)
-
-include_directories(include include/tdep src)
 
 enable_language(ASM)
 
@@ -55,3 +52,10 @@ src/elf64.c
 src/os-linux.c
 src/x86_64/Los-linux.c
 )
+
+target_compile_definitions (unwind PRIVATE HAVE_CONFIG_H=1 _XOPEN_SOURCE _GNU_SOURCE)
+target_compile_options (unwind PRIVATE -Wno-visibility -Wno-header-guard)
+
+target_include_directories (unwind PUBLIC include)
+target_include_directories (unwind PRIVATE include/tdep)
+target_include_directories (unwind PRIVATE src)

contrib/libzookeeper/CMakeLists.txt

@@ -1,8 +1,3 @@
-add_definitions(-DHAVE_CONFIG_H -DTHREADED)
-
-if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
-	add_definitions(-Wno-unused-but-set-variable)
-endif()
 
 add_library (zookeeper_mt
 src/zookeeper.c
@@ -15,6 +10,12 @@ src/hashtable/hashtable.c
 src/hashtable/hashtable_itr.c
 )
 
+target_compile_definitions (zookeeper_mt PRIVATE HAVE_CONFIG_H THREADED)
+
+if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+    target_compile_options (zookeeper_mt PRIVATE -Wno-unused-but-set-variable)
+endif()
+
 target_include_directories (zookeeper_mt
     PRIVATE include/zookeeper
     PRIVATE src

contrib/libzstd/CMakeLists.txt

@@ -51,30 +51,42 @@ MESSAGE(STATUS "ZSTD VERSION ${LIBVER_MAJOR}.${LIBVER_MINOR}.${LIBVER_RELEASE}")
 
 SET(Sources
         ${LIBRARY_DIR}/common/entropy_common.c
-        ${LIBRARY_DIR}/common/zstd_common.c
-        ${LIBRARY_DIR}/common/xxhash.c
+        ${LIBRARY_DIR}/common/error_private.c
         ${LIBRARY_DIR}/common/fse_decompress.c
+        ${LIBRARY_DIR}/common/pool.c
+        ${LIBRARY_DIR}/common/threading.c
+        ${LIBRARY_DIR}/common/xxhash.c
+        ${LIBRARY_DIR}/common/zstd_common.c
         ${LIBRARY_DIR}/compress/fse_compress.c
         ${LIBRARY_DIR}/compress/huf_compress.c
-        ${LIBRARY_DIR}/compress/zbuff_compress.c
         ${LIBRARY_DIR}/compress/zstd_compress.c
+        ${LIBRARY_DIR}/compress/zstdmt_compress.c
         ${LIBRARY_DIR}/decompress/huf_decompress.c
-        ${LIBRARY_DIR}/decompress/zbuff_decompress.c
         ${LIBRARY_DIR}/decompress/zstd_decompress.c
+        ${LIBRARY_DIR}/deprecated/zbuff_common.c
+        ${LIBRARY_DIR}/deprecated/zbuff_compress.c
+        ${LIBRARY_DIR}/deprecated/zbuff_decompress.c
+        ${LIBRARY_DIR}/dictBuilder/cover.c
         ${LIBRARY_DIR}/dictBuilder/divsufsort.c
         ${LIBRARY_DIR}/dictBuilder/zdict.c)
 
 SET(Headers
         ${LIBRARY_DIR}/common/bitstream.h
         ${LIBRARY_DIR}/common/error_private.h
-        ${LIBRARY_DIR}/common/error_public.h
         ${LIBRARY_DIR}/common/fse.h
         ${LIBRARY_DIR}/common/huf.h
         ${LIBRARY_DIR}/common/mem.h
-        ${LIBRARY_DIR}/common/zbuff.h
+        ${LIBRARY_DIR}/common/pool.h
+        ${LIBRARY_DIR}/common/threading.h
+        ${LIBRARY_DIR}/common/xxhash.h
+        ${LIBRARY_DIR}/common/zstd_errors.h
         ${LIBRARY_DIR}/common/zstd_internal.h
-        ${LIBRARY_DIR}/zstd.h
-        ${LIBRARY_DIR}/dictBuilder/zdict.h)
+        ${LIBRARY_DIR}/compress/zstdmt_compress.h
+        ${LIBRARY_DIR}/compress/zstd_opt.h
+        ${LIBRARY_DIR}/deprecated/zbuff.h
+        ${LIBRARY_DIR}/dictBuilder/divsufsort.h
+        ${LIBRARY_DIR}/dictBuilder/zdict.h
+        ${LIBRARY_DIR}/zstd.h)
 
 SET(ZSTD_LEGACY_SUPPORT true)
 
@@ -84,11 +96,25 @@ IF (ZSTD_LEGACY_SUPPORT)
     ADD_DEFINITIONS(-D ZSTD_LEGACY_SUPPORT=1)
 
     SET(Sources ${Sources}
-            ${LIBRARY_LEGACY_DIR}/zstd_v06.c)
+            ${LIBRARY_LEGACY_DIR}/zstd_v01.c
+            ${LIBRARY_LEGACY_DIR}/zstd_v02.c
+            ${LIBRARY_LEGACY_DIR}/zstd_v03.c
+            ${LIBRARY_LEGACY_DIR}/zstd_v04.c
+            ${LIBRARY_LEGACY_DIR}/zstd_v05.c
+            ${LIBRARY_LEGACY_DIR}/zstd_v06.c
+            ${LIBRARY_LEGACY_DIR}/zstd_v07.c)
 
     SET(Headers ${Headers}
             ${LIBRARY_LEGACY_DIR}/zstd_legacy.h
-            ${LIBRARY_LEGACY_DIR}/zstd_v06.h)
+            ${LIBRARY_LEGACY_DIR}/zstd_v01.h
+            ${LIBRARY_LEGACY_DIR}/zstd_v02.h
+            ${LIBRARY_LEGACY_DIR}/zstd_v03.h
+            ${LIBRARY_LEGACY_DIR}/zstd_v04.h
+            ${LIBRARY_LEGACY_DIR}/zstd_v05.h
+            ${LIBRARY_LEGACY_DIR}/zstd_v06.h
+            ${LIBRARY_LEGACY_DIR}/zstd_v07.h)
 ENDIF (ZSTD_LEGACY_SUPPORT)
 
 ADD_LIBRARY(zstd ${Sources} ${Headers})
+
+target_include_directories (zstd PUBLIC include/zstd) 

contrib/libzstd/README

@@ -1 +1 @@
-https://github.com/facebook/zstd/tree/v1.1.0
+https://github.com/facebook/zstd/tree/v1.3.1

contrib/libzstd/include/zstd/common/bitstream.h

@@ -2,7 +2,7 @@
    bitstream
    Part of FSE library
    header file (to include)
-   Copyright (C) 2013-2016, Yann Collet.
+   Copyright (C) 2013-2017, Yann Collet.
 
    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
 
@@ -39,7 +39,6 @@
 extern "C" {
 #endif
 
-
 /*
 *  This API consists of small unitary functions, which must be inlined for best performance.
 *  Since link-time-optimization is not available for all compilers,
@@ -53,6 +52,18 @@ extern "C" {
 #include "error_private.h"  /* error codes and messages */
 
 
+/*-*************************************
+*  Debug
+***************************************/
+#if defined(BIT_DEBUG) && (BIT_DEBUG>=1)
+#  include <assert.h>
+#else
+#  ifndef assert
+#    define assert(condition) ((void)0)
+#  endif
+#endif
+
+
 /*=========================================
 *  Target specific
 =========================================*/
@@ -60,6 +71,10 @@ extern "C" {
 #  include <immintrin.h>   /* support for bextr (experimental) */
 #endif
 
+#define STREAM_ACCUMULATOR_MIN_32  25
+#define STREAM_ACCUMULATOR_MIN_64  57
+#define STREAM_ACCUMULATOR_MIN    ((U32)(MEM_32bits() ? STREAM_ACCUMULATOR_MIN_32 : STREAM_ACCUMULATOR_MIN_64))
+
 
 /*-******************************************
 *  bitStream encoding API (write forward)
@@ -71,7 +86,7 @@ extern "C" {
 typedef struct
 {
     size_t bitContainer;
-    int    bitPos;
+    unsigned bitPos;
     char*  startPtr;
     char*  ptr;
     char*  endPtr;
@@ -109,6 +124,7 @@ typedef struct
     unsigned bitsConsumed;
     const char* ptr;
     const char* start;
+    const char* limitPtr;
 } BIT_DStream_t;
 
 typedef enum { BIT_DStream_unfinished = 0,
@@ -160,7 +176,10 @@ MEM_STATIC unsigned BIT_highbit32 (register U32 val)
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* Use GCC Intrinsic */
     return 31 - __builtin_clz (val);
 #   else   /* Software version */
-    static const unsigned DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
+    static const unsigned DeBruijnClz[32] = { 0,  9,  1, 10, 13, 21,  2, 29,
+                                             11, 14, 16, 18, 22, 25,  3, 30,
+                                              8, 12, 20, 28, 15, 17, 24,  7,
+                                             19, 27, 23,  6, 26,  5,  4, 31 };
     U32 v = val;
     v |= v >> 1;
     v |= v >> 2;
@@ -172,31 +191,36 @@ MEM_STATIC unsigned BIT_highbit32 (register U32 val)
 }
 
 /*=====    Local Constants   =====*/
-static const unsigned BIT_mask[] = { 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, 0x1FF, 0x3FF, 0x7FF, 0xFFF, 0x1FFF, 0x3FFF, 0x7FFF, 0xFFFF, 0x1FFFF, 0x3FFFF, 0x7FFFF, 0xFFFFF, 0x1FFFFF, 0x3FFFFF, 0x7FFFFF,  0xFFFFFF, 0x1FFFFFF, 0x3FFFFFF };   /* up to 26 bits */
+static const unsigned BIT_mask[] = { 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F,
+                                    0xFF, 0x1FF, 0x3FF, 0x7FF, 0xFFF, 0x1FFF, 0x3FFF, 0x7FFF,
+                                    0xFFFF, 0x1FFFF, 0x3FFFF, 0x7FFFF, 0xFFFFF, 0x1FFFFF, 0x3FFFFF, 0x7FFFFF,
+                                    0xFFFFFF, 0x1FFFFFF, 0x3FFFFFF };   /* up to 26 bits */
 
 
 /*-**************************************************************
 *  bitStream encoding
 ****************************************************************/
 /*! BIT_initCStream() :
- *  `dstCapacity` must be > sizeof(void*)
+ *  `dstCapacity` must be > sizeof(size_t)
  *  @return : 0 if success,
               otherwise an error code (can be tested using ERR_isError() ) */
-MEM_STATIC size_t BIT_initCStream(BIT_CStream_t* bitC, void* startPtr, size_t dstCapacity)
+MEM_STATIC size_t BIT_initCStream(BIT_CStream_t* bitC,
+                                  void* startPtr, size_t dstCapacity)
 {
     bitC->bitContainer = 0;
     bitC->bitPos = 0;
     bitC->startPtr = (char*)startPtr;
     bitC->ptr = bitC->startPtr;
-    bitC->endPtr = bitC->startPtr + dstCapacity - sizeof(bitC->ptr);
-    if (dstCapacity <= sizeof(bitC->ptr)) return ERROR(dstSize_tooSmall);
+    bitC->endPtr = bitC->startPtr + dstCapacity - sizeof(bitC->bitContainer);
+    if (dstCapacity <= sizeof(bitC->bitContainer)) return ERROR(dstSize_tooSmall);
     return 0;
 }
 
 /*! BIT_addBits() :
     can add up to 26 bits into `bitC`.
     Does not check for register overflow ! */
-MEM_STATIC void BIT_addBits(BIT_CStream_t* bitC, size_t value, unsigned nbBits)
+MEM_STATIC void BIT_addBits(BIT_CStream_t* bitC,
+                            size_t value, unsigned nbBits)
 {
     bitC->bitContainer |= (value & BIT_mask[nbBits]) << bitC->bitPos;
     bitC->bitPos += nbBits;
@@ -204,34 +228,42 @@ MEM_STATIC void BIT_addBits(BIT_CStream_t* bitC, size_t value, unsigned nbBits)
 
 /*! BIT_addBitsFast() :
  *  works only if `value` is _clean_, meaning all high bits above nbBits are 0 */
-MEM_STATIC void BIT_addBitsFast(BIT_CStream_t* bitC, size_t value, unsigned nbBits)
+MEM_STATIC void BIT_addBitsFast(BIT_CStream_t* bitC,
+                                size_t value, unsigned nbBits)
 {
+    assert((value>>nbBits) == 0);
     bitC->bitContainer |= value << bitC->bitPos;
     bitC->bitPos += nbBits;
 }
 
 /*! BIT_flushBitsFast() :
+ *  assumption : bitContainer has not overflowed
  *  unsafe version; does not check buffer overflow */
 MEM_STATIC void BIT_flushBitsFast(BIT_CStream_t* bitC)
 {
     size_t const nbBytes = bitC->bitPos >> 3;
+    assert( bitC->bitPos <= (sizeof(bitC->bitContainer)*8) );
     MEM_writeLEST(bitC->ptr, bitC->bitContainer);
     bitC->ptr += nbBytes;
+    assert(bitC->ptr <= bitC->endPtr);
     bitC->bitPos &= 7;
-    bitC->bitContainer >>= nbBytes*8;   /* if bitPos >= sizeof(bitContainer)*8 --> undefined behavior */
+    bitC->bitContainer >>= nbBytes*8;
 }
 
 /*! BIT_flushBits() :
+ *  assumption : bitContainer has not overflowed
  *  safe version; check for buffer overflow, and prevents it.
- *  note : does not signal buffer overflow. This will be revealed later on using BIT_closeCStream() */
+ *  note : does not signal buffer overflow.
+ *  overflow will be revealed later on using BIT_closeCStream() */
 MEM_STATIC void BIT_flushBits(BIT_CStream_t* bitC)
 {
     size_t const nbBytes = bitC->bitPos >> 3;
+    assert( bitC->bitPos <= (sizeof(bitC->bitContainer)*8) );
     MEM_writeLEST(bitC->ptr, bitC->bitContainer);
     bitC->ptr += nbBytes;
     if (bitC->ptr > bitC->endPtr) bitC->ptr = bitC->endPtr;
     bitC->bitPos &= 7;
-    bitC->bitContainer >>= nbBytes*8;   /* if bitPos >= sizeof(bitContainer)*8 --> undefined behavior */
+    bitC->bitContainer >>= nbBytes*8;
 }
 
 /*! BIT_closeCStream() :
@@ -241,9 +273,7 @@ MEM_STATIC size_t BIT_closeCStream(BIT_CStream_t* bitC)
 {
     BIT_addBitsFast(bitC, 1, 1);   /* endMark */
     BIT_flushBits(bitC);
-
-    if (bitC->ptr >= bitC->endPtr) return 0; /* doesn't fit within authorized budget : cancel */
-
+    if (bitC->ptr >= bitC->endPtr) return 0; /* overflow detected */
     return (bitC->ptr - bitC->startPtr) + (bitC->bitPos > 0);
 }
 
@@ -261,26 +291,39 @@ MEM_STATIC size_t BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, si
 {
     if (srcSize < 1) { memset(bitD, 0, sizeof(*bitD)); return ERROR(srcSize_wrong); }
 
-    if (srcSize >=  sizeof(bitD->bitContainer)) {  /* normal case */
     bitD->start = (const char*)srcBuffer;
+    bitD->limitPtr = bitD->start + sizeof(bitD->bitContainer);
+
+    if (srcSize >=  sizeof(bitD->bitContainer)) {  /* normal case */
         bitD->ptr   = (const char*)srcBuffer + srcSize - sizeof(bitD->bitContainer);
         bitD->bitContainer = MEM_readLEST(bitD->ptr);
         { BYTE const lastByte = ((const BYTE*)srcBuffer)[srcSize-1];
-          bitD->bitsConsumed = lastByte ? 8 - BIT_highbit32(lastByte) : 0;
+          bitD->bitsConsumed = lastByte ? 8 - BIT_highbit32(lastByte) : 0;  /* ensures bitsConsumed is always set */
           if (lastByte == 0) return ERROR(GENERIC); /* endMark not present */ }
     } else {
-        bitD->start = (const char*)srcBuffer;
         bitD->ptr   = bitD->start;
         bitD->bitContainer = *(const BYTE*)(bitD->start);
         switch(srcSize)
         {
 	    case 7: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[6]) << (sizeof(bitD->bitContainer)*8 - 16);
+	            /* fall-through */
+
 	    case 6: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[5]) << (sizeof(bitD->bitContainer)*8 - 24);
+	            /* fall-through */
+
 	    case 5: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[4]) << (sizeof(bitD->bitContainer)*8 - 32);
+	            /* fall-through */
+
 	    case 4: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[3]) << 24;
+	            /* fall-through */
+
 	    case 3: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[2]) << 16;
+	            /* fall-through */
+
 	    case 2: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[1]) <<  8;
-            default:;
+	            /* fall-through */
+
+            default: break;
         }
         { BYTE const lastByte = ((const BYTE*)srcBuffer)[srcSize-1];
           bitD->bitsConsumed = lastByte ? 8 - BIT_highbit32(lastByte) : 0;
@@ -298,7 +341,7 @@ MEM_STATIC size_t BIT_getUpperBits(size_t bitContainer, U32 const start)
 
 MEM_STATIC size_t BIT_getMiddleBits(size_t bitContainer, U32 const start, U32 const nbBits)
 {
-#if defined(__BMI__) && defined(__GNUC__)   /* experimental */
+#if defined(__BMI__) && defined(__GNUC__) && __GNUC__*1000+__GNUC_MINOR__ >= 4008  /* experimental */
 #  if defined(__x86_64__)
     if (sizeof(bitContainer)==8)
         return _bextr_u64(bitContainer, start, nbBits);
@@ -327,17 +370,18 @@ MEM_STATIC size_t BIT_getLowerBits(size_t bitContainer, U32 const nbBits)
 #if defined(__BMI__) && defined(__GNUC__)   /* experimental; fails if bitD->bitsConsumed + nbBits > sizeof(bitD->bitContainer)*8 */
     return BIT_getMiddleBits(bitD->bitContainer, (sizeof(bitD->bitContainer)*8) - bitD->bitsConsumed - nbBits, nbBits);
 #else
-    U32 const bitMask = sizeof(bitD->bitContainer)*8 - 1;
-    return ((bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> 1) >> ((bitMask-nbBits) & bitMask);
+    U32 const regMask = sizeof(bitD->bitContainer)*8 - 1;
+    return ((bitD->bitContainer << (bitD->bitsConsumed & regMask)) >> 1) >> ((regMask-nbBits) & regMask);
 #endif
 }
 
 /*! BIT_lookBitsFast() :
-*   unsafe version; only works only if nbBits >= 1 */
+ *  unsafe version; only works if nbBits >= 1 */
 MEM_STATIC size_t BIT_lookBitsFast(const BIT_DStream_t* bitD, U32 nbBits)
 {
-    U32 const bitMask = sizeof(bitD->bitContainer)*8 - 1;
-    return (bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> (((bitMask+1)-nbBits) & bitMask);
+    U32 const regMask = sizeof(bitD->bitContainer)*8 - 1;
+    assert(nbBits >= 1);
+    return (bitD->bitContainer << (bitD->bitsConsumed & regMask)) >> (((regMask+1)-nbBits) & regMask);
 }
 
 MEM_STATIC void BIT_skipBits(BIT_DStream_t* bitD, U32 nbBits)
@@ -362,21 +406,22 @@ MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
 MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, U32 nbBits)
 {
     size_t const value = BIT_lookBitsFast(bitD, nbBits);
+    assert(nbBits >= 1);
     BIT_skipBits(bitD, nbBits);
     return value;
 }
 
 /*! BIT_reloadDStream() :
-*   Refill `BIT_DStream_t` from src buffer previously defined (see BIT_initDStream() ).
+*   Refill `bitD` from buffer previously set in BIT_initDStream() .
 *   This function is safe, it guarantees it will not read beyond src buffer.
 *   @return : status of `BIT_DStream_t` internal register.
-              if status == unfinished, internal register is filled with >= (sizeof(bitD->bitContainer)*8 - 7) bits */
+              if status == BIT_DStream_unfinished, internal register is filled with >= (sizeof(bitD->bitContainer)*8 - 7) bits */
 MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD)
 {
-	if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* should not happen => corruption detected */
+    if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* overflow detected, like end of stream */
         return BIT_DStream_overflow;
 
-    if (bitD->ptr >= bitD->start + sizeof(bitD->bitContainer)) {
+    if (bitD->ptr >= bitD->limitPtr) {
         bitD->ptr -= bitD->bitsConsumed >> 3;
         bitD->bitsConsumed &= 7;
         bitD->bitContainer = MEM_readLEST(bitD->ptr);
@@ -386,6 +431,7 @@ MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD)
         if (bitD->bitsConsumed < sizeof(bitD->bitContainer)*8) return BIT_DStream_endOfBuffer;
         return BIT_DStream_completed;
     }
+    /* start < ptr < limitPtr */
     {   U32 nbBytes = bitD->bitsConsumed >> 3;
         BIT_DStream_status result = BIT_DStream_unfinished;
         if (bitD->ptr - nbBytes < bitD->start) {
@@ -394,7 +440,7 @@ MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD)
         }
         bitD->ptr -= nbBytes;
         bitD->bitsConsumed -= nbBytes*8;
-        bitD->bitContainer = MEM_readLEST(bitD->ptr);   /* reminder : srcSize > sizeof(bitD) */
+        bitD->bitContainer = MEM_readLEST(bitD->ptr);   /* reminder : srcSize > sizeof(bitD->bitContainer), otherwise bitD->ptr == bitD->start */
         return result;
     }
 }

contrib/libzstd/include/zstd/common/entropy_common.c

@@ -43,27 +43,21 @@
 #include "huf.h"
 
 
-/*-****************************************
-*  FSE Error Management
-******************************************/
-unsigned FSE_isError(size_t code) { return ERR_isError(code); }
+/*===   Version   ===*/
+unsigned FSE_versionNumber(void) { return FSE_VERSION_NUMBER; }
 
+
+/*===   Error Management   ===*/
+unsigned FSE_isError(size_t code) { return ERR_isError(code); }
 const char* FSE_getErrorName(size_t code) { return ERR_getErrorName(code); }
 
-
-/* **************************************************************
-*  HUF Error Management
-****************************************************************/
 unsigned HUF_isError(size_t code) { return ERR_isError(code); }
-
 const char* HUF_getErrorName(size_t code) { return ERR_getErrorName(code); }
 
 
 /*-**************************************************************
 *  FSE NCount encoding-decoding
 ****************************************************************/
-static short FSE_abs(short a) { return (short)(a<0 ? -a : a); }
-
 size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSVPtr, unsigned* tableLogPtr,
                  const void* headerBuffer, size_t hbSize)
 {
@@ -117,21 +111,21 @@ size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSVPtr, unsigned* t
             } else {
                 bitStream >>= 2;
         }   }
-        {   short const max = (short)((2*threshold-1)-remaining);
-            short count;
+        {   int const max = (2*threshold-1) - remaining;
+            int count;
 
             if ((bitStream & (threshold-1)) < (U32)max) {
-                count = (short)(bitStream & (threshold-1));
+                count = bitStream & (threshold-1);
                 bitCount += nbBits-1;
             } else {
-                count = (short)(bitStream & (2*threshold-1));
+                count = bitStream & (2*threshold-1);
                 if (count >= threshold) count -= max;
                 bitCount += nbBits;
             }
 
             count--;   /* extra accuracy */
-            remaining -= FSE_abs(count);
-            normalizedCounter[charnum++] = count;
+            remaining -= count < 0 ? -count : count;   /* -1 means +1 */
+            normalizedCounter[charnum++] = (short)count;
             previous0 = !count;
             while (remaining < threshold) {
                 nbBits--;
@@ -159,6 +153,7 @@ size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSVPtr, unsigned* t
 /*! HUF_readStats() :
     Read compact Huffman tree, saved by HUF_writeCTable().
     `huffWeight` is destination buffer.
+    `rankStats` is assumed to be a table of at least HUF_TABLELOG_MAX U32.
     @return : size read from `src` , or an error Code .
     Note : Needed by HUF_readCTable() and HUF_readDTableX?() .
 */
@@ -168,9 +163,11 @@ size_t HUF_readStats(BYTE* huffWeight, size_t hwSize, U32* rankStats,
 {
     U32 weightTotal;
     const BYTE* ip = (const BYTE*) src;
-    size_t iSize = ip[0];
+    size_t iSize;
     size_t oSize;
 
+    if (!srcSize) return ERROR(srcSize_wrong);
+    iSize = ip[0];
     /* memset(huffWeight, 0, hwSize);   *//* is not necessary, even though some analyzer complain ... */
 
     if (iSize >= 128) {  /* special header */
@@ -185,23 +182,25 @@ size_t HUF_readStats(BYTE* huffWeight, size_t hwSize, U32* rankStats,
                 huffWeight[n+1] = ip[n/2] & 15;
     }   }   }
     else  {   /* header compressed with FSE (normal case) */
+        FSE_DTable fseWorkspace[FSE_DTABLE_SIZE_U32(6)];  /* 6 is max possible tableLog for HUF header (maybe even 5, to be tested) */
         if (iSize+1 > srcSize) return ERROR(srcSize_wrong);
-        oSize = FSE_decompress(huffWeight, hwSize-1, ip+1, iSize);   /* max (hwSize-1) values decoded, as last one is implied */
+        oSize = FSE_decompress_wksp(huffWeight, hwSize-1, ip+1, iSize, fseWorkspace, 6);   /* max (hwSize-1) values decoded, as last one is implied */
         if (FSE_isError(oSize)) return oSize;
     }
 
     /* collect weight stats */
-    memset(rankStats, 0, (HUF_TABLELOG_ABSOLUTEMAX + 1) * sizeof(U32));
+    memset(rankStats, 0, (HUF_TABLELOG_MAX + 1) * sizeof(U32));
     weightTotal = 0;
     {   U32 n; for (n=0; n<oSize; n++) {
-            if (huffWeight[n] >= HUF_TABLELOG_ABSOLUTEMAX) return ERROR(corruption_detected);
+            if (huffWeight[n] >= HUF_TABLELOG_MAX) return ERROR(corruption_detected);
             rankStats[huffWeight[n]]++;
             weightTotal += (1 << huffWeight[n]) >> 1;
     }   }
+    if (weightTotal == 0) return ERROR(corruption_detected);
 
     /* get last non-null symbol weight (implied, total must be 2^n) */
     {   U32 const tableLog = BIT_highbit32(weightTotal) + 1;
-        if (tableLog > HUF_TABLELOG_ABSOLUTEMAX) return ERROR(corruption_detected);
+        if (tableLog > HUF_TABLELOG_MAX) return ERROR(corruption_detected);
         *tableLogPtr = tableLog;
         /* determine last weight */
         {   U32 const total = 1 << tableLog;

contrib/libzstd/include/zstd/common/error_private.c

@@ -0,0 +1,47 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+/* The purpose of this file is to have a single list of error strings embedded in binary */
+
+#include "error_private.h"
+
+const char* ERR_getErrorString(ERR_enum code)
+{
+    static const char* const notErrorCode = "Unspecified error code";
+    switch( code )
+    {
+    case PREFIX(no_error): return "No error detected";
+    case PREFIX(GENERIC):  return "Error (generic)";
+    case PREFIX(prefix_unknown): return "Unknown frame descriptor";
+    case PREFIX(version_unsupported): return "Version not supported";
+    case PREFIX(parameter_unknown): return "Unknown parameter type";
+    case PREFIX(frameParameter_unsupported): return "Unsupported frame parameter";
+    case PREFIX(frameParameter_unsupportedBy32bits): return "Frame parameter unsupported in 32-bits mode";
+    case PREFIX(frameParameter_windowTooLarge): return "Frame requires too much memory for decoding";
+    case PREFIX(compressionParameter_unsupported): return "Compression parameter is not supported";
+    case PREFIX(compressionParameter_outOfBound): return "Compression parameter is out of bound";
+    case PREFIX(init_missing): return "Context should be init first";
+    case PREFIX(memory_allocation): return "Allocation error : not enough memory";
+    case PREFIX(stage_wrong): return "Operation not authorized at current processing stage";
+    case PREFIX(dstSize_tooSmall): return "Destination buffer is too small";
+    case PREFIX(srcSize_wrong): return "Src size is incorrect";
+    case PREFIX(corruption_detected): return "Corrupted block detected";
+    case PREFIX(checksum_wrong): return "Restored data doesn't match checksum";
+    case PREFIX(tableLog_tooLarge): return "tableLog requires too much memory : unsupported";
+    case PREFIX(maxSymbolValue_tooLarge): return "Unsupported max Symbol Value : too large";
+    case PREFIX(maxSymbolValue_tooSmall): return "Specified maxSymbolValue is too small";
+    case PREFIX(dictionary_corrupted): return "Dictionary is corrupted";
+    case PREFIX(dictionary_wrong): return "Dictionary mismatch";
+    case PREFIX(dictionaryCreation_failed): return "Cannot create Dictionary from provided samples";
+    case PREFIX(frameIndex_tooLarge): return "Frame index is too large";
+    case PREFIX(seekableIO): return "An I/O error occurred when reading/seeking";
+    case PREFIX(maxCode):
+    default: return notErrorCode;
+    }
+}

contrib/libzstd/include/zstd/common/error_private.h

@@ -21,7 +21,7 @@ extern "C" {
 *  Dependencies
 ******************************************/
 #include <stddef.h>        /* size_t */
-#include "error_public.h"  /* enum list */
+#include "zstd_errors.h"  /* enum list */
 
 
 /* ****************************************
@@ -62,35 +62,7 @@ ERR_STATIC ERR_enum ERR_getErrorCode(size_t code) { if (!ERR_isError(code)) retu
 *  Error Strings
 ******************************************/
 
-ERR_STATIC const char* ERR_getErrorString(ERR_enum code)
-{
-    static const char* notErrorCode = "Unspecified error code";
-    switch( code )
-    {
-    case PREFIX(no_error): return "No error detected";
-    case PREFIX(GENERIC):  return "Error (generic)";
-    case PREFIX(prefix_unknown): return "Unknown frame descriptor";
-    case PREFIX(version_unsupported): return "Version not supported";
-    case PREFIX(parameter_unknown): return "Unknown parameter type";
-    case PREFIX(frameParameter_unsupported): return "Unsupported frame parameter";
-    case PREFIX(frameParameter_unsupportedBy32bits): return "Frame parameter unsupported in 32-bits mode";
-    case PREFIX(compressionParameter_unsupported): return "Compression parameter is out of bound";
-    case PREFIX(init_missing): return "Context should be init first";
-    case PREFIX(memory_allocation): return "Allocation error : not enough memory";
-    case PREFIX(stage_wrong): return "Operation not authorized at current processing stage";
-    case PREFIX(dstSize_tooSmall): return "Destination buffer is too small";
-    case PREFIX(srcSize_wrong): return "Src size incorrect";
-    case PREFIX(corruption_detected): return "Corrupted block detected";
-    case PREFIX(checksum_wrong): return "Restored data doesn't match checksum";
-    case PREFIX(tableLog_tooLarge): return "tableLog requires too much memory : unsupported";
-    case PREFIX(maxSymbolValue_tooLarge): return "Unsupported max Symbol Value : too large";
-    case PREFIX(maxSymbolValue_tooSmall): return "Specified maxSymbolValue is too small";
-    case PREFIX(dictionary_corrupted): return "Dictionary is corrupted";
-    case PREFIX(dictionary_wrong): return "Dictionary mismatch";
-    case PREFIX(maxCode):
-    default: return notErrorCode;
-    }
-}
+const char* ERR_getErrorString(ERR_enum code);   /* error_private.c */
 
 ERR_STATIC const char* ERR_getErrorName(size_t code)
 {

contrib/libzstd/include/zstd/common/error_public.h

@@ -1,59 +0,0 @@
-/**
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
- * All rights reserved.
- *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
- */
-
-#ifndef ERROR_PUBLIC_H_MODULE
-#define ERROR_PUBLIC_H_MODULE
-
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
-/*===== dependency =====*/
-#include <stddef.h>   /* size_t */
-
-
-/*-****************************************
-*  error codes list
-******************************************/
-typedef enum {
-  ZSTD_error_no_error,
-  ZSTD_error_GENERIC,
-  ZSTD_error_prefix_unknown,
-  ZSTD_error_version_unsupported,
-  ZSTD_error_parameter_unknown,
-  ZSTD_error_frameParameter_unsupported,
-  ZSTD_error_frameParameter_unsupportedBy32bits,
-  ZSTD_error_compressionParameter_unsupported,
-  ZSTD_error_init_missing,
-  ZSTD_error_memory_allocation,
-  ZSTD_error_stage_wrong,
-  ZSTD_error_dstSize_tooSmall,
-  ZSTD_error_srcSize_wrong,
-  ZSTD_error_corruption_detected,
-  ZSTD_error_checksum_wrong,
-  ZSTD_error_tableLog_tooLarge,
-  ZSTD_error_maxSymbolValue_tooLarge,
-  ZSTD_error_maxSymbolValue_tooSmall,
-  ZSTD_error_dictionary_corrupted,
-  ZSTD_error_dictionary_wrong,
-  ZSTD_error_maxCode
-} ZSTD_ErrorCode;
-
-/*! ZSTD_getErrorCode() :
-    convert a `size_t` function result into a `ZSTD_ErrorCode` enum type,
-    which can be used to compare directly with enum list published into "error_public.h" */
-ZSTD_ErrorCode ZSTD_getErrorCode(size_t functionResult);
-const char* ZSTD_getErrorString(ZSTD_ErrorCode code);
-
-
-#if defined (__cplusplus)
-}
-#endif
-
-#endif /* ERROR_PUBLIC_H_MODULE */

contrib/libzstd/include/zstd/common/fse.h

@@ -45,6 +45,32 @@ extern "C" {
 #include <stddef.h>    /* size_t, ptrdiff_t */
 
 
+/*-*****************************************
+*  FSE_PUBLIC_API : control library symbols visibility
+******************************************/
+#if defined(FSE_DLL_EXPORT) && (FSE_DLL_EXPORT==1) && defined(__GNUC__) && (__GNUC__ >= 4)
+#  define FSE_PUBLIC_API __attribute__ ((visibility ("default")))
+#elif defined(FSE_DLL_EXPORT) && (FSE_DLL_EXPORT==1)   /* Visual expected */
+#  define FSE_PUBLIC_API __declspec(dllexport)
+#elif defined(FSE_DLL_IMPORT) && (FSE_DLL_IMPORT==1)
+#  define FSE_PUBLIC_API __declspec(dllimport) /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/
+#else
+#  define FSE_PUBLIC_API
+#endif
+
+/*------   Version   ------*/
+#define FSE_VERSION_MAJOR    0
+#define FSE_VERSION_MINOR    9
+#define FSE_VERSION_RELEASE  0
+
+#define FSE_LIB_VERSION FSE_VERSION_MAJOR.FSE_VERSION_MINOR.FSE_VERSION_RELEASE
+#define FSE_QUOTE(str) #str
+#define FSE_EXPAND_AND_QUOTE(str) FSE_QUOTE(str)
+#define FSE_VERSION_STRING FSE_EXPAND_AND_QUOTE(FSE_LIB_VERSION)
+
+#define FSE_VERSION_NUMBER  (FSE_VERSION_MAJOR *100*100 + FSE_VERSION_MINOR *100 + FSE_VERSION_RELEASE)
+FSE_PUBLIC_API unsigned FSE_versionNumber(void);   /**< library version number; to be used when checking dll version */
+
 /*-****************************************
 *  FSE simple functions
 ******************************************/
@@ -56,7 +82,7 @@ extern "C" {
                      if return == 1, srcData is a single byte symbol * srcSize times. Use RLE compression instead.
                      if FSE_isError(return), compression failed (more details using FSE_getErrorName())
 */
-size_t FSE_compress(void* dst, size_t dstCapacity,
+FSE_PUBLIC_API size_t FSE_compress(void* dst, size_t dstCapacity,
                              const void* src, size_t srcSize);
 
 /*! FSE_decompress():
@@ -69,18 +95,18 @@ size_t FSE_compress(void* dst, size_t dstCapacity,
     Why ? : making this distinction requires a header.
     Header management is intentionally delegated to the user layer, which can better manage special cases.
 */
-size_t FSE_decompress(void* dst,  size_t dstCapacity,
+FSE_PUBLIC_API size_t FSE_decompress(void* dst,  size_t dstCapacity,
                                const void* cSrc, size_t cSrcSize);
 
 
 /*-*****************************************
 *  Tool functions
 ******************************************/
-size_t FSE_compressBound(size_t size);       /* maximum compressed size */
+FSE_PUBLIC_API size_t FSE_compressBound(size_t size);       /* maximum compressed size */
 
 /* Error Management */
-unsigned    FSE_isError(size_t code);        /* tells if a return value is an error code */
-const char* FSE_getErrorName(size_t code);   /* provides error code string (useful for debugging) */
+FSE_PUBLIC_API unsigned    FSE_isError(size_t code);        /* tells if a return value is an error code */
+FSE_PUBLIC_API const char* FSE_getErrorName(size_t code);   /* provides error code string (useful for debugging) */
 
 
 /*-*****************************************
@@ -94,7 +120,7 @@ const char* FSE_getErrorName(size_t code);   /* provides error code string (usef
                      if return == 1, srcData is a single byte symbol * srcSize times. Use RLE compression.
                      if FSE_isError(return), it's an error code.
 */
-size_t FSE_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
+FSE_PUBLIC_API size_t FSE_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
 
 
 /*-*****************************************
@@ -127,50 +153,50 @@ or to save and provide normalized distribution using external method.
     @return : the count of the most frequent symbol (which is not identified).
               if return == srcSize, there is only one symbol.
               Can also return an error code, which can be tested with FSE_isError(). */
-size_t FSE_count(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize);
+FSE_PUBLIC_API size_t FSE_count(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize);
 
 /*! FSE_optimalTableLog():
     dynamically downsize 'tableLog' when conditions are met.
     It saves CPU time, by using smaller tables, while preserving or even improving compression ratio.
     @return : recommended tableLog (necessarily <= 'maxTableLog') */
-unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue);
+FSE_PUBLIC_API unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue);
 
 /*! FSE_normalizeCount():
     normalize counts so that sum(count[]) == Power_of_2 (2^tableLog)
     'normalizedCounter' is a table of short, of minimum size (maxSymbolValue+1).
     @return : tableLog,
               or an errorCode, which can be tested using FSE_isError() */
-size_t FSE_normalizeCount(short* normalizedCounter, unsigned tableLog, const unsigned* count, size_t srcSize, unsigned maxSymbolValue);
+FSE_PUBLIC_API size_t FSE_normalizeCount(short* normalizedCounter, unsigned tableLog, const unsigned* count, size_t srcSize, unsigned maxSymbolValue);
 
 /*! FSE_NCountWriteBound():
     Provides the maximum possible size of an FSE normalized table, given 'maxSymbolValue' and 'tableLog'.
     Typically useful for allocation purpose. */
-size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog);
+FSE_PUBLIC_API size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog);
 
 /*! FSE_writeNCount():
     Compactly save 'normalizedCounter' into 'buffer'.
     @return : size of the compressed table,
               or an errorCode, which can be tested using FSE_isError(). */
-size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
+FSE_PUBLIC_API size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
 
 
 /*! Constructor and Destructor of FSE_CTable.
     Note that FSE_CTable size depends on 'tableLog' and 'maxSymbolValue' */
 typedef unsigned FSE_CTable;   /* don't allocate that. It's only meant to be more restrictive than void* */
-FSE_CTable* FSE_createCTable (unsigned tableLog, unsigned maxSymbolValue);
-void        FSE_freeCTable (FSE_CTable* ct);
+FSE_PUBLIC_API FSE_CTable* FSE_createCTable (unsigned tableLog, unsigned maxSymbolValue);
+FSE_PUBLIC_API void        FSE_freeCTable (FSE_CTable* ct);
 
 /*! FSE_buildCTable():
     Builds `ct`, which must be already allocated, using FSE_createCTable().
     @return : 0, or an errorCode, which can be tested using FSE_isError() */
-size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
+FSE_PUBLIC_API size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
 
 /*! FSE_compress_usingCTable():
     Compress `src` using `ct` into `dst` which must be already allocated.
     @return : size of compressed data (<= `dstCapacity`),
               or 0 if compressed data could not fit into `dst`,
               or an errorCode, which can be tested using FSE_isError() */
-size_t FSE_compress_usingCTable (void* dst, size_t dstCapacity, const void* src, size_t srcSize, const FSE_CTable* ct);
+FSE_PUBLIC_API size_t FSE_compress_usingCTable (void* dst, size_t dstCapacity, const void* src, size_t srcSize, const FSE_CTable* ct);
 
 /*!
 Tutorial :
@@ -223,25 +249,25 @@ If there is an error, the function will return an ErrorCode (which can be tested
     @return : size read from 'rBuffer',
               or an errorCode, which can be tested using FSE_isError().
               maxSymbolValuePtr[0] and tableLogPtr[0] will also be updated with their respective values */
-size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSymbolValuePtr, unsigned* tableLogPtr, const void* rBuffer, size_t rBuffSize);
+FSE_PUBLIC_API size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSymbolValuePtr, unsigned* tableLogPtr, const void* rBuffer, size_t rBuffSize);
 
 /*! Constructor and Destructor of FSE_DTable.
     Note that its size depends on 'tableLog' */
 typedef unsigned FSE_DTable;   /* don't allocate that. It's just a way to be more restrictive than void* */
-FSE_DTable* FSE_createDTable(unsigned tableLog);
-void        FSE_freeDTable(FSE_DTable* dt);
+FSE_PUBLIC_API FSE_DTable* FSE_createDTable(unsigned tableLog);
+FSE_PUBLIC_API void        FSE_freeDTable(FSE_DTable* dt);
 
 /*! FSE_buildDTable():
     Builds 'dt', which must be already allocated, using FSE_createDTable().
     return : 0, or an errorCode, which can be tested using FSE_isError() */
-size_t FSE_buildDTable (FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
+FSE_PUBLIC_API size_t FSE_buildDTable (FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
 
 /*! FSE_decompress_usingDTable():
     Decompress compressed source `cSrc` of size `cSrcSize` using `dt`
     into `dst` which must be already allocated.
     @return : size of regenerated data (necessarily <= `dstCapacity`),
               or an errorCode, which can be tested using FSE_isError() */
-size_t FSE_decompress_usingDTable(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, const FSE_DTable* dt);
+FSE_PUBLIC_API size_t FSE_decompress_usingDTable(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, const FSE_DTable* dt);
 
 /*!
 Tutorial :
@@ -286,45 +312,84 @@ If there is an error, the function will return an error code, which can be teste
 #define FSE_BLOCKBOUND(size) (size + (size>>7))
 #define FSE_COMPRESSBOUND(size) (FSE_NCOUNTBOUND + FSE_BLOCKBOUND(size))   /* Macro version, useful for static allocation */
 
-/* It is possible to statically allocate FSE CTable/DTable as a table of unsigned using below macros */
+/* It is possible to statically allocate FSE CTable/DTable as a table of FSE_CTable/FSE_DTable using below macros */
 #define FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue)   (1 + (1<<(maxTableLog-1)) + ((maxSymbolValue+1)*2))
 #define FSE_DTABLE_SIZE_U32(maxTableLog)                   (1 + (1<<maxTableLog))
 
+/* or use the size to malloc() space directly. Pay attention to alignment restrictions though */
+#define FSE_CTABLE_SIZE(maxTableLog, maxSymbolValue)   (FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) * sizeof(FSE_CTable))
+#define FSE_DTABLE_SIZE(maxTableLog)                   (FSE_DTABLE_SIZE_U32(maxTableLog) * sizeof(FSE_DTable))
+
 
 /* *****************************************
 *  FSE advanced API
 *******************************************/
+/* FSE_count_wksp() :
+ * Same as FSE_count(), but using an externally provided scratch buffer.
+ * `workSpace` size must be table of >= `1024` unsigned
+ */
+size_t FSE_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+                 const void* source, size_t sourceSize, unsigned* workSpace);
+
+/** FSE_countFast() :
+ *  same as FSE_count(), but blindly trusts that all byte values within src are <= *maxSymbolValuePtr
+ */
 size_t FSE_countFast(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize);
-/**< same as FSE_count(), but blindly trusts that all byte values within src are <= *maxSymbolValuePtr  */
+
+/* FSE_countFast_wksp() :
+ * Same as FSE_countFast(), but using an externally provided scratch buffer.
+ * `workSpace` must be a table of minimum `1024` unsigned
+ */
+size_t FSE_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize, unsigned* workSpace);
+
+/*! FSE_count_simple
+ * Same as FSE_countFast(), but does not use any additional memory (not even on stack).
+ * This function is unsafe, and will segfault if any value within `src` is `> *maxSymbolValuePtr` (presuming it's also the size of `count`).
+*/
+size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize);
+
+
 
 unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue, unsigned minus);
 /**< same as FSE_optimalTableLog(), which used `minus==2` */
 
+/* FSE_compress_wksp() :
+ * Same as FSE_compress2(), but using an externally allocated scratch buffer (`workSpace`).
+ * FSE_WKSP_SIZE_U32() provides the minimum size required for `workSpace` as a table of FSE_CTable.
+ */
+#define FSE_WKSP_SIZE_U32(maxTableLog, maxSymbolValue)   ( FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue) + ((maxTableLog > 12) ? (1 << (maxTableLog - 2)) : 1024) )
+size_t FSE_compress_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);
+
 size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits);
-/**< build a fake FSE_CTable, designed to not compress an input, where each symbol uses nbBits */
+/**< build a fake FSE_CTable, designed for a flat distribution, where each symbol uses nbBits */
 
 size_t FSE_buildCTable_rle (FSE_CTable* ct, unsigned char symbolValue);
 /**< build a fake FSE_CTable, designed to compress always the same symbolValue */
 
+/* FSE_buildCTable_wksp() :
+ * Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`).
+ * `wkspSize` must be >= `(1<<tableLog)`.
+ */
+size_t FSE_buildCTable_wksp(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);
+
 size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits);
-/**< build a fake FSE_DTable, designed to read an uncompressed bitstream where each symbol uses nbBits */
+/**< build a fake FSE_DTable, designed to read a flat distribution where each symbol uses nbBits */
 
 size_t FSE_buildDTable_rle (FSE_DTable* dt, unsigned char symbolValue);
 /**< build a fake FSE_DTable, designed to always generate the same symbolValue */
 
+size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, FSE_DTable* workSpace, unsigned maxLog);
+/**< same as FSE_decompress(), using an externally allocated `workSpace` produced with `FSE_DTABLE_SIZE_U32(maxLog)` */
+
 
 /* *****************************************
 *  FSE symbol compression API
 *******************************************/
 /*!
    This API consists of small unitary functions, which highly benefit from being inlined.
-   You will want to enable link-time-optimization to ensure these functions are properly inlined in your binary.
-   Visual seems to do it automatically.
-   For gcc or clang, you'll need to add -flto flag at compilation and linking stages.
-   If none of these solutions is applicable, include "fse.c" directly.
+   Hence their body are included in next section.
 */
-typedef struct
-{
+typedef struct {
     ptrdiff_t   value;
     const void* stateTable;
     const void* symbolTT;
@@ -384,8 +449,7 @@ If there is an error, it returns an errorCode (which can be tested using FSE_isE
 /* *****************************************
 *  FSE symbol decompression API
 *******************************************/
-typedef struct
-{
+typedef struct {
     size_t      state;
     const void* table;   /* precise table may vary, depending on U16 */
 } FSE_DState_t;
@@ -490,9 +554,9 @@ MEM_STATIC void FSE_initCState2(FSE_CState_t* statePtr, const FSE_CTable* ct, U3
 
 MEM_STATIC void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* statePtr, U32 symbol)
 {
-    const FSE_symbolCompressionTransform symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol];
+    FSE_symbolCompressionTransform const symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol];
     const U16* const stateTable = (const U16*)(statePtr->stateTable);
-    U32 nbBitsOut  = (U32)((statePtr->value + symbolTT.deltaNbBits) >> 16);
+    U32 const nbBitsOut  = (U32)((statePtr->value + symbolTT.deltaNbBits) >> 16);
     BIT_addBits(bitC, statePtr->value, nbBitsOut);
     statePtr->value = stateTable[ (statePtr->value >> nbBitsOut) + symbolTT.deltaFindState];
 }
@@ -503,6 +567,7 @@ MEM_STATIC void FSE_flushCState(BIT_CStream_t* bitC, const FSE_CState_t* statePt
     BIT_flushBits(bitC);
 }
 
