Disable error as warning

Add one more #include<functional>
Add some include<functional>
2026-07-08 07:05:23 +08:00 · 2019-11-05 10:57:56 -08:00 · 2019-11-05 10:56:44 -08:00 · 2019-10-31 14:34:54 -07:00 · 2019-10-31 14:34:53 -07:00 · 2017-07-17 22:21:04 -07:00
310 changed files with 18245 additions and 5667 deletions
@@ -38,7 +38,7 @@ coverage/COVERAGE_REPORT
 .gdbhistory
 package/
 .phutil_module_cache
-unity
+unity.a
 tags
 rocksdb_dump
 rocksdb_undump
@@ -52,6 +52,8 @@ java/include/org_rocksdb_*.h
 .idea/
 *.iml

+rocksdb.cc
+rocksdb.h
 unity.cc
 java/crossbuild/.vagrant
 .vagrant/
@@ -9,17 +9,32 @@ matrix:
      addons:
         apt:
            sources: ['ubuntu-toolchain-r-test', 'llvm-toolchain-precise-3.6']
-            packages: ['clang-3.6', 'zlib1g-dev', 'libbz2-dev', 'libsnappy-dev']
+            packages: ['clang-3.6', 'clang-format-3.6', 'zlib1g-dev', 'libbz2-dev', 'libsnappy-dev', 'curl']
    - os: osx
      compiler: clang

+install:
+  # Build gflags
+  # TODO(noetzli): Remove when gflags available through Travis
+  - pushd /tmp/ && curl -L https://github.com/gflags/gflags/archive/v2.1.2.tar.gz -o gflags.tar.gz && tar xfz gflags.tar.gz && cd gflags-2.1.2 && cmake . && make && popd
+  # Download clang-format-diff.py to check source code formatting
+  - pushd /tmp/ && curl -L http://llvm.org/svn/llvm-project/cfe/trunk/tools/clang-format/clang-format-diff.py -o clang-format-diff.py && chmod +x clang-format-diff.py && popd
+
 before_script:
-    - if [ -n "${COMPILER}" ]; then CXX=${COMPILER}; fi
-    - if [[ $(uname -s) == 'Darwin' ]]; then brew install gflags snappy; fi
+  # Add gflags to include/library paths
+  # TODO(noetzli): Remove when gflags available through Travis
+  - export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/tmp/gflags-2.1.2/lib"
+  - export LIBRARY_PATH="$LIBRARY_PATH:/tmp/gflags-2.1.2/lib"
+  - export CPLUS_INCLUDE_PATH="$CPLUS_INCLUDE_PATH:/tmp/gflags-2.1.2/include"
+  - if [ -n "${COMPILER}" ]; then CXX=${COMPILER}; fi
+  - if [[ "${TRAVIS_OS_NAME}" == 'osx' ]]; then brew install gflags snappy; fi
+  - ulimit -n 2000 || true

 # Lousy hack to disable use and testing of fallocate, which doesn't behave quite
 # as EnvPosixTest::AllocateTest expects within the Travis OpenVZ environment.
-script: OPT=-DTRAVIS V=1 make -j4 check && OPT=-DTRAVIS V=1 make clean jclean rocksdbjava jtest
+script:
+  - if [[ "${TRAVIS_OS_NAME}" == 'linux' ]]; then OPT=-DTRAVIS CLANG_FORMAT_DIFF=/tmp/clang-format-diff.py make format || true; fi
+  - OPT=-DTRAVIS V=1 make -j4 check && OPT=-DTRAVIS V=1 make clean jclean rocksdbjava jtest && make clean && OPT="-DTRAVIS -DROCKSDB_LITE" V=1 make -j4 check

 notifications:
    email:
@@ -13,11 +13,16 @@
 #        cd build
 # 3. Run cmake to generate project files for Windows, add more options to enable required third-party libraries.
 #    See thirdparty.inc for more information.
-#        cmake -G "Visual Studio 12 Win64" .. <more options>
-# 4. Then build the project in debug mode (you may want to add /m:<N> flag to run msbuild in <N> parallel threads)
-#        msbuild ALL_BUILD.vcxproj
-# 5. And release mode (/m:<N> is also supported)
-#        msbuild ALL_BUILD.vcxproj /p:Configuration=Release
+#        sample command: cmake -G "Visual Studio 12 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 ..
+# 4. Then build the project in debug mode (you may want to add /m[:<N>] flag to run msbuild in <N> parallel threads
+#                                          or simply /m ot use all avail cores)
+#        msbuild rocksdb.sln
+#
+#        rocksdb.sln build features exclusions of test only code in Release. If you build ALL_BUILD then everything
+#        will be attempted but test only code does not build in Release mode.
+#
+# 5. And release mode (/m[:<N>] is also supported)
+#        msbuild rocksdb.sln /p:Configuration=Release
 #

 cmake_minimum_required(VERSION 2.6)
@@ -31,7 +36,14 @@ string(REGEX REPLACE "(..)/(..)/..(..).*" "\\1/\\2/\\3" DATE ${DATE})
 string(REGEX REPLACE "(..):(.....).*" " \\1:\\2" TIME ${TIME})
 string(CONCAT GIT_DATE_TIME ${DATE} ${TIME})

-execute_process(COMMAND $ENV{COMSPEC} " /C git rev-parse HEAD 2>nil" OUTPUT_VARIABLE GIT_SHA)
+find_package(Git)
+
+if (GIT_FOUND AND EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/.git")
+    execute_process(COMMAND $ENV{COMSPEC} /C ${GIT_EXECUTABLE} -C ${CMAKE_CURRENT_SOURCE_DIR} rev-parse HEAD OUTPUT_VARIABLE GIT_SHA)
+else()
+    set(GIT_SHA 0)
+endif()
+
 string(REGEX REPLACE "[^0-9a-f]+" "" GIT_SHA ${GIT_SHA})

 set(BUILD_VERSION_CC ${CMAKE_CURRENT_SOURCE_DIR}/util/build_version.cc)
@@ -45,19 +57,34 @@ add_custom_command(OUTPUT ${BUILD_VERSION_CC}

 add_custom_target(GenerateBuildVersion DEPENDS ${BUILD_VERSION_CC})

-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zi /nologo /W3 /WX /EHsc /GS /fp:precise /Zc:wchar_t /Zc:forScope /Gd /TP /errorReport:queue")
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /wd4018 /wd4100 /wd4101 /wd4127 /wd4189 /wd4200 /wd4244 /wd4267 /wd4296 /wd4305 /wd4307 /wd4309 /wd4512 /wd4701 /wd4702 /wd4800 /wd4804 /wd4996")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zi /nologo  /EHsc /GS /Gd /GR /GF /fp:precise /Zc:wchar_t /Zc:forScope /errorReport:queue")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /W3 /WX /wd4127 /wd4244 /wd4267 /wd4800 /wd4804 /wd4996")

-set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /Od /RTC1 /Gm /MDd")
-set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /O2 /Oi /Gm- /Gy /MD")
+# Used to run CI build and tests so we can run faster
+set(OPTIMIZE_DEBUG_DEFAULT 0)        # Debug build is unoptimized by default use -DOPTDBG=1 to optimize

+if(DEFINED OPTDBG)
+   set(OPTIMIZE_DEBUG ${OPTDBG})
+else()
+   set(OPTIMIZE_DEBUG ${OPTIMIZE_DEBUG_DEFAULT})
+endif()
+
+if((${OPTIMIZE_DEBUG} EQUAL 1))
+   message(STATUS "Debug optimization is enabled")
+   set(CMAKE_CXX_FLAGS_DEBUG "/Oxt /MDd")
+else()
+   set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /Od /RTC1 /Gm /MDd")
+endif()
+
+set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /Oxt /Zp8 /Gm- /Gy /MD")
+
+set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} /DEBUG")
 set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} /DEBUG")

 add_definitions(-DWIN32 -DOS_WIN -D_MBCS -DWIN64)

 include_directories(${PROJECT_SOURCE_DIR})
 include_directories(${PROJECT_SOURCE_DIR}/include)
-include_directories(${PROJECT_SOURCE_DIR}/port)
 include_directories(${PROJECT_SOURCE_DIR}/third-party/gtest-1.7.0/fused-src)

 set(ROCKSDB_LIBS rocksdblib${ARTIFACT_SUFFIX})
@@ -68,6 +95,7 @@ set(LIBS ${ROCKSDB_LIBS} ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})

 add_subdirectory(third-party/gtest-1.7.0/fused-src/gtest)

+# Main library source code
 set(SOURCES
        db/builder.cc
        db/c.cc
@@ -115,6 +143,9 @@ set(SOURCES
        db/write_batch_base.cc
        db/write_controller.cc
        db/write_thread.cc
+        memtable/hash_cuckoo_rep.cc
+        memtable/hash_linklist_rep.cc
+        memtable/hash_skiplist_rep.cc
        port/stack_trace.cc
        port/win/env_win.cc
        port/win/port_win.cc
@@ -138,6 +169,7 @@ set(SOURCES
        table/get_context.cc
        table/iterator.cc
        table/merger.cc
+        table/sst_file_writer.cc
        table/meta_blocks.cc
        table/mock_table.cc
        table/plain_table_builder.cc
@@ -147,6 +179,8 @@ set(SOURCES
        table/plain_table_reader.cc
        table/table_properties.cc
        table/two_level_iterator.cc
+        tools/sst_dump_tool.cc
+        tools/dump/db_dump_tool.cc
        util/arena.cc
        util/auto_roll_logger.cc
        util/bloom.cc
@@ -158,7 +192,6 @@ set(SOURCES
        util/crc32c.cc
        util/db_info_dumper.cc
        util/delete_scheduler_impl.cc
-        util/db_test_util.cc
        util/dynamic_bloom.cc
        util/env.cc
        util/env_hdfs.cc
@@ -167,14 +200,11 @@ set(SOURCES
        util/file_reader_writer.cc
        util/filter_policy.cc
        util/hash.cc
-        util/hash_cuckoo_rep.cc
-        util/hash_linklist_rep.cc
-        util/hash_skiplist_rep.cc
        util/histogram.cc
        util/instrumented_mutex.cc
        util/iostats_context.cc
-        util/ldb_cmd.cc
-        util/ldb_tool.cc
+        tools/ldb_cmd.cc
+        tools/ldb_tool.cc
        util/logging.cc
        util/log_buffer.cc
        util/memenv.cc
@@ -184,12 +214,13 @@ set(SOURCES
        util/options.cc
        util/options_builder.cc
        util/options_helper.cc
+        util/options_parser.cc
+        util/options_sanity_check.cc
        util/perf_context.cc
        util/perf_level.cc
        util/rate_limiter.cc
        util/skiplistrep.cc
        util/slice.cc
-        util/sst_dump_tool.cc
        util/statistics.cc
        util/status.cc
        util/status_message.cc
@@ -200,7 +231,6 @@ set(SOURCES
        util/thread_local.cc
        util/thread_status_impl.cc
        util/thread_status_updater.cc
-        util/thread_status_updater_debug.cc
        util/thread_status_util.cc
        util/thread_status_util_debug.cc
        util/vectorrep.cc
@@ -214,6 +244,7 @@ set(SOURCES
        utilities/flashcache/flashcache.cc
        utilities/geodb/geodb_impl.cc
        utilities/leveldb_options/leveldb_options.cc
+        utilities/memory/memory_util.cc
        utilities/merge_operators/string_append/stringappend.cc
        utilities/merge_operators/string_append/stringappend2.cc
        utilities/merge_operators/put.cc
@@ -234,6 +265,13 @@ set(SOURCES
        utilities/write_batch_with_index/write_batch_with_index_internal.cc
 )

+# For test util library that is build only in DEBUG mode
+# and linked to tests. Add test only code that is not #ifdefed for Release here.
+set(TESTUTIL_SOURCE
+    db/db_test_util.cc
+    util/thread_status_updater_debug.cc
+)
+
 add_library(rocksdblib${ARTIFACT_SUFFIX} ${SOURCES})
 set_target_properties(rocksdblib${ARTIFACT_SUFFIX} PROPERTIES COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/rocksdblib${ARTIFACT_SUFFIX}.pdb")
 add_dependencies(rocksdblib${ARTIFACT_SUFFIX} GenerateBuildVersion)
@@ -248,6 +286,7 @@ set(APPS
        db/memtablerep_bench.cc
        table/table_reader_bench.cc
        tools/db_stress.cc
+        tools/write_stress.cc
        tools/db_repl_stress.cc
        tools/sst_dump.cc
        tools/dump/rocksdb_dump.cc
@@ -255,8 +294,9 @@ set(APPS
        util/cache_bench.cc
 )

+set(C_TESTS db/c_test.c)
+
 set(TESTS
-        db/c_test.c
        db/column_family_test.cc
        db/compact_files_test.cc
        db/compaction_job_test.cc
@@ -275,6 +315,7 @@ set(TESTS
        db/db_universal_compaction_test.cc
        db/db_wal_test.cc
        db/db_tailing_iter_test.cc
+        db/db_table_properties_test.cc
        db/dbformat_test.cc
        db/deletefile_test.cc
        db/fault_injection_test.cc
@@ -283,6 +324,7 @@ set(TESTS
        db/flush_job_test.cc
        db/listener_test.cc
        db/log_test.cc
+        db/manual_compaction_test.cc
        db/memtable_list_test.cc
        db/merge_test.cc
        db/merge_helper_test.cc
@@ -307,7 +349,9 @@ set(TESTS
        table/merger_test.cc
        table/table_test.cc
        tools/db_sanity_test.cc
+        tools/ldb_cmd_test.cc
        tools/reduce_levels_test.cc
+        tools/sst_dump_test.cc
        util/arena_test.cc
        util/autovector_test.cc
        util/auto_roll_logger_test.cc
@@ -322,13 +366,11 @@ set(TESTS
        util/file_reader_writer_test.cc
        util/heap_test.cc
        util/histogram_test.cc
-        util/manual_compaction_test.cc
        util/memenv_test.cc
        util/mock_env_test.cc
        util/options_test.cc
        util/rate_limiter_test.cc
        util/slice_transform_test.cc
-        util/sst_dump_test.cc
        util/thread_list_test.cc
        util/thread_local_test.cc
        utilities/backupable/backupable_db_test.cc
@@ -336,6 +378,7 @@ set(TESTS
        utilities/document/document_db_test.cc
        utilities/document/json_document_test.cc
        utilities/geodb/geodb_test.cc
+        utilities/memory/memory_test.cc
        utilities/merge_operators/string_append/stringappend_test.cc
        utilities/redis/redis_lists_test.cc
        utilities/spatialdb/spatial_db_test.cc
@@ -346,12 +389,51 @@ set(TESTS
        utilities/write_batch_with_index/write_batch_with_index_test.cc
 )

-set(EXES ${APPS} ${TESTS})
+set(EXES ${APPS})

 foreach(sourcefile ${EXES})
    string(REPLACE ".cc" "" exename ${sourcefile})
-    string(REPLACE ".c" "" exename ${exename})
    string(REGEX REPLACE "^((.+)/)+" "" exename ${exename})
    add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile})
    target_link_libraries(${exename}${ARTIFACT_SUFFIX} ${LIBS})
 endforeach(sourcefile ${EXES})
+
+# test utilities are only build in debug
+set(TESTUTILLIB testutillib${ARTIFACT_SUFFIX})
+add_library(${TESTUTILLIB} STATIC ${TESTUTIL_SOURCE})
+set_target_properties(${TESTUTILLIB} PROPERTIES COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/testutillib${ARTIFACT_SUFFIX}.pdb")
+set_target_properties(${TESTUTILLIB}
+      PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
+      )
+
+# Tests are excluded from Release builds
+set(TEST_EXES ${TESTS})
+
+foreach(sourcefile ${TEST_EXES})
+    string(REPLACE ".cc" "" exename ${sourcefile})
+    string(REGEX REPLACE "^((.+)/)+" "" exename ${exename})
+    add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile})
+    set_target_properties(${exename}${ARTIFACT_SUFFIX}
+      PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
+      )
+    target_link_libraries(${exename}${ARTIFACT_SUFFIX} ${LIBS} testutillib${ARTIFACT_SUFFIX})
+endforeach(sourcefile ${TEST_EXES})
+
+# C executables must link to a shared object
+set(C_TEST_EXES ${C_TESTS})
+
+foreach(sourcefile ${C_TEST_EXES})
+    string(REPLACE ".c" "" exename ${sourcefile})
+    string(REGEX REPLACE "^((.+)/)+" "" exename ${exename})
+    add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile})
+    set_target_properties(${exename}${ARTIFACT_SUFFIX}
+      PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
+      EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
+      )
+    target_link_libraries(${exename}${ARTIFACT_SUFFIX} rocksdb${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX})
+endforeach(sourcefile ${C_TEST_EXES})
@@ -1,9 +1,34 @@
 # Rocksdb Change Log

+## 4.2.0 (11/9/2015)
+### New Features
+* Introduce CreateLoggerFromOptions(), this function create a Logger for provided DBOptions. 
+* Add GetAggregatedIntProperty(), which returns the sum of the GetIntProperty of all the column families. 
+* Add MemoryUtil in rocksdb/utilities/memory.h.  It currently offers a way to get the memory usage by type from a list rocksdb instances.
+### Public API Changes
+* CompactionFilter::Context includes information of Column Family ID
+* The need-compaction hint given by TablePropertiesCollector::NeedCompact() will be persistent and recoverable after DB recovery. This introduces a breaking format change. If you use this experimental feature, including NewCompactOnDeletionCollectorFactory() in the new version, you may not be able to directly downgrade the DB back to version 4.0 or lower.
+* TablePropertiesCollectorFactory::CreateTablePropertiesCollector() now takes an option Context, containing the information of column family ID for the file being written.
+* Remove DefaultCompactionFilterFactory.
+
+## 4.1.0 (10/8/2015)
+### New Features
+* Added single delete operation as a more efficient way to delete keys that have not been overwritten.
+* Added experimental AddFile() to DB interface that allow users to add files created by SstFileWriter into an empty Database, see include/rocksdb/sst_file_writer.h and DB::AddFile() for more info.
+* Added support for opening SST files with .ldb suffix which enables opening LevelDB databases.
+* CompactionFilter now supports filtering of merge operands and merge results.
+
+### Public API Changes
+* Added SingleDelete() to the DB interface.
+* Added AddFile() to DB interface.
+* Added SstFileWriter class.
+* CompactionFilter has a new method FilterMergeOperand() that RocksDB applies to every merge operand during compaction to decide whether to filter the operand.
+* We removed CompactionFilterV2 interfaces from include/rocksdb/compaction_filter.h. The functionality was deprecated already in version 3.13.
+
 ## 4.0.0 (9/9/2015)
 ### New Features
 * Added support for transactions.  See include/rocksdb/utilities/transaction.h for more info.
-* DB::GetProperty() now accept "rocksdb.aggregated-table-properties" and "rocksdb.aggregated-table-properties-at-levelN", in which case it returns aggregated table properties of the target column family, or the aggregated table properties of the specified level N if the "at-level" version is used.
+* DB::GetProperty() now accepts "rocksdb.aggregated-table-properties" and "rocksdb.aggregated-table-properties-at-levelN", in which case it returns aggregated table properties of the target column family, or the aggregated table properties of the specified level N if the "at-level" version is used.
 * Add compression option kZSTDNotFinalCompression for people to experiment ZSTD although its format is not finalized.
 * We removed the need for LATEST_BACKUP file in BackupEngine. We still keep writing it when we create new backups (because of backward compatibility), but we don't read it anymore.

@@ -1,19 +1,24 @@
 ## Compilation

+**Important**: If you plan to run RocksDB in production, don't compile using default 
+`make` or `make all`. That will compile RocksDB in debug mode, which is much slower
+than release mode.
+
 RocksDB's library should be able to compile without any dependency installed,
 although we recommend installing some compression libraries (see below).
 We do depend on newer gcc/clang with C++11 support.

 There are few options when compiling RocksDB:

-* [recommended] `make static_lib` will compile librocksdb.a, RocksDB static library.
+* [recommended] `make static_lib` will compile librocksdb.a, RocksDB static library. Compiles static library in release mode.

-* `make shared_lib` will compile librocksdb.so, RocksDB shared library.
+* `make shared_lib` will compile librocksdb.so, RocksDB shared library. Compiles shared library in release mode.

-* `make check` will compile and run all the unit tests
+* `make check` will compile and run all the unit tests. `make check` will compile RocksDB in debug mode.

 * `make all` will compile our static library, and all our tools and unit tests. Our tools
-depend on gflags. You will need to have gflags installed to run `make all`.
+depend on gflags. You will need to have gflags installed to run `make all`. This will compile RocksDB in debug mode. Don't
+use binaries compiled by `make all` in production.

 * By default the binary we produce is optimized for the platform you're compiling on
 (-march=native). If you want to build a portable binary, add 'PORTABLE=1' before
@@ -82,4 +87,5 @@ your make commands, like this: `PORTABLE=1 make static_lib`
  * Run: `TARGET_OS=IOS make static_lib`. When building the project which uses rocksdb iOS library, make sure to define two important pre-processing macros: `ROCKSDB_LITE` and `IOS_CROSS_COMPILE`.

 * **Windows**:
+  * For building with MS Visual Studio 13 you will need Update 4 installed.
  * Read and follow the instructions at CMakeLists.txt
@@ -33,12 +33,22 @@ quoted_perl_command = $(subst ','\'',$(perl_command))
 # with debug level 0. To compile with level 0, run `make shared_lib`,
 # `make install-shared`, `make static_lib`, `make install-static` or
 # `make install`
-DEBUG_LEVEL=1
+
+# Set the default DEBUG_LEVEL to 1
+DEBUG_LEVEL?=1

 ifeq ($(MAKECMDGOALS),dbg)
 	DEBUG_LEVEL=2
 endif

+ifeq ($(MAKECMDGOALS),clean)
+	DEBUG_LEVEL=0
+endif
+
+ifeq ($(MAKECMDGOALS),release)
+	DEBUG_LEVEL=0
+endif
+
 ifeq ($(MAKECMDGOALS),shared_lib)
 	DEBUG_LEVEL=0
 endif
@@ -63,6 +73,14 @@ ifeq ($(MAKECMDGOALS),rocksdbjavastatic)
 	DEBUG_LEVEL=0
 endif

+ifeq ($(MAKECMDGOALS),rocksdbjavastaticrelease)
+	DEBUG_LEVEL=0
+endif
+
+ifeq ($(MAKECMDGOALS),rocksdbjavastaticpublish)
+	DEBUG_LEVEL=0
+endif
+
 # compile with -O2 if debug level is not 2
 ifneq ($(DEBUG_LEVEL), 2)
 OPT += -O2 -fno-omit-frame-pointer
@@ -76,6 +94,8 @@ endif
 ifeq ($(DEBUG_LEVEL),0)
 OPT += -DNDEBUG
 DISABLE_WARNING_AS_ERROR=1
+else
+$(warning Warning: Compiling in debug mode. Don't use the resulting binary in production)
 endif

 #-----------------------------------------------
@@ -167,10 +187,6 @@ default: all
 WARNING_FLAGS = -W -Wextra -Wall -Wsign-compare -Wshadow \
  -Wno-unused-parameter

-ifndef DISABLE_WARNING_AS_ERROR
-	WARNING_FLAGS += -Werror
-endif
-
 CFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CCFLAGS) $(OPT)
 CXXFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CXXFLAGS) $(OPT) -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers

@@ -206,13 +222,13 @@ util/build_version.cc: FORCE
 	else mv -f $@-t $@; fi

 LIBOBJECTS = $(LIB_SOURCES:.cc=.o)
+LIBOBJECTS += $(TOOL_SOURCES:.cc=.o)
 MOCKOBJECTS = $(MOCK_SOURCES:.cc=.o)

 GTEST = $(GTEST_DIR)/gtest/gtest-all.o
 TESTUTIL = ./util/testutil.o
 TESTHARNESS = ./util/testharness.o $(TESTUTIL) $(MOCKOBJECTS) $(GTEST)
 VALGRIND_ERROR = 2
-VALGRIND_DIR = build_tools/VALGRIND_LOGS
 VALGRIND_VER := $(join $(VALGRIND_VER),valgrind)

 VALGRIND_OPTS = --error-exitcode=$(VALGRIND_ERROR) --leak-check=full
@@ -228,6 +244,7 @@ TESTS = \
 	db_tailing_iter_test \
 	db_universal_compaction_test \
 	db_wal_test \
+	db_table_properties_test \
 	block_hash_index_test \
 	autovector_test \
 	column_family_test \
@@ -259,6 +276,7 @@ TESTS = \
 	mock_env_test \
 	memtable_list_test \
 	merge_helper_test \
+	memory_test \
 	merge_test \
 	merger_test \
 	redis_test \
@@ -295,6 +313,7 @@ TESTS = \
 	flush_job_test \
 	wal_manager_test \
 	listener_test \
+	compaction_iterator_test \
 	compaction_job_test \
 	thread_list_test \
 	sst_dump_test \
@@ -305,7 +324,8 @@ TESTS = \
 	heap_test \
 	compact_on_deletion_collector_test \
 	compaction_job_stats_test \
-	transaction_test
+	transaction_test \
+	ldb_cmd_test

 SUBSET :=  $(shell echo $(TESTS) |sed s/^.*$(ROCKSDBTESTS_START)/$(ROCKSDBTESTS_START)/)

@@ -313,6 +333,7 @@ TOOLS = \
 	sst_dump \
 	db_sanity_test \
 	db_stress \
+	write_stress \
 	ldb \
 	db_repl_stress \
 	rocksdb_dump \
@@ -320,10 +341,14 @@ TOOLS = \

 BENCHMARKS = db_bench table_reader_bench cache_bench memtablerep_bench

-# The library name is configurable since we are maintaining libraries of both
-# debug/release mode.
+# if user didn't config LIBNAME, set the default
 ifeq ($(LIBNAME),)
+# we should only run rocksdb in production with DEBUG_LEVEL 0
+ifeq ($(DEBUG_LEVEL),0)
        LIBNAME=librocksdb
+else
+        LIBNAME=librocksdb_debug
+endif
 endif
 LIBRARY = ${LIBNAME}.a

@@ -369,7 +394,7 @@ $(SHARED3): $(SHARED4)
 endif

 $(SHARED4):
-	$(CXX) $(PLATFORM_SHARED_LDFLAGS)$(SHARED3) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(LIB_SOURCES) \
+	$(CXX) $(PLATFORM_SHARED_LDFLAGS)$(SHARED3) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(LIB_SOURCES) $(TOOL_SOURCES) \
 		$(LDFLAGS) -o $@

 endif  # PLATFORM_SHARED_EXT
@@ -393,7 +418,7 @@ dbg: $(LIBRARY) $(BENCHMARKS) tools $(TESTS)
 # creates static library and programs
 release:
 	$(MAKE) clean
-	OPT="-DNDEBUG -O2" $(MAKE) static_lib tools db_bench
+	DEBUG_LEVEL=0 $(MAKE) static_lib tools db_bench

 coverage:
 	$(MAKE) clean
@@ -542,8 +567,10 @@ check: all
 	      echo "===== Running $$t"; ./$$t || exit 1; done;          \
 	fi
 	rm -rf $(TMPD)
+ifeq ($(filter -DROCKSDB_LITE,$(OPT)),)
 	python tools/ldb_test.py
 	sh tools/rocksdb_dump_test.sh
+endif

 check_some: $(SUBSET) ldb_tests
 	for t in $(SUBSET); do echo "===== Running $$t"; ./$$t || exit 1; done
@@ -555,12 +582,12 @@ ldb_tests: ldb
 crash_test: whitebox_crash_test blackbox_crash_test

 blackbox_crash_test: db_stress
-	python -u tools/db_crashtest.py -s
-	python -u tools/db_crashtest.py
+	python -u tools/db_crashtest.py --simple blackbox 
+	python -u tools/db_crashtest.py blackbox

 whitebox_crash_test: db_stress
-	python -u tools/db_crashtest2.py -s
-	python -u tools/db_crashtest2.py
+	python -u tools/db_crashtest.py --simple whitebox
+	python -u tools/db_crashtest.py whitebox

 asan_check:
 	$(MAKE) clean
@@ -573,20 +600,12 @@ asan_crash_test:
 	$(MAKE) clean

 valgrind_check: $(TESTS)
-	mkdir -p $(VALGRIND_DIR)
-	echo TESTS THAT HAVE VALGRIND ERRORS > $(VALGRIND_DIR)/valgrind_failed_tests; \
-	echo TIMES in seconds TAKEN BY TESTS ON VALGRIND > $(VALGRIND_DIR)/valgrind_tests_times; \
 	for t in $(filter-out skiplist_test,$(TESTS)); do \
-		stime=`date '+%s'`; \
 		$(VALGRIND_VER) $(VALGRIND_OPTS) ./$$t; \
 		ret_code=$$?; \
-		if [ $$ret_code -eq $(VALGRIND_ERROR) ] ; then \
-			echo $$t >> $(VALGRIND_DIR)/valgrind_failed_tests; \
-		elif [ $$ret_code -ne 0 ]; then \
+		if [ $$ret_code -ne 0 ]; then \
 			exit $$ret_code; \
 		fi; \
-		etime=`date '+%s'`; \
-		echo $$t $$((etime - stime)) >> $(VALGRIND_DIR)/valgrind_tests_times; \
 	done

 analyze: clean
@@ -599,14 +618,22 @@ CLEAN_FILES += unity.cc
 unity.cc: Makefile
 	rm -f $@ $@-t
 	for source_file in $(LIB_SOURCES); do \
-		echo "#include <$$source_file>" >> $@-t; \
+		echo "#include \"$$source_file\"" >> $@-t; \
 	done
-	echo 'int main(int argc, char** argv){ return 0; }' >> $@-t
 	chmod a=r $@-t
 	mv $@-t $@

-unity: unity.o
+unity.a: unity.o
+	$(AM_V_AR)rm -f $@
+	$(AM_V_at)$(AR) $(ARFLAGS) $@ unity.o
+
+# try compiling db_test with unity
+unity_test: db/db_test.o db/db_test_util.o $(TESTHARNESS) unity.a
 	$(AM_LINK)
+	./unity_test
+
+rocksdb.h rocksdb.cc: build_tools/amalgamate.py Makefile $(LIB_SOURCES) unity.cc
+	build_tools/amalgamate.py -I. -i./include unity.cc -x include/rocksdb/c.h -H rocksdb.h -o rocksdb.cc

 clean:
 	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(LIBRARY) $(SHARED)
@@ -647,6 +674,9 @@ block_hash_index_test: table/block_hash_index_test.o $(LIBOBJECTS) $(TESTHARNESS
 db_stress: tools/db_stress.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)

+write_stress: tools/write_stress.o $(LIBOBJECTS) $(TESTUTIL)
+	$(AM_LINK)
+
 db_sanity_test: tools/db_sanity_test.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)

@@ -701,34 +731,37 @@ crc32c_test: util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS)
 slice_transform_test: util/slice_transform_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_test: db/db_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_test: db/db_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_log_iter_test: db/db_log_iter_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_log_iter_test: db/db_log_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_compaction_filter_test: db/db_compaction_filter_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_compaction_filter_test: db/db_compaction_filter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_compaction_test: db/db_compaction_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_compaction_test: db/db_compaction_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_dynamic_level_test: db/db_dynamic_level_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_dynamic_level_test: db/db_dynamic_level_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_inplace_update_test: db/db_inplace_update_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_inplace_update_test: db/db_inplace_update_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_tailing_iter_test: db/db_tailing_iter_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_tailing_iter_test: db/db_tailing_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 db_iter_test: db/db_iter_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_universal_compaction_test: db/db_universal_compaction_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_universal_compaction_test: db/db_universal_compaction_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_wal_test: db/db_wal_test.o util/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_wal_test: db/db_wal_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+db_table_properties_test: db/db_table_properties_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 log_write_bench: util/log_write_bench.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -773,6 +806,9 @@ write_batch_with_index_test: utilities/write_batch_with_index/write_batch_with_i
 flush_job_test: db/flush_job_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+compaction_iterator_test: db/compaction_iterator_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 compaction_job_test: db/compaction_job_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -851,6 +887,9 @@ write_controller_test: db/write_controller_test.o $(LIBOBJECTS) $(TESTHARNESS)
 merge_helper_test: db/merge_helper_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+memory_test: utilities/memory/memory_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 merge_test: db/merge_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -893,7 +932,7 @@ options_test: util/options_test.o $(LIBOBJECTS) $(TESTHARNESS)
 event_logger_test: util/event_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-sst_dump_test: util/sst_dump_test.o $(LIBOBJECTS) $(TESTHARNESS)
+sst_dump_test: tools/sst_dump_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 memenv_test : util/memenv_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -905,7 +944,7 @@ optimistic_transaction_test: utilities/transactions/optimistic_transaction_test.
 mock_env_test : util/mock_env_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-manual_compaction_test: util/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
+manual_compaction_test: db/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 filelock_test: util/filelock_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -929,6 +968,9 @@ transaction_test: utilities/transactions/transaction_test.o $(LIBOBJECTS) $(TEST
 sst_dump: tools/sst_dump.o $(LIBOBJECTS)
 	$(AM_LINK)

+ldb_cmd_test: tools/ldb_cmd_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 ldb: tools/ldb.o $(LIBOBJECTS)
 	$(AM_LINK)

@@ -1021,21 +1063,23 @@ liblz4.a:
 	   cd lz4-r127/lib && make CFLAGS='-fPIC' all
 	   cp lz4-r127/lib/liblz4.a .

-# A version of each $(LIBOBJECTS) compiled with -fPIC
-java_libobjects = $(patsubst %,jl/%,$(LIBOBJECTS))
-CLEAN_FILES += jl
+# A version of each $(LIBOBJECTS) compiled with -fPIC and a fixed set of static compression libraries
+java_static_libobjects = $(patsubst %,jls/%,$(LIBOBJECTS))
+CLEAN_FILES += jls

-$(java_libobjects): jl/%.o: %.cc
-	$(AM_V_CC)mkdir -p $(@D)
-	@$(CXX) $(CXXFLAGS) -fPIC -c $< -o $@ $(COVERAGEFLAGS)
+JAVA_STATIC_FLAGS = -DZLIB -DBZIP2 -DSNAPPY -DLZ4
+JAVA_STATIC_INCLUDES = -I./zlib-1.2.8 -I./bzip2-1.0.6 -I./snappy-1.1.1 -I./lz4-r127/lib

-rocksdbjavastatic: $(java_libobjects) libz.a libbz2.a libsnappy.a liblz4.a
+$(java_static_libobjects): jls/%.o: %.cc libz.a libbz2.a libsnappy.a liblz4.a
+	$(AM_V_CC)mkdir -p $(@D) && $(CXX) $(CXXFLAGS) $(JAVA_STATIC_FLAGS) $(JAVA_STATIC_INCLUDES) -fPIC -c $< -o $@ $(COVERAGEFLAGS)
+
+rocksdbjavastatic: $(java_static_libobjects)
 	cd java;$(MAKE) javalib;
 	rm -f ./java/target/$(ROCKSDBJNILIB)
 	$(CXX) $(CXXFLAGS) -I./java/. $(JAVA_INCLUDE) -shared -fPIC \
 	  -o ./java/target/$(ROCKSDBJNILIB) $(JNI_NATIVE_SOURCES) \
-	  $(java_libobjects) $(COVERAGEFLAGS) \
-	  libz.a libbz2.a libsnappy.a liblz4.a $(LDFLAGS)
+	  $(java_static_libobjects) $(COVERAGEFLAGS) \
+	  libz.a libbz2.a libsnappy.a liblz4.a $(JAVA_STATIC_LDFLAGS)
 	cd java/target;strip -S -x $(ROCKSDBJNILIB)
 	cd java;jar -cf target/$(ROCKSDB_JAR) HISTORY*.md
 	cd java/target;jar -uf $(ROCKSDB_JAR) $(ROCKSDBJNILIB)
@@ -1046,7 +1090,7 @@ rocksdbjavastatic: $(java_libobjects) libz.a libbz2.a libsnappy.a liblz4.a
 rocksdbjavastaticrelease: rocksdbjavastatic
 	cd java/crossbuild && vagrant destroy -f && vagrant up linux32 && vagrant halt linux32 && vagrant up linux64 && vagrant halt linux64
 	cd java;jar -cf target/$(ROCKSDB_JAR_ALL) HISTORY*.md
-	cd java;jar -uf target/$(ROCKSDB_JAR_ALL) librocksdbjni-*.so librocksdbjni-*.jnilib
+	cd java/target;jar -uf $(ROCKSDB_JAR_ALL) librocksdbjni-*.so librocksdbjni-*.jnilib
 	cd java/target/classes;jar -uf ../$(ROCKSDB_JAR_ALL) org/rocksdb/*.class org/rocksdb/util/*.class

 rocksdbjavastaticpublish: rocksdbjavastaticrelease
@@ -1057,13 +1101,20 @@ rocksdbjavastaticpublish: rocksdbjavastaticrelease
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar -Dclassifier=osx
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar

+# A version of each $(LIBOBJECTS) compiled with -fPIC
+java_libobjects = $(patsubst %,jl/%,$(LIBOBJECTS))
+CLEAN_FILES += jl
+
+$(java_libobjects): jl/%.o: %.cc
+	$(AM_V_CC)mkdir -p $(@D) && $(CXX) $(CXXFLAGS) -fPIC -c $< -o $@ $(COVERAGEFLAGS)
+
 rocksdbjava: $(java_libobjects)
-	cd java;$(MAKE) javalib;
-	rm -f ./java/target/$(ROCKSDBJNILIB)
-	$(CXX) $(CXXFLAGS) -I./java/. $(JAVA_INCLUDE) -shared -fPIC -o ./java/target/$(ROCKSDBJNILIB) $(JNI_NATIVE_SOURCES) $(java_libobjects) $(JAVA_LDFLAGS) $(COVERAGEFLAGS)
-	cd java;jar -cf target/$(ROCKSDB_JAR) HISTORY*.md
-	cd java/target;jar -uf $(ROCKSDB_JAR) $(ROCKSDBJNILIB)
-	cd java/target/classes;jar -uf ../$(ROCKSDB_JAR) org/rocksdb/*.class org/rocksdb/util/*.class
+	$(AM_V_GEN)cd java;$(MAKE) javalib;
+	$(AM_V_at)rm -f ./java/target/$(ROCKSDBJNILIB)
+	$(AM_V_at)$(CXX) $(CXXFLAGS) -I./java/. $(JAVA_INCLUDE) -shared -fPIC -o ./java/target/$(ROCKSDBJNILIB) $(JNI_NATIVE_SOURCES) $(java_libobjects) $(JAVA_LDFLAGS) $(COVERAGEFLAGS)
+	$(AM_V_at)cd java;jar -cf target/$(ROCKSDB_JAR) HISTORY*.md
+	$(AM_V_at)cd java/target;jar -uf $(ROCKSDB_JAR) $(ROCKSDBJNILIB)
+	$(AM_V_at)cd java/target/classes;jar -uf ../$(ROCKSDB_JAR) org/rocksdb/*.class org/rocksdb/util/*.class

 jclean:
 	cd java;$(MAKE) clean;
@@ -5,6 +5,7 @@ before_build:
 - cmake -G "Visual Studio 12 Win64" ..
 - cd ..
 build:
-  project: build\ALL_BUILD.vcxproj
+  project: build\rocksdb.sln
+  parallel: true
  verbosity: minimal
 test: off
@@ -0,0 +1,22 @@
+version: 1.0.{build}
+before_build:
+- md %APPVEYOR_BUILD_FOLDER%\build
+- cd %APPVEYOR_BUILD_FOLDER%\build
+- cmake -G "Visual Studio 12 Win64" -DOPTDBG=1 ..
+- cd ..
+build:
+  project: build\rocksdb.sln
+  parallel: true
+  verbosity: minimal
+test:
+test_script:
+- ps: build_tools\run_ci_db_test.ps1
+notifications:
+  - provider: Email
+    to:
+      - svmtrocksdb@microsoft.com
+    subject: "Build {{status}}"
+    message: "{{message}}, {{commitId}}, ..."
+    on_build_success: false
+    on_build_failure: true
+    on_build_status_changed: true
@@ -88,6 +88,9 @@ class FbcodeCppLinter extends ArcanistLinter {
  }

  private function getCppLintOutput($path) {
+     if (!array_key_exists($path, $this->rawLintOutput)) {
+       return array();
+     }
    list($output) = $this->rawLintOutput[$path];

    $msgs = array();
@@ -0,0 +1,110 @@
+#!/usr/bin/python
+
+# amalgamate.py creates an amalgamation from a unity build.
+# It can be run with either Python 2 or 3.
+# An amalgamation consists of a header that includes the contents of all public
+# headers and a source file that includes the contents of all source files and
+# private headers.
+#
+# This script works by starting with the unity build file and recursively expanding
+# #include directives. If the #include is found in a public include directory,
+# that header is expanded into the amalgamation header.
+#
+# A particular header is only expanded once, so this script will
+# break if there are multiple inclusions of the same header that are expected to
+# expand differently. Similarly, this type of code causes issues:
+#
+# #ifdef FOO
+#   #include "bar.h"
+#   // code here
+# #else
+#   #include "bar.h"            // oops, doesn't get expanded
+#   // different code here
+# #endif
+#
+# The solution is to move the include out of the #ifdef.
+
+from __future__ import print_function
+
+import argparse
+from os import path
+import re
+import sys
+
+include_re = re.compile('^[ \t]*#include[ \t]+"(.*)"[ \t]*$')
+included = set()
+excluded = set()
+
+def find_header(name, abs_path, include_paths):
+    samedir = path.join(path.dirname(abs_path), name)
+    if path.exists(samedir):
+        return samedir
+    for include_path in include_paths:
+        include_path = path.join(include_path, name)
+        if path.exists(include_path):
+            return include_path
+    return None
+
+def expand_include(include_path, f, abs_path, source_out, header_out, include_paths, public_include_paths):
+    if include_path in included:
+        return False
+
+    included.add(include_path)
+    with open(include_path) as f:
+        print('#line 1 "{}"'.format(include_path), file=source_out)
+        process_file(f, include_path, source_out, header_out, include_paths, public_include_paths)
+    return True
+
+def process_file(f, abs_path, source_out, header_out, include_paths, public_include_paths):
+    for (line, text) in enumerate(f):
+        m = include_re.match(text)
+        if m:
+            filename = m.groups()[0]
+            # first check private headers
+            include_path = find_header(filename, abs_path, include_paths)
+            if include_path:
+                if include_path in excluded:
+                    source_out.write(text)
+                    expanded = False
+                else:
+                    expanded = expand_include(include_path, f, abs_path, source_out, header_out, include_paths, public_include_paths)
+            else:
+                # now try public headers
+                include_path = find_header(filename, abs_path, public_include_paths)
+                if include_path:
+                    # found public header
+                    expanded = False
+                    if include_path in excluded:
+                        source_out.write(text)
+                    else:
+                        expand_include(include_path, f, abs_path, header_out, None, public_include_paths, [])
+                else:
+                    sys.exit("unable to find {}, included in {} on line {}".format(filename, abs_path, line))
+
+            if expanded:
+                print('#line {} "{}"'.format(line+1, abs_path), file=source_out)
+        elif text != "#pragma once\n":
+            source_out.write(text)
+
+def main():
+    parser = argparse.ArgumentParser(description="Transform a unity build into an amalgamation")
+    parser.add_argument("source", help="source file")
+    parser.add_argument("-I", action="append", dest="include_paths", help="include paths for private headers")
+    parser.add_argument("-i", action="append", dest="public_include_paths", help="include paths for public headers")
+    parser.add_argument("-x", action="append", dest="excluded", help="excluded header files")
+    parser.add_argument("-o", dest="source_out", help="output C++ file", required=True)
+    parser.add_argument("-H", dest="header_out", help="output C++ header file", required=True)
+    args = parser.parse_args()
+
+    include_paths = list(map(path.abspath, args.include_paths or []))
+    public_include_paths = list(map(path.abspath, args.public_include_paths or []))
+    excluded.update(map(path.abspath, args.excluded or []))
+    filename = args.source
+    abs_path = path.abspath(filename)
+    with open(filename) as f, open(args.source_out, 'w') as source_out, open(args.header_out, 'w') as header_out:
+        print('#line 1 "{}"'.format(filename), file=source_out)
+        print('#include "{}"'.format(header_out.name), file=source_out)
+        process_file(f, abs_path, source_out, header_out, include_paths, public_include_paths)
+
+if __name__ == "__main__":
+    main()
@@ -8,6 +8,7 @@
 #   CXX                         C++ Compiler path
 #   PLATFORM_LDFLAGS            Linker flags
 #   JAVA_LDFLAGS                Linker flags for RocksDBJava
+#   JAVA_STATIC_LDFLAGS         Linker flags for RocksDBJava static build
 #   PLATFORM_SHARED_EXT         Extension for shared libraries
 #   PLATFORM_SHARED_LDFLAGS     Flags for building shared library
 #   PLATFORM_SHARED_CFLAGS      Flags for compiling objects for shared library
@@ -44,19 +45,14 @@ fi
 # we depend on C++11
 PLATFORM_CXXFLAGS="-std=c++11"
 # we currently depend on POSIX platform
-COMMON_FLAGS="-DROCKSDB_PLATFORM_POSIX"
+COMMON_FLAGS="-DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX"

 # Default to fbcode gcc on internal fb machines
 if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
    FBCODE_BUILD="true"
    # If we're compiling with TSAN we need pic build
    PIC_BUILD=$COMPILE_WITH_TSAN
-    if [ -z "$ROCKSDB_FBCODE_BUILD_WITH_481" ]; then
-      source "$PWD/build_tools/fbcode_config.sh"
-    else
-      # we need this to build with MySQL. Don't use for other purposes.
-      source "$PWD/build_tools/fbcode_config4.8.1.sh"
-    fi
+    source "$PWD/build_tools/fbcode_config.sh"
 fi

 # Delete existing output, if it exists
@@ -181,6 +177,7 @@ esac

 PLATFORM_CXXFLAGS="$PLATFORM_CXXFLAGS ${CXXFLAGS}"
 JAVA_LDFLAGS="$PLATFORM_LDFLAGS"
+JAVA_STATIC_LDFLAGS="$PLATFORM_LDFLAGS"

 if [ "$CROSS_COMPILE" = "true" -o "$FBCODE_BUILD" = "true" ]; then
    # Cross-compiling; do not try any compilation tests.
@@ -213,7 +210,7 @@ EOF
    fi

    # Test whether gflags library is installed
-    # http://code.google.com/p/gflags/
+    # http://gflags.github.io/gflags/
    # check if the namespace is gflags
    $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null << EOF
      #include <gflags/gflags.h>
@@ -374,6 +371,7 @@ echo "CXX=$CXX" >> "$OUTPUT"
 echo "PLATFORM=$PLATFORM" >> "$OUTPUT"
 echo "PLATFORM_LDFLAGS=$PLATFORM_LDFLAGS" >> "$OUTPUT"
 echo "JAVA_LDFLAGS=$JAVA_LDFLAGS" >> "$OUTPUT"
+echo "JAVA_STATIC_LDFLAGS=$JAVA_STATIC_LDFLAGS" >> "$OUTPUT"
 echo "VALGRIND_VER=$VALGRIND_VER" >> "$OUTPUT"
 echo "PLATFORM_CCFLAGS=$PLATFORM_CCFLAGS" >> "$OUTPUT"
 echo "PLATFORM_CXXFLAGS=$PLATFORM_CXXFLAGS" >> "$OUTPUT"
@@ -0,0 +1,19 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+GCC_BASE=/mnt/gvfs/third-party2/gcc/7331085db891a2ef4a88a48a751d834e8d68f4cb/7.x/centos7-native/b2ef2b6
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/963d9aeda70cc4779885b1277484fe7544a04e3e/9.0.0/platform007/9e92d53/
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/6ace84e956873d53638c738b6f65f3f469cca74c/7.x/platform007/5620abc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/192b0f42d63dcf6210d6ceae387b49af049e6e0c/2.26/platform007/f259413
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/7f9bdaada18f59bc27ec2b0871eb8a6144343aef/1.1.3/platform007/ca4da3d
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/2d9f0b9a4274cc21f61272a9e89bdb859bce8f1f/1.2.8/platform007/ca4da3d
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/dc49a21c5fceec6456a7a28a94dcd16690af1337/1.0.6/platform007/ca4da3d
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/0f607f8fc442ea7d6b876931b1898bb573d5e5da/1.9.1/platform007/ca4da3d
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/ca22bc441a4eb709e9e0b1f9fec9750fed7b31c5/1.4.x/platform007/15a3614
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/0b9929d2588991c65a57168bf88aff2db87c5d48/2.2.0/platform007/ca4da3d
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/c26f08f47ac35fc31da2633b7da92d6b863246eb/master/platform007/c26c002
+NUMA_BASE=/mnt/gvfs/third-party2/numa/3f3fb57a5ccc5fd21c66416c0b83e0aa76a05376/2.0.11/platform007/ca4da3d
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/40c73d874898b386a71847f1b99115d93822d11f/1.4/platform007/6f3e0a9
+TBB_BASE=/mnt/gvfs/third-party2/tbb/4ce8e8dba77cdbd81b75d6f0c32fd7a1b76a11ec/2018_U5/platform007/ca4da3d
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/fb251ecd2f5ae16f8671f7014c246e52a748fe0b/fb/platform007/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/ab9f09bba370e7066cafd4eb59752db93f2e8312/2.29.1/platform007/15a3614
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d42d152a15636529b0861ec493927200ebebca8e/3.15.0/platform007/ca4da3d
+LUA_BASE=/mnt/gvfs/third-party2/lua/f0cd714433206d5139df61659eb7b28b1dea6683/5.3.4/platform007/5007832
@@ -6,128 +6,147 @@
 # Environment variables that change the behavior of this script:
 # PIC_BUILD -- if true, it will only take pic versions of libraries from fbcode. libraries that don't have pic variant will not be included

+
+BASEDIR=`dirname $BASH_SOURCE`
+source "$BASEDIR/dependencies.sh"
+
 CFLAGS=""

-# location of libgcc
-LIBGCC_BASE="/mnt/gvfs/third-party2/libgcc/0473c80518a10d6efcbe24c5eeca3fb4ec9b519c/4.9.x/gcc-4.9-glibc-2.20/e1a7e4e"
-LIBGCC_INCLUDE="$LIBGCC_BASE/include"
-LIBGCC_LIBS=" -L $LIBGCC_BASE/libs"
+# libgcc
+LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/7.3.0"
+LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"

-# location of glibc
-GLIBC_REV=7397bed99280af5d9543439cdb7d018af7542720
-GLIBC_INCLUDE="/mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.20/gcc-4.9-glibc-2.20/99df8fc/include"
-GLIBC_LIBS=" -L /mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.20/gcc-4.9-glibc-2.20/99df8fc/lib"
-
-SNAPPY_INCLUDE=" -I /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/include/"
+# glibc
+GLIBC_INCLUDE="$GLIBC_BASE/include"
+GLIBC_LIBS=" -L $GLIBC_BASE/lib"

+# snappy
+SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
 if test -z $PIC_BUILD; then
-  SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/lib/libsnappy.a"
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"
 else
-  SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/lib/libsnappy_pic.a"
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy_pic.a"
 fi
-
 CFLAGS+=" -DSNAPPY"

 if test -z $PIC_BUILD; then
  # location of zlib headers and libraries
-  ZLIB_INCLUDE=" -I /mnt/gvfs/third-party2/zlib/feb983d9667f4cf5e9da07ce75abc824764b67a1/1.2.8/gcc-4.9-glibc-2.20/4230243/include/"
-  ZLIB_LIBS=" /mnt/gvfs/third-party2/zlib/feb983d9667f4cf5e9da07ce75abc824764b67a1/1.2.8/gcc-4.9-glibc-2.20/4230243/lib/libz.a"
+  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
+  ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"
  CFLAGS+=" -DZLIB"

  # location of bzip headers and libraries
-  BZIP_INCLUDE=" -I /mnt/gvfs/third-party2/bzip2/af004cceebb2dfd173ca29933ea5915e727aad2f/1.0.6/gcc-4.9-glibc-2.20/4230243/include/"
-  BZIP_LIBS=" /mnt/gvfs/third-party2/bzip2/af004cceebb2dfd173ca29933ea5915e727aad2f/1.0.6/gcc-4.9-glibc-2.20/4230243/lib/libbz2.a"
+  BZIP_INCLUDE=" -I $BZIP2_BASE/include/"
+  BZIP_LIBS=" $BZIP2_BASE/lib/libbz2.a"
  CFLAGS+=" -DBZIP2"

-  LZ4_INCLUDE=" -I /mnt/gvfs/third-party2/lz4/79d2943e2dd7208a3e0b06cf95e9f85f05fe9e1b/r124/gcc-4.9-glibc-2.20/4230243/include/"
-  LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/79d2943e2dd7208a3e0b06cf95e9f85f05fe9e1b/r124/gcc-4.9-glibc-2.20/4230243/lib/liblz4.a"
+  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
+  LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
  CFLAGS+=" -DLZ4"
-
-  ZSTD_REV=8df2d01673ae6afcc8c8d16fec862b2d67ecc1e9
-  ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.1.1/gcc-4.8.1-glibc-2.17/c3f970a/include"
-  ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.1.1/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"
-  CFLAGS+=" -DZSTD"
 fi

+ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
+if test -z $PIC_BUILD; then
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd.a"
+else
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd_pic.a"
+fi
+CFLAGS+=" -DZSTD"
+
 # location of gflags headers and libraries
-GFLAGS_INCLUDE=" -I /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/include/"
+GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
 if test -z $PIC_BUILD; then
-  GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/lib/libgflags.a"
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"
 else
-  GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/lib/libgflags_pic.a"
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags_pic.a"
 fi
-CFLAGS+=" -DGFLAGS=google"
+CFLAGS+=" -DGFLAGS=gflags"

 # location of jemalloc
-JEMALLOC_INCLUDE=" -I /mnt/gvfs/third-party2/jemalloc/bcd68e5e419efa4e61b9486d6854564d6d75a0b5/3.6.0/gcc-4.9-glibc-2.20/2aafc78/include/"
-JEMALLOC_LIB=" /mnt/gvfs/third-party2/jemalloc/bcd68e5e419efa4e61b9486d6854564d6d75a0b5/3.6.0/gcc-4.9-glibc-2.20/2aafc78/lib/libjemalloc.a"
+JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
+JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc.a"

 if test -z $PIC_BUILD; then
  # location of numa
-  NUMA_INCLUDE=" -I /mnt/gvfs/third-party2/numa/bbefc39ecbf31d0ca184168eb613ef8d397790ee/2.0.8/gcc-4.9-glibc-2.20/4230243/include/"
-  NUMA_LIB=" /mnt/gvfs/third-party2/numa/bbefc39ecbf31d0ca184168eb613ef8d397790ee/2.0.8/gcc-4.9-glibc-2.20/4230243/lib/libnuma.a"
+  NUMA_INCLUDE=" -I $NUMA_BASE/include/"
+  NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"
  CFLAGS+=" -DNUMA"

  # location of libunwind
-  LIBUNWIND="/mnt/gvfs/third-party2/libunwind/1de3b75e0afedfe5585b231bbb340ec7a1542335/1.1/gcc-4.9-glibc-2.20/34235e8/lib/libunwind.a"
+  LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
 fi

+# location of TBB
+TBB_INCLUDE=" -isystem $TBB_BASE/include/"
+if test -z $PIC_BUILD; then
+  TBB_LIBS="$TBB_BASE/lib/libtbb.a"
+else
+  TBB_LIBS="$TBB_BASE/lib/libtbb_pic.a"
+fi
+CFLAGS+=" -DTBB"
+
 # use Intel SSE support for checksum calculations
 export USE_SSE=1

-BINUTILS="/mnt/gvfs/third-party2/binutils/0b6ad0c88ddd903333a48ae8bff134efac468e4a/2.25/centos6-native/da39a3e/bin"
+BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"

-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE"

-GCC_BASE="/mnt/gvfs/third-party2/gcc/1c67a0b88f64d4d9ced0382d141c76aaa7d62fba/4.9.x/centos6-native/1317bc4"
 STDLIBS="-L $GCC_BASE/lib64"

-CLANG_BASE="/mnt/gvfs/third-party2/clang/d81444dd214df3d2466734de45bb264a0486acc3/dev"
-CLANG_BIN="$CLANG_BASE/centos6-native/af4b1a0/bin"
+CLANG_BIN="$CLANG_BASE/bin"
+CLANG_LIB="$CLANG_BASE/lib"
+CLANG_SRC="$CLANG_BASE/../../src"
+
 CLANG_ANALYZER="$CLANG_BIN/clang++"
-CLANG_SCAN_BUILD="$CLANG_BASE/src/clang/tools/scan-build/scan-build"
+CLANG_SCAN_BUILD="$CLANG_SRC/llvm/tools/clang/tools/scan-build/bin/scan-build"

 if [ -z "$USE_CLANG" ]; then
  # gcc
  CC="$GCC_BASE/bin/gcc"
  CXX="$GCC_BASE/bin/g++"
-  
+
  CFLAGS+=" -B$BINUTILS/gold"
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  CFLAGS+=" -isystem $GLIBC_INCLUDE"
+  JEMALLOC=1
 else
-  # clang 
-  CLANG_INCLUDE="$CLANG_BASE/gcc-4.9-glibc-2.20/74c386f/lib/clang/dev/include/"
+  # clang
+  CLANG_INCLUDE="$CLANG_LIB/clang/stable/include"
  CC="$CLANG_BIN/clang"
  CXX="$CLANG_BIN/clang++"

-  KERNEL_HEADERS_INCLUDE="/mnt/gvfs/third-party2/kernel-headers/ffd14f660a43c4b92717986b1bba66722ef089d0/3.2.18_70_fbk11_00129_gc8882d0/gcc-4.9-glibc-2.20/da39a3e/include"
+  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"

  CFLAGS+=" -B$BINUTILS/gold -nostdinc -nostdlib"
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.9.x "
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.9.x/x86_64-facebook-linux "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x/x86_64-facebook-linux "
  CFLAGS+=" -isystem $GLIBC_INCLUDE"
  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
  CFLAGS+=" -isystem $CLANG_INCLUDE"
  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE/linux "
  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE "
+  CFLAGS+=" -Wno-expansion-to-defined "
  CXXFLAGS="-nostdinc++"
 fi

 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
 CXXFLAGS+=" $CFLAGS"

-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB"
-EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/gcc-4.9-glibc-2.20/lib/ld.so"
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
+EXEC_LDFLAGS+=" -B$BINUTILS/gold"
+EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
 EXEC_LDFLAGS+=" $LIBUNWIND"
-EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/gcc-4.9-glibc-2.20/lib"
+EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform007/lib"
+# required by libtbb
+EXEC_LDFLAGS+=" -ldl"

 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"

-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"

-VALGRIND_VER="/mnt/gvfs/third-party2/valgrind/6c45ef049cbf11c2df593addb712cd891049e737/3.10.0/gcc-4.9-glibc-2.20/4230243/bin/"
+VALGRIND_VER="$VALGRIND_BASE/bin/"

-export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD
+export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD 
@@ -30,9 +30,9 @@ LZ4_REV=065ec7e38fe83329031f6668c43bef83eff5808b
 LZ4_INCLUDE=" -I /mnt/gvfs/third-party2/lz4/$LZ4_REV/r108/gcc-4.8.1-glibc-2.17/c3f970a/include"
 LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/$LZ4_REV/r108/gcc-4.8.1-glibc-2.17/c3f970a/lib/liblz4.a"

-ZSTD_REV=8df2d01673ae6afcc8c8d16fec862b2d67ecc1e9
-ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.1.1/gcc-4.8.1-glibc-2.17/c3f970a/include"
-ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.1.1/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"
+ZSTD_REV=810b81b4705def5243e998b54701f3c504e4009e
+ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/include"
+ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"

 # location of gflags headers and libraries
 GFLAGS_INCLUDE=" -I /mnt/gvfs/third-party2/gflags/1ad047a6e6f6673991918ecadc670868205a243a/1.6/gcc-4.8.1-glibc-2.17/c3f970a/include/"
@@ -91,7 +91,7 @@ else
 fi

 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
 CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4 -DZSTD -DNUMA"
 CXXFLAGS+=" $CFLAGS"

@@ -79,6 +79,12 @@ echo -e "Detect lines that doesn't follow the format rules:\r"
 echo "$diffs" | 
  sed -e "s/\(^-.*$\)/`echo -e \"$COLOR_RED\1$COLOR_END\"`/" |
  sed -e "s/\(^+.*$\)/`echo -e \"$COLOR_GREEN\1$COLOR_END\"`/"
+
+if [[ "$OPT" == *"-DTRAVIS"* ]]
+then
+  exit 1
+fi
+
 echo -e "Would you like to fix the format automatically (y/n): \c"