+
 /* ======    Decompression    ====== */
 
 typedef struct {
@@ -581,14 +646,19 @@ MEM_STATIC unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr)
 *  Increasing memory usage improves compression ratio
 *  Reduced memory usage can improve speed, due to cache effect
 *  Recommended max value is 14, for 16KB, which nicely fits into Intel x86 L1 cache */
-#define FSE_MAX_MEMORY_USAGE 14
-#define FSE_DEFAULT_MEMORY_USAGE 13
+#ifndef FSE_MAX_MEMORY_USAGE
+#  define FSE_MAX_MEMORY_USAGE 14
+#endif
+#ifndef FSE_DEFAULT_MEMORY_USAGE
+#  define FSE_DEFAULT_MEMORY_USAGE 13
+#endif
 
 /*!FSE_MAX_SYMBOL_VALUE :
 *  Maximum symbol value authorized.
 *  Required for proper stack allocation */
-#define FSE_MAX_SYMBOL_VALUE 255
-
+#ifndef FSE_MAX_SYMBOL_VALUE
+#  define FSE_MAX_SYMBOL_VALUE 255
+#endif
 
 /* **************************************************************
 *  template functions type & suffix

contrib/libzstd/include/zstd/common/fse_decompress.c

@@ -59,7 +59,6 @@
 ****************************************************************/
 #include <stdlib.h>     /* malloc, free, qsort */
 #include <string.h>     /* memcpy, memset */
-#include <stdio.h>      /* printf (debug) */
 #include "bitstream.h"
 #define FSE_STATIC_LINKING_ONLY
 #include "fse.h"
@@ -75,12 +74,6 @@
 #define CHECK_F(f) { size_t const e = f; if (FSE_isError(e)) return e; }
 
 
-/* **************************************************************
-*  Complex types
-****************************************************************/
-typedef U32 DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
-
-
 /* **************************************************************
 *  Templates
 ****************************************************************/
@@ -300,28 +293,34 @@ size_t FSE_decompress_usingDTable(void* dst, size_t originalSize,
 }
 
 
-size_t FSE_decompress(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize)
+size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, FSE_DTable* workSpace, unsigned maxLog)
 {
     const BYTE* const istart = (const BYTE*)cSrc;
     const BYTE* ip = istart;
     short counting[FSE_MAX_SYMBOL_VALUE+1];
-    DTable_max_t dt;   /* Static analyzer seems unable to understand this table will be properly initialized later */
     unsigned tableLog;
     unsigned maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
 
-    if (cSrcSize<2) return ERROR(srcSize_wrong);   /* too small input size */
-
     /* normal FSE decoding mode */
-    {   size_t const NCountLength = FSE_readNCount (counting, &maxSymbolValue, &tableLog, istart, cSrcSize);
+    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 */
+    //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;
-    }
 
-    CHECK_F( FSE_buildDTable (dt, counting, maxSymbolValue, tableLog) );
+    CHECK_F( FSE_buildDTable (workSpace, counting, maxSymbolValue, tableLog) );
 
-    return FSE_decompress_usingDTable (dst, maxDstSize, ip, cSrcSize, dt);   /* always return, even if it is an error code */
+    return FSE_decompress_usingDTable (dst, dstCapacity, ip, cSrcSize, workSpace);   /* always return, even if it is an error code */
+}
+
+
+typedef FSE_DTable DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
+
+size_t FSE_decompress(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize)
+{
+    DTable_max_t dt;   /* Static analyzer seems unable to understand this table will be properly initialized later */
+    return FSE_decompress_wksp(dst, dstCapacity, cSrc, cSrcSize, dt, FSE_MAX_TABLELOG);
 }
 
 

contrib/libzstd/include/zstd/common/huf.h

@@ -43,6 +43,21 @@ extern "C" {
 #include <stddef.h>    /* size_t */
 
 
+/* *** library symbols visibility *** */
+/* Note : when linking with -fvisibility=hidden on gcc, or by default on Visual,
+ *        HUF symbols remain "private" (internal symbols for library only).
+ *        Set macro FSE_DLL_EXPORT to 1 if you want HUF symbols visible on DLL interface */
+#if defined(FSE_DLL_EXPORT) && (FSE_DLL_EXPORT==1) && defined(__GNUC__) && (__GNUC__ >= 4)
+#  define HUF_PUBLIC_API __attribute__ ((visibility ("default")))
+#elif defined(FSE_DLL_EXPORT) && (FSE_DLL_EXPORT==1)   /* Visual expected */
+#  define HUF_PUBLIC_API __declspec(dllexport)
+#elif defined(FSE_DLL_IMPORT) && (FSE_DLL_IMPORT==1)
+#  define HUF_PUBLIC_API __declspec(dllimport)  /* not required, just to generate faster code (saves a function pointer load from IAT and an indirect jump) */
+#else
+#  define HUF_PUBLIC_API
+#endif
+
+
 /* *** simple functions *** */
 /**
 HUF_compress() :
@@ -55,42 +70,68 @@ HUF_compress() :
                      if return == 1, srcData is a single repeated byte symbol (RLE compression).
                      if HUF_isError(return), compression failed (more details using HUF_getErrorName())
 */
-size_t HUF_compress(void* dst, size_t dstCapacity,
+HUF_PUBLIC_API size_t HUF_compress(void* dst, size_t dstCapacity,
                              const void* src, size_t srcSize);
 
 /**
 HUF_decompress() :
     Decompress HUF data from buffer 'cSrc', of size 'cSrcSize',
     into already allocated buffer 'dst', of minimum size 'dstSize'.
-    `dstSize` : **must** be the ***exact*** size of original (uncompressed) data.
+    `originalSize` : **must** be the ***exact*** size of original (uncompressed) data.
     Note : in contrast with FSE, HUF_decompress can regenerate
            RLE (cSrcSize==1) and uncompressed (cSrcSize==dstSize) data,
            because it knows size to regenerate.
-    @return : size of regenerated data (== dstSize),
+    @return : size of regenerated data (== originalSize),
               or an error code, which can be tested using HUF_isError()
 */
-size_t HUF_decompress(void* dst,  size_t dstSize,
+HUF_PUBLIC_API size_t HUF_decompress(void* dst,  size_t originalSize,
                                const void* cSrc, size_t cSrcSize);
 
 
-/* ****************************************
-*  Tool functions
-******************************************/
-#define HUF_BLOCKSIZE_MAX (128 * 1024)
-size_t HUF_compressBound(size_t size);       /**< maximum compressed size (worst case) */
+/* ***   Tool functions *** */
+#define HUF_BLOCKSIZE_MAX (128 * 1024)                  /**< maximum input size for a single block compressed with HUF_compress */
+HUF_PUBLIC_API size_t HUF_compressBound(size_t size);   /**< maximum compressed size (worst case) */
 
 /* Error Management */
-unsigned    HUF_isError(size_t code);        /**< tells if a return value is an error code */
-const char* HUF_getErrorName(size_t code);   /**< provides error code string (useful for debugging) */
+HUF_PUBLIC_API unsigned    HUF_isError(size_t code);       /**< tells if a return value is an error code */
+HUF_PUBLIC_API const char* HUF_getErrorName(size_t code);  /**< provides error code string (useful for debugging) */
 
 
 /* ***   Advanced function   *** */
 
 /** HUF_compress2() :
-*   Same as HUF_compress(), but offers direct control over `maxSymbolValue` and `tableLog` */
-size_t HUF_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
+ *  Same as HUF_compress(), but offers direct control over `maxSymbolValue` and `tableLog`.
+ *  `tableLog` must be `<= HUF_TABLELOG_MAX` . */
+HUF_PUBLIC_API size_t HUF_compress2 (void* dst, size_t dstCapacity, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
+
+/** 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 following macro */
+#define HUF_WORKSPACE_SIZE (6 << 10)
+#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, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);
+
+/**
+ *  The minimum workspace size for the `workSpace` used in
+ *  HUF_readDTableX2_wksp() and HUF_readDTableX4_wksp().
+ *
+ *  The space used depends on HUF_TABLELOG_MAX, ranging from ~1500 bytes when
+ *  HUF_TABLE_LOG_MAX=12 to ~1850 bytes when HUF_TABLE_LOG_MAX=15.
+ *  Buffer overflow errors may potentially occur if code modifications result in
+ *  a required workspace size greater than that specified in the following
+ *  macro.
+ */
+#define HUF_DECOMPRESS_WORKSPACE_SIZE (2 << 10)
+#define HUF_DECOMPRESS_WORKSPACE_SIZE_U32 (HUF_DECOMPRESS_WORKSPACE_SIZE / sizeof(U32))
 
 
+/* ******************************************************************
+ *  WARNING !!
+ *  The following section contains advanced and experimental definitions
+ *  which shall never be used in the context of dll
+ *  because they are not guaranteed to remain stable in the future.
+ *  Only consider them in association with static linking.
+ *******************************************************************/
 #ifdef HUF_STATIC_LINKING_ONLY
 
 /* *** Dependencies *** */
@@ -98,10 +139,11 @@ size_t HUF_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize
 
 
 /* *** Constants *** */
-#define HUF_TABLELOG_ABSOLUTEMAX  16   /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
 #define HUF_TABLELOG_MAX      12       /* max configured tableLog (for static allocation); can be modified up to HUF_ABSOLUTEMAX_TABLELOG */
 #define HUF_TABLELOG_DEFAULT  11       /* tableLog by default, when not specified */
 #define HUF_SYMBOLVALUE_MAX  255
+
+#define HUF_TABLELOG_ABSOLUTEMAX  15   /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
 #if (HUF_TABLELOG_MAX > HUF_TABLELOG_ABSOLUTEMAX)
 #  error "HUF_TABLELOG_MAX is too large !"
 #endif
@@ -112,12 +154,14 @@ size_t HUF_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize
 ******************************************/
 /* HUF buffer bounds */
 #define HUF_CTABLEBOUND 129
-#define HUF_BLOCKBOUND(size) (size + (size>>8) + 8)   /* only true if incompressible pre-filtered with fast heuristic */
+#define HUF_BLOCKBOUND(size) (size + (size>>8) + 8)   /* only true when incompressible is pre-filtered with fast heuristic */
 #define HUF_COMPRESSBOUND(size) (HUF_CTABLEBOUND + HUF_BLOCKBOUND(size))   /* Macro version, useful for static allocation */
 
 /* static allocation of HUF's Compression Table */
+#define HUF_CTABLE_SIZE_U32(maxSymbolValue)   ((maxSymbolValue)+1)   /* Use tables of U32, for proper alignment */
+#define HUF_CTABLE_SIZE(maxSymbolValue)       (HUF_CTABLE_SIZE_U32(maxSymbolValue) * sizeof(U32))
 #define HUF_CREATE_STATIC_CTABLE(name, maxSymbolValue) \
-    U32 name##hb[maxSymbolValue+1]; \
+    U32 name##hb[HUF_CTABLE_SIZE_U32(maxSymbolValue)]; \
     void* name##hv = &(name##hb); \
     HUF_CElt* name = (HUF_CElt*)(name##hv)   /* no final ; */
 
@@ -125,9 +169,9 @@ size_t HUF_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize
 typedef U32 HUF_DTable;
 #define HUF_DTABLE_SIZE(maxTableLog)   (1 + (1<<(maxTableLog)))
 #define HUF_CREATE_STATIC_DTABLEX2(DTable, maxTableLog) \
-        HUF_DTable DTable[HUF_DTABLE_SIZE((maxTableLog)-1)] = { ((U32)((maxTableLog)-1)*0x1000001) }
+        HUF_DTable DTable[HUF_DTABLE_SIZE((maxTableLog)-1)] = { ((U32)((maxTableLog)-1) * 0x01000001) }
 #define HUF_CREATE_STATIC_DTABLEX4(DTable, maxTableLog) \
-        HUF_DTable DTable[HUF_DTABLE_SIZE(maxTableLog)] = { ((U32)(maxTableLog)*0x1000001) }
+        HUF_DTable DTable[HUF_DTABLE_SIZE(maxTableLog)] = { ((U32)(maxTableLog) * 0x01000001) }
 
 
 /* ****************************************
@@ -138,12 +182,11 @@ size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cS
 
 size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< decodes RLE and uncompressed */
 size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); /**< considers RLE and uncompressed as errors */
+size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize); /**< considers RLE and uncompressed as errors */
 size_t HUF_decompress4X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
+size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< single-symbol decoder */
 size_t HUF_decompress4X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
-
-size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
-size_t HUF_decompress1X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
-size_t HUF_decompress1X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
+size_t HUF_decompress4X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< double-symbols decoder */
 
 
 /* ****************************************
@@ -168,6 +211,23 @@ 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);
 
+typedef enum {
+   HUF_repeat_none,  /**< Cannot use the previous table */
+   HUF_repeat_check, /**< Can use the previous table but it must be checked. Note : The previous table must have been constructed by HUF_compress{1, 4}X_repeat */
+   HUF_repeat_valid  /**< Can use the previous table and it is asumed to be valid */
+ } HUF_repeat;
+/** HUF_compress4X_repeat() :
+*   Same as HUF_compress4X_wksp(), but considers using hufTable if *repeat != HUF_repeat_none.
+*   If it uses hufTable it does not modify hufTable or repeat.
+*   If it doesn't, it sets *repeat = HUF_repeat_none, and it sets hufTable to the table used.
+*   If preferRepeat then the old table will always be used if valid. */
+size_t HUF_compress4X_repeat(void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize, HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat);  /**< `workSpace` must be a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
+
+/** 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 1024 unsigned.
+ */
+size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize);
 
 /*! HUF_readStats() :
     Read compact Huffman tree, saved by HUF_writeCTable().
@@ -198,7 +258,9 @@ HUF_decompress() does the following:
 U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize);
 
 size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize);
+size_t HUF_readDTableX2_wksp (HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize);
 size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize);
+size_t HUF_readDTableX4_wksp (HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize);
 
 size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
 size_t HUF_decompress4X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
@@ -208,16 +270,29 @@ size_t HUF_decompress4X4_usingDTable(void* dst, size_t maxDstSize, const void* c
 /* single stream variants */
 
 size_t HUF_compress1X (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
+size_t HUF_compress1X_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);  /**< `workSpace` must be a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
 size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable);
+/** HUF_compress1X_repeat() :
+*   Same as HUF_compress1X_wksp(), but considers using hufTable if *repeat != HUF_repeat_none.
+*   If it uses hufTable it does not modify hufTable or repeat.
+*   If it doesn't, it sets *repeat = HUF_repeat_none, and it sets hufTable to the table used.
+*   If preferRepeat then the old table will always be used if valid. */
+size_t HUF_compress1X_repeat(void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize, HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat);  /**< `workSpace` must be a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
 
 size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* single-symbol decoder */
 size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* double-symbol decoder */
 
-size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
+size_t HUF_decompress1X_DCtx_wksp (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);
+size_t HUF_decompress1X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
+size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< single-symbol decoder */
+size_t HUF_decompress1X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
+size_t HUF_decompress1X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< double-symbols decoder */
+
+size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);   /**< automatic selection of sing or double symbol decoder, based on DTable */
 size_t HUF_decompress1X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
 size_t HUF_decompress1X4_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
 
-
 #endif /* HUF_STATIC_LINKING_ONLY */
 
 

contrib/libzstd/include/zstd/common/mem.h

@@ -39,7 +39,7 @@ extern "C" {
 #endif
 
 /* code only tested on 32 and 64 bits systems */
-#define MEM_STATIC_ASSERT(c)   { enum { XXH_static_assert = 1/(int)(!!(c)) }; }
+#define MEM_STATIC_ASSERT(c)   { enum { MEM_static_assert = 1/(int)(!!(c)) }; }
 MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (sizeof(size_t)==8)); }
 
 
@@ -55,6 +55,8 @@ MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (size
   typedef   int32_t S32;
   typedef  uint64_t U64;
   typedef   int64_t S64;
+  typedef  intptr_t iPtrDiff;
+  typedef uintptr_t uPtrDiff;
 #else
   typedef unsigned char      BYTE;
   typedef unsigned short      U16;
@@ -63,6 +65,8 @@ MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (size
   typedef   signed int        S32;
   typedef unsigned long long  U64;
   typedef   signed long long  S64;
+  typedef ptrdiff_t      iPtrDiff;
+  typedef size_t         uPtrDiff;
 #endif
 
 
@@ -74,19 +78,18 @@ MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (size
  * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
  * The below switch allow to select different access method for improved performance.
  * Method 0 (default) : use `memcpy()`. Safe and portable.
- * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
+ * Method 1 : `__packed` statement. It depends on compiler extension (i.e., not portable).
  *            This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
  * Method 2 : direct access. This method is portable but violate C standard.
  *            It can generate buggy code on targets depending on alignment.
- *            In some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
+ *            In some circumstances, it's the only known way to get the most performance (i.e. GCC + ARMv6)
  * See http://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
  * Prefer these methods in priority order (0 > 1 > 2)
  */
 #ifndef MEM_FORCE_MEMORY_ACCESS   /* can be defined externally, on command line for example */
 #  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
 #    define MEM_FORCE_MEMORY_ACCESS 2
-#  elif defined(__INTEL_COMPILER) /*|| defined(_MSC_VER)*/ || \
-  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) ))
+#  elif defined(__INTEL_COMPILER) || defined(__GNUC__)
 #    define MEM_FORCE_MEMORY_ACCESS 1
 #  endif
 #endif
@@ -180,7 +183,7 @@ MEM_STATIC U32 MEM_swap32(U32 in)
 {
 #if defined(_MSC_VER)     /* Visual Studio */
     return _byteswap_ulong(in);
-#elif defined (__GNUC__)
+#elif defined (__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 403)
     return __builtin_bswap32(in);
 #else
     return  ((in << 24) & 0xff000000 ) |
@@ -194,7 +197,7 @@ MEM_STATIC U64 MEM_swap64(U64 in)
 {
 #if defined(_MSC_VER)     /* Visual Studio */
     return _byteswap_uint64(in);
-#elif defined (__GNUC__)
+#elif defined (__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 403)
     return __builtin_bswap64(in);
 #else
     return  ((in << 56) & 0xff00000000000000ULL) |
@@ -349,20 +352,6 @@ MEM_STATIC void MEM_writeBEST(void* memPtr, size_t val)
 }
 
 
-/* function safe only for comparisons */
-MEM_STATIC U32 MEM_readMINMATCH(const void* memPtr, U32 length)
-{
-    switch (length)
-    {
-    default :
-    case 4 : return MEM_read32(memPtr);
-    case 3 : if (MEM_isLittleEndian())
-                return MEM_read32(memPtr)<<8;
-             else
-                return MEM_read32(memPtr)>>8;
-    }
-}
-
 #if defined (__cplusplus)
 }
 #endif

contrib/libzstd/include/zstd/common/pool.c

@@ -0,0 +1,206 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+/* ======   Dependencies   ======= */
+#include <stddef.h>  /* size_t */
+#include <stdlib.h>  /* malloc, calloc, free */
+#include "pool.h"
+
+/* ======   Compiler specifics   ====== */
+#if defined(_MSC_VER)
+#  pragma warning(disable : 4204)        /* disable: C4204: non-constant aggregate initializer */
+#endif
+
+
+#ifdef ZSTD_MULTITHREAD
+
+#include "threading.h"   /* pthread adaptation */
+
+/* A job is a function and an opaque argument */
+typedef struct POOL_job_s {
+  POOL_function function;
+  void *opaque;
+} POOL_job;
+
+struct POOL_ctx_s {
+    /* Keep track of the threads */
+    pthread_t *threads;
+    size_t numThreads;
+
+    /* The queue is a circular buffer */
+    POOL_job *queue;
+    size_t queueHead;
+    size_t queueTail;
+    size_t queueSize;
+    /* The mutex protects the queue */
+    pthread_mutex_t queueMutex;
+    /* Condition variable for pushers to wait on when the queue is full */
+    pthread_cond_t queuePushCond;
+    /* Condition variables for poppers to wait on when the queue is empty */
+    pthread_cond_t queuePopCond;
+    /* Indicates if the queue is shutting down */
+    int shutdown;
+};
+
+/* POOL_thread() :
+   Work thread for the thread pool.
+   Waits for jobs and executes them.
+   @returns : NULL on failure else non-null.
+*/
+static void* POOL_thread(void* opaque) {
+    POOL_ctx* const ctx = (POOL_ctx*)opaque;
+    if (!ctx) { return NULL; }
+    for (;;) {
+        /* Lock the mutex and wait for a non-empty queue or until shutdown */
+        pthread_mutex_lock(&ctx->queueMutex);
+        while (ctx->queueHead == ctx->queueTail && !ctx->shutdown) {
+            pthread_cond_wait(&ctx->queuePopCond, &ctx->queueMutex);
+        }
+        /* empty => shutting down: so stop */
+        if (ctx->queueHead == ctx->queueTail) {
+            pthread_mutex_unlock(&ctx->queueMutex);
+            return opaque;
+        }
+        /* Pop a job off the queue */
+        {   POOL_job const job = ctx->queue[ctx->queueHead];
+            ctx->queueHead = (ctx->queueHead + 1) % ctx->queueSize;
+            /* Unlock the mutex, signal a pusher, and run the job */
+            pthread_mutex_unlock(&ctx->queueMutex);
+            pthread_cond_signal(&ctx->queuePushCond);
+            job.function(job.opaque);
+        }
+    }
+    /* Unreachable */
+}
+
+POOL_ctx *POOL_create(size_t numThreads, size_t queueSize) {
+    POOL_ctx *ctx;
+    /* Check the parameters */
+    if (!numThreads || !queueSize) { return NULL; }
+    /* Allocate the context and zero initialize */
+    ctx = (POOL_ctx *)calloc(1, sizeof(POOL_ctx));
+    if (!ctx) { return NULL; }
+    /* Initialize the job queue.
+     * It needs one extra space since one space is wasted to differentiate empty
+     * and full queues.
+     */
+    ctx->queueSize = queueSize + 1;
+    ctx->queue = (POOL_job *)malloc(ctx->queueSize * sizeof(POOL_job));
+    ctx->queueHead = 0;
+    ctx->queueTail = 0;
+    pthread_mutex_init(&ctx->queueMutex, NULL);
+    pthread_cond_init(&ctx->queuePushCond, NULL);
+    pthread_cond_init(&ctx->queuePopCond, NULL);
+    ctx->shutdown = 0;
+    /* Allocate space for the thread handles */
+    ctx->threads = (pthread_t *)malloc(numThreads * sizeof(pthread_t));
+    ctx->numThreads = 0;
+    /* Check for errors */
+    if (!ctx->threads || !ctx->queue) { POOL_free(ctx); return NULL; }
+    /* Initialize the threads */
+    {   size_t i;
+        for (i = 0; i < numThreads; ++i) {
+            if (pthread_create(&ctx->threads[i], NULL, &POOL_thread, ctx)) {
+                ctx->numThreads = i;
+                POOL_free(ctx);
+                return NULL;
+        }   }
+        ctx->numThreads = numThreads;
+    }
+    return ctx;
+}
+
+/*! POOL_join() :
+    Shutdown the queue, wake any sleeping threads, and join all of the threads.
+*/
+static void POOL_join(POOL_ctx *ctx) {
+    /* Shut down the queue */
+    pthread_mutex_lock(&ctx->queueMutex);
+    ctx->shutdown = 1;
+    pthread_mutex_unlock(&ctx->queueMutex);
+    /* Wake up sleeping threads */
+    pthread_cond_broadcast(&ctx->queuePushCond);
+    pthread_cond_broadcast(&ctx->queuePopCond);
+    /* Join all of the threads */
+    {   size_t i;
+        for (i = 0; i < ctx->numThreads; ++i) {
+            pthread_join(ctx->threads[i], NULL);
+    }   }
+}
+
+void POOL_free(POOL_ctx *ctx) {
+    if (!ctx) { return; }
+    POOL_join(ctx);
+    pthread_mutex_destroy(&ctx->queueMutex);
+    pthread_cond_destroy(&ctx->queuePushCond);
+    pthread_cond_destroy(&ctx->queuePopCond);
+    if (ctx->queue) free(ctx->queue);
+    if (ctx->threads) free(ctx->threads);
+    free(ctx);
+}
+
+size_t POOL_sizeof(POOL_ctx *ctx) {
+    if (ctx==NULL) return 0;  /* supports sizeof NULL */
+    return sizeof(*ctx)
+        + ctx->queueSize * sizeof(POOL_job)
+        + ctx->numThreads * sizeof(pthread_t);
+}
+
+void POOL_add(void *ctxVoid, POOL_function function, void *opaque) {
+    POOL_ctx *ctx = (POOL_ctx *)ctxVoid;
+    if (!ctx) { return; }
+
+    pthread_mutex_lock(&ctx->queueMutex);
+    {   POOL_job const job = {function, opaque};
+        /* Wait until there is space in the queue for the new job */
+        size_t newTail = (ctx->queueTail + 1) % ctx->queueSize;
+        while (ctx->queueHead == newTail && !ctx->shutdown) {
+          pthread_cond_wait(&ctx->queuePushCond, &ctx->queueMutex);
+          newTail = (ctx->queueTail + 1) % ctx->queueSize;
+        }
+        /* The queue is still going => there is space */
+        if (!ctx->shutdown) {
+            ctx->queue[ctx->queueTail] = job;
+            ctx->queueTail = newTail;
+        }
+    }
+    pthread_mutex_unlock(&ctx->queueMutex);
+    pthread_cond_signal(&ctx->queuePopCond);
+}
+
+#else  /* ZSTD_MULTITHREAD  not defined */
+/* No multi-threading support */
+
+/* We don't need any data, but if it is empty malloc() might return NULL. */
+struct POOL_ctx_s {
+  int data;
+};
+
+POOL_ctx *POOL_create(size_t numThreads, size_t queueSize) {
+  (void)numThreads;
+  (void)queueSize;
+  return (POOL_ctx *)malloc(sizeof(POOL_ctx));
+}
+
+void POOL_free(POOL_ctx *ctx) {
+  if (ctx) free(ctx);
+}
+
+void POOL_add(void *ctx, POOL_function function, void *opaque) {
+  (void)ctx;
+  function(opaque);
+}
+
+size_t POOL_sizeof(POOL_ctx *ctx) {
+    if (ctx==NULL) return 0;  /* supports sizeof NULL */
+    return sizeof(*ctx);
+}
+
+#endif  /* ZSTD_MULTITHREAD */

contrib/libzstd/include/zstd/common/pool.h

@@ -0,0 +1,61 @@
+/**
+ * Copyright (c) 2016-present, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+#ifndef POOL_H
+#define POOL_H
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+#include <stddef.h>   /* size_t */
+
+typedef struct POOL_ctx_s POOL_ctx;
+
+/*! POOL_create() :
+    Create a thread pool with at most `numThreads` threads.
+    `numThreads` must be at least 1.
+    The maximum number of queued jobs before blocking is `queueSize`.
+    `queueSize` must be at least 1.
+    @return : The POOL_ctx pointer on success else NULL.
+*/
+POOL_ctx *POOL_create(size_t numThreads, size_t queueSize);
+
+/*! POOL_free() :
+    Free a thread pool returned by POOL_create().
+*/
+void POOL_free(POOL_ctx *ctx);
+
+/*! POOL_sizeof() :
+    return memory usage of pool returned by POOL_create().
+*/
+size_t POOL_sizeof(POOL_ctx *ctx);
+
+/*! POOL_function :
+    The function type that can be added to a thread pool.
+*/
+typedef void (*POOL_function)(void *);
+/*! POOL_add_function :
+    The function type for a generic thread pool add function.
+*/
+typedef void (*POOL_add_function)(void *, POOL_function, void *);
+
+/*! POOL_add() :
+    Add the job `function(opaque)` to the thread pool.
+    Possibly blocks until there is room in the queue.
+    Note : The function may be executed asynchronously, so `opaque` must live until the function has been completed.
+*/
+void POOL_add(void *ctx, POOL_function function, void *opaque);
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif

contrib/libzstd/include/zstd/common/threading.c

@@ -0,0 +1,79 @@
+/**
+ * Copyright (c) 2016 Tino Reichardt
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ *
+ * You can contact the author at:
+ * - zstdmt source repository: https://github.com/mcmilk/zstdmt
+ */
+
+/**
+ * This file will hold wrapper for systems, which do not support pthreads
+ */
+
+/* When ZSTD_MULTITHREAD is not defined, this file would become an empty translation unit.
+* Include some ISO C header code to prevent this and portably avoid related warnings.
+* (Visual C++: C4206 / GCC: -Wpedantic / Clang: -Wempty-translation-unit)
+*/
+#include <stddef.h>
+
+
+#if defined(ZSTD_MULTITHREAD) && defined(_WIN32)
+
+/**
+ * Windows minimalist Pthread Wrapper, based on :
+ * http://www.cse.wustl.edu/~schmidt/win32-cv-1.html
+ */
+
+
+/* ===  Dependencies  === */
+#include <process.h>
+#include <errno.h>
+#include "threading.h"
+
+
+/* ===  Implementation  === */
+
+static unsigned __stdcall worker(void *arg)
+{
+    pthread_t* const thread = (pthread_t*) arg;
+    thread->arg = thread->start_routine(thread->arg);
+    return 0;
+}
+
+int pthread_create(pthread_t* thread, const void* unused,
+            void* (*start_routine) (void*), void* arg)
+{
+    (void)unused;
+    thread->arg = arg;
+    thread->start_routine = start_routine;
+    thread->handle = (HANDLE) _beginthreadex(NULL, 0, worker, thread, 0, NULL);
+
+    if (!thread->handle)
+        return errno;
+    else
+        return 0;
+}
+
+int _pthread_join(pthread_t * thread, void **value_ptr)
+{
+    DWORD result;
+
+    if (!thread->handle) return 0;
+
+    result = WaitForSingleObject(thread->handle, INFINITE);
+    switch (result) {
+    case WAIT_OBJECT_0:
+        if (value_ptr) *value_ptr = thread->arg;
+        return 0;
+    case WAIT_ABANDONED:
+        return EINVAL;
+    default:
+        return GetLastError();
+    }
+}
+
+#endif   /* ZSTD_MULTITHREAD */

contrib/libzstd/include/zstd/common/threading.h

@@ -0,0 +1,104 @@
+
+/**
+ * Copyright (c) 2016 Tino Reichardt
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ *
+ * You can contact the author at:
+ * - zstdmt source repository: https://github.com/mcmilk/zstdmt
+ */
+
+#ifndef THREADING_H_938743
+#define THREADING_H_938743
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#if defined(ZSTD_MULTITHREAD) && defined(_WIN32)
+
+/**
+ * Windows minimalist Pthread Wrapper, based on :
+ * http://www.cse.wustl.edu/~schmidt/win32-cv-1.html
+ */
+#ifdef WINVER
+#  undef WINVER
+#endif
+#define WINVER       0x0600
+
+#ifdef _WIN32_WINNT
+#  undef _WIN32_WINNT
+#endif
+#define _WIN32_WINNT 0x0600
+
+#ifndef WIN32_LEAN_AND_MEAN
+#  define WIN32_LEAN_AND_MEAN
+#endif
+
+#include <windows.h>
+
+/* mutex */
+#define pthread_mutex_t           CRITICAL_SECTION
+#define pthread_mutex_init(a,b)   InitializeCriticalSection((a))
+#define pthread_mutex_destroy(a)  DeleteCriticalSection((a))
+#define pthread_mutex_lock(a)     EnterCriticalSection((a))
+#define pthread_mutex_unlock(a)   LeaveCriticalSection((a))
+
+/* condition variable */
+#define pthread_cond_t             CONDITION_VARIABLE
+#define pthread_cond_init(a, b)    InitializeConditionVariable((a))
+#define pthread_cond_destroy(a)    /* No delete */
+#define pthread_cond_wait(a, b)    SleepConditionVariableCS((a), (b), INFINITE)
+#define pthread_cond_signal(a)     WakeConditionVariable((a))
+#define pthread_cond_broadcast(a)  WakeAllConditionVariable((a))
+
+/* pthread_create() and pthread_join() */
+typedef struct {
+    HANDLE handle;
+    void* (*start_routine)(void*);
+    void* arg;
+} pthread_t;
+
+int pthread_create(pthread_t* thread, const void* unused,
+                   void* (*start_routine) (void*), void* arg);
+
+#define pthread_join(a, b) _pthread_join(&(a), (b))
+int _pthread_join(pthread_t* thread, void** value_ptr);
+
+/**
+ * add here more wrappers as required
+ */
+
+
+#elif defined(ZSTD_MULTITHREAD)   /* posix assumed ; need a better detection method */
+/* ===   POSIX Systems   === */
+#  include <pthread.h>
+
+#else  /* ZSTD_MULTITHREAD not defined */
+/* No multithreading support */
+
+#define pthread_mutex_t int   /* #define rather than typedef, as sometimes pthread support is implicit, resulting in duplicated symbols */
+#define pthread_mutex_init(a,b)
+#define pthread_mutex_destroy(a)
+#define pthread_mutex_lock(a)
+#define pthread_mutex_unlock(a)
+
+#define pthread_cond_t int
+#define pthread_cond_init(a,b)
+#define pthread_cond_destroy(a)
+#define pthread_cond_wait(a,b)
+#define pthread_cond_signal(a)
+#define pthread_cond_broadcast(a)
+
+/* do not use pthread_t */
+
+#endif /* ZSTD_MULTITHREAD */
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* THREADING_H_938743 */

contrib/libzstd/include/zstd/common/xxhash.c

@@ -104,7 +104,9 @@ static void  XXH_free  (void* p)  { free(p); }
 #include <string.h>
 static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); }
 
-#define XXH_STATIC_LINKING_ONLY
+#ifndef XXH_STATIC_LINKING_ONLY
+#  define XXH_STATIC_LINKING_ONLY
+#endif
 #include "xxhash.h"
 
 

contrib/libzstd/include/zstd/common/xxhash.h

@@ -64,16 +64,12 @@ XXH64       13.8 GB/s            1.9 GB/s
 XXH32        6.8 GB/s            6.0 GB/s
 */
 
-#ifndef XXHASH_H_5627135585666179
-#define XXHASH_H_5627135585666179 1
-
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
-#ifndef XXH_NAMESPACE
-#  define XXH_NAMESPACE ZSTD_  /* Zstandard specific */
-#endif
+#ifndef XXHASH_H_5627135585666179
+#define XXHASH_H_5627135585666179 1
 
 
 /* ****************************
@@ -242,6 +238,11 @@ XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* restrict dst_state, const XXH
 /* **************************
 *  Canonical representation
 ****************************/
+/* Default result type for XXH functions are primitive unsigned 32 and 64 bits.
+*  The canonical representation uses human-readable write convention, aka big-endian (large digits first).
+*  These functions allow transformation of hash result into and from its canonical format.
+*  This way, hash values can be written into a file / memory, and remain comparable on different systems and programs.
+*/
 typedef struct { unsigned char digest[4]; } XXH32_canonical_t;
 typedef struct { unsigned char digest[8]; } XXH64_canonical_t;
 
@@ -251,14 +252,9 @@ XXH_PUBLIC_API void XXH64_canonicalFromHash(XXH64_canonical_t* dst, XXH64_hash_t
 XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src);
 XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src);
 
-/* Default result type for XXH functions are primitive unsigned 32 and 64 bits.
-*  The canonical representation uses human-readable write convention, aka big-endian (large digits first).
-*  These functions allow transformation of hash result into and from its canonical format.
-*  This way, hash values can be written into a file / memory, and remain comparable on different systems and programs.
-*/
+#endif /* XXHASH_H_5627135585666179 */
 
 
-#ifdef XXH_STATIC_LINKING_ONLY
 
 /* ================================================================================================
    This section contains definitions which are not guaranteed to remain stable.
@@ -266,6 +262,8 @@ XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src
    They shall only be used with static linking.
    Never use these definitions in association with dynamic linking !
 =================================================================================================== */
+#if defined(XXH_STATIC_LINKING_ONLY) && !defined(XXH_STATIC_H_3543687687345)
+#define XXH_STATIC_H_3543687687345
 
 /* These definitions are only meant to allow allocation of XXH state
    statically, on stack, or in a struct for example.
@@ -299,11 +297,9 @@ XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src
 #    include "xxhash.c"   /* include xxhash functions as `static`, for inlining */
 #  endif
 
-#endif /* XXH_STATIC_LINKING_ONLY */
+#endif /* XXH_STATIC_LINKING_ONLY && XXH_STATIC_H_3543687687345 */
 
 
 #if defined (__cplusplus)
 }
 #endif
-
-#endif /* XXHASH_H_5627135585666179 */

contrib/libzstd/include/zstd/common/zstd_common.c

@@ -12,17 +12,19 @@
 /*-*************************************
 *  Dependencies
 ***************************************/
-#include <stdlib.h>         /* malloc */
+#include <stdlib.h>      /* malloc, calloc, free */
+#include <string.h>      /* memset */
 #include "error_private.h"
 #define ZSTD_STATIC_LINKING_ONLY
-#include "zstd.h"           /* declaration of ZSTD_isError, ZSTD_getErrorName, ZSTD_getErrorCode, ZSTD_getErrorString, ZSTD_versionNumber */
-#include "zbuff.h"          /* declaration of ZBUFF_isError, ZBUFF_getErrorName */
+#include "zstd.h"
 
 
 /*-****************************************
 *  Version
 ******************************************/
-unsigned ZSTD_versionNumber (void) { return ZSTD_VERSION_NUMBER; }
+unsigned ZSTD_versionNumber(void) { return ZSTD_VERSION_NUMBER; }
+
+const char* ZSTD_versionString(void) { return ZSTD_VERSION_STRING; }
 
 
 /*-****************************************
@@ -42,42 +44,37 @@ ZSTD_ErrorCode ZSTD_getErrorCode(size_t code) { return ERR_getErrorCode(code); }
 
 /*! ZSTD_getErrorString() :
 *   provides error code string from enum */
-const char* ZSTD_getErrorString(ZSTD_ErrorCode code) { return ERR_getErrorName(code); }
-
-
-/* **************************************************************
-*  ZBUFF Error Management
-****************************************************************/
-unsigned ZBUFF_isError(size_t errorCode) { return ERR_isError(errorCode); }
-
-const char* ZBUFF_getErrorName(size_t errorCode) { return ERR_getErrorName(errorCode); }
-
+const char* ZSTD_getErrorString(ZSTD_ErrorCode code) { return ERR_getErrorString(code); }
 
 
 /*=**************************************************************
 *  Custom allocator
 ****************************************************************/
-/* default uses stdlib */
-void* ZSTD_defaultAllocFunction(void* opaque, size_t size)
-{
-    void* address = malloc(size);
-    (void)opaque;
-    return address;
-}
-
-void ZSTD_defaultFreeFunction(void* opaque, void* address)
-{
-    (void)opaque;
-    free(address);
-}
-
 void* ZSTD_malloc(size_t size, ZSTD_customMem customMem)
 {
+    if (customMem.customAlloc)
         return customMem.customAlloc(customMem.opaque, size);
+    return malloc(size);
+}
+
+void* ZSTD_calloc(size_t size, ZSTD_customMem customMem)
+{
+    if (customMem.customAlloc) {
+        /* calloc implemented as malloc+memset;
+         * not as efficient as calloc, but next best guess for custom malloc */
+        void* const ptr = customMem.customAlloc(customMem.opaque, size);
+        memset(ptr, 0, size);
+        return ptr;
+    }
+    return calloc(1, size);
 }
 
 void ZSTD_free(void* ptr, ZSTD_customMem customMem)
 {
-    if (ptr!=NULL)
+    if (ptr!=NULL) {
+        if (customMem.customFree)
             customMem.customFree(customMem.opaque, ptr);
+        else
+            free(ptr);
+    }
 }

contrib/libzstd/include/zstd/common/zstd_errors.h

@@ -0,0 +1,83 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_ERRORS_H_398273423
+#define ZSTD_ERRORS_H_398273423
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*===== dependency =====*/
+#include <stddef.h>   /* size_t */
+
+
+/* =====   ZSTDERRORLIB_API : control library symbols visibility   ===== */
+#ifndef ZSTDERRORLIB_VISIBILITY
+#  if defined(__GNUC__) && (__GNUC__ >= 4)
+#    define ZSTDERRORLIB_VISIBILITY __attribute__ ((visibility ("default")))
+#  else
+#    define ZSTDERRORLIB_VISIBILITY
+#  endif
+#endif
+#if defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
+#  define ZSTDERRORLIB_API __declspec(dllexport) ZSTDERRORLIB_VISIBILITY
+#elif defined(ZSTD_DLL_IMPORT) && (ZSTD_DLL_IMPORT==1)
+#  define ZSTDERRORLIB_API __declspec(dllimport) ZSTDERRORLIB_VISIBILITY /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/
+#else
+#  define ZSTDERRORLIB_API ZSTDERRORLIB_VISIBILITY
+#endif
+
+/*-****************************************
+ *  error codes list
+ *  note : this API is still considered unstable
+ *         it should not be used with a dynamic library
+ *         only static linking is allowed
+ ******************************************/
+typedef enum {
+  ZSTD_error_no_error,
+  ZSTD_error_GENERIC,
+  ZSTD_error_prefix_unknown,
+  ZSTD_error_version_unsupported,
+  ZSTD_error_parameter_unknown,
+  ZSTD_error_frameParameter_unsupported,
+  ZSTD_error_frameParameter_unsupportedBy32bits,
+  ZSTD_error_frameParameter_windowTooLarge,
+  ZSTD_error_compressionParameter_unsupported,
+  ZSTD_error_compressionParameter_outOfBound,
+  ZSTD_error_init_missing,
+  ZSTD_error_memory_allocation,
+  ZSTD_error_stage_wrong,
+  ZSTD_error_dstSize_tooSmall,
+  ZSTD_error_srcSize_wrong,
+  ZSTD_error_corruption_detected,
+  ZSTD_error_checksum_wrong,
+  ZSTD_error_tableLog_tooLarge,
+  ZSTD_error_maxSymbolValue_tooLarge,
+  ZSTD_error_maxSymbolValue_tooSmall,
+  ZSTD_error_dictionary_corrupted,
+  ZSTD_error_dictionary_wrong,
+  ZSTD_error_dictionaryCreation_failed,
+  ZSTD_error_frameIndex_tooLarge,
+  ZSTD_error_seekableIO,
+  ZSTD_error_maxCode
+} ZSTD_ErrorCode;
+
+/*! ZSTD_getErrorCode() :
+    convert a `size_t` function result into a `ZSTD_ErrorCode` enum type,
+    which can be used to compare with enum list published above */
+ZSTDERRORLIB_API ZSTD_ErrorCode ZSTD_getErrorCode(size_t functionResult);
+ZSTDERRORLIB_API const char* ZSTD_getErrorString(ZSTD_ErrorCode code);
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_ERRORS_H_398273423 */

contrib/libzstd/include/zstd/common/zstd_internal.h

@@ -16,9 +16,10 @@
 #ifdef _MSC_VER    /* Visual Studio */
 #  define FORCE_INLINE static __forceinline
 #  include <intrin.h>                    /* For Visual 2005 */
-#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
-#  pragma warning(disable : 4324)        /* disable: C4324: padded structure */
 #  pragma warning(disable : 4100)        /* disable: C4100: unreferenced formal parameter */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#  pragma warning(disable : 4204)        /* disable: C4204: non-constant aggregate initializer */
+#  pragma warning(disable : 4324)        /* disable: C4324: padded structure */
 #else
 #  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
 #    ifdef __GNUC__
@@ -31,6 +32,16 @@
 #  endif /* __STDC_VERSION__ */
 #endif
 
+#ifdef _MSC_VER
+#  define FORCE_NOINLINE static __declspec(noinline)
+#else
+#  ifdef __GNUC__
+#    define FORCE_NOINLINE static __attribute__((__noinline__))
+#  else
+#    define FORCE_NOINLINE static
+#  endif
+#endif
+
 
 /*-*************************************
 *  Dependencies
@@ -39,11 +50,50 @@
 #include "error_private.h"
 #define ZSTD_STATIC_LINKING_ONLY
 #include "zstd.h"
+#ifndef XXH_STATIC_LINKING_ONLY
+#  define XXH_STATIC_LINKING_ONLY  /* XXH64_state_t */
+#endif
+#include "xxhash.h"                /* XXH_reset, update, digest */
+
+
+/*-*************************************
+*  Debug
+***************************************/
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=1)
+#  include <assert.h>
+#else
+#  ifndef assert
+#    define assert(condition) ((void)0)
+#  endif
+#endif
+
+#define ZSTD_STATIC_ASSERT(c) { enum { ZSTD_static_assert = 1/(int)(!!(c)) }; }
+
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=2)
+#  include <stdio.h>
+/* recommended values for ZSTD_DEBUG display levels :
+ * 1 : no display, enables assert() only
+ * 2 : reserved for currently active debugging path
+ * 3 : events once per object lifetime (CCtx, CDict)
+ * 4 : events once per frame
+ * 5 : events once per block
+ * 6 : events once per sequence (*very* verbose) */
+#  define DEBUGLOG(l, ...) {                         \
+                if (l<=ZSTD_DEBUG) {                 \
+                    fprintf(stderr, __FILE__ ": ");  \
+                    fprintf(stderr, __VA_ARGS__);    \
+                    fprintf(stderr, " \n");          \
+            }   }
+#else
+#  define DEBUGLOG(l, ...)      {}    /* disabled */
+#endif
 
 
 /*-*************************************
 *  shared macros
 ***************************************/
+#undef MIN
+#undef MAX
 #define MIN(a,b) ((a)<(b) ? (a) : (b))
 #define MAX(a,b) ((a)>(b) ? (a) : (b))
 #define CHECK_F(f) { size_t const errcod = f; if (ERR_isError(errcod)) return errcod; }  /* check and Forward error code */
@@ -54,7 +104,6 @@
 *  Common constants
 ***************************************/
 #define ZSTD_OPT_NUM    (1<<12)
-#define ZSTD_DICT_MAGIC  0xEC30A437   /* v0.7+ */
 
 #define ZSTD_REP_NUM      3                 /* number of repcodes */
 #define ZSTD_REP_CHECK    (ZSTD_REP_NUM)    /* number of repcodes to check by the optimal parser */
@@ -90,7 +139,6 @@ typedef enum { set_basic, set_rle, set_compressed, set_repeat } symbolEncodingTy
 #define LONGNBSEQ 0x7F00
 
 #define MINMATCH 3
-#define EQUAL_READ32 4
 
 #define Litbits  8
 #define MaxLit ((1<<Litbits) - 1)
@@ -137,7 +185,7 @@ static void ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
 /*! ZSTD_wildcopy() :
 *   custom version of memcpy(), can copy up to 7 bytes too many (8 bytes if length==0) */
 #define WILDCOPY_OVERLENGTH 8
-MEM_STATIC void ZSTD_wildcopy(void* dst, const void* src, size_t length)
+MEM_STATIC void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length)
 {
     const BYTE* ip = (const BYTE*)src;
     BYTE* op = (BYTE*)dst;
@@ -212,6 +260,7 @@ typedef struct {
     U32  log2litSum;
     U32  log2offCodeSum;
     U32  factor;
+    U32  staticPrices;
     U32  cachedPrice;
     U32  cachedLitLength;
     const BYTE* cachedLiterals;
@@ -219,13 +268,10 @@ typedef struct {
 
 const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx);
 void ZSTD_seqToCodes(const seqStore_t* seqStorePtr);
-int ZSTD_isSkipFrame(ZSTD_DCtx* dctx);
 
 /* custom memory allocation functions */
-void* ZSTD_defaultAllocFunction(void* opaque, size_t size);
-void ZSTD_defaultFreeFunction(void* opaque, void* address);
-static const ZSTD_customMem defaultCustomMem = { ZSTD_defaultAllocFunction, ZSTD_defaultFreeFunction, NULL };
 void* ZSTD_malloc(size_t size, ZSTD_customMem customMem);
+void* ZSTD_calloc(size_t size, ZSTD_customMem customMem);
 void ZSTD_free(void* ptr, ZSTD_customMem customMem);
 
 
@@ -254,4 +300,35 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val)
 }
 
 
+/* hidden functions */
+
+/* ZSTD_invalidateRepCodes() :
+ * ensures next compression will not use repcodes from previous block.
+ * Note : only works with regular variant;
+ *        do not use with extDict variant ! */
+void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx);
+
+
+/*! ZSTD_initCStream_internal() :
+ *  Private use only. Init streaming operation.
+ *  expects params to be valid.
+ *  must receive dict, or cdict, or none, but not both.
+ *  @return : 0, or an error code */
+size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
+                     const void* dict, size_t dictSize,
+                     const ZSTD_CDict* cdict,
+                     ZSTD_parameters params, unsigned long long pledgedSrcSize);
+
+/*! ZSTD_compressStream_generic() :
+ *  Private use only. To be called from zstdmt_compress.c in single-thread mode. */
+size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
+                                   ZSTD_outBuffer* output,
+                                   ZSTD_inBuffer* input,
+                                   ZSTD_EndDirective const flushMode);
+
+/*! ZSTD_getParamsFromCDict() :
+ *  as the name implies */
+ZSTD_parameters ZSTD_getParamsFromCDict(const ZSTD_CDict* cdict);
+
+
 #endif   /* ZSTD_CCOMMON_H_MODULE */

contrib/libzstd/include/zstd/compress/fse_compress.c

@@ -70,12 +70,6 @@
 #define FSE_STATIC_ASSERT(c) { enum { FSE_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
 