 # Make sure under any mode, we can read user input.
@@ -2,22 +2,62 @@
 # This script is executed by Sandcastle
 # to determine next steps to run

+# Usage:
+# EMAIL=<email> ONCALL=<email> TRIGGER=<trigger> SUBSCRIBER=<email> rocks_ci.py <test-name>
+#
+# Input         Value
+# -------------------------------------------------------------------------
+# EMAIL         Email address to report on trigger conditions
+# ONCAL         Email address to raise a task on failure
+# TRIGGER       Trigger conditions for email. Valid values are fail, warn, all
+# SUBSCRIBER    Email addresss to add as subscriber for task
+#
+
+#
+# Report configuration
+#
+REPORT_EMAIL=
+if [ ! -z $EMAIL ]; then
+  if [ -z $TRIGGER ]; then
+    TRIGGER="fail"
+  fi
+
+  REPORT_EMAIL="
+  {
+      'type':'email',
+      'triggers': [ '$TRIGGER' ],
+      'emails':['$EMAIL']
+  },"
+fi
+
+CREATE_TASK=
+if [ ! -z $ONCALL ]; then
+  CREATE_TASK="
+  {
+      'type':'task',
+      'triggers':[ 'fail' ],
+      'priority':0,
+      'subscribers':[ '$SUBSCRIBER' ],
+      'tags':[ 'rocksdb', 'ci' ],
+  },"
+fi
+
+REPORT=
+if [[ ! -z $REPORT_EMAIL || ! -z $CREATE_TASK ]]; then
+  REPORT="'report': [
+    $REPORT_EMAIL
+    $CREATE_TASK
+  ]"
+fi
+
 #
 # Helper variables
 #
-REPORT_EMAIL="'report': [
-    {
-        'type':'email',
-        'triggers':['fail'],
-        'emails':['leveldb@fb.com']
-    }
-]"
-
 CLEANUP_ENV="
 {
-  'name':'Cleanup environment',
-  'shell':'rm -rf /dev/shm/rocksdb && mkdir /dev/shm/rocksdb && make clean',
-  'user':'root'
+    'name':'Cleanup environment',
+    'shell':'rm -rf /dev/shm/rocksdb && mkdir /dev/shm/rocksdb && make clean',
+    'user':'root'
 }"

 DEBUG="OPT=-g"
@@ -27,6 +67,32 @@ ASAN="COMPILE_WITH_ASAN=1"
 CLANG="USE_CLANG=1"
 LITE="OPT=-DROCKSDB_LITE"
 TSAN="COMPILE_WITH_TSAN=1"
+DISABLE_JEMALLOC="DISABLE_JEMALLOC=1"
+PARSER="'parser':'egrep \'Failure|^#|Abort\''"
+
+ARTIFACTS=" 'artifacts': [
+    {
+        'name':'database',
+        'paths':[ '/dev/shm/rocksdb' ],
+    }
+]"
+
+#
+# A mechanism to disable tests temporarily
+#
+DISABLE_COMMANDS="[
+    {
+        'name':'Disable test',
+        'oncall':'$ONCALL',
+        'steps': [
+            {
+              'name':'Job disabled. Please contact test owner',
+              'shell':'exit 1',
+              'user':'root'
+            },
+        ],
+    }
+]"

 #
 # RocksDB unit test in parallel
@@ -36,21 +102,24 @@ TSAN="COMPILE_WITH_TSAN=1"
 PARALLEL_UNIT_TEST_COMMANDS="[
    {
        'name':'Rocksdb Parallel Unit Test',
+        'oncall':'$ONCALL',
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Build and test RocksDB debug version',
-                'shell':'$SHM $DEBUG make check 2>&1 > /dev/null || cat t/log-*',
-                'user':'root'
+                'shell':'$DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || cat t/log-*',
+                'user':'root',
+                $PARSER
            },
            $CLEANUP_ENV,
            {
                'name':'Build and test RocksDB debug version under gcc-4.8.1',
-                'shell':'$SHM $GCC_481 $DEBUG make check 2>&1 > /dev/null || cat t/log-*',
-                'user':'root'
+                'shell':'$GCC_481 $DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || cat t/log-*',
+                'user':'root',
+                $PARSER
            },
        ],
-        $REPORT_EMAIL
+        $REPORT
    }
 ]"

@@ -60,15 +129,17 @@ PARALLEL_UNIT_TEST_COMMANDS="[
 UNIT_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test',
+        'oncall':'$ONCALL',
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Build and test RocksDB debug version',
                'shell':'$SHM $DEBUG make J=1 check',
-                'user':'root'
+                'user':'root',
+                $PARSER
            },
        ],
-        $REPORT_EMAIL
+        $REPORT
    }
 ]"

@@ -78,15 +149,17 @@ UNIT_TEST_COMMANDS="[
 UNIT_TEST_COMMANDS_481="[
    {
        'name':'Rocksdb Unit Test on GCC 4.8.1',
+        'oncall':'$ONCALL',
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Build and test RocksDB debug version',
                'shell':'$SHM $GCC_481 $DEBUG make J=1 check',
-                'user':'root'
+                'user':'root',
+                $PARSER
            },
        ],
-        $REPORT_EMAIL
+        $REPORT
    }
 ]"

@@ -96,15 +169,77 @@ UNIT_TEST_COMMANDS_481="[
 CLANG_UNIT_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test',
+        'oncall':'$ONCALL',
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Build and test RocksDB debug',
-                'shell':'$CLANG $SHM $DEBUG make check',
-                'user':'root'
+                'shell':'$CLANG $SHM $DEBUG make J=1 check',
+                'user':'root',
+                $PARSER
            },
        ],
-        $REPORT_EMAIL
+        $REPORT
+    }
+]"
+
+#
+# RocksDB analyze
+#
+CLANG_ANALYZE_COMMANDS="[
+    {
+        'name':'Rocksdb analyze',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'RocksDB build and analyze',
+                'shell':'$CLANG $SHM $DEBUG make J=1 analyze',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB code coverage
+#
+CODE_COV_COMMANDS="[
+    {
+        'name':'Rocksdb Unit Test Code Coverage',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build, test and collect code coverage info',
+                'shell':'$SHM $DEBUG make J=1 coverage',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB unity
+#
+UNITY_COMMANDS="[
+    {
+        'name':'Rocksdb Unity',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build, test unity test',
+                'shell':'$SHM $DEBUG V=1 make J=1 unity_test',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
    }
 ]"

@@ -114,15 +249,37 @@ CLANG_UNIT_TEST_COMMANDS="[
 LITE_BUILD_COMMANDS="[
    {
        'name':'Rocksdb Lite build',
+        'oncall':'$ONCALL',
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Build RocksDB debug version',
                'shell':'$LITE $DEBUG make J=1 static_lib',
-                'user':'root'
+                'user':'root',
+                $PARSER
            },
        ],
-        $REPORT_EMAIL
+        $REPORT
+    }
+]"
+
+#
+# RocksDB lite tests
+#
+LITE_UNIT_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb Lite Unit Test',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build RocksDB debug version',
+                'shell':'$SHM $LITE $DEBUG make J=1 check',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
    }
 ]"

@@ -132,20 +289,47 @@ LITE_BUILD_COMMANDS="[
 STRESS_CRASH_TEST_COMMANDS="[
    {
        'name':'Rocksdb Stress/Crash Test',
+        'oncall':'$ONCALL',
+        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Build and run RocksDB debug stress tests',
                'shell':'$SHM $DEBUG make J=1 db_stress',
-                'user':'root'
+                'user':'root',
+                $PARSER
            },
            {
                'name':'Build and run RocksDB debug crash tests',
+                'timeout': 86400,
                'shell':'$SHM $DEBUG make J=1 crash_test',
-                'user':'root'
+                'user':'root',
+                $PARSER
            }
        ],
-        $REPORT_EMAIL
+        $ARTIFACTS,
+        $REPORT
+    }
+]"
+
+# RocksDB write stress test.
+# We run on disk device on purpose (i.e. no $SHM)
+# because we want to add some randomness to fsync commands
+WRITE_STRESS_COMMANDS="[
+    {
+        'name':'Rocksdb Write Stress Test',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build and run RocksDB write stress tests',
+                'shell':'make write_stress && python tools/write_stress_runner.py --runtime_sec=3600 --db=/tmp/rocksdb_write_stress',
+                'user':'root',
+                $PARSER
+            }
+        ],
+        'artifacts': [{'name': 'database', 'paths': ['/tmp/rocksdb_write_stress']}],
+        $REPORT
    }
 ]"

@@ -156,15 +340,17 @@ STRESS_CRASH_TEST_COMMANDS="[
 ASAN_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test under ASAN',
+        'oncall':'$ONCALL',
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Test RocksDB debug under ASAN',
                'shell':'set -o pipefail && $SHM $ASAN $DEBUG make J=1 asan_check |& /usr/facebook/ops/scripts/asan_symbolize.py -d',
-                'user':'root'
+                'user':'root',
+                $PARSER
            }
        ],
-        $REPORT_EMAIL
+        $REPORT
    }
 ]"

@@ -174,15 +360,19 @@ ASAN_TEST_COMMANDS="[
 ASAN_CRASH_TEST_COMMANDS="[
    {
        'name':'Rocksdb crash test under ASAN',
+        'oncall':'$ONCALL',
+        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Build and run RocksDB debug asan_crash_test',
+                'timeout': 86400,
                'shell':'$SHM $DEBUG make J=1 asan_crash_test',
-                'user':'root'
+                'user':'root',
+                $PARSER
            },
        ],
-        $REPORT_EMAIL
+        $REPORT
    }
 ]"

@@ -192,33 +382,213 @@ ASAN_CRASH_TEST_COMMANDS="[
 VALGRIND_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test under valgrind',
+        'oncall':'$ONCALL',
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Run RocksDB debug unit tests',
-                'shell':'$SHM $DEBUG ./build_tools/valgrind_test.sh',
-                'user':'root'
+                'shell':'$DISABLE_JEMALLOC $SHM $DEBUG make valgrind_check',
+                'user':'root',
+                $PARSER
            },
        ],
-        $REPORT_EMAIL
+        $REPORT
    }
 ]"

 #
 # RocksDB test under TSAN
 #
-TSAN_TEST_COMMANDS="[
+TSAN_UNIT_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test under TSAN',
+        'oncall':'$ONCALL',
        'steps': [
            $CLEANUP_ENV,
            {
                'name':'Run RocksDB debug unit test',
                'shell':'set -o pipefail && $SHM $DEBUG $TSAN make J=1 check',
-                'user':'root'
+                'user':'root',
+                $PARSER
            },
        ],
-        $REPORT_EMAIL
+        $REPORT
+    }
+]"
+
+#
+# RocksDB crash test under TSAN
+#
+TSAN_CRASH_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb Crash Test under TSAN',
+        'oncall':'$ONCALL',
+        'timeout': 86400,
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Compile and run',
+                'timeout': 86400,
+                'shell':'set -o pipefail && $SHM $DEBUG $TSAN make J=1 crash_test',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+TSAN_CRASH_TEST_COMMANDS=$DISABLE_COMMANDS
+
+#
+# RocksDB format compatible
+#
+
+run_format_compatible()
+{
+  export TEST_TMPDIR=/dev/shm/rocksdb
+  rm -rf /dev/shm/rocksdb
+  mkdir /dev/shm/rocksdb
+
+  echo '
+  if [ -e "build_tools/build_detect_platform" ]
+  then
+    sed "s/tcmalloc/nothingnothingnothing/g" build_tools/build_detect_platform > $TEST_TMPDIR/temp_build_file
+    rm -rf build_tools/build_detect_platform
+    cp $TEST_TMPDIR/temp_build_file build_tools/build_detect_platform
+    chmod +x build_tools/build_detect_platform
+  fi
+
+  if [ -e "build_detect_platform" ]
+  then
+    sed "s/tcmalloc/nothingnothingnothing/g" build_detect_platform > $TEST_TMPDIR/temp_build_file
+    rm -rf build_detect_platform 
+    cp $TEST_TMPDIR/temp_build_file build_detect_platform
+    chmod +x build_detect_platform 
+  fi
+
+  make ldb -j32
+
+  if [ -e "build_detect_platform" ]
+  then
+    git checkout -- build_detect_platform
+  fi
+
+  if [ -e "build_tools/build_detect_platform" ]
+  then
+    git checkout -- build_tools/build_detect_platform
+  fi
+  ' > temp_build_ldb.sh
+
+  sed "s/make ldb -j32/source temp_build_ldb.sh/g" tools/check_format_compatible.sh > tools/temp_check_format_compatible.sh
+  chmod +x tools/temp_check_format_compatible.sh
+  tools/temp_check_format_compatible.sh
+}
+
+FORMAT_COMPATIBLE_COMMANDS="[
+    {
+        'name':'Rocksdb Format Compatible tests',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Run RocksDB debug unit test',
+                'shell':'build_tools/rocksdb-lego-determinator run_format_compatible',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB no compression
+#
+run_no_compression()
+{
+  export TEST_TMPDIR=/dev/shm/rocksdb
+  rm -rf /dev/shm/rocksdb
+  mkdir /dev/shm/rocksdb
+  make clean
+  cat build_tools/fbcode_config.sh | grep -iv dzlib | grep -iv dlz4 | grep -iv dsnappy | grep -iv dbzip2 > .tmp.fbcode_config.sh
+  mv .tmp.fbcode_config.sh build_tools/fbcode_config.sh
+  cat Makefile | grep -v tools/ldb_test.py > .tmp.Makefile
+  mv .tmp.Makefile Makefile
+  make $DEBUG J=1 check
+}
+
+NO_COMPRESSION_COMMANDS="[
+    {
+        'name':'Rocksdb No Compression tests',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Run RocksDB debug unit test',
+                'shell':'build_tools/rocksdb-lego-determinator run_no_compression',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB regression
+#
+run_regression()
+{
+  time -v bash -vx ./build_tools/regression_build_test.sh $(mktemp -d  $WORKSPACE/leveldb.XXXX) $(mktemp leveldb_test_stats.XXXX)
+
+  # ======= report size to ODS ========
+
+  # parameters: $1 -- key, $2 -- value
+  function send_size_to_ods {
+    curl -s "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=rocksdb.build_size.$1&value=$2" \
+      --connect-timeout 60
+  }
+
+  # === normal build ===
+  make clean
+  make -j$(nproc) static_lib
+  send_size_to_ods static_lib $(stat --printf="%s" librocksdb.a)
+  strip librocksdb.a
+  send_size_to_ods static_lib_stripped $(stat --printf="%s" librocksdb.a)
+
+  make -j$(nproc) shared_lib
+  send_size_to_ods shared_lib $(stat --printf="%s" `readlink -f librocksdb.so`)
+  strip `readlink -f librocksdb.so`
+  send_size_to_ods shared_lib_stripped $(stat --printf="%s" `readlink -f librocksdb.so`)
+
+  # === lite build ===
+  make clean
+  OPT=-DROCKSDB_LITE make -j$(nproc) static_lib
+  send_size_to_ods static_lib_lite $(stat --printf="%s" librocksdb.a)
+  strip librocksdb.a
+  send_size_to_ods static_lib_lite_stripped $(stat --printf="%s" librocksdb.a)
+
+  OPT=-DROCKSDB_LITE make -j$(nproc) shared_lib
+  send_size_to_ods shared_lib_lite $(stat --printf="%s" `readlink -f librocksdb.so`)
+  strip `readlink -f librocksdb.so`
+  send_size_to_ods shared_lib_lite_stripped $(stat --printf="%s" `readlink -f librocksdb.so`)
+}
+
+REGRESSION_COMMANDS="[
+    {
+        'name':'Rocksdb regression commands',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Make and run script',
+                'shell':'build_tools/rocksdb-lego-determinator run_regression',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
    }
 ]"

@@ -235,12 +605,27 @@ case $1 in
  clang_unit)
    echo $CLANG_UNIT_TEST_COMMANDS
    ;;
+  clang_analyze)
+    echo $CLANG_ANALYZE_COMMANDS
+    ;;
+  code_cov)
+    echo $CODE_COV_COMMANDS
+    ;;
+  unity)
+    echo $UNITY_COMMANDS
+    ;;
  lite)
    echo $LITE_BUILD_COMMANDS
    ;;
+  lite_test)
+    echo $LITE_UNIT_TEST_COMMANDS
+    ;;
  stress_crash)
    echo $STRESS_CRASH_TEST_COMMANDS
    ;;
+  write_stress)
+    echo $WRITE_STRESS_COMMANDS
+    ;;
  asan)
    echo $ASAN_TEST_COMMANDS
    ;;
@@ -251,7 +636,28 @@ case $1 in
    echo $VALGRIND_TEST_COMMANDS
    ;;
  tsan)
-    echo $TSAN_TEST_COMMANDS
+    echo $TSAN_UNIT_TEST_COMMANDS
+    ;;
+  tsan_crash)
+    echo $TSAN_CRASH_TEST_COMMANDS
+    ;;
+  format_compatible)
+    echo $FORMAT_COMPATIBLE_COMMANDS
+    ;;
+  run_format_compatible)
+    run_format_compatible
+    ;;
+  no_compression)
+    echo $NO_COMPRESSION_COMMANDS
+    ;;
+  run_no_compression)
+    run_no_compression
+    ;;
+  regression)
+    echo $REGRESSION_COMMANDS
+    ;;
+  run_regression)
+    run_regression
    ;;
  *)
    echo "Invalid determinator command"
@@ -0,0 +1,273 @@
+# This script enables you running RocksDB tests by running
+# All the tests in paralell and utilizing all the cores
+# For db_test the script first lists and parses the tests
+# and then fires them up in parallel using async PS Job functionality
+# Run the script from the enlistment
+Param(
+  [switch]$EnableJE = $false,  # Use je executable
+  [string]$WorkFolder = "",  # Direct tests to use that folder
+  [int]$Limit = -1, # -1 means run all otherwise limit for testing purposes
+  [string]$Exclude = "", # Expect a comma separated list, no spaces
+  [string]$Run = "db_test",  # Run db_test|tests
+   # Number of async tasks that would run concurrently. Recommend a number below 64.
+   # However, CPU utlization really depends on the storage media. Recommend ram based disk.
+  [int]$Concurrency = 62
+)
+
+# Folders and commands must be fullpath to run assuming
+# the current folder is at the root of the git enlistment
+Get-Date
+
+# If running under Appveyor assume that root
+[string]$Appveyor = $Env:APPVEYOR_BUILD_FOLDER
+if($Appveyor -ne "") {
+    $RootFolder = $Appveyor
+} else {
+    $RootFolder = $PSScriptRoot -replace '\\build_tools', ''
+}
+
+$LogFolder = -Join($RootFolder, "\db_logs\")
+$BinariesFolder = -Join($RootFolder, "\build\Debug\")
+
+if($WorkFolder -eq "") {
+
+    # If TEST_TMPDIR is set use it    
+    [string]$var = $Env:TEST_TMPDIR
+    if($var -eq "") {
+        $WorkFolder = -Join($RootFolder, "\db_tests\")
+        $Env:TEST_TMPDIR = $WorkFolder
+    } else {
+        $WorkFolder = $var
+    }
+} else {
+# Override from a command line
+  $Env:TEST_TMPDIR = $WorkFolder
+}
+
+# Use JEMALLOC executables
+if($EnableJE) {
+    $db_test = -Join ($BinariesFolder, "db_test_je.exe")
+} else {
+    $db_test = -Join ($BinariesFolder, "db_test.exe")
+}
+
+Write-Output "Root: $RootFolder, WorkFolder: $WorkFolder"
+Write-Output "Binaries: $BinariesFolder exe: $db_test"
+
+#Exclusions that we do not want to run
+$ExcludeTests = New-Object System.Collections.Generic.HashSet[string]
+
+
+if($Exclude -ne "") {
+    Write-Host "Exclude: $Exclude"
+    $l = $Exclude -split ','
+    ForEach($t in $l) { $ExcludeTests.Add($t) | Out-Null }
+}
+
+# Create test directories in the current folder
+md -Path $WorkFolder -ErrorAction Ignore | Out-Null
+md -Path $LogFolder -ErrorAction Ignore | Out-Null
+
+# Extract the names of its tests by running db_test with --gtest_list_tests.
+# This filter removes the "#"-introduced comments, and expands to
+# fully-qualified names by changing input like this:
+#
+#   DBTest.
+#     Empty
+#     WriteEmptyBatch
+#   MultiThreaded/MultiThreadedDBTest.
+#     MultiThreaded/0  # GetParam() = 0
+#     MultiThreaded/1  # GetParam() = 1
+#
+# into this:
+#
+#   DBTest.Empty
+#   DBTest.WriteEmptyBatch
+#   MultiThreaded/MultiThreadedDBTest.MultiThreaded/0
+#   MultiThreaded/MultiThreadedDBTest.MultiThreaded/1
+# Output into the parameter in a form TestName -> Log File Name
+function Normalize-DbTests($HashTable) {
+
+    $Tests = @()
+# Run db_test to get a list of tests and store it into $a array
+    &$db_test --gtest_list_tests | tee -Variable Tests | Out-Null
+
+    # Current group
+    $Group=""
+
+    ForEach( $l in $Tests) {
+      # Trailing dot is a test group
+      if( $l -match "\.$") {
+        $Group = $l
+      }  else {
+        # Otherwise it is a test name, remove leading space
+        $test = $l -replace '^\s+',''
+        # remove trailing comment if any and create a log name
+        $test = $test -replace '\s+\#.*',''
+        $test = "$Group$test"
+
+        if($ExcludeTests.Contains($test)) {
+            continue
+        }
+
+        $test_log = $test -replace '[\./]','_'
+        $test_log += ".log"
+
+        # Add to a hashtable
+        $HashTable.Add($test, $test_log);
+      }
+    }
+}
+
+# The function scans build\Debug folder to discover
+# Test executables. It then populates a table with
+# Test executable name -> Log file
+function Discover-TestBinaries($HashTable) {
+
+    $Exclusions = @("db_test*", "db_sanity_test*")
+    if($EnableJE) {
+        $p = -join ($BinariesFolder, "*_test_je.exe")
+    } else {
+        $p = -join ($BinariesFolder, "*_test.exe")
+    }
+
+    dir -Path $p -Exclude $Exclusions | ForEach-Object {
+       $t = ($_.Name) -replace '.exe$', ''
+       if($ExcludeTests.Contains($t)) {
+            continue
+       }
+       $test_log = -join ($t, ".log")
+       $HashTable.Add($t, $test_log)
+    }
+}
+
+$TestToLog = [ordered]@{}
+
+if($Run -ceq "db_test") {
+    Normalize-DbTests -HashTable $TestToLog
+} elseif($Run -ceq "tests") {
+    Discover-TestBinaries -HashTable $TestToLog
+} else {
+    Write-Warning "Invalid -Run option value"
+    exit 2
+}
+
+
+Write-Host "Attempting to start: " ($TestToLog.Count) " tests"
+
+# Invoke a test with a filter and redirect all output
+$InvokeTestCase = {
+    param($exe, $test, $log);
+    &$exe --gtest_filter=$test > $log 2>&1
+}
+
+# Invoke all tests and redirect output
+$InvokeTestAsync = {
+    param($exe, $log)
+    &$exe > $log 2>&1
+}
+
+$jobs = @()
+$JobToLog = @{}
+# Test limiting factor here
+$count = 0
+
+[bool]$success = $true;
+
+# Wait for all to finish and get the results
+while(($JobToLog.Count -gt 0) -or
+      ($TestToLog.Count -gt 0)) {
+
+    # Make sure we have maximum concurrent jobs running if anything
+    # and the $Limit either not set or allows to proceed
+    while(($JobToLog.Count -lt $Concurrency) -and
+          (($TestToLog.Count -gt 0) -and
+          (($Limit -lt 0) -or ($count -lt $Limit)))) {
+
+
+        # We only need the first key
+        foreach($key in $TestToLog.keys) {
+            $k = $key
+            break
+        }
+
+        Write-Host "Starting $k"
+        $log_path = -join ($LogFolder, ($TestToLog.$k))
+
+        if($Run -ceq "db_test") {
+          $job = Start-Job -Name $k -ScriptBlock $InvokeTestCase -ArgumentList @($db_test,$k,$log_path)
+        } else {
+          [string]$Exe =  -Join ($BinariesFolder, $k)
+           $job = Start-Job -Name $k -ScriptBlock $InvokeTestAsync -ArgumentList @($exe,$log_path)
+        }
+
+        $JobToLog.Add($job, $log_path)
+        $TestToLog.Remove($k)
+
+        ++$count
+    }
+
+    if($JobToLog.Count -lt 1) {
+      break
+    }
+
+    $jobs = @()
+    foreach($k in $JobToLog.Keys) { $jobs += $k }
+
+    $completed = Wait-Job -Job $jobs -Any
+    $log = $JobToLog[$completed]
+    $JobToLog.Remove($completed)
+
+    $message = -join @($completed.Name, " State: ", ($completed.State))
+
+    $log_content = @(Get-Content $log)
+
+    if($completed.State -ne "Completed") {
+        $success = $false
+        Write-Warning $message
+        $log_content | Write-Warning
+    } else {
+        # Scan the log. If we find PASSED and no occurence of FAILED
+        # then it is a success
+        [bool]$pass_found = $false
+        ForEach($l in $log_content) {
+
+            if(($l -match "^\[\s+FAILED") -or
+               ($l -match "Assertion failed:")) {
+                $pass_found = $false
+                break
+            }
+
+            if(($l -match "^\[\s+PASSED") -or
+               ($l -match " : PASSED$") -or
+                ($l -match "^PASSED") -or
+                ($l -match "Passed all tests!") ) {
+                $pass_found = $true
+            }
+        }
+
+        if(!$pass_found) {
+            $success = $false;
+            Write-Warning $message
+            $log_content | Write-Warning
+        } else {
+            Write-Host $message
+        }
+    }
+
+    # Remove cached job info from the system
+    # Should be no output
+    Receive-Job -Job $completed | Out-Null
+}
+
+Get-Date
+
+if(!$success) {
+# This does not succeed killing off jobs quick
+# So we simply exit
+#    Remove-Job -Job $jobs -Force
+# indicate failure using this exit code
+    exit 12345
+ }
+
+ 
@@ -1,15 +0,0 @@
-#!/bin/bash
-#A shell script for Jenknis to run valgrind on rocksdb tests
-#Returns 0 on success when there are no failed tests 
-
-VALGRIND_DIR=build_tools/VALGRIND_LOGS
-make clean
-make -j$(nproc) valgrind_check
-NUM_FAILED_TESTS=$((`wc -l $VALGRIND_DIR/valgrind_failed_tests | awk '{print $1}'` - 1))
-if [ $NUM_FAILED_TESTS -lt 1 ]; then
-  echo No tests have valgrind errors
-  exit 0
-else
-  cat $VALGRIND_DIR/valgrind_failed_tests
-  exit 1
-fi
@@ -26,6 +26,7 @@
 #include "rocksdb/options.h"
 #include "rocksdb/table.h"
 #include "table/block_based_table_builder.h"
+#include "table/internal_iterator.h"
 #include "util/file_reader_writer.h"
 #include "util/iostats_context_imp.h"
 #include "util/stop_watch.h"
@@ -40,22 +41,25 @@ TableBuilder* NewTableBuilder(
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    WritableFileWriter* file, const CompressionType compression_type,
+    uint32_t column_family_id, WritableFileWriter* file,
+    const CompressionType compression_type,
    const CompressionOptions& compression_opts, const bool skip_filters) {
  return ioptions.table_factory->NewTableBuilder(
      TableBuilderOptions(ioptions, internal_comparator,
                          int_tbl_prop_collector_factories, compression_type,
                          compression_opts, skip_filters),
-      file);
+      column_family_id, file);
 }

 Status BuildTable(
    const std::string& dbname, Env* env, const ImmutableCFOptions& ioptions,
-    const EnvOptions& env_options, TableCache* table_cache, Iterator* iter,
-    FileMetaData* meta, const InternalKeyComparator& internal_comparator,
+    const EnvOptions& env_options, TableCache* table_cache,
+    InternalIterator* iter, FileMetaData* meta,
+    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    std::vector<SequenceNumber> snapshots, const CompressionType compression,
+    uint32_t column_family_id, std::vector<SequenceNumber> snapshots,
+    const CompressionType compression,
    const CompressionOptions& compression_opts, bool paranoid_file_checks,
    InternalStats* internal_stats, const Env::IOPriority io_priority,
    TableProperties* table_properties) {
@@ -72,7 +76,7 @@ Status BuildTable(
    unique_ptr<WritableFileWriter> file_writer;
    {
      unique_ptr<WritableFile> file;
-      s = env->NewWritableFile(fname, &file, env_options);
+      s = NewWritableFile(env, fname, &file, env_options);
      if (!s.ok()) {
        return s;
      }
@@ -82,13 +86,14 @@ Status BuildTable(

      builder = NewTableBuilder(
          ioptions, internal_comparator, int_tbl_prop_collector_factories,
-          file_writer.get(), compression, compression_opts);
+          column_family_id, file_writer.get(), compression, compression_opts);
    }

-    MergeHelper merge(internal_comparator.user_comparator(),
-                      ioptions.merge_operator, ioptions.info_log,
+    MergeHelper merge(env, internal_comparator.user_comparator(),
+                      ioptions.merge_operator, nullptr, ioptions.info_log,
                      ioptions.min_partial_merge_operands,
-                      true /* internal key corruption is not ok */);
+                      true /* internal key corruption is not ok */,
+                      snapshots.empty() ? 0 : snapshots.back());

    CompactionIterator c_iter(iter, internal_comparator.user_comparator(),
                              &merge, kMaxSequenceNumber, &snapshots, env,
@@ -103,20 +108,21 @@ Status BuildTable(
      // TODO(noetzli): Update stats after flush, too.
      if (io_priority == Env::IO_HIGH &&
          IOSTATS(bytes_written) >= kReportFlushIOStatsEvery) {
-        ThreadStatusUtil::IncreaseThreadOperationProperty(
+        ThreadStatusUtil::SetThreadOperationProperty(
            ThreadStatus::FLUSH_BYTES_WRITTEN, IOSTATS(bytes_written));
-        IOSTATS_RESET(bytes_written);
      }
    }