 
-/* **************************************************************
-*  Complex types
-****************************************************************/
-typedef U32 CTable_max_t[FSE_CTABLE_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)];
-
-
 /* **************************************************************
 *  Templates
 ****************************************************************/
@@ -100,7 +94,13 @@ typedef U32 CTable_max_t[FSE_CTABLE_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VA
 
 
 /* Function templates */
-size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+
+/* FSE_buildCTable_wksp() :
+ * Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`).
+ * wkspSize should be sized to handle worst case situation, which is `1<<max_tableLog * sizeof(FSE_FUNCTION_TYPE)`
+ * workSpace must also be properly aligned with FSE_FUNCTION_TYPE requirements
+ */
+size_t FSE_buildCTable_wksp(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize)
 {
     U32 const tableSize = 1 << tableLog;
     U32 const tableMask = tableSize - 1;
@@ -111,10 +111,11 @@ size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned
     U32 const step = FSE_TABLESTEP(tableSize);
     U32 cumul[FSE_MAX_SYMBOL_VALUE+2];
 
-    FSE_FUNCTION_TYPE tableSymbol[FSE_MAX_TABLESIZE]; /* memset() is not necessary, even if static analyzer complain about it */
+    FSE_FUNCTION_TYPE* const tableSymbol = (FSE_FUNCTION_TYPE*)workSpace;
     U32 highThreshold = tableSize-1;
 
     /* CTable header */
+    if (((size_t)1 << tableLog) * sizeof(FSE_FUNCTION_TYPE) > wkspSize) return ERROR(tableLog_tooLarge);
     tableU16[-2] = (U16) tableLog;
     tableU16[-1] = (U16) maxSymbolValue;
 
@@ -181,6 +182,13 @@ size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned
 }
 
 
+size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+{
+    FSE_FUNCTION_TYPE tableSymbol[FSE_MAX_TABLESIZE];   /* memset() is not necessary, even if static analyzer complain about it */
+    return FSE_buildCTable_wksp(ct, normalizedCounter, maxSymbolValue, tableLog, tableSymbol, sizeof(tableSymbol));
+}
+
+
 
 #ifndef FSE_COMMONDEFS_ONLY
 
@@ -189,12 +197,10 @@ size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned
 ****************************************************************/
 size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog)
 {
-    size_t maxHeaderSize = (((maxSymbolValue+1) * tableLog) >> 3) + 3;
+    size_t const maxHeaderSize = (((maxSymbolValue+1) * tableLog) >> 3) + 3;
     return maxSymbolValue ? maxHeaderSize : FSE_NCOUNTBOUND;  /* maxSymbolValue==0 ? use default */
 }
 
-static short FSE_abs(short a) { return (short)(a<0 ? -a : a); }
-
 static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
                                        const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
                                        unsigned writeIsSafe)
@@ -250,16 +256,16 @@ static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
                 bitStream >>= 16;
                 bitCount -= 16;
         }   }
-        {   short count = normalizedCounter[charnum++];
-            const short max = (short)((2*threshold-1)-remaining);
-            remaining -= FSE_abs(count);
-            if (remaining<1) return ERROR(GENERIC);
+        {   int count = normalizedCounter[charnum++];
+            int const max = (2*threshold-1)-remaining;
+            remaining -= count < 0 ? -count : count;
             count++;   /* +1 for extra accuracy */
             if (count>=threshold) count += max;   /* [0..max[ [max..threshold[ (...) [threshold+max 2*threshold[ */
             bitStream += count << bitCount;
             bitCount  += nbBits;
             bitCount  -= (count<max);
             previous0  = (count==1);
+            if (remaining<1) return ERROR(GENERIC);
             while (remaining<threshold) nbBits--, threshold>>=1;
         }
         if (bitCount>16) {
@@ -285,7 +291,7 @@ static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
 
 size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
 {
-    if (tableLog > FSE_MAX_TABLELOG) return ERROR(GENERIC);   /* Unsupported */
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);   /* Unsupported */
     if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC);   /* Unsupported */
 
     if (bufferSize < FSE_NCountWriteBound(maxSymbolValue, tableLog))
@@ -300,13 +306,13 @@ size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalized
 *  Counting histogram
 ****************************************************************/
 /*! FSE_count_simple
-    This function just counts byte values within `src`,
-    and store the histogram into table `count`.
-    This function is unsafe : it doesn't check that all values within `src` can fit into `count`.
+    This function counts byte values within `src`, and store the histogram into table `count`.
+    It doesn't use any additional memory.
+    But this function is unsafe : it doesn't check that all values within `src` can fit into `count`.
     For this reason, prefer using a table `count` with 256 elements.
     @return : count of most numerous element
 */
-static size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
+size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
                         const void* src, size_t srcSize)
 {
     const BYTE* ip = (const BYTE*)src;
@@ -314,7 +320,6 @@ static size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
     unsigned maxSymbolValue = *maxSymbolValuePtr;
     unsigned max=0;
 
-
     memset(count, 0, (maxSymbolValue+1)*sizeof(*count));
     if (srcSize==0) { *maxSymbolValuePtr = 0; return 0; }
 
@@ -329,20 +334,24 @@ static size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
 }
 
 
-static size_t FSE_count_parallel(unsigned* count, unsigned* maxSymbolValuePtr,
+/* FSE_count_parallel_wksp() :
+ * Same as FSE_count_parallel(), but using an externally provided scratch buffer.
+ * `workSpace` size must be a minimum of `1024 * sizeof(unsigned)`` */
+static size_t FSE_count_parallel_wksp(
+                                unsigned* count, unsigned* maxSymbolValuePtr,
                                 const void* source, size_t sourceSize,
-                                unsigned checkMax)
+                                unsigned checkMax, unsigned* const workSpace)
 {
     const BYTE* ip = (const BYTE*)source;
     const BYTE* const iend = ip+sourceSize;
     unsigned maxSymbolValue = *maxSymbolValuePtr;
     unsigned max=0;
+    U32* const Counting1 = workSpace;
+    U32* const Counting2 = Counting1 + 256;
+    U32* const Counting3 = Counting2 + 256;
+    U32* const Counting4 = Counting3 + 256;
 
-
-    U32 Counting1[256] = { 0 };
-    U32 Counting2[256] = { 0 };
-    U32 Counting3[256] = { 0 };
-    U32 Counting4[256] = { 0 };
+    memset(Counting1, 0, 4*256*sizeof(unsigned));
 
     /* safety checks */
     if (!sourceSize) {
@@ -391,28 +400,48 @@ static size_t FSE_count_parallel(unsigned* count, unsigned* maxSymbolValuePtr,
     {   U32 s; for (s=0; s<=maxSymbolValue; s++) {
             count[s] = Counting1[s] + Counting2[s] + Counting3[s] + Counting4[s];
             if (count[s] > max) max = count[s];
-    }}
+    }   }
 
     while (!count[maxSymbolValue]) maxSymbolValue--;
     *maxSymbolValuePtr = maxSymbolValue;
     return (size_t)max;
 }
 
+/* FSE_countFast_wksp() :
+ * Same as FSE_countFast(), but using an externally provided scratch buffer.
+ * `workSpace` size must be table of >= `1024` unsigned */
+size_t FSE_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+                     const void* source, size_t sourceSize, unsigned* workSpace)
+{
+    if (sourceSize < 1500) return FSE_count_simple(count, maxSymbolValuePtr, source, sourceSize);
+    return FSE_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, 0, workSpace);
+}
+
 /* fast variant (unsafe : won't check if src contains values beyond count[] limit) */
 size_t FSE_countFast(unsigned* count, unsigned* maxSymbolValuePtr,
                      const void* source, size_t sourceSize)
 {
-    if (sourceSize < 1500) return FSE_count_simple(count, maxSymbolValuePtr, source, sourceSize);
-    return FSE_count_parallel(count, maxSymbolValuePtr, source, sourceSize, 0);
+    unsigned tmpCounters[1024];
+    return FSE_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, tmpCounters);
+}
+
+/* FSE_count_wksp() :
+ * Same as FSE_count(), but using an externally provided scratch buffer.
+ * `workSpace` size must be table of >= `1024` unsigned */
+size_t FSE_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
+                 const void* source, size_t sourceSize, unsigned* workSpace)
+{
+    if (*maxSymbolValuePtr < 255)
+        return FSE_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, 1, workSpace);
+    *maxSymbolValuePtr = 255;
+    return FSE_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, workSpace);
 }
 
 size_t FSE_count(unsigned* count, unsigned* maxSymbolValuePtr,
-                 const void* source, size_t sourceSize)
+                 const void* src, size_t srcSize)
 {
-    if (*maxSymbolValuePtr <255)
-        return FSE_count_parallel(count, maxSymbolValuePtr, source, sourceSize, 1);
-    *maxSymbolValuePtr = 255;
-    return FSE_countFast(count, maxSymbolValuePtr, source, sourceSize);
+    unsigned tmpCounters[1024];
+    return FSE_count_wksp(count, maxSymbolValuePtr, src, srcSize, tmpCounters);
 }
 
 
@@ -428,14 +457,10 @@ size_t FSE_count(unsigned* count, unsigned* maxSymbolValuePtr,
     `FSE_symbolCompressionTransform symbolTT[maxSymbolValue+1];`  // This size is variable
 Allocation is manual (C standard does not support variable-size structures).
 */
-
 size_t FSE_sizeof_CTable (unsigned maxSymbolValue, unsigned tableLog)
 {
-    size_t size;
-    FSE_STATIC_ASSERT((size_t)FSE_CTABLE_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)*4 >= sizeof(CTable_max_t));   /* A compilation error here means FSE_CTABLE_SIZE_U32 is not large enough */
-    if (tableLog > FSE_MAX_TABLELOG) return ERROR(GENERIC);
-    size = FSE_CTABLE_SIZE_U32 (tableLog, maxSymbolValue) * sizeof(U32);
-    return size;
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
+    return FSE_CTABLE_SIZE_U32 (tableLog, maxSymbolValue) * sizeof(U32);
 }
 
 FSE_CTable* FSE_createCTable (unsigned maxSymbolValue, unsigned tableLog)
@@ -481,12 +506,13 @@ unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxS
 
 static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count, size_t total, U32 maxSymbolValue)
 {
+    short const NOT_YET_ASSIGNED = -2;
     U32 s;
     U32 distributed = 0;
     U32 ToDistribute;
 
     /* Init */
-    U32 lowThreshold = (U32)(total >> tableLog);
+    U32 const lowThreshold = (U32)(total >> tableLog);
     U32 lowOne = (U32)((total * 3) >> (tableLog + 1));
 
     for (s=0; s<=maxSymbolValue; s++) {
@@ -506,7 +532,8 @@ static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count,
             total -= count[s];
             continue;
         }
-        norm[s]=-2;
+
+        norm[s]=NOT_YET_ASSIGNED;
     }
     ToDistribute = (1 << tableLog) - distributed;
 
@@ -514,7 +541,7 @@ static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count,
         /* risk of rounding to zero */
         lowOne = (U32)((total * 3) / (ToDistribute * 2));
         for (s=0; s<=maxSymbolValue; s++) {
-            if ((norm[s] == -2) && (count[s] <= lowOne)) {
+            if ((norm[s] == NOT_YET_ASSIGNED) && (count[s] <= lowOne)) {
                 norm[s] = 1;
                 distributed++;
                 total -= count[s];
@@ -534,17 +561,23 @@ static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count,
         return 0;
     }
 
-    {
-        U64 const vStepLog = 62 - tableLog;
+    if (total == 0) {
+        /* all of the symbols were low enough for the lowOne or lowThreshold */
+        for (s=0; ToDistribute > 0; s = (s+1)%(maxSymbolValue+1))
+            if (norm[s] > 0) ToDistribute--, norm[s]++;
+        return 0;
+    }
+
+    {   U64 const vStepLog = 62 - tableLog;
         U64 const mid = (1ULL << (vStepLog-1)) - 1;
         U64 const rStep = ((((U64)1<<vStepLog) * ToDistribute) + mid) / total;   /* scale on remaining */
         U64 tmpTotal = mid;
         for (s=0; s<=maxSymbolValue; s++) {
-            if (norm[s]==-2) {
-                U64 end = tmpTotal + (count[s] * rStep);
-                U32 sStart = (U32)(tmpTotal >> vStepLog);
-                U32 sEnd = (U32)(end >> vStepLog);
-                U32 weight = sEnd - sStart;
+            if (norm[s]==NOT_YET_ASSIGNED) {
+                U64 const end = tmpTotal + (count[s] * rStep);
+                U32 const sStart = (U32)(tmpTotal >> vStepLog);
+                U32 const sEnd = (U32)(end >> vStepLog);
+                U32 const weight = sEnd - sStart;
                 if (weight < 1)
                     return ERROR(GENERIC);
                 norm[s] = (short)weight;
@@ -566,7 +599,6 @@ size_t FSE_normalizeCount (short* normalizedCounter, unsigned tableLog,
     if (tableLog < FSE_minTableLog(total, maxSymbolValue)) return ERROR(GENERIC);   /* Too small tableLog, compression potentially impossible */
 
     {   U32 const rtbTable[] = {     0, 473195, 504333, 520860, 550000, 700000, 750000, 830000 };
-
         U64 const scale = 62 - tableLog;
         U64 const step = ((U64)1<<62) / total;   /* <== here, one division ! */
         U64 const vStep = 1ULL<<(scale-20);
@@ -594,7 +626,7 @@ size_t FSE_normalizeCount (short* normalizedCounter, unsigned tableLog,
         }   }
         if (-stillToDistribute >= (normalizedCounter[largest] >> 1)) {
             /* corner case, need another normalization method */
-            size_t errorCode = FSE_normalizeM2(normalizedCounter, tableLog, count, total, maxSymbolValue);
+            size_t const errorCode = FSE_normalizeM2(normalizedCounter, tableLog, count, total, maxSymbolValue);
             if (FSE_isError(errorCode)) return errorCode;
         }
         else normalizedCounter[largest] += (short)stillToDistribute;
@@ -643,17 +675,15 @@ size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits)
 
     /* Build Symbol Transformation Table */
     {   const U32 deltaNbBits = (nbBits << 16) - (1 << nbBits);
-
         for (s=0; s<=maxSymbolValue; s++) {
             symbolTT[s].deltaNbBits = deltaNbBits;
             symbolTT[s].deltaFindState = s-1;
     }   }
 
-
     return 0;
 }
 
-/* fake FSE_CTable, for rle (100% always same symbol) input */
+/* fake FSE_CTable, for rle input (always same symbol) */
 size_t FSE_buildCTable_rle (FSE_CTable* ct, BYTE symbolValue)
 {
     void* ptr = ct;
@@ -685,14 +715,13 @@ static size_t FSE_compress_usingCTable_generic (void* dst, size_t dstSize,
     const BYTE* const iend = istart + srcSize;
     const BYTE* ip=iend;
 
-
     BIT_CStream_t bitC;
     FSE_CState_t CState1, CState2;
 
     /* init */
     if (srcSize <= 2) return 0;
-    { size_t const errorCode = BIT_initCStream(&bitC, dst, dstSize);
-      if (FSE_isError(errorCode)) return 0; }
+    { size_t const initError = BIT_initCStream(&bitC, dst, dstSize);
+      if (FSE_isError(initError)) return 0; /* not enough space available to write a bitstream */ }
 
 #define FSE_FLUSHBITS(s)  (fast ? BIT_flushBitsFast(s) : BIT_flushBits(s))
 
@@ -715,7 +744,7 @@ static size_t FSE_compress_usingCTable_generic (void* dst, size_t dstSize,
     }
 
     /* 2 or 4 encoding per loop */
-    for ( ; ip>istart ; ) {
+    while ( ip>istart ) {
 
         FSE_encodeSymbol(&bitC, &CState2, *--ip);
 
@@ -741,7 +770,7 @@ size_t FSE_compress_usingCTable (void* dst, size_t dstSize,
                            const void* src, size_t srcSize,
                            const FSE_CTable* ct)
 {
-    const unsigned fast = (dstSize >= FSE_BLOCKBOUND(srcSize));
+    unsigned const fast = (dstSize >= FSE_BLOCKBOUND(srcSize));
 
     if (fast)
         return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 1);
@@ -752,58 +781,76 @@ size_t FSE_compress_usingCTable (void* dst, size_t dstSize,
 
 size_t FSE_compressBound(size_t size) { return FSE_COMPRESSBOUND(size); }
 
-size_t FSE_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog)
-{
-    const BYTE* const istart = (const BYTE*) src;
-    const BYTE* ip = istart;
+#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return f
+#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)`.
+ */
+size_t FSE_compress_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize)
+{
     BYTE* const ostart = (BYTE*) dst;
     BYTE* op = ostart;
     BYTE* const oend = ostart + dstSize;
 
     U32   count[FSE_MAX_SYMBOL_VALUE+1];
     S16   norm[FSE_MAX_SYMBOL_VALUE+1];
-    CTable_max_t ct;
-    size_t errorCode;
+    FSE_CTable* CTable = (FSE_CTable*)workSpace;
+    size_t const CTableSize = FSE_CTABLE_SIZE_U32(tableLog, maxSymbolValue);
+    void* scratchBuffer = (void*)(CTable + CTableSize);
+    size_t const scratchBufferSize = wkspSize - (CTableSize * sizeof(FSE_CTable));
 
     /* init conditions */
-    if (srcSize <= 1) return 0;  /* Uncompressible */
+    if (wkspSize < FSE_WKSP_SIZE_U32(tableLog, maxSymbolValue)) return ERROR(tableLog_tooLarge);
+    if (srcSize <= 1) return 0;  /* Not compressible */
     if (!maxSymbolValue) maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
     if (!tableLog) tableLog = FSE_DEFAULT_TABLELOG;
 
     /* Scan input and build symbol stats */
-    errorCode = FSE_count (count, &maxSymbolValue, ip, srcSize);
-    if (FSE_isError(errorCode)) return errorCode;
-    if (errorCode == srcSize) return 1;
-    if (errorCode == 1) return 0;   /* each symbol only present once */
-    if (errorCode < (srcSize >> 7)) return 0;   /* Heuristic : not compressible enough */
+    {   CHECK_V_F(maxCount, FSE_count_wksp(count, &maxSymbolValue, src, srcSize, (unsigned*)scratchBuffer) );
+        if (maxCount == srcSize) return 1;   /* only a single symbol in src : rle */
+        if (maxCount == 1) return 0;         /* each symbol present maximum once => not compressible */
+        if (maxCount < (srcSize >> 7)) return 0;   /* Heuristic : not compressible enough */
+    }
 
     tableLog = FSE_optimalTableLog(tableLog, srcSize, maxSymbolValue);
-    errorCode = FSE_normalizeCount (norm, tableLog, count, srcSize, maxSymbolValue);
-    if (FSE_isError(errorCode)) return errorCode;
+    CHECK_F( FSE_normalizeCount(norm, tableLog, count, srcSize, maxSymbolValue) );
 
     /* Write table description header */
-    errorCode = FSE_writeNCount (op, oend-op, norm, maxSymbolValue, tableLog);
-    if (FSE_isError(errorCode)) return errorCode;
-    op += errorCode;
+    {   CHECK_V_F(nc_err, FSE_writeNCount(op, oend-op, norm, maxSymbolValue, tableLog) );
+        op += nc_err;
+    }
 
     /* Compress */
-    errorCode = FSE_buildCTable (ct, norm, maxSymbolValue, tableLog);
-    if (FSE_isError(errorCode)) return errorCode;
-    errorCode = FSE_compress_usingCTable(op, oend - op, ip, srcSize, ct);
-    if (errorCode == 0) return 0;   /* not enough space for compressed data */
-    op += errorCode;
+    CHECK_F( FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, scratchBuffer, scratchBufferSize) );
+    {   CHECK_V_F(cSize, FSE_compress_usingCTable(op, oend - op, src, srcSize, CTable) );
+        if (cSize == 0) return 0;   /* not enough space for compressed data */
+        op += cSize;
+    }
 
     /* check compressibility */
-    if ( (size_t)(op-ostart) >= srcSize-1 )
-        return 0;
+    if ( (size_t)(op-ostart) >= srcSize-1 ) return 0;
 
     return op-ostart;
 }
 
-size_t FSE_compress (void* dst, size_t dstSize, const void* src, size_t srcSize)
+typedef struct {
+    FSE_CTable CTable_max[FSE_CTABLE_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)];
+    BYTE scratchBuffer[1 << FSE_MAX_TABLELOG];
+} fseWkspMax_t;
+
+size_t FSE_compress2 (void* dst, size_t dstCapacity, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog)
 {
-    return FSE_compress2(dst, dstSize, src, (U32)srcSize, FSE_MAX_SYMBOL_VALUE, FSE_DEFAULT_TABLELOG);
+    fseWkspMax_t scratchBuffer;
+    FSE_STATIC_ASSERT(sizeof(scratchBuffer) >= FSE_WKSP_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE));   /* compilation failures here means scratchBuffer is not large enough */
+    if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
+    return FSE_compress_wksp(dst, dstCapacity, src, srcSize, maxSymbolValue, tableLog, &scratchBuffer, sizeof(scratchBuffer));
+}
+
+size_t FSE_compress (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+{
+    return FSE_compress2(dst, dstCapacity, src, srcSize, FSE_MAX_SYMBOL_VALUE, FSE_DEFAULT_TABLELOG);
 }
 
 

contrib/libzstd/include/zstd/compress/huf_compress.c

@@ -56,6 +56,8 @@
 *  Error Management
 ****************************************************************/
 #define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return f
+#define CHECK_F(f)   { CHECK_V_F(_var_err__, f); }
 
 
 /* **************************************************************
@@ -70,31 +72,73 @@ unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxS
 /* *******************************************************
 *  HUF : Huffman block compression
 *********************************************************/
+/* HUF_compressWeights() :
+ * Same as FSE_compress(), but dedicated to huff0's weights compression.
+ * The use case needs much less stack memory.
+ * Note : all elements within weightTable are supposed to be <= HUF_TABLELOG_MAX.
+ */
+#define MAX_FSE_TABLELOG_FOR_HUFF_HEADER 6
+size_t HUF_compressWeights (void* dst, size_t dstSize, const void* weightTable, size_t wtSize)
+{
+    BYTE* const ostart = (BYTE*) dst;
+    BYTE* op = ostart;
+    BYTE* const oend = ostart + dstSize;
+
+    U32 maxSymbolValue = HUF_TABLELOG_MAX;
+    U32 tableLog = MAX_FSE_TABLELOG_FOR_HUFF_HEADER;
+
+    FSE_CTable CTable[FSE_CTABLE_SIZE_U32(MAX_FSE_TABLELOG_FOR_HUFF_HEADER, HUF_TABLELOG_MAX)];
+    BYTE scratchBuffer[1<<MAX_FSE_TABLELOG_FOR_HUFF_HEADER];
+
+    U32 count[HUF_TABLELOG_MAX+1];
+    S16 norm[HUF_TABLELOG_MAX+1];
+
+    /* init conditions */
+    if (wtSize <= 1) return 0;  /* Not compressible */
+
+    /* Scan input and build symbol stats */
+    {   CHECK_V_F(maxCount, FSE_count_simple(count, &maxSymbolValue, weightTable, wtSize) );
+        if (maxCount == wtSize) return 1;   /* only a single symbol in src : rle */
+        if (maxCount == 1) return 0;         /* each symbol present maximum once => not compressible */
+    }
+
+    tableLog = FSE_optimalTableLog(tableLog, wtSize, maxSymbolValue);
+    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) );
+        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) );
+        if (cSize == 0) return 0;   /* not enough space for compressed data */
+        op += cSize;
+    }
+
+    return op-ostart;
+}
+
+
 struct HUF_CElt_s {
   U16  val;
   BYTE nbBits;
 };   /* typedef'd to HUF_CElt within "huf.h" */
 
-typedef struct nodeElt_s {
-    U32 count;
-    U16 parent;
-    BYTE byte;
-    BYTE nbBits;
-} nodeElt;
-
 /*! HUF_writeCTable() :
-    `CTable` : huffman tree to save, using huf representation.
+    `CTable` : Huffman tree to save, using huf representation.
     @return : size of saved CTable */
 size_t HUF_writeCTable (void* dst, size_t maxDstSize,
                         const HUF_CElt* CTable, U32 maxSymbolValue, U32 huffLog)
 {
-    BYTE bitsToWeight[HUF_TABLELOG_MAX + 1];
+    BYTE bitsToWeight[HUF_TABLELOG_MAX + 1];   /* precomputed conversion table */
     BYTE huffWeight[HUF_SYMBOLVALUE_MAX];
     BYTE* op = (BYTE*)dst;
     U32 n;
 
      /* check conditions */
-    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(GENERIC);
+    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
 
     /* convert to weight */
     bitsToWeight[0] = 0;
@@ -103,38 +147,33 @@ size_t HUF_writeCTable (void* dst, size_t maxDstSize,
     for (n=0; n<maxSymbolValue; n++)
         huffWeight[n] = bitsToWeight[CTable[n].nbBits];
 
-    {   size_t const size = FSE_compress(op+1, maxDstSize-1, huffWeight, maxSymbolValue);
-        if (FSE_isError(size)) return size;
-        if ((size>1) & (size < maxSymbolValue/2)) {   /* FSE compressed */
-            op[0] = (BYTE)size;
-            return size+1;
-        }
-    }
+    /* attempt weights compression by FSE */
+    {   CHECK_V_F(hSize, HUF_compressWeights(op+1, maxDstSize-1, huffWeight, maxSymbolValue) );
+        if ((hSize>1) & (hSize < maxSymbolValue/2)) {   /* FSE compressed */
+            op[0] = (BYTE)hSize;
+            return hSize+1;
+    }   }
 
-    /* raw values */
-    if (maxSymbolValue > (256-128)) return ERROR(GENERIC);   /* should not happen */
+    /* write raw values as 4-bits (max : 15) */
+    if (maxSymbolValue > (256-128)) return ERROR(GENERIC);   /* should not happen : likely means source cannot be compressed */
     if (((maxSymbolValue+1)/2) + 1 > maxDstSize) return ERROR(dstSize_tooSmall);   /* not enough space within dst buffer */
     op[0] = (BYTE)(128 /*special case*/ + (maxSymbolValue-1));
-    huffWeight[maxSymbolValue] = 0;   /* to be sure it doesn't cause issue in final combination */
+    huffWeight[maxSymbolValue] = 0;   /* to be sure it doesn't cause msan issue in final combination */
     for (n=0; n<maxSymbolValue; n+=2)
         op[(n/2)+1] = (BYTE)((huffWeight[n] << 4) + huffWeight[n+1]);
     return ((maxSymbolValue+1)/2) + 1;
-
 }
 
 
 size_t HUF_readCTable (HUF_CElt* CTable, U32 maxSymbolValue, const void* src, size_t srcSize)
 {
-    BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1];
+    BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1];   /* init not required, even though some static analyzer may complain */
     U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];   /* large enough for values from 0 to 16 */
     U32 tableLog = 0;
-    size_t readSize;
     U32 nbSymbols = 0;
-    /*memset(huffWeight, 0, sizeof(huffWeight));*/   /* is not necessary, even though some analyzer complain ... */
 
     /* get symbol weights */
-    readSize = HUF_readStats(huffWeight, HUF_SYMBOLVALUE_MAX+1, rankVal, &nbSymbols, &tableLog, src, srcSize);
-    if (HUF_isError(readSize)) return readSize;
+    CHECK_V_F(readSize, HUF_readStats(huffWeight, HUF_SYMBOLVALUE_MAX+1, rankVal, &nbSymbols, &tableLog, src, srcSize));
 
     /* check result */
     if (tableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
@@ -155,12 +194,13 @@ size_t HUF_readCTable (HUF_CElt* CTable, U32 maxSymbolValue, const void* src, si
     }   }
 
     /* fill val */
-    {   U16 nbPerRank[HUF_TABLELOG_MAX+1] = {0};
-        U16 valPerRank[HUF_TABLELOG_MAX+1] = {0};
+    {   U16 nbPerRank[HUF_TABLELOG_MAX+2]  = {0};  /* support w=0=>n=tableLog+1 */
+        U16 valPerRank[HUF_TABLELOG_MAX+2] = {0};
         { U32 n; for (n=0; n<nbSymbols; n++) nbPerRank[CTable[n].nbBits]++; }
         /* determine stating value per rank */
+        valPerRank[tableLog+1] = 0;   /* for w==0 */
         {   U16 min = 0;
-            U32 n; for (n=HUF_TABLELOG_MAX; n>0; n--) {
+            U32 n; for (n=tableLog; n>0; n--) {  /* start at n=tablelog <-> w=1 */
                 valPerRank[n] = min;     /* get starting value within each rank */
                 min += nbPerRank[n];
                 min >>= 1;
@@ -173,6 +213,13 @@ size_t HUF_readCTable (HUF_CElt* CTable, U32 maxSymbolValue, const void* src, si
 }
 
 
+typedef struct nodeElt_s {
+    U32 count;
+    U16 parent;
+    BYTE byte;
+    BYTE nbBits;
+} nodeElt;
+
 static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
 {
     const U32 largestBits = huffNode[lastNonNull].nbBits;
@@ -219,7 +266,8 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
                         if (highTotal <= lowTotal) break;
                 }   }
                 /* only triggered when no more rank 1 symbol left => find closest one (note : there is necessarily at least one !) */
-                while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))  /* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
+                /* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
+                while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))
                     nBitsToDecrease ++;
                 totalCost -= 1 << (nBitsToDecrease-1);
                 if (rankLast[nBitsToDecrease-1] == noSymbol)
@@ -278,20 +326,26 @@ static void HUF_sort(nodeElt* huffNode, const U32* count, U32 maxSymbolValue)
 }
 
 
+/** 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 1024 unsigned.
+ */
 #define STARTNODE (HUF_SYMBOLVALUE_MAX+1)
-size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits)
+typedef nodeElt huffNodeTable[2*HUF_SYMBOLVALUE_MAX+1 +1];
+size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
 {
-    nodeElt huffNode0[2*HUF_SYMBOLVALUE_MAX+1 +1];
-    nodeElt* huffNode = huffNode0 + 1;
+    nodeElt* const huffNode0 = (nodeElt*)workSpace;
+    nodeElt* const huffNode = huffNode0+1;
     U32 n, nonNullRank;
     int lowS, lowN;
     U16 nodeNb = STARTNODE;
     U32 nodeRoot;
 
     /* safety checks */
+    if (wkspSize < sizeof(huffNodeTable)) return ERROR(GENERIC);   /* workSpace is not large enough */
     if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
     if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(GENERIC);
-    memset(huffNode0, 0, sizeof(huffNode0));
+    memset(huffNode0, 0, sizeof(huffNodeTable));
 
     /* sort, decreasing order */
     HUF_sort(huffNode, count, maxSymbolValue);
@@ -304,7 +358,7 @@ size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U3
     huffNode[lowS].parent = huffNode[lowS-1].parent = nodeNb;
     nodeNb++; lowS-=2;
     for (n=nodeNb; n<=nodeRoot; n++) huffNode[n].count = (U32)(1U<<30);
-    huffNode0[0].count = (U32)(1U<<31);
+    huffNode0[0].count = (U32)(1U<<31);  /* fake entry, strong barrier */
 
     /* create parents */
     while (nodeNb <= nodeRoot) {
@@ -347,6 +401,34 @@ size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U3
     return maxNbBits;
 }
 
+/** HUF_buildCTable() :
+ *  Note : count is used before tree is written, so they can safely overlap
+ */
+size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits)
+{
+    huffNodeTable nodeTable;
+    return HUF_buildCTable_wksp(tree, count, maxSymbolValue, maxNbBits, nodeTable, sizeof(nodeTable));
+}
+
+static size_t HUF_estimateCompressedSize(HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue)
+{
+    size_t nbBits = 0;
+    int s;
+    for (s = 0; s <= (int)maxSymbolValue; ++s) {
+        nbBits += CTable[s].nbBits * count[s];
+    }
+    return nbBits >> 3;
+}
+
+static int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue) {
+  int bad = 0;
+  int s;
+  for (s = 0; s <= (int)maxSymbolValue; ++s) {
+    bad |= (count[s] != 0) & (CTable[s].nbBits == 0);
+  }
+  return !bad;
+}
+
 static void HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable)
 {
     BIT_addBitsFast(bitCPtr, CTable[symbol].val, CTable[symbol].nbBits);
@@ -374,20 +456,23 @@ size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, si
 
     /* init */
     if (dstSize < 8) return 0;   /* not enough space to compress */
-    { size_t const errorCode = BIT_initCStream(&bitC, op, oend-op);
-      if (HUF_isError(errorCode)) return 0; }
+    { size_t const initErr = BIT_initCStream(&bitC, op, oend-op);
+      if (HUF_isError(initErr)) return 0; }
 
     n = srcSize & ~3;  /* join to mod 4 */
     switch (srcSize & 3)
     {
         case 3 : HUF_encodeSymbol(&bitC, ip[n+ 2], CTable);
                  HUF_FLUSHBITS_2(&bitC);
+		 /* fall-through */
         case 2 : HUF_encodeSymbol(&bitC, ip[n+ 1], CTable);
                  HUF_FLUSHBITS_1(&bitC);
+		 /* fall-through */
         case 1 : HUF_encodeSymbol(&bitC, ip[n+ 0], CTable);
                  HUF_FLUSHBITS(&bitC);
-        case 0 :
-        default: ;
+		 /* fall-through */
+        case 0 : /* fall-through */
+        default: break;
     }
 
     for (; n>0; n-=4) {  /* note : n&3==0 at this stage */
@@ -418,32 +503,28 @@ size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, si
     if (srcSize < 12) return 0;   /* no saving possible : too small input */
     op += 6;   /* jumpTable */
 
-    {   size_t const cSize = HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable);
-        if (HUF_isError(cSize)) return cSize;
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
         if (cSize==0) return 0;
         MEM_writeLE16(ostart, (U16)cSize);
         op += cSize;
     }
 
     ip += segmentSize;
-    {   size_t const cSize = HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable);
-        if (HUF_isError(cSize)) return cSize;
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
         if (cSize==0) return 0;
         MEM_writeLE16(ostart+2, (U16)cSize);
         op += cSize;
     }
 
     ip += segmentSize;
-    {   size_t const cSize = HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable);
-        if (HUF_isError(cSize)) return cSize;
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
         if (cSize==0) return 0;
         MEM_writeLE16(ostart+4, (U16)cSize);
         op += cSize;
     }
 
     ip += segmentSize;
-    {   size_t const cSize = HUF_compress1X_usingCTable(op, oend-op, ip, iend-ip, CTable);
-        if (HUF_isError(cSize)) return cSize;
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, iend-ip, CTable) );
         if (cSize==0) return 0;
         op += cSize;
     }
@@ -452,20 +533,43 @@ size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, si
 }
 
 
+static size_t HUF_compressCTable_internal(
+                BYTE* const ostart, BYTE* op, BYTE* const oend,
+                const void* src, size_t srcSize,
+                unsigned singleStream, const HUF_CElt* CTable)
+{
+    size_t const cSize = singleStream ?
+                         HUF_compress1X_usingCTable(op, oend - op, src, srcSize, CTable) :
+                         HUF_compress4X_usingCTable(op, oend - op, src, srcSize, CTable);
+    if (HUF_isError(cSize)) { return cSize; }
+    if (cSize==0) { return 0; }   /* uncompressible */
+    op += cSize;
+    /* check compressibility */
+    if ((size_t)(op-ostart) >= srcSize-1) { return 0; }
+    return op-ostart;
+}
+
+
+/* `workSpace` must a table of at least 1024 unsigned */
 static size_t HUF_compress_internal (
                 void* dst, size_t dstSize,
                 const void* src, size_t srcSize,
                 unsigned maxSymbolValue, unsigned huffLog,
-                unsigned singleStream)
+                unsigned singleStream,
+                void* workSpace, size_t wkspSize,
+                HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat)
 {
     BYTE* const ostart = (BYTE*)dst;
     BYTE* const oend = ostart + dstSize;
     BYTE* op = ostart;
 
-    U32 count[HUF_SYMBOLVALUE_MAX+1];
-    HUF_CElt CTable[HUF_SYMBOLVALUE_MAX+1];
+    U32* count;
+    size_t const countSize = sizeof(U32) * (HUF_SYMBOLVALUE_MAX + 1);
+    HUF_CElt* CTable;
+    size_t const CTableSize = sizeof(HUF_CElt) * (HUF_SYMBOLVALUE_MAX + 1);
 
     /* checks & inits */
+    if (wkspSize < sizeof(huffNodeTable) + countSize + CTableSize) return ERROR(GENERIC);
     if (!srcSize) return 0;  /* Uncompressed (note : 1 means rle, so first byte must be correct) */
     if (!dstSize) return 0;  /* cannot fit within dst budget */
     if (srcSize > HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong);   /* current block size limit */
@@ -473,59 +577,111 @@ static size_t HUF_compress_internal (
     if (!maxSymbolValue) maxSymbolValue = HUF_SYMBOLVALUE_MAX;
     if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT;
 
+    count = (U32*)workSpace;
+    workSpace = (BYTE*)workSpace + countSize;
+    wkspSize -= countSize;
+    CTable = (HUF_CElt*)workSpace;
+    workSpace = (BYTE*)workSpace + CTableSize;
+    wkspSize -= CTableSize;
+
+    /* Heuristic : If we don't need to check the validity of the old table use the old table for small inputs */
+    if (preferRepeat && repeat && *repeat == HUF_repeat_valid) {
+        return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
+    }
+
     /* Scan input and build symbol stats */
-    {   size_t const largest = FSE_count (count, &maxSymbolValue, (const BYTE*)src, srcSize);
-        if (HUF_isError(largest)) return largest;
+    {   CHECK_V_F(largest, FSE_count_wksp (count, &maxSymbolValue, (const BYTE*)src, srcSize, (U32*)workSpace) );
         if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; }   /* single symbol, rle */
         if (largest <= (srcSize >> 7)+1) return 0;   /* Fast heuristic : not compressible enough */
     }
 
+    /* Check validity of previous table */
+    if (repeat && *repeat == HUF_repeat_check && !HUF_validateCTable(oldHufTable, count, maxSymbolValue)) {
+        *repeat = HUF_repeat_none;
+    }
+    /* Heuristic : use existing table for small inputs */
+    if (preferRepeat && repeat && *repeat != HUF_repeat_none) {
+        return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
+    }
+
     /* Build Huffman Tree */
     huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
-    {   size_t const maxBits = HUF_buildCTable (CTable, count, maxSymbolValue, huffLog);
-        if (HUF_isError(maxBits)) return maxBits;
+    {   CHECK_V_F(maxBits, HUF_buildCTable_wksp (CTable, count, maxSymbolValue, huffLog, workSpace, wkspSize) );
         huffLog = (U32)maxBits;
+        /* Zero the unused symbols so we can check it for validity */
+        memset(CTable + maxSymbolValue + 1, 0, CTableSize - (maxSymbolValue + 1) * sizeof(HUF_CElt));
     }
 
     /* Write table description header */
-    {   size_t const hSize = HUF_writeCTable (op, dstSize, CTable, maxSymbolValue, huffLog);
-        if (HUF_isError(hSize)) return hSize;
-        if (hSize + 12 >= srcSize) return 0;   /* not useful to try compression */
+    {   CHECK_V_F(hSize, HUF_writeCTable (op, dstSize, CTable, maxSymbolValue, huffLog) );
+        /* Check if using the previous table will be beneficial */
+        if (repeat && *repeat != HUF_repeat_none) {
+            size_t const oldSize = HUF_estimateCompressedSize(oldHufTable, count, maxSymbolValue);
+            size_t const newSize = HUF_estimateCompressedSize(CTable, count, maxSymbolValue);
+            if (oldSize <= hSize + newSize || hSize + 12 >= srcSize) {
+                return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
+            }
+        }
+        /* Use the new table */
+        if (hSize + 12ul >= srcSize) { return 0; }
         op += hSize;
+        if (repeat) { *repeat = HUF_repeat_none; }
+        if (oldHufTable) { memcpy(oldHufTable, CTable, CTableSize); } /* Save the new table */
     }
-
-    /* Compress */
-    {   size_t const cSize = (singleStream) ?
-                            HUF_compress1X_usingCTable(op, oend - op, src, srcSize, CTable) :   /* single segment */
-                            HUF_compress4X_usingCTable(op, oend - op, src, srcSize, CTable);
-        if (HUF_isError(cSize)) return cSize;
-        if (cSize==0) return 0;   /* uncompressible */
-        op += cSize;
-    }
-
-    /* check compressibility */
-    if ((size_t)(op-ostart) >= srcSize-1)
-        return 0;
-
-    return op-ostart;
+    return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, CTable);
 }
 
 
+size_t HUF_compress1X_wksp (void* dst, size_t dstSize,
+                      const void* src, size_t srcSize,
+                      unsigned maxSymbolValue, unsigned huffLog,
+                      void* workSpace, size_t wkspSize)
+{
+    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1 /* single stream */, workSpace, wkspSize, NULL, NULL, 0);
+}
+
+size_t HUF_compress1X_repeat (void* dst, size_t dstSize,
+                      const void* src, size_t srcSize,
+                      unsigned maxSymbolValue, unsigned huffLog,
+                      void* workSpace, size_t wkspSize,
+                      HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat)
+{
+    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1 /* single stream */, workSpace, wkspSize, hufTable, repeat, preferRepeat);
+}
+
 size_t HUF_compress1X (void* dst, size_t dstSize,
                  const void* src, size_t srcSize,
                  unsigned maxSymbolValue, unsigned huffLog)
 {
-    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1);
+    unsigned workSpace[1024];
+    return HUF_compress1X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
+}
+
+size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
+                      const void* src, size_t srcSize,
+                      unsigned maxSymbolValue, unsigned huffLog,
+                      void* workSpace, size_t wkspSize)
+{
+    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0 /* 4 streams */, workSpace, wkspSize, NULL, NULL, 0);
+}
+
+size_t HUF_compress4X_repeat (void* dst, size_t dstSize,
+                      const void* src, size_t srcSize,
+                      unsigned maxSymbolValue, unsigned huffLog,
+                      void* workSpace, size_t wkspSize,
+                      HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat)
+{
+    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0 /* 4 streams */, workSpace, wkspSize, hufTable, repeat, preferRepeat);
 }
 
 size_t HUF_compress2 (void* dst, size_t dstSize,
                 const void* src, size_t srcSize,
                 unsigned maxSymbolValue, unsigned huffLog)
 {
-    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0);
+    unsigned workSpace[1024];
+    return HUF_compress4X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
 }
 