    // Finish and check for builder errors
+    bool empty = builder->NumEntries() == 0;
    s = c_iter.status();
-    if (s.ok()) {
-      s = builder->Finish();
-    } else {
+    if (!s.ok() || empty) {
      builder->Abandon();
+    } else {
+      s = builder->Finish();
    }
-    if (s.ok()) {
+
+    if (s.ok() && !empty) {
      meta->fd.file_size = builder->FileSize();
      meta->marked_for_compaction = builder->NeedCompact();
      assert(meta->fd.GetFileSize() > 0);
@@ -127,28 +133,27 @@ Status BuildTable(
    delete builder;

    // Finish and check for file errors
-    if (s.ok() && !ioptions.disable_data_sync) {
+    if (s.ok() && !empty && !ioptions.disable_data_sync) {
      StopWatch sw(env, ioptions.statistics, TABLE_SYNC_MICROS);
      file_writer->Sync(ioptions.use_fsync);
    }
-    if (s.ok()) {
+    if (s.ok() && !empty) {
      s = file_writer->Close();
    }

-    if (s.ok()) {
+    if (s.ok() && !empty) {
      // Verify that the table is usable
-      Iterator* it = table_cache->NewIterator(
+      std::unique_ptr<InternalIterator> it(table_cache->NewIterator(
          ReadOptions(), env_options, internal_comparator, meta->fd, nullptr,
          (internal_stats == nullptr) ? nullptr
                                      : internal_stats->GetFileReadHist(0),
-          false);
+          false));
      s = it->status();
      if (s.ok() && paranoid_file_checks) {
-        for (it->SeekToFirst(); it->Valid(); it->Next()) {}
+        for (it->SeekToFirst(); it->Valid(); it->Next()) {
+        }
        s = it->status();
      }
-
-      delete it;
    }
  }

@@ -157,9 +162,7 @@ Status BuildTable(
    s = iter->status();
  }

-  if (s.ok() && meta->fd.GetFileSize() > 0) {
-    // Keep it
-  } else {
+  if (!s.ok() || meta->fd.GetFileSize() == 0) {
    env->DeleteFile(fname);
  }
  return s;
@@ -31,13 +31,15 @@ class VersionEdit;
 class TableBuilder;
 class WritableFileWriter;
 class InternalStats;
+class InternalIterator;

 TableBuilder* NewTableBuilder(
    const ImmutableCFOptions& options,
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    WritableFileWriter* file, const CompressionType compression_type,
+    uint32_t column_family_id, WritableFileWriter* file,
+    const CompressionType compression_type,
    const CompressionOptions& compression_opts,
    const bool skip_filters = false);

@@ -48,11 +50,13 @@ TableBuilder* NewTableBuilder(
 // zero, and no Table file will be produced.
 extern Status BuildTable(
    const std::string& dbname, Env* env, const ImmutableCFOptions& options,
-    const EnvOptions& env_options, TableCache* table_cache, Iterator* iter,
-    FileMetaData* meta, const InternalKeyComparator& internal_comparator,
+    const EnvOptions& env_options, TableCache* table_cache,
+    InternalIterator* iter, FileMetaData* meta,
+    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    std::vector<SequenceNumber> snapshots, const CompressionType compression,
+    uint32_t column_family_id, std::vector<SequenceNumber> snapshots,
+    const CompressionType compression,
    const CompressionOptions& compression_opts, bool paranoid_file_checks,
    InternalStats* internal_stats,
    const Env::IOPriority io_priority = Env::IO_HIGH,
@@ -1288,6 +1288,11 @@ void rocksdb_block_based_options_set_cache_index_and_filter_blocks(
  options->rep.cache_index_and_filter_blocks = v;
 }

+void rocksdb_block_based_options_set_skip_table_builder_flush(
+    rocksdb_block_based_table_options_t* options, unsigned char v) {
+  options->rep.skip_table_builder_flush = v;
+}
+
 void rocksdb_options_set_block_based_table_factory(
    rocksdb_options_t *opt,
    rocksdb_block_based_table_options_t* table_options) {
@@ -1687,6 +1692,11 @@ void rocksdb_options_set_keep_log_file_num(rocksdb_options_t* opt, size_t v) {
  opt->rep.keep_log_file_num = v;
 }

+void rocksdb_options_set_recycle_log_file_num(rocksdb_options_t* opt,
+                                              size_t v) {
+  opt->rep.recycle_log_file_num = v;
+}
+
 void rocksdb_options_set_soft_rate_limit(rocksdb_options_t* opt, double v) {
  opt->rep.soft_rate_limit = v;
 }
@@ -27,9 +27,9 @@
 #include "db/version_set.h"
 #include "db/write_controller.h"
 #include "db/writebuffer.h"
+#include "memtable/hash_skiplist_rep.h"
 #include "util/autovector.h"
 #include "util/compression.h"
-#include "util/hash_skiplist_rep.h"
 #include "util/options_helper.h"
 #include "util/thread_status_util.h"
 #include "util/xfunc.h"
@@ -446,15 +446,33 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
    } else if (vstorage->l0_delay_trigger_count() >=
               mutable_cf_options.level0_stop_writes_trigger) {
      write_controller_token_ = write_controller->GetStopToken();
-      internal_stats_->AddCFStats(InternalStats::LEVEL0_NUM_FILES, 1);
+      internal_stats_->AddCFStats(InternalStats::LEVEL0_NUM_FILES_TOTAL, 1);
+      if (compaction_picker_->IsLevel0CompactionInProgress()) {
+        internal_stats_->AddCFStats(
+            InternalStats::LEVEL0_NUM_FILES_WITH_COMPACTION, 1);
+      }
      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
          "[%s] Stopping writes because we have %d level-0 files",
          name_.c_str(), vstorage->l0_delay_trigger_count());
+    } else if (mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
+               vstorage->estimated_compaction_needed_bytes() >=
+                   mutable_cf_options.hard_pending_compaction_bytes_limit) {
+      write_controller_token_ = write_controller->GetStopToken();
+      internal_stats_->AddCFStats(
+          InternalStats::HARD_PENDING_COMPACTION_BYTES_LIMIT, 1);
+      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+          "[%s] Stopping writes because estimated pending compaction "
+          "bytes exceed %" PRIu64,
+          name_.c_str(), vstorage->estimated_compaction_needed_bytes());
    } else if (mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
               vstorage->l0_delay_trigger_count() >=
                   mutable_cf_options.level0_slowdown_writes_trigger) {
      write_controller_token_ = write_controller->GetDelayToken();
-      internal_stats_->AddCFStats(InternalStats::LEVEL0_SLOWDOWN, 1);
+      internal_stats_->AddCFStats(InternalStats::LEVEL0_SLOWDOWN_TOTAL, 1);
+      if (compaction_picker_->IsLevel0CompactionInProgress()) {
+        internal_stats_->AddCFStats(
+            InternalStats::LEVEL0_SLOWDOWN_WITH_COMPACTION, 1);
+      }
      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
          "[%s] Stalling writes because we have %d level-0 files",
          name_.c_str(), vstorage->l0_delay_trigger_count());
@@ -10,6 +10,7 @@
 #include <algorithm>
 #include <vector>
 #include <string>
+#include <thread>

 #include "db/db_impl.h"
 #include "rocksdb/db.h"
@@ -19,6 +20,7 @@
 #include "util/testharness.h"
 #include "util/testutil.h"
 #include "util/coding.h"
+#include "util/sync_point.h"
 #include "utilities/merge_operators.h"

 namespace rocksdb {
@@ -1196,6 +1198,67 @@ TEST_F(ColumnFamilyTest, ReadDroppedColumnFamily) {
  }
 }

+TEST_F(ColumnFamilyTest, FlushAndDropRaceCondition) {
+  db_options_.create_missing_column_families = true;
+  Open({"default", "one"});
+  ColumnFamilyOptions options;
+  options.level0_file_num_compaction_trigger = 100;
+  options.level0_slowdown_writes_trigger = 200;
+  options.level0_stop_writes_trigger = 200;
+  options.max_write_buffer_number = 20;
+  options.write_buffer_size = 100000;  // small write buffer size
+  Reopen({options, options});
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"VersionSet::LogAndApply::ColumnFamilyDrop:1"
+        "FlushJob::InstallResults"},
+       {"FlushJob::InstallResults",
+        "VersionSet::LogAndApply::ColumnFamilyDrop:2", }});
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  test::SleepingBackgroundTask sleeping_task;
+
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
+                 Env::Priority::HIGH);
+
+  // 1MB should create ~10 files for each CF
+  int kKeysNum = 10000;
+  PutRandomData(1, kKeysNum, 100);
+
+  std::vector<std::thread> threads;
+  threads.emplace_back([&] { ASSERT_OK(db_->DropColumnFamily(handles_[1])); });
+
+  sleeping_task.WakeUp();
+  sleeping_task.WaitUntilDone();
+  sleeping_task.Reset();
+  // now we sleep again. this is just so we're certain that flush job finished
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
+                 Env::Priority::HIGH);
+  sleeping_task.WakeUp();
+  sleeping_task.WaitUntilDone();
+
+  {
+    // Since we didn't delete CF handle, RocksDB's contract guarantees that
+    // we're still able to read dropped CF
+    std::unique_ptr<Iterator> iterator(
+        db_->NewIterator(ReadOptions(), handles_[1]));
+    int count = 0;
+    for (iterator->SeekToFirst(); iterator->Valid(); iterator->Next()) {
+      ASSERT_OK(iterator->status());
+      ++count;
+    }
+    ASSERT_OK(iterator->status());
+    ASSERT_EQ(count, kKeysNum);
+  }
+  for (auto& t : threads) {
+    t.join();
+  }
+
+  Close();
+  Destroy();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -40,6 +40,47 @@ void Compaction::SetInputVersion(Version* _input_version) {
  edit_.SetColumnFamily(cfd_->GetID());
 }

+void Compaction::GetBoundaryKeys(
+    VersionStorageInfo* vstorage,
+    const std::vector<CompactionInputFiles>& inputs, Slice* smallest_user_key,
+    Slice* largest_user_key) {
+  bool initialized = false;
+  const Comparator* ucmp = vstorage->InternalComparator()->user_comparator();
+  for (uint32_t i = 0; i < inputs.size(); ++i) {
+    if (inputs[i].files.empty()) {
+      continue;
+    }
+    if (inputs[i].level == 0) {
+      // we need to consider all files on level 0
+      for (const auto* f : inputs[i].files) {
+        const Slice& start_user_key = f->smallest.user_key();
+        if (!initialized ||
+            ucmp->Compare(start_user_key, *smallest_user_key) < 0) {
+          *smallest_user_key = start_user_key;
+        }
+        const Slice& end_user_key = f->largest.user_key();
+        if (!initialized ||
+            ucmp->Compare(end_user_key, *largest_user_key) > 0) {
+          *largest_user_key = end_user_key;
+        }
+        initialized = true;
+      }
+    } else {
+      // we only need to consider the first and last file
+      const Slice& start_user_key = inputs[i].files[0]->smallest.user_key();
+      if (!initialized ||
+          ucmp->Compare(start_user_key, *smallest_user_key) < 0) {
+        *smallest_user_key = start_user_key;
+      }
+      const Slice& end_user_key = inputs[i].files.back()->largest.user_key();
+      if (!initialized || ucmp->Compare(end_user_key, *largest_user_key) > 0) {
+        *largest_user_key = end_user_key;
+      }
+      initialized = true;
+    }
+  }
+}
+
 // helper function to determine if compaction is creating files at the
 // bottommost level
 bool Compaction::IsBottommostLevel(
@@ -50,15 +91,40 @@ bool Compaction::IsBottommostLevel(
    return false;
  }

-  // checks whether there are files living beyond the output_level.
+  Slice smallest_key, largest_key;
+  GetBoundaryKeys(vstorage, inputs, &smallest_key, &largest_key);
+
+  // Checks whether there are files living beyond the output_level.
+  // If lower levels have files, it checks for overlap between files
+  // if the compaction process and those files.
+  // Bottomlevel optimizations can be made if there are no files in
+  // lower levels or if there is no overlap with the files in
+  // the lower levels.
  for (int i = output_level + 1; i < vstorage->num_levels(); i++) {
-    if (vstorage->NumLevelFiles(i) > 0) {
+    // It is not the bottommost level if there are files in higher
+    // levels when the output level is 0 or if there are files in
+    // higher levels which overlap with files to be compacted.
+    // output_level == 0 means that we want it to be considered
+    // s the bottommost level only if the last file on the level
+    // is a part of the files to be compacted - this is verified by
+    // the first if condition in this function
+    if (vstorage->NumLevelFiles(i) > 0 &&
+        (output_level == 0 ||
+         vstorage->OverlapInLevel(i, &smallest_key, &largest_key))) {
      return false;
    }
  }
  return true;
 }

+// test function to validate the functionality of IsBottommostLevel()
+// function -- determines if compaction with inputs and storage is bottommost
+bool Compaction::TEST_IsBottommostLevel(
+    int output_level, VersionStorageInfo* vstorage,
+    const std::vector<CompactionInputFiles>& inputs) {
+  return IsBottommostLevel(output_level, vstorage, inputs);
+}
+
 bool Compaction::IsFullCompaction(
    VersionStorageInfo* vstorage,
    const std::vector<CompactionInputFiles>& inputs) {
@@ -373,6 +439,7 @@ std::unique_ptr<CompactionFilter> Compaction::CreateCompactionFilter() const {
  CompactionFilter::Context context;
  context.is_full_compaction = is_full_compaction_;
  context.is_manual_compaction = is_manual_compaction_;
+  context.column_family_id = cfd_->GetID();
  return cfd_->ioptions()->compaction_filter_factory->CreateCompactionFilter(
      context);
 }
@@ -204,15 +204,35 @@ class Compaction {
  // Should this compaction be broken up into smaller ones run in parallel?
  bool ShouldFormSubcompactions() const;

+  // test function to validate the functionality of IsBottommostLevel()
+  // function -- determines if compaction with inputs and storage is bottommost
+  static bool TEST_IsBottommostLevel(
+      int output_level, VersionStorageInfo* vstorage,
+      const std::vector<CompactionInputFiles>& inputs);
+
+  TablePropertiesCollection GetOutputTableProperties() const {
+    return output_table_properties_;
+  }
+
+  void SetOutputTableProperties(TablePropertiesCollection tp) {
+    output_table_properties_ = std::move(tp);
+  }
+
 private:
  // mark (or clear) all files that are being compacted
  void MarkFilesBeingCompacted(bool mark_as_compacted);

+  // get the smallest and largest key present in files to be compacted
+  static void GetBoundaryKeys(VersionStorageInfo* vstorage,
+                              const std::vector<CompactionInputFiles>& inputs,
+                              Slice* smallest_key, Slice* largest_key);
+
  // helper function to determine if compaction with inputs and storage is
  // bottommost
  static bool IsBottommostLevel(
      int output_level, VersionStorageInfo* vstorage,
      const std::vector<CompactionInputFiles>& inputs);
+
  static bool IsFullCompaction(VersionStorageInfo* vstorage,
                               const std::vector<CompactionInputFiles>& inputs);

@@ -261,6 +281,9 @@ class Compaction {

  // Does input compression match the output compression?
  bool InputCompressionMatchesOutput() const;
+
+  // table properties of output files
+  TablePropertiesCollection output_table_properties_;
 };

 // Utility function
@@ -6,22 +6,21 @@
 //  of patent rights can be found in the PATENTS file in the same directory.

 #include "db/compaction_iterator.h"
+#include "table/internal_iterator.h"

 namespace rocksdb {

 CompactionIterator::CompactionIterator(
-    Iterator* input, const Comparator* cmp, MergeHelper* merge_helper,
+    InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
    SequenceNumber last_sequence, std::vector<SequenceNumber>* snapshots,
-    Env* env, bool expect_valid_internal_key, Statistics* stats,
-    Compaction* compaction, const CompactionFilter* compaction_filter,
-    LogBuffer* log_buffer)
+    Env* env, bool expect_valid_internal_key, Compaction* compaction,
+    const CompactionFilter* compaction_filter, LogBuffer* log_buffer)
    : input_(input),
      cmp_(cmp),
      merge_helper_(merge_helper),
      snapshots_(snapshots),
      env_(env),
      expect_valid_internal_key_(expect_valid_internal_key),
-      stats_(stats),
      compaction_(compaction),
      compaction_filter_(compaction_filter),
      log_buffer_(log_buffer),
@@ -71,6 +70,10 @@ void CompactionIterator::Next() {
      // MergeUntil stops when it encounters a corrupt key and does not
      // include them in the result, so we expect the keys here to be valid.
      assert(valid_key);
+      // Keep current_key_ in sync.
+      current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+      key_ = current_key_.GetKey();
+      ikey_.user_key = current_key_.GetUserKey();
      valid_ = true;
    } else {
      // MergeHelper moves the iterator to the first record after the merged
@@ -79,8 +82,11 @@ void CompactionIterator::Next() {
      NextFromInput();
    }
  } else {
-    // Only advance the input iterator if there is no merge output.
-    input_->Next();
+    // Only advance the input iterator if there is no merge output and the
+    // iterator is not already at the next record.
+    if (!at_next_) {
+      input_->Next();
+    }
    NextFromInput();
  }

@@ -88,9 +94,10 @@ void CompactionIterator::Next() {
 }

 void CompactionIterator::NextFromInput() {
+  at_next_ = false;
  valid_ = false;

-  while (input_->Valid()) {
+  while (!valid_ && input_->Valid()) {
    key_ = input_->key();
    value_ = input_->value();
    iter_stats_.num_input_records++;
@@ -100,10 +107,11 @@ void CompactionIterator::NextFromInput() {
      // and let the caller decide what to do with it.
      // TODO(noetzli): We should have a more elegant solution for this.
      if (expect_valid_internal_key_) {
-        assert(!"corrupted internal key is not expected");
+        assert(!"Corrupted internal key not expected.");
+        status_ = Status::Corruption("Corrupted internal key not expected.");
        break;
      }
-      current_user_key_.Clear();
+      key_ = current_key_.SetKey(key_);
      has_current_user_key_ = false;
      current_user_key_sequence_ = kMaxSequenceNumber;
      current_user_key_snapshot_ = 0;
@@ -113,22 +121,26 @@ void CompactionIterator::NextFromInput() {
    }

    // Update input statistics
-    if (ikey_.type == kTypeDeletion) {
+    if (ikey_.type == kTypeDeletion || ikey_.type == kTypeSingleDeletion) {
      iter_stats_.num_input_deletion_records++;
    }
    iter_stats_.total_input_raw_key_bytes += key_.size();
    iter_stats_.total_input_raw_value_bytes += value_.size();

+    // Check whether the user key changed. After this if statement current_key_
+    // is a copy of the current input key (maybe converted to a delete by the
+    // compaction filter). ikey_.user_key is pointing to the copy.
    if (!has_current_user_key_ ||
-        cmp_->Compare(ikey_.user_key, current_user_key_.GetKey()) != 0) {
+        !cmp_->Equal(ikey_.user_key, current_user_key_)) {
      // First occurrence of this user key
-      current_user_key_.SetKey(ikey_.user_key);
+      key_ = current_key_.SetKey(key_, &ikey_);
+      current_user_key_ = ikey_.user_key;
      has_current_user_key_ = true;
      current_user_key_sequence_ = kMaxSequenceNumber;
      current_user_key_snapshot_ = 0;
      // apply the compaction filter to the first occurrence of the user key
      if (compaction_filter_ != nullptr && ikey_.type == kTypeValue &&
-          (visible_at_tip_ || latest_snapshot_)) {
+          (visible_at_tip_ || ikey_.sequence > latest_snapshot_)) {
        // If the user has specified a compaction filter and the sequence
        // number is greater than any external snapshot, then invoke the
        // filter. If the return value of the compaction filter is true,
@@ -145,13 +157,9 @@ void CompactionIterator::NextFromInput() {
              env_ != nullptr ? timer.ElapsedNanos() : 0;
        }
        if (to_delete) {
-          // make a copy of the original key and convert it to a delete
-          delete_key_.SetInternalKey(ExtractUserKey(key_), ikey_.sequence,
-                                     kTypeDeletion);
-          // anchor the key again
-          key_ = delete_key_.GetKey();
-          // needed because ikey_ is backed by key
-          ParseInternalKey(key_, &ikey_);
+          // convert the current key to a delete
+          ikey_.type = kTypeDeletion;
+          current_key_.UpdateInternalKey(ikey_.sequence, kTypeDeletion);
          // no value associated with delete
          value_.clear();
          iter_stats_.num_record_drop_user++;
@@ -159,6 +167,12 @@ void CompactionIterator::NextFromInput() {
          value_ = compaction_filter_value_;
        }
      }
+    } else {
+      // Update the current key to reflect the new sequence number/type without
+      // copying the user key.
+      current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+      key_ = current_key_.GetKey();
+      ikey_.user_key = current_key_.GetUserKey();
    }

    // If there are no snapshots, then this kv affect visibility at tip.
@@ -173,7 +187,58 @@ void CompactionIterator::NextFromInput() {
        visible_at_tip_ ? visible_at_tip_ : findEarliestVisibleSnapshot(
                                                ikey_.sequence, &prev_snapshot);

-    if (last_snapshot == current_user_key_snapshot_) {
+    if (ikey_.type == kTypeSingleDeletion) {
+      ParsedInternalKey next_ikey;
+      input_->Next();
+
+      // Check whether the current key is valid, not corrupt and the same
+      // as the single delete.
+      if (input_->Valid() && ParseInternalKey(input_->key(), &next_ikey) &&
+          cmp_->Equal(ikey_.user_key, next_ikey.user_key)) {
+        // Mixing single deletes and merges is not supported. Consecutive
+        // single deletes are not valid.
+        if (next_ikey.type != kTypeValue) {
+          assert(false);
+          status_ =
+              Status::InvalidArgument("Put expected after single delete.");
+          break;
+        }
+
+        // Check whether the current key belongs to the same snapshot as the
+        // single delete.
+        if (prev_snapshot == 0 || next_ikey.sequence > prev_snapshot) {
+          // Found the matching value, we can drop the single delete and the
+          // value.
+          ++iter_stats_.num_record_drop_hidden;
+          ++iter_stats_.num_record_drop_obsolete;
+          input_->Next();
+        } else {
+          // We hit the next snapshot without hitting a put, so the iterator
+          // returns the single delete.
+          valid_ = true;
+        }
+      } else {
+        // We are at the end of the input, could not parse the next key, or hit
+        // the next key. The iterator returns the single delete if the key
+        // possibly exists beyond the current output level.  We set
+        // has_current_user_key to false so that if the iterator is at the next
+        // key, we do not compare it again against the previous key at the next
+        // iteration. If the next key is corrupt, we return before the
+        // comparison, so the value of has_current_user_key does not matter.
+        has_current_user_key_ = false;
+        if (compaction_ != nullptr &&
+            compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
+                                                       &level_ptrs_)) {
+          ++iter_stats_.num_record_drop_obsolete;
+        } else {
+          valid_ = true;
+        }
+      }
+
+      if (valid_) {
+        at_next_ = true;
+      }
+    } else if (last_snapshot == current_user_key_snapshot_) {
      // If the earliest snapshot is which this key is visible in
      // is the same as the visibility of a previous instance of the
      // same key, then this kv is not visible in any snapshot.
@@ -181,6 +246,7 @@ void CompactionIterator::NextFromInput() {
      // TODO: why not > ?
      assert(last_sequence >= current_user_key_sequence_);
      ++iter_stats_.num_record_drop_hidden;  // (A)
+      input_->Next();
    } else if (compaction_ != nullptr && ikey_.type == kTypeDeletion &&
               ikey_.sequence <= earliest_snapshot_ &&
               compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
@@ -197,6 +263,7 @@ void CompactionIterator::NextFromInput() {
      //     few iterations of this loop (by rule (A) above).
      // Therefore this deletion marker is obsolete and can be dropped.
      ++iter_stats_.num_record_drop_obsolete;
+      input_->Next();
    } else if (ikey_.type == kTypeMerge) {
      if (!merge_helper_->HasOperator()) {
        LogToBuffer(log_buffer_, "Options::merge_operator is null.");
@@ -209,27 +276,33 @@ void CompactionIterator::NextFromInput() {
      // have hit (A)
      // We encapsulate the merge related state machine in a different
      // object to minimize change to the existing flow.
-      merge_helper_->MergeUntil(input_, prev_snapshot, bottommost_level_,
-                                stats_, env_);
+      merge_helper_->MergeUntil(input_, prev_snapshot, bottommost_level_);
      merge_out_iter_.SeekToFirst();

-      // NOTE: key, value, and ikey_ refer to old entries.
-      //       These will be correctly set below.
-      key_ = merge_out_iter_.key();
-      value_ = merge_out_iter_.value();
-      bool valid_key __attribute__((__unused__)) =
-          ParseInternalKey(key_, &ikey_);
-      // MergeUntil stops when it encounters a corrupt key and does not
-      // include them in the result, so we expect the keys here to valid.
-      assert(valid_key);
-      valid_ = true;
-      break;
+      if (merge_out_iter_.Valid()) {
+        // NOTE: key, value, and ikey_ refer to old entries.
+        //       These will be correctly set below.
+        key_ = merge_out_iter_.key();
+        value_ = merge_out_iter_.value();
+        bool valid_key __attribute__((__unused__)) =
+            ParseInternalKey(key_, &ikey_);
+        // MergeUntil stops when it encounters a corrupt key and does not
+        // include them in the result, so we expect the keys here to valid.
+        assert(valid_key);
+        // Keep current_key_ in sync.
+        current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+        key_ = current_key_.GetKey();
+        ikey_.user_key = current_key_.GetUserKey();
+        valid_ = true;
+      } else {
+        // all merge operands were filtered out. reset the user key, since the
+        // batch consumed by the merge operator should not shadow any keys
+        // coming after the merges
+        has_current_user_key_ = false;
+      }
    } else {
      valid_ = true;
-      break;
    }
-
-    input_->Next();
  }
 }

@@ -240,12 +313,9 @@ void CompactionIterator::PrepareOutput() {
  // then we can squash the seqno to zero.
  if (bottommost_level_ && valid_ && ikey_.sequence < earliest_snapshot_ &&
      ikey_.type != kTypeMerge) {
-    assert(ikey_.type != kTypeDeletion);
-    // make a copy because updating in place would cause problems
-    // with the priority queue that is managing the input key iterator
-    updated_key_.assign(key_.data(), key_.size());
-    UpdateInternalKey(&updated_key_, (uint64_t)0, ikey_.type);
-    key_ = Slice(updated_key_);
+    assert(ikey_.type != kTypeDeletion && ikey_.type != kTypeSingleDeletion);
+    ikey_.sequence = 0;
+    current_key_.UpdateInternalKey(0, ikey_.type);
  }
 }

@@ -37,11 +37,10 @@ struct CompactionIteratorStats {

 class CompactionIterator {
 public:
-  CompactionIterator(Iterator* input, const Comparator* cmp,
+  CompactionIterator(InternalIterator* input, const Comparator* cmp,
                     MergeHelper* merge_helper, SequenceNumber last_sequence,
                     std::vector<SequenceNumber>* snapshots, Env* env,
                     bool expect_valid_internal_key,
-                     Statistics* stats = nullptr,
                     Compaction* compaction = nullptr,
                     const CompactionFilter* compaction_filter = nullptr,
                     LogBuffer* log_buffer = nullptr);
@@ -64,7 +63,7 @@ class CompactionIterator {
  const Status& status() const { return status_; }
  const ParsedInternalKey& ikey() const { return ikey_; }
  bool Valid() const { return valid_; }
-  Slice user_key() const { return current_user_key_.GetKey(); }
+  const Slice& user_key() const { return current_user_key_; }
  const CompactionIteratorStats& iter_stats() const { return iter_stats_; }

 private:
@@ -85,13 +84,12 @@ class CompactionIterator {
  inline SequenceNumber findEarliestVisibleSnapshot(
      SequenceNumber in, SequenceNumber* prev_snapshot);

-  Iterator* input_;
+  InternalIterator* input_;
  const Comparator* cmp_;
  MergeHelper* merge_helper_;
  const std::vector<SequenceNumber>* snapshots_;
  Env* env_;
  bool expect_valid_internal_key_;
-  Statistics* stats_;
  Compaction* compaction_;
  const CompactionFilter* compaction_filter_;
  LogBuffer* log_buffer_;
@@ -102,18 +100,32 @@ class CompactionIterator {
  SequenceNumber latest_snapshot_;

  // State
+  //
+  // Points to a copy of the current compaction iterator output (current_key_)
+  // if valid_.
  Slice key_;
+  // Points to the value in the underlying iterator that corresponds to the
+  // current output.
  Slice value_;
+  // The status is OK unless compaction iterator encounters a merge operand
+  // while not having a merge operator defined.
  Status status_;
+  // Stores the user key, sequence number and type of the current compaction
+  // iterator output (or current key in the underlying iterator during
+  // NextFromInput()).
  ParsedInternalKey ikey_;
+  // Stores whether ikey_.user_key is valid. If set to false, the user key is
+  // not compared against the current key in the underlying iterator.
  bool has_current_user_key_ = false;
-  IterKey current_user_key_;
+  bool at_next_ = false;  // If false, the iterator
+  // Holds a copy of the current compaction iterator output (or current key in
+  // the underlying iterator during NextFromInput()).
+  IterKey current_key_;
+  Slice current_user_key_;
  SequenceNumber current_user_key_sequence_;
  SequenceNumber current_user_key_snapshot_;
  MergeOutputIterator merge_out_iter_;
-  std::string updated_key_;
  std::string compaction_filter_value_;
-  IterKey delete_key_;
  // "level_ptrs" holds indices that remember which file of an associated
  // level we were last checking during the last call to compaction->
  // KeyNotExistsBeyondOutputLevel(). This allows future calls to the function
@@ -0,0 +1,71 @@
+//  Copyright (c) 2013, 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.
+
+#include "db/compaction_iterator.h"
+#include "util/testharness.h"
+#include "util/testutil.h"
+
+namespace rocksdb {
+
+class CompactionIteratorTest : public testing::Test {
+ public:
+  CompactionIteratorTest() : cmp_(BytewiseComparator()), snapshots_({}) {}
+
+  void InitIterator(const std::vector<std::string>& ks,
+                    const std::vector<std::string>& vs,
+                    SequenceNumber last_sequence) {
+    merge_helper_.reset(new MergeHelper(Env::Default(), cmp_, nullptr, nullptr,
+                                        nullptr, 0U, false, 0));
+    iter_.reset(new test::VectorIterator(ks, vs));
+    iter_->SeekToFirst();
+    c_iter_.reset(new CompactionIterator(iter_.get(), cmp_, merge_helper_.get(),
+                                         last_sequence, &snapshots_,
+                                         Env::Default(), false));
+  }
+
+  const Comparator* cmp_;
+  std::vector<SequenceNumber> snapshots_;
+  std::unique_ptr<MergeHelper> merge_helper_;
+  std::unique_ptr<test::VectorIterator> iter_;
+  std::unique_ptr<CompactionIterator> c_iter_;
+};
+
+// It is possible that the output of the compaction iterator is empty even if
+// the input is not.
+TEST_F(CompactionIteratorTest, EmptyResult) {
+  InitIterator({test::KeyStr("a", 5, kTypeSingleDeletion),
+                test::KeyStr("a", 3, kTypeValue)},
+               {"", "val"}, 5);
+  c_iter_->SeekToFirst();
+  ASSERT_FALSE(c_iter_->Valid());
+}
+
+// If there is a corruption after a single deletion, the corrupted key should
+// be preserved.
+TEST_F(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
+  InitIterator({test::KeyStr("a", 5, kTypeSingleDeletion),
+                test::KeyStr("a", 3, kTypeValue, true),
+                test::KeyStr("b", 10, kTypeValue)},
+               {"", "val", "val2"}, 10);
+  c_iter_->SeekToFirst();
+  ASSERT_TRUE(c_iter_->Valid());
+  ASSERT_EQ(test::KeyStr("a", 5, kTypeSingleDeletion),
+            c_iter_->key().ToString());
+  c_iter_->Next();
+  ASSERT_TRUE(c_iter_->Valid());
+  ASSERT_EQ(test::KeyStr("a", 3, kTypeValue, true), c_iter_->key().ToString());
+  c_iter_->Next();
+  ASSERT_TRUE(c_iter_->Valid());
+  ASSERT_EQ(test::KeyStr("b", 10, kTypeValue), c_iter_->key().ToString());
+  c_iter_->Next();
+  ASSERT_FALSE(c_iter_->Valid());
+}
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -77,6 +77,7 @@ struct CompactionJob::SubcompactionState {
  struct Output {
    FileMetaData meta;
    bool finished;
+    std::shared_ptr<const TableProperties> table_properties;
  };