-
 size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
 {
     return HUF_compress2(dst, maxDstSize, src, (U32)srcSize, 255, HUF_TABLELOG_DEFAULT);

contrib/libzstd/include/zstd/compress/zbuff_compress.c

@@ -1,319 +0,0 @@
-/**
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
- * All rights reserved.
- *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
- */
-
-
-
-/* *************************************
-*  Dependencies
-***************************************/
-#include <stdlib.h>
-#include "error_private.h"
-#include "zstd_internal.h"  /* MIN, ZSTD_BLOCKHEADERSIZE, defaultCustomMem */
-#define ZBUFF_STATIC_LINKING_ONLY
-#include "zbuff.h"
-
-
-/* *************************************
-*  Constants
-***************************************/
-static size_t const ZBUFF_endFrameSize = ZSTD_BLOCKHEADERSIZE;
-
-
-/*-***********************************************************
-*  Streaming compression
-*
-*  A ZBUFF_CCtx object is required to track streaming operation.
-*  Use ZBUFF_createCCtx() and ZBUFF_freeCCtx() to create/release resources.
-*  Use ZBUFF_compressInit() to start a new compression operation.
-*  ZBUFF_CCtx objects can be reused multiple times.
-*
-*  Use ZBUFF_compressContinue() repetitively to consume your input.
-*  *srcSizePtr and *dstCapacityPtr can be any size.
-*  The function will report how many bytes were read or written by modifying *srcSizePtr and *dstCapacityPtr.
-*  Note that it may not consume the entire input, in which case it's up to the caller to call again the function with remaining input.
-*  The content of dst will be overwritten (up to *dstCapacityPtr) at each function call, so save its content if it matters or change dst .
-*  @return : a hint to preferred nb of bytes to use as input for next function call (it's only a hint, to improve latency)
-*            or an error code, which can be tested using ZBUFF_isError().
-*
-*  ZBUFF_compressFlush() can be used to instruct ZBUFF to compress and output whatever remains within its buffer.
-*  Note that it will not output more than *dstCapacityPtr.
-*  Therefore, some content might still be left into its internal buffer if dst buffer is too small.
-*  @return : nb of bytes still present into internal buffer (0 if it's empty)
-*            or an error code, which can be tested using ZBUFF_isError().
-*
-*  ZBUFF_compressEnd() instructs to finish a frame.
-*  It will perform a flush and write frame epilogue.
-*  Similar to ZBUFF_compressFlush(), it may not be able to output the entire internal buffer content if *dstCapacityPtr is too small.
-*  @return : nb of bytes still present into internal buffer (0 if it's empty)
-*            or an error code, which can be tested using ZBUFF_isError().
-*
-*  Hint : recommended buffer sizes (not compulsory)
-*  input : ZSTD_BLOCKSIZE_MAX (128 KB), internal unit size, it improves latency to use this value.
-*  output : ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + ZBUFF_endFrameSize : ensures it's always possible to write/flush/end a full block at best speed.
-* ***********************************************************/
-
-typedef enum { ZBUFFcs_init, ZBUFFcs_load, ZBUFFcs_flush, ZBUFFcs_final } ZBUFF_cStage;
-
-/* *** Resources *** */
-struct ZBUFF_CCtx_s {
-    ZSTD_CCtx* zc;
-    char*  inBuff;
-    size_t inBuffSize;
-    size_t inToCompress;
-    size_t inBuffPos;
-    size_t inBuffTarget;
-    size_t blockSize;
-    char*  outBuff;
-    size_t outBuffSize;
-    size_t outBuffContentSize;
-    size_t outBuffFlushedSize;
-    ZBUFF_cStage stage;
-    U32    checksum;
-    U32    frameEnded;
-    ZSTD_customMem customMem;
-};   /* typedef'd tp ZBUFF_CCtx within "zbuff.h" */
-
-ZBUFF_CCtx* ZBUFF_createCCtx(void)
-{
-    return ZBUFF_createCCtx_advanced(defaultCustomMem);
-}
-
-ZBUFF_CCtx* ZBUFF_createCCtx_advanced(ZSTD_customMem customMem)
-{
-    ZBUFF_CCtx* zbc;
-
-    if (!customMem.customAlloc && !customMem.customFree)
-        customMem = defaultCustomMem;
-
-    if (!customMem.customAlloc || !customMem.customFree)
-        return NULL;
-
-    zbc = (ZBUFF_CCtx*)customMem.customAlloc(customMem.opaque, sizeof(ZBUFF_CCtx));
-    if (zbc==NULL) return NULL;
-    memset(zbc, 0, sizeof(ZBUFF_CCtx));
-    memcpy(&zbc->customMem, &customMem, sizeof(ZSTD_customMem));
-    zbc->zc = ZSTD_createCCtx_advanced(customMem);
-    if (zbc->zc == NULL) { ZBUFF_freeCCtx(zbc); return NULL; }
-    return zbc;
-}
-
-size_t ZBUFF_freeCCtx(ZBUFF_CCtx* zbc)
-{
-    if (zbc==NULL) return 0;   /* support free on NULL */
-    ZSTD_freeCCtx(zbc->zc);
-    if (zbc->inBuff) zbc->customMem.customFree(zbc->customMem.opaque, zbc->inBuff);
-    if (zbc->outBuff) zbc->customMem.customFree(zbc->customMem.opaque, zbc->outBuff);
-    zbc->customMem.customFree(zbc->customMem.opaque, zbc);
-    return 0;
-}
-
-
-/* ======   Initialization   ====== */
-
-size_t ZBUFF_compressInit_advanced(ZBUFF_CCtx* zbc,
-                                   const void* dict, size_t dictSize,
-                                   ZSTD_parameters params, unsigned long long pledgedSrcSize)
-{
-    /* allocate buffers */
-    {   size_t const neededInBuffSize = (size_t)1 << params.cParams.windowLog;
-        if (zbc->inBuffSize < neededInBuffSize) {
-            zbc->inBuffSize = neededInBuffSize;
-            zbc->customMem.customFree(zbc->customMem.opaque, zbc->inBuff);   /* should not be necessary */
-            zbc->inBuff = (char*)zbc->customMem.customAlloc(zbc->customMem.opaque, neededInBuffSize);
-            if (zbc->inBuff == NULL) return ERROR(memory_allocation);
-        }
-        zbc->blockSize = MIN(ZSTD_BLOCKSIZE_ABSOLUTEMAX, neededInBuffSize);
-    }
-    if (zbc->outBuffSize < ZSTD_compressBound(zbc->blockSize)+1) {
-        zbc->outBuffSize = ZSTD_compressBound(zbc->blockSize)+1;
-        zbc->customMem.customFree(zbc->customMem.opaque, zbc->outBuff);   /* should not be necessary */
-        zbc->outBuff = (char*)zbc->customMem.customAlloc(zbc->customMem.opaque, zbc->outBuffSize);
-        if (zbc->outBuff == NULL) return ERROR(memory_allocation);
-    }
-
-    { size_t const errorCode = ZSTD_compressBegin_advanced(zbc->zc, dict, dictSize, params, pledgedSrcSize);
-      if (ZSTD_isError(errorCode)) return errorCode; }
-
-    zbc->inToCompress = 0;
-    zbc->inBuffPos = 0;
-    zbc->inBuffTarget = zbc->blockSize;
-    zbc->outBuffContentSize = zbc->outBuffFlushedSize = 0;
-    zbc->stage = ZBUFFcs_load;
-    zbc->checksum = params.fParams.checksumFlag > 0;
-    zbc->frameEnded = 0;
-    return 0;   /* ready to go */
-}
-
-
-size_t ZBUFF_compressInitDictionary(ZBUFF_CCtx* zbc, const void* dict, size_t dictSize, int compressionLevel)
-{
-    ZSTD_parameters const params = ZSTD_getParams(compressionLevel, 0, dictSize);
-    return ZBUFF_compressInit_advanced(zbc, dict, dictSize, params, 0);
-}
-
-size_t ZBUFF_compressInit(ZBUFF_CCtx* zbc, int compressionLevel)
-{
-    return ZBUFF_compressInitDictionary(zbc, NULL, 0, compressionLevel);
-}
-
-
-/* internal util function */
-MEM_STATIC size_t ZBUFF_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
-{
-    size_t const length = MIN(dstCapacity, srcSize);
-    memcpy(dst, src, length);
-    return length;
-}
-
-
-/* ======   Compression   ====== */
-
-typedef enum { zbf_gather, zbf_flush, zbf_end } ZBUFF_flush_e;
-
-static size_t ZBUFF_compressContinue_generic(ZBUFF_CCtx* zbc,
-                              void* dst, size_t* dstCapacityPtr,
-                        const void* src, size_t* srcSizePtr,
-                              ZBUFF_flush_e const flush)
-{
-    U32 someMoreWork = 1;
-    const char* const istart = (const char*)src;
-    const char* const iend = istart + *srcSizePtr;
-    const char* ip = istart;
-    char* const ostart = (char*)dst;
-    char* const oend = ostart + *dstCapacityPtr;
-    char* op = ostart;
-
-    while (someMoreWork) {
-        switch(zbc->stage)
-        {
-        case ZBUFFcs_init: return ERROR(init_missing);   /* call ZBUFF_compressInit() first ! */
-
-        case ZBUFFcs_load:
-            /* complete inBuffer */
-            {   size_t const toLoad = zbc->inBuffTarget - zbc->inBuffPos;
-                size_t const loaded = ZBUFF_limitCopy(zbc->inBuff + zbc->inBuffPos, toLoad, ip, iend-ip);
-                zbc->inBuffPos += loaded;
-                ip += loaded;
-                if ( (zbc->inBuffPos==zbc->inToCompress) || (!flush && (toLoad != loaded)) ) {
-                    someMoreWork = 0; break;  /* not enough input to get a full block : stop there, wait for more */
-            }   }
-            /* compress current block (note : this stage cannot be stopped in the middle) */
-            {   void* cDst;
-                size_t cSize;
-                size_t const iSize = zbc->inBuffPos - zbc->inToCompress;
-                size_t oSize = oend-op;
-                if (oSize >= ZSTD_compressBound(iSize))
-                    cDst = op;   /* compress directly into output buffer (avoid flush stage) */
-                else
-                    cDst = zbc->outBuff, oSize = zbc->outBuffSize;
-                cSize = (flush == zbf_end) ?
-                        ZSTD_compressEnd(zbc->zc, cDst, oSize, zbc->inBuff + zbc->inToCompress, iSize) :
-                        ZSTD_compressContinue(zbc->zc, cDst, oSize, zbc->inBuff + zbc->inToCompress, iSize);
-                if (ZSTD_isError(cSize)) return cSize;
-                if (flush == zbf_end) zbc->frameEnded = 1;
-                /* prepare next block */
-                zbc->inBuffTarget = zbc->inBuffPos + zbc->blockSize;
-                if (zbc->inBuffTarget > zbc->inBuffSize)
-                    zbc->inBuffPos = 0, zbc->inBuffTarget = zbc->blockSize;   /* note : inBuffSize >= blockSize */
-                zbc->inToCompress = zbc->inBuffPos;
-                if (cDst == op) { op += cSize; break; }   /* no need to flush */
-                zbc->outBuffContentSize = cSize;
-                zbc->outBuffFlushedSize = 0;
-                zbc->stage = ZBUFFcs_flush;   /* continue to flush stage */
-            }
-
-        case ZBUFFcs_flush:
-            {   size_t const toFlush = zbc->outBuffContentSize - zbc->outBuffFlushedSize;
-                size_t const flushed = ZBUFF_limitCopy(op, oend-op, zbc->outBuff + zbc->outBuffFlushedSize, toFlush);
-                op += flushed;
-                zbc->outBuffFlushedSize += flushed;
-                if (toFlush!=flushed) { someMoreWork = 0; break; } /* dst too small to store flushed data : stop there */
-                zbc->outBuffContentSize = zbc->outBuffFlushedSize = 0;
-                zbc->stage = ZBUFFcs_load;
-                break;
-            }
-
-        case ZBUFFcs_final:
-            someMoreWork = 0;   /* do nothing */
-            break;
-
-        default:
-            return ERROR(GENERIC);   /* impossible */
-        }
-    }
-
-    *srcSizePtr = ip - istart;
-    *dstCapacityPtr = op - ostart;
-    if (zbc->frameEnded) return 0;
-    {   size_t hintInSize = zbc->inBuffTarget - zbc->inBuffPos;
-        if (hintInSize==0) hintInSize = zbc->blockSize;
-        return hintInSize;
-    }
-}
-
-size_t ZBUFF_compressContinue(ZBUFF_CCtx* zbc,
-                              void* dst, size_t* dstCapacityPtr,
-                        const void* src, size_t* srcSizePtr)
-{
-    return ZBUFF_compressContinue_generic(zbc, dst, dstCapacityPtr, src, srcSizePtr, zbf_gather);
-}
-
-
-
-/* ======   Finalize   ====== */
-
-size_t ZBUFF_compressFlush(ZBUFF_CCtx* zbc, void* dst, size_t* dstCapacityPtr)
-{
-    size_t srcSize = 0;
-    ZBUFF_compressContinue_generic(zbc, dst, dstCapacityPtr, &srcSize, &srcSize, zbf_flush);  /* use a valid src address instead of NULL */
-    return zbc->outBuffContentSize - zbc->outBuffFlushedSize;
-}
-
-
-size_t ZBUFF_compressEnd(ZBUFF_CCtx* zbc, void* dst, size_t* dstCapacityPtr)
-{
-    BYTE* const ostart = (BYTE*)dst;
-    BYTE* const oend = ostart + *dstCapacityPtr;
-    BYTE* op = ostart;
-
-    if (zbc->stage != ZBUFFcs_final) {
-        /* flush whatever remains */
-        size_t outSize = *dstCapacityPtr;
-        size_t srcSize = 0;
-        size_t const notEnded = ZBUFF_compressContinue_generic(zbc, dst, &outSize, &srcSize, &srcSize, zbf_end);  /* use a valid address instead of NULL */
-        size_t const remainingToFlush = zbc->outBuffContentSize - zbc->outBuffFlushedSize;
-        op += outSize;
-        if (remainingToFlush) {
-            *dstCapacityPtr = op-ostart;
-            return remainingToFlush + ZBUFF_endFrameSize + (zbc->checksum * 4);
-        }
-        /* create epilogue */
-        zbc->stage = ZBUFFcs_final;
-        zbc->outBuffContentSize = !notEnded ? 0 :
-            ZSTD_compressEnd(zbc->zc, zbc->outBuff, zbc->outBuffSize, NULL, 0);  /* write epilogue into outBuff */
-    }
-
-    /* flush epilogue */
-    {   size_t const toFlush = zbc->outBuffContentSize - zbc->outBuffFlushedSize;
-        size_t const flushed = ZBUFF_limitCopy(op, oend-op, zbc->outBuff + zbc->outBuffFlushedSize, toFlush);
-        op += flushed;
-        zbc->outBuffFlushedSize += flushed;
-        *dstCapacityPtr = op-ostart;
-        if (toFlush==flushed) zbc->stage = ZBUFFcs_init;  /* end reached */
-        return toFlush - flushed;
-    }
-}
-
-
-
-/* *************************************
-*  Tool functions
-***************************************/
-size_t ZBUFF_recommendedCInSize(void)  { return ZSTD_BLOCKSIZE_ABSOLUTEMAX; }
-size_t ZBUFF_recommendedCOutSize(void) { return ZSTD_compressBound(ZSTD_BLOCKSIZE_ABSOLUTEMAX) + ZSTD_blockHeaderSize + ZBUFF_endFrameSize; }

contrib/libzstd/include/zstd/compress/zstd_opt.h

@@ -1,5 +1,5 @@
 /**
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-present, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under the BSD-style license found in the
@@ -15,7 +15,9 @@
 #define ZSTD_OPT_H_91842398743
 
 
-#define ZSTD_FREQ_DIV   5
+#define ZSTD_LITFREQ_ADD    2
+#define ZSTD_FREQ_DIV       4
+#define ZSTD_MAX_PRICE      (1<<30)
 
 /*-*************************************
 *  Price functions for optimal parser
@@ -30,22 +32,33 @@ FORCE_INLINE void ZSTD_setLog2Prices(seqStore_t* ssPtr)
 }
 
 
-MEM_STATIC void ZSTD_rescaleFreqs(seqStore_t* ssPtr)
+MEM_STATIC void ZSTD_rescaleFreqs(seqStore_t* ssPtr, const BYTE* src, size_t srcSize)
 {
     unsigned u;
 
     ssPtr->cachedLiterals = NULL;
     ssPtr->cachedPrice = ssPtr->cachedLitLength = 0;
+    ssPtr->staticPrices = 0;
 
     if (ssPtr->litLengthSum == 0) {
-        ssPtr->litSum = (2<<Litbits);
+        if (srcSize <= 1024) ssPtr->staticPrices = 1;
+
+        assert(ssPtr->litFreq!=NULL);
+        for (u=0; u<=MaxLit; u++)
+            ssPtr->litFreq[u] = 0;
+        for (u=0; u<srcSize; u++)
+            ssPtr->litFreq[src[u]]++;
+
+        ssPtr->litSum = 0;
         ssPtr->litLengthSum = MaxLL+1;
         ssPtr->matchLengthSum = MaxML+1;
         ssPtr->offCodeSum = (MaxOff+1);
-        ssPtr->matchSum = (2<<Litbits);
+        ssPtr->matchSum = (ZSTD_LITFREQ_ADD<<Litbits);
 
-        for (u=0; u<=MaxLit; u++)
-            ssPtr->litFreq[u] = 2;
+        for (u=0; u<=MaxLit; u++) {
+            ssPtr->litFreq[u] = 1 + (ssPtr->litFreq[u]>>ZSTD_FREQ_DIV);
+            ssPtr->litSum += ssPtr->litFreq[u];
+        }
         for (u=0; u<=MaxLL; u++)
             ssPtr->litLengthFreq[u] = 1;
         for (u=0; u<=MaxML; u++)
@@ -60,11 +73,11 @@ MEM_STATIC void ZSTD_rescaleFreqs(seqStore_t* ssPtr)
         ssPtr->litSum = 0;
 
         for (u=0; u<=MaxLit; u++) {
-            ssPtr->litFreq[u] = 1 + (ssPtr->litFreq[u]>>ZSTD_FREQ_DIV);
+            ssPtr->litFreq[u] = 1 + (ssPtr->litFreq[u]>>(ZSTD_FREQ_DIV+1));
             ssPtr->litSum += ssPtr->litFreq[u];
         }
         for (u=0; u<=MaxLL; u++) {
-            ssPtr->litLengthFreq[u] = 1 + (ssPtr->litLengthFreq[u]>>ZSTD_FREQ_DIV);
+            ssPtr->litLengthFreq[u] = 1 + (ssPtr->litLengthFreq[u]>>(ZSTD_FREQ_DIV+1));
             ssPtr->litLengthSum += ssPtr->litLengthFreq[u];
         }
         for (u=0; u<=MaxML; u++) {
@@ -72,6 +85,7 @@ MEM_STATIC void ZSTD_rescaleFreqs(seqStore_t* ssPtr)
             ssPtr->matchLengthSum += ssPtr->matchLengthFreq[u];
             ssPtr->matchSum += ssPtr->matchLengthFreq[u] * (u + 3);
         }
+        ssPtr->matchSum *= ZSTD_LITFREQ_ADD;
         for (u=0; u<=MaxOff; u++) {
             ssPtr->offCodeFreq[u] = 1 + (ssPtr->offCodeFreq[u]>>ZSTD_FREQ_DIV);
             ssPtr->offCodeSum += ssPtr->offCodeFreq[u];
@@ -86,6 +100,9 @@ FORCE_INLINE U32 ZSTD_getLiteralPrice(seqStore_t* ssPtr, U32 litLength, const BY
 {
     U32 price, u;
 
+    if (ssPtr->staticPrices)
+        return ZSTD_highbit32((U32)litLength+1) + (litLength*6);
+
     if (litLength == 0)
         return ssPtr->log2litLengthSum - ZSTD_highbit32(ssPtr->litLengthFreq[0]+1);
 
@@ -120,11 +137,17 @@ FORCE_INLINE U32 ZSTD_getLiteralPrice(seqStore_t* ssPtr, U32 litLength, const BY
 }
 
 
-FORCE_INLINE U32 ZSTD_getPrice(seqStore_t* seqStorePtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength)
+FORCE_INLINE U32 ZSTD_getPrice(seqStore_t* seqStorePtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength, const int ultra)
 {
     /* offset */
+    U32 price;
     BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
-    U32 price = offCode + seqStorePtr->log2offCodeSum - ZSTD_highbit32(seqStorePtr->offCodeFreq[offCode]+1);
+
+    if (seqStorePtr->staticPrices)
+        return ZSTD_getLiteralPrice(seqStorePtr, litLength, literals) + ZSTD_highbit32((U32)matchLength+1) + 16 + offCode;
+
+    price = offCode + seqStorePtr->log2offCodeSum - ZSTD_highbit32(seqStorePtr->offCodeFreq[offCode]+1);
+    if (!ultra && offCode >= 20) price += (offCode-19)*2;
 
     /* match Length */
     {   const BYTE ML_deltaCode = 36;
@@ -141,9 +164,9 @@ MEM_STATIC void ZSTD_updatePrice(seqStore_t* seqStorePtr, U32 litLength, const B
     U32 u;
 
     /* literals */
-    seqStorePtr->litSum += litLength;
+    seqStorePtr->litSum += litLength*ZSTD_LITFREQ_ADD;
     for (u=0; u < litLength; u++)
-        seqStorePtr->litFreq[literals[u]]++;
+        seqStorePtr->litFreq[literals[u]] += ZSTD_LITFREQ_ADD;
 
     /* literal Length */
     {   const BYTE LL_deltaCode = 19;
@@ -171,7 +194,7 @@ MEM_STATIC void ZSTD_updatePrice(seqStore_t* seqStorePtr, U32 litLength, const B
 
 #define SET_PRICE(pos, mlen_, offset_, litlen_, price_)   \
     {                                                 \
-        while (last_pos < pos)  { opt[last_pos+1].price = 1<<30; last_pos++; } \
+        while (last_pos < pos)  { opt[last_pos+1].price = ZSTD_MAX_PRICE; last_pos++; } \
         opt[pos].mlen = mlen_;                         \
         opt[pos].off = offset_;                        \
         opt[pos].litlen = litlen_;                     \
@@ -179,9 +202,23 @@ MEM_STATIC void ZSTD_updatePrice(seqStore_t* seqStorePtr, U32 litLength, const B
     }
 
 
+/* function safe only for comparisons */
+MEM_STATIC U32 ZSTD_readMINMATCH(const void* memPtr, U32 length)
+{
+    switch (length)
+    {
+    default :
+    case 4 : return MEM_read32(memPtr);
+    case 3 : if (MEM_isLittleEndian())
+                return MEM_read32(memPtr)<<8;
+             else
+                return MEM_read32(memPtr)>>8;
+    }
+}
+
 
 /* Update hashTable3 up to ip (excluded)
-   Assumption : always within prefix (ie. not within extDict) */
+   Assumption : always within prefix (i.e. not within extDict) */
 FORCE_INLINE
 U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* zc, const BYTE* ip)
 {
@@ -212,12 +249,12 @@ static U32 ZSTD_insertBtAndGetAllMatches (
 {
     const BYTE* const base = zc->base;
     const U32 current = (U32)(ip-base);
-    const U32 hashLog = zc->params.cParams.hashLog;
+    const U32 hashLog = zc->appliedParams.cParams.hashLog;
     const size_t h  = ZSTD_hashPtr(ip, hashLog, mls);
     U32* const hashTable = zc->hashTable;
     U32 matchIndex  = hashTable[h];
     U32* const bt   = zc->chainTable;
-    const U32 btLog = zc->params.cParams.chainLog - 1;
+    const U32 btLog = zc->appliedParams.cParams.chainLog - 1;
     const U32 btMask= (1U << btLog) - 1;
     size_t commonLengthSmaller=0, commonLengthLarger=0;
     const BYTE* const dictBase = zc->dictBase;
@@ -245,7 +282,7 @@ static U32 ZSTD_insertBtAndGetAllMatches (
                 if (match[bestLength] == ip[bestLength]) currentMl = ZSTD_count(ip, match, iLimit);
             } else {
                 match = dictBase + matchIndex3;
-                if (MEM_readMINMATCH(match, MINMATCH) == MEM_readMINMATCH(ip, MINMATCH))    /* assumption : matchIndex3 <= dictLimit-4 (by table construction) */
+                if (ZSTD_readMINMATCH(match, MINMATCH) == ZSTD_readMINMATCH(ip, MINMATCH))    /* assumption : matchIndex3 <= dictLimit-4 (by table construction) */
                     currentMl = ZSTD_count_2segments(ip+MINMATCH, match+MINMATCH, iLimit, dictEnd, prefixStart) + MINMATCH;
             }
 
@@ -338,6 +375,7 @@ static U32 ZSTD_BtGetAllMatches_selectMLS (
     default :
     case 4 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
     case 5 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
+    case 7 :
     case 6 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
     }
 }
@@ -365,6 +403,7 @@ static U32 ZSTD_BtGetAllMatches_selectMLS_extDict (
     default :
     case 4 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
     case 5 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
+    case 7 :
     case 6 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
     }
 }
@@ -375,7 +414,7 @@ static U32 ZSTD_BtGetAllMatches_selectMLS_extDict (
 *********************************/
 FORCE_INLINE
 void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
-                                    const void* src, size_t srcSize)
+                                    const void* src, size_t srcSize, const int ultra)
 {
     seqStore_t* seqStorePtr = &(ctx->seqStore);
     const BYTE* const istart = (const BYTE*)src;
@@ -386,10 +425,10 @@ void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
     const BYTE* const base = ctx->base;
     const BYTE* const prefixStart = base + ctx->dictLimit;
 
-    const U32 maxSearches = 1U << ctx->params.cParams.searchLog;
-    const U32 sufficient_len = ctx->params.cParams.targetLength;
-    const U32 mls = ctx->params.cParams.searchLength;
-    const U32 minMatch = (ctx->params.cParams.searchLength == 3) ? 3 : 4;
+    const U32 maxSearches = 1U << ctx->appliedParams.cParams.searchLog;
+    const U32 sufficient_len = ctx->appliedParams.cParams.targetLength;
+    const U32 mls = ctx->appliedParams.cParams.searchLength;
+    const U32 minMatch = (ctx->appliedParams.cParams.searchLength == 3) ? 3 : 4;
 
     ZSTD_optimal_t* opt = seqStorePtr->priceTable;
     ZSTD_match_t* matches = seqStorePtr->matchTable;
@@ -398,10 +437,9 @@ void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
 
     /* init */
     ctx->nextToUpdate3 = ctx->nextToUpdate;
-    ZSTD_rescaleFreqs(seqStorePtr);
+    ZSTD_rescaleFreqs(seqStorePtr, (const BYTE*)src, srcSize);
     ip += (ip==prefixStart);
     { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=ctx->rep[i]; }
-    inr = ip;
 
     /* Match Loop */
     while (ip < ilimit) {
@@ -414,9 +452,9 @@ void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
         /* check repCode */
         {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
             for (i=(ip == anchor); i<last_i; i++) {
-                const S32 repCur = ((i==ZSTD_REP_MOVE_OPT) && (ip==anchor)) ? (rep[0] - 1) : rep[i];
+                const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : rep[i];
                 if ( (repCur > 0) && (repCur < (S32)(ip-prefixStart))
-                    && (MEM_readMINMATCH(ip, minMatch) == MEM_readMINMATCH(ip - repCur, minMatch))) {
+                    && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repCur, minMatch))) {
                     mlen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repCur, iend) + minMatch;
                     if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
                         best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
@@ -424,7 +462,7 @@ void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
                     }
                     best_off = i - (ip == anchor);
                     do {
-                        price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH);
+                        price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
                         if (mlen > last_pos || price < opt[mlen].price)
                             SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
                         mlen--;
@@ -449,7 +487,7 @@ void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
             mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
             best_mlen = matches[u].len;
             while (mlen <= best_mlen) {
-                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH);
+                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
                 if (mlen > last_pos || price < opt[mlen].price)
                     SET_PRICE(mlen, mlen, matches[u].off, litlen, price);   /* note : macro modifies last_pos */
                 mlen++;
@@ -499,9 +537,9 @@ void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
             best_mlen = minMatch;
             {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
                 for (i=(opt[cur].mlen != 1); i<last_i; i++) {  /* check rep */
-                    const S32 repCur = ((i==ZSTD_REP_MOVE_OPT) && (opt[cur].mlen != 1)) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
+                    const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
                     if ( (repCur > 0) && (repCur < (S32)(inr-prefixStart))
-                       && (MEM_readMINMATCH(inr, minMatch) == MEM_readMINMATCH(inr - repCur, minMatch))) {
+                       && (ZSTD_readMINMATCH(inr, minMatch) == ZSTD_readMINMATCH(inr - repCur, minMatch))) {
                        mlen = (U32)ZSTD_count(inr+minMatch, inr+minMatch - repCur, iend) + minMatch;
 
                        if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
@@ -510,21 +548,20 @@ void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
                        }
 
                        best_off = i - (opt[cur].mlen != 1);
-
-                       if (opt[cur].mlen == 1) {
-                            litlen = opt[cur].litlen;
-                            if (cur > litlen) {
-                                price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, inr-litlen, best_off, mlen - MINMATCH);
-                            } else
-                                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH);
-                        } else {
-                            litlen = 0;
-                            price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, best_off, mlen - MINMATCH);
-                        }
-
                        if (mlen > best_mlen) best_mlen = mlen;
 
                        do {
+                           if (opt[cur].mlen == 1) {
+                                litlen = opt[cur].litlen;
+                                if (cur > litlen) {
+                                    price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
+                                } else
+                                    price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                            } else {
+                                litlen = 0;
+                                price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
+                            }
+
                             if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
                                 SET_PRICE(cur + mlen, mlen, i, litlen, price);
                             mlen--;
@@ -549,12 +586,12 @@ void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
                     if (opt[cur].mlen == 1) {
                         litlen = opt[cur].litlen;
                         if (cur > litlen)
-                            price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH);
+                            price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
                         else
-                            price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH);
+                            price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
                     } else {
                         litlen = 0;
-                        price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH);
+                        price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
                     }
 
                     if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
@@ -600,7 +637,7 @@ _storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
                 offset--;
             } else {
                 if (offset != 0) {
-                    best_off = ((offset==ZSTD_REP_MOVE_OPT) && (litLength==0)) ? (rep[0] - 1) : (rep[offset]);
+                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
                     if (offset != 1) rep[2] = rep[1];
                     rep[1] = rep[0];
                     rep[0] = best_off;
@@ -614,7 +651,7 @@ _storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
     }    }   /* for (cur=0; cur < last_pos; ) */
 
     /* Save reps for next block */
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) ctx->savedRep[i] = rep[i]; }
+    { int i; for (i=0; i<ZSTD_REP_NUM; i++) ctx->repToConfirm[i] = rep[i]; }
 
     /* Last Literals */
     {   size_t const lastLLSize = iend - anchor;
@@ -626,7 +663,7 @@ _storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
 
 FORCE_INLINE
 void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
-                                     const void* src, size_t srcSize)
+                                     const void* src, size_t srcSize, const int ultra)
 {
     seqStore_t* seqStorePtr = &(ctx->seqStore);
     const BYTE* const istart = (const BYTE*)src;
@@ -641,10 +678,10 @@ void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
     const BYTE* const dictBase = ctx->dictBase;
     const BYTE* const dictEnd  = dictBase + dictLimit;
 
-    const U32 maxSearches = 1U << ctx->params.cParams.searchLog;
-    const U32 sufficient_len = ctx->params.cParams.targetLength;
-    const U32 mls = ctx->params.cParams.searchLength;
-    const U32 minMatch = (ctx->params.cParams.searchLength == 3) ? 3 : 4;
+    const U32 maxSearches = 1U << ctx->appliedParams.cParams.searchLog;
+    const U32 sufficient_len = ctx->appliedParams.cParams.targetLength;
+    const U32 mls = ctx->appliedParams.cParams.searchLength;
+    const U32 minMatch = (ctx->appliedParams.cParams.searchLength == 3) ? 3 : 4;
 
     ZSTD_optimal_t* opt = seqStorePtr->priceTable;
     ZSTD_match_t* matches = seqStorePtr->matchTable;
@@ -655,9 +692,8 @@ void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
     { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=ctx->rep[i]; }
 
     ctx->nextToUpdate3 = ctx->nextToUpdate;
-    ZSTD_rescaleFreqs(seqStorePtr);
+    ZSTD_rescaleFreqs(seqStorePtr, (const BYTE*)src, srcSize);
     ip += (ip==prefixStart);
-    inr = ip;
 
     /* Match Loop */
     while (ip < ilimit) {
@@ -666,19 +702,18 @@ void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
         U32 current = (U32)(ip-base);
         memset(opt, 0, sizeof(ZSTD_optimal_t));
         last_pos = 0;
-        inr = ip;
         opt[0].litlen = (U32)(ip - anchor);
 
         /* check repCode */
         {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
             for (i = (ip==anchor); i<last_i; i++) {
-                const S32 repCur = ((i==ZSTD_REP_MOVE_OPT) && (ip==anchor)) ? (rep[0] - 1) : rep[i];
+                const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : rep[i];
                 const U32 repIndex = (U32)(current - repCur);
                 const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
                 const BYTE* const repMatch = repBase + repIndex;
                 if ( (repCur > 0 && repCur <= (S32)current)
                    && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
-                   && (MEM_readMINMATCH(ip, minMatch) == MEM_readMINMATCH(repMatch, minMatch)) ) {
+                   && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
                     /* repcode detected we should take it */
                     const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
                     mlen = (U32)ZSTD_count_2segments(ip+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
@@ -691,7 +726,7 @@ void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
                     best_off = i - (ip==anchor);
                     litlen = opt[0].litlen;
                     do {
-                        price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH);
+                        price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
                         if (mlen > last_pos || price < opt[mlen].price)
                             SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
                         mlen--;
@@ -721,7 +756,7 @@ void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
             best_mlen = matches[u].len;
             litlen = opt[0].litlen;
             while (mlen <= best_mlen) {
-                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH);
+                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
                 if (mlen > last_pos || price < opt[mlen].price)
                     SET_PRICE(mlen, mlen, matches[u].off, litlen, price);
                 mlen++;
@@ -765,17 +800,16 @@ void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
                 opt[cur].rep[0] = ((opt[cur].off==ZSTD_REP_MOVE_OPT) && (mlen != 1)) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
             }
 
-            best_mlen = 0;
-
+            best_mlen = minMatch;
             {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
                 for (i = (mlen != 1); i<last_i; i++) {
-                    const S32 repCur = ((i==ZSTD_REP_MOVE_OPT) && (opt[cur].mlen != 1)) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
+                    const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
                     const U32 repIndex = (U32)(current+cur - repCur);
                     const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
                     const BYTE* const repMatch = repBase + repIndex;
                     if ( (repCur > 0 && repCur <= (S32)(current+cur))
                       && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
-                      && (MEM_readMINMATCH(inr, minMatch) == MEM_readMINMATCH(repMatch, minMatch)) ) {
+                      && (ZSTD_readMINMATCH(inr, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
                         /* repcode detected */
                         const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
                         mlen = (U32)ZSTD_count_2segments(inr+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
@@ -786,20 +820,20 @@ void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
                         }
 
                         best_off = i - (opt[cur].mlen != 1);
+                        if (mlen > best_mlen) best_mlen = mlen;
+
+                        do {
                             if (opt[cur].mlen == 1) {
                                 litlen = opt[cur].litlen;
                                 if (cur > litlen) {
-                                price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, inr-litlen, best_off, mlen - MINMATCH);
+                                    price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
                                 } else
-                                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH);
+                                    price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
                             } else {
                                 litlen = 0;
-                            price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, best_off, mlen - MINMATCH);
+                                price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
                             }
 
-                        best_mlen = mlen;
-
-                        do {
                             if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
                                 SET_PRICE(cur + mlen, mlen, i, litlen, price);
                             mlen--;
@@ -808,30 +842,28 @@ void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
 
             match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, inr, iend, maxSearches, mls, matches, minMatch);
 
-            if (match_num > 0 && matches[match_num-1].len > sufficient_len) {
+            if (match_num > 0 && (matches[match_num-1].len > sufficient_len || cur + matches[match_num-1].len >= ZSTD_OPT_NUM)) {
                 best_mlen = matches[match_num-1].len;
                 best_off = matches[match_num-1].off;
                 last_pos = cur + 1;
                 goto _storeSequence;
             }
 
-            best_mlen = (best_mlen > minMatch) ? best_mlen : minMatch;
-
             /* set prices using matches at position = cur */
             for (u = 0; u < match_num; u++) {
                 mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-                best_mlen = (cur + matches[u].len < ZSTD_OPT_NUM) ? matches[u].len : ZSTD_OPT_NUM - cur;
+                best_mlen = matches[u].len;
 
                 while (mlen <= best_mlen) {
                     if (opt[cur].mlen == 1) {
                         litlen = opt[cur].litlen;
                         if (cur > litlen)
-                            price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH);
+                            price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
                         else
-                            price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH);
+                            price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
                     } else {
                         litlen = 0;
-                        price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH);
+                        price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
                     }
 
                     if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
@@ -877,7 +909,7 @@ _storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
                 offset--;
             } else {
                 if (offset != 0) {
-                    best_off = ((offset==ZSTD_REP_MOVE_OPT) && (litLength==0)) ? (rep[0] - 1) : (rep[offset]);
+                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
                     if (offset != 1) rep[2] = rep[1];
                     rep[1] = rep[0];
                     rep[0] = best_off;
@@ -892,7 +924,7 @@ _storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
     }    }   /* for (cur=0; cur < last_pos; ) */
 
     /* Save reps for next block */
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) ctx->savedRep[i] = rep[i]; }
+    { int i; for (i=0; i<ZSTD_REP_NUM; i++) ctx->repToConfirm[i] = rep[i]; }
 
     /* Last Literals */
     {   size_t lastLLSize = iend - anchor;