  // State kept for output being generated
@@ -316,6 +317,8 @@ void CompactionJob::Prepare() {
      Slice* end = i == boundaries_.size() ? nullptr : &boundaries_[i];
      compact_->sub_compact_states.emplace_back(c, start, end, sizes_[i]);
    }
+    MeasureTime(stats_, NUM_SUBCOMPACTIONS_SCHEDULED,
+                compact_->sub_compact_states.size());
  } else {
    compact_->sub_compact_states.emplace_back(c, nullptr, nullptr);
  }
@@ -485,6 +488,16 @@ Status CompactionJob::Run() {
    }
  }

+  TablePropertiesCollection tp;
+  for (const auto& state : compact_->sub_compact_states) {
+    for (const auto& output : state.outputs) {
+      auto fn = TableFileName(db_options_.db_paths, output.meta.fd.GetNumber(),
+                              output.meta.fd.GetPathId());
+      tp[fn] = output.table_properties;
+    }
+  }
+  compact_->compaction->SetOutputTableProperties(std::move(tp));
+
  // Finish up all book-keeping to unify the subcompaction results
  AggregateStatistics();
  UpdateCompactionStats();
@@ -540,8 +553,9 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options,
  UpdateCompactionJobStats(stats);

  auto stream = event_logger_->LogToBuffer(log_buffer_);
-  stream << "job" << job_id_ << "event"
-         << "compaction_finished"
+  stream << "job" << job_id_
+         << "event" << "compaction_finished"
+         << "compaction_time_micros" << compaction_stats_.micros
         << "output_level" << compact_->compaction->output_level()
         << "num_output_files" << compact_->NumOutputFiles()
         << "total_output_size" << compact_->total_bytes
@@ -571,7 +585,7 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options,

 void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  assert(sub_compact != nullptr);
-  std::unique_ptr<Iterator> input(
+  std::unique_ptr<InternalIterator> input(
      versions_->MakeInputIterator(sub_compact->compaction));

  AutoThreadOperationStageUpdater stage_updater(
@@ -587,17 +601,13 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  if (measure_io_stats_) {
    prev_perf_level = GetPerfLevel();
    SetPerfLevel(PerfLevel::kEnableTime);
-    prev_write_nanos = iostats_context.write_nanos;
-    prev_fsync_nanos = iostats_context.fsync_nanos;
-    prev_range_sync_nanos = iostats_context.range_sync_nanos;
-    prev_prepare_write_nanos = iostats_context.prepare_write_nanos;
+    prev_write_nanos = IOSTATS(write_nanos);
+    prev_fsync_nanos = IOSTATS(fsync_nanos);
+    prev_range_sync_nanos = IOSTATS(range_sync_nanos);
+    prev_prepare_write_nanos = IOSTATS(prepare_write_nanos);
  }

  ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
-  MergeHelper merge(cfd->user_comparator(), cfd->ioptions()->merge_operator,
-                    db_options_.info_log.get(),
-                    cfd->ioptions()->min_partial_merge_operands,
-                    false /* internal key corruption is expected */);
  auto compaction_filter = cfd->ioptions()->compaction_filter;
  std::unique_ptr<CompactionFilter> compaction_filter_from_factory = nullptr;
  if (compaction_filter == nullptr) {
@@ -605,6 +615,13 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
        sub_compact->compaction->CreateCompactionFilter();
    compaction_filter = compaction_filter_from_factory.get();
  }
+  MergeHelper merge(
+      env_, cfd->user_comparator(), cfd->ioptions()->merge_operator,
+      compaction_filter, db_options_.info_log.get(),
+      cfd->ioptions()->min_partial_merge_operands,
+      false /* internal key corruption is expected */,
+      existing_snapshots_.empty() ? 0 : existing_snapshots_.back(),
+      compact_->compaction->level(), db_options_.statistics.get());

  TEST_SYNC_POINT("CompactionJob::Run():Inprogress");

@@ -621,8 +638,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  Status status;
  sub_compact->c_iter.reset(new CompactionIterator(
      input.get(), cfd->user_comparator(), &merge, versions_->LastSequence(),
-      &existing_snapshots_, env_, false, db_options_.statistics.get(),
-      sub_compact->compaction, compaction_filter));
+      &existing_snapshots_, env_, false, sub_compact->compaction,
+      compaction_filter));
  auto c_iter = sub_compact->c_iter.get();
  c_iter->SeekToFirst();
  const auto& c_iter_stats = c_iter->iter_stats();
@@ -711,13 +728,13 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {

  if (measure_io_stats_) {
    sub_compact->compaction_job_stats.file_write_nanos +=
-        iostats_context.write_nanos - prev_write_nanos;
+        IOSTATS(write_nanos) - prev_write_nanos;
    sub_compact->compaction_job_stats.file_fsync_nanos +=
-        iostats_context.fsync_nanos - prev_fsync_nanos;
+        IOSTATS(fsync_nanos) - prev_fsync_nanos;
    sub_compact->compaction_job_stats.file_range_sync_nanos +=
-        iostats_context.range_sync_nanos - prev_range_sync_nanos;
+        IOSTATS(range_sync_nanos) - prev_range_sync_nanos;
    sub_compact->compaction_job_stats.file_prepare_write_nanos +=
-        iostats_context.prepare_write_nanos - prev_prepare_write_nanos;
+        IOSTATS(prepare_write_nanos) - prev_prepare_write_nanos;
    if (prev_perf_level != PerfLevel::kEnableTime) {
      SetPerfLevel(prev_perf_level);
    }
@@ -794,7 +811,7 @@ Status CompactionJob::FinishCompactionOutputFile(
  if (s.ok() && current_entries > 0) {
    // Verify that the table is usable
    ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
-    Iterator* iter = cfd->table_cache()->NewIterator(
+    InternalIterator* iter = cfd->table_cache()->NewIterator(
        ReadOptions(), env_options_, cfd->internal_comparator(), meta->fd,
        nullptr, cfd->internal_stats()->GetFileReadHist(
                     compact_->compaction->output_level()),
@@ -808,7 +825,10 @@ Status CompactionJob::FinishCompactionOutputFile(

    delete iter;
    if (s.ok()) {
-      TableFileCreationInfo info(sub_compact->builder->GetTableProperties());
+      auto tp = sub_compact->builder->GetTableProperties();
+      sub_compact->current_output()->table_properties =
+          std::make_shared<TableProperties>(tp);
+      TableFileCreationInfo info(std::move(tp));
      info.db_name = dbname_;
      info.cf_name = cfd->GetName();
      info.file_path =
@@ -891,7 +911,7 @@ Status CompactionJob::OpenCompactionOutputFile(
  unique_ptr<WritableFile> writable_file;
  std::string fname = TableFileName(db_options_.db_paths, file_number,
                                    sub_compact->compaction->output_path_id());
-  Status s = env_->NewWritableFile(fname, &writable_file, env_options_);
+  Status s = NewWritableFile(env_, fname, &writable_file, env_options_);
  if (!s.ok()) {
    Log(InfoLogLevel::ERROR_LEVEL, db_options_.info_log,
        "[%s] [JOB %d] OpenCompactionOutputFiles for table #%" PRIu64
@@ -921,8 +941,8 @@ Status CompactionJob::OpenCompactionOutputFile(
      cfd->ioptions()->optimize_filters_for_hits && bottommost_level_;
  sub_compact->builder.reset(NewTableBuilder(
      *cfd->ioptions(), cfd->internal_comparator(),
-      cfd->int_tbl_prop_collector_factories(), sub_compact->outfile.get(),
-      sub_compact->compaction->output_compression(),
+      cfd->int_tbl_prop_collector_factories(), cfd->GetID(),
+      sub_compact->outfile.get(), sub_compact->compaction->output_compression(),
      cfd->ioptions()->compression_opts, skip_filters));
  LogFlush(db_options_.info_log);
  return s;
@@ -35,9 +35,9 @@
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/transaction_log.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/autovector.h"
 #include "util/event_logger.h"
-#include "util/scoped_arena_iterator.h"
 #include "util/stop_watch.h"
 #include "util/thread_local.h"

@@ -27,6 +27,7 @@
 #include "db/job_context.h"
 #include "db/version_set.h"
 #include "db/write_batch_internal.h"
+#include "memtable/hash_linklist_rep.h"
 #include "port/stack_trace.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/compaction_filter.h"
@@ -47,14 +48,13 @@
 #include "table/block_based_table_factory.h"
 #include "table/mock_table.h"
 #include "table/plain_table_factory.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/compression.h"
 #include "util/hash.h"
-#include "util/hash_linklist_rep.h"
 #include "util/logging.h"
 #include "util/mock_env.h"
 #include "util/mutexlock.h"
 #include "util/rate_limiter.h"
-#include "util/scoped_arena_iterator.h"
 #include "util/statistics.h"
 #include "util/string_util.h"
 #include "util/sync_point.h"
@@ -64,7 +64,7 @@
 #include "util/xfunc.h"
 #include "utilities/merge_operators.h"

-#if !defined(IOS_CROSS_COMPILE) && (!defined(NDEBUG) || !defined(OS_WIN))
+#if !defined(IOS_CROSS_COMPILE)
 #ifndef ROCKSDB_LITE
 namespace rocksdb {

@@ -1014,8 +1014,8 @@ TEST_P(CompactionJobStatsTest, UniversalCompactionTest) {
        &rnd, start_key, start_key + key_base - 1,
        kKeySize, kValueSize, key_interval,
        compression_ratio, 1);
+    reinterpret_cast<DBImpl*>(db_)->TEST_WaitForCompact();
  }
-  reinterpret_cast<DBImpl*>(db_)->TEST_WaitForCompact();
  ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 0U);
 }

@@ -1029,6 +1029,14 @@ int main(int argc, char** argv) {
  return RUN_ALL_TESTS();
 }

+#else
+#include <stdio.h>
+
+int main(int argc, char** argv) {
+  fprintf(stderr, "SKIPPED, not supported in ROCKSDB_LITE\n");
+  return 0;
+}
+
 #endif  // !ROCKSDB_LITE

 #else
@@ -3,6 +3,8 @@
 //  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 ROCKSDB_LITE
+
 #include <algorithm>
 #include <map>
 #include <string>
@@ -182,7 +184,7 @@ class CompactionJobTest : public testing::Test {
    return expected_results;
  }

-  void NewDB(std::shared_ptr<MergeOperator> merge_operator = nullptr) {
+  void NewDB() {
    VersionEdit new_db;
    new_db.SetLogNumber(0);
    new_db.SetNextFile(2);
@@ -196,7 +198,7 @@ class CompactionJobTest : public testing::Test {
    unique_ptr<WritableFileWriter> file_writer(
        new WritableFileWriter(std::move(file), env_options_));
    {
-      log::Writer log(std::move(file_writer));
+      log::Writer log(std::move(file_writer), 0, false);
      std::string record;
      new_db.EncodeTo(&record);
      s = log.AddRecord(record);
@@ -207,7 +209,8 @@ class CompactionJobTest : public testing::Test {

    std::vector<ColumnFamilyDescriptor> column_families;
    cf_options_.table_factory = mock_table_factory_;
-    cf_options_.merge_operator = merge_operator;
+    cf_options_.merge_operator = merge_op_;
+    cf_options_.compaction_filter = compaction_filter_.get();
    column_families.emplace_back(kDefaultColumnFamilyName, cf_options_);

    EXPECT_OK(versions_->Recover(column_families, false));
@@ -215,7 +218,8 @@ class CompactionJobTest : public testing::Test {
  }

  void RunCompaction(const std::vector<std::vector<FileMetaData*>>& input_files,
-                     const stl_wrappers::KVMap& expected_results) {
+                     const stl_wrappers::KVMap& expected_results,
+                     const std::vector<SequenceNumber>& snapshots = {}) {
    auto cfd = versions_->GetColumnFamilySet()->GetDefault();

    size_t num_input_files = 0;
@@ -241,9 +245,9 @@ class CompactionJobTest : public testing::Test {
    EventLogger event_logger(db_options_.info_log.get());
    CompactionJob compaction_job(0, &compaction, db_options_, env_options_,
                                 versions_.get(), &shutting_down_, &log_buffer,
-                                 nullptr, nullptr, nullptr, {}, table_cache_,
-                                 &event_logger, false, false, dbname_,
-                                 &compaction_job_stats_);
+                                 nullptr, nullptr, nullptr, snapshots,
+                                 table_cache_, &event_logger, false, false,
+                                 dbname_, &compaction_job_stats_);

    VerifyInitializationOfCompactionJobStats(compaction_job_stats_);

@@ -257,10 +261,16 @@ class CompactionJobTest : public testing::Test {
                                     &mutex_));
    mutex_.Unlock();

-    ASSERT_GE(compaction_job_stats_.elapsed_micros, 0U);
-    ASSERT_EQ(compaction_job_stats_.num_input_files, num_input_files);
-    ASSERT_EQ(compaction_job_stats_.num_output_files, 1U);
-    mock_table_factory_->AssertLatestFile(expected_results);
+    if (expected_results.size() == 0) {
+      ASSERT_GE(compaction_job_stats_.elapsed_micros, 0U);
+      ASSERT_EQ(compaction_job_stats_.num_input_files, num_input_files);
+      ASSERT_EQ(compaction_job_stats_.num_output_files, 0U);
+    } else {
+      ASSERT_GE(compaction_job_stats_.elapsed_micros, 0U);
+      ASSERT_EQ(compaction_job_stats_.num_input_files, num_input_files);
+      ASSERT_EQ(compaction_job_stats_.num_output_files, 1U);
+      mock_table_factory_->AssertLatestFile(expected_results);
+    }
  }

  Env* env_;
@@ -278,6 +288,8 @@ class CompactionJobTest : public testing::Test {
  std::shared_ptr<mock::MockTableFactory> mock_table_factory_;
  CompactionJobStats compaction_job_stats_;
  ColumnFamilyData* cfd_;
+  std::unique_ptr<CompactionFilter> compaction_filter_;
+  std::shared_ptr<MergeOperator> merge_op_;
 };

 TEST_F(CompactionJobTest, Simple) {
@@ -296,7 +308,7 @@ TEST_F(CompactionJobTest, SimpleCorrupted) {
  auto expected_results = CreateTwoFiles(true);
  auto cfd = versions_->GetColumnFamilySet()->GetDefault();
  auto files = cfd->current()->storage_info()->LevelFiles(0);
-  RunCompaction({ files }, expected_results);
+  RunCompaction({files}, expected_results);
  ASSERT_EQ(compaction_job_stats_.num_corrupt_keys, 400U);
 }

@@ -316,7 +328,7 @@ TEST_F(CompactionJobTest, SimpleDeletion) {

  SetLastSequence(4U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
-  RunCompaction({ files }, expected_results);
+  RunCompaction({files}, expected_results);
 }

 TEST_F(CompactionJobTest, SimpleOverwrite) {
@@ -338,7 +350,7 @@ TEST_F(CompactionJobTest, SimpleOverwrite) {

  SetLastSequence(4U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
-  RunCompaction({ files }, expected_results);
+  RunCompaction({files}, expected_results);
 }

 TEST_F(CompactionJobTest, SimpleNonLastLevel) {
@@ -367,12 +379,12 @@ TEST_F(CompactionJobTest, SimpleNonLastLevel) {
  SetLastSequence(6U);
  auto lvl0_files = cfd_->current()->storage_info()->LevelFiles(0);
  auto lvl1_files = cfd_->current()->storage_info()->LevelFiles(1);
-  RunCompaction({ lvl0_files, lvl1_files }, expected_results);
+  RunCompaction({lvl0_files, lvl1_files}, expected_results);
 }

 TEST_F(CompactionJobTest, SimpleMerge) {
-  auto merge_op = MergeOperators::CreateStringAppendOperator();
-  NewDB(merge_op);
+  merge_op_ = MergeOperators::CreateStringAppendOperator();
+  NewDB();

  auto file1 = mock::MakeMockFile({
      {KeyStr("a", 5U, kTypeMerge), "5"},
@@ -391,12 +403,12 @@ TEST_F(CompactionJobTest, SimpleMerge) {

  SetLastSequence(5U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
-  RunCompaction({ files }, expected_results);
+  RunCompaction({files}, expected_results);
 }

 TEST_F(CompactionJobTest, NonAssocMerge) {
-  auto merge_op = MergeOperators::CreateStringAppendTESTOperator();
-  NewDB(merge_op);
+  merge_op_ = MergeOperators::CreateStringAppendTESTOperator();
+  NewDB();

  auto file1 = mock::MakeMockFile({
      {KeyStr("a", 5U, kTypeMerge), "5"},
@@ -416,7 +428,267 @@ TEST_F(CompactionJobTest, NonAssocMerge) {

  SetLastSequence(5U);
  auto files = cfd_->current()->storage_info()->LevelFiles(0);
-  RunCompaction({ files }, expected_results);
+  RunCompaction({files}, expected_results);
+}
+
+// Filters merge operands with value 10.
+TEST_F(CompactionJobTest, MergeOperandFilter) {
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();
+  compaction_filter_.reset(new test::FilterNumber(10U));
+  NewDB();
+
+  auto file1 = mock::MakeMockFile(
+      {{KeyStr("a", 5U, kTypeMerge), test::EncodeInt(5U)},
+       {KeyStr("a", 4U, kTypeMerge), test::EncodeInt(10U)},  // Filtered
+       {KeyStr("a", 3U, kTypeMerge), test::EncodeInt(3U)}});
+  AddMockFile(file1);
+
+  auto file2 = mock::MakeMockFile({
+      {KeyStr("b", 2U, kTypeMerge), test::EncodeInt(2U)},
+      {KeyStr("b", 1U, kTypeMerge), test::EncodeInt(10U)}  // Filtered
+  });
+  AddMockFile(file2);
+
+  auto expected_results =
+      mock::MakeMockFile({{KeyStr("a", 0U, kTypeValue), test::EncodeInt(8U)},
+                          {KeyStr("b", 2U, kTypeMerge), test::EncodeInt(2U)}});
+
+  SetLastSequence(5U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results);
+}
+
+TEST_F(CompactionJobTest, FilterSomeMergeOperands) {
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();
+  compaction_filter_.reset(new test::FilterNumber(10U));
+  NewDB();
+
+  auto file1 = mock::MakeMockFile(
+      {{KeyStr("a", 5U, kTypeMerge), test::EncodeInt(5U)},
+       {KeyStr("a", 4U, kTypeMerge), test::EncodeInt(10U)},  // Filtered
+       {KeyStr("a", 3U, kTypeValue), test::EncodeInt(5U)},
+       {KeyStr("d", 8U, kTypeMerge), test::EncodeInt(10U)}});
+  AddMockFile(file1);
+
+  auto file2 =
+      mock::MakeMockFile({{KeyStr("b", 2U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("b", 1U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("c", 2U, kTypeMerge), test::EncodeInt(3U)},
+                          {KeyStr("c", 1U, kTypeValue), test::EncodeInt(7U)},
+                          {KeyStr("d", 1U, kTypeValue), test::EncodeInt(6U)}});
+  AddMockFile(file2);
+
+  auto file3 =
+      mock::MakeMockFile({{KeyStr("a", 1U, kTypeMerge), test::EncodeInt(3U)}});
+  AddMockFile(file3, 2);
+
+  auto expected_results = mock::MakeMockFile({
+      {KeyStr("a", 5U, kTypeValue), test::EncodeInt(10U)},
+      {KeyStr("c", 2U, kTypeValue), test::EncodeInt(10U)},
+      {KeyStr("d", 1U, kTypeValue), test::EncodeInt(6U)}
+      // b does not appear because the operands are filtered
+  });
+
+  SetLastSequence(5U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results);
+}
+
+// Test where all operands/merge results are filtered out.
+TEST_F(CompactionJobTest, FilterAllMergeOperands) {
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();
+  compaction_filter_.reset(new test::FilterNumber(10U));
+  NewDB();
+
+  auto file1 =
+      mock::MakeMockFile({{KeyStr("a", 11U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("a", 10U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("a", 9U, kTypeMerge), test::EncodeInt(10U)}});
+  AddMockFile(file1);
+
+  auto file2 =
+      mock::MakeMockFile({{KeyStr("b", 8U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("b", 7U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("b", 6U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("b", 5U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("b", 4U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("b", 3U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("b", 2U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("c", 2U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("c", 1U, kTypeMerge), test::EncodeInt(10U)}});
+  AddMockFile(file2);
+
+  auto file3 =
+      mock::MakeMockFile({{KeyStr("a", 2U, kTypeMerge), test::EncodeInt(10U)},
+                          {KeyStr("b", 1U, kTypeMerge), test::EncodeInt(10U)}});
+  AddMockFile(file3, 2);
+
+  SetLastSequence(11U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+
+  stl_wrappers::KVMap empty_map;
+  RunCompaction({files}, empty_map);
+}
+
+TEST_F(CompactionJobTest, SimpleSingleDelete) {
+  NewDB();
+
+  auto file1 = mock::MakeMockFile({
+      {KeyStr("a", 5U, kTypeDeletion), ""},
+      {KeyStr("b", 6U, kTypeSingleDeletion), ""},
+  });
+  AddMockFile(file1);
+
+  auto file2 = mock::MakeMockFile({{KeyStr("a", 3U, kTypeValue), "val"},
+                                   {KeyStr("b", 4U, kTypeValue), "val"}});
+  AddMockFile(file2);
+
+  auto file3 = mock::MakeMockFile({
+      {KeyStr("a", 1U, kTypeValue), "val"},
+  });
+  AddMockFile(file3, 2);
+
+  auto expected_results =
+      mock::MakeMockFile({{KeyStr("a", 5U, kTypeDeletion), ""}});
+
+  SetLastSequence(6U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results);
+}
+
+TEST_F(CompactionJobTest, SingleDeleteSnapshots) {
+  NewDB();
+
+  auto file1 = mock::MakeMockFile({{KeyStr("A", 12U, kTypeSingleDeletion), ""},
+                                   {KeyStr("a", 12U, kTypeSingleDeletion), ""},
+                                   {KeyStr("b", 21U, kTypeSingleDeletion), ""},
+                                   {KeyStr("c", 22U, kTypeSingleDeletion), ""},
+                                   {KeyStr("d", 9U, kTypeSingleDeletion), ""}});
+  AddMockFile(file1);
+
+  auto file2 = mock::MakeMockFile({{KeyStr("0", 2U, kTypeSingleDeletion), ""},
+                                   {KeyStr("a", 11U, kTypeValue), "val1"},
+                                   {KeyStr("b", 11U, kTypeValue), "val2"},
+                                   {KeyStr("c", 21U, kTypeValue), "val3"},
+                                   {KeyStr("d", 8U, kTypeValue), "val4"},
+                                   {KeyStr("e", 2U, kTypeSingleDeletion), ""}});
+  AddMockFile(file2);
+
+  auto file3 = mock::MakeMockFile({{KeyStr("A", 1U, kTypeValue), "val"},
+                                   {KeyStr("e", 1U, kTypeValue), "val"}});
+  AddMockFile(file3, 2);
+
+  auto expected_results =
+      mock::MakeMockFile({{KeyStr("A", 12U, kTypeSingleDeletion), ""},
+                          {KeyStr("b", 21U, kTypeSingleDeletion), ""},
+                          {KeyStr("b", 11U, kTypeValue), "val2"},
+                          {KeyStr("e", 2U, kTypeSingleDeletion), ""}});
+
+  SetLastSequence(22U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results, {10U, 20U});
+}
+
+TEST_F(CompactionJobTest, SingleDeleteZeroSeq) {
+  NewDB();
+
+  auto file1 = mock::MakeMockFile({
+      {KeyStr("A", 10U, kTypeSingleDeletion), ""},
+      {KeyStr("dummy", 5U, kTypeValue), "val2"},
+  });
+  AddMockFile(file1);
+
+  auto file2 = mock::MakeMockFile({
+      {KeyStr("A", 0U, kTypeValue), "val"},
+  });
+  AddMockFile(file2);
+
+  auto expected_results = mock::MakeMockFile({
+      {KeyStr("dummy", 0U, kTypeValue), "val2"},
+  });
+
+  SetLastSequence(22U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results, {});
+}
+
+TEST_F(CompactionJobTest, MultiSingleDelete) {
+  // Tests three scenarios involving multiple single delete/put pairs:
+  //
+  // A: Put Snapshot SDel Put SDel -> Put Snapshot SDel
+  // B: Put SDel Put SDel -> (Removed)
+  // C: SDel Put SDel Snapshot Put -> Snapshot Put
+  // D: (Put) SDel Snapshot Put SDel -> (Put) SDel Snapshot
+  NewDB();
+
+  auto file1 = mock::MakeMockFile({
+      {KeyStr("A", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("A", 13U, kTypeValue), "val5"},
+      {KeyStr("A", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("B", 14U, kTypeSingleDeletion), ""},
+      {KeyStr("B", 13U, kTypeValue), "val2"},
+      {KeyStr("C", 14U, kTypeValue), "val3"},
+      {KeyStr("D", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("D", 11U, kTypeValue), "val4"},
+  });
+  AddMockFile(file1);
+
+  auto file2 = mock::MakeMockFile({
+      {KeyStr("A", 10U, kTypeValue), "val"},
+      {KeyStr("B", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("B", 11U, kTypeValue), "val2"},
+      {KeyStr("C", 10U, kTypeSingleDeletion), ""},
+      {KeyStr("C", 9U, kTypeValue), "val6"},
+      {KeyStr("C", 8U, kTypeSingleDeletion), ""},
+      {KeyStr("D", 10U, kTypeSingleDeletion), ""},
+  });
+  AddMockFile(file2);
+
+  auto file3 = mock::MakeMockFile({
+      {KeyStr("D", 11U, kTypeValue), "val"},
+  });
+  AddMockFile(file3, 2);
+
+  auto expected_results = mock::MakeMockFile({
+      {KeyStr("A", 12U, kTypeSingleDeletion), ""},
+      {KeyStr("A", 10U, kTypeValue), "val"},
+      {KeyStr("C", 14U, kTypeValue), "val3"},
+      {KeyStr("D", 10U, kTypeSingleDeletion), ""},
+  });
+
+  SetLastSequence(22U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results, {10U});
+}
+
+// This test documents the behavior where a corrupt key follows a deletion or a
+// single deletion and the (single) deletion gets removed while the corrupt key
+// gets written out. TODO(noetzli): We probably want a better way to treat
+// corrupt keys.
+TEST_F(CompactionJobTest, CorruptionAfterDeletion) {
+  NewDB();
+
+  auto file1 =
+      mock::MakeMockFile({{test::KeyStr("A", 6U, kTypeValue), "val3"},
+                          {test::KeyStr("a", 5U, kTypeDeletion), ""},
+                          {test::KeyStr("a", 4U, kTypeValue, true), "val"}});
+  AddMockFile(file1);
+
+  auto file2 =
+      mock::MakeMockFile({{test::KeyStr("b", 3U, kTypeSingleDeletion), ""},
+                          {test::KeyStr("b", 2U, kTypeValue, true), "val"},
+                          {test::KeyStr("c", 1U, kTypeValue), "val2"}});
+  AddMockFile(file2);
+
+  auto expected_results =
+      mock::MakeMockFile({{test::KeyStr("A", 0U, kTypeValue), "val3"},
+                          {test::KeyStr("a", 0U, kTypeValue, true), "val"},
+                          {test::KeyStr("b", 0U, kTypeValue, true), "val"},
+                          {test::KeyStr("c", 0U, kTypeValue), "val2"}});
+
+  SetLastSequence(6U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results);
 }

 }  // namespace rocksdb
@@ -425,3 +697,14 @@ int main(int argc, char** argv) {
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
 }
+
+#else
+#include <stdio.h>
+
+int main(int argc, char** argv) {
+  fprintf(stderr,
+          "SKIPPED as CompactionJobStats is not supported in ROCKSDB_LITE\n");
+  return 0;
+}
+
+#endif  // ROCKSDB_LITE
@@ -1028,7 +1028,7 @@ bool LevelCompactionPicker::PickCompactionBySize(VersionStorageInfo* vstorage,

  // Pick the largest file in this level that is not already
  // being compacted
-  const std::vector<int>& file_size = vstorage->FilesBySize(level);
+  const std::vector<int>& file_size = vstorage->FilesByCompactionPri(level);
  const std::vector<FileMetaData*>& level_files = vstorage->LevelFiles(level);

  // record the first file that is not yet compacted
@@ -1039,11 +1039,6 @@ bool LevelCompactionPicker::PickCompactionBySize(VersionStorageInfo* vstorage,
    int index = file_size[i];
    auto* f = level_files[index];

-    assert((i == file_size.size() - 1) ||
-           (i >= VersionStorageInfo::kNumberFilesToSort - 1) ||
-           (f->compensated_file_size >=
-            level_files[file_size[i + 1]]->compensated_file_size));
-
    // do not pick a file to compact if it is being compacted
    // from n-1 level.
    if (f->being_compacted) {
@@ -1322,7 +1317,13 @@ Compaction* UniversalCompactionPicker::PickCompaction(
                              &largest_seqno);
      if (is_first) {
        is_first = false;
-      } else {
+      } else if (prev_smallest_seqno > 0) {
+        // A level is considered as the bottommost level if there are
+        // no files in higher levels or if files in higher levels do
+        // not overlap with the files being compacted. Sequence numbers
+        // of files in bottommost level can be set to 0 to help
+        // compression. As a result, the following assert may not hold
+        // if the prev_smallest_seqno is 0.
        assert(prev_smallest_seqno > largest_seqno);
      }
      prev_smallest_seqno = smallest_seqno;
@@ -110,6 +110,11 @@ class CompactionPicker {
  // overlapping.
  bool IsInputNonOverlapping(Compaction* c);

+  // Is there currently a compaction involving level 0 taking place
+  bool IsLevel0CompactionInProgress() const {
+    return !level0_compactions_in_progress_.empty();
+  }
+
 protected:
  int NumberLevels() const { return ioptions_.num_levels; }

@@ -3,9 +3,12 @@
 //  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.

+#include "db/compaction.h"
 #include "db/compaction_picker.h"
 #include <limits>
 #include <string>
+#include <utility>
+
 #include "util/logging.h"
 #include "util/string_util.h"
 #include "util/testharness.h"
@@ -35,6 +38,11 @@ class CompactionPickerTest : public testing::Test {
  CompactionOptionsFIFO fifo_options_;
  std::unique_ptr<VersionStorageInfo> vstorage_;
  std::vector<std::unique_ptr<FileMetaData>> files_;
+  // does not own FileMetaData
+  std::unordered_map<uint32_t, std::pair<FileMetaData*, int>> file_map_;
+  // input files to compaction process.
+  std::vector<CompactionInputFiles> input_files_;
+  int compaction_level_start_;

  CompactionPickerTest()
      : ucmp_(BytewiseComparator()),
@@ -66,6 +74,8 @@ class CompactionPickerTest : public testing::Test {
  void DeleteVersionStorage() {
    vstorage_.reset();
    files_.clear();
+    file_map_.clear();
+    input_files_.clear();
  }

  void Add(int level, uint32_t file_number, const char* smallest,
@@ -83,11 +93,29 @@ class CompactionPickerTest : public testing::Test {
    f->refs = 0;
    vstorage_->AddFile(level, f);
    files_.emplace_back(f);
+    file_map_.insert({file_number, {f, level}});
+  }
+
+  void SetCompactionInputFilesLevels(int level_count, int start_level) {
+    input_files_.resize(level_count);
+    for (int i = 0; i < level_count; ++i) {
+      input_files_[i].level = start_level + i;
+    }
+    compaction_level_start_ = start_level;
+  }
+
+  void AddToCompactionFiles(uint32_t file_number) {
+    auto iter = file_map_.find(file_number);
+    assert(iter != file_map_.end());
+    int level = iter->second.second;
+    assert(level < vstorage_->num_levels());
+    input_files_[level - compaction_level_start_].files.emplace_back(
+        iter->second.first);
  }

  void UpdateVersionStorageInfo() {
    vstorage_->CalculateBaseBytes(ioptions_, mutable_cf_options_);
-    vstorage_->UpdateFilesBySize();
+    vstorage_->UpdateFilesByCompactionPri(mutable_cf_options_);
    vstorage_->UpdateNumNonEmptyLevels();
    vstorage_->GenerateFileIndexer();
    vstorage_->GenerateLevelFilesBrief();
@@ -637,6 +665,164 @@ TEST_F(CompactionPickerTest, EstimateCompactionBytesNeededDynamicLevel) {
            vstorage_->estimated_compaction_needed_bytes());
 }

+TEST_F(CompactionPickerTest, IsBottommostLevelTest) {
+  // case 1: Higher levels are empty
+  NewVersionStorage(6, kCompactionStyleLevel);
+  Add(0, 1U, "a", "m");
+  Add(0, 2U, "c", "z");
+  Add(1, 3U, "d", "e");
+  Add(1, 4U, "l", "p");
+  Add(2, 5U, "g", "i");
+  Add(2, 6U, "x", "z");
+  UpdateVersionStorageInfo();
+  SetCompactionInputFilesLevels(2, 1);
+  AddToCompactionFiles(3U);
+  AddToCompactionFiles(5U);
+  bool result =
+      Compaction::TEST_IsBottommostLevel(2, vstorage_.get(), input_files_);
+  ASSERT_TRUE(result);
+
+  // case 2: Higher levels have no overlap
+  NewVersionStorage(6, kCompactionStyleLevel);
+  Add(0, 1U, "a", "m");
+  Add(0, 2U, "c", "z");
+  Add(1, 3U, "d", "e");
+  Add(1, 4U, "l", "p");
+  Add(2, 5U, "g", "i");
+  Add(2, 6U, "x", "z");
+  Add(3, 7U, "k", "p");
+  Add(3, 8U, "t", "w");
+  Add(4, 9U, "a", "b");
+  Add(5, 10U, "c", "cc");
+  UpdateVersionStorageInfo();
+  SetCompactionInputFilesLevels(2, 1);
+  AddToCompactionFiles(3U);
+  AddToCompactionFiles(5U);
+  result = Compaction::TEST_IsBottommostLevel(2, vstorage_.get(), input_files_);
+  ASSERT_TRUE(result);
+
+  // case 3.1: Higher levels (level 3) have overlap
+  NewVersionStorage(6, kCompactionStyleLevel);
+  Add(0, 1U, "a", "m");
+  Add(0, 2U, "c", "z");
+  Add(1, 3U, "d", "e");
+  Add(1, 4U, "l", "p");
+  Add(2, 5U, "g", "i");
+  Add(2, 6U, "x", "z");
+  Add(3, 7U, "e", "g");
+  Add(3, 8U, "h", "k");
+  Add(4, 9U, "a", "b");
+  Add(5, 10U, "c", "cc");
+  UpdateVersionStorageInfo();
+  SetCompactionInputFilesLevels(2, 1);
+  AddToCompactionFiles(3U);
+  AddToCompactionFiles(5U);
+  result = Compaction::TEST_IsBottommostLevel(2, vstorage_.get(), input_files_);
+  ASSERT_FALSE(result);
+
+  // case 3.2: Higher levels (level 5) have overlap
+  DeleteVersionStorage();
+  NewVersionStorage(6, kCompactionStyleLevel);
+  Add(0, 1U, "a", "m");
+  Add(0, 2U, "c", "z");
+  Add(1, 3U, "d", "e");
+  Add(1, 4U, "l", "p");
+  Add(2, 5U, "g", "i");
+  Add(2, 6U, "x", "z");
+  Add(3, 7U, "j", "k");
+  Add(3, 8U, "l", "m");
+  Add(4, 9U, "a", "b");
+  Add(5, 10U, "c", "cc");
+  Add(5, 11U, "h", "k");
+  Add(5, 12U, "y", "yy");
+  Add(5, 13U, "z", "zz");
+  UpdateVersionStorageInfo();
+  SetCompactionInputFilesLevels(2, 1);
+  AddToCompactionFiles(3U);
+  AddToCompactionFiles(5U);
+  result = Compaction::TEST_IsBottommostLevel(2, vstorage_.get(), input_files_);
+  ASSERT_FALSE(result);
+
+  // case 3.3: Higher levels (level 5) have overlap, but it's only overlapping
+  // one key ("d")
+  NewVersionStorage(6, kCompactionStyleLevel);
+  Add(0, 1U, "a", "m");
+  Add(0, 2U, "c", "z");
+  Add(1, 3U, "d", "e");
+  Add(1, 4U, "l", "p");
+  Add(2, 5U, "g", "i");
+  Add(2, 6U, "x", "z");
+  Add(3, 7U, "j", "k");
+  Add(3, 8U, "l", "m");
+  Add(4, 9U, "a", "b");
+  Add(5, 10U, "c", "cc");
+  Add(5, 11U, "ccc", "d");
+  Add(5, 12U, "y", "yy");
+  Add(5, 13U, "z", "zz");
+  UpdateVersionStorageInfo();
+  SetCompactionInputFilesLevels(2, 1);
+  AddToCompactionFiles(3U);
+  AddToCompactionFiles(5U);
+  result = Compaction::TEST_IsBottommostLevel(2, vstorage_.get(), input_files_);
+  ASSERT_FALSE(result);
+
+  // Level 0 files overlap
+  NewVersionStorage(6, kCompactionStyleLevel);
+  Add(0, 1U, "s", "t");
+  Add(0, 2U, "a", "m");
+  Add(0, 3U, "b", "z");
+  Add(0, 4U, "e", "f");
+  Add(5, 10U, "y", "z");
+  UpdateVersionStorageInfo();
+  SetCompactionInputFilesLevels(1, 0);
+  AddToCompactionFiles(1U);
+  AddToCompactionFiles(2U);
+  AddToCompactionFiles(3U);
+  AddToCompactionFiles(4U);
+  result = Compaction::TEST_IsBottommostLevel(2, vstorage_.get(), input_files_);
+  ASSERT_FALSE(result);
+
+  // Level 0 files don't overlap
+  NewVersionStorage(6, kCompactionStyleLevel);
+  Add(0, 1U, "s", "t");
+  Add(0, 2U, "a", "m");
+  Add(0, 3U, "b", "k");
+  Add(0, 4U, "e", "f");
+  Add(5, 10U, "y", "z");
+  UpdateVersionStorageInfo();
+  SetCompactionInputFilesLevels(1, 0);
+  AddToCompactionFiles(1U);
+  AddToCompactionFiles(2U);
+  AddToCompactionFiles(3U);
+  AddToCompactionFiles(4U);
+  result = Compaction::TEST_IsBottommostLevel(2, vstorage_.get(), input_files_);
+  ASSERT_TRUE(result);
+
+  // Level 1 files overlap
+  NewVersionStorage(6, kCompactionStyleLevel);
+  Add(0, 1U, "s", "t");
+  Add(0, 2U, "a", "m");
+  Add(0, 3U, "b", "k");
+  Add(0, 4U, "e", "f");
+  Add(1, 5U, "a", "m");
+  Add(1, 6U, "n", "o");
+  Add(1, 7U, "w", "y");
+  Add(5, 10U, "y", "z");
+  UpdateVersionStorageInfo();
+  SetCompactionInputFilesLevels(2, 0);
+  AddToCompactionFiles(1U);
+  AddToCompactionFiles(2U);
+  AddToCompactionFiles(3U);
+  AddToCompactionFiles(4U);
+  AddToCompactionFiles(5U);
+  AddToCompactionFiles(6U);
+  AddToCompactionFiles(7U);
+  result = Compaction::TEST_IsBottommostLevel(2, vstorage_.get(), input_files_);
+  ASSERT_FALSE(result);
+
+  DeleteVersionStorage();
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -7,10 +7,11 @@
 #include <map>
 #include <string>

+#include "memtable/stl_wrappers.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "util/hash.h"
-#include "util/stl_wrappers.h"
+#include "util/kv_map.h"
 #include "util/string_util.h"
 #include "util/testharness.h"
 #include "util/testutil.h"
@@ -7,6 +7,8 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

+#ifndef ROCKSDB_LITE
+
 #include "rocksdb/db.h"

 #include <errno.h>
@@ -85,10 +87,14 @@ class CorruptionTest : public testing::Test {
    ASSERT_OK(::rocksdb::RepairDB(dbname_, options_));
  }

-  void Build(int n) {
+  void Build(int n, int flush_every = 0) {
    std::string key_space, value_space;
    WriteBatch batch;
    for (int i = 0; i < n; i++) {
+      if (flush_every != 0 && i != 0 && i % flush_every == 0) {
+        DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
+        dbi->TEST_FlushMemTable();
+      }
      //if ((i % 100) == 0) fprintf(stderr, "@ %d of %d\n", i, n);
      Slice key = Key(i, &key_space);
      batch.Clear();
@@ -295,13 +301,21 @@ TEST_F(CorruptionTest, TableFile) {
 }

 TEST_F(CorruptionTest, TableFileIndexData) {
-  Build(10000);  // Enough to build multiple Tables
+  Options options;
+  // very big, we'll trigger flushes manually
+  options.write_buffer_size = 100 * 1024 * 1024;
+  Reopen(&options);
+  // build 2 tables, flush at 5000
+  Build(10000, 5000);
  DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
  dbi->TEST_FlushMemTable();

+  // corrupt an index block of an entire file
  Corrupt(kTableFile, -2000, 500);
  Reopen();
-  Check(5000, 9999);
+  // one full file should be readable, since only one was corrupted
+  // the other file should be fully non-readable, since index was corrupted
+  Check(5000, 5000);
 }

 TEST_F(CorruptionTest, MissingDescriptor) {
@@ -467,3 +481,13 @@ int main(int argc, char** argv) {
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
 }
+
+#else
+#include <stdio.h>
+
+int main(int argc, char** argv) {
+  fprintf(stderr, "SKIPPED as RepairDB() is not supported in ROCKSDB_LITE\n");
+  return 0;
+}
+
+#endif  // !ROCKSDB_LITE
@@ -110,7 +110,8 @@ DEFINE_string(benchmarks,
              "uncompress,"
              "acquireload,"
              "fillseekseq,"
-              "randomtransaction",
+              "randomtransaction,"
+              "randomreplacekeys",

              "Comma-separated list of operations to run in the specified"
              " order. Available benchmarks:\n"
@@ -161,6 +162,8 @@ DEFINE_string(benchmarks,
              "them by seeking to each key\n"
              "\trandomtransaction     -- execute N random transactions and "
              "verify correctness\n"
+              "\trandomreplacekeys     -- randomly replaces N keys by deleting "
+              "the old version and putting the new version\n\n"
              "Meta operations:\n"
              "\tcompact     -- Compact the entire DB\n"
              "\tstats       -- Print DB stats\n"
@@ -311,6 +314,10 @@ static rocksdb::CompactionStyle FLAGS_compaction_style_e;
 DEFINE_int32(compaction_style, (int32_t) rocksdb::Options().compaction_style,
             "style of compaction: level-based vs universal");

+static rocksdb::CompactionPri FLAGS_compaction_pri_e;
+DEFINE_int32(compaction_pri, (int32_t)rocksdb::Options().compaction_pri,
+             "priority of files to compaction: by size or by data age");
+
 DEFINE_int32(universal_size_ratio, 0,
             "Percentage flexibility while comparing file size"
             " (for universal compaction only).");
@@ -357,11 +364,24 @@ DEFINE_int32(open_files, rocksdb::Options().max_open_files,
             "Maximum number of files to keep open at the same time"
             " (use default if == 0)");

+DEFINE_int32(file_opening_threads, rocksdb::Options().max_file_opening_threads,
+             "If open_files is set to -1, this option set the number of "
+             "threads that will be used to open files during DB::Open()");
+
 DEFINE_int32(new_table_reader_for_compaction_inputs, true,
             "If true, uses a separate file handle for compaction inputs");

 DEFINE_int32(compaction_readahead_size, 0, "Compaction readahead size");

+DEFINE_int32(random_access_max_buffer_size, 1024 * 1024,
+             "Maximum windows randomaccess buffer size");
+
+DEFINE_int32(writable_file_max_buffer_size, 1024 * 1024, 
+             "Maximum write buffer for Writeable File");
+
+DEFINE_int32(skip_table_builder_flush, false, "Skip flushing block in "
+             "table builder ");
+
 DEFINE_int32(bloom_bits, -1, "Bloom filter bits per key. Negative means"
             " use default settings.");
 DEFINE_int32(memtable_bloom_bits, 0, "Bloom filter bits per key for memtable. "
@@ -475,6 +495,7 @@ DEFINE_int32(deletepercent, 2, "Percentage of deletes out of reads/writes/"
 DEFINE_uint64(delete_obsolete_files_period_micros, 0,
              "Ignored. Left here for backward compatibility");

+#ifndef ROCKSDB_LITE
 DEFINE_bool(optimistic_transaction_db, false,
            "Open a OptimisticTransactionDB instance. "
            "Required for randomtransaction benchmark.");
@@ -498,6 +519,7 @@ DEFINE_int32(transaction_sleep, 0,
 DEFINE_uint64(transaction_lock_timeout, 100,
              "If using a transaction_db, specifies the lock wait timeout in"
              " milliseconds before failing a transaction waiting on a lock");
+#endif  // ROCKSDB_LITE

 DEFINE_bool(compaction_measure_io_stats, false,
            "Measure times spents on I/Os while in compactions. ");
@@ -611,9 +633,10 @@ static bool ValidateRateLimit(const char* flagname, double value) {
 }
 DEFINE_double(soft_rate_limit, 0.0, "");

-DEFINE_double(hard_rate_limit, 0.0, "When not equal to 0 this make threads "
-              "sleep at each stats reporting interval until the compaction"
-              " score for all levels is less than or equal to this value.");
+DEFINE_double(hard_rate_limit, 0.0, "DEPRECATED");
+
+DEFINE_uint64(hard_pending_compaction_bytes_limit, 128ull * 1024 * 1024 * 1024,
+              "Stop writes if pending compaction bytes exceed this number");

 DEFINE_uint64(delayed_write_rate, 2097152u,
              "Limited bytes allowed to DB when soft_rate_limit or "
@@ -633,7 +656,9 @@ DEFINE_int32(max_grandparent_overlap_factor, 10, "Control maximum bytes of "
             "overlaps in grandparent (i.e., level+2) before we stop building a"
             " single file in a level->level+1 compaction.");

+#ifndef ROCKSDB_LITE
 DEFINE_bool(readonly, false, "Run read only benchmarks.");
+#endif  // ROCKSDB_LITE

 DEFINE_bool(disable_auto_compactions, false, "Do not auto trigger compactions");

@@ -687,6 +712,13 @@ DEFINE_uint64(wal_bytes_per_sync,  rocksdb::Options().wal_bytes_per_sync,
 DEFINE_bool(filter_deletes, false, " On true, deletes use bloom-filter and drop"
            " the delete if key not present");

+DEFINE_bool(use_single_deletes, true,
+            "Use single deletes (used in RandomReplaceKeys only).");
+
+DEFINE_double(stddev, 2000.0,
+              "Standard deviation of normal distribution used for picking keys"
+              " (used in RandomReplaceKeys only).");
+
 DEFINE_int32(max_successive_merges, 0, "Maximum number of successive merge"
             " operations on a key in the memtable");

@@ -898,6 +930,7 @@ class ReportFileOpEnv : public EnvWrapper {
        return rv;
      }

+      Status Truncate(uint64_t size) override { return target_->Truncate(size); }
      Status Close() override { return target_->Close(); }
      Status Flush() override { return target_->Flush(); }
      Status Sync() override { return target_->Sync(); }
@@ -963,7 +996,9 @@ static void AppendWithSpace(std::string* str, Slice msg) {
 struct DBWithColumnFamilies {
  std::vector<ColumnFamilyHandle*> cfh;
  DB* db;
+#ifndef ROCKSDB_LITE
  OptimisticTransactionDB* opt_txn_db;
+#endif  // ROCKSDB_LITE
  std::atomic<size_t> num_created;  // Need to be updated after all the
                                    // new entries in cfh are set.
  size_t num_hot;  // Number of column families to be queried at each moment.
@@ -971,7 +1006,12 @@ struct DBWithColumnFamilies {
                   // Column families will be created and used to be queried.
  port::Mutex create_cf_mutex;  // Only one thread can execute CreateNewCf()

-  DBWithColumnFamilies() : db(nullptr), opt_txn_db(nullptr) {
+  DBWithColumnFamilies()
+      : db(nullptr)
+#ifndef ROCKSDB_LITE
+        , opt_txn_db(nullptr)
+#endif  // ROCKSDB_LITE
+  {
    cfh.clear();
    num_created = 0;
    num_hot = 0;
@@ -980,7 +1020,9 @@ struct DBWithColumnFamilies {
  DBWithColumnFamilies(const DBWithColumnFamilies& other)
      : cfh(other.cfh),
        db(other.db),
+#ifndef ROCKSDB_LITE
        opt_txn_db(other.opt_txn_db),
+#endif  // ROCKSDB_LITE
        num_created(other.num_created.load()),
        num_hot(other.num_hot) {}

@@ -988,13 +1030,18 @@ struct DBWithColumnFamilies {
    std::for_each(cfh.begin(), cfh.end(),
                  [](ColumnFamilyHandle* cfhi) { delete cfhi; });
    cfh.clear();
+#ifndef ROCKSDB_LITE
    if (opt_txn_db) {
      delete opt_txn_db;
      opt_txn_db = nullptr;
    } else {
      delete db;
+      db = nullptr;
    }
+#else
+    delete db;
    db = nullptr;
+#endif  // ROCKSDB_LITE
  }

  ColumnFamilyHandle* GetCfh(int64_t rand_num) {
@@ -1920,9 +1967,14 @@ class Benchmark {
        method = &Benchmark::Compress;
      } else if (name == "uncompress") {
        method = &Benchmark::Uncompress;
+#ifndef ROCKSDB_LITE
      } else if (name == "randomtransaction") {
        method = &Benchmark::RandomTransaction;
        post_process_method = &Benchmark::RandomTransactionVerify;
+#endif  // ROCKSDB_LITE
+      } else if (name == "randomreplacekeys") {
+        fresh_db = true;
+        method = &Benchmark::RandomReplaceKeys;
      } else if (name == "stats") {
        PrintStats("rocksdb.stats");
      } else if (name == "levelstats") {
@@ -2233,6 +2285,7 @@ class Benchmark {
    options.max_subcompactions = static_cast<uint32_t>(FLAGS_subcompactions);
    options.max_background_flushes = FLAGS_max_background_flushes;
    options.compaction_style = FLAGS_compaction_style_e;
+    options.compaction_pri = FLAGS_compaction_pri_e;
    if (FLAGS_prefix_size != 0) {
      options.prefix_extractor.reset(
          NewFixedPrefixTransform(FLAGS_prefix_size));
@@ -2247,9 +2300,12 @@ class Benchmark {
    options.memtable_prefix_bloom_bits = FLAGS_memtable_bloom_bits;
    options.bloom_locality = FLAGS_bloom_locality;
    options.max_open_files = FLAGS_open_files;
+    options.max_file_opening_threads = FLAGS_file_opening_threads;
    options.new_table_reader_for_compaction_inputs =
        FLAGS_new_table_reader_for_compaction_inputs;
    options.compaction_readahead_size = FLAGS_compaction_readahead_size;
+    options.random_access_max_buffer_size = FLAGS_random_access_max_buffer_size;
+    options.writable_file_max_buffer_size = FLAGS_writable_file_max_buffer_size;
    options.statistics = dbstats;
    if (FLAGS_enable_io_prio) {
      FLAGS_env->LowerThreadPoolIOPriority(Env::LOW);
@@ -2392,6 +2448,8 @@ class Benchmark {
      block_based_options.block_size = FLAGS_block_size;
      block_based_options.block_restart_interval = FLAGS_block_restart_interval;
      block_based_options.filter_policy = filter_policy_;
+      block_based_options.skip_table_builder_flush = 
+          FLAGS_skip_table_builder_flush;
      block_based_options.format_version = 2;
      options.table_factory.reset(
          NewBlockBasedTableFactory(block_based_options));
@@ -2430,6 +2488,8 @@ class Benchmark {
    }
    options.soft_rate_limit = FLAGS_soft_rate_limit;
    options.hard_rate_limit = FLAGS_hard_rate_limit;
+    options.hard_pending_compaction_bytes_limit =
+        FLAGS_hard_pending_compaction_bytes_limit;
    options.delayed_write_rate = FLAGS_delayed_write_rate;
    options.rate_limit_delay_max_milliseconds =
      FLAGS_rate_limit_delay_max_milliseconds;
@@ -2491,10 +2551,12 @@ class Benchmark {
          NewGenericRateLimiter(FLAGS_rate_limiter_bytes_per_sec));
    }

+#ifndef ROCKSDB_LITE
    if (FLAGS_readonly && FLAGS_transaction_db) {
      fprintf(stderr, "Cannot use readonly flag with transaction_db\n");
      exit(1);
    }
+#endif  // ROCKSDB_LITE

    if (FLAGS_num_multi_db <= 1) {
      OpenDb(options, FLAGS_db, &db_);
@@ -2527,6 +2589,7 @@ class Benchmark {
        column_families.push_back(ColumnFamilyDescriptor(
              ColumnFamilyName(i), ColumnFamilyOptions(options)));
      }
+#ifndef ROCKSDB_LITE
      if (FLAGS_readonly) {
        s = DB::OpenForReadOnly(options, db_name, column_families,
            &db->cfh, &db->db);
@@ -2547,9 +2610,13 @@ class Benchmark {
      } else {
        s = DB::Open(options, db_name, column_families, &db->cfh, &db->db);
      }
+#else
+      s = DB::Open(options, db_name, column_families, &db->cfh, &db->db);
+#endif  // ROCKSDB_LITE
      db->cfh.resize(FLAGS_num_column_families);
      db->num_created = num_hot;
      db->num_hot = num_hot;
+#ifndef ROCKSDB_LITE
    } else if (FLAGS_readonly) {
      s = DB::OpenForReadOnly(options, db_name, &db->db);
    } else if (FLAGS_optimistic_transaction_db) {
@@ -2564,7 +2631,7 @@ class Benchmark {
      if (s.ok()) {
        db->db = ptr;
      }
-
+#endif  // ROCKSDB_LITE
    } else {
      s = DB::Open(options, db_name, &db->db);
    }
@@ -3609,6 +3676,7 @@ class Benchmark {
    }
  }

+#ifndef ROCKSDB_LITE
  // This benchmark stress tests Transactions.  For a given --duration (or
  // total number of --writes, a Transaction will perform a read-modify-write
  // to increment the value of a key in each of N(--transaction-sets) sets of
@@ -3841,6 +3909,67 @@ class Benchmark {

    fprintf(stdout, "RandomTransactionVerify Success!\n");
  }
+#endif  // ROCKSDB_LITE
+
+  // Writes and deletes random keys without overwriting keys.
+  //
+  // This benchmark is intended to partially replicate the behavior of MyRocks
+  // secondary indices: All data is stored in keys and updates happen by
+  // deleting the old version of the key and inserting the new version.
+  void RandomReplaceKeys(ThreadState* thread) {
+    std::unique_ptr<const char[]> key_guard;
+    Slice key = AllocateKey(&key_guard);
+    std::vector<uint32_t> counters(FLAGS_numdistinct, 0);
+    size_t max_counter = 50;
+    RandomGenerator gen;
+
+    Status s;
+    DB* db = SelectDB(thread);
+    for (int64_t i = 0; i < FLAGS_numdistinct; i++) {
+      GenerateKeyFromInt(i * max_counter, FLAGS_num, &key);
+      s = db->Put(write_options_, key, gen.Generate(value_size_));
+      if (!s.ok()) {
+        fprintf(stderr, "Operation failed: %s\n", s.ToString().c_str());
+        exit(1);
+      }
+    }
+
+    db->GetSnapshot();
+
+    std::default_random_engine generator;
+    std::normal_distribution<double> distribution(FLAGS_numdistinct / 2.0,
+                                                  FLAGS_stddev);
+    Duration duration(FLAGS_duration, FLAGS_num);
+    while (!duration.Done(1)) {
+      int64_t rnd_id = static_cast<int64_t>(distribution(generator));
+      int64_t key_id = std::max(std::min(FLAGS_numdistinct - 1, rnd_id),
+                                static_cast<int64_t>(0));
+      GenerateKeyFromInt(key_id * max_counter + counters[key_id], FLAGS_num,
+                         &key);
+      s = FLAGS_use_single_deletes ? db->SingleDelete(write_options_, key)
+                                   : db->Delete(write_options_, key);
+      if (s.ok()) {
+        counters[key_id] = (counters[key_id] + 1) % max_counter;
+        GenerateKeyFromInt(key_id * max_counter + counters[key_id], FLAGS_num,
+                           &key);
+        s = db->Put(write_options_, key, Slice());
+      }
+
+      if (!s.ok()) {
+        fprintf(stderr, "Operation failed: %s\n", s.ToString().c_str());
+        exit(1);
+      }
+
+      thread->stats.FinishedOps(nullptr, db, 1);
+    }
+
+    char msg[200];
+    snprintf(msg, sizeof(msg),
+             "use single deletes: %d, "
+             "standard deviation: %lf\n",
+             FLAGS_use_single_deletes, FLAGS_stddev);
+    thread->stats.AddMessage(msg);
+  }

  void Compact(ThreadState* thread) {
    DB* db = SelectDB(thread);
@@ -3880,6 +4009,7 @@ int main(int argc, char** argv) {
  if (FLAGS_statistics) {
    dbstats = rocksdb::CreateDBStatistics();
  }
+  FLAGS_compaction_pri_e = (rocksdb::CompactionPri)FLAGS_compaction_pri;

  std::vector<std::string> fanout = rocksdb::StringSplit(
      FLAGS_max_bytes_for_level_multiplier_additional, ',');
@@ -7,8 +7,8 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"

 namespace rocksdb {

@@ -97,8 +97,11 @@ class ChangeFilter : public CompactionFilter {

 class KeepFilterFactory : public CompactionFilterFactory {
 public:
-  explicit KeepFilterFactory(bool check_context = false)
-      : check_context_(check_context) {}
+  explicit KeepFilterFactory(bool check_context = false,
+                             bool check_context_cf_id = false)
+      : check_context_(check_context),
+        check_context_cf_id_(check_context_cf_id),
+        compaction_filter_created_(false) {}

  virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
      const CompactionFilter::Context& context) override {
@@ -106,13 +109,22 @@ class KeepFilterFactory : public CompactionFilterFactory {
      EXPECT_EQ(expect_full_compaction_.load(), context.is_full_compaction);
      EXPECT_EQ(expect_manual_compaction_.load(), context.is_manual_compaction);
    }
+    if (check_context_cf_id_) {
+      EXPECT_EQ(expect_cf_id_.load(), context.column_family_id);
+    }
+    compaction_filter_created_ = true;
    return std::unique_ptr<CompactionFilter>(new KeepFilter());
  }