contrib/libzstd/include/zstd/compress/zstdmt_compress.c

@@ -0,0 +1,955 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+/* ======   Tuning parameters   ====== */
+#define ZSTDMT_NBTHREADS_MAX 128
+
+
+/* ======   Compiler specifics   ====== */
+#if defined(_MSC_VER)
+#  pragma warning(disable : 4204)   /* disable: C4204: non-constant aggregate initializer */
+#endif
+
+
+/* ======   Dependencies   ====== */
+#include <string.h>      /* memcpy, memset */
+#include "pool.h"        /* threadpool */
+#include "threading.h"   /* mutex */
+#include "zstd_internal.h"  /* MIN, ERROR, ZSTD_*, ZSTD_highbit32 */
+#include "zstdmt_compress.h"
+
+
+/* ======   Debug   ====== */
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=2)
+
+#  include <stdio.h>
+#  include <unistd.h>
+#  include <sys/times.h>
+#  define DEBUGLOGRAW(l, ...) if (l<=ZSTD_DEBUG) { fprintf(stderr, __VA_ARGS__); }
+
+#  define DEBUG_PRINTHEX(l,p,n) {            \
+    unsigned debug_u;                        \
+    for (debug_u=0; debug_u<(n); debug_u++)  \
+        DEBUGLOGRAW(l, "%02X ", ((const unsigned char*)(p))[debug_u]); \
+    DEBUGLOGRAW(l, " \n");                   \
+}
+
+static unsigned long long GetCurrentClockTimeMicroseconds(void)
+{
+   static clock_t _ticksPerSecond = 0;
+   if (_ticksPerSecond <= 0) _ticksPerSecond = sysconf(_SC_CLK_TCK);
+
+   { struct tms junk; clock_t newTicks = (clock_t) times(&junk);
+     return ((((unsigned long long)newTicks)*(1000000))/_ticksPerSecond); }
+}
+
+#define MUTEX_WAIT_TIME_DLEVEL 6
+#define PTHREAD_MUTEX_LOCK(mutex) {               \
+    if (ZSTD_DEBUG>=MUTEX_WAIT_TIME_DLEVEL) {   \
+        unsigned long long const beforeTime = GetCurrentClockTimeMicroseconds(); \
+        pthread_mutex_lock(mutex);                \
+        {   unsigned long long const afterTime = GetCurrentClockTimeMicroseconds(); \
+            unsigned long long const elapsedTime = (afterTime-beforeTime); \
+            if (elapsedTime > 1000) {  /* or whatever threshold you like; I'm using 1 millisecond here */ \
+                DEBUGLOG(MUTEX_WAIT_TIME_DLEVEL, "Thread took %llu microseconds to acquire mutex %s \n", \
+                   elapsedTime, #mutex);          \
+        }   }                                     \
+    } else pthread_mutex_lock(mutex);             \
+}
+
+#else
+
+#  define PTHREAD_MUTEX_LOCK(m) pthread_mutex_lock(m)
+#  define DEBUG_PRINTHEX(l,p,n) {}
+
+#endif
+
+
+/* =====   Buffer Pool   ===== */
+
+typedef struct buffer_s {
+    void* start;
+    size_t size;
+} buffer_t;
+
+static const buffer_t g_nullBuffer = { NULL, 0 };
+
+typedef struct ZSTDMT_bufferPool_s {
+    unsigned totalBuffers;
+    unsigned nbBuffers;
+    ZSTD_customMem cMem;
+    buffer_t bTable[1];   /* variable size */
+} ZSTDMT_bufferPool;
+
+static ZSTDMT_bufferPool* ZSTDMT_createBufferPool(unsigned nbThreads, ZSTD_customMem cMem)
+{
+    unsigned const maxNbBuffers = 2*nbThreads + 2;
+    ZSTDMT_bufferPool* const bufPool = (ZSTDMT_bufferPool*)ZSTD_calloc(
+        sizeof(ZSTDMT_bufferPool) + (maxNbBuffers-1) * sizeof(buffer_t), cMem);
+    if (bufPool==NULL) return NULL;
+    bufPool->totalBuffers = maxNbBuffers;
+    bufPool->nbBuffers = 0;
+    bufPool->cMem = cMem;
+    return bufPool;
+}
+
+static void ZSTDMT_freeBufferPool(ZSTDMT_bufferPool* bufPool)
+{
+    unsigned u;
+    if (!bufPool) return;   /* compatibility with free on NULL */
+    for (u=0; u<bufPool->totalBuffers; u++)
+        ZSTD_free(bufPool->bTable[u].start, bufPool->cMem);
+    ZSTD_free(bufPool, bufPool->cMem);
+}
+
+/* only works at initialization, not during compression */
+static size_t ZSTDMT_sizeof_bufferPool(ZSTDMT_bufferPool* bufPool)
+{
+    size_t const poolSize = sizeof(*bufPool)
+                            + (bufPool->totalBuffers - 1) * sizeof(buffer_t);
+    unsigned u;
+    size_t totalBufferSize = 0;
+    for (u=0; u<bufPool->totalBuffers; u++)
+        totalBufferSize += bufPool->bTable[u].size;
+
+    return poolSize + totalBufferSize;
+}
+
+/** ZSTDMT_getBuffer() :
+ *  assumption : invocation from main thread only ! */
+static buffer_t ZSTDMT_getBuffer(ZSTDMT_bufferPool* pool, size_t bSize)
+{
+    if (pool->nbBuffers) {   /* try to use an existing buffer */
+        buffer_t const buf = pool->bTable[--(pool->nbBuffers)];
+        size_t const availBufferSize = buf.size;
+        if ((availBufferSize >= bSize) & (availBufferSize <= 10*bSize))
+            /* large enough, but not too much */
+            return buf;
+        /* size conditions not respected : scratch this buffer, create new one */
+        ZSTD_free(buf.start, pool->cMem);
+    }
+    /* create new buffer */
+    {   buffer_t buffer;
+        void* const start = ZSTD_malloc(bSize, pool->cMem);
+        if (start==NULL) bSize = 0;
+        buffer.start = start;   /* note : start can be NULL if malloc fails ! */
+        buffer.size = bSize;
+        return buffer;
+    }
+}
+
+/* store buffer for later re-use, up to pool capacity */
+static void ZSTDMT_releaseBuffer(ZSTDMT_bufferPool* pool, buffer_t buf)
+{
+    if (buf.start == NULL) return;   /* release on NULL */
+    if (pool->nbBuffers < pool->totalBuffers) {
+        pool->bTable[pool->nbBuffers++] = buf;   /* store for later re-use */
+        return;
+    }
+    /* Reached bufferPool capacity (should not happen) */
+    ZSTD_free(buf.start, pool->cMem);
+}
+
+
+/* =====   CCtx Pool   ===== */
+
+typedef struct {
+    unsigned totalCCtx;
+    unsigned availCCtx;
+    ZSTD_customMem cMem;
+    ZSTD_CCtx* cctx[1];   /* variable size */
+} ZSTDMT_CCtxPool;
+
+/* assumption : CCtxPool invocation only from main thread */
+
+/* note : all CCtx borrowed from the pool should be released back to the pool _before_ freeing the pool */
+static void ZSTDMT_freeCCtxPool(ZSTDMT_CCtxPool* pool)
+{
+    unsigned u;
+    for (u=0; u<pool->totalCCtx; u++)
+        ZSTD_freeCCtx(pool->cctx[u]);  /* note : compatible with free on NULL */
+    ZSTD_free(pool, pool->cMem);
+}
+
+/* ZSTDMT_createCCtxPool() :
+ * implies nbThreads >= 1 , checked by caller ZSTDMT_createCCtx() */
+static ZSTDMT_CCtxPool* ZSTDMT_createCCtxPool(unsigned nbThreads,
+                                              ZSTD_customMem cMem)
+{
+    ZSTDMT_CCtxPool* const cctxPool = (ZSTDMT_CCtxPool*) ZSTD_calloc(
+        sizeof(ZSTDMT_CCtxPool) + (nbThreads-1)*sizeof(ZSTD_CCtx*), cMem);
+    if (!cctxPool) return NULL;
+    cctxPool->cMem = cMem;
+    cctxPool->totalCCtx = nbThreads;
+    cctxPool->availCCtx = 1;   /* at least one cctx for single-thread mode */
+    cctxPool->cctx[0] = ZSTD_createCCtx_advanced(cMem);
+    if (!cctxPool->cctx[0]) { ZSTDMT_freeCCtxPool(cctxPool); return NULL; }
+    DEBUGLOG(3, "cctxPool created, with %u threads", nbThreads);
+    return cctxPool;
+}
+
+/* only works during initialization phase, not during compression */
+static size_t ZSTDMT_sizeof_CCtxPool(ZSTDMT_CCtxPool* cctxPool)
+{
+    unsigned const nbThreads = cctxPool->totalCCtx;
+    size_t const poolSize = sizeof(*cctxPool)
+                            + (nbThreads-1)*sizeof(ZSTD_CCtx*);
+    unsigned u;
+    size_t totalCCtxSize = 0;
+    for (u=0; u<nbThreads; u++)
+        totalCCtxSize += ZSTD_sizeof_CCtx(cctxPool->cctx[u]);
+
+    return poolSize + totalCCtxSize;
+}
+
+static ZSTD_CCtx* ZSTDMT_getCCtx(ZSTDMT_CCtxPool* pool)
+{
+    if (pool->availCCtx) {
+        pool->availCCtx--;
+        return pool->cctx[pool->availCCtx];
+    }
+    return ZSTD_createCCtx();   /* note : can be NULL, when creation fails ! */
+}
+
+static void ZSTDMT_releaseCCtx(ZSTDMT_CCtxPool* pool, ZSTD_CCtx* cctx)
+{
+    if (cctx==NULL) return;   /* compatibility with release on NULL */
+    if (pool->availCCtx < pool->totalCCtx)
+        pool->cctx[pool->availCCtx++] = cctx;
+    else
+        /* pool overflow : should not happen, since totalCCtx==nbThreads */
+        ZSTD_freeCCtx(cctx);
+}
+
+
+/* =====   Thread worker   ===== */
+
+typedef struct {
+    buffer_t buffer;
+    size_t filled;
+} inBuff_t;
+
+typedef struct {
+    ZSTD_CCtx* cctx;
+    buffer_t src;
+    const void* srcStart;
+    size_t   srcSize;
+    size_t   dictSize;
+    buffer_t dstBuff;
+    size_t   cSize;
+    size_t   dstFlushed;
+    unsigned firstChunk;
+    unsigned lastChunk;
+    unsigned jobCompleted;
+    unsigned jobScanned;
+    pthread_mutex_t* jobCompleted_mutex;
+    pthread_cond_t* jobCompleted_cond;
+    ZSTD_parameters params;
+    const ZSTD_CDict* cdict;
+    unsigned long long fullFrameSize;
+} ZSTDMT_jobDescription;
+
+/* ZSTDMT_compressChunk() : POOL_function type */
+void ZSTDMT_compressChunk(void* jobDescription)
+{
+    ZSTDMT_jobDescription* const job = (ZSTDMT_jobDescription*)jobDescription;
+    const void* const src = (const char*)job->srcStart + job->dictSize;
+    buffer_t const dstBuff = job->dstBuff;
+    DEBUGLOG(5, "job (first:%u) (last:%u) : dictSize %u, srcSize %u",
+                 job->firstChunk, job->lastChunk, (U32)job->dictSize, (U32)job->srcSize);
+    if (job->cdict) {  /* should only happen for first segment */
+        size_t const initError = ZSTD_compressBegin_usingCDict_advanced(job->cctx, job->cdict, job->params.fParams, job->fullFrameSize);
+        DEBUGLOG(5, "using CDict");
+        if (ZSTD_isError(initError)) { job->cSize = initError; goto _endJob; }
+    } else {  /* srcStart points at reloaded section */
+        if (!job->firstChunk) job->params.fParams.contentSizeFlag = 0;  /* ensure no srcSize control */
+        {   size_t const dictModeError = ZSTD_setCCtxParameter(job->cctx, ZSTD_p_forceRawDict, 1);  /* Force loading dictionary in "content-only" mode (no header analysis) */
+            size_t const initError = ZSTD_compressBegin_advanced(job->cctx, job->srcStart, job->dictSize, job->params, job->fullFrameSize);
+            if (ZSTD_isError(initError) || ZSTD_isError(dictModeError)) { job->cSize = initError; goto _endJob; }
+            ZSTD_setCCtxParameter(job->cctx, ZSTD_p_forceWindow, 1);
+    }   }
+    if (!job->firstChunk) {  /* flush and overwrite frame header when it's not first segment */
+        size_t const hSize = ZSTD_compressContinue(job->cctx, dstBuff.start, dstBuff.size, src, 0);
+        if (ZSTD_isError(hSize)) { job->cSize = hSize; goto _endJob; }
+        ZSTD_invalidateRepCodes(job->cctx);
+    }
+
+    DEBUGLOG(5, "Compressing : ");
+    DEBUG_PRINTHEX(4, job->srcStart, 12);
+    job->cSize = (job->lastChunk) ?
+                 ZSTD_compressEnd     (job->cctx, dstBuff.start, dstBuff.size, src, job->srcSize) :
+                 ZSTD_compressContinue(job->cctx, dstBuff.start, dstBuff.size, src, job->srcSize);
+    DEBUGLOG(5, "compressed %u bytes into %u bytes   (first:%u) (last:%u)",
+                (unsigned)job->srcSize, (unsigned)job->cSize, job->firstChunk, job->lastChunk);
+    DEBUGLOG(5, "dstBuff.size : %u ; => %s", (U32)dstBuff.size, ZSTD_getErrorName(job->cSize));
+
+_endJob:
+    PTHREAD_MUTEX_LOCK(job->jobCompleted_mutex);
+    job->jobCompleted = 1;
+    job->jobScanned = 0;
+    pthread_cond_signal(job->jobCompleted_cond);
+    pthread_mutex_unlock(job->jobCompleted_mutex);
+}
+
+
+/* ------------------------------------------ */
+/* =====   Multi-threaded compression   ===== */
+/* ------------------------------------------ */
+
+struct ZSTDMT_CCtx_s {
+    POOL_ctx* factory;
+    ZSTDMT_jobDescription* jobs;
+    ZSTDMT_bufferPool* buffPool;
+    ZSTDMT_CCtxPool* cctxPool;
+    pthread_mutex_t jobCompleted_mutex;
+    pthread_cond_t jobCompleted_cond;
+    size_t targetSectionSize;
+    size_t marginSize;
+    size_t inBuffSize;
+    size_t dictSize;
+    size_t targetDictSize;
+    inBuff_t inBuff;
+    ZSTD_parameters params;
+    XXH64_state_t xxhState;
+    unsigned nbThreads;
+    unsigned jobIDMask;
+    unsigned doneJobID;
+    unsigned nextJobID;
+    unsigned frameEnded;
+    unsigned allJobsCompleted;
+    unsigned overlapRLog;
+    unsigned long long frameContentSize;
+    size_t sectionSize;
+    ZSTD_customMem cMem;
+    ZSTD_CDict* cdictLocal;
+    const ZSTD_CDict* cdict;
+};
+
+static ZSTDMT_jobDescription* ZSTDMT_allocJobsTable(U32* nbJobsPtr, ZSTD_customMem cMem)
+{
+    U32 const nbJobsLog2 = ZSTD_highbit32(*nbJobsPtr) + 1;
+    U32 const nbJobs = 1 << nbJobsLog2;
+    *nbJobsPtr = nbJobs;
+    return (ZSTDMT_jobDescription*) ZSTD_calloc(
+                            nbJobs * sizeof(ZSTDMT_jobDescription), cMem);
+}
+
+ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbThreads, ZSTD_customMem cMem)
+{
+    ZSTDMT_CCtx* mtctx;
+    U32 nbJobs = nbThreads + 2;
+    DEBUGLOG(3, "ZSTDMT_createCCtx_advanced");
+
+    if ((nbThreads < 1) | (nbThreads > ZSTDMT_NBTHREADS_MAX)) return NULL;
+    if ((cMem.customAlloc!=NULL) ^ (cMem.customFree!=NULL))
+        /* invalid custom allocator */
+        return NULL;
+
+    mtctx = (ZSTDMT_CCtx*) ZSTD_calloc(sizeof(ZSTDMT_CCtx), cMem);
+    if (!mtctx) return NULL;
+    mtctx->cMem = cMem;
+    mtctx->nbThreads = nbThreads;
+    mtctx->allJobsCompleted = 1;
+    mtctx->sectionSize = 0;
+    mtctx->overlapRLog = 3;
+    mtctx->factory = POOL_create(nbThreads, 1);
+    mtctx->jobs = ZSTDMT_allocJobsTable(&nbJobs, cMem);
+    mtctx->jobIDMask = nbJobs - 1;
+    mtctx->buffPool = ZSTDMT_createBufferPool(nbThreads, cMem);
+    mtctx->cctxPool = ZSTDMT_createCCtxPool(nbThreads, cMem);
+    if (!mtctx->factory | !mtctx->jobs | !mtctx->buffPool | !mtctx->cctxPool) {
+        ZSTDMT_freeCCtx(mtctx);
+        return NULL;
+    }
+    pthread_mutex_init(&mtctx->jobCompleted_mutex, NULL);   /* Todo : check init function return */
+    pthread_cond_init(&mtctx->jobCompleted_cond, NULL);
+    DEBUGLOG(3, "mt_cctx created, for %u threads", nbThreads);
+    return mtctx;
+}
+
+ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbThreads)
+{
+    return ZSTDMT_createCCtx_advanced(nbThreads, ZSTD_defaultCMem);
+}
+
+/* ZSTDMT_releaseAllJobResources() :
+ * note : ensure all workers are killed first ! */
+static void ZSTDMT_releaseAllJobResources(ZSTDMT_CCtx* mtctx)
+{
+    unsigned jobID;
+    DEBUGLOG(3, "ZSTDMT_releaseAllJobResources");
+    for (jobID=0; jobID <= mtctx->jobIDMask; jobID++) {
+        ZSTDMT_releaseBuffer(mtctx->buffPool, mtctx->jobs[jobID].dstBuff);
+        mtctx->jobs[jobID].dstBuff = g_nullBuffer;
+        ZSTDMT_releaseBuffer(mtctx->buffPool, mtctx->jobs[jobID].src);
+        mtctx->jobs[jobID].src = g_nullBuffer;
+        ZSTDMT_releaseCCtx(mtctx->cctxPool, mtctx->jobs[jobID].cctx);
+        mtctx->jobs[jobID].cctx = NULL;
+    }
+    memset(mtctx->jobs, 0, (mtctx->jobIDMask+1)*sizeof(ZSTDMT_jobDescription));
+    ZSTDMT_releaseBuffer(mtctx->buffPool, mtctx->inBuff.buffer);
+    mtctx->inBuff.buffer = g_nullBuffer;
+    mtctx->allJobsCompleted = 1;
+}
+
+size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx)
+{
+    if (mtctx==NULL) return 0;   /* compatible with free on NULL */
+    POOL_free(mtctx->factory);
+    if (!mtctx->allJobsCompleted) ZSTDMT_releaseAllJobResources(mtctx); /* stop workers first */
+    ZSTDMT_freeBufferPool(mtctx->buffPool);  /* release job resources into pools first */
+    ZSTD_free(mtctx->jobs, mtctx->cMem);
+    ZSTDMT_freeCCtxPool(mtctx->cctxPool);
+    ZSTD_freeCDict(mtctx->cdictLocal);
+    pthread_mutex_destroy(&mtctx->jobCompleted_mutex);
+    pthread_cond_destroy(&mtctx->jobCompleted_cond);
+    ZSTD_free(mtctx, mtctx->cMem);
+    return 0;
+}
+
+size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx)
+{
+    if (mtctx == NULL) return 0;   /* supports sizeof NULL */
+    return sizeof(*mtctx)
+        + POOL_sizeof(mtctx->factory)
+        + ZSTDMT_sizeof_bufferPool(mtctx->buffPool)
+        + (mtctx->jobIDMask+1) * sizeof(ZSTDMT_jobDescription)
+        + ZSTDMT_sizeof_CCtxPool(mtctx->cctxPool)
+        + ZSTD_sizeof_CDict(mtctx->cdictLocal);
+}
+
+size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSDTMT_parameter parameter, unsigned value)
+{
+    switch(parameter)
+    {
+    case ZSTDMT_p_sectionSize :
+        mtctx->sectionSize = value;
+        return 0;
+    case ZSTDMT_p_overlapSectionLog :
+        DEBUGLOG(5, "ZSTDMT_p_overlapSectionLog : %u", value);
+        mtctx->overlapRLog = (value >= 9) ? 0 : 9 - value;
+        return 0;
+    default :
+        return ERROR(compressionParameter_unsupported);
+    }
+}
+
+
+/* ------------------------------------------ */
+/* =====   Multi-threaded compression   ===== */
+/* ------------------------------------------ */
+
+static unsigned computeNbChunks(size_t srcSize, unsigned windowLog, unsigned nbThreads) {
+    size_t const chunkSizeTarget = (size_t)1 << (windowLog + 2);
+    size_t const chunkMaxSize = chunkSizeTarget << 2;
+    size_t const passSizeMax = chunkMaxSize * nbThreads;
+    unsigned const multiplier = (unsigned)(srcSize / passSizeMax) + 1;
+    unsigned const nbChunksLarge = multiplier * nbThreads;
+    unsigned const nbChunksMax = (unsigned)(srcSize / chunkSizeTarget) + 1;
+    unsigned const nbChunksSmall = MIN(nbChunksMax, nbThreads);
+    return (multiplier>1) ? nbChunksLarge : nbChunksSmall;
+}
+
+
+size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
+                           void* dst, size_t dstCapacity,
+                     const void* src, size_t srcSize,
+                     const ZSTD_CDict* cdict,
+                           ZSTD_parameters const params,
+                           unsigned overlapRLog)
+{
+    size_t const overlapSize = (overlapRLog>=9) ? 0 : (size_t)1 << (params.cParams.windowLog - overlapRLog);
+    unsigned nbChunks = computeNbChunks(srcSize, params.cParams.windowLog, mtctx->nbThreads);
+    size_t const proposedChunkSize = (srcSize + (nbChunks-1)) / nbChunks;
+    size_t const avgChunkSize = ((proposedChunkSize & 0x1FFFF) < 0x7FFF) ? proposedChunkSize + 0xFFFF : proposedChunkSize;   /* avoid too small last block */
+    const char* const srcStart = (const char*)src;
+    size_t remainingSrcSize = srcSize;
+    unsigned const compressWithinDst = (dstCapacity >= ZSTD_compressBound(srcSize)) ? nbChunks : (unsigned)(dstCapacity / ZSTD_compressBound(avgChunkSize));  /* presumes avgChunkSize >= 256 KB, which should be the case */
+    size_t frameStartPos = 0, dstBufferPos = 0;
+
+    DEBUGLOG(4, "nbChunks  : %2u   (chunkSize : %u bytes)   ", nbChunks, (U32)avgChunkSize);
+    if (nbChunks==1) {   /* fallback to single-thread mode */
+        ZSTD_CCtx* const cctx = mtctx->cctxPool->cctx[0];
+        if (cdict) return ZSTD_compress_usingCDict_advanced(cctx, dst, dstCapacity, src, srcSize, cdict, params.fParams);
+        return ZSTD_compress_advanced(cctx, dst, dstCapacity, src, srcSize, NULL, 0, params);
+    }
+    assert(avgChunkSize >= 256 KB);  /* condition for ZSTD_compressBound(A) + ZSTD_compressBound(B) <= ZSTD_compressBound(A+B), which is useful to avoid allocating extra buffers */
+
+    if (nbChunks > mtctx->jobIDMask+1) {  /* enlarge job table */
+        U32 nbJobs = nbChunks;
+        ZSTD_free(mtctx->jobs, mtctx->cMem);
+        mtctx->jobIDMask = 0;
+        mtctx->jobs = ZSTDMT_allocJobsTable(&nbJobs, mtctx->cMem);
+        if (mtctx->jobs==NULL) return ERROR(memory_allocation);
+        mtctx->jobIDMask = nbJobs - 1;
+    }
+
+    {   unsigned u;
+        for (u=0; u<nbChunks; u++) {
+            size_t const chunkSize = MIN(remainingSrcSize, avgChunkSize);
+            size_t const dstBufferCapacity = ZSTD_compressBound(chunkSize);
+            buffer_t const dstAsBuffer = { (char*)dst + dstBufferPos, dstBufferCapacity };
+            buffer_t const dstBuffer = u < compressWithinDst ? dstAsBuffer : ZSTDMT_getBuffer(mtctx->buffPool, dstBufferCapacity);
+            ZSTD_CCtx* const cctx = ZSTDMT_getCCtx(mtctx->cctxPool);
+            size_t dictSize = u ? overlapSize : 0;
+
+            if ((cctx==NULL) || (dstBuffer.start==NULL)) {
+                mtctx->jobs[u].cSize = ERROR(memory_allocation);   /* job result */
+                mtctx->jobs[u].jobCompleted = 1;
+                nbChunks = u+1;   /* only wait and free u jobs, instead of initially expected nbChunks ones */
+                break;   /* let's wait for previous jobs to complete, but don't start new ones */
+            }
+
+            mtctx->jobs[u].srcStart = srcStart + frameStartPos - dictSize;
+            mtctx->jobs[u].dictSize = dictSize;
+            mtctx->jobs[u].srcSize = chunkSize;
+            mtctx->jobs[u].cdict = mtctx->nextJobID==0 ? cdict : NULL;
+            mtctx->jobs[u].fullFrameSize = srcSize;
+            mtctx->jobs[u].params = params;
+            /* do not calculate checksum within sections, but write it in header for first section */
+            if (u!=0) mtctx->jobs[u].params.fParams.checksumFlag = 0;
+            mtctx->jobs[u].dstBuff = dstBuffer;
+            mtctx->jobs[u].cctx = cctx;
+            mtctx->jobs[u].firstChunk = (u==0);
+            mtctx->jobs[u].lastChunk = (u==nbChunks-1);
+            mtctx->jobs[u].jobCompleted = 0;
+            mtctx->jobs[u].jobCompleted_mutex = &mtctx->jobCompleted_mutex;
+            mtctx->jobs[u].jobCompleted_cond = &mtctx->jobCompleted_cond;
+
+            DEBUGLOG(5, "posting job %u   (%u bytes)", u, (U32)chunkSize);
+            DEBUG_PRINTHEX(6, mtctx->jobs[u].srcStart, 12);
+            POOL_add(mtctx->factory, ZSTDMT_compressChunk, &mtctx->jobs[u]);
+
+            frameStartPos += chunkSize;
+            dstBufferPos += dstBufferCapacity;
+            remainingSrcSize -= chunkSize;
+    }   }
+
+    /* collect result */
+    {   unsigned chunkID;
+        size_t error = 0, dstPos = 0;
+        for (chunkID=0; chunkID<nbChunks; chunkID++) {
+            DEBUGLOG(5, "waiting for chunk %u ", chunkID);
+            PTHREAD_MUTEX_LOCK(&mtctx->jobCompleted_mutex);
+            while (mtctx->jobs[chunkID].jobCompleted==0) {
+                DEBUGLOG(5, "waiting for jobCompleted signal from chunk %u", chunkID);
+                pthread_cond_wait(&mtctx->jobCompleted_cond, &mtctx->jobCompleted_mutex);
+            }
+            pthread_mutex_unlock(&mtctx->jobCompleted_mutex);
+            DEBUGLOG(5, "ready to write chunk %u ", chunkID);
+
+            ZSTDMT_releaseCCtx(mtctx->cctxPool, mtctx->jobs[chunkID].cctx);
+            mtctx->jobs[chunkID].cctx = NULL;
+            mtctx->jobs[chunkID].srcStart = NULL;
+            {   size_t const cSize = mtctx->jobs[chunkID].cSize;
+                if (ZSTD_isError(cSize)) error = cSize;
+                if ((!error) && (dstPos + cSize > dstCapacity)) error = ERROR(dstSize_tooSmall);
+                if (chunkID) {   /* note : chunk 0 is written directly at dst, which is correct position */
+                    if (!error)
+                        memmove((char*)dst + dstPos, mtctx->jobs[chunkID].dstBuff.start, cSize);  /* may overlap when chunk compressed within dst */
+                    if (chunkID >= compressWithinDst) {  /* chunk compressed into its own buffer, which must be released */
+                        DEBUGLOG(5, "releasing buffer %u>=%u", chunkID, compressWithinDst);
+                        ZSTDMT_releaseBuffer(mtctx->buffPool, mtctx->jobs[chunkID].dstBuff);
+                    }
+                    mtctx->jobs[chunkID].dstBuff = g_nullBuffer;
+                }
+                dstPos += cSize ;
+            }
+        }
+        if (!error) DEBUGLOG(4, "compressed size : %u  ", (U32)dstPos);
+        return error ? error : dstPos;
+    }
+}
+
+
+size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
+                           void* dst, size_t dstCapacity,
+                     const void* src, size_t srcSize,
+                           int compressionLevel)
+{
+    U32 const overlapRLog = (compressionLevel >= ZSTD_maxCLevel()) ? 0 : 3;
+    ZSTD_parameters params = ZSTD_getParams(compressionLevel, srcSize, 0);
+    params.fParams.contentSizeFlag = 1;
+    return ZSTDMT_compress_advanced(mtctx, dst, dstCapacity, src, srcSize, NULL, params, overlapRLog);
+}
+
+
+/* ====================================== */
+/* =======      Streaming API     ======= */
+/* ====================================== */
+
+static void ZSTDMT_waitForAllJobsCompleted(ZSTDMT_CCtx* zcs)
+{
+    DEBUGLOG(4, "ZSTDMT_waitForAllJobsCompleted");
+    while (zcs->doneJobID < zcs->nextJobID) {
+        unsigned const jobID = zcs->doneJobID & zcs->jobIDMask;
+        PTHREAD_MUTEX_LOCK(&zcs->jobCompleted_mutex);
+        while (zcs->jobs[jobID].jobCompleted==0) {
+            DEBUGLOG(5, "waiting for jobCompleted signal from chunk %u", zcs->doneJobID);   /* we want to block when waiting for data to flush */
+            pthread_cond_wait(&zcs->jobCompleted_cond, &zcs->jobCompleted_mutex);
+        }
+        pthread_mutex_unlock(&zcs->jobCompleted_mutex);
+        zcs->doneJobID++;
+    }
+}
+
+
+/** ZSTDMT_initCStream_internal() :
+ *  internal usage only */
+size_t ZSTDMT_initCStream_internal(ZSTDMT_CCtx* zcs,
+                    const void* dict, size_t dictSize, const ZSTD_CDict* cdict,
+                    ZSTD_parameters params, unsigned long long pledgedSrcSize)
+{
+    DEBUGLOG(4, "ZSTDMT_initCStream_internal");
+    /* params are supposed to be fully validated at this point */
+    assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
+    assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
+
+    if (zcs->nbThreads==1) {
+        DEBUGLOG(4, "single thread mode");
+        return ZSTD_initCStream_internal(zcs->cctxPool->cctx[0],
+                                dict, dictSize, cdict,
+                                params, pledgedSrcSize);
+    }
+
+    if (zcs->allJobsCompleted == 0) {   /* previous compression not correctly finished */
+        ZSTDMT_waitForAllJobsCompleted(zcs);
+        ZSTDMT_releaseAllJobResources(zcs);
+        zcs->allJobsCompleted = 1;
+    }
+
+    zcs->params = params;
+    zcs->frameContentSize = pledgedSrcSize;
+    if (dict) {
+        DEBUGLOG(4,"cdictLocal: %08X", (U32)(size_t)zcs->cdictLocal);
+        ZSTD_freeCDict(zcs->cdictLocal);
+        zcs->cdictLocal = ZSTD_createCDict_advanced(dict, dictSize,
+                                                    0 /* byRef */, ZSTD_dm_auto,   /* note : a loadPrefix becomes an internal CDict */
+                                                    params.cParams, zcs->cMem);
+        zcs->cdict = zcs->cdictLocal;
+        if (zcs->cdictLocal == NULL) return ERROR(memory_allocation);
+    } else {
+        DEBUGLOG(4,"cdictLocal: %08X", (U32)(size_t)zcs->cdictLocal);
+        ZSTD_freeCDict(zcs->cdictLocal);
+        zcs->cdictLocal = NULL;
+        zcs->cdict = cdict;
+    }
+
+    zcs->targetDictSize = (zcs->overlapRLog>=9) ? 0 : (size_t)1 << (zcs->params.cParams.windowLog - zcs->overlapRLog);
+    DEBUGLOG(4, "overlapRLog : %u ", zcs->overlapRLog);
+    DEBUGLOG(4, "overlap Size : %u KB", (U32)(zcs->targetDictSize>>10));
+    zcs->targetSectionSize = zcs->sectionSize ? zcs->sectionSize : (size_t)1 << (zcs->params.cParams.windowLog + 2);
+    zcs->targetSectionSize = MAX(ZSTDMT_SECTION_SIZE_MIN, zcs->targetSectionSize);
+    zcs->targetSectionSize = MAX(zcs->targetDictSize, zcs->targetSectionSize);
+    DEBUGLOG(4, "Section Size : %u KB", (U32)(zcs->targetSectionSize>>10));
+    zcs->marginSize = zcs->targetSectionSize >> 2;
+    zcs->inBuffSize = zcs->targetDictSize + zcs->targetSectionSize + zcs->marginSize;
+    zcs->inBuff.buffer = ZSTDMT_getBuffer(zcs->buffPool, zcs->inBuffSize);
+    if (zcs->inBuff.buffer.start == NULL) return ERROR(memory_allocation);
+    zcs->inBuff.filled = 0;
+    zcs->dictSize = 0;
+    zcs->doneJobID = 0;
+    zcs->nextJobID = 0;
+    zcs->frameEnded = 0;
+    zcs->allJobsCompleted = 0;
+    if (params.fParams.checksumFlag) XXH64_reset(&zcs->xxhState, 0);
+    return 0;
+}
+
+size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
+                                const void* dict, size_t dictSize,
+                                ZSTD_parameters params, unsigned long long pledgedSrcSize)
+{
+    DEBUGLOG(5, "ZSTDMT_initCStream_advanced");
+    return ZSTDMT_initCStream_internal(mtctx, dict, dictSize, NULL, params, pledgedSrcSize);
+}
+
+size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
+                               const ZSTD_CDict* cdict,
+                                     ZSTD_frameParameters fParams,
+                                     unsigned long long pledgedSrcSize)
+{
+    ZSTD_parameters params = ZSTD_getParamsFromCDict(cdict);
+    if (cdict==NULL) return ERROR(dictionary_wrong);   /* method incompatible with NULL cdict */
+    params.fParams = fParams;
+    return ZSTDMT_initCStream_internal(mtctx, NULL, 0 /*dictSize*/, cdict,
+                                        params, pledgedSrcSize);
+}
+
+
+/* ZSTDMT_resetCStream() :
+ * pledgedSrcSize is optional and can be zero == unknown */
+size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* zcs, unsigned long long pledgedSrcSize)
+{
+    if (zcs->nbThreads==1)
+        return ZSTD_resetCStream(zcs->cctxPool->cctx[0], pledgedSrcSize);
+    return ZSTDMT_initCStream_internal(zcs, NULL, 0, 0, zcs->params, pledgedSrcSize);
+}
+
+size_t ZSTDMT_initCStream(ZSTDMT_CCtx* zcs, int compressionLevel) {
+    ZSTD_parameters const params = ZSTD_getParams(compressionLevel, 0, 0);
+    return ZSTDMT_initCStream_internal(zcs, NULL, 0, NULL, params, 0);
+}
+
+
+static size_t ZSTDMT_createCompressionJob(ZSTDMT_CCtx* zcs, size_t srcSize, unsigned endFrame)
+{
+    size_t const dstBufferCapacity = ZSTD_compressBound(srcSize);
+    buffer_t const dstBuffer = ZSTDMT_getBuffer(zcs->buffPool, dstBufferCapacity);
+    ZSTD_CCtx* const cctx = ZSTDMT_getCCtx(zcs->cctxPool);
+    unsigned const jobID = zcs->nextJobID & zcs->jobIDMask;
+
+    if ((cctx==NULL) || (dstBuffer.start==NULL)) {
+        zcs->jobs[jobID].jobCompleted = 1;
+        zcs->nextJobID++;
+        ZSTDMT_waitForAllJobsCompleted(zcs);
+        ZSTDMT_releaseAllJobResources(zcs);
+        return ERROR(memory_allocation);
+    }
+
+    DEBUGLOG(4, "preparing job %u to compress %u bytes with %u preload ",
+                zcs->nextJobID, (U32)srcSize, (U32)zcs->dictSize);
+    zcs->jobs[jobID].src = zcs->inBuff.buffer;
+    zcs->jobs[jobID].srcStart = zcs->inBuff.buffer.start;
+    zcs->jobs[jobID].srcSize = srcSize;
+    zcs->jobs[jobID].dictSize = zcs->dictSize;
+    assert(zcs->inBuff.filled >= srcSize + zcs->dictSize);
+    zcs->jobs[jobID].params = zcs->params;
+    /* do not calculate checksum within sections, but write it in header for first section */
+    if (zcs->nextJobID) zcs->jobs[jobID].params.fParams.checksumFlag = 0;
+    zcs->jobs[jobID].cdict = zcs->nextJobID==0 ? zcs->cdict : NULL;
+    zcs->jobs[jobID].fullFrameSize = zcs->frameContentSize;
+    zcs->jobs[jobID].dstBuff = dstBuffer;
+    zcs->jobs[jobID].cctx = cctx;
+    zcs->jobs[jobID].firstChunk = (zcs->nextJobID==0);
+    zcs->jobs[jobID].lastChunk = endFrame;
+    zcs->jobs[jobID].jobCompleted = 0;
+    zcs->jobs[jobID].dstFlushed = 0;
+    zcs->jobs[jobID].jobCompleted_mutex = &zcs->jobCompleted_mutex;
+    zcs->jobs[jobID].jobCompleted_cond = &zcs->jobCompleted_cond;
+
+    /* get a new buffer for next input */
+    if (!endFrame) {
+        size_t const newDictSize = MIN(srcSize + zcs->dictSize, zcs->targetDictSize);
+        DEBUGLOG(5, "ZSTDMT_createCompressionJob::endFrame = %u", endFrame);
+        zcs->inBuff.buffer = ZSTDMT_getBuffer(zcs->buffPool, zcs->inBuffSize);
+        if (zcs->inBuff.buffer.start == NULL) {   /* not enough memory to allocate next input buffer */
+            zcs->jobs[jobID].jobCompleted = 1;
+            zcs->nextJobID++;
+            ZSTDMT_waitForAllJobsCompleted(zcs);
+            ZSTDMT_releaseAllJobResources(zcs);
+            return ERROR(memory_allocation);
+        }
+        DEBUGLOG(5, "inBuff currently filled to %u", (U32)zcs->inBuff.filled);
+        zcs->inBuff.filled -= srcSize + zcs->dictSize - newDictSize;
+        DEBUGLOG(5, "new job : inBuff filled to %u, with %u dict and %u src",
+                    (U32)zcs->inBuff.filled, (U32)newDictSize,
+                    (U32)(zcs->inBuff.filled - newDictSize));
+        memmove(zcs->inBuff.buffer.start,
+            (const char*)zcs->jobs[jobID].srcStart + zcs->dictSize + srcSize - newDictSize,
+            zcs->inBuff.filled);
+        DEBUGLOG(5, "new inBuff pre-filled");
+        zcs->dictSize = newDictSize;
+    } else {   /* if (endFrame==1) */
+        DEBUGLOG(5, "ZSTDMT_createCompressionJob::endFrame = %u", endFrame);
+        zcs->inBuff.buffer = g_nullBuffer;
+        zcs->inBuff.filled = 0;
+        zcs->dictSize = 0;
+        zcs->frameEnded = 1;
+        if (zcs->nextJobID == 0)
+            /* single chunk exception : checksum is calculated directly within worker thread */
+            zcs->params.fParams.checksumFlag = 0;
+    }
+
+    DEBUGLOG(4, "posting job %u : %u bytes  (end:%u) (note : doneJob = %u=>%u)",
+                zcs->nextJobID,
+                (U32)zcs->jobs[jobID].srcSize,
+                zcs->jobs[jobID].lastChunk,
+                zcs->doneJobID,
+                zcs->doneJobID & zcs->jobIDMask);
+    POOL_add(zcs->factory, ZSTDMT_compressChunk, &zcs->jobs[jobID]);   /* this call is blocking when thread worker pool is exhausted */
+    zcs->nextJobID++;
+    return 0;
+}
+
+
+/* ZSTDMT_flushNextJob() :
+ * output : will be updated with amount of data flushed .
+ * blockToFlush : if >0, the function will block and wait if there is no data available to flush .
+ * @return : amount of data remaining within internal buffer, 1 if unknown but > 0, 0 if no more, or an error code */
+static size_t ZSTDMT_flushNextJob(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output, unsigned blockToFlush)
+{
+    unsigned const wJobID = zcs->doneJobID & zcs->jobIDMask;
+    if (zcs->doneJobID == zcs->nextJobID) return 0;   /* all flushed ! */
+    PTHREAD_MUTEX_LOCK(&zcs->jobCompleted_mutex);
+    while (zcs->jobs[wJobID].jobCompleted==0) {
+        DEBUGLOG(5, "waiting for jobCompleted signal from job %u", zcs->doneJobID);
+        if (!blockToFlush) { pthread_mutex_unlock(&zcs->jobCompleted_mutex); return 0; }  /* nothing ready to be flushed => skip */
+        pthread_cond_wait(&zcs->jobCompleted_cond, &zcs->jobCompleted_mutex);  /* block when nothing available to flush */
+    }
+    pthread_mutex_unlock(&zcs->jobCompleted_mutex);
+    /* compression job completed : output can be flushed */
+    {   ZSTDMT_jobDescription job = zcs->jobs[wJobID];
+        if (!job.jobScanned) {
+            if (ZSTD_isError(job.cSize)) {
+                DEBUGLOG(5, "compression error detected ");
+                ZSTDMT_waitForAllJobsCompleted(zcs);
+                ZSTDMT_releaseAllJobResources(zcs);
+                return job.cSize;
+            }
+            ZSTDMT_releaseCCtx(zcs->cctxPool, job.cctx);
+            zcs->jobs[wJobID].cctx = NULL;
+            DEBUGLOG(5, "zcs->params.fParams.checksumFlag : %u ", zcs->params.fParams.checksumFlag);
+            if (zcs->params.fParams.checksumFlag) {
+                XXH64_update(&zcs->xxhState, (const char*)job.srcStart + job.dictSize, job.srcSize);
+                if (zcs->frameEnded && (zcs->doneJobID+1 == zcs->nextJobID)) {  /* write checksum at end of last section */
+                    U32 const checksum = (U32)XXH64_digest(&zcs->xxhState);
+                    DEBUGLOG(5, "writing checksum : %08X \n", checksum);
+                    MEM_writeLE32((char*)job.dstBuff.start + job.cSize, checksum);
+                    job.cSize += 4;
+                    zcs->jobs[wJobID].cSize += 4;
+            }   }
+            ZSTDMT_releaseBuffer(zcs->buffPool, job.src);
+            zcs->jobs[wJobID].srcStart = NULL;
+            zcs->jobs[wJobID].src = g_nullBuffer;
+            zcs->jobs[wJobID].jobScanned = 1;
+        }
+        {   size_t const toWrite = MIN(job.cSize - job.dstFlushed, output->size - output->pos);
+            DEBUGLOG(5, "Flushing %u bytes from job %u ", (U32)toWrite, zcs->doneJobID);
+            memcpy((char*)output->dst + output->pos, (const char*)job.dstBuff.start + job.dstFlushed, toWrite);
+            output->pos += toWrite;
+            job.dstFlushed += toWrite;
+        }
+        if (job.dstFlushed == job.cSize) {   /* output buffer fully flushed => move to next one */
+            ZSTDMT_releaseBuffer(zcs->buffPool, job.dstBuff);
+            zcs->jobs[wJobID].dstBuff = g_nullBuffer;
+            zcs->jobs[wJobID].jobCompleted = 0;
+            zcs->doneJobID++;
+        } else {
+            zcs->jobs[wJobID].dstFlushed = job.dstFlushed;
+        }
+        /* return value : how many bytes left in buffer ; fake it to 1 if unknown but >0 */
+        if (job.cSize > job.dstFlushed) return (job.cSize - job.dstFlushed);
+        if (zcs->doneJobID < zcs->nextJobID) return 1;   /* still some buffer to flush */
+        zcs->allJobsCompleted = zcs->frameEnded;   /* frame completed and entirely flushed */
+        return 0;   /* everything flushed */
+}   }
+
+
+/** ZSTDMT_compressStream_generic() :
+ *  internal use only
+ *  assumption : output and input are valid (pos <= size)
+ * @return : minimum amount of data remaining to flush, 0 if none */
+size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
+                                     ZSTD_outBuffer* output,
+                                     ZSTD_inBuffer* input,
+                                     ZSTD_EndDirective endOp)
+{
+    size_t const newJobThreshold = mtctx->dictSize + mtctx->targetSectionSize + mtctx->marginSize;
+    assert(output->pos <= output->size);
+    assert(input->pos  <= input->size);
+    if ((mtctx->frameEnded) && (endOp==ZSTD_e_continue)) {
+        /* current frame being ended. Only flush/end are allowed. Or start new frame with init */
+        return ERROR(stage_wrong);
+    }
+    if (mtctx->nbThreads==1) {
+        return ZSTD_compressStream_generic(mtctx->cctxPool->cctx[0], output, input, endOp);
+    }
+
+    /* single-pass shortcut (note : this is blocking-mode) */
+    if ( (mtctx->nextJobID==0)      /* just started */
+      && (mtctx->inBuff.filled==0)  /* nothing buffered */
+      && (endOp==ZSTD_e_end)        /* end order */
+      && (output->size - output->pos >= ZSTD_compressBound(input->size - input->pos)) ) { /* enough room */
+        size_t const cSize = ZSTDMT_compress_advanced(mtctx,
+                (char*)output->dst + output->pos, output->size - output->pos,
+                (const char*)input->src + input->pos, input->size - input->pos,
+                mtctx->cdict, mtctx->params, mtctx->overlapRLog);
+        if (ZSTD_isError(cSize)) return cSize;
+        input->pos = input->size;
+        output->pos += cSize;
+        ZSTDMT_releaseBuffer(mtctx->buffPool, mtctx->inBuff.buffer);  /* was allocated in initStream */
+        mtctx->allJobsCompleted = 1;
+        mtctx->frameEnded = 1;
+        return 0;
+    }
+
+    /* fill input buffer */
+    if ((input->src) && (mtctx->inBuff.buffer.start)) {   /* support NULL input */
+        size_t const toLoad = MIN(input->size - input->pos, mtctx->inBuffSize - mtctx->inBuff.filled);
+        DEBUGLOG(2, "inBuff:%08X;  inBuffSize=%u;  ToCopy=%u", (U32)(size_t)mtctx->inBuff.buffer.start, (U32)mtctx->inBuffSize, (U32)toLoad);
+        memcpy((char*)mtctx->inBuff.buffer.start + mtctx->inBuff.filled, (const char*)input->src + input->pos, toLoad);
+        input->pos += toLoad;
+        mtctx->inBuff.filled += toLoad;
+    }
+
+    if ( (mtctx->inBuff.filled >= newJobThreshold)  /* filled enough : let's compress */
+      && (mtctx->nextJobID <= mtctx->doneJobID + mtctx->jobIDMask) ) {   /* avoid overwriting job round buffer */
+        CHECK_F( ZSTDMT_createCompressionJob(mtctx, mtctx->targetSectionSize, 0 /* endFrame */) );
+    }
+
+    /* check for potential compressed data ready to be flushed */
+    CHECK_F( ZSTDMT_flushNextJob(mtctx, output, (mtctx->inBuff.filled == mtctx->inBuffSize) /* blockToFlush */) ); /* block if it wasn't possible to create new job due to saturation */
+
+    if (input->pos < input->size)  /* input not consumed : do not flush yet */
+        endOp = ZSTD_e_continue;
+
+    switch(endOp)
+    {
+        case ZSTD_e_flush:
+            return ZSTDMT_flushStream(mtctx, output);
+        case ZSTD_e_end:
+            return ZSTDMT_endStream(mtctx, output);
+        case ZSTD_e_continue:
+            return 1;
+        default:
+            return ERROR(GENERIC);   /* invalid endDirective */
+    }
+}
+
+
+size_t ZSTDMT_compressStream(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
+{
+    CHECK_F( ZSTDMT_compressStream_generic(zcs, output, input, ZSTD_e_continue) );
+
+    /* recommended next input size : fill current input buffer */
+    return zcs->inBuffSize - zcs->inBuff.filled;   /* note : could be zero when input buffer is fully filled and no more availability to create new job */
+}
+
+
+static size_t ZSTDMT_flushStream_internal(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output, unsigned endFrame)
+{
+    size_t const srcSize = zcs->inBuff.filled - zcs->dictSize;
+
+    if ( ((srcSize > 0) || (endFrame && !zcs->frameEnded))
+       && (zcs->nextJobID <= zcs->doneJobID + zcs->jobIDMask) ) {
+        CHECK_F( ZSTDMT_createCompressionJob(zcs, srcSize, endFrame) );
+    }
+
+    /* check if there is any data available to flush */
+    return ZSTDMT_flushNextJob(zcs, output, 1 /* blockToFlush */);
+}
+
+
+size_t ZSTDMT_flushStream(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output)
+{
+    DEBUGLOG(5, "ZSTDMT_flushStream");
+    if (zcs->nbThreads==1)
+        return ZSTD_flushStream(zcs->cctxPool->cctx[0], output);
+    return ZSTDMT_flushStream_internal(zcs, output, 0 /* endFrame */);
+}
+
+size_t ZSTDMT_endStream(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output)
+{
+    DEBUGLOG(4, "ZSTDMT_endStream");
+    if (zcs->nbThreads==1)
+        return ZSTD_endStream(zcs->cctxPool->cctx[0], output);
+    return ZSTDMT_flushStream_internal(zcs, output, 1 /* endFrame */);
+}

contrib/libzstd/include/zstd/compress/zstdmt_compress.h

@@ -0,0 +1,114 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+ #ifndef ZSTDMT_COMPRESS_H
+ #define ZSTDMT_COMPRESS_H
+
+ #if defined (__cplusplus)
+ extern "C" {
+ #endif
+
+
+/* Note : All prototypes defined in this file are labelled experimental.
+ *        No guarantee of API continuity is provided on any of them.
+ *        In fact, the expectation is that these prototypes will be replaced
+ *        by ZSTD_compress_generic() API in the near future */
+
+/* ===   Dependencies   === */
+#include <stddef.h>                /* size_t */
+#define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_parameters */
+#include "zstd.h"            /* ZSTD_inBuffer, ZSTD_outBuffer, ZSTDLIB_API */
+
+
+/* ===   Memory management   === */
+typedef struct ZSTDMT_CCtx_s ZSTDMT_CCtx;
+ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbThreads);
+ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbThreads,
+                                                    ZSTD_customMem cMem);
+ZSTDLIB_API size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx);
+
+ZSTDLIB_API size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx);
+
+
+/* ===   Simple buffer-to-butter one-pass function   === */
+
+ZSTDLIB_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
+                                       void* dst, size_t dstCapacity,
+                                 const void* src, size_t srcSize,
+                                       int compressionLevel);
+
+
+
+/* ===   Streaming functions   === */
+
+ZSTDLIB_API size_t ZSTDMT_initCStream(ZSTDMT_CCtx* mtctx, int compressionLevel);
+ZSTDLIB_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize);    /**< pledgedSrcSize is optional and can be zero == unknown */
+
+ZSTDLIB_API size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input);
+
+ZSTDLIB_API size_t ZSTDMT_flushStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output);   /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
+ZSTDLIB_API size_t ZSTDMT_endStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output);     /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
+
+
+/* ===   Advanced functions and parameters  === */
+
+#ifndef ZSTDMT_SECTION_SIZE_MIN
+#  define ZSTDMT_SECTION_SIZE_MIN (1U << 20)   /* 1 MB - Minimum size of each compression job */
+#endif
+
+ZSTDLIB_API size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
+                                           void* dst, size_t dstCapacity,
+                                     const void* src, size_t srcSize,
+                                     const ZSTD_CDict* cdict,
+                                           ZSTD_parameters const params,
+                                           unsigned overlapRLog);
+
+ZSTDLIB_API size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
+                                        const void* dict, size_t dictSize,   /* dict can be released after init, a local copy is preserved within zcs */
+                                        ZSTD_parameters params,
+                                        unsigned long long pledgedSrcSize);  /* pledgedSrcSize is optional and can be zero == unknown */
+
+ZSTDLIB_API size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
+                                        const ZSTD_CDict* cdict,
+                                        ZSTD_frameParameters fparams,
+                                        unsigned long long pledgedSrcSize);  /* note : zero means empty */
+
+/* ZSDTMT_parameter :
+ * List of parameters that can be set using ZSTDMT_setMTCtxParameter() */
+typedef enum {
+    ZSTDMT_p_sectionSize,        /* size of input "section". Each section is compressed in parallel. 0 means default, which is dynamically determined within compression functions */
+    ZSTDMT_p_overlapSectionLog   /* Log of overlapped section; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window */
+} ZSDTMT_parameter;
+
+/* ZSTDMT_setMTCtxParameter() :
+ * allow setting individual parameters, one at a time, among a list of enums defined in ZSTDMT_parameter.
+ * The function must be called typically after ZSTD_createCCtx().
+ * Parameters not explicitly reset by ZSTDMT_init*() remain the same in consecutive compression sessions.
+ * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
+ZSTDLIB_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSDTMT_parameter parameter, unsigned value);
+
+
+/*! ZSTDMT_compressStream_generic() :
+ *  Combines ZSTDMT_compressStream() with ZSTDMT_flushStream() or ZSTDMT_endStream()
+ *  depending on flush directive.
+ * @return : minimum amount of data still to be flushed
+ *           0 if fully flushed
+ *           or an error code */
+ZSTDLIB_API size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
+                                                ZSTD_outBuffer* output,
+                                                ZSTD_inBuffer* input,
+                                                ZSTD_EndDirective endOp);
+
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif   /* ZSTDMT_COMPRESS_H */

contrib/libzstd/include/zstd/decompress/huf_decompress.c

@@ -35,16 +35,19 @@
 /* **************************************************************
 *  Compiler specifics
 ****************************************************************/
-#if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
-/* inline is defined */
-#elif defined(_MSC_VER) || defined(__GNUC__)
-#  define inline __inline
-#else
-#  define inline /* disable inline */
-#endif
-
 #ifdef _MSC_VER    /* Visual Studio */
+#  define FORCE_INLINE static __forceinline
 #  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#else
+#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#    ifdef __GNUC__
+#      define FORCE_INLINE static inline __attribute__((always_inline))
+#    else
+#      define FORCE_INLINE static inline
+#    endif
+#  else
+#    define FORCE_INLINE static
+#  endif /* __STDC_VERSION__ */
 #endif
 
 
@@ -64,6 +67,12 @@
 #define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
 
 
+/* **************************************************************
+*  Byte alignment for workSpace management
+****************************************************************/
+#define HUF_ALIGN(x, a) HUF_ALIGN_MASK((x), (a) - 1)
+#define HUF_ALIGN_MASK(x, mask) (((x) + (mask)) & ~(mask))
+
 /*-***************************/
 /*  generic DTableDesc       */
 /*-***************************/
@@ -84,16 +93,28 @@ static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
 
 typedef struct { BYTE byte; BYTE nbBits; } HUF_DEltX2;   /* single-symbol decoding */
 
-size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize)
+size_t HUF_readDTableX2_wksp(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize)
 {
-    BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1];
-    U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];   /* large enough for values from 0 to 16 */
     U32 tableLog = 0;
     U32 nbSymbols = 0;
     size_t iSize;
     void* const dtPtr = DTable + 1;
     HUF_DEltX2* const dt = (HUF_DEltX2*)dtPtr;
 
+    U32* rankVal;
+    BYTE* huffWeight;
+    size_t spaceUsed32 = 0;
+
+    rankVal = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_ABSOLUTEMAX + 1;
+    huffWeight = (BYTE *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
+
+    if ((spaceUsed32 << 2) > wkspSize)
+        return ERROR(tableLog_tooLarge);
+    workSpace = (U32 *)workSpace + spaceUsed32;
+    wkspSize -= (spaceUsed32 << 2);
+
     HUF_STATIC_ASSERT(sizeof(DTableDesc) == sizeof(HUF_DTable));
     /* memset(huffWeight, 0, sizeof(huffWeight)); */   /* is not necessary, even though some analyzer complain ... */
 
@@ -102,16 +123,16 @@ size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize)
 
     /* Table header */
     {   DTableDesc dtd = HUF_getDTableDesc(DTable);
-        if (tableLog > (U32)(dtd.maxTableLog+1)) return ERROR(tableLog_tooLarge);   /* DTable too small, huffman tree cannot fit in */
+        if (tableLog > (U32)(dtd.maxTableLog+1)) return ERROR(tableLog_tooLarge);   /* DTable too small, Huffman tree cannot fit in */
         dtd.tableType = 0;
         dtd.tableLog = (BYTE)tableLog;
         memcpy(DTable, &dtd, sizeof(dtd));
     }
 
-    /* Prepare ranks */
+    /* Calculate starting value for each rank */
     {   U32 n, nextRankStart = 0;
         for (n=1; n<tableLog+1; n++) {
-            U32 current = nextRankStart;
+            U32 const current = nextRankStart;
             nextRankStart += (rankVal[n] << (n-1));
             rankVal[n] = current;
     }   }
@@ -121,17 +142,24 @@ size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize)
         for (n=0; n<nbSymbols; n++) {
             U32 const w = huffWeight[n];
             U32 const length = (1 << w) >> 1;
-            U32 i;
+            U32 u;
             HUF_DEltX2 D;
             D.byte = (BYTE)n; D.nbBits = (BYTE)(tableLog + 1 - w);
-            for (i = rankVal[w]; i < rankVal[w] + length; i++)
-                dt[i] = D;
+            for (u = rankVal[w]; u < rankVal[w] + length; u++)
+                dt[u] = D;
             rankVal[w] += length;
     }   }
 
     return iSize;
 }
 
+size_t HUF_readDTableX2(HUF_DTable* DTable, const void* src, size_t srcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_readDTableX2_wksp(DTable, src, srcSize,
+                                 workSpace, sizeof(workSpace));
+}
+
 
 static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, const U32 dtLog)
 {
@@ -152,7 +180,7 @@ static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, con
     if (MEM_64bits()) \
         HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
 
-static inline size_t HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX2* const dt, const U32 dtLog)
+FORCE_INLINE size_t HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX2* const dt, const U32 dtLog)
 {
     BYTE* const pStart = p;
 
@@ -209,11 +237,13 @@ size_t HUF_decompress1X2_usingDTable(
     return HUF_decompress1X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
 }
 
-size_t HUF_decompress1X2_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t const hSize = HUF_readDTableX2 (DCtx, cSrc, cSrcSize);
+    size_t const hSize = HUF_readDTableX2_wksp(DCtx, cSrc, cSrcSize, workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
@@ -221,6 +251,15 @@ size_t HUF_decompress1X2_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, cons
     return HUF_decompress1X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
 }
 
+
+size_t HUF_decompress1X2_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X2_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
+
 size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
@@ -332,11 +371,14 @@ size_t HUF_decompress4X2_usingDTable(
 }
 
 
-size_t HUF_decompress4X2_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t const hSize = HUF_readDTableX2 (dctx, cSrc, cSrcSize);
+    size_t const hSize = HUF_readDTableX2_wksp (dctx, cSrc, cSrcSize,
+                                                workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
@@ -344,6 +386,13 @@ size_t HUF_decompress4X2_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, cons
     return HUF_decompress4X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, dctx);
 }
 
+
+size_t HUF_decompress4X2_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
 size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
@@ -358,13 +407,15 @@ typedef struct { U16 sequence; BYTE nbBits; BYTE length; } HUF_DEltX4;  /* doubl
 
 typedef struct { BYTE symbol; BYTE weight; } sortedSymbol_t;
 
+/* HUF_fillDTableX4Level2() :
+ * `rankValOrigin` must be a table of at least (HUF_TABLELOG_MAX + 1) U32 */
 static void HUF_fillDTableX4Level2(HUF_DEltX4* DTable, U32 sizeLog, const U32 consumed,
                            const U32* rankValOrigin, const int minWeight,
                            const sortedSymbol_t* sortedSymbols, const U32 sortedListSize,
                            U32 nbBitsBaseline, U16 baseSeq)
 {
     HUF_DEltX4 DElt;
-    U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];
+    U32 rankVal[HUF_TABLELOG_MAX + 1];
 
     /* get pre-calculated rankVal */
     memcpy(rankVal, rankValOrigin, sizeof(rankVal));
@@ -398,14 +449,15 @@ static void HUF_fillDTableX4Level2(HUF_DEltX4* DTable, U32 sizeLog, const U32 co
     }   }
 }
 
-typedef U32 rankVal_t[HUF_TABLELOG_ABSOLUTEMAX][HUF_TABLELOG_ABSOLUTEMAX + 1];
+typedef U32 rankValCol_t[HUF_TABLELOG_MAX + 1];
+typedef rankValCol_t rankVal_t[HUF_TABLELOG_MAX];
 
 static void HUF_fillDTableX4(HUF_DEltX4* DTable, const U32 targetLog,
                            const sortedSymbol_t* sortedList, const U32 sortedListSize,
                            const U32* rankStart, rankVal_t rankValOrigin, const U32 maxWeight,
                            const U32 nbBitsBaseline)
 {
-    U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];
+    U32 rankVal[HUF_TABLELOG_MAX + 1];
     const int scaleLog = nbBitsBaseline - targetLog;   /* note : targetLog >= srcLog, hence scaleLog <= 1 */
     const U32 minBits  = nbBitsBaseline - maxWeight;
     U32 s;
@@ -442,23 +494,46 @@ static void HUF_fillDTableX4(HUF_DEltX4* DTable, const U32 targetLog,
     }
 }
 
-size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize)
+size_t HUF_readDTableX4_wksp(HUF_DTable* DTable, const void* src,
+                             size_t srcSize, void* workSpace,
+                             size_t wkspSize)
 {
-    BYTE weightList[HUF_SYMBOLVALUE_MAX + 1];
-    sortedSymbol_t sortedSymbol[HUF_SYMBOLVALUE_MAX + 1];
-    U32 rankStats[HUF_TABLELOG_ABSOLUTEMAX + 1] = { 0 };
-    U32 rankStart0[HUF_TABLELOG_ABSOLUTEMAX + 2] = { 0 };
-    U32* const rankStart = rankStart0+1;
-    rankVal_t rankVal;
     U32 tableLog, maxW, sizeOfSort, nbSymbols;
     DTableDesc dtd = HUF_getDTableDesc(DTable);
     U32 const maxTableLog = dtd.maxTableLog;
     size_t iSize;
     void* dtPtr = DTable+1;   /* force compiler to avoid strict-aliasing */
     HUF_DEltX4* const dt = (HUF_DEltX4*)dtPtr;
+    U32 *rankStart;
 
-    HUF_STATIC_ASSERT(sizeof(HUF_DEltX4) == sizeof(HUF_DTable));   /* if compilation fails here, assertion is false */
-    if (maxTableLog > HUF_TABLELOG_ABSOLUTEMAX) return ERROR(tableLog_tooLarge);
+    rankValCol_t* rankVal;
+    U32* rankStats;
+    U32* rankStart0;
+    sortedSymbol_t* sortedSymbol;
+    BYTE* weightList;
+    size_t spaceUsed32 = 0;
+
+    rankVal = (rankValCol_t *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += (sizeof(rankValCol_t) * HUF_TABLELOG_MAX) >> 2;
+    rankStats = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_MAX + 1;
+    rankStart0 = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_MAX + 2;
+    sortedSymbol = (sortedSymbol_t *)workSpace + (spaceUsed32 * sizeof(U32)) / sizeof(sortedSymbol_t);
+    spaceUsed32 += HUF_ALIGN(sizeof(sortedSymbol_t) * (HUF_SYMBOLVALUE_MAX + 1), sizeof(U32)) >> 2;
+    weightList = (BYTE *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
+
+    if ((spaceUsed32 << 2) > wkspSize)
+        return ERROR(tableLog_tooLarge);
+    workSpace = (U32 *)workSpace + spaceUsed32;
+    wkspSize -= (spaceUsed32 << 2);
+
+    rankStart = rankStart0 + 1;
+    memset(rankStats, 0, sizeof(U32) * (2 * HUF_TABLELOG_MAX + 2 + 1));
+
+    HUF_STATIC_ASSERT(sizeof(HUF_DEltX4) == sizeof(HUF_DTable));   /* if compiler fails here, assertion is wrong */
+    if (maxTableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
     /* memset(weightList, 0, sizeof(weightList)); */  /* is not necessary, even though some analyzer complain ... */
 
     iSize = HUF_readStats(weightList, HUF_SYMBOLVALUE_MAX + 1, rankStats, &nbSymbols, &tableLog, src, srcSize);
@@ -522,6 +597,12 @@ size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize)
     return iSize;
 }
 
+size_t HUF_readDTableX4(HUF_DTable* DTable, const void* src, size_t srcSize)
+{
+  U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+  return HUF_readDTableX4_wksp(DTable, src, srcSize,
+                               workSpace, sizeof(workSpace));
+}
 
 static U32 HUF_decodeSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
 {
@@ -540,7 +621,8 @@ static U32 HUF_decodeLastSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DE
         if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8)) {
             BIT_skipBits(DStream, dt[val].nbBits);
             if (DStream->bitsConsumed > (sizeof(DStream->bitContainer)*8))
-                DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);   /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
+                /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
+                DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);
     }   }
     return 1;
 }
@@ -557,7 +639,7 @@ static U32 HUF_decodeLastSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DE
     if (MEM_64bits()) \
         ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
 