+  bool compaction_filter_created() const { return compaction_filter_created_; }
+
  virtual const char* Name() const override { return "KeepFilterFactory"; }
  bool check_context_;
+  bool check_context_cf_id_;
  std::atomic_bool expect_full_compaction_;
  std::atomic_bool expect_manual_compaction_;
+  std::atomic<uint32_t> expect_cf_id_;
+  bool compaction_filter_created_;
 };

 class DeleteFilterFactory : public CompactionFilterFactory {
@@ -174,6 +186,7 @@ class ChangeFilterFactory : public CompactionFilterFactory {
  virtual const char* Name() const override { return "ChangeFilterFactory"; }
 };

+#ifndef ROCKSDB_LITE
 TEST_F(DBTestCompactionFilter, CompactionFilter) {
  Options options = CurrentOptions();
  options.max_open_files = -1;
@@ -216,7 +229,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilter) {
  Arena arena;
  {
    ScopedArenaIterator iter(
-        dbfull()->TEST_NewInternalIterator(&arena, handles_[1]));
+        dbfull()->NewInternalIterator(&arena, handles_[1]));
    iter->SeekToFirst();
    ASSERT_OK(iter->status());
    while (iter->Valid()) {
@@ -304,7 +317,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilter) {
  count = 0;
  {
    ScopedArenaIterator iter(
-        dbfull()->TEST_NewInternalIterator(&arena, handles_[1]));
+        dbfull()->NewInternalIterator(&arena, handles_[1]));
    iter->SeekToFirst();
    ASSERT_OK(iter->status());
    while (iter->Valid()) {
@@ -350,6 +363,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilterDeletesAll) {

  delete itr;
 }
+#endif  // ROCKSDB_LITE

 TEST_F(DBTestCompactionFilter, CompactionFilterWithValueChange) {
  do {
@@ -481,8 +495,9 @@ TEST_F(DBTestCompactionFilter, CompactionFilterWithMergeOperator) {
  ASSERT_EQ(newvalue, four);
 }

+#ifndef ROCKSDB_LITE
 TEST_F(DBTestCompactionFilter, CompactionFilterContextManual) {
-  KeepFilterFactory* filter = new KeepFilterFactory();
+  KeepFilterFactory* filter = new KeepFilterFactory(true, true);

  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
@@ -504,22 +519,24 @@ TEST_F(DBTestCompactionFilter, CompactionFilterContextManual) {
    // be triggered.
    num_keys_per_file /= 2;
  }
+  dbfull()->TEST_WaitForCompact();

  // Force a manual compaction
  cfilter_count = 0;
  filter->expect_manual_compaction_.store(true);
-  filter->expect_full_compaction_.store(false);  // Manual compaction always
-                                                 // set this flag.
+  filter->expect_full_compaction_.store(true);
+  filter->expect_cf_id_.store(0);
  dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr);
  ASSERT_EQ(cfilter_count, 700);
  ASSERT_EQ(NumSortedRuns(0), 1);
+  ASSERT_TRUE(filter->compaction_filter_created());

  // Verify total number of keys is correct after manual compaction.
  {
    int count = 0;
    int total = 0;
    Arena arena;
-    ScopedArenaIterator iter(dbfull()->TEST_NewInternalIterator(&arena));
+    ScopedArenaIterator iter(dbfull()->NewInternalIterator(&arena));
    iter->SeekToFirst();
    ASSERT_OK(iter->status());
    while (iter->Valid()) {
@@ -536,15 +553,79 @@ TEST_F(DBTestCompactionFilter, CompactionFilterContextManual) {
    ASSERT_EQ(count, 1);
  }
 }
+#endif  // ROCKSDB_LITE
+
+TEST_F(DBTestCompactionFilter, CompactionFilterContextCfId) {
+  KeepFilterFactory* filter = new KeepFilterFactory(false, true);
+  filter->expect_cf_id_.store(1);
+
+  Options options = CurrentOptions();
+  options.compaction_filter_factory.reset(filter);
+  options.compression = kNoCompression;
+  options.level0_file_num_compaction_trigger = 2;
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  int num_keys_per_file = 400;
+  for (int j = 0; j < 3; j++) {
+    // Write several keys.
+    const std::string value(10, 'x');
+    for (int i = 0; i < num_keys_per_file; i++) {
+      char key[100];
+      snprintf(key, sizeof(key), "B%08d%02d", i, j);
+      Put(1, key, value);
+    }
+    Flush(1);
+    // Make sure next file is much smaller so automatic compaction will not
+    // be triggered.
+    num_keys_per_file /= 2;
+  }
+  dbfull()->TEST_WaitForCompact();
+
+  ASSERT_TRUE(filter->compaction_filter_created());
+}
+
+#ifndef ROCKSDB_LITE
+// Compaction filters should only be applied to records that are newer than the
+// latest snapshot. This test inserts records and applies a delete filter.
+TEST_F(DBTestCompactionFilter, CompactionFilterSnapshot) {
+  Options options;
+  options.compaction_filter_factory = std::make_shared<DeleteFilterFactory>();
+  options.disable_auto_compactions = true;
+  options.create_if_missing = true;
+  options = CurrentOptions(options);
+  DestroyAndReopen(options);
+
+  // Put some data.
+  const Snapshot* snapshot = nullptr;
+  for (int table = 0; table < 4; ++table) {
+    for (int i = 0; i < 10; ++i) {
+      Put(ToString(table * 100 + i), "val");
+    }
+    Flush();
+
+    if (table == 0) {
+      snapshot = db_->GetSnapshot();
+    }
+  }
+  assert(snapshot != nullptr);
+
+  cfilter_count = 0;
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  // The filter should delete 10 records.
+  ASSERT_EQ(30U, cfilter_count);
+
+  // Release the snapshot and compact again -> now all records should be
+  // removed.
+  db_->ReleaseSnapshot(snapshot);
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_EQ(0U, CountLiveFiles());
+}
+#endif  // ROCKSDB_LITE

 }  // namespace rocksdb

 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
-#else
-  return 0;
-#endif
 }
@@ -7,15 +7,14 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/experimental.h"
-#include "util/db_test_util.h"
 #include "util/sync_point.h"
 namespace rocksdb {

 // SYNC_POINT is not supported in released Windows mode.
-#if !(defined NDEBUG) || !defined(OS_WIN)
-
+#if !defined(ROCKSDB_LITE)

 class DBCompactionTest : public DBTestBase {
 public:
@@ -66,51 +65,6 @@ class OnFileDeletionListener : public EventListener {
  std::string expected_file_name_;
 };

-class SleepingBackgroundTask {
- public:
-  SleepingBackgroundTask()
-      : bg_cv_(&mutex_), should_sleep_(true), done_with_sleep_(false) {}
-  void DoSleep() {
-    MutexLock l(&mutex_);
-    while (should_sleep_) {
-      bg_cv_.Wait();
-    }
-    done_with_sleep_ = true;
-    bg_cv_.SignalAll();
-  }
-  void WakeUp() {
-    MutexLock l(&mutex_);
-    should_sleep_ = false;
-    bg_cv_.SignalAll();
-  }
-  void WaitUntilDone() {
-    MutexLock l(&mutex_);
-    while (!done_with_sleep_) {
-      bg_cv_.Wait();
-    }
-  }
-  bool WokenUp() {
-    MutexLock l(&mutex_);
-    return should_sleep_ == false;
-  }
-
-  void Reset() {
-    MutexLock l(&mutex_);
-    should_sleep_ = true;
-    done_with_sleep_ = false;
-  }
-
-  static void DoSleepTask(void* arg) {
-    reinterpret_cast<SleepingBackgroundTask*>(arg)->DoSleep();
-  }
-
- private:
-  port::Mutex mutex_;
-  port::CondVar bg_cv_;  // Signalled when background work finishes
-  bool should_sleep_;
-  bool done_with_sleep_;
-};
-
 static const int kCDTValueSize = 1000;
 static const int kCDTKeysPerBuffer = 4;
 static const int kCDTNumLevels = 8;
@@ -1273,6 +1227,7 @@ TEST_F(DBCompactionTest, L0_CompactionBug_Issue44_b) {
 TEST_P(DBCompactionTestWithParam, ManualCompaction) {
  Options options = CurrentOptions();
  options.max_subcompactions = max_subcompactions_;
+  options.statistics = rocksdb::CreateDBStatistics();
  CreateAndReopenWithCF({"pikachu"}, options);

  // iter - 0 with 7 levels
@@ -1304,7 +1259,14 @@ TEST_P(DBCompactionTestWithParam, ManualCompaction) {
    // Compact all
    MakeTables(1, "a", "z", 1);
    ASSERT_EQ("1,0,2", FilesPerLevel(1));
+
+    uint64_t prev_block_cache_add =
+        options.statistics->getTickerCount(BLOCK_CACHE_ADD);
    db_->CompactRange(CompactRangeOptions(), handles_[1], nullptr, nullptr);
+    // Verify manual compaction doesn't fill block cache
+    ASSERT_EQ(prev_block_cache_add,
+              options.statistics->getTickerCount(BLOCK_CACHE_ADD));
+
    ASSERT_EQ("0,0,1", FilesPerLevel(1));

    if (iter == 0) {
@@ -1312,6 +1274,7 @@ TEST_P(DBCompactionTestWithParam, ManualCompaction) {
      options.max_background_flushes = 0;
      options.num_levels = 3;
      options.create_if_missing = true;
+      options.statistics = rocksdb::CreateDBStatistics();
      DestroyAndReopen(options);
      CreateAndReopenWithCF({"pikachu"}, options);
    }
@@ -1493,8 +1456,8 @@ TEST_P(DBCompactionTestWithParam, PartialCompactionFailure) {
  env_->SetBackgroundThreads(1, Env::HIGH);
  env_->SetBackgroundThreads(1, Env::LOW);
  // stop the compaction thread until we simulate the file creation failure.
-  SleepingBackgroundTask sleeping_task_low;
-  env_->Schedule(&SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
+  test::SleepingBackgroundTask sleeping_task_low;
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
                 Env::Priority::LOW);

  options.env = env_;
@@ -1586,8 +1549,8 @@ TEST_P(DBCompactionTestWithParam, DeleteMovedFileAfterCompaction) {
    ASSERT_EQ("0,1", FilesPerLevel(0));

    // block compactions
-    SleepingBackgroundTask sleeping_task;
-    env_->Schedule(&SleepingBackgroundTask::DoSleepTask, &sleeping_task,
+    test::SleepingBackgroundTask sleeping_task;
+    env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
                   Env::Priority::LOW);

    options.max_bytes_for_level_base = 1024 * 1024;  // 1 MB
@@ -1889,11 +1852,11 @@ TEST_P(DBCompactionTestWithParam, ForceBottommostLevelCompaction) {

 INSTANTIATE_TEST_CASE_P(DBCompactionTestWithParam, DBCompactionTestWithParam,
                        ::testing::Values(1, 4));
-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif // !defined(ROCKSDB_LITE)
 }  // namespace rocksdb

 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
@@ -10,10 +10,10 @@
 // Introduction of SyncPoint effectively disabled building and running this test
 // in Release build.
 // which is a pity, it is a good test
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)

+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"

 namespace rocksdb {
 class DBTestDynamicLevel : public DBTestBase {
@@ -484,10 +484,10 @@ TEST_F(DBTestDynamicLevel, MigrateToDynamicLevelMaxBytesBase) {
 }
 }  // namespace rocksdb

-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif  // !defined(ROCKSDB_LITE)

 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
@@ -12,7 +12,6 @@
 #include <deque>
 #include <limits>
 #include <list>
-#include <list>
 #include <set>
 #include <string>
 #include <utility>
@@ -37,12 +36,11 @@
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/transaction_log.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/autovector.h"
 #include "util/event_logger.h"
 #include "util/hash.h"
-#include "util/hash.h"
 #include "util/instrumented_mutex.h"
-#include "util/scoped_arena_iterator.h"
 #include "util/stop_watch.h"
 #include "util/thread_local.h"

@@ -53,10 +51,10 @@ class TableCache;
 class Version;
 class VersionEdit;
 class VersionSet;
-class CompactionFilterV2;
 class Arena;
 class WriteCallback;
 struct JobContext;
+struct ExternalSstFileInfo;

 class DBImpl : public DB {
 public:
@@ -76,6 +74,10 @@ class DBImpl : public DB {
  virtual Status Delete(const WriteOptions& options,
                        ColumnFamilyHandle* column_family,
                        const Slice& key) override;
+  using DB::SingleDelete;
+  virtual Status SingleDelete(const WriteOptions& options,
+                              ColumnFamilyHandle* column_family,
+                              const Slice& key) override;
  using DB::Write;
  virtual Status Write(const WriteOptions& options,
                       WriteBatch* updates) override;
@@ -120,6 +122,9 @@ class DBImpl : public DB {
  using DB::GetIntProperty;
  virtual bool GetIntProperty(ColumnFamilyHandle* column_family,
                              const Slice& property, uint64_t* value) override;
+  using DB::GetAggregatedIntProperty;
+  virtual bool GetAggregatedIntProperty(const Slice& property,
+                                        uint64_t* aggregated_value) override;
  using DB::GetApproximateSizes;
  virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
                                   const Range* range, int n, uint64_t* sizes,
@@ -136,6 +141,9 @@ class DBImpl : public DB {
                              const int output_level,
                              const int output_path_id = -1) override;

+  virtual Status PauseBackgroundWork() override;
+  virtual Status ContinueBackgroundWork() override;
+
  using DB::SetOptions;
  Status SetOptions(
      ColumnFamilyHandle* column_family,
@@ -223,6 +231,13 @@ class DBImpl : public DB {
  Status GetLatestSequenceForKeyFromMemtable(SuperVersion* sv, const Slice& key,
                                             SequenceNumber* seq);

+  using DB::AddFile;
+  virtual Status AddFile(ColumnFamilyHandle* column_family,
+                         const ExternalSstFileInfo* file_info,
+                         bool move_file) override;
+  virtual Status AddFile(ColumnFamilyHandle* column_family,
+                         const std::string& file_path, bool move_file) override;
+
 #endif  // ROCKSDB_LITE

  // checks if all live files exist on file system and that their file sizes
@@ -236,7 +251,13 @@ class DBImpl : public DB {
                             const Slice* begin, const Slice* end,
                             bool disallow_trivial_move = false);

-#ifndef ROCKSDB_LITE
+  // Return an internal iterator over the current state of the database.
+  // The keys of this iterator are internal keys (see format.h).
+  // The returned iterator should be deleted when no longer needed.
+  InternalIterator* NewInternalIterator(
+      Arena* arena, ColumnFamilyHandle* column_family = nullptr);
+
+#ifndef NDEBUG
  // Extra methods (for testing) that are not in the public DB interface
  // Implemented in db_impl_debug.cc

@@ -254,12 +275,6 @@ class DBImpl : public DB {
  // Wait for any compaction
  Status TEST_WaitForCompact();

-  // Return an internal iterator over the current state of the database.
-  // The keys of this iterator are internal keys (see format.h).
-  // The returned iterator should be deleted when no longer needed.
-  Iterator* TEST_NewInternalIterator(
-      Arena* arena, ColumnFamilyHandle* column_family = nullptr);
-
  // Return the maximum overlapping data (in bytes) at next level for any
  // file at a level >= 1.
  int64_t TEST_MaxNextLevelOverlappingBytes(ColumnFamilyHandle* column_family =
@@ -293,7 +308,13 @@ class DBImpl : public DB {

  uint64_t TEST_LogfileNumber();

-#endif  // ROCKSDB_LITE
+  // Returns column family name to ImmutableCFOptions map.
+  Status TEST_GetAllImmutableCFOptions(
+      std::unordered_map<std::string, const ImmutableCFOptions*>* iopts_map);
+
+  Cache* TEST_table_cache() { return table_cache_.get(); }
+
+#endif  // NDEBUG

  // Returns the list of live files in 'live' and the list
  // of all files in the filesystem in 'candidate_files'.
@@ -351,6 +372,20 @@ class DBImpl : public DB {
  // Same as above, should called without mutex held and not on write thread.
  ColumnFamilyHandle* GetColumnFamilyHandleUnlocked(uint32_t column_family_id);

+  // Returns the number of currently running flushes.
+  // REQUIREMENT: mutex_ must be held when calling this function.
+  int num_running_flushes() {
+    mutex_.AssertHeld();
+    return num_running_flushes_;
+  }
+
+  // Returns the number of currently running compactions.
+  // REQUIREMENT: mutex_ must be held when calling this function.
+  int num_running_compactions() {
+    mutex_.AssertHeld();
+    return num_running_compactions_;
+  }
+
 protected:
  Env* const env_;
  const std::string dbname_;
@@ -358,12 +393,14 @@ class DBImpl : public DB {
  const DBOptions db_options_;
  Statistics* stats_;

-  Iterator* NewInternalIterator(const ReadOptions&, ColumnFamilyData* cfd,
-                                SuperVersion* super_version, Arena* arena);
+  InternalIterator* NewInternalIterator(const ReadOptions&,
+                                        ColumnFamilyData* cfd,
+                                        SuperVersion* super_version,
+                                        Arena* arena);

  void NotifyOnFlushCompleted(ColumnFamilyData* cfd, FileMetaData* file_meta,
                              const MutableCFOptions& mutable_cf_options,
-                              int job_id);
+                              int job_id, TableProperties prop);

  void NotifyOnCompactionCompleted(ColumnFamilyData* cfd,
                                   Compaction *c, const Status &st,
@@ -528,6 +565,8 @@ class DBImpl : public DB {
  // * whenever there is an error in background flush or compaction
  InstrumentedCondVar bg_cv_;
  uint64_t logfile_number_;
+  std::deque<uint64_t>
+      log_recycle_files;  // a list of log files that we can recycle
  bool log_dir_synced_;
  bool log_empty_;
  ColumnFamilyHandleImpl* default_cf_handle_;
@@ -673,6 +712,9 @@ class DBImpl : public DB {
  // count how many background compactions are running or have been scheduled
  int bg_compaction_scheduled_;

+  // stores the number of compactions are currently running
+  int num_running_compactions_;
+
  // If non-zero, MaybeScheduleFlushOrCompaction() will only schedule manual
  // compactions (if manual_compaction_ is not null). This mechanism enables
  // manual compactions to wait until all other compactions are finished.
@@ -681,6 +723,9 @@ class DBImpl : public DB {
  // number of background memtable flush jobs, submitted to the HIGH pool
  int bg_flush_scheduled_;

+  // stores the number of flushes are currently running
+  int num_running_flushes_;
+
  // Information for a manual compaction
  struct ManualCompaction {
    ColumnFamilyData* cfd;
@@ -736,8 +781,11 @@ class DBImpl : public DB {
  // Unified interface for logging events
  EventLogger event_logger_;

-  // A value of true temporarily disables scheduling of background work
-  bool bg_work_gate_closed_;
+  // A value of > 0 temporarily disables scheduling of background work
+  int bg_work_paused_;
+
+  // A value of > 0 temporarily disables scheduling of background compaction
+  int bg_compaction_paused_;

  // Guard against multiple concurrent refitting
  bool refitting_level_;
@@ -776,6 +824,10 @@ class DBImpl : public DB {
  virtual Status GetPropertiesOfAllTables(ColumnFamilyHandle* column_family,
                                          TablePropertiesCollection* props)
      override;
+  virtual Status GetPropertiesOfTablesInRange(
+      ColumnFamilyHandle* column_family, const Range* range, std::size_t n,
+      TablePropertiesCollection* props) override;
+
 #endif  // ROCKSDB_LITE

  // Function that Get and KeyMayExist call with no_io true or false
@@ -784,9 +836,10 @@ class DBImpl : public DB {
                 const Slice& key, std::string* value,
                 bool* value_found = nullptr);

-  bool GetIntPropertyInternal(ColumnFamilyHandle* column_family,
+  bool GetIntPropertyInternal(ColumnFamilyData* cfd,
                              DBPropertyType property_type,
-                              bool need_out_of_mutex, uint64_t* value);
+                              bool need_out_of_mutex, bool is_locked,
+                              uint64_t* value);
 };

 // Sanitize db options.  The caller should delete result.info_log if
@@ -7,7 +7,7 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

-#ifndef ROCKSDB_LITE
+#ifndef NDEBUG

 #include "db/db_impl.h"
 #include "util/thread_status_updater.h"
@@ -19,23 +19,6 @@ uint64_t DBImpl::TEST_GetLevel0TotalSize() {
  return default_cf_handle_->cfd()->current()->storage_info()->NumLevelBytes(0);
 }

-Iterator* DBImpl::TEST_NewInternalIterator(Arena* arena,
-                                           ColumnFamilyHandle* column_family) {
-  ColumnFamilyData* cfd;
-  if (column_family == nullptr) {
-    cfd = default_cf_handle_->cfd();
-  } else {
-    auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
-    cfd = cfh->cfd();
-  }
-
-  mutex_.Lock();
-  SuperVersion* super_version = cfd->GetSuperVersion()->Ref();
-  mutex_.Unlock();
-  ReadOptions roptions;
-  return NewInternalIterator(roptions, cfd, super_version, arena);
-}
-
 int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes(
    ColumnFamilyHandle* column_family) {
  ColumnFamilyData* cfd;
@@ -152,5 +135,24 @@ uint64_t DBImpl::TEST_LogfileNumber() {
  return logfile_number_;
 }

+Status DBImpl::TEST_GetAllImmutableCFOptions(
+    std::unordered_map<std::string, const ImmutableCFOptions*>* iopts_map) {
+  std::vector<std::string> cf_names;
+  std::vector<const ImmutableCFOptions*> iopts;
+  {
+    InstrumentedMutexLock l(&mutex_);
+    for (auto cfd : *versions_->GetColumnFamilySet()) {
+      cf_names.push_back(cfd->GetName());
+      iopts.push_back(cfd->ioptions());
+    }
+  }
+  iopts_map->clear();
+  for (size_t i = 0; i < cf_names.size(); ++i) {
+    iopts_map->insert({cf_names[i], iopts[i]});
+  }
+
+  return Status::OK();
+}
+
 }  // namespace rocksdb
-#endif  // ROCKSDB_LITE
+#endif  // NDEBUG
@@ -53,6 +53,12 @@ class DBImplReadOnly : public DBImpl {
                        const Slice& key) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
+  using DBImpl::SingleDelete;
+  virtual Status SingleDelete(const WriteOptions& options,
+                              ColumnFamilyHandle* column_family,
+                              const Slice& key) override {
+    return Status::NotSupported("Not supported operation in read only mode.");
+  }
  virtual Status Write(const WriteOptions& options,
                       WriteBatch* updates) override {
    return Status::NotSupported("Not supported operation in read only mode.");
@@ -94,6 +100,11 @@ class DBImplReadOnly : public DBImpl {
    return Status::NotSupported("Not supported operation in read only mode.");
  }

+  using DBImpl::SyncWAL;
+  virtual Status SyncWAL() override {
+    return Status::NotSupported("Not supported operation in read only mode.");
+  }
+
 private:
  friend class DB;

@@ -6,8 +6,8 @@
 // Copyright (c) 2011 The LevelDB Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"

 namespace rocksdb {

@@ -15,11 +15,12 @@

 #include "db/filename.h"
 #include "db/dbformat.h"
+#include "port/port.h"
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
 #include "rocksdb/iterator.h"
 #include "rocksdb/merge_operator.h"
-#include "port/port.h"
+#include "table/internal_iterator.h"
 #include "util/arena.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
@@ -58,10 +59,11 @@ class DBIter: public Iterator {
    kReverse
  };

-  DBIter(Env* env, const ImmutableCFOptions& ioptions,
-         const Comparator* cmp, Iterator* iter, SequenceNumber s,
-         bool arena_mode, uint64_t max_sequential_skip_in_iterations,
-         const Slice* iterate_upper_bound = nullptr)
+  DBIter(Env* env, const ImmutableCFOptions& ioptions, const Comparator* cmp,
+         InternalIterator* iter, SequenceNumber s, bool arena_mode,
+         uint64_t max_sequential_skip_in_iterations,
+         const Slice* iterate_upper_bound = nullptr,
+         bool prefix_same_as_start = false)
      : arena_mode_(arena_mode),
        env_(env),
        logger_(ioptions.info_log),
@@ -73,7 +75,8 @@ class DBIter: public Iterator {
        valid_(false),
        current_entry_is_merged_(false),
        statistics_(ioptions.statistics),
-        iterate_upper_bound_(iterate_upper_bound) {
+        iterate_upper_bound_(iterate_upper_bound),
+        prefix_same_as_start_(prefix_same_as_start) {
    RecordTick(statistics_, NO_ITERATORS);
    prefix_extractor_ = ioptions.prefix_extractor;
    max_skip_ = max_sequential_skip_in_iterations;
@@ -83,10 +86,10 @@ class DBIter: public Iterator {
    if (!arena_mode_) {
      delete iter_;
    } else {
-      iter_->~Iterator();
+      iter_->~InternalIterator();
    }
  }
-  virtual void SetIter(Iterator* iter) {
+  virtual void SetIter(InternalIterator* iter) {
    assert(iter_ == nullptr);
    iter_ = iter;
  }
@@ -142,7 +145,7 @@ class DBIter: public Iterator {
  Logger* logger_;
  const Comparator* const user_comparator_;
  const MergeOperator* const user_merge_operator_;
-  Iterator* iter_;
+  InternalIterator* iter_;
  SequenceNumber const sequence_;

  Status status_;
@@ -154,6 +157,8 @@ class DBIter: public Iterator {
  Statistics* statistics_;
  uint64_t max_skip_;
  const Slice* iterate_upper_bound_;
+  IterKey prefix_start_;
+  bool prefix_same_as_start_;

  // No copying allowed
  DBIter(const DBIter&);
@@ -189,6 +194,18 @@ void DBIter::Next() {
    return;
  }
  FindNextUserEntry(true /* skipping the current user key */);
+  if (statistics_ != nullptr) {
+    RecordTick(statistics_, NUMBER_DB_NEXT);
+    if (valid_) {
+      RecordTick(statistics_, NUMBER_DB_NEXT_FOUND);
+      RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
+    }
+  }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_ &&
+      prefix_extractor_->Transform(saved_key_.GetKey())
+              .compare(prefix_start_.GetKey()) != 0) {
+    valid_ = false;
+  }
 }

 // PRE: saved_key_ has the current user key if skipping
@@ -228,6 +245,7 @@ void DBIter::FindNextUserEntryInternal(bool skipping) {
        } else {
          switch (ikey.type) {
            case kTypeDeletion:
+            case kTypeSingleDeletion:
              // Arrange to skip all upcoming entries for this key since
              // they are hidden by this deletion.
              saved_key_.SetKey(ikey.user_key);
@@ -301,16 +319,12 @@ void DBIter::MergeValuesNewToOld() {
    if (!user_comparator_->Equal(ikey.user_key, saved_key_.GetKey())) {
      // hit the next user key, stop right here
      break;
-    }
-
-    if (kTypeDeletion == ikey.type) {
+    } else if (kTypeDeletion == ikey.type || kTypeSingleDeletion == ikey.type) {
      // hit a delete with the same user key, stop right here
      // iter_ is positioned after delete
      iter_->Next();
      break;
-    }
-
-    if (kTypeValue == ikey.type) {
+    } else if (kTypeValue == ikey.type) {
      // hit a put, merge the put value with operands and store the
      // final result in saved_value_. We are done!
      // ignore corruption if there is any.
@@ -326,13 +340,13 @@ void DBIter::MergeValuesNewToOld() {
      // iter_ is positioned after put
      iter_->Next();
      return;
-    }
-
-    if (kTypeMerge == ikey.type) {
+    } else if (kTypeMerge == ikey.type) {
      // hit a merge, add the value as an operand and run associative merge.
      // when complete, add result to operands and continue.
      const Slice& val = iter_->value();
      operands.push_front(val.ToString());
+    } else {
+      assert(false);
    }
  }

@@ -355,6 +369,18 @@ void DBIter::Prev() {
    ReverseToBackward();
  }
  PrevInternal();
+  if (statistics_ != nullptr) {
+    RecordTick(statistics_, NUMBER_DB_PREV);
+    if (valid_) {
+      RecordTick(statistics_, NUMBER_DB_PREV_FOUND);
+      RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
+    }
+  }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_ &&
+      prefix_extractor_->Transform(saved_key_.GetKey())
+              .compare(prefix_start_.GetKey()) != 0) {
+    valid_ = false;
+  }
 }

 void DBIter::ReverseToBackward() {
@@ -419,12 +445,14 @@ void DBIter::PrevInternal() {
 }

 // This function checks, if the entry with biggest sequence_number <= sequence_
-// is non kTypeDeletion. If it's not, we save value in saved_value_
+// is non kTypeDeletion or kTypeSingleDeletion. If it's not, we save value in
+// saved_value_
 bool DBIter::FindValueForCurrentKey() {
  assert(iter_->Valid());
  // Contains operands for merge operator.
  std::deque<std::string> operands;
-  // last entry before merge (could be kTypeDeletion or kTypeValue)
+  // last entry before merge (could be kTypeDeletion, kTypeSingleDeletion or
+  // kTypeValue)
  ValueType last_not_merge_type = kTypeDeletion;
  ValueType last_key_entry_type = kTypeDeletion;

@@ -447,8 +475,9 @@ bool DBIter::FindValueForCurrentKey() {
        last_not_merge_type = kTypeValue;
        break;
      case kTypeDeletion:
+      case kTypeSingleDeletion:
        operands.clear();
-        last_not_merge_type = kTypeDeletion;
+        last_not_merge_type = last_key_entry_type;
        PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
        break;
      case kTypeMerge:
@@ -468,6 +497,7 @@ bool DBIter::FindValueForCurrentKey() {

  switch (last_key_entry_type) {
    case kTypeDeletion:
+    case kTypeSingleDeletion:
      valid_ = false;
      return false;
    case kTypeMerge:
@@ -516,7 +546,8 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
  ParsedInternalKey ikey;
  FindParseableKey(&ikey, kForward);

-  if (ikey.type == kTypeValue || ikey.type == kTypeDeletion) {
+  if (ikey.type == kTypeValue || ikey.type == kTypeDeletion ||
+      ikey.type == kTypeSingleDeletion) {
    if (ikey.type == kTypeValue) {
      saved_value_ = iter_->value().ToString();
      valid_ = true;
@@ -539,7 +570,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {

  if (!iter_->Valid() ||
      !user_comparator_->Equal(ikey.user_key, saved_key_.GetKey()) ||
-      ikey.type == kTypeDeletion) {
+      ikey.type == kTypeDeletion || ikey.type == kTypeSingleDeletion) {
    {
      StopWatchNano timer(env_, statistics_ != nullptr);
      PERF_TIMER_GUARD(merge_operator_time_nanos);
@@ -637,13 +668,23 @@ void DBIter::Seek(const Slice& target) {
    iter_->Seek(saved_key_.GetKey());
  }