-static inline size_t HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const HUF_DEltX4* const dt, const U32 dtLog)
+FORCE_INLINE size_t HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const HUF_DEltX4* const dt, const U32 dtLog)
 {
     BYTE* const pStart = p;
 
@@ -621,11 +703,14 @@ size_t HUF_decompress1X4_usingDTable(
     return HUF_decompress1X4_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
 }
 
-size_t HUF_decompress1X4_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress1X4_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t const hSize = HUF_readDTableX4 (DCtx, cSrc, cSrcSize);
+    size_t const hSize = HUF_readDTableX4_wksp(DCtx, cSrc, cSrcSize,
+                                               workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
@@ -633,6 +718,15 @@ size_t HUF_decompress1X4_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, cons
     return HUF_decompress1X4_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
 }
 
+
+size_t HUF_decompress1X4_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X4_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
+
 size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX4(DTable, HUF_TABLELOG_MAX);
@@ -743,11 +837,14 @@ size_t HUF_decompress4X4_usingDTable(
 }
 
 
-size_t HUF_decompress4X4_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress4X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t hSize = HUF_readDTableX4 (dctx, cSrc, cSrcSize);
+    size_t hSize = HUF_readDTableX4_wksp(dctx, cSrc, cSrcSize,
+                                         workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
@@ -755,6 +852,15 @@ size_t HUF_decompress4X4_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, cons
     return HUF_decompress4X4_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx);
 }
 
+
+size_t HUF_decompress4X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X4_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
+
 size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX4(DTable, HUF_TABLELOG_MAX);
@@ -856,19 +962,32 @@ size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const
     }
 }
 
-size_t HUF_decompress4X_hufOnly (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X_hufOnly_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                         workSpace, sizeof(workSpace));
+}
+
+
+size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst,
+                                     size_t dstSize, const void* cSrc,
+                                     size_t cSrcSize, void* workSpace,
+                                     size_t wkspSize)
 {
     /* validation checks */
     if (dstSize == 0) return ERROR(dstSize_tooSmall);
     if ((cSrcSize >= dstSize) || (cSrcSize <= 1)) return ERROR(corruption_detected);   /* invalid */
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
-        return algoNb ? HUF_decompress4X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
-                        HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+        return algoNb ? HUF_decompress4X4_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize):
+                        HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize);
     }
 }
 
-size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress1X_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                  const void* cSrc, size_t cSrcSize,
+                                  void* workSpace, size_t wkspSize)
 {
     /* validation checks */
     if (dstSize == 0) return ERROR(dstSize_tooSmall);
@@ -877,7 +996,17 @@ size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const
     if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
-        return algoNb ? HUF_decompress1X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
-                        HUF_decompress1X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+        return algoNb ? HUF_decompress1X4_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                                cSrcSize, workSpace, wkspSize):
+                        HUF_decompress1X2_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                                cSrcSize, workSpace, wkspSize);
     }
 }
+
+size_t HUF_decompress1X_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
+                             const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                      workSpace, sizeof(workSpace));
+}

contrib/libzstd/include/zstd/decompress/zbuff_decompress.c

@@ -1,252 +0,0 @@
-/**
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
- * All rights reserved.
- *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
- */
-
-
-
-/* *************************************
-*  Dependencies
-***************************************/
-#include <stdlib.h>
-#include "error_private.h"
-#include "zstd_internal.h"  /* MIN, ZSTD_blockHeaderSize, ZSTD_BLOCKSIZE_MAX */
-#define ZBUFF_STATIC_LINKING_ONLY
-#include "zbuff.h"
-
-
-typedef enum { ZBUFFds_init, ZBUFFds_loadHeader,
-               ZBUFFds_read, ZBUFFds_load, ZBUFFds_flush } ZBUFF_dStage;
-
-/* *** Resource management *** */
-struct ZBUFF_DCtx_s {
-    ZSTD_DCtx* zd;
-    ZSTD_frameParams fParams;
-    ZBUFF_dStage stage;
-    char*  inBuff;
-    size_t inBuffSize;
-    size_t inPos;
-    char*  outBuff;
-    size_t outBuffSize;
-    size_t outStart;
-    size_t outEnd;
-    size_t blockSize;
-    BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
-    size_t lhSize;
-    ZSTD_customMem customMem;
-};   /* typedef'd to ZBUFF_DCtx within "zbuff.h" */
-
-
-ZBUFF_DCtx* ZBUFF_createDCtx(void)
-{
-    return ZBUFF_createDCtx_advanced(defaultCustomMem);
-}
-
-ZBUFF_DCtx* ZBUFF_createDCtx_advanced(ZSTD_customMem customMem)
-{
-    ZBUFF_DCtx* zbd;
-
-    if (!customMem.customAlloc && !customMem.customFree)
-        customMem = defaultCustomMem;
-
-    if (!customMem.customAlloc || !customMem.customFree)
-        return NULL;
-
-    zbd = (ZBUFF_DCtx*)customMem.customAlloc(customMem.opaque, sizeof(ZBUFF_DCtx));
-    if (zbd==NULL) return NULL;
-    memset(zbd, 0, sizeof(ZBUFF_DCtx));
-    memcpy(&zbd->customMem, &customMem, sizeof(ZSTD_customMem));
-    zbd->zd = ZSTD_createDCtx_advanced(customMem);
-    if (zbd->zd == NULL) { ZBUFF_freeDCtx(zbd); return NULL; }
-    zbd->stage = ZBUFFds_init;
-    return zbd;
-}
-
-size_t ZBUFF_freeDCtx(ZBUFF_DCtx* zbd)
-{
-    if (zbd==NULL) return 0;   /* support free on null */
-    ZSTD_freeDCtx(zbd->zd);
-    if (zbd->inBuff) zbd->customMem.customFree(zbd->customMem.opaque, zbd->inBuff);
-    if (zbd->outBuff) zbd->customMem.customFree(zbd->customMem.opaque, zbd->outBuff);
-    zbd->customMem.customFree(zbd->customMem.opaque, zbd);
-    return 0;
-}
-
-
-/* *** Initialization *** */
-
-size_t ZBUFF_decompressInitDictionary(ZBUFF_DCtx* zbd, const void* dict, size_t dictSize)
-{
-    zbd->stage = ZBUFFds_loadHeader;
-    zbd->lhSize = zbd->inPos = zbd->outStart = zbd->outEnd = 0;
-    return ZSTD_decompressBegin_usingDict(zbd->zd, dict, dictSize);
-}
-
-size_t ZBUFF_decompressInit(ZBUFF_DCtx* zbd)
-{
-    return ZBUFF_decompressInitDictionary(zbd, NULL, 0);
-}
-
-
-/* internal util function */
-MEM_STATIC size_t ZBUFF_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
-{
-    size_t const length = MIN(dstCapacity, srcSize);
-    memcpy(dst, src, length);
-    return length;
-}
-
-
-/* *** Decompression *** */
-
-size_t ZBUFF_decompressContinue(ZBUFF_DCtx* zbd,
-                                void* dst, size_t* dstCapacityPtr,
-                          const void* src, size_t* srcSizePtr)
-{
-    const char* const istart = (const char*)src;
-    const char* const iend = istart + *srcSizePtr;
-    const char* ip = istart;
-    char* const ostart = (char*)dst;
-    char* const oend = ostart + *dstCapacityPtr;
-    char* op = ostart;
-    U32 someMoreWork = 1;
-
-    while (someMoreWork) {
-        switch(zbd->stage)
-        {
-        case ZBUFFds_init :
-            return ERROR(init_missing);
-
-        case ZBUFFds_loadHeader :
-            {   size_t const hSize = ZSTD_getFrameParams(&(zbd->fParams), zbd->headerBuffer, zbd->lhSize);
-                if (ZSTD_isError(hSize)) return hSize;
-                if (hSize != 0) {   /* need more input */
-                    size_t const toLoad = hSize - zbd->lhSize;   /* if hSize!=0, hSize > zbd->lhSize */
-                    if (toLoad > (size_t)(iend-ip)) {   /* not enough input to load full header */
-                        memcpy(zbd->headerBuffer + zbd->lhSize, ip, iend-ip);
-                        zbd->lhSize += iend-ip;
-                        *dstCapacityPtr = 0;
-                        return (hSize - zbd->lhSize) + ZSTD_blockHeaderSize;   /* remaining header bytes + next block header */
-                    }
-                    memcpy(zbd->headerBuffer + zbd->lhSize, ip, toLoad); zbd->lhSize = hSize; ip += toLoad;
-                    break;
-            }   }
-
-            /* Consume header */
-            {   size_t const h1Size = ZSTD_nextSrcSizeToDecompress(zbd->zd);  /* == ZSTD_frameHeaderSize_min */
-                size_t const h1Result = ZSTD_decompressContinue(zbd->zd, NULL, 0, zbd->headerBuffer, h1Size);
-                if (ZSTD_isError(h1Result)) return h1Result;   /* should not happen : already checked */
-                if (h1Size < zbd->lhSize) {   /* long header */
-                    size_t const h2Size = ZSTD_nextSrcSizeToDecompress(zbd->zd);
-                    size_t const h2Result = ZSTD_decompressContinue(zbd->zd, NULL, 0, zbd->headerBuffer+h1Size, h2Size);
-                    if (ZSTD_isError(h2Result)) return h2Result;
-            }   }
-
-            zbd->fParams.windowSize = MAX(zbd->fParams.windowSize, 1U << ZSTD_WINDOWLOG_ABSOLUTEMIN);
-
-            /* Frame header instruct buffer sizes */
-            {   size_t const blockSize = MIN(zbd->fParams.windowSize, ZSTD_BLOCKSIZE_ABSOLUTEMAX);
-                size_t const neededOutSize = zbd->fParams.windowSize + blockSize;
-                zbd->blockSize = blockSize;
-                if (zbd->inBuffSize < blockSize) {
-                    zbd->customMem.customFree(zbd->customMem.opaque, zbd->inBuff);
-                    zbd->inBuffSize = blockSize;
-                    zbd->inBuff = (char*)zbd->customMem.customAlloc(zbd->customMem.opaque, blockSize);
-                    if (zbd->inBuff == NULL) return ERROR(memory_allocation);
-                }
-                if (zbd->outBuffSize < neededOutSize) {
-                    zbd->customMem.customFree(zbd->customMem.opaque, zbd->outBuff);
-                    zbd->outBuffSize = neededOutSize;
-                    zbd->outBuff = (char*)zbd->customMem.customAlloc(zbd->customMem.opaque, neededOutSize);
-                    if (zbd->outBuff == NULL) return ERROR(memory_allocation);
-            }   }
-            zbd->stage = ZBUFFds_read;
-            /* pass-through */
-
-        case ZBUFFds_read:
-            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zbd->zd);
-                if (neededInSize==0) {  /* end of frame */
-                    zbd->stage = ZBUFFds_init;
-                    someMoreWork = 0;
-                    break;
-                }
-                if ((size_t)(iend-ip) >= neededInSize) {  /* decode directly from src */
-                    const int isSkipFrame = ZSTD_isSkipFrame(zbd->zd);
-                    size_t const decodedSize = ZSTD_decompressContinue(zbd->zd,
-                        zbd->outBuff + zbd->outStart, (isSkipFrame ? 0 : zbd->outBuffSize - zbd->outStart),
-                        ip, neededInSize);
-                    if (ZSTD_isError(decodedSize)) return decodedSize;
-                    ip += neededInSize;
-                    if (!decodedSize && !isSkipFrame) break;   /* this was just a header */
-                    zbd->outEnd = zbd->outStart +  decodedSize;
-                    zbd->stage = ZBUFFds_flush;
-                    break;
-                }
-                if (ip==iend) { someMoreWork = 0; break; }   /* no more input */
-                zbd->stage = ZBUFFds_load;
-                /* pass-through */
-            }
-
-        case ZBUFFds_load:
-            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zbd->zd);
-                size_t const toLoad = neededInSize - zbd->inPos;   /* should always be <= remaining space within inBuff */
-                size_t loadedSize;
-                if (toLoad > zbd->inBuffSize - zbd->inPos) return ERROR(corruption_detected);   /* should never happen */
-                loadedSize = ZBUFF_limitCopy(zbd->inBuff + zbd->inPos, toLoad, ip, iend-ip);
-                ip += loadedSize;
-                zbd->inPos += loadedSize;
-                if (loadedSize < toLoad) { someMoreWork = 0; break; }   /* not enough input, wait for more */
-
-                /* decode loaded input */
-                {  const int isSkipFrame = ZSTD_isSkipFrame(zbd->zd);
-                   size_t const decodedSize = ZSTD_decompressContinue(zbd->zd,
-                        zbd->outBuff + zbd->outStart, zbd->outBuffSize - zbd->outStart,
-                        zbd->inBuff, neededInSize);
-                    if (ZSTD_isError(decodedSize)) return decodedSize;
-                    zbd->inPos = 0;   /* input is consumed */
-                    if (!decodedSize && !isSkipFrame) { zbd->stage = ZBUFFds_read; break; }   /* this was just a header */
-                    zbd->outEnd = zbd->outStart +  decodedSize;
-                    zbd->stage = ZBUFFds_flush;
-                    /* pass-through */
-            }   }
-
-        case ZBUFFds_flush:
-            {   size_t const toFlushSize = zbd->outEnd - zbd->outStart;
-                size_t const flushedSize = ZBUFF_limitCopy(op, oend-op, zbd->outBuff + zbd->outStart, toFlushSize);
-                op += flushedSize;
-                zbd->outStart += flushedSize;
-                if (flushedSize == toFlushSize) {  /* flush completed */
-                    zbd->stage = ZBUFFds_read;
-                    if (zbd->outStart + zbd->blockSize > zbd->outBuffSize)
-                        zbd->outStart = zbd->outEnd = 0;
-                    break;
-                }
-                /* cannot flush everything */
-                someMoreWork = 0;
-                break;
-            }
-        default: return ERROR(GENERIC);   /* impossible */
-    }   }
-
-    /* result */
-    *srcSizePtr = ip-istart;
-    *dstCapacityPtr = op-ostart;
-    {   size_t nextSrcSizeHint = ZSTD_nextSrcSizeToDecompress(zbd->zd);
-        if (!nextSrcSizeHint) return (zbd->outEnd != zbd->outStart);   /* return 0 only if fully flushed too */
-        nextSrcSizeHint += ZSTD_blockHeaderSize * (ZSTD_nextInputType(zbd->zd) == ZSTDnit_block);
-        if (zbd->inPos > nextSrcSizeHint) return ERROR(GENERIC);   /* should never happen */
-        nextSrcSizeHint -= zbd->inPos;   /* already loaded*/
-        return nextSrcSizeHint;
-    }
-}
-
-
-/* *************************************
-*  Tool functions
-***************************************/
-size_t ZBUFF_recommendedDInSize(void)  { return ZSTD_BLOCKSIZE_ABSOLUTEMAX + ZSTD_blockHeaderSize /* block header size*/ ; }
-size_t ZBUFF_recommendedDOutSize(void) { return ZSTD_BLOCKSIZE_ABSOLUTEMAX; }

contrib/libzstd/include/zstd/deprecated/zbuff.h

@@ -9,35 +9,52 @@
 
 /* ***************************************************************
 *  NOTES/WARNINGS
-*****************************************************************/
-/* The streaming API defined here will soon be deprecated by the
-* new one in 'zstd.h'; consider migrating towards newer streaming
-* API. See 'lib/README.md'.
-*****************************************************************/
+******************************************************************/
+/* The streaming API defined here is deprecated.
+ * Consider migrating towards ZSTD_compressStream() API in `zstd.h`
+ * See 'lib/README.md'.
+ *****************************************************************/
 
-#ifndef ZSTD_BUFFERED_H_23987
-#define ZSTD_BUFFERED_H_23987
 
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
+#ifndef ZSTD_BUFFERED_H_23987
+#define ZSTD_BUFFERED_H_23987
+
 /* *************************************
 *  Dependencies
 ***************************************/
 #include <stddef.h>      /* size_t */
+#include "zstd.h"        /* ZSTD_CStream, ZSTD_DStream, ZSTDLIB_API */
 
 
 /* ***************************************************************
 *  Compiler specifics
 *****************************************************************/
-/* ZSTD_DLL_EXPORT :
-*  Enable exporting of functions when building a Windows DLL */
-#if defined(_WIN32) && defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
-#  define ZSTDLIB_API __declspec(dllexport)
+/* Deprecation warnings */
+/* Should these warnings be a problem,
+   it is generally possible to disable them,
+   typically with -Wno-deprecated-declarations for gcc
+   or _CRT_SECURE_NO_WARNINGS in Visual.
+   Otherwise, it's also possible to define ZBUFF_DISABLE_DEPRECATE_WARNINGS */
+#ifdef ZBUFF_DISABLE_DEPRECATE_WARNINGS
+#  define ZBUFF_DEPRECATED(message) ZSTDLIB_API  /* disable deprecation warnings */
 #else
-#  define ZSTDLIB_API
-#endif
+#  if defined (__cplusplus) && (__cplusplus >= 201402) /* C++14 or greater */
+#    define ZBUFF_DEPRECATED(message) [[deprecated(message)]] ZSTDLIB_API
+#  elif (defined(__GNUC__) && (__GNUC__ >= 5)) || defined(__clang__)
+#    define ZBUFF_DEPRECATED(message) ZSTDLIB_API __attribute__((deprecated(message)))
+#  elif defined(__GNUC__) && (__GNUC__ >= 3)
+#    define ZBUFF_DEPRECATED(message) ZSTDLIB_API __attribute__((deprecated))
+#  elif defined(_MSC_VER)
+#    define ZBUFF_DEPRECATED(message) ZSTDLIB_API __declspec(deprecated(message))
+#  else
+#    pragma message("WARNING: You need to implement ZBUFF_DEPRECATED for this compiler")
+#    define ZBUFF_DEPRECATED(message) ZSTDLIB_API
+#  endif
+#endif /* ZBUFF_DISABLE_DEPRECATE_WARNINGS */
 
 
 /* *************************************
@@ -49,16 +66,16 @@ extern "C" {
 *  ZBUFF and ZSTD are 100% interoperable,
 *  frames created by one can be decoded by the other one */
 
-typedef struct ZBUFF_CCtx_s ZBUFF_CCtx;
-ZSTDLIB_API ZBUFF_CCtx* ZBUFF_createCCtx(void);
-ZSTDLIB_API size_t      ZBUFF_freeCCtx(ZBUFF_CCtx* cctx);
+typedef ZSTD_CStream ZBUFF_CCtx;
+ZBUFF_DEPRECATED("use ZSTD_createCStream") ZBUFF_CCtx* ZBUFF_createCCtx(void);
+ZBUFF_DEPRECATED("use ZSTD_freeCStream")   size_t      ZBUFF_freeCCtx(ZBUFF_CCtx* cctx);
 
-ZSTDLIB_API size_t ZBUFF_compressInit(ZBUFF_CCtx* cctx, int compressionLevel);
-ZSTDLIB_API size_t ZBUFF_compressInitDictionary(ZBUFF_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel);
+ZBUFF_DEPRECATED("use ZSTD_initCStream")           size_t ZBUFF_compressInit(ZBUFF_CCtx* cctx, int compressionLevel);
+ZBUFF_DEPRECATED("use ZSTD_initCStream_usingDict") size_t ZBUFF_compressInitDictionary(ZBUFF_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel);
 
-ZSTDLIB_API size_t ZBUFF_compressContinue(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr, const void* src, size_t* srcSizePtr);
-ZSTDLIB_API size_t ZBUFF_compressFlush(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr);
-ZSTDLIB_API size_t ZBUFF_compressEnd(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr);
+ZBUFF_DEPRECATED("use ZSTD_compressStream") size_t ZBUFF_compressContinue(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr, const void* src, size_t* srcSizePtr);
+ZBUFF_DEPRECATED("use ZSTD_flushStream")    size_t ZBUFF_compressFlush(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr);
+ZBUFF_DEPRECATED("use ZSTD_endStream")      size_t ZBUFF_compressEnd(ZBUFF_CCtx* cctx, void* dst, size_t* dstCapacityPtr);
 
 /*-*************************************************
 *  Streaming compression - howto
@@ -101,14 +118,14 @@ ZSTDLIB_API size_t ZBUFF_compressEnd(ZBUFF_CCtx* cctx, void* dst, size_t* dstCap
 * **************************************************/
 
 
-typedef struct ZBUFF_DCtx_s ZBUFF_DCtx;
-ZSTDLIB_API ZBUFF_DCtx* ZBUFF_createDCtx(void);
-ZSTDLIB_API size_t      ZBUFF_freeDCtx(ZBUFF_DCtx* dctx);
+typedef ZSTD_DStream ZBUFF_DCtx;
+ZBUFF_DEPRECATED("use ZSTD_createDStream") ZBUFF_DCtx* ZBUFF_createDCtx(void);
+ZBUFF_DEPRECATED("use ZSTD_freeDStream")   size_t      ZBUFF_freeDCtx(ZBUFF_DCtx* dctx);
 
-ZSTDLIB_API size_t ZBUFF_decompressInit(ZBUFF_DCtx* dctx);
-ZSTDLIB_API size_t ZBUFF_decompressInitDictionary(ZBUFF_DCtx* dctx, const void* dict, size_t dictSize);
+ZBUFF_DEPRECATED("use ZSTD_initDStream")           size_t ZBUFF_decompressInit(ZBUFF_DCtx* dctx);
+ZBUFF_DEPRECATED("use ZSTD_initDStream_usingDict") size_t ZBUFF_decompressInitDictionary(ZBUFF_DCtx* dctx, const void* dict, size_t dictSize);
 
-ZSTDLIB_API size_t ZBUFF_decompressContinue(ZBUFF_DCtx* dctx,
+ZBUFF_DEPRECATED("use ZSTD_decompressStream") size_t ZBUFF_decompressContinue(ZBUFF_DCtx* dctx,
                                             void* dst, size_t* dstCapacityPtr,
                                       const void* src, size_t* srcSizePtr);
 
@@ -141,18 +158,22 @@ ZSTDLIB_API size_t ZBUFF_decompressContinue(ZBUFF_DCtx* dctx,
 /* *************************************
 *  Tool functions
 ***************************************/
-ZSTDLIB_API unsigned ZBUFF_isError(size_t errorCode);
-ZSTDLIB_API const char* ZBUFF_getErrorName(size_t errorCode);
+ZBUFF_DEPRECATED("use ZSTD_isError")      unsigned ZBUFF_isError(size_t errorCode);
+ZBUFF_DEPRECATED("use ZSTD_getErrorName") const char* ZBUFF_getErrorName(size_t errorCode);
 
 /** Functions below provide recommended buffer sizes for Compression or Decompression operations.
 *   These sizes are just hints, they tend to offer better latency */
-ZSTDLIB_API size_t ZBUFF_recommendedCInSize(void);
-ZSTDLIB_API size_t ZBUFF_recommendedCOutSize(void);
-ZSTDLIB_API size_t ZBUFF_recommendedDInSize(void);
-ZSTDLIB_API size_t ZBUFF_recommendedDOutSize(void);
+ZBUFF_DEPRECATED("use ZSTD_CStreamInSize")  size_t ZBUFF_recommendedCInSize(void);
+ZBUFF_DEPRECATED("use ZSTD_CStreamOutSize") size_t ZBUFF_recommendedCOutSize(void);
+ZBUFF_DEPRECATED("use ZSTD_DStreamInSize")  size_t ZBUFF_recommendedDInSize(void);
+ZBUFF_DEPRECATED("use ZSTD_DStreamOutSize") size_t ZBUFF_recommendedDOutSize(void);
+
+#endif  /* ZSTD_BUFFERED_H_23987 */
 
 
 #ifdef ZBUFF_STATIC_LINKING_ONLY
+#ifndef ZBUFF_STATIC_H_30298098432
+#define ZBUFF_STATIC_H_30298098432
 
 /* ====================================================================================
  * The definitions in this section are considered experimental.
@@ -169,23 +190,23 @@ ZSTDLIB_API size_t ZBUFF_recommendedDOutSize(void);
 /*--- Custom memory allocator ---*/
 /*! ZBUFF_createCCtx_advanced() :
  *  Create a ZBUFF compression context using external alloc and free functions */
-ZSTDLIB_API ZBUFF_CCtx* ZBUFF_createCCtx_advanced(ZSTD_customMem customMem);
+ZBUFF_DEPRECATED("use ZSTD_createCStream_advanced") ZBUFF_CCtx* ZBUFF_createCCtx_advanced(ZSTD_customMem customMem);
 
 /*! ZBUFF_createDCtx_advanced() :
  *  Create a ZBUFF decompression context using external alloc and free functions */
-ZSTDLIB_API ZBUFF_DCtx* ZBUFF_createDCtx_advanced(ZSTD_customMem customMem);
+ZBUFF_DEPRECATED("use ZSTD_createDStream_advanced") ZBUFF_DCtx* ZBUFF_createDCtx_advanced(ZSTD_customMem customMem);
 
 
 /*--- Advanced Streaming Initialization ---*/
-ZSTDLIB_API size_t ZBUFF_compressInit_advanced(ZBUFF_CCtx* zbc,
+ZBUFF_DEPRECATED("use ZSTD_initDStream_usingDict") size_t ZBUFF_compressInit_advanced(ZBUFF_CCtx* zbc,
                                                const void* dict, size_t dictSize,
                                                ZSTD_parameters params, unsigned long long pledgedSrcSize);
 
+
+#endif    /* ZBUFF_STATIC_H_30298098432 */
 #endif    /* ZBUFF_STATIC_LINKING_ONLY */
 
 
 #if defined (__cplusplus)
 }
 #endif
-
-#endif  /* ZSTD_BUFFERED_H_23987 */

contrib/libzstd/include/zstd/deprecated/zbuff_common.c

@@ -0,0 +1,26 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include "error_private.h"
+#include "zbuff.h"
+
+/*-****************************************
+*  ZBUFF Error Management  (deprecated)
+******************************************/
+
+/*! ZBUFF_isError() :
+*   tells if a return value is an error code */
+unsigned ZBUFF_isError(size_t errorCode) { return ERR_isError(errorCode); }
+/*! ZBUFF_getErrorName() :
+*   provides error code string from function result (useful for debugging) */
+const char* ZBUFF_getErrorName(size_t errorCode) { return ERR_getErrorName(errorCode); }
+

contrib/libzstd/include/zstd/deprecated/zbuff_compress.c

@@ -0,0 +1,145 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+
+/* *************************************
+*  Dependencies
+***************************************/
+#define ZBUFF_STATIC_LINKING_ONLY
+#include "zbuff.h"
+
+
+/*-***********************************************************
+*  Streaming compression
+*
+*  A ZBUFF_CCtx object is required to track streaming operation.
+*  Use ZBUFF_createCCtx() and ZBUFF_freeCCtx() to create/release resources.
+*  Use ZBUFF_compressInit() to start a new compression operation.
+*  ZBUFF_CCtx objects can be reused multiple times.
+*
+*  Use ZBUFF_compressContinue() repetitively to consume your input.
+*  *srcSizePtr and *dstCapacityPtr can be any size.
+*  The function will report how many bytes were read or written by modifying *srcSizePtr and *dstCapacityPtr.
+*  Note that it may not consume the entire input, in which case it's up to the caller to call again the function with remaining input.
+*  The content of dst will be overwritten (up to *dstCapacityPtr) at each function call, so save its content if it matters or change dst .
+*  @return : a hint to preferred nb of bytes to use as input for next function call (it's only a hint, to improve latency)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  ZBUFF_compressFlush() can be used to instruct ZBUFF to compress and output whatever remains within its buffer.
+*  Note that it will not output more than *dstCapacityPtr.
+*  Therefore, some content might still be left into its internal buffer if dst buffer is too small.
+*  @return : nb of bytes still present into internal buffer (0 if it's empty)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  ZBUFF_compressEnd() instructs to finish a frame.
+*  It will perform a flush and write frame epilogue.
+*  Similar to ZBUFF_compressFlush(), it may not be able to output the entire internal buffer content if *dstCapacityPtr is too small.
+*  @return : nb of bytes still present into internal buffer (0 if it's empty)
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  Hint : recommended buffer sizes (not compulsory)
+*  input : ZSTD_BLOCKSIZE_MAX (128 KB), internal unit size, it improves latency to use this value.
+*  output : ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + ZBUFF_endFrameSize : ensures it's always possible to write/flush/end a full block at best speed.
+* ***********************************************************/
+
+ZBUFF_CCtx* ZBUFF_createCCtx(void)
+{
+    return ZSTD_createCStream();
+}
+
+ZBUFF_CCtx* ZBUFF_createCCtx_advanced(ZSTD_customMem customMem)
+{
+    return ZSTD_createCStream_advanced(customMem);
+}
+
+size_t ZBUFF_freeCCtx(ZBUFF_CCtx* zbc)
+{
+    return ZSTD_freeCStream(zbc);
+}
+
+
+/* ======   Initialization   ====== */
+
+size_t ZBUFF_compressInit_advanced(ZBUFF_CCtx* zbc,
+                                   const void* dict, size_t dictSize,
+                                   ZSTD_parameters params, unsigned long long pledgedSrcSize)
+{
+    return ZSTD_initCStream_advanced(zbc, dict, dictSize, params, pledgedSrcSize);
+}
+
+
+size_t ZBUFF_compressInitDictionary(ZBUFF_CCtx* zbc, const void* dict, size_t dictSize, int compressionLevel)
+{
+    return ZSTD_initCStream_usingDict(zbc, dict, dictSize, compressionLevel);
+}
+
+size_t ZBUFF_compressInit(ZBUFF_CCtx* zbc, int compressionLevel)
+{
+    return ZSTD_initCStream(zbc, compressionLevel);
+}
+
+/* ======   Compression   ====== */
+
+
+size_t ZBUFF_compressContinue(ZBUFF_CCtx* zbc,
+                              void* dst, size_t* dstCapacityPtr,
+                        const void* src, size_t* srcSizePtr)
+{
+    size_t result;
+    ZSTD_outBuffer outBuff;
+    ZSTD_inBuffer inBuff;
+    outBuff.dst = dst;
+    outBuff.pos = 0;
+    outBuff.size = *dstCapacityPtr;
+    inBuff.src = src;
+    inBuff.pos = 0;
+    inBuff.size = *srcSizePtr;
+    result = ZSTD_compressStream(zbc, &outBuff, &inBuff);
+    *dstCapacityPtr = outBuff.pos;
+    *srcSizePtr = inBuff.pos;
+    return result;
+}
+
+
+
+/* ======   Finalize   ====== */
+
+size_t ZBUFF_compressFlush(ZBUFF_CCtx* zbc, void* dst, size_t* dstCapacityPtr)
+{
+    size_t result;
+    ZSTD_outBuffer outBuff;
+    outBuff.dst = dst;
+    outBuff.pos = 0;
+    outBuff.size = *dstCapacityPtr;
+    result = ZSTD_flushStream(zbc, &outBuff);
+    *dstCapacityPtr = outBuff.pos;
+    return result;
+}
+
+
+size_t ZBUFF_compressEnd(ZBUFF_CCtx* zbc, void* dst, size_t* dstCapacityPtr)
+{
+    size_t result;
+    ZSTD_outBuffer outBuff;
+    outBuff.dst = dst;
+    outBuff.pos = 0;
+    outBuff.size = *dstCapacityPtr;
+    result = ZSTD_endStream(zbc, &outBuff);
+    *dstCapacityPtr = outBuff.pos;
+    return result;
+}
+
+
+
+/* *************************************
+*  Tool functions
+***************************************/
+size_t ZBUFF_recommendedCInSize(void)  { return ZSTD_CStreamInSize(); }
+size_t ZBUFF_recommendedCOutSize(void) { return ZSTD_CStreamOutSize(); }

contrib/libzstd/include/zstd/deprecated/zbuff_decompress.c

@@ -0,0 +1,74 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+
+
+/* *************************************
+*  Dependencies
+***************************************/
+#define ZBUFF_STATIC_LINKING_ONLY
+#include "zbuff.h"
+
+
+ZBUFF_DCtx* ZBUFF_createDCtx(void)
+{
+    return ZSTD_createDStream();
+}
+
+ZBUFF_DCtx* ZBUFF_createDCtx_advanced(ZSTD_customMem customMem)
+{
+    return ZSTD_createDStream_advanced(customMem);
+}
+
+size_t ZBUFF_freeDCtx(ZBUFF_DCtx* zbd)
+{
+    return ZSTD_freeDStream(zbd);
+}
+
+
+/* *** Initialization *** */
+
+size_t ZBUFF_decompressInitDictionary(ZBUFF_DCtx* zbd, const void* dict, size_t dictSize)
+{
+    return ZSTD_initDStream_usingDict(zbd, dict, dictSize);
+}
+
+size_t ZBUFF_decompressInit(ZBUFF_DCtx* zbd)
+{
+    return ZSTD_initDStream(zbd);
+}
+
+
+/* *** Decompression *** */
+
+size_t ZBUFF_decompressContinue(ZBUFF_DCtx* zbd,
+                                void* dst, size_t* dstCapacityPtr,
+                          const void* src, size_t* srcSizePtr)
+{
+    ZSTD_outBuffer outBuff;
+    ZSTD_inBuffer inBuff;
+    size_t result;
+    outBuff.dst  = dst;
+    outBuff.pos  = 0;
+    outBuff.size = *dstCapacityPtr;
+    inBuff.src  = src;
+    inBuff.pos  = 0;
+    inBuff.size = *srcSizePtr;
+    result = ZSTD_decompressStream(zbd, &outBuff, &inBuff);
+    *dstCapacityPtr = outBuff.pos;
+    *srcSizePtr = inBuff.pos;
+    return result;
+}
+
+
+/* *************************************
+*  Tool functions
+***************************************/
+size_t ZBUFF_recommendedDInSize(void)  { return ZSTD_DStreamInSize(); }
+size_t ZBUFF_recommendedDOutSize(void) { return ZSTD_DStreamOutSize(); }

contrib/libzstd/include/zstd/dictBuilder/zdict.c

@@ -11,8 +11,9 @@
 /*-**************************************
 *  Tuning parameters
 ****************************************/
+#define MINRATIO 4   /* minimum nb of apparition to be selected in dictionary */
 #define ZDICT_MAX_SAMPLES_SIZE (2000U << 20)
-#define ZDICT_MIN_SAMPLES_SIZE 512
+#define ZDICT_MIN_SAMPLES_SIZE (ZDICT_CONTENTSIZE_MIN * MINRATIO)
 
 
 /*-**************************************
@@ -36,12 +37,11 @@
 #include <time.h>          /* clock */
 
 #include "mem.h"           /* read */
-#include "error_private.h"
 #include "fse.h"           /* FSE_normalizeCount, FSE_writeNCount */
 #define HUF_STATIC_LINKING_ONLY
-#include "huf.h"
+#include "huf.h"           /* HUF_buildCTable, HUF_writeCTable */
 #include "zstd_internal.h" /* includes zstd.h */
-#include "xxhash.h"
+#include "xxhash.h"        /* XXH64 */
 #include "divsufsort.h"
 #ifndef ZDICT_STATIC_LINKING_ONLY
 #  define ZDICT_STATIC_LINKING_ONLY
@@ -60,11 +60,8 @@
 
 #define NOISELENGTH 32
 
-#define MINRATIO 4
-static const int g_compressionLevel_default = 5;
+static const int g_compressionLevel_default = 6;
 static const U32 g_selectivity_default = 9;
-static const size_t g_provision_entropySize = 200;
-static const size_t g_min_fast_dictContent = 192;
 
 
 /*-*************************************
@@ -97,7 +94,7 @@ const char* ZDICT_getErrorName(size_t errorCode) { return ERR_getErrorName(error
 unsigned ZDICT_getDictID(const void* dictBuffer, size_t dictSize)
 {
     if (dictSize < 8) return 0;
-    if (MEM_readLE32(dictBuffer) != ZSTD_DICT_MAGIC) return 0;
+    if (MEM_readLE32(dictBuffer) != ZSTD_MAGIC_DICTIONARY) return 0;
     return MEM_readLE32((const char*)dictBuffer + 4);
 }
 
@@ -364,26 +361,41 @@ static dictItem ZDICT_analyzePos(
 }
 
 
+static int isIncluded(const void* in, const void* container, size_t length)
+{
+    const char* const ip = (const char*) in;
+    const char* const into = (const char*) container;
+    size_t u;
+
+    for (u=0; u<length; u++) {  /* works because end of buffer is a noisy guard band */
+        if (ip[u] != into[u]) break;
+    }
+
+    return u==length;
+}
+
 /*! ZDICT_checkMerge
     check if dictItem can be merged, do it if possible
     @return : id of destination elt, 0 if not merged
 */
-static U32 ZDICT_checkMerge(dictItem* table, dictItem elt, U32 eltNbToSkip)
+static U32 ZDICT_tryMerge(dictItem* table, dictItem elt, U32 eltNbToSkip, const void* buffer)
 {
     const U32 tableSize = table->pos;
-    const U32 max = elt.pos + (elt.length-1);
+    const U32 eltEnd = elt.pos + elt.length;
+    const char* const buf = (const char*) buffer;
 
     /* tail overlap */
     U32 u; for (u=1; u<tableSize; u++) {
         if (u==eltNbToSkip) continue;
-        if ((table[u].pos > elt.pos) && (table[u].pos < max)) {  /* overlap */
+        if ((table[u].pos > elt.pos) && (table[u].pos <= eltEnd)) {  /* overlap, existing > new */
             /* append */
-            U32 addedLength = table[u].pos - elt.pos;
+            U32 const addedLength = table[u].pos - elt.pos;
             table[u].length += addedLength;
             table[u].pos = elt.pos;
             table[u].savings += elt.savings * addedLength / elt.length;   /* rough approx */
-            table[u].savings += elt.length / 8;    /* rough approx */
+            table[u].savings += elt.length / 8;    /* rough approx bonus */
             elt = table[u];
+            /* sort : improve rank */
             while ((u>1) && (table[u-1].savings < elt.savings))
             table[u] = table[u-1], u--;
             table[u] = elt;
@@ -393,20 +405,33 @@ static U32 ZDICT_checkMerge(dictItem* table, dictItem elt, U32 eltNbToSkip)
     /* front overlap */
     for (u=1; u<tableSize; u++) {
         if (u==eltNbToSkip) continue;
-        if ((table[u].pos + table[u].length > elt.pos) && (table[u].pos < elt.pos)) {  /* overlap */
+
+        if ((table[u].pos + table[u].length >= elt.pos) && (table[u].pos < elt.pos)) {  /* overlap, existing < new */
             /* append */
-            int addedLength = (elt.pos + elt.length) - (table[u].pos + table[u].length);
-            table[u].savings += elt.length / 8;    /* rough approx */
-            if (addedLength > 0) {   /* otherwise, already included */
+            int const addedLength = (int)eltEnd - (table[u].pos + table[u].length);
+            table[u].savings += elt.length / 8;    /* rough approx bonus */
+            if (addedLength > 0) {   /* otherwise, elt fully included into existing */
                 table[u].length += addedLength;
                 table[u].savings += elt.savings * addedLength / elt.length;   /* rough approx */
             }
+            /* sort : improve rank */
             elt = table[u];
             while ((u>1) && (table[u-1].savings < elt.savings))
                 table[u] = table[u-1], u--;
             table[u] = elt;
             return u;
-    }   }
+        }
+
+        if (MEM_read64(buf + table[u].pos) == MEM_read64(buf + elt.pos + 1)) {
+            if (isIncluded(buf + table[u].pos, buf + elt.pos + 1, table[u].length)) {
+                size_t const addedLength = MAX( (int)elt.length - (int)table[u].length , 1 );
+                table[u].pos = elt.pos;
+                table[u].savings += (U32)(elt.savings * addedLength / elt.length);
+                table[u].length = MIN(elt.length, table[u].length + 1);
+                return u;
+            }
+        }
+    }
 
     return 0;
 }
@@ -424,14 +449,14 @@ static void ZDICT_removeDictItem(dictItem* table, U32 id)
 }
 
 
-static void ZDICT_insertDictItem(dictItem* table, U32 maxSize, dictItem elt)
+static void ZDICT_insertDictItem(dictItem* table, U32 maxSize, dictItem elt, const void* buffer)
 {
     /* merge if possible */
-    U32 mergeId = ZDICT_checkMerge(table, elt, 0);
+    U32 mergeId = ZDICT_tryMerge(table, elt, 0, buffer);
     if (mergeId) {
         U32 newMerge = 1;
         while (newMerge) {
-            newMerge = ZDICT_checkMerge(table, table[mergeId], mergeId);
+            newMerge = ZDICT_tryMerge(table, table[mergeId], mergeId, buffer);
             if (newMerge) ZDICT_removeDictItem(table, mergeId);
             mergeId = newMerge;
         }
@@ -462,7 +487,7 @@ static U32 ZDICT_dictSize(const dictItem* dictList)
 }
 
 
-static size_t ZDICT_trainBuffer(dictItem* dictList, U32 dictListSize,
+static size_t ZDICT_trainBuffer_legacy(dictItem* dictList, U32 dictListSize,
                             const void* const buffer, size_t bufferSize,   /* buffer must end with noisy guard band */
                             const size_t* fileSizes, unsigned nbFiles,
                             U32 minRatio, U32 notificationLevel)
@@ -479,7 +504,7 @@ static size_t ZDICT_trainBuffer(dictItem* dictList, U32 dictListSize,
 #   define DISPLAYUPDATE(l, ...) if (notificationLevel>=l) { \
             if (ZDICT_clockSpan(displayClock) > refreshRate)  \
             { displayClock = clock(); DISPLAY(__VA_ARGS__); \
-            if (notificationLevel>=4) fflush(stdout); } }
+            if (notificationLevel>=4) fflush(stderr); } }
 
     /* init */
     DISPLAYLEVEL(2, "\r%70s\r", "");   /* clean display line */
@@ -520,7 +545,7 @@ static size_t ZDICT_trainBuffer(dictItem* dictList, U32 dictListSize,
             if (doneMarks[cursor]) { cursor++; continue; }
             solution = ZDICT_analyzePos(doneMarks, suffix, reverseSuffix[cursor], buffer, minRatio, notificationLevel);
             if (solution.length==0) { cursor++; continue; }
-            ZDICT_insertDictItem(dictList, dictListSize, solution);
+            ZDICT_insertDictItem(dictList, dictListSize, solution, buffer);
             cursor += solution.length;
             DISPLAYUPDATE(2, "\r%4.2f %% \r", (double)cursor / bufferSize * 100);
     }   }
@@ -551,7 +576,7 @@ typedef struct
 {
     ZSTD_CCtx* ref;
     ZSTD_CCtx* zc;
-    void* workPlace;   /* must be ZSTD_BLOCKSIZE_ABSOLUTEMAX allocated */
+    void* workPlace;   /* must be ZSTD_BLOCKSIZE_MAX allocated */
 } EStats_ress_t;
 
 #define MAXREPOFFSET 1024
@@ -560,15 +585,15 @@ static void ZDICT_countEStats(EStats_ress_t esr, ZSTD_parameters params,
                             U32* countLit, U32* offsetcodeCount, U32* matchlengthCount, U32* litlengthCount, U32* repOffsets,
                             const void* src, size_t srcSize, U32 notificationLevel)
 {
-    size_t const blockSizeMax = MIN (ZSTD_BLOCKSIZE_ABSOLUTEMAX, 1 << params.cParams.windowLog);
+    size_t const blockSizeMax = MIN (ZSTD_BLOCKSIZE_MAX, 1 << params.cParams.windowLog);
     size_t cSize;
 
     if (srcSize > blockSizeMax) srcSize = blockSizeMax;   /* protection vs large samples */
     {  size_t const errorCode = ZSTD_copyCCtx(esr.zc, esr.ref, 0);
             if (ZSTD_isError(errorCode)) { DISPLAYLEVEL(1, "warning : ZSTD_copyCCtx failed \n"); return; }
     }
-    cSize = ZSTD_compressBlock(esr.zc, esr.workPlace, ZSTD_BLOCKSIZE_ABSOLUTEMAX, src, srcSize);
-    if (ZSTD_isError(cSize)) { DISPLAYLEVEL(1, "warning : could not compress sample size %u \n", (U32)srcSize); return; }
+    cSize = ZSTD_compressBlock(esr.zc, esr.workPlace, ZSTD_BLOCKSIZE_MAX, src, srcSize);
+    if (ZSTD_isError(cSize)) { DISPLAYLEVEL(3, "warning : could not compress sample size %u \n", (U32)srcSize); return; }
 
     if (cSize) {  /* if == 0; block is not compressible */
         const seqStore_t* seqStorePtr = ZSTD_getSeqStore(esr.zc);
@@ -609,17 +634,6 @@ static void ZDICT_countEStats(EStats_ress_t esr, ZSTD_parameters params,
     }   }   }
 }
 
-/*
-static size_t ZDICT_maxSampleSize(const size_t* fileSizes, unsigned nbFiles)
-{
-    unsigned u;
-    size_t max=0;
-    for (u=0; u<nbFiles; u++)
-        if (max < fileSizes[u]) max = fileSizes[u];
-    return max;
-}
-*/
-
 static size_t ZDICT_totalSampleSize(const size_t* fileSizes, unsigned nbFiles)
 {
     size_t total=0;
@@ -675,26 +689,26 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     /* init */
     esr.ref = ZSTD_createCCtx();
     esr.zc = ZSTD_createCCtx();
-    esr.workPlace = malloc(ZSTD_BLOCKSIZE_ABSOLUTEMAX);
+    esr.workPlace = malloc(ZSTD_BLOCKSIZE_MAX);
     if (!esr.ref || !esr.zc || !esr.workPlace) {
         eSize = ERROR(memory_allocation);
         DISPLAYLEVEL(1, "Not enough memory \n");
         goto _cleanup;
     }
     if (offcodeMax>OFFCODE_MAX) { eSize = ERROR(dictionary_wrong); goto _cleanup; }   /* too large dictionary */
-    for (u=0; u<256; u++) countLit[u]=1;   /* any character must be described */
-    for (u=0; u<=offcodeMax; u++) offcodeCount[u]=1;
-    for (u=0; u<=MaxML; u++) matchLengthCount[u]=1;
-    for (u=0; u<=MaxLL; u++) litLengthCount[u]=1;
+    for (u=0; u<256; u++) countLit[u] = 1;   /* any character must be described */
+    for (u=0; u<=offcodeMax; u++) offcodeCount[u] = 1;
+    for (u=0; u<=MaxML; u++) matchLengthCount[u] = 1;
+    for (u=0; u<=MaxLL; u++) litLengthCount[u] = 1;
     memset(repOffset, 0, sizeof(repOffset));
     repOffset[1] = repOffset[4] = repOffset[8] = 1;
     memset(bestRepOffset, 0, sizeof(bestRepOffset));
-    if (compressionLevel==0) compressionLevel=g_compressionLevel_default;
+    if (compressionLevel==0) compressionLevel = g_compressionLevel_default;
     params = ZSTD_getParams(compressionLevel, averageSampleSize, dictBufferSize);
     {   size_t const beginResult = ZSTD_compressBegin_advanced(esr.ref, dictBuffer, dictBufferSize, params, 0);
         if (ZSTD_isError(beginResult)) {
+            DISPLAYLEVEL(1, "error : ZSTD_compressBegin_advanced() failed : %s \n", ZSTD_getErrorName(beginResult));
             eSize = ERROR(GENERIC);
-            DISPLAYLEVEL(1, "error : ZSTD_compressBegin_advanced failed \n");
             goto _cleanup;
     }   }
 
@@ -811,7 +825,6 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     MEM_writeLE32(dstPtr+4, repStartValue[1]);
     MEM_writeLE32(dstPtr+8, repStartValue[2]);
 #endif
-    dstPtr += 12;
     eSize += 12;
 
 _cleanup:
@@ -823,26 +836,66 @@ _cleanup:
 }
 
 
-size_t ZDICT_addEntropyTablesFromBuffer_advanced(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
+
+size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
+                          const void* customDictContent, size_t dictContentSize,
                           const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
                           ZDICT_params_t params)
 {
     size_t hSize;
+#define HBUFFSIZE 256   /* should prove large enough for all entropy headers */
+    BYTE header[HBUFFSIZE];
     int const compressionLevel = (params.compressionLevel <= 0) ? g_compressionLevel_default : params.compressionLevel;
     U32 const notificationLevel = params.notificationLevel;
 
+    /* check conditions */
+    if (dictBufferCapacity < dictContentSize) return ERROR(dstSize_tooSmall);
+    if (dictContentSize < ZDICT_CONTENTSIZE_MIN) return ERROR(srcSize_wrong);
+    if (dictBufferCapacity < ZDICT_DICTSIZE_MIN) return ERROR(dstSize_tooSmall);
+
     /* dictionary header */
-    MEM_writeLE32(dictBuffer, ZSTD_DICT_MAGIC);
-    {   U64 const randomID = XXH64((char*)dictBuffer + dictBufferCapacity - dictContentSize, dictContentSize, 0);
+    MEM_writeLE32(header, ZSTD_MAGIC_DICTIONARY);
+    {   U64 const randomID = XXH64(customDictContent, dictContentSize, 0);
         U32 const compliantID = (randomID % ((1U<<31)-32768)) + 32768;
         U32 const dictID = params.dictID ? params.dictID : compliantID;
-        MEM_writeLE32((char*)dictBuffer+4, dictID);
+        MEM_writeLE32(header+4, dictID);
     }
     hSize = 8;
 
     /* entropy tables */
     DISPLAYLEVEL(2, "\r%70s\r", "");   /* clean display line */
     DISPLAYLEVEL(2, "statistics ... \n");
+    {   size_t const eSize = ZDICT_analyzeEntropy(header+hSize, HBUFFSIZE-hSize,
+                                  compressionLevel,
+                                  samplesBuffer, samplesSizes, nbSamples,
+                                  customDictContent, dictContentSize,
+                                  notificationLevel);
+        if (ZDICT_isError(eSize)) return eSize;
+        hSize += eSize;
+    }
+
+    /* copy elements in final buffer ; note : src and dst buffer can overlap */
+    if (hSize + dictContentSize > dictBufferCapacity) dictContentSize = dictBufferCapacity - hSize;
+    {   size_t const dictSize = hSize + dictContentSize;
+        char* dictEnd = (char*)dictBuffer + dictSize;
+        memmove(dictEnd - dictContentSize, customDictContent, dictContentSize);
+        memcpy(dictBuffer, header, hSize);
+        return dictSize;
+    }
+}
+
+
+size_t ZDICT_addEntropyTablesFromBuffer_advanced(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
+                                                 const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
+                                                 ZDICT_params_t params)
+{
+    int const compressionLevel = (params.compressionLevel <= 0) ? g_compressionLevel_default : params.compressionLevel;
+    U32 const notificationLevel = params.notificationLevel;
+    size_t hSize = 8;
+
+    /* calculate entropy tables */
+    DISPLAYLEVEL(2, "\r%70s\r", "");   /* clean display line */
+    DISPLAYLEVEL(2, "statistics ... \n");
     {   size_t const eSize = ZDICT_analyzeEntropy((char*)dictBuffer+hSize, dictBufferCapacity-hSize,
                                   compressionLevel,
                                   samplesBuffer, samplesSizes, nbSamples,
@@ -852,6 +905,13 @@ size_t ZDICT_addEntropyTablesFromBuffer_advanced(void* dictBuffer, size_t dictCo
         hSize += eSize;
     }
 
+    /* add dictionary header (after entropy tables) */
+    MEM_writeLE32(dictBuffer, ZSTD_MAGIC_DICTIONARY);
+    {   U64 const randomID = XXH64((char*)dictBuffer + dictBufferCapacity - dictContentSize, dictContentSize, 0);
+        U32 const compliantID = (randomID % ((1U<<31)-32768)) + 32768;
+        U32 const dictID = params.dictID ? params.dictID : compliantID;
+        MEM_writeLE32((char*)dictBuffer+4, dictID);
+    }
 
     if (hSize + dictContentSize < dictBufferCapacity)
         memmove((char*)dictBuffer + hSize, (char*)dictBuffer + dictBufferCapacity - dictContentSize, dictContentSize);
@@ -859,14 +919,14 @@ size_t ZDICT_addEntropyTablesFromBuffer_advanced(void* dictBuffer, size_t dictCo
 }
 
 
-/*! ZDICT_trainFromBuffer_unsafe() :
+/*! ZDICT_trainFromBuffer_unsafe_legacy() :
 *   Warning : `samplesBuffer` must be followed by noisy guard band.
 *   @return : size of dictionary, or an error code which can be tested with ZDICT_isError()
 */
-size_t ZDICT_trainFromBuffer_unsafe(
+size_t ZDICT_trainFromBuffer_unsafe_legacy(
                             void* dictBuffer, size_t maxDictSize,
                             const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
-                            ZDICT_params_t params)
+                            ZDICT_legacy_params_t params)
 {
     U32 const dictListSize = MAX(MAX(DICTLISTSIZE_DEFAULT, nbSamples), (U32)(maxDictSize/16));
     dictItem* const dictList = (dictItem*)malloc(dictListSize * sizeof(*dictList));
@@ -875,33 +935,35 @@ size_t ZDICT_trainFromBuffer_unsafe(
     size_t const targetDictSize = maxDictSize;
     size_t const samplesBuffSize = ZDICT_totalSampleSize(samplesSizes, nbSamples);
     size_t dictSize = 0;
-    U32 const notificationLevel = params.notificationLevel;
+    U32 const notificationLevel = params.zParams.notificationLevel;
 
     /* checks */
     if (!dictList) return ERROR(memory_allocation);
-    if (maxDictSize <= g_provision_entropySize + g_min_fast_dictContent) { free(dictList); return ERROR(dstSize_tooSmall); }
-    if (samplesBuffSize < ZDICT_MIN_SAMPLES_SIZE) { free(dictList); return 0; }   /* not enough source to create dictionary */
+    if (maxDictSize < ZDICT_DICTSIZE_MIN) { free(dictList); return ERROR(dstSize_tooSmall); }   /* requested dictionary size is too small */
+    if (samplesBuffSize < ZDICT_MIN_SAMPLES_SIZE) { free(dictList); return ERROR(dictionaryCreation_failed); }   /* not enough source to create dictionary */
 
     /* init */
     ZDICT_initDictItem(dictList);
 
     /* build dictionary */
-    ZDICT_trainBuffer(dictList, dictListSize,
+    ZDICT_trainBuffer_legacy(dictList, dictListSize,
                        samplesBuffer, samplesBuffSize,
                        samplesSizes, nbSamples,
                        minRep, notificationLevel);
 
     /* display best matches */
-    if (params.notificationLevel>= 3) {
+    if (params.zParams.notificationLevel>= 3) {
         U32 const nb = MIN(25, dictList[0].pos);
         U32 const dictContentSize = ZDICT_dictSize(dictList);
         U32 u;
-        DISPLAYLEVEL(3, "\n %u segments found, of total size %u \n", dictList[0].pos, dictContentSize);
-        DISPLAYLEVEL(3, "list %u best segments \n", nb);
-        for (u=1; u<=nb; u++) {
-            U32 pos = dictList[u].pos;
-            U32 length = dictList[u].length;
-            U32 printedLength = MIN(40, length);
+        DISPLAYLEVEL(3, "\n %u segments found, of total size %u \n", dictList[0].pos-1, dictContentSize);
+        DISPLAYLEVEL(3, "list %u best segments \n", nb-1);
+        for (u=1; u<nb; u++) {
+            U32 const pos = dictList[u].pos;
+            U32 const length = dictList[u].length;
+            U32 const printedLength = MIN(40, length);
+            if ((pos > samplesBuffSize) || ((pos + length) > samplesBuffSize))
+                return ERROR(GENERIC);   /* should never happen */
             DISPLAYLEVEL(3, "%3u:%3u bytes at pos %8u, savings %7u bytes |",
                          u, length, pos, dictList[u].savings);
             ZDICT_printHex((const char*)samplesBuffer+pos, printedLength);
@@ -911,14 +973,15 @@ size_t ZDICT_trainFromBuffer_unsafe(
 
     /* create dictionary */
     {   U32 dictContentSize = ZDICT_dictSize(dictList);
-        if (dictContentSize < targetDictSize/3) {
+        if (dictContentSize < ZDICT_CONTENTSIZE_MIN) { free(dictList); return ERROR(dictionaryCreation_failed); }   /* dictionary content too small */
+        if (dictContentSize < targetDictSize/4) {
             DISPLAYLEVEL(2, "!  warning : selected content significantly smaller than requested (%u < %u) \n", dictContentSize, (U32)maxDictSize);
+            if (samplesBuffSize < 10 * targetDictSize)
+                DISPLAYLEVEL(2, "!  consider increasing the number of samples (total size : %u MB)\n", (U32)(samplesBuffSize>>20));
             if (minRep > MINRATIO) {
                 DISPLAYLEVEL(2, "!  consider increasing selectivity to produce larger dictionary (-s%u) \n", selectivity+1);
                 DISPLAYLEVEL(2, "!  note : larger dictionaries are not necessarily better, test its efficiency on samples \n");
             }
-            if (samplesBuffSize < 10 * targetDictSize)
-                DISPLAYLEVEL(2, "!  consider increasing the number of samples (total size : %u MB)\n", (U32)(samplesBuffSize>>20));
         }
 
         if ((dictContentSize > targetDictSize*3) && (nbSamples > 2*MINRATIO) && (selectivity>1)) {
@@ -926,7 +989,7 @@ size_t ZDICT_trainFromBuffer_unsafe(
             while ((nbSamples >> proposedSelectivity) <= MINRATIO) { proposedSelectivity--; }
             DISPLAYLEVEL(2, "!  note : calculated dictionary significantly larger than requested (%u > %u) \n", dictContentSize, (U32)maxDictSize);
             DISPLAYLEVEL(2, "!  consider increasing dictionary size, or produce denser dictionary (-s%u) \n", proposedSelectivity);
-            DISPLAYLEVEL(2, "!  always test dictionary efficiency on samples \n");
+            DISPLAYLEVEL(2, "!  always test dictionary efficiency on real samples \n");
         }
 
         /* limit dictionary size */
@@ -952,7 +1015,7 @@ size_t ZDICT_trainFromBuffer_unsafe(
 
         dictSize = ZDICT_addEntropyTablesFromBuffer_advanced(dictBuffer, dictContentSize, maxDictSize,
                                                              samplesBuffer, samplesSizes, nbSamples,
-                                                             params);
+                                                             params.zParams);
     }
 
     /* clean up */
@@ -963,9 +1026,9 @@ size_t ZDICT_trainFromBuffer_unsafe(
 
 /* issue : samplesBuffer need to be followed by a noisy guard band.
 *  work around : duplicate the buffer, and add the noise */
-size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacity,
+size_t ZDICT_trainFromBuffer_legacy(void* dictBuffer, size_t dictBufferCapacity,
                               const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
-                                      ZDICT_params_t params)
+                              ZDICT_legacy_params_t params)
 {
     size_t result;
     void* newBuff;
@@ -978,10 +1041,9 @@ size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacit
     memcpy(newBuff, samplesBuffer, sBuffSize);
     ZDICT_fillNoise((char*)newBuff + sBuffSize, NOISELENGTH);   /* guard band, for end of buffer condition */
 
-    result = ZDICT_trainFromBuffer_unsafe(
-                                        dictBuffer, dictBufferCapacity,
-                                        newBuff, samplesSizes, nbSamples,
-                                        params);
+    result =
+        ZDICT_trainFromBuffer_unsafe_legacy(dictBuffer, dictBufferCapacity, newBuff,
+                                            samplesSizes, nbSamples, params);
     free(newBuff);
     return result;
 }
@@ -990,11 +1052,13 @@ size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacit
 size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
                              const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples)
 {
-    ZDICT_params_t params;
+    ZDICT_cover_params_t params;
     memset(&params, 0, sizeof(params));
-    return ZDICT_trainFromBuffer_advanced(dictBuffer, dictBufferCapacity,
-                                          samplesBuffer, samplesSizes, nbSamples,
-                                          params);
+    params.d = 8;
+    params.steps = 4;
+    return ZDICT_optimizeTrainFromBuffer_cover(dictBuffer, dictBufferCapacity,
+                                               samplesBuffer, samplesSizes,
+                                               nbSamples, &params);
 }
 
 size_t ZDICT_addEntropyTablesFromBuffer(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,

contrib/libzstd/include/zstd/dictBuilder/zdict.h

@@ -19,30 +19,37 @@ extern "C" {
 #include <stddef.h>  /* size_t */
 
 
-/*======  Export for Windows  ======*/
-/*!
-*  ZSTD_DLL_EXPORT :
-*  Enable exporting of functions when building a Windows DLL
-*/
-#if defined(_WIN32) && defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
-#  define ZDICTLIB_API __declspec(dllexport)
+/* =====   ZDICTLIB_API : control library symbols visibility   ===== */
+#ifndef ZDICTLIB_VISIBILITY
+#  if defined(__GNUC__) && (__GNUC__ >= 4)
+#    define ZDICTLIB_VISIBILITY __attribute__ ((visibility ("default")))
+#  else
+#    define ZDICTLIB_VISIBILITY
+#  endif
+#endif
+#if defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
+#  define ZDICTLIB_API __declspec(dllexport) ZDICTLIB_VISIBILITY
+#elif defined(ZSTD_DLL_IMPORT) && (ZSTD_DLL_IMPORT==1)
+#  define ZDICTLIB_API __declspec(dllimport) ZDICTLIB_VISIBILITY /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/
 #else
-#  define ZDICTLIB_API
+#  define ZDICTLIB_API ZDICTLIB_VISIBILITY
 #endif
 
 
-/*! ZDICT_trainFromBuffer() :
-    Train a dictionary from an array of samples.
-    Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
-    supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
-    The resulting dictionary will be saved into `dictBuffer`.
-    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
-              or an error code, which can be tested with ZDICT_isError().
-    Tips : In general, a reasonable dictionary has a size of ~ 100 KB.
-           It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
-           In general, it's recommended to provide a few thousands samples, but this can vary a lot.
-           It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
-*/
+/*! ZDICT_trainFromBuffer():
+ * Train a dictionary from an array of samples.
+ * Uses ZDICT_optimizeTrainFromBuffer_cover() single-threaded, with d=8 and steps=4.
+ * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ * The resulting dictionary will be saved into `dictBuffer`.
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+ *           or an error code, which can be tested with ZDICT_isError().
+ * Note: ZDICT_trainFromBuffer() requires about 9 bytes of memory for each input byte.
+ * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
+ */
 ZDICTLIB_API size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
                        const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
 
@@ -64,44 +71,136 @@ ZDICTLIB_API const char* ZDICT_getErrorName(size_t errorCode);
  * ==================================================================================== */
 
 typedef struct {
-    unsigned selectivityLevel;   /* 0 means default; larger => select more => larger dictionary */
     int      compressionLevel;   /* 0 means default; target a specific zstd compression level */
     unsigned notificationLevel;  /* Write to stderr; 0 = none (default); 1 = errors; 2 = progression; 3 = details; 4 = debug; */
     unsigned dictID;             /* 0 means auto mode (32-bits random value); other : force dictID value */
-    unsigned reserved[2];        /* reserved space for future parameters */
 } ZDICT_params_t;
 
+/*! ZDICT_cover_params_t:
+ *  For all values 0 means default.
+ *  k and d are the only required parameters.
+ */
+typedef struct {
+    unsigned k;                  /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
+    unsigned d;                  /* dmer size : constraint: 0 < d <= k : Reasonable range [6, 16] */
+    unsigned steps;              /* Number of steps : Only used for optimization : 0 means default (32) : Higher means more parameters checked */
+    unsigned nbThreads;          /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
+    ZDICT_params_t zParams;
+} ZDICT_cover_params_t;
 
-/*! ZDICT_trainFromBuffer_advanced() :
-    Same as ZDICT_trainFromBuffer() with control over more parameters.
-    `parameters` is optional and can be provided with values set to 0 to mean "default".
-    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferSize`),
-              or an error code, which can be tested by ZDICT_isError().
-    note : ZDICT_trainFromBuffer_advanced() will send notifications into stderr if instructed to, using notificationLevel>0.
-*/
-size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacity,
+
+/*! ZDICT_trainFromBuffer_cover():
+ * Train a dictionary from an array of samples using the COVER algorithm.
+ * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ * The resulting dictionary will be saved into `dictBuffer`.
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+ *           or an error code, which can be tested with ZDICT_isError().
+ * Note: ZDICT_trainFromBuffer_cover() requires about 9 bytes of memory for each input byte.
+ * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
+ */
+ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
+    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
+    const size_t *samplesSizes, unsigned nbSamples,
+    ZDICT_cover_params_t parameters);
+
+/*! ZDICT_optimizeTrainFromBuffer_cover():
+ * The same requirements as above hold for all the parameters except `parameters`.
+ * This function tries many parameter combinations and picks the best parameters.
+ * `*parameters` is filled with the best parameters found, and the dictionary
+ * constructed with those parameters is stored in `dictBuffer`.
+ *
+ * All of the parameters d, k, steps are optional.
+ * If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8, 10, 12, 14, 16}.
+ * if steps is zero it defaults to its default value.
+ * If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [16, 2048].
+ *
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+ *           or an error code, which can be tested with ZDICT_isError().
+ *           On success `*parameters` contains the parameters selected.
+ * Note: ZDICT_optimizeTrainFromBuffer_cover() requires about 8 bytes of memory for each input byte and additionally another 5 bytes of memory for each byte of memory for each thread.
+ */
+ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
+    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
+    const size_t *samplesSizes, unsigned nbSamples,
+    ZDICT_cover_params_t *parameters);
+
+/*! ZDICT_finalizeDictionary():
+ * Given a custom content as a basis for dictionary, and a set of samples,
+ * finalize dictionary by adding headers and statistics.
+ *
+ * Samples must be stored concatenated in a flat buffer `samplesBuffer`,
+ * supplied with an array of sizes `samplesSizes`, providing the size of each sample in order.
+ *
+ * dictContentSize must be >= ZDICT_CONTENTSIZE_MIN bytes.
+ * maxDictSize must be >= dictContentSize, and must be >= ZDICT_DICTSIZE_MIN bytes.
+ *
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`),
+ *           or an error code, which can be tested by ZDICT_isError().
+ * Note: ZDICT_finalizeDictionary() will push notifications into stderr if instructed to, using notificationLevel>0.
+ * Note 2: dictBuffer and dictContent can overlap
+ */
+#define ZDICT_CONTENTSIZE_MIN 128
+#define ZDICT_DICTSIZE_MIN    256
+ZDICTLIB_API size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
+                                const void* dictContent, size_t dictContentSize,
                                 const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
                                 ZDICT_params_t parameters);
 
+typedef struct {
+    unsigned selectivityLevel;   /* 0 means default; larger => select more => larger dictionary */
+    ZDICT_params_t zParams;
+} ZDICT_legacy_params_t;
 
-/*! ZDICT_addEntropyTablesFromBuffer() :
+/*! ZDICT_trainFromBuffer_legacy():
+ * Train a dictionary from an array of samples.
+ * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ * The resulting dictionary will be saved into `dictBuffer`.
+ * `parameters` is optional and can be provided with values set to 0 to mean "default".
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+ *           or an error code, which can be tested with ZDICT_isError().
+ * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
+ * Note: ZDICT_trainFromBuffer_legacy() will send notifications into stderr if instructed to, using notificationLevel>0.
+ */
+ZDICTLIB_API size_t ZDICT_trainFromBuffer_legacy(
+    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
+    const size_t *samplesSizes, unsigned nbSamples, ZDICT_legacy_params_t parameters);
 
-    Given a content-only dictionary (built using any 3rd party algorithm),
-    add entropy tables computed from an array of samples.
-    Samples must be stored concatenated in a flat buffer `samplesBuffer`,
-    supplied with an array of sizes `samplesSizes`, providing the size of each sample in order.
+/* Deprecation warnings */
+/* It is generally possible to disable deprecation warnings from compiler,
+   for example with -Wno-deprecated-declarations for gcc
+   or _CRT_SECURE_NO_WARNINGS in Visual.
+   Otherwise, it's also possible to manually define ZDICT_DISABLE_DEPRECATE_WARNINGS */
+#ifdef ZDICT_DISABLE_DEPRECATE_WARNINGS
+#  define ZDICT_DEPRECATED(message) ZDICTLIB_API   /* disable deprecation warnings */
+#else
+#  define ZDICT_GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
+#  if defined (__cplusplus) && (__cplusplus >= 201402) /* C++14 or greater */
+#    define ZDICT_DEPRECATED(message) [[deprecated(message)]] ZDICTLIB_API
+#  elif (ZDICT_GCC_VERSION >= 405) || defined(__clang__)
+#    define ZDICT_DEPRECATED(message) ZDICTLIB_API __attribute__((deprecated(message)))
+#  elif (ZDICT_GCC_VERSION >= 301)
+#    define ZDICT_DEPRECATED(message) ZDICTLIB_API __attribute__((deprecated))
+#  elif defined(_MSC_VER)
+#    define ZDICT_DEPRECATED(message) ZDICTLIB_API __declspec(deprecated(message))
+#  else
+#    pragma message("WARNING: You need to implement ZDICT_DEPRECATED for this compiler")
+#    define ZDICT_DEPRECATED(message) ZDICTLIB_API
+#  endif
+#endif /* ZDICT_DISABLE_DEPRECATE_WARNINGS */
 
-    The input dictionary content must be stored *at the end* of `dictBuffer`.
-    Its size is `dictContentSize`.
-    The resulting dictionary with added entropy tables will be *written back to `dictBuffer`*,
-    starting from its beginning.
-    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`).
-*/
+ZDICT_DEPRECATED("use ZDICT_finalizeDictionary() instead")
 size_t ZDICT_addEntropyTablesFromBuffer(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
                                   const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
 
 
-
 #endif   /* ZDICT_STATIC_LINKING_ONLY */
 
 #if defined (__cplusplus)

contrib/libzstd/include/zstd/legacy/zstd_legacy.h

@@ -7,8 +7,6 @@
  * of patent rights can be found in the PATENTS file in the same directory.
  */
 
-/// Milovidov: we had only used version 6 in ClickHouse.
-
 #ifndef ZSTD_LEGACY_H
 #define ZSTD_LEGACY_H
 
@@ -22,8 +20,33 @@ extern "C" {
 #include "mem.h"            /* MEM_STATIC */
 #include "error_private.h"  /* ERROR */
 #include "zstd.h"           /* ZSTD_inBuffer, ZSTD_outBuffer */
-#include "zstd_v06.h"
 
+#if !defined (ZSTD_LEGACY_SUPPORT) || (ZSTD_LEGACY_SUPPORT == 0)
+#  undef ZSTD_LEGACY_SUPPORT
+#  define ZSTD_LEGACY_SUPPORT 8
+#endif
+
+#if (ZSTD_LEGACY_SUPPORT <= 1)
+#  include "zstd_v01.h"
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 2)
+#  include "zstd_v02.h"
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 3)
+#  include "zstd_v03.h"
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 4)
+#  include "zstd_v04.h"
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 5)
+#  include "zstd_v05.h"
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 6)
+#  include "zstd_v06.h"
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 7)
+#  include "zstd_v07.h"
+#endif
 
 /** ZSTD_isLegacy() :
     @return : > 0 if supported by legacy decoder. 0 otherwise.
@@ -36,7 +59,27 @@ MEM_STATIC unsigned ZSTD_isLegacy(const void* src, size_t srcSize)
     magicNumberLE = MEM_readLE32(src);
     switch(magicNumberLE)
     {
+#if (ZSTD_LEGACY_SUPPORT <= 1)
+        case ZSTDv01_magicNumberLE:return 1;
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 2)
+        case ZSTDv02_magicNumber : return 2;
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 3)
+        case ZSTDv03_magicNumber : return 3;
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 4)
+        case ZSTDv04_magicNumber : return 4;
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 5)
+        case ZSTDv05_MAGICNUMBER : return 5;
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 6)
         case ZSTDv06_MAGICNUMBER : return 6;
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 7)
+        case ZSTDv07_MAGICNUMBER : return 7;
+#endif
         default : return 0;
     }
 }
@@ -45,12 +88,31 @@ MEM_STATIC unsigned ZSTD_isLegacy(const void* src, size_t srcSize)
 MEM_STATIC unsigned long long ZSTD_getDecompressedSize_legacy(const void* src, size_t srcSize)
 {
     U32 const version = ZSTD_isLegacy(src, srcSize);
+    if (version < 5) return 0;  /* no decompressed size in frame header, or not a legacy format */
+#if (ZSTD_LEGACY_SUPPORT <= 5)
+    if (version==5) {
+        ZSTDv05_parameters fParams;
+        size_t const frResult = ZSTDv05_getFrameParams(&fParams, src, srcSize);
+        if (frResult != 0) return 0;
+        return fParams.srcSize;
+    }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 6)
     if (version==6) {
         ZSTDv06_frameParams fParams;
         size_t const frResult = ZSTDv06_getFrameParams(&fParams, src, srcSize);
         if (frResult != 0) return 0;
         return fParams.frameContentSize;
     }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 7)
+    if (version==7) {
+        ZSTDv07_frameParams fParams;
+        size_t const frResult = ZSTDv07_getFrameParams(&fParams, src, srcSize);
+        if (frResult != 0) return 0;
+        return fParams.frameContentSize;
+    }
+#endif
     return 0;   /* should not be possible */
 }
 
@@ -63,6 +125,33 @@ MEM_STATIC size_t ZSTD_decompressLegacy(
     U32 const version = ZSTD_isLegacy(src, compressedSize);
     switch(version)
     {
+#if (ZSTD_LEGACY_SUPPORT <= 1)
+        case 1 :
+            return ZSTDv01_decompress(dst, dstCapacity, src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 2)
+        case 2 :
+            return ZSTDv02_decompress(dst, dstCapacity, src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 3)
+        case 3 :
+            return ZSTDv03_decompress(dst, dstCapacity, src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 4)
+        case 4 :
+            return ZSTDv04_decompress(dst, dstCapacity, src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 5)
+        case 5 :
+            {   size_t result;
+                ZSTDv05_DCtx* const zd = ZSTDv05_createDCtx();
+                if (zd==NULL) return ERROR(memory_allocation);
+                result = ZSTDv05_decompress_usingDict(zd, dst, dstCapacity, src, compressedSize, dict, dictSize);
+                ZSTDv05_freeDCtx(zd);
+                return result;
+            }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 6)
         case 6 :
             {   size_t result;
                 ZSTDv06_DCtx* const zd = ZSTDv06_createDCtx();
@@ -71,19 +160,82 @@ MEM_STATIC size_t ZSTD_decompressLegacy(
                 ZSTDv06_freeDCtx(zd);
                 return result;
             }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 7)
+        case 7 :
+            {   size_t result;
+                ZSTDv07_DCtx* const zd = ZSTDv07_createDCtx();
+                if (zd==NULL) return ERROR(memory_allocation);
+                result = ZSTDv07_decompress_usingDict(zd, dst, dstCapacity, src, compressedSize, dict, dictSize);
+                ZSTDv07_freeDCtx(zd);
+                return result;
+            }
+#endif
         default :
             return ERROR(prefix_unknown);
     }
 }
 
+MEM_STATIC size_t ZSTD_findFrameCompressedSizeLegacy(const void *src,
+                                             size_t compressedSize)
+{
+    U32 const version = ZSTD_isLegacy(src, compressedSize);
+    switch(version)
+    {
+#if (ZSTD_LEGACY_SUPPORT <= 1)
+        case 1 :
+            return ZSTDv01_findFrameCompressedSize(src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 2)
+        case 2 :
+            return ZSTDv02_findFrameCompressedSize(src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 3)
+        case 3 :
+            return ZSTDv03_findFrameCompressedSize(src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 4)
+        case 4 :
+            return ZSTDv04_findFrameCompressedSize(src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 5)
+        case 5 :
+            return ZSTDv05_findFrameCompressedSize(src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 6)
+        case 6 :
+            return ZSTDv06_findFrameCompressedSize(src, compressedSize);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 7)
+        case 7 :
+            return ZSTDv07_findFrameCompressedSize(src, compressedSize);
+#endif
+        default :
+            return ERROR(prefix_unknown);
+    }
+}
 
 MEM_STATIC size_t ZSTD_freeLegacyStreamContext(void* legacyContext, U32 version)
 {
     switch(version)
     {
         default :
+        case 1 :
+        case 2 :
+        case 3 :
             return ERROR(version_unsupported);
+#if (ZSTD_LEGACY_SUPPORT <= 4)
+        case 4 : return ZBUFFv04_freeDCtx((ZBUFFv04_DCtx*)legacyContext);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 5)
+        case 5 : return ZBUFFv05_freeDCtx((ZBUFFv05_DCtx*)legacyContext);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 6)
         case 6 : return ZBUFFv06_freeDCtx((ZBUFFv06_DCtx*)legacyContext);
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 7)
+        case 7 : return ZBUFFv07_freeDCtx((ZBUFFv07_DCtx*)legacyContext);
+#endif
     }
 }
 
@@ -95,7 +247,32 @@ MEM_STATIC size_t ZSTD_initLegacyStream(void** legacyContext, U32 prevVersion, U
     switch(newVersion)
     {
         default :
+        case 1 :
+        case 2 :
+        case 3 :
             return 0;
+#if (ZSTD_LEGACY_SUPPORT <= 4)
+        case 4 :
+        {
+            ZBUFFv04_DCtx* dctx = (prevVersion != newVersion) ? ZBUFFv04_createDCtx() : (ZBUFFv04_DCtx*)*legacyContext;
+            if (dctx==NULL) return ERROR(memory_allocation);
+            ZBUFFv04_decompressInit(dctx);
+            ZBUFFv04_decompressWithDictionary(dctx, dict, dictSize);
+            *legacyContext = dctx;
+            return 0;
+        }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 5)
+        case 5 :
+        {
+            ZBUFFv05_DCtx* dctx = (prevVersion != newVersion) ? ZBUFFv05_createDCtx() : (ZBUFFv05_DCtx*)*legacyContext;
+            if (dctx==NULL) return ERROR(memory_allocation);
+            ZBUFFv05_decompressInitDictionary(dctx, dict, dictSize);
+            *legacyContext = dctx;
+            return 0;
+        }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 6)
         case 6 :
         {
             ZBUFFv06_DCtx* dctx = (prevVersion != newVersion) ? ZBUFFv06_createDCtx() : (ZBUFFv06_DCtx*)*legacyContext;
@@ -104,6 +281,17 @@ MEM_STATIC size_t ZSTD_initLegacyStream(void** legacyContext, U32 prevVersion, U
             *legacyContext = dctx;
             return 0;
         }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 7)
+        case 7 :
+        {
+            ZBUFFv07_DCtx* dctx = (prevVersion != newVersion) ? ZBUFFv07_createDCtx() : (ZBUFFv07_DCtx*)*legacyContext;
+            if (dctx==NULL) return ERROR(memory_allocation);
+            ZBUFFv07_decompressInitDictionary(dctx, dict, dictSize);
+            *legacyContext = dctx;
+            return 0;
+        }
+#endif
     }
 }
 
@@ -115,7 +303,39 @@ MEM_STATIC size_t ZSTD_decompressLegacyStream(void* legacyContext, U32 version,
     switch(version)
     {
         default :
+        case 1 :
+        case 2 :
+        case 3 :
             return ERROR(version_unsupported);
+#if (ZSTD_LEGACY_SUPPORT <= 4)
+        case 4 :
+            {
+                ZBUFFv04_DCtx* dctx = (ZBUFFv04_DCtx*) legacyContext;
+                const void* src = (const char*)input->src + input->pos;
+                size_t readSize = input->size - input->pos;
+                void* dst = (char*)output->dst + output->pos;
+                size_t decodedSize = output->size - output->pos;
+                size_t const hintSize = ZBUFFv04_decompressContinue(dctx, dst, &decodedSize, src, &readSize);
+                output->pos += decodedSize;
+                input->pos += readSize;
+                return hintSize;
+            }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 5)
+        case 5 :
+            {
+                ZBUFFv05_DCtx* dctx = (ZBUFFv05_DCtx*) legacyContext;
+                const void* src = (const char*)input->src + input->pos;
+                size_t readSize = input->size - input->pos;
+                void* dst = (char*)output->dst + output->pos;
+                size_t decodedSize = output->size - output->pos;
+                size_t const hintSize = ZBUFFv05_decompressContinue(dctx, dst, &decodedSize, src, &readSize);
+                output->pos += decodedSize;
+                input->pos += readSize;
+                return hintSize;
+            }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 6)
         case 6 :
             {
                 ZBUFFv06_DCtx* dctx = (ZBUFFv06_DCtx*) legacyContext;
@@ -128,6 +348,21 @@ MEM_STATIC size_t ZSTD_decompressLegacyStream(void* legacyContext, U32 version,
                 input->pos += readSize;
                 return hintSize;
             }
+#endif
+#if (ZSTD_LEGACY_SUPPORT <= 7)
+        case 7 :
+            {
+                ZBUFFv07_DCtx* dctx = (ZBUFFv07_DCtx*) legacyContext;
+                const void* src = (const char*)input->src + input->pos;
+                size_t readSize = input->size - input->pos;
+                void* dst = (char*)output->dst + output->pos;
+                size_t decodedSize = output->size - output->pos;
+                size_t const hintSize = ZBUFFv07_decompressContinue(dctx, dst, &decodedSize, src, &readSize);
+                output->pos += decodedSize;
+                input->pos += readSize;
+                return hintSize;
+            }
+#endif
     }
 }
 

contrib/libzstd/include/zstd/legacy/zstd_v01.h

@@ -0,0 +1,88 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_V01_H_28739879432
+#define ZSTD_V01_H_28739879432
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/* *************************************
+*  Includes
+***************************************/
+#include <stddef.h>   /* size_t */
+
+
+/* *************************************
+*  Simple one-step function
+***************************************/
+/**
+ZSTDv01_decompress() : decompress ZSTD frames compliant with v0.1.x format
+    compressedSize : is the exact source size
+    maxOriginalSize : is the size of the 'dst' buffer, which must be already allocated.
+                      It must be equal or larger than originalSize, otherwise decompression will fail.
+    return : the number of bytes decompressed into destination buffer (originalSize)
+             or an errorCode if it fails (which can be tested using ZSTDv01_isError())
+*/
+size_t ZSTDv01_decompress( void* dst, size_t maxOriginalSize,
+                     const void* src, size_t compressedSize);
+
+/**
+ZSTDv01_getFrameSrcSize() : get the source length of a ZSTD frame compliant with v0.1.x format
+    compressedSize : The size of the 'src' buffer, at least as large as the frame pointed to by 'src'
+    return : the number of bytes that would be read to decompress this frame
+             or an errorCode if it fails (which can be tested using ZSTDv01_isError())
+*/
+size_t ZSTDv01_findFrameCompressedSize(const void* src, size_t compressedSize);
+
+/**
+ZSTDv01_isError() : tells if the result of ZSTDv01_decompress() is an error
+*/
+unsigned ZSTDv01_isError(size_t code);
+
+
+/* *************************************
+*  Advanced functions
+***************************************/
+typedef struct ZSTDv01_Dctx_s ZSTDv01_Dctx;
+ZSTDv01_Dctx* ZSTDv01_createDCtx(void);
+size_t ZSTDv01_freeDCtx(ZSTDv01_Dctx* dctx);
+
+size_t ZSTDv01_decompressDCtx(void* ctx,
+                              void* dst, size_t maxOriginalSize,
+                        const void* src, size_t compressedSize);
+
+/* *************************************
+*  Streaming functions
+***************************************/
+size_t ZSTDv01_resetDCtx(ZSTDv01_Dctx* dctx);
+
+size_t ZSTDv01_nextSrcSizeToDecompress(ZSTDv01_Dctx* dctx);
+size_t ZSTDv01_decompressContinue(ZSTDv01_Dctx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize);
+/**
+  Use above functions alternatively.
+  ZSTD_nextSrcSizeToDecompress() tells how much bytes to provide as 'srcSize' to ZSTD_decompressContinue().
+  ZSTD_decompressContinue() will use previous data blocks to improve compression if they are located prior to current block.
+  Result is the number of bytes regenerated within 'dst'.
+  It can be zero, which is not an error; it just means ZSTD_decompressContinue() has decoded some header.
+*/
+
+/* *************************************
+*  Prefix - version detection
+***************************************/
+#define ZSTDv01_magicNumber   0xFD2FB51E   /* Big Endian version */
+#define ZSTDv01_magicNumberLE 0x1EB52FFD   /* Little Endian version */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_V01_H_28739879432 */

contrib/libzstd/include/zstd/legacy/zstd_v02.h

@@ -0,0 +1,87 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_V02_H_4174539423
+#define ZSTD_V02_H_4174539423
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/* *************************************
+*  Includes
+***************************************/
+#include <stddef.h>   /* size_t */
+
+
+/* *************************************
+*  Simple one-step function
+***************************************/
+/**
+ZSTDv02_decompress() : decompress ZSTD frames compliant with v0.2.x format
+    compressedSize : is the exact source size
+    maxOriginalSize : is the size of the 'dst' buffer, which must be already allocated.
+                      It must be equal or larger than originalSize, otherwise decompression will fail.
+    return : the number of bytes decompressed into destination buffer (originalSize)
+             or an errorCode if it fails (which can be tested using ZSTDv01_isError())
+*/
+size_t ZSTDv02_decompress( void* dst, size_t maxOriginalSize,
+                     const void* src, size_t compressedSize);
+
+/**
+ZSTDv02_getFrameSrcSize() : get the source length of a ZSTD frame compliant with v0.2.x format
+    compressedSize : The size of the 'src' buffer, at least as large as the frame pointed to by 'src'
+    return : the number of bytes that would be read to decompress this frame
+             or an errorCode if it fails (which can be tested using ZSTDv02_isError())
+*/
+size_t ZSTDv02_findFrameCompressedSize(const void* src, size_t compressedSize);
+
+/**
+ZSTDv02_isError() : tells if the result of ZSTDv02_decompress() is an error
+*/
+unsigned ZSTDv02_isError(size_t code);
+
+
+/* *************************************
+*  Advanced functions
+***************************************/
+typedef struct ZSTDv02_Dctx_s ZSTDv02_Dctx;
+ZSTDv02_Dctx* ZSTDv02_createDCtx(void);
+size_t ZSTDv02_freeDCtx(ZSTDv02_Dctx* dctx);
+
+size_t ZSTDv02_decompressDCtx(void* ctx,
+                              void* dst, size_t maxOriginalSize,
+                        const void* src, size_t compressedSize);
+
+/* *************************************
+*  Streaming functions
+***************************************/
+size_t ZSTDv02_resetDCtx(ZSTDv02_Dctx* dctx);
+
+size_t ZSTDv02_nextSrcSizeToDecompress(ZSTDv02_Dctx* dctx);
+size_t ZSTDv02_decompressContinue(ZSTDv02_Dctx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize);
+/**
+  Use above functions alternatively.
+  ZSTD_nextSrcSizeToDecompress() tells how much bytes to provide as 'srcSize' to ZSTD_decompressContinue().
+  ZSTD_decompressContinue() will use previous data blocks to improve compression if they are located prior to current block.
+  Result is the number of bytes regenerated within 'dst'.
+  It can be zero, which is not an error; it just means ZSTD_decompressContinue() has decoded some header.
+*/
+
+/* *************************************
+*  Prefix - version detection
+***************************************/
+#define ZSTDv02_magicNumber 0xFD2FB522   /* v0.2 */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_V02_H_4174539423 */

contrib/libzstd/include/zstd/legacy/zstd_v03.h

@@ -0,0 +1,87 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_V03_H_298734209782
+#define ZSTD_V03_H_298734209782
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/* *************************************
+*  Includes
+***************************************/
+#include <stddef.h>   /* size_t */
+
+
+/* *************************************
+*  Simple one-step function
+***************************************/
+/**
+ZSTDv03_decompress() : decompress ZSTD frames compliant with v0.3.x format
+    compressedSize : is the exact source size
+    maxOriginalSize : is the size of the 'dst' buffer, which must be already allocated.
+                      It must be equal or larger than originalSize, otherwise decompression will fail.
+    return : the number of bytes decompressed into destination buffer (originalSize)
+             or an errorCode if it fails (which can be tested using ZSTDv01_isError())
+*/
+size_t ZSTDv03_decompress( void* dst, size_t maxOriginalSize,
+                     const void* src, size_t compressedSize);
+
+/**
+ZSTDv03_getFrameSrcSize() : get the source length of a ZSTD frame compliant with v0.3.x format
+    compressedSize : The size of the 'src' buffer, at least as large as the frame pointed to by 'src'
+    return : the number of bytes that would be read to decompress this frame
+             or an errorCode if it fails (which can be tested using ZSTDv03_isError())
+*/
+size_t ZSTDv03_findFrameCompressedSize(const void* src, size_t compressedSize);
+
+    /**
+ZSTDv03_isError() : tells if the result of ZSTDv03_decompress() is an error
+*/
+unsigned ZSTDv03_isError(size_t code);
+
+
+/* *************************************
+*  Advanced functions
+***************************************/
+typedef struct ZSTDv03_Dctx_s ZSTDv03_Dctx;
+ZSTDv03_Dctx* ZSTDv03_createDCtx(void);
+size_t ZSTDv03_freeDCtx(ZSTDv03_Dctx* dctx);
+
+size_t ZSTDv03_decompressDCtx(void* ctx,
+                              void* dst, size_t maxOriginalSize,
+                        const void* src, size_t compressedSize);
+
+/* *************************************
+*  Streaming functions
+***************************************/
+size_t ZSTDv03_resetDCtx(ZSTDv03_Dctx* dctx);
+
+size_t ZSTDv03_nextSrcSizeToDecompress(ZSTDv03_Dctx* dctx);
+size_t ZSTDv03_decompressContinue(ZSTDv03_Dctx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize);
+/**
+  Use above functions alternatively.
+  ZSTD_nextSrcSizeToDecompress() tells how much bytes to provide as 'srcSize' to ZSTD_decompressContinue().
+  ZSTD_decompressContinue() will use previous data blocks to improve compression if they are located prior to current block.
+  Result is the number of bytes regenerated within 'dst'.
+  It can be zero, which is not an error; it just means ZSTD_decompressContinue() has decoded some header.
+*/
+
+/* *************************************
+*  Prefix - version detection
+***************************************/
+#define ZSTDv03_magicNumber 0xFD2FB523   /* v0.3 */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_V03_H_298734209782 */

contrib/libzstd/include/zstd/legacy/zstd_v04.h

@@ -0,0 +1,136 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTD_V04_H_91868324769238
+#define ZSTD_V04_H_91868324769238
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/* *************************************
+*  Includes
+***************************************/
+#include <stddef.h>   /* size_t */
+
+
+/* *************************************
+*  Simple one-step function
+***************************************/
+/**
+ZSTDv04_decompress() : decompress ZSTD frames compliant with v0.4.x format
+    compressedSize : is the exact source size
+    maxOriginalSize : is the size of the 'dst' buffer, which must be already allocated.
+                      It must be equal or larger than originalSize, otherwise decompression will fail.
+    return : the number of bytes decompressed into destination buffer (originalSize)
+             or an errorCode if it fails (which can be tested using ZSTDv01_isError())
+*/
+size_t ZSTDv04_decompress( void* dst, size_t maxOriginalSize,
+                     const void* src, size_t compressedSize);
+
+/**
+ZSTDv04_getFrameSrcSize() : get the source length of a ZSTD frame compliant with v0.4.x format
+    compressedSize : The size of the 'src' buffer, at least as large as the frame pointed to by 'src'
+    return : the number of bytes that would be read to decompress this frame
+             or an errorCode if it fails (which can be tested using ZSTDv04_isError())
+*/
+size_t ZSTDv04_findFrameCompressedSize(const void* src, size_t compressedSize);
+
+/**
+ZSTDv04_isError() : tells if the result of ZSTDv04_decompress() is an error
+*/
+unsigned ZSTDv04_isError(size_t code);
+
+
+/* *************************************
+*  Advanced functions
+***************************************/
+typedef struct ZSTDv04_Dctx_s ZSTDv04_Dctx;
+ZSTDv04_Dctx* ZSTDv04_createDCtx(void);
+size_t ZSTDv04_freeDCtx(ZSTDv04_Dctx* dctx);
+
+size_t ZSTDv04_decompressDCtx(ZSTDv04_Dctx* dctx,
+                              void* dst, size_t maxOriginalSize,
+                        const void* src, size_t compressedSize);
+
+
+/* *************************************
+*  Direct Streaming
+***************************************/
+size_t ZSTDv04_resetDCtx(ZSTDv04_Dctx* dctx);
+
+size_t ZSTDv04_nextSrcSizeToDecompress(ZSTDv04_Dctx* dctx);
+size_t ZSTDv04_decompressContinue(ZSTDv04_Dctx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize);
+/**
+  Use above functions alternatively.
+  ZSTD_nextSrcSizeToDecompress() tells how much bytes to provide as 'srcSize' to ZSTD_decompressContinue().
+  ZSTD_decompressContinue() will use previous data blocks to improve compression if they are located prior to current block.
+  Result is the number of bytes regenerated within 'dst'.
+  It can be zero, which is not an error; it just means ZSTD_decompressContinue() has decoded some header.
+*/
+
+
+/* *************************************
+*  Buffered Streaming
+***************************************/
+typedef struct ZBUFFv04_DCtx_s ZBUFFv04_DCtx;
+ZBUFFv04_DCtx* ZBUFFv04_createDCtx(void);
+size_t         ZBUFFv04_freeDCtx(ZBUFFv04_DCtx* dctx);
+
+size_t ZBUFFv04_decompressInit(ZBUFFv04_DCtx* dctx);
+size_t ZBUFFv04_decompressWithDictionary(ZBUFFv04_DCtx* dctx, const void* dict, size_t dictSize);
+
+size_t ZBUFFv04_decompressContinue(ZBUFFv04_DCtx* dctx, void* dst, size_t* maxDstSizePtr, const void* src, size_t* srcSizePtr);
+
+/** ************************************************
+*  Streaming decompression
+*
+*  A ZBUFF_DCtx object is required to track streaming operation.
+*  Use ZBUFF_createDCtx() and ZBUFF_freeDCtx() to create/release resources.
+*  Use ZBUFF_decompressInit() to start a new decompression operation.
+*  ZBUFF_DCtx objects can be reused multiple times.
+*
+*  Optionally, a reference to a static dictionary can be set, using ZBUFF_decompressWithDictionary()
+*  It must be the same content as the one set during compression phase.
+*  Dictionary content must remain accessible during the decompression process.
+*
+*  Use ZBUFF_decompressContinue() repetitively to consume your input.
+*  *srcSizePtr and *maxDstSizePtr can be any size.
+*  The function will report how many bytes were read or written by modifying *srcSizePtr and *maxDstSizePtr.
+*  Note that it may not consume the entire input, in which case it's up to the caller to present remaining input again.
+*  The content of dst will be overwritten (up to *maxDstSizePtr) at each function call, so save its content if it matters or change dst.
+*  @return : a hint to preferred nb of bytes to use as input for next function call (it's only a hint, to improve latency)
+*            or 0 when a frame is completely decoded
+*            or an error code, which can be tested using ZBUFF_isError().
+*
+*  Hint : recommended buffer sizes (not compulsory) : ZBUFF_recommendedDInSize / ZBUFF_recommendedDOutSize
+*  output : ZBUFF_recommendedDOutSize==128 KB block size is the internal unit, it ensures it's always possible to write a full block when it's decoded.
+*  input : ZBUFF_recommendedDInSize==128Kb+3; just follow indications from ZBUFF_decompressContinue() to minimize latency. It should always be <= 128 KB + 3 .
+* **************************************************/
+unsigned ZBUFFv04_isError(size_t errorCode);
+const char* ZBUFFv04_getErrorName(size_t errorCode);
+
+
+/** The below functions provide recommended buffer sizes for Compression or Decompression operations.
+*   These sizes are not compulsory, they just tend to offer better latency */
+size_t ZBUFFv04_recommendedDInSize(void);
+size_t ZBUFFv04_recommendedDOutSize(void);
+
+
+/* *************************************
+*  Prefix - version detection
+***************************************/
+#define ZSTDv04_magicNumber 0xFD2FB524   /* v0.4 */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_V04_H_91868324769238 */

contrib/libzstd/include/zstd/legacy/zstd_v05.h

@@ -0,0 +1,156 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTDv05_H
+#define ZSTDv05_H
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include <stddef.h>   /* size_t */
+#include "mem.h"      /* U64, U32 */
+
+
+/* *************************************
+*  Simple functions
+***************************************/
+/*! ZSTDv05_decompress() :
+    `compressedSize` : is the _exact_ size of the compressed blob, otherwise decompression will fail.
+    `dstCapacity` must be large enough, equal or larger than originalSize.
+    @return : the number of bytes decompressed into `dst` (<= `dstCapacity`),
+              or an errorCode if it fails (which can be tested using ZSTDv05_isError()) */
+size_t ZSTDv05_decompress( void* dst, size_t dstCapacity,
+                     const void* src, size_t compressedSize);
+
+/**
+ZSTDv05_getFrameSrcSize() : get the source length of a ZSTD frame
+    compressedSize : The size of the 'src' buffer, at least as large as the frame pointed to by 'src'
+    return : the number of bytes that would be read to decompress this frame
+             or an errorCode if it fails (which can be tested using ZSTDv05_isError())
+*/
+size_t ZSTDv05_findFrameCompressedSize(const void* src, size_t compressedSize);
+
+/* *************************************
+*  Helper functions
+***************************************/
+/* Error Management */
+unsigned    ZSTDv05_isError(size_t code);          /*!< tells if a `size_t` function result is an error code */
+const char* ZSTDv05_getErrorName(size_t code);     /*!< provides readable string for an error code */
+
+
+/* *************************************
+*  Explicit memory management
+***************************************/
+/** Decompression context */
+typedef struct ZSTDv05_DCtx_s ZSTDv05_DCtx;
+ZSTDv05_DCtx* ZSTDv05_createDCtx(void);
+size_t ZSTDv05_freeDCtx(ZSTDv05_DCtx* dctx);      /*!< @return : errorCode */
+
+/** ZSTDv05_decompressDCtx() :
+*   Same as ZSTDv05_decompress(), but requires an already allocated ZSTDv05_DCtx (see ZSTDv05_createDCtx()) */
+size_t ZSTDv05_decompressDCtx(ZSTDv05_DCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+
+
+/*-***********************
+*  Simple Dictionary API
+*************************/
+/*! ZSTDv05_decompress_usingDict() :
+*   Decompression using a pre-defined Dictionary content (see dictBuilder).
+*   Dictionary must be identical to the one used during compression, otherwise regenerated data will be corrupted.
+*   Note : dict can be NULL, in which case, it's equivalent to ZSTDv05_decompressDCtx() */
+size_t ZSTDv05_decompress_usingDict(ZSTDv05_DCtx* dctx,
+                                            void* dst, size_t dstCapacity,
+                                      const void* src, size_t srcSize,
+                                      const void* dict,size_t dictSize);
+
+/*-************************
+*  Advanced Streaming API
+***************************/
+typedef enum { ZSTDv05_fast, ZSTDv05_greedy, ZSTDv05_lazy, ZSTDv05_lazy2, ZSTDv05_btlazy2, ZSTDv05_opt, ZSTDv05_btopt } ZSTDv05_strategy;
+typedef struct {
+    U64 srcSize;
+    U32 windowLog;     /* the only useful information to retrieve */
+    U32 contentLog; U32 hashLog; U32 searchLog; U32 searchLength; U32 targetLength; ZSTDv05_strategy strategy;
+} ZSTDv05_parameters;
+size_t ZSTDv05_getFrameParams(ZSTDv05_parameters* params, const void* src, size_t srcSize);
+
+size_t ZSTDv05_decompressBegin_usingDict(ZSTDv05_DCtx* dctx, const void* dict, size_t dictSize);
+void   ZSTDv05_copyDCtx(ZSTDv05_DCtx* dstDCtx, const ZSTDv05_DCtx* srcDCtx);
+size_t ZSTDv05_nextSrcSizeToDecompress(ZSTDv05_DCtx* dctx);
+size_t ZSTDv05_decompressContinue(ZSTDv05_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+
+
+/*-***********************
+*  ZBUFF API
+*************************/
+typedef struct ZBUFFv05_DCtx_s ZBUFFv05_DCtx;
+ZBUFFv05_DCtx* ZBUFFv05_createDCtx(void);
+size_t         ZBUFFv05_freeDCtx(ZBUFFv05_DCtx* dctx);
+
+size_t ZBUFFv05_decompressInit(ZBUFFv05_DCtx* dctx);
+size_t ZBUFFv05_decompressInitDictionary(ZBUFFv05_DCtx* dctx, const void* dict, size_t dictSize);
+
+size_t ZBUFFv05_decompressContinue(ZBUFFv05_DCtx* dctx,
+                                            void* dst, size_t* dstCapacityPtr,
+                                      const void* src, size_t* srcSizePtr);
+
+/*-***************************************************************************
+*  Streaming decompression
+*
+*  A ZBUFFv05_DCtx object is required to track streaming operations.
+*  Use ZBUFFv05_createDCtx() and ZBUFFv05_freeDCtx() to create/release resources.
+*  Use ZBUFFv05_decompressInit() to start a new decompression operation,
+*   or ZBUFFv05_decompressInitDictionary() if decompression requires a dictionary.
+*  Note that ZBUFFv05_DCtx objects can be reused multiple times.
+*
+*  Use ZBUFFv05_decompressContinue() repetitively to consume your input.
+*  *srcSizePtr and *dstCapacityPtr can be any size.
+*  The function will report how many bytes were read or written by modifying *srcSizePtr and *dstCapacityPtr.
+*  Note that it may not consume the entire input, in which case it's up to the caller to present remaining input again.
+*  The content of @dst will be overwritten (up to *dstCapacityPtr) at each function call, so save its content if it matters or change @dst.
+*  @return : a hint to preferred nb of bytes to use as input for next function call (it's only a hint, to help latency)
+*            or 0 when a frame is completely decoded
+*            or an error code, which can be tested using ZBUFFv05_isError().
+*
+*  Hint : recommended buffer sizes (not compulsory) : ZBUFFv05_recommendedDInSize() / ZBUFFv05_recommendedDOutSize()
+*  output : ZBUFFv05_recommendedDOutSize==128 KB block size is the internal unit, it ensures it's always possible to write a full block when decoded.
+*  input  : ZBUFFv05_recommendedDInSize==128Kb+3; just follow indications from ZBUFFv05_decompressContinue() to minimize latency. It should always be <= 128 KB + 3 .
+* *******************************************************************************/
+
+
+/* *************************************
+*  Tool functions
+***************************************/
+unsigned ZBUFFv05_isError(size_t errorCode);
+const char* ZBUFFv05_getErrorName(size_t errorCode);
+
+/** Functions below provide recommended buffer sizes for Compression or Decompression operations.
+*   These sizes are just hints, and tend to offer better latency */
+size_t ZBUFFv05_recommendedDInSize(void);
+size_t ZBUFFv05_recommendedDOutSize(void);
+
+
+
+/*-*************************************
+*  Constants
+***************************************/
+#define ZSTDv05_MAGICNUMBER 0xFD2FB525   /* v0.5 */
+
+
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif  /* ZSTDv0505_H */

contrib/libzstd/include/zstd/legacy/zstd_v06.c

@@ -537,7 +537,7 @@ static void ZSTDv06_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
 /*! ZSTDv06_wildcopy() :
 *   custom version of memcpy(), can copy up to 7 bytes too many (8 bytes if length==0) */
 #define WILDCOPY_OVERLENGTH 8
-MEM_STATIC void ZSTDv06_wildcopy(void* dst, const void* src, size_t length)
+MEM_STATIC void ZSTDv06_wildcopy(void* dst, const void* src, ptrdiff_t length)
 {
     const BYTE* ip = (const BYTE*)src;
     BYTE* op = (BYTE*)dst;
@@ -910,13 +910,13 @@ MEM_STATIC size_t BITv06_initDStream(BITv06_DStream_t* bitD, const void* srcBuff
         bitD->bitContainer = *(const BYTE*)(bitD->start);
         switch(srcSize)
         {
-            case 7: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[6]) << (sizeof(bitD->bitContainer)*8 - 16);
-            case 6: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[5]) << (sizeof(bitD->bitContainer)*8 - 24);
-            case 5: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[4]) << (sizeof(bitD->bitContainer)*8 - 32);
-            case 4: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[3]) << 24;
-            case 3: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[2]) << 16;
-            case 2: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[1]) <<  8;
-            default:;
+            case 7: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[6]) << (sizeof(bitD->bitContainer)*8 - 16);/* fall-through */
+            case 6: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[5]) << (sizeof(bitD->bitContainer)*8 - 24);/* fall-through */
+            case 5: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[4]) << (sizeof(bitD->bitContainer)*8 - 32);/* fall-through */
+            case 4: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[3]) << 24; /* fall-through */
+            case 3: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[2]) << 16; /* fall-through */
+            case 2: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[1]) <<  8; /* fall-through */
+            default: break;
         }
         { BYTE const lastByte = ((const BYTE*)srcBuffer)[srcSize-1];
           if (lastByte == 0) return ERROR(GENERIC);   /* endMark not present */
@@ -1932,9 +1932,11 @@ MEM_STATIC size_t HUFv06_readStats(BYTE* huffWeight, size_t hwSize, U32* rankSta
 {
     U32 weightTotal;
     const BYTE* ip = (const BYTE*) src;
-    size_t iSize = ip[0];
+    size_t iSize;
     size_t oSize;
 
+    if (!srcSize) return ERROR(srcSize_wrong);
+    iSize = ip[0];
     //memset(huffWeight, 0, hwSize);   /* is not necessary, even though some analyzer complain ... */
 
     if (iSize >= 128)  { /* special header */
@@ -1969,6 +1971,7 @@ MEM_STATIC size_t HUFv06_readStats(BYTE* huffWeight, size_t hwSize, U32* rankSta
             rankStats[huffWeight[n]]++;
             weightTotal += (1 << huffWeight[n]) >> 1;
     }   }
+    if (weightTotal == 0) return ERROR(corruption_detected);
 
     /* get last non-null symbol weight (implied, total must be 2^n) */
     {   U32 const tableLog = BITv06_highbit32(weightTotal) + 1;
@@ -2890,7 +2893,6 @@ struct ZSTDv06_DCtx_s
     ZSTDv06_dStage stage;
     U32 flagRepeatTable;
     const BYTE* litPtr;
-    size_t litBufSize;
     size_t litSize;
     BYTE litBuffer[ZSTDv06_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH];
     BYTE headerBuffer[ZSTDv06_FRAMEHEADERSIZE_MAX];
@@ -3167,8 +3169,8 @@ size_t ZSTDv06_decodeLiteralsBlock(ZSTDv06_DCtx* dctx,
                 return ERROR(corruption_detected);
 
             dctx->litPtr = dctx->litBuffer;
-            dctx->litBufSize = ZSTDv06_BLOCKSIZE_MAX+8;
             dctx->litSize = litSize;
+            memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
             return litCSize + lhSize;
         }
     case IS_PCH:
@@ -3183,13 +3185,14 @@ size_t ZSTDv06_decodeLiteralsBlock(ZSTDv06_DCtx* dctx,
             lhSize=3;
             litSize  = ((istart[0] & 15) << 6) + (istart[1] >> 2);
             litCSize = ((istart[1] &  3) << 8) + istart[2];
+            if (litCSize + lhSize > srcSize) return ERROR(corruption_detected);
 
             {   size_t const errorCode = HUFv06_decompress1X4_usingDTable(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->hufTableX4);
                 if (HUFv06_isError(errorCode)) return ERROR(corruption_detected);
             }
             dctx->litPtr = dctx->litBuffer;
-            dctx->litBufSize = ZSTDv06_BLOCKSIZE_MAX+WILDCOPY_OVERLENGTH;
             dctx->litSize = litSize;
+            memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
             return litCSize + lhSize;
         }
     case IS_RAW:
@@ -3213,13 +3216,12 @@ size_t ZSTDv06_decodeLiteralsBlock(ZSTDv06_DCtx* dctx,
                 if (litSize+lhSize > srcSize) return ERROR(corruption_detected);
                 memcpy(dctx->litBuffer, istart+lhSize, litSize);
                 dctx->litPtr = dctx->litBuffer;
-                dctx->litBufSize = ZSTDv06_BLOCKSIZE_MAX+8;
                 dctx->litSize = litSize;
+                memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
                 return lhSize+litSize;
             }
             /* direct reference into compressed stream */
             dctx->litPtr = istart+lhSize;
-            dctx->litBufSize = srcSize-lhSize;
             dctx->litSize = litSize;
             return lhSize+litSize;
         }
@@ -3241,9 +3243,8 @@ size_t ZSTDv06_decodeLiteralsBlock(ZSTDv06_DCtx* dctx,
                 break;
             }
             if (litSize > ZSTDv06_BLOCKSIZE_MAX) return ERROR(corruption_detected);
-            memset(dctx->litBuffer, istart[lhSize], litSize);
+            memset(dctx->litBuffer, istart[lhSize], litSize + WILDCOPY_OVERLENGTH);
             dctx->litPtr = dctx->litBuffer;
-            dctx->litBufSize = ZSTDv06_BLOCKSIZE_MAX+WILDCOPY_OVERLENGTH;
             dctx->litSize = litSize;
             return lhSize+1;
         }
@@ -3302,11 +3303,14 @@ size_t ZSTDv06_decodeSeqHeaders(int* nbSeqPtr,
     {   int nbSeq = *ip++;
         if (!nbSeq) { *nbSeqPtr=0; return 1; }
         if (nbSeq > 0x7F) {
-            if (nbSeq == 0xFF)
+            if (nbSeq == 0xFF) {
+                if (ip+2 > iend) return ERROR(srcSize_wrong);
                 nbSeq = MEM_readLE16(ip) + LONGNBSEQ, ip+=2;
-            else
+            } else {
+                if (ip >= iend) return ERROR(srcSize_wrong);
                 nbSeq = ((nbSeq-0x80)<<8) + *ip++;
             }
+        }
         *nbSeqPtr = nbSeq;
     }
 
@@ -3431,7 +3435,7 @@ static void ZSTDv06_decodeSequence(seq_t* seq, seqState_t* seqState)
 
 size_t ZSTDv06_execSequence(BYTE* op,
                                 BYTE* const oend, seq_t sequence,
-                                const BYTE** litPtr, const BYTE* const litLimit_8,
+                                const BYTE** litPtr, const BYTE* const litLimit,
                                 const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
 {
     BYTE* const oLitEnd = op + sequence.litLength;
@@ -3444,7 +3448,7 @@ size_t ZSTDv06_execSequence(BYTE* op,
     /* check */
     if (oLitEnd > oend_8) return ERROR(dstSize_tooSmall);   /* last match must start at a minimum distance of 8 from oend */
     if (oMatchEnd > oend) return ERROR(dstSize_tooSmall);   /* overwrite beyond dst buffer */
-    if (iLitEnd > litLimit_8) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
+    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
 
     /* copy Literals */
     ZSTDv06_wildcopy(op, *litPtr, sequence.litLength);   /* note : oLitEnd <= oend-8 : no risk of overwrite beyond oend */
@@ -3466,7 +3470,12 @@ size_t ZSTDv06_execSequence(BYTE* op,
             op = oLitEnd + length1;
             sequence.matchLength -= length1;
             match = base;
+            if (op > oend_8 || sequence.matchLength < MINMATCH) {
+              while (op < oMatchEnd) *op++ = *match++;
+              return sequenceLength;
+            }
     }   }
+    /* Requirement: op <= oend_8 */
 
     /* match within prefix */
     if (sequence.offset < 8) {
@@ -3494,7 +3503,7 @@ size_t ZSTDv06_execSequence(BYTE* op,
         }
         while (op < oMatchEnd) *op++ = *match++;
     } else {
-        ZSTDv06_wildcopy(op, match, sequence.matchLength-8);   /* works even if matchLength < 8 */
+        ZSTDv06_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8);   /* works even if matchLength < 8 */
     }
     return sequenceLength;
 }
@@ -3511,7 +3520,6 @@ static size_t ZSTDv06_decompressSequences(
     BYTE* const oend = ostart + maxDstSize;
     BYTE* op = ostart;
     const BYTE* litPtr = dctx->litPtr;
-    const BYTE* const litLimit_8 = litPtr + dctx->litBufSize - 8;
     const BYTE* const litEnd = litPtr + dctx->litSize;
     FSEv06_DTable* DTableLL = dctx->LLTable;
     FSEv06_DTable* DTableML = dctx->MLTable;
@@ -3555,7 +3563,7 @@ static size_t ZSTDv06_decompressSequences(
                        pos, (U32)sequence.litLength, (U32)sequence.matchLength, (U32)sequence.offset);
 #endif
 
-            {   size_t const oneSeqSize = ZSTDv06_execSequence(op, oend, sequence, &litPtr, litLimit_8, base, vBase, dictEnd);
+            {   size_t const oneSeqSize = ZSTDv06_execSequence(op, oend, sequence, &litPtr, litEnd, base, vBase, dictEnd);
                 if (ZSTDv06_isError(oneSeqSize)) return oneSeqSize;
                 op += oneSeqSize;
         }   }
@@ -3721,6 +3729,37 @@ size_t ZSTDv06_decompress(void* dst, size_t dstCapacity, const void* src, size_t
 #endif
 }
 
+size_t ZSTDv06_findFrameCompressedSize(const void* src, size_t srcSize)
+{
+    const BYTE* ip = (const BYTE*)src;
+    size_t remainingSize = srcSize;
+    blockProperties_t blockProperties = { bt_compressed, 0 };
+
+    /* Frame Header */
+    {   size_t const frameHeaderSize = ZSTDv06_frameHeaderSize(src, ZSTDv06_frameHeaderSize_min);
+        if (ZSTDv06_isError(frameHeaderSize)) return frameHeaderSize;
+        if (MEM_readLE32(src) != ZSTDv06_MAGICNUMBER) return ERROR(prefix_unknown);
+        if (srcSize < frameHeaderSize+ZSTDv06_blockHeaderSize) return ERROR(srcSize_wrong);
+        ip += frameHeaderSize; remainingSize -= frameHeaderSize;
+    }
+
+    /* Loop on each block */
+    while (1) {
+        size_t const cBlockSize = ZSTDv06_getcBlockSize(ip, remainingSize, &blockProperties);
+        if (ZSTDv06_isError(cBlockSize)) return cBlockSize;
+
+        ip += ZSTDv06_blockHeaderSize;
+        remainingSize -= ZSTDv06_blockHeaderSize;
+        if (cBlockSize > remainingSize) return ERROR(srcSize_wrong);
+
+        if (cBlockSize == 0) break;   /* bt_end */
+
+        ip += cBlockSize;
+        remainingSize -= cBlockSize;
+    }
+
+    return ip - (const BYTE*)src;
+}
 
 /*_******************************
 *  Streaming Decompression API
@@ -3750,7 +3789,7 @@ size_t ZSTDv06_decompressContinue(ZSTDv06_DCtx* dctx, void* dst, size_t dstCapac
             return 0;
         }
         dctx->expected = 0;   /* not necessary to copy more */
-
+	/* fall-through */
     case ZSTDds_decodeFrameHeader:
         {   size_t result;
             memcpy(dctx->headerBuffer + ZSTDv06_frameHeaderSize_min, src, dctx->expected);
@@ -3822,9 +3861,10 @@ static size_t ZSTDv06_loadEntropy(ZSTDv06_DCtx* dctx, const void* dict, size_t d
     dictSize -= hSize;
 
     {   short offcodeNCount[MaxOff+1];
-        U32 offcodeMaxValue=MaxOff, offcodeLog=OffFSELog;
+        U32 offcodeMaxValue=MaxOff, offcodeLog;
         offcodeHeaderSize = FSEv06_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dict, dictSize);
         if (FSEv06_isError(offcodeHeaderSize)) return ERROR(dictionary_corrupted);
+        if (offcodeLog > OffFSELog) return ERROR(dictionary_corrupted);
         { size_t const errorCode = FSEv06_buildDTable(dctx->OffTable, offcodeNCount, offcodeMaxValue, offcodeLog);
           if (FSEv06_isError(errorCode)) return ERROR(dictionary_corrupted); }
         dict = (const char*)dict + offcodeHeaderSize;
@@ -3832,9 +3872,10 @@ static size_t ZSTDv06_loadEntropy(ZSTDv06_DCtx* dctx, const void* dict, size_t d
     }
 
     {   short matchlengthNCount[MaxML+1];
-        unsigned matchlengthMaxValue = MaxML, matchlengthLog = MLFSELog;
+        unsigned matchlengthMaxValue = MaxML, matchlengthLog;
         matchlengthHeaderSize = FSEv06_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dict, dictSize);
         if (FSEv06_isError(matchlengthHeaderSize)) return ERROR(dictionary_corrupted);
+        if (matchlengthLog > MLFSELog) return ERROR(dictionary_corrupted);
         { size_t const errorCode = FSEv06_buildDTable(dctx->MLTable, matchlengthNCount, matchlengthMaxValue, matchlengthLog);
           if (FSEv06_isError(errorCode)) return ERROR(dictionary_corrupted); }
         dict = (const char*)dict + matchlengthHeaderSize;
@@ -3842,9 +3883,10 @@ static size_t ZSTDv06_loadEntropy(ZSTDv06_DCtx* dctx, const void* dict, size_t d
     }
 
     {   short litlengthNCount[MaxLL+1];
-        unsigned litlengthMaxValue = MaxLL, litlengthLog = LLFSELog;
+        unsigned litlengthMaxValue = MaxLL, litlengthLog;
         litlengthHeaderSize = FSEv06_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dict, dictSize);
         if (FSEv06_isError(litlengthHeaderSize)) return ERROR(dictionary_corrupted);
+        if (litlengthLog > LLFSELog) return ERROR(dictionary_corrupted);
         { size_t const errorCode = FSEv06_buildDTable(dctx->LLTable, litlengthNCount, litlengthMaxValue, litlengthLog);
           if (FSEv06_isError(errorCode)) return ERROR(dictionary_corrupted); }
     }
@@ -4066,7 +4108,7 @@ size_t ZBUFFv06_decompressContinue(ZBUFFv06_DCtx* zbd,
                     zbd->inBuff = (char*)malloc(blockSize);
                     if (zbd->inBuff == NULL) return ERROR(memory_allocation);
                 }
-                {   size_t const neededOutSize = ((size_t)1 << zbd->fParams.windowLog) + blockSize;
+                {   size_t const neededOutSize = ((size_t)1 << zbd->fParams.windowLog) + blockSize + WILDCOPY_OVERLENGTH * 2;
                     if (zbd->outBuffSize < neededOutSize) {
                         free(zbd->outBuff);
                         zbd->outBuffSize = neededOutSize;
@@ -4074,7 +4116,7 @@ size_t ZBUFFv06_decompressContinue(ZBUFFv06_DCtx* zbd,
                         if (zbd->outBuff == NULL) return ERROR(memory_allocation);
             }   }   }
             zbd->stage = ZBUFFds_read;
-
+	    /* fall-through */
         case ZBUFFds_read:
             {   size_t const neededInSize = ZSTDv06_nextSrcSizeToDecompress(zbd->zd);
                 if (neededInSize==0) {  /* end of frame */
@@ -4096,7 +4138,7 @@ size_t ZBUFFv06_decompressContinue(ZBUFFv06_DCtx* zbd,
                 if (ip==iend) { notDone = 0; break; }   /* no more input */
                 zbd->stage = ZBUFFds_load;
             }
-
+	    /* fall-through */
         case ZBUFFds_load:
             {   size_t const neededInSize = ZSTDv06_nextSrcSizeToDecompress(zbd->zd);
                 size_t const toLoad = neededInSize - zbd->inPos;   /* should always be <= remaining space within inBuff */
@@ -4117,8 +4159,9 @@ size_t ZBUFFv06_decompressContinue(ZBUFFv06_DCtx* zbd,
                     zbd->outEnd = zbd->outStart +  decodedSize;
                     zbd->stage = ZBUFFds_flush;
                     // break; /* ZBUFFds_flush follows */
-            }   }
-
+                }
+	    }
+	    /* fall-through */
         case ZBUFFds_flush:
             {   size_t const toFlushSize = zbd->outEnd - zbd->outStart;
                 size_t const flushedSize = ZBUFFv06_limitCopy(op, oend-op, zbd->outBuff + zbd->outStart, toFlushSize);

contrib/libzstd/include/zstd/legacy/zstd_v06.h

@@ -41,6 +41,13 @@ extern "C" {
 ZSTDLIBv06_API size_t ZSTDv06_decompress( void* dst, size_t dstCapacity,
                                     const void* src, size_t compressedSize);
 
+/**
+ZSTDv06_getFrameSrcSize() : get the source length of a ZSTD frame
+    compressedSize : The size of the 'src' buffer, at least as large as the frame pointed to by 'src'
+    return : the number of bytes that would be read to decompress this frame
+             or an errorCode if it fails (which can be tested using ZSTDv06_isError())
+*/
+size_t ZSTDv06_findFrameCompressedSize(const void* src, size_t compressedSize);
 
 /* *************************************
 *  Helper functions

contrib/libzstd/include/zstd/legacy/zstd_v07.h

@@ -0,0 +1,181 @@
+/**
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree. An additional grant
+ * of patent rights can be found in the PATENTS file in the same directory.
+ */
+
+#ifndef ZSTDv07_H_235446
+#define ZSTDv07_H_235446
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*======  Dependency  ======*/
+#include <stddef.h>   /* size_t */
+
+
+/*======  Export for Windows  ======*/
+/*!
+*  ZSTDv07_DLL_EXPORT :
+*  Enable exporting of functions when building a Windows DLL
+*/
+#if defined(_WIN32) && defined(ZSTDv07_DLL_EXPORT) && (ZSTDv07_DLL_EXPORT==1)
+#  define ZSTDLIBv07_API __declspec(dllexport)
+#else
+#  define ZSTDLIBv07_API
+#endif
+
+
+/* *************************************
+*  Simple API
+***************************************/
+/*! ZSTDv07_getDecompressedSize() :
+*   @return : decompressed size if known, 0 otherwise.
+       note 1 : if `0`, follow up with ZSTDv07_getFrameParams() to know precise failure cause.
+       note 2 : decompressed size could be wrong or intentionally modified !
+                always ensure results fit within application's authorized limits */
+unsigned long long ZSTDv07_getDecompressedSize(const void* src, size_t srcSize);
+
+/*! ZSTDv07_decompress() :
+    `compressedSize` : must be _exact_ size of compressed input, otherwise decompression will fail.
+    `dstCapacity` must be equal or larger than originalSize.
+    @return : the number of bytes decompressed into `dst` (<= `dstCapacity`),
+              or an errorCode if it fails (which can be tested using ZSTDv07_isError()) */
+ZSTDLIBv07_API size_t ZSTDv07_decompress( void* dst, size_t dstCapacity,
+                                    const void* src, size_t compressedSize);
+
+/**
+ZSTDv07_getFrameSrcSize() : get the source length of a ZSTD frame
+    compressedSize : The size of the 'src' buffer, at least as large as the frame pointed to by 'src'
+    return : the number of bytes that would be read to decompress this frame
+             or an errorCode if it fails (which can be tested using ZSTDv07_isError())
+*/
+size_t ZSTDv07_findFrameCompressedSize(const void* src, size_t compressedSize);
+
+/*======  Helper functions  ======*/
+ZSTDLIBv07_API unsigned    ZSTDv07_isError(size_t code);          /*!< tells if a `size_t` function result is an error code */
+ZSTDLIBv07_API const char* ZSTDv07_getErrorName(size_t code);     /*!< provides readable string from an error code */
+
+
+/*-*************************************
+*  Explicit memory management
+***************************************/
+/** Decompression context */
+typedef struct ZSTDv07_DCtx_s ZSTDv07_DCtx;
+ZSTDLIBv07_API ZSTDv07_DCtx* ZSTDv07_createDCtx(void);
+ZSTDLIBv07_API size_t     ZSTDv07_freeDCtx(ZSTDv07_DCtx* dctx);      /*!< @return : errorCode */
+
+/** ZSTDv07_decompressDCtx() :
+*   Same as ZSTDv07_decompress(), requires an allocated ZSTDv07_DCtx (see ZSTDv07_createDCtx()) */
+ZSTDLIBv07_API size_t ZSTDv07_decompressDCtx(ZSTDv07_DCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+
+
+/*-************************
+*  Simple dictionary API
+***************************/
+/*! ZSTDv07_decompress_usingDict() :
+*   Decompression using a pre-defined Dictionary content (see dictBuilder).
+*   Dictionary must be identical to the one used during compression.
+*   Note : This function load the dictionary, resulting in a significant startup time */
+ZSTDLIBv07_API size_t ZSTDv07_decompress_usingDict(ZSTDv07_DCtx* dctx,
+                                                   void* dst, size_t dstCapacity,
+                                             const void* src, size_t srcSize,
+                                             const void* dict,size_t dictSize);
+
+
+/*-**************************
+*  Advanced Dictionary API
+****************************/
+/*! ZSTDv07_createDDict() :
+*   Create a digested dictionary, ready to start decompression operation without startup delay.
+*   `dict` can be released after creation */
+typedef struct ZSTDv07_DDict_s ZSTDv07_DDict;
+ZSTDLIBv07_API ZSTDv07_DDict* ZSTDv07_createDDict(const void* dict, size_t dictSize);
+ZSTDLIBv07_API size_t      ZSTDv07_freeDDict(ZSTDv07_DDict* ddict);
+
+/*! ZSTDv07_decompress_usingDDict() :
+*   Decompression using a pre-digested Dictionary
+*   Faster startup than ZSTDv07_decompress_usingDict(), recommended when same dictionary is used multiple times. */
+ZSTDLIBv07_API size_t ZSTDv07_decompress_usingDDict(ZSTDv07_DCtx* dctx,
+                                                    void* dst, size_t dstCapacity,
+                                              const void* src, size_t srcSize,
+                                              const ZSTDv07_DDict* ddict);
+
+typedef struct {
+    unsigned long long frameContentSize;
+    unsigned windowSize;
+    unsigned dictID;
+    unsigned checksumFlag;
+} ZSTDv07_frameParams;
+
+ZSTDLIBv07_API size_t ZSTDv07_getFrameParams(ZSTDv07_frameParams* fparamsPtr, const void* src, size_t srcSize);   /**< doesn't consume input */
+
+
+
+
+/* *************************************
+*  Streaming functions
+***************************************/
+typedef struct ZBUFFv07_DCtx_s ZBUFFv07_DCtx;
+ZSTDLIBv07_API ZBUFFv07_DCtx* ZBUFFv07_createDCtx(void);
+ZSTDLIBv07_API size_t      ZBUFFv07_freeDCtx(ZBUFFv07_DCtx* dctx);
+
+ZSTDLIBv07_API size_t ZBUFFv07_decompressInit(ZBUFFv07_DCtx* dctx);
+ZSTDLIBv07_API size_t ZBUFFv07_decompressInitDictionary(ZBUFFv07_DCtx* dctx, const void* dict, size_t dictSize);
+
+ZSTDLIBv07_API size_t ZBUFFv07_decompressContinue(ZBUFFv07_DCtx* dctx,
+                                            void* dst, size_t* dstCapacityPtr,
+                                      const void* src, size_t* srcSizePtr);
+
+/*-***************************************************************************
+*  Streaming decompression howto
+*
+*  A ZBUFFv07_DCtx object is required to track streaming operations.
+*  Use ZBUFFv07_createDCtx() and ZBUFFv07_freeDCtx() to create/release resources.
+*  Use ZBUFFv07_decompressInit() to start a new decompression operation,
+*   or ZBUFFv07_decompressInitDictionary() if decompression requires a dictionary.
+*  Note that ZBUFFv07_DCtx objects can be re-init multiple times.
+*
+*  Use ZBUFFv07_decompressContinue() repetitively to consume your input.
+*  *srcSizePtr and *dstCapacityPtr can be any size.
+*  The function will report how many bytes were read or written by modifying *srcSizePtr and *dstCapacityPtr.
+*  Note that it may not consume the entire input, in which case it's up to the caller to present remaining input again.
+*  The content of `dst` will be overwritten (up to *dstCapacityPtr) at each function call, so save its content if it matters, or change `dst`.
+*  @return : a hint to preferred nb of bytes to use as input for next function call (it's only a hint, to help latency),
+*            or 0 when a frame is completely decoded,
+*            or an error code, which can be tested using ZBUFFv07_isError().
+*
+*  Hint : recommended buffer sizes (not compulsory) : ZBUFFv07_recommendedDInSize() and ZBUFFv07_recommendedDOutSize()
+*  output : ZBUFFv07_recommendedDOutSize== 128 KB block size is the internal unit, it ensures it's always possible to write a full block when decoded.
+*  input  : ZBUFFv07_recommendedDInSize == 128KB + 3;
+*           just follow indications from ZBUFFv07_decompressContinue() to minimize latency. It should always be <= 128 KB + 3 .
+* *******************************************************************************/
+
+
+/* *************************************
+*  Tool functions
+***************************************/
+ZSTDLIBv07_API unsigned ZBUFFv07_isError(size_t errorCode);
+ZSTDLIBv07_API const char* ZBUFFv07_getErrorName(size_t errorCode);
+
+/** Functions below provide recommended buffer sizes for Compression or Decompression operations.
+*   These sizes are just hints, they tend to offer better latency */
+ZSTDLIBv07_API size_t ZBUFFv07_recommendedDInSize(void);
+ZSTDLIBv07_API size_t ZBUFFv07_recommendedDOutSize(void);
+
+
+/*-*************************************
+*  Constants
+***************************************/
+#define ZSTDv07_MAGICNUMBER            0xFD2FB527   /* v0.7 */
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif  /* ZSTDv07_H_235446 */

dbms/CMakeLists.txt

@@ -28,6 +28,7 @@ add_subdirectory (src)
 add_library(string_utils
     src/Common/StringUtils.h
     src/Common/StringUtils.cpp)
+target_include_directories (string_utils PRIVATE ${DBMS_INCLUDE_DIR})
 
 set(dbms_headers)
 set(dbms_sources)
@@ -150,7 +151,7 @@ if (NOT CMAKE_BUILD_TYPE STREQUAL "Debug")
         PROPERTIES COMPILE_FLAGS -g0)
 endif ()
 
-if (NOT ARM)
+if (NOT ARCH_ARM)
     set (LINK_LIBRARIES_ONLY_ON_X86_64 cpuid)
 endif()
 
@@ -163,12 +164,13 @@ endif()
 target_link_libraries (dbms
     common
     ${MYSQLXX_LIBRARY}
-    cityhash farmhash metrohash
+    ${FARMHASH_LIBRARIES}
+    ${METROHASH_LIBRARIES}
     ${LZ4_LIBRARY}
     ${ZSTD_LIBRARY}
     ${ZOOKEEPER_LIBRARY}
     string_utils
-    ${DOUBLE_CONVERSION_LIBRARY}
+    ${DOUBLE_CONVERSION_LIBRARIES}
     ${ZLIB_LIBRARIES}
     ${LINK_LIBRARIES_ONLY_ON_X86_64}
     ${RE2_LIBRARY}
@@ -176,7 +178,7 @@ target_link_libraries (dbms
     ${OPENSSL_CRYPTO_LIBRARY}
     ${Boost_SYSTEM_LIBRARY}
     ${Poco_Data_LIBRARY}
-    btrie
+    ${BTRIE_LIBRARIES}
 )
 
 if (Poco_DataODBC_FOUND)
@@ -200,19 +202,18 @@ endif ()
 target_link_libraries (dbms
     ${PLATFORM_LIBS}
     ${CMAKE_DL_LIBS}
-    ${LTDL_LIB}
     ${CMAKE_THREAD_LIBS_INIT}
 )
 
-target_include_directories (dbms BEFORE PRIVATE ${CPUID_INCLUDE_DIR})
-target_include_directories (dbms BEFORE PUBLIC ${DIVIDE_INCLUDE_DIR})
-target_include_directories (dbms BEFORE PRIVATE ${BTRIE_INCLUDE_DIR})
-target_include_directories (dbms BEFORE PRIVATE ${CITYHASH_INCLUDE_DIR})
-target_include_directories (dbms PUBLIC ${MYSQLXX_INCLUDE_DIR})
-target_include_directories (dbms PRIVATE ${POCOEXT_INCLUDE_DIR})
-target_include_directories (dbms PRIVATE ${COMMON_INCLUDE_DIR})
+target_include_directories (dbms BEFORE PRIVATE ${DIVIDE_INCLUDE_DIR})
+target_include_directories (dbms BEFORE PRIVATE ${SPARCEHASH_INCLUDE_DIR})
 target_include_directories (dbms PUBLIC ${DBMS_INCLUDE_DIR})
 
+# only for copy_headers.sh:
+target_include_directories (dbms PRIVATE ${COMMON_INCLUDE_DIR})
+target_include_directories (dbms BEFORE PRIVATE ${DOUBLE_CONVERSION_INCLUDE_DIR})
+
+
 if (ENABLE_TESTS)
     add_subdirectory (tests)
     # attach all dbms gtest sources

dbms/cmake/find_vectorclass.cmake

@@ -3,14 +3,13 @@ option (ENABLE_VECTORCLASS "Faster math functions with vectorclass lib" OFF)
 
 if (ENABLE_VECTORCLASS)
 
-set (VECTORCLASS_INCLUDE_PATHS "${ClickHouse_SOURCE_DIR}/contrib/vectorclass" CACHE STRING "Path of vectorclass library")
-find_path (VECTORCLASS_INCLUDE_DIR NAMES vectorf128.h PATHS ${VECTORCLASS_INCLUDE_PATHS})
+    set (VECTORCLASS_INCLUDE_PATHS "${ClickHouse_SOURCE_DIR}/contrib/vectorclass" CACHE STRING "Path of vectorclass library")
+    find_path (VECTORCLASS_INCLUDE_DIR NAMES vectorf128.h PATHS ${VECTORCLASS_INCLUDE_PATHS})
 
-if (VECTORCLASS_INCLUDE_DIR)
+    if (VECTORCLASS_INCLUDE_DIR)
         set (USE_VECTORCLASS 1)
-    include_directories (BEFORE ${VECTORCLASS_INCLUDE_DIR})
-endif ()
+    endif ()
 
-message (STATUS "Using vectorclass=${ENABLE_VECTORCLASS}: ${VECTORCLASS_INCLUDE_DIR}")
+    message (STATUS "Using vectorclass=${USE_VECTORCLASS}: ${VECTORCLASS_INCLUDE_DIR}")
 
 endif ()

dbms/cmake/version.cmake

@@ -1,6 +1,6 @@
 # This strings autochanged from release_lib.sh:
-set(VERSION_DESCRIBE v1.1.54265-testing)
-set(VERSION_REVISION 54265)
+set(VERSION_DESCRIBE v1.1.54281-testing)
+set(VERSION_REVISION 54281)
 # end of autochange
 
 set (VERSION_MAJOR 1)

dbms/src/AggregateFunctions/AggregateFunctionFactory.cpp

@@ -1,13 +1,17 @@
 #include <AggregateFunctions/AggregateFunctionFactory.h>
-#include <IO/WriteBuffer.h>
-#include <IO/WriteHelpers.h>
+
 #include <DataTypes/DataTypeAggregateFunction.h>
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeNullable.h>
+#include <IO/WriteBuffer.h>
+#include <IO/WriteHelpers.h>
+#include <Interpreters/Context.h>
+
 #include <Common/StringUtils.h>
-#include <Poco/String.h>
 #include <Common/typeid_cast.h>
 
+#include <Poco/String.h>
+
 
 namespace DB
 {

dbms/src/AggregateFunctions/AggregateFunctionFactory.h

@@ -1,14 +1,22 @@
 #pragma once
 
-#include <unordered_map>
 #include <AggregateFunctions/IAggregateFunction.h>
+
 #include <ext/singleton.h>
 
+#include <functional>
+#include <memory>
+#include <string>
+#include <unordered_map>
+#include <vector>
+
 
 namespace DB
 {
 
+class Context;
 class IDataType;
+
 using DataTypePtr = std::shared_ptr<IDataType>;
 using DataTypes = std::vector<DataTypePtr>;
 
@@ -19,22 +27,8 @@ class AggregateFunctionFactory final : public ext::singleton<AggregateFunctionFa
 {
     friend class StorageSystemFunctions;
 
-private:
-    /// No std::function, for smaller object size and less indirection.
-    using Creator = AggregateFunctionPtr(*)(const String & name, const DataTypes & argument_types, const Array & parameters);
-    using AggregateFunctions = std::unordered_map<String, Creator>;
-
 public:
-
-    AggregateFunctionPtr get(
-        const String & name,
-        const DataTypes & argument_types,
-        const Array & parameters = {},
-        int recursion_level = 0) const;
-
-    AggregateFunctionPtr tryGet(const String & name, const DataTypes & argument_types, const Array & parameters = {}) const;
-
-    bool isAggregateFunctionName(const String & name, int recursion_level = 0) const;
+    using Creator = std::function<AggregateFunctionPtr(const String &, const DataTypes &, const Array &)>;
 
     /// For compatibility with SQL, it's possible to specify that certain aggregate function name is case insensitive.
     enum CaseSensitiveness
@@ -43,11 +37,29 @@ public:
         CaseInsensitive
     };
 
-    /// Register an aggregate function by its name.
-    void registerFunction(const String & name, Creator creator, CaseSensitiveness case_sensitiveness = CaseSensitive);
+    /// Register a function by its name.
+    /// No locking, you must register all functions before usage of get.
+    void registerFunction(
+        const String & name,
+        Creator creator,
+        CaseSensitiveness case_sensitiveness = CaseSensitive);
+
+    /// Throws an exception if not found.
+    AggregateFunctionPtr get(
+        const String & name,
+        const DataTypes & argument_types,
+        const Array & parameters = {},
+        int recursion_level = 0) const;
+
+    /// Returns nullptr if not found.
+    AggregateFunctionPtr tryGet(
+        const String & name,
+        const DataTypes & argument_types,
+        const Array & parameters = {}) const;
+
+    bool isAggregateFunctionName(const String & name, int recursion_level = 0) const;
 
 private:
-
     AggregateFunctionPtr getImpl(
         const String & name,
         const DataTypes & argument_types,
@@ -55,6 +67,8 @@ private:
         int recursion_level) const;
 
 private:
+    using AggregateFunctions = std::unordered_map<String, Creator>;
+
     AggregateFunctions aggregate_functions;
 
     /// Case insensitive aggregate functions will be additionally added here with lowercased name.

dbms/src/AggregateFunctions/AggregateFunctionGroupArray.cpp

@@ -8,6 +8,28 @@ namespace DB
 namespace
 {
 
+template <template <typename, typename> class AggregateFunctionTemplate, class Data, typename ... TArgs>
+static IAggregateFunction * createWithNumericOrTimeType(const IDataType & argument_type, TArgs && ... args)
+{
+         if (typeid_cast<const DataTypeDate     *>(&argument_type)) return new AggregateFunctionTemplate<UInt16, Data>(std::forward<TArgs>(args)...);
+    else if (typeid_cast<const DataTypeDateTime *>(&argument_type)) return new AggregateFunctionTemplate<UInt32, Data>(std::forward<TArgs>(args)...);
+    else return createWithNumericType<AggregateFunctionTemplate, Data, TArgs...>(argument_type, std::forward<TArgs>(args)...);
+}
+
+
+template <typename has_limit, typename ... TArgs>
+inline AggregateFunctionPtr createAggregateFunctionGroupArrayImpl(const DataTypePtr & argument_type, TArgs ... args)
+{
+    if (auto res = createWithNumericOrTimeType<GroupArrayNumericImpl, has_limit>(*argument_type, argument_type, std::forward<TArgs>(args)...))
+        return AggregateFunctionPtr(res);
+
+    if (typeid_cast<const DataTypeString *>(argument_type.get()))
+        return std::make_shared<GroupArrayGeneralListImpl<GroupArrayListNodeString, has_limit::value>>(std::forward<TArgs>(args)...);
+
+    return std::make_shared<GroupArrayGeneralListImpl<GroupArrayListNodeGeneral, has_limit::value>>(std::forward<TArgs>(args)...);
+};
+
+
 static AggregateFunctionPtr createAggregateFunctionGroupArray(const std::string & name, const DataTypes & argument_types, const Array & parameters)
 {
     if (argument_types.size() != 1)
@@ -15,7 +37,7 @@ static AggregateFunctionPtr createAggregateFunctionGroupArray(const std::string
             ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
 
     bool limit_size = false;
-    UInt64 max_elems = 0;
+    UInt64 max_elems = std::numeric_limits<UInt64>::max();
 
     if (parameters.empty())
     {
@@ -39,23 +61,9 @@ static AggregateFunctionPtr createAggregateFunctionGroupArray(const std::string
             ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
 
     if (!limit_size)
-    {
-        if (auto res = createWithNumericType<GroupArrayNumericImpl, std::false_type>(*argument_types[0]))
-            return AggregateFunctionPtr(res);
-        else if (typeid_cast<const DataTypeString *>(argument_types[0].get()))
-            return std::make_shared<GroupArrayGeneralListImpl<NodeString, false>>();
+        return createAggregateFunctionGroupArrayImpl<std::false_type>(argument_types[0]);
     else
-            return std::make_shared<GroupArrayGeneralListImpl<NodeGeneral, false>>();
-    }
-    else
-    {
-        if (auto res = createWithNumericType<GroupArrayNumericImpl, std::true_type>(*argument_types[0], max_elems))
-            return AggregateFunctionPtr(res);
-        else if (typeid_cast<const DataTypeString *>(argument_types[0].get()))
-            return std::make_shared<GroupArrayGeneralListImpl<NodeString, true>>(max_elems);
-        else
-            return std::make_shared<GroupArrayGeneralListImpl<NodeGeneral, true>>(max_elems);
-    }
+        return createAggregateFunctionGroupArrayImpl<std::true_type>(argument_types[0], max_elems);
 }
 
 }

dbms/src/AggregateFunctions/AggregateFunctionGroupArray.h

@@ -31,9 +31,6 @@ namespace ErrorCodes
 }
 
 
-namespace
-{
-
 /// A particular case is an implementation for numeric types.
 template <typename T>
 struct GroupArrayNumericData
@@ -51,16 +48,18 @@ class GroupArrayNumericImpl final
     : public IUnaryAggregateFunction<GroupArrayNumericData<T>, GroupArrayNumericImpl<T, Tlimit_num_elems>>
 {
     static constexpr bool limit_num_elems = Tlimit_num_elems::value;
+    DataTypePtr data_type;
     UInt64 max_elems;
 
 public:
-    GroupArrayNumericImpl(UInt64 max_elems_ = std::numeric_limits<UInt64>::max()) : max_elems(max_elems_) {}
+    explicit GroupArrayNumericImpl(const DataTypePtr & data_type_, UInt64 max_elems_ = std::numeric_limits<UInt64>::max())
+            : data_type(data_type_), max_elems(max_elems_) {}
 
     String getName() const override { return "groupArray"; }
 
     DataTypePtr getReturnType() const override
     {
-        return std::make_shared<DataTypeArray>(std::make_shared<DataTypeNumber<T>>());
+        return std::make_shared<DataTypeArray>(data_type);
     }
 
     void setArgument(const DataTypePtr & argument) {}
@@ -144,12 +143,10 @@ public:
 /// General case
 
 
-/// Nodes used to implement linked list for stoarge of groupArray states
-struct NodeString;
-struct NodeGeneral;
+/// Nodes used to implement a linked list for storage of groupArray states
 
 template <typename Node>
-struct NodeBase
+struct GroupArrayListNodeBase
 {
     Node * next;
     UInt64 size; // size of payload
@@ -157,7 +154,7 @@ struct NodeBase
     /// Returns pointer to actual payload
     char * data()
     {
-        static_assert(sizeof(NodeBase) == sizeof(Node));
+        static_assert(sizeof(GroupArrayListNodeBase) == sizeof(Node));
         return reinterpret_cast<char *>(this) + sizeof(Node);
     }
 
@@ -187,9 +184,9 @@ struct NodeBase
     }
 };
 
-struct NodeString : public NodeBase<NodeString>
+struct GroupArrayListNodeString : public GroupArrayListNodeBase<GroupArrayListNodeString>
 {
-    using Node = NodeString;
+    using Node = GroupArrayListNodeString;
 
     /// Create node from string
     static Node * allocate(const IColumn & column, size_t row_num, Arena * arena)
@@ -210,9 +207,9 @@ struct NodeString : public NodeBase<NodeString>
     }
 };
 
-struct NodeGeneral : public NodeBase<NodeGeneral>
+struct GroupArrayListNodeGeneral : public GroupArrayListNodeBase<GroupArrayListNodeGeneral>
 {
-    using Node = NodeGeneral;
+    using Node = GroupArrayListNodeGeneral;
 
     static Node * allocate(const IColumn & column, size_t row_num, Arena * arena)
     {
@@ -265,7 +262,7 @@ public:
     void setParameters(const Array & params) override
     {
         if (!limit_num_elems && !params.empty())
-            throw Exception("This instatintion of " + getName() + "aggregate function doesn't accept any parameters. It is a bug.", ErrorCodes::LOGICAL_ERROR);
+            throw Exception("This instantiation of " + getName() + "aggregate function doesn't accept any parameters. It is a bug.", ErrorCodes::LOGICAL_ERROR);
     }
 
     void setArgument(const DataTypePtr & argument)
@@ -393,7 +390,7 @@ public:
 
         auto & column_data = column_array.getData();
 
-        if (std::is_same<Node, NodeString>::value)
+        if (std::is_same<Node, GroupArrayListNodeString>::value)
         {
             auto & string_offsets = static_cast<ColumnString &>(column_data).getOffsets();
             string_offsets.reserve(string_offsets.size() + data(place).elems);
@@ -413,9 +410,6 @@ public:
     }
 };
 
-}
-
-
 #undef AGGREGATE_FUNCTION_GROUP_ARRAY_MAX_ARRAY_SIZE
 
 }