+  RecordTick(statistics_, NUMBER_DB_SEEK);
  if (iter_->Valid()) {
    direction_ = kForward;
    ClearSavedValue();
    FindNextUserEntry(false /* not skipping */);
+    if (statistics_ != nullptr) {
+      if (valid_) {
+        RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
+        RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
+      }
+    }
  } else {
    valid_ = false;
  }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
+    prefix_start_.SetKey(prefix_extractor_->Transform(target));
+  }
 }

 void DBIter::SeekToFirst() {
@@ -660,11 +701,21 @@ void DBIter::SeekToFirst() {
    iter_->SeekToFirst();
  }

+  RecordTick(statistics_, NUMBER_DB_SEEK);
  if (iter_->Valid()) {
    FindNextUserEntry(false /* not skipping */);
+    if (statistics_ != nullptr) {
+      if (valid_) {
+        RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
+        RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
+      }
+    }
  } else {
    valid_ = false;
  }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
+    prefix_start_.SetKey(prefix_extractor_->Transform(saved_key_.GetKey()));
+  }
 }

 void DBIter::SeekToLast() {
@@ -703,24 +754,35 @@ void DBIter::SeekToLast() {
    }
  }
  PrevInternal();
+  if (statistics_ != nullptr) {
+    RecordTick(statistics_, NUMBER_DB_SEEK);
+    if (valid_) {
+      RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
+      RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
+    }
+  }
+  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
+    prefix_start_.SetKey(prefix_extractor_->Transform(saved_key_.GetKey()));
+  }
 }

 Iterator* NewDBIterator(Env* env, const ImmutableCFOptions& ioptions,
                        const Comparator* user_key_comparator,
-                        Iterator* internal_iter,
+                        InternalIterator* internal_iter,
                        const SequenceNumber& sequence,
                        uint64_t max_sequential_skip_in_iterations,
-                        const Slice* iterate_upper_bound) {
+                        const Slice* iterate_upper_bound,
+                        bool prefix_same_as_start) {
  return new DBIter(env, ioptions, user_key_comparator, internal_iter, sequence,
                    false, max_sequential_skip_in_iterations,
-                    iterate_upper_bound);
+                    iterate_upper_bound, prefix_same_as_start);
 }

 ArenaWrappedDBIter::~ArenaWrappedDBIter() { db_iter_->~DBIter(); }

 void ArenaWrappedDBIter::SetDBIter(DBIter* iter) { db_iter_ = iter; }

-void ArenaWrappedDBIter::SetIterUnderDBIter(Iterator* iter) {
+void ArenaWrappedDBIter::SetIterUnderDBIter(InternalIterator* iter) {
  static_cast<DBIter*>(db_iter_)->SetIter(iter);
 }

@@ -742,16 +804,16 @@ void ArenaWrappedDBIter::RegisterCleanup(CleanupFunction function, void* arg1,

 ArenaWrappedDBIter* NewArenaWrappedDbIterator(
    Env* env, const ImmutableCFOptions& ioptions,
-    const Comparator* user_key_comparator,
-    const SequenceNumber& sequence,
+    const Comparator* user_key_comparator, const SequenceNumber& sequence,
    uint64_t max_sequential_skip_in_iterations,
-    const Slice* iterate_upper_bound) {
+    const Slice* iterate_upper_bound, bool prefix_same_as_start) {
  ArenaWrappedDBIter* iter = new ArenaWrappedDBIter();
  Arena* arena = iter->GetArena();
  auto mem = arena->AllocateAligned(sizeof(DBIter));
-  DBIter* db_iter = new (mem) DBIter(env, ioptions, user_key_comparator,
-      nullptr, sequence, true, max_sequential_skip_in_iterations,
-      iterate_upper_bound);
+  DBIter* db_iter =
+      new (mem) DBIter(env, ioptions, user_key_comparator, nullptr, sequence,
+                       true, max_sequential_skip_in_iterations,
+                       iterate_upper_bound, prefix_same_as_start);

  iter->SetDBIter(db_iter);

@@ -10,6 +10,7 @@
 #pragma once
 #include <stdint.h>
 #include "rocksdb/db.h"
+#include "rocksdb/iterator.h"
 #include "db/dbformat.h"
 #include "util/arena.h"
 #include "util/autovector.h"
@@ -18,18 +19,18 @@ namespace rocksdb {

 class Arena;
 class DBIter;
+class InternalIterator;

 // Return a new iterator that converts internal keys (yielded by
 // "*internal_iter") that were live at the specified "sequence" number
 // into appropriate user keys.
-extern Iterator* NewDBIterator(
-    Env* env,
-    const ImmutableCFOptions& options,
-    const Comparator *user_key_comparator,
-    Iterator* internal_iter,
-    const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations,
-    const Slice* iterate_upper_bound = nullptr);
+extern Iterator* NewDBIterator(Env* env, const ImmutableCFOptions& options,
+                               const Comparator* user_key_comparator,
+                               InternalIterator* internal_iter,
+                               const SequenceNumber& sequence,
+                               uint64_t max_sequential_skip_in_iterations,
+                               const Slice* iterate_upper_bound = nullptr,
+                               bool prefix_same_as_start = false);

 // A wrapper iterator which wraps DB Iterator and the arena, with which the DB
 // iterator is supposed be allocated. This class is used as an entry point of
@@ -50,7 +51,7 @@ class ArenaWrappedDBIter : public Iterator {

  // Set the internal iterator wrapped inside the DB Iterator. Usually it is
  // a merging iterator.
-  virtual void SetIterUnderDBIter(Iterator* iter);
+  virtual void SetIterUnderDBIter(InternalIterator* iter);
  virtual bool Valid() const override;
  virtual void SeekToFirst() override;
  virtual void SeekToLast() override;
@@ -60,6 +61,7 @@ class ArenaWrappedDBIter : public Iterator {
  virtual Slice key() const override;
  virtual Slice value() const override;
  virtual Status status() const override;
+
  void RegisterCleanup(CleanupFunction function, void* arg1, void* arg2);

 private:
@@ -70,8 +72,9 @@ class ArenaWrappedDBIter : public Iterator {
 // Generate the arena wrapped iterator class.
 extern ArenaWrappedDBIter* NewArenaWrappedDbIterator(
    Env* env, const ImmutableCFOptions& options,
-    const Comparator* user_key_comparator,
-    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
-    const Slice* iterate_upper_bound = nullptr);
+    const Comparator* user_key_comparator, const SequenceNumber& sequence,
+    uint64_t max_sequential_skip_in_iterations,
+    const Slice* iterate_upper_bound = nullptr,
+    bool prefix_same_as_start = false);

 }  // namespace rocksdb
@@ -29,7 +29,7 @@ static uint64_t TestGetTickerCount(const Options& options,
  return options.statistics->getTickerCount(ticker_type);
 }

-class TestIterator : public Iterator {
+class TestIterator : public InternalIterator {
 public:
  explicit TestIterator(const Comparator* comparator)
      : initialized_(false),
@@ -38,16 +38,20 @@ class TestIterator : public Iterator {
        iter_(0),
        cmp(comparator) {}

-  void AddMerge(std::string argkey, std::string argvalue) {
-    Add(argkey, kTypeMerge, argvalue);
+  void AddPut(std::string argkey, std::string argvalue) {
+    Add(argkey, kTypeValue, argvalue);
  }

  void AddDeletion(std::string argkey) {
    Add(argkey, kTypeDeletion, std::string());
  }

-  void AddPut(std::string argkey, std::string argvalue) {
-    Add(argkey, kTypeValue, argvalue);
+  void AddSingleDeletion(std::string argkey) {
+    Add(argkey, kTypeSingleDeletion, std::string());
+  }
+
+  void AddMerge(std::string argkey, std::string argvalue) {
+    Add(argkey, kTypeMerge, argvalue);
  }

  void Add(std::string argkey, ValueType type, std::string argvalue) {
@@ -891,6 +895,8 @@ TEST_F(DBIteratorTest, DBIterator1) {
  db_iter->Next();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "b");
+  db_iter->Next();
+  ASSERT_FALSE(db_iter->Valid());
 }

 TEST_F(DBIteratorTest, DBIterator2) {
@@ -1787,6 +1793,58 @@ TEST_F(DBIteratorTest, SeekToLastOccurrenceSeq0) {
  ASSERT_FALSE(db_iter->Valid());
 }

+TEST_F(DBIteratorTest, DBIterator11) {
+  Options options;
+  options.merge_operator = MergeOperators::CreateFromStringId("stringappend");
+
+  TestIterator* internal_iter = new TestIterator(BytewiseComparator());
+  internal_iter->AddPut("a", "0");
+  internal_iter->AddPut("b", "0");
+  internal_iter->AddSingleDeletion("b");
+  internal_iter->AddMerge("a", "1");
+  internal_iter->AddMerge("b", "2");
+  internal_iter->Finish();
+
+  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 1,
+      options.max_sequential_skip_in_iterations));
+  db_iter->SeekToFirst();
+  ASSERT_TRUE(db_iter->Valid());
+  ASSERT_EQ(db_iter->key().ToString(), "a");
+  ASSERT_EQ(db_iter->value().ToString(), "0");
+  db_iter->Next();
+  ASSERT_TRUE(db_iter->Valid());
+  ASSERT_EQ(db_iter->key().ToString(), "b");
+  db_iter->Next();
+  ASSERT_FALSE(db_iter->Valid());
+}
+
+TEST_F(DBIteratorTest, DBIterator12) {
+  Options options;
+  options.merge_operator = nullptr;
+
+  TestIterator* internal_iter = new TestIterator(BytewiseComparator());
+  internal_iter->AddPut("a", "1");
+  internal_iter->AddPut("b", "2");
+  internal_iter->AddPut("c", "3");
+  internal_iter->AddSingleDeletion("b");
+  internal_iter->Finish();
+
+  std::unique_ptr<Iterator> db_iter(
+      NewDBIterator(env_, ImmutableCFOptions(options), BytewiseComparator(),
+                    internal_iter, 10, 0));
+  db_iter->SeekToLast();
+  ASSERT_TRUE(db_iter->Valid());
+  ASSERT_EQ(db_iter->key().ToString(), "c");
+  ASSERT_EQ(db_iter->value().ToString(), "3");
+  db_iter->Prev();
+  ASSERT_TRUE(db_iter->Valid());
+  ASSERT_EQ(db_iter->key().ToString(), "a");
+  ASSERT_EQ(db_iter->value().ToString(), "1");
+  db_iter->Prev();
+  ASSERT_FALSE(db_iter->Valid());
+}
+
 class DBIterWithMergeIterTest : public testing::Test {
 public:
  DBIterWithMergeIterTest()
@@ -1806,11 +1864,12 @@ class DBIterWithMergeIterTest : public testing::Test {
    internal_iter2_->Add("d", kTypeValue, "7", 3u);
    internal_iter2_->Finish();

-    std::vector<Iterator*> child_iters;
+    std::vector<InternalIterator*> child_iters;
    child_iters.push_back(internal_iter1_);
    child_iters.push_back(internal_iter2_);
    InternalKeyComparator icomp(BytewiseComparator());
-    Iterator* merge_iter = NewMergingIterator(&icomp_, &child_iters[0], 2u);
+    InternalIterator* merge_iter =
+        NewMergingIterator(&icomp_, &child_iters[0], 2u);

    db_iter_.reset(NewDBIterator(env_, ImmutableCFOptions(options_),
                                 BytewiseComparator(), merge_iter,
@@ -10,10 +10,10 @@
 // Introduction of SyncPoint effectively disabled building and running this test
 // in Release build.
 // which is a pity, it is a good test
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)

+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"

 namespace rocksdb {

@@ -277,10 +277,10 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorBlobs) {
 }
 }  // namespace rocksdb

-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif  // !defined(ROCKSDB_LITE)

 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
@@ -0,0 +1,213 @@
+//  Copyright (c) 2013, 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.
+//
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include <unordered_set>
+#include <vector>
+
+#include "db/db_test_util.h"
+#include "port/stack_trace.h"
+#include "rocksdb/db.h"
+#include "util/testharness.h"
+#include "util/testutil.h"
+
+#ifndef ROCKSDB_LITE
+
+namespace rocksdb {
+
+// A helper function that ensures the table properties returned in
+// `GetPropertiesOfAllTablesTest` is correct.
+// This test assumes entries size is different for each of the tables.
+namespace {
+
+void VerifyTableProperties(DB* db, uint64_t expected_entries_size) {
+  TablePropertiesCollection props;
+  ASSERT_OK(db->GetPropertiesOfAllTables(&props));
+
+  ASSERT_EQ(4U, props.size());
+  std::unordered_set<uint64_t> unique_entries;
+
+  // Indirect test
+  uint64_t sum = 0;
+  for (const auto& item : props) {
+    unique_entries.insert(item.second->num_entries);
+    sum += item.second->num_entries;
+  }
+
+  ASSERT_EQ(props.size(), unique_entries.size());
+  ASSERT_EQ(expected_entries_size, sum);
+}
+}  // namespace
+
+class DBTablePropertiesTest : public DBTestBase {
+ public:
+  DBTablePropertiesTest() : DBTestBase("/db_table_properties_test") {}
+  TablePropertiesCollection TestGetPropertiesOfTablesInRange(
+      std::vector<Range> ranges, std::size_t* num_properties = nullptr,
+      std::size_t* num_files = nullptr);
+};
+
+TEST_F(DBTablePropertiesTest, GetPropertiesOfAllTablesTest) {
+  Options options = CurrentOptions();
+  options.level0_file_num_compaction_trigger = 8;
+  Reopen(options);
+  // Create 4 tables
+  for (int table = 0; table < 4; ++table) {
+    for (int i = 0; i < 10 + table; ++i) {
+      db_->Put(WriteOptions(), ToString(table * 100 + i), "val");
+    }
+    db_->Flush(FlushOptions());
+  }
+
+  // 1. Read table properties directly from file
+  Reopen(options);
+  VerifyTableProperties(db_, 10 + 11 + 12 + 13);
+
+  // 2. Put two tables to table cache and
+  Reopen(options);
+  // fetch key from 1st and 2nd table, which will internally place that table to
+  // the table cache.
+  for (int i = 0; i < 2; ++i) {
+    Get(ToString(i * 100 + 0));
+  }
+
+  VerifyTableProperties(db_, 10 + 11 + 12 + 13);
+
+  // 3. Put all tables to table cache
+  Reopen(options);
+  // fetch key from 1st and 2nd table, which will internally place that table to
+  // the table cache.
+  for (int i = 0; i < 4; ++i) {
+    Get(ToString(i * 100 + 0));
+  }
+  VerifyTableProperties(db_, 10 + 11 + 12 + 13);
+}
+
+TablePropertiesCollection
+DBTablePropertiesTest::TestGetPropertiesOfTablesInRange(
+    std::vector<Range> ranges, std::size_t* num_properties,
+    std::size_t* num_files) {
+  // run the query
+  TablePropertiesCollection props;
+  EXPECT_OK(db_->GetPropertiesOfTablesInRange(
+      db_->DefaultColumnFamily(), &ranges[0], ranges.size(), &props));
+
+  // Make sure that we've received properties for those and for those files
+  // only which fall within requested ranges
+  std::vector<LiveFileMetaData> vmd;
+  db_->GetLiveFilesMetaData(&vmd);
+  for (auto md : vmd) {
+    std::string fn = md.db_path + md.name;
+    bool in_range = false;
+    for (auto r : ranges) {
+      // smallestkey < limit && largestkey >= start
+      if (r.limit.compare(md.smallestkey) >= 0 &&
+          r.start.compare(md.largestkey) <= 0) {
+        in_range = true;
+        EXPECT_GT(props.count(fn), 0);
+      }
+    }
+    if (!in_range) {
+      EXPECT_EQ(props.count(fn), 0);
+    }
+  }
+
+  if (num_properties) {
+    *num_properties = props.size();
+  }
+
+  if (num_files) {
+    *num_files = vmd.size();
+  }
+  return props;
+}
+
+TEST_F(DBTablePropertiesTest, GetPropertiesOfTablesInRange) {
+  // Fixed random sead
+  Random rnd(301);
+
+  Options options;
+  options.create_if_missing = true;
+  options.write_buffer_size = 4096;
+  options.max_write_buffer_number = 8;
+  options.level0_file_num_compaction_trigger = 2;
+  options.level0_slowdown_writes_trigger = 2;
+  options.level0_stop_writes_trigger = 4;
+  options.target_file_size_base = 2048;
+  options.max_bytes_for_level_base = 10240;
+  options.max_bytes_for_level_multiplier = 4;
+  options.soft_rate_limit = 1.1;
+  options.num_levels = 8;
+
+  DestroyAndReopen(options);
+
+  // build a decent LSM
+  for (int i = 0; i < 10000; i++) {
+    ASSERT_OK(Put(test::RandomKey(&rnd, 5), RandomString(&rnd, 102)));
+  }
+  Flush();
+  db_->PauseBackgroundWork();
+
+  // Ensure that we have at least L0, L1 and L2
+  ASSERT_GT(NumTableFilesAtLevel(0), 0);
+  ASSERT_GT(NumTableFilesAtLevel(1), 0);
+  ASSERT_GT(NumTableFilesAtLevel(2), 0);
+
+  // Query the largest range
+  std::size_t num_properties, num_files;
+  TestGetPropertiesOfTablesInRange(
+      {Range(test::RandomKey(&rnd, 5, test::RandomKeyType::SMALLEST),
+             test::RandomKey(&rnd, 5, test::RandomKeyType::LARGEST))},
+      &num_properties, &num_files);
+  ASSERT_EQ(num_properties, num_files);
+
+  // Query the empty range
+  TestGetPropertiesOfTablesInRange(
+      {Range(test::RandomKey(&rnd, 5, test::RandomKeyType::LARGEST),
+             test::RandomKey(&rnd, 5, test::RandomKeyType::SMALLEST))},
+      &num_properties, &num_files);
+  ASSERT_GT(num_files, 0);
+  ASSERT_EQ(num_properties, 0);
+
+  // Query the middle rangee
+  TestGetPropertiesOfTablesInRange(
+      {Range(test::RandomKey(&rnd, 5, test::RandomKeyType::MIDDLE),
+             test::RandomKey(&rnd, 5, test::RandomKeyType::LARGEST))},
+      &num_properties, &num_files);
+  ASSERT_GT(num_files, 0);
+  ASSERT_GT(num_files, num_properties);
+  ASSERT_GT(num_properties, 0);
+
+  // Query a bunch of random ranges
+  for (int j = 0; j < 100; j++) {
+    // create a bunch of ranges
+    std::vector<std::string> random_keys;
+    auto n = 2 * rnd.Uniform(50);
+    for (uint32_t i = 0; i < n; ++i) {
+      random_keys.push_back(test::RandomKey(&rnd, 5));
+    }
+
+    std::vector<Range> ranges;
+    auto it = random_keys.begin();
+    while (it != random_keys.end()) {
+      ranges.push_back(Range(*it, *(it + 1)));
+      it += 2;
+    }
+
+    TestGetPropertiesOfTablesInRange(std::move(ranges));
+  }
+}
+}  // namespace rocksdb
+
+#endif  // ROCKSDB_LITE
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -10,11 +10,11 @@
 // Introduction of SyncPoint effectively disabled building and running this test
 // in Release build.
 // which is a pity, it is a good test
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)

+#include "db/db_test_util.h"
 #include "db/forward_iterator.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"

 namespace rocksdb {

@@ -646,10 +646,10 @@ TEST_F(DBTestTailingIterator, ManagedTailingIteratorSeekToSame) {

 }  // namespace rocksdb

-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif  // !defined(ROCKSDB_LITE)

 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
@@ -7,7 +7,7 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

-#include "util/db_test_util.h"
+#include "db/db_test_util.h"

 namespace rocksdb {

@@ -38,11 +38,10 @@ SpecialEnv::SpecialEnv(Env* base)
  table_write_callback_ = nullptr;
 }

-
-DBTestBase::DBTestBase(const std::string path) : option_config_(kDefault),
-           mem_env_(!getenv("MEM_ENV") ? nullptr :
-                                         new MockEnv(Env::Default())),
-           env_(new SpecialEnv(mem_env_ ? mem_env_ : Env::Default())) {
+DBTestBase::DBTestBase(const std::string path)
+    : option_config_(kDefault),
+      mem_env_(!getenv("MEM_ENV") ? nullptr : new MockEnv(Env::Default())),
+      env_(new SpecialEnv(mem_env_ ? mem_env_ : Env::Default())) {
  env_->SetBackgroundThreads(1, Env::LOW);
  env_->SetBackgroundThreads(1, Env::HIGH);
  dbname_ = test::TmpDir(env_) + path;
@@ -59,12 +58,9 @@ DBTestBase::DBTestBase(const std::string path) : option_config_(kDefault),
 }

 DBTestBase::~DBTestBase() {
-// SyncPoint is not supported in Released Windows Mode.
-#if !(defined NDEBUG) || !defined(OS_WIN)
  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
  rocksdb::SyncPoint::GetInstance()->LoadDependency({});
  rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
  Close();
  Options options;
  options.db_paths.emplace_back(dbname_, 0);
@@ -79,6 +75,23 @@ DBTestBase::~DBTestBase() {
 // test.  Return false if there are no more configurations to test.
 bool DBTestBase::ChangeOptions(int skip_mask) {
  for (option_config_++; option_config_ < kEnd; option_config_++) {
+#ifdef ROCKSDB_LITE
+    // These options are not supported in ROCKSDB_LITE
+    if (option_config_ == kHashSkipList ||
+        option_config_ == kPlainTableFirstBytePrefix ||
+        option_config_ == kPlainTableCappedPrefix ||
+        option_config_ == kPlainTableCappedPrefixNonMmap ||
+        option_config_ == kPlainTableAllBytesPrefix ||
+        option_config_ == kVectorRep || option_config_ == kHashLinkList ||
+        option_config_ == kHashCuckoo ||
+        option_config_ == kUniversalCompaction ||
+        option_config_ == kUniversalCompactionMultiLevel ||
+        option_config_ == kUniversalSubcompactions ||
+        option_config_ == kFIFOCompaction) {
+      continue;
+    }
+#endif
+
    if ((skip_mask & kSkipDeletesFilterFirst) &&
        option_config_ == kDeletesFilterFirst) {
      continue;
@@ -92,14 +105,14 @@ bool DBTestBase::ChangeOptions(int skip_mask) {
      continue;
    }
    if ((skip_mask & kSkipNoSeekToLast) &&
-        (option_config_ == kHashLinkList ||
-         option_config_ == kHashSkipList)) {;
+        (option_config_ == kHashLinkList || option_config_ == kHashSkipList)) {
      continue;
    }
    if ((skip_mask & kSkipPlainTable) &&
        (option_config_ == kPlainTableAllBytesPrefix ||
         option_config_ == kPlainTableFirstBytePrefix ||
-         option_config_ == kPlainTableCappedPrefix)) {
+         option_config_ == kPlainTableCappedPrefix ||
+         option_config_ == kPlainTableCappedPrefixNonMmap)) {
      continue;
    }
    if ((skip_mask & kSkipHashIndex) &&
@@ -114,8 +127,7 @@ bool DBTestBase::ChangeOptions(int skip_mask) {
        option_config_ == kFIFOCompaction) {
      continue;
    }
-    if ((skip_mask & kSkipMmapReads) &&
-        option_config_ == kWalDirAndMmapReads) {
+    if ((skip_mask & kSkipMmapReads) && option_config_ == kWalDirAndMmapReads) {
      continue;
    }
    break;
@@ -204,10 +216,10 @@ Options DBTestBase::CurrentOptions(
  BlockBasedTableOptions table_options;
  bool set_block_based_table_factory = true;
  switch (option_config_) {
+#ifndef ROCKSDB_LITE
    case kHashSkipList:
      options.prefix_extractor.reset(NewFixedPrefixTransform(1));
-      options.memtable_factory.reset(
-          NewHashSkipListRepFactory(16));
+      options.memtable_factory.reset(NewHashSkipListRepFactory(16));
      break;
    case kPlainTableFirstBytePrefix:
      options.table_factory.reset(new PlainTableFactory());
@@ -223,6 +235,13 @@ Options DBTestBase::CurrentOptions(
      options.max_sequential_skip_in_iterations = 999999;
      set_block_based_table_factory = false;
      break;
+    case kPlainTableCappedPrefixNonMmap:
+      options.table_factory.reset(new PlainTableFactory());
+      options.prefix_extractor.reset(NewCappedPrefixTransform(8));
+      options.allow_mmap_reads = false;
+      options.max_sequential_skip_in_iterations = 999999;
+      set_block_based_table_factory = false;
+      break;
    case kPlainTableAllBytesPrefix:
      options.table_factory.reset(new PlainTableFactory());
      options.prefix_extractor.reset(NewNoopTransform());
@@ -230,6 +249,19 @@ Options DBTestBase::CurrentOptions(
      options.max_sequential_skip_in_iterations = 999999;
      set_block_based_table_factory = false;
      break;
+    case kVectorRep:
+      options.memtable_factory.reset(new VectorRepFactory(100));
+      break;
+    case kHashLinkList:
+      options.prefix_extractor.reset(NewFixedPrefixTransform(1));
+      options.memtable_factory.reset(
+          NewHashLinkListRepFactory(4, 0, 3, true, 4));
+      break;
+    case kHashCuckoo:
+      options.memtable_factory.reset(
+          NewHashCuckooRepFactory(options.write_buffer_size));
+      break;
+#endif  // ROCKSDB_LITE
    case kMergePut:
      options.merge_operator = MergeOperators::CreatePutOperator();
      break;
@@ -266,18 +298,6 @@ Options DBTestBase::CurrentOptions(
    case kDeletesFilterFirst:
      options.filter_deletes = true;
      break;
-    case kVectorRep:
-      options.memtable_factory.reset(new VectorRepFactory(100));
-      break;
-    case kHashLinkList:
-      options.prefix_extractor.reset(NewFixedPrefixTransform(1));
-      options.memtable_factory.reset(
-          NewHashLinkListRepFactory(4, 0, 3, true, 4));
-      break;
-    case kHashCuckoo:
-      options.memtable_factory.reset(
-          NewHashCuckooRepFactory(options.write_buffer_size));
-      break;
    case kUniversalCompaction:
      options.compaction_style = kCompactionStyleUniversal;
      options.num_levels = 1;
@@ -288,7 +308,7 @@ Options DBTestBase::CurrentOptions(
      break;
    case kCompressedBlockCache:
      options.allow_mmap_writes = true;
-      table_options.block_cache_compressed = NewLRUCache(8*1024*1024);
+      table_options.block_cache_compressed = NewLRUCache(8 * 1024 * 1024);
      break;
    case kInfiniteMaxOpenFiles:
      options.max_open_files = -1;
@@ -320,6 +340,10 @@ Options DBTestBase::CurrentOptions(
      options.row_cache = NewLRUCache(1024 * 1024);
      break;
    }
+    case kRecycleLogFiles: {
+      options.recycle_log_file_num = 2;
+      break;
+    }
    case kLevelSubcompactions: {
      options.max_subcompactions = 4;
      break;
@@ -347,7 +371,7 @@ Options DBTestBase::CurrentOptions(
 }

 void DBTestBase::CreateColumnFamilies(const std::vector<std::string>& cfs,
-                          const Options& options) {
+                                      const Options& options) {
  ColumnFamilyOptions cf_opts(options);
  size_t cfi = handles_.size();
  handles_.resize(cfi + cfs.size());
@@ -357,7 +381,7 @@ void DBTestBase::CreateColumnFamilies(const std::vector<std::string>& cfs,
 }

 void DBTestBase::CreateAndReopenWithCF(const std::vector<std::string>& cfs,
-                           const Options& options) {
+                                       const Options& options) {
  CreateColumnFamilies(cfs, options);
  std::vector<std::string> cfs_plus_default = cfs;
  cfs_plus_default.insert(cfs_plus_default.begin(), kDefaultColumnFamilyName);
@@ -365,18 +389,17 @@ void DBTestBase::CreateAndReopenWithCF(const std::vector<std::string>& cfs,
 }

 void DBTestBase::ReopenWithColumnFamilies(const std::vector<std::string>& cfs,
-                              const std::vector<Options>& options) {
+                                          const std::vector<Options>& options) {
  ASSERT_OK(TryReopenWithColumnFamilies(cfs, options));
 }

 void DBTestBase::ReopenWithColumnFamilies(const std::vector<std::string>& cfs,
-                              const Options& options) {
+                                          const Options& options) {
  ASSERT_OK(TryReopenWithColumnFamilies(cfs, options));
 }

 Status DBTestBase::TryReopenWithColumnFamilies(
-    const std::vector<std::string>& cfs,
-    const std::vector<Options>& options) {
+    const std::vector<std::string>& cfs, const std::vector<Options>& options) {
  Close();
  EXPECT_EQ(cfs.size(), options.size());
  std::vector<ColumnFamilyDescriptor> column_families;
@@ -388,8 +411,7 @@ Status DBTestBase::TryReopenWithColumnFamilies(
 }

 Status DBTestBase::TryReopenWithColumnFamilies(
-    const std::vector<std::string>& cfs,
-    const Options& options) {
+    const std::vector<std::string>& cfs, const Options& options) {
  Close();
  std::vector<Options> v_opts(cfs.size(), options);
  return TryReopenWithColumnFamilies(cfs, v_opts);
@@ -446,7 +468,7 @@ Status DBTestBase::Put(const Slice& k, const Slice& v, WriteOptions wo) {
 }

 Status DBTestBase::Put(int cf, const Slice& k, const Slice& v,
-           WriteOptions wo) {
+                       WriteOptions wo) {
  if (kMergePut == option_config_) {
    return db_->Merge(wo, handles_[cf], k, v);
  } else {
@@ -462,6 +484,14 @@ Status DBTestBase::Delete(int cf, const std::string& k) {
  return db_->Delete(WriteOptions(), handles_[cf], k);
 }

+Status DBTestBase::SingleDelete(const std::string& k) {
+  return db_->SingleDelete(WriteOptions(), k);
+}
+
+Status DBTestBase::SingleDelete(int cf, const std::string& k) {
+  return db_->SingleDelete(WriteOptions(), handles_[cf], k);
+}
+
 std::string DBTestBase::Get(const std::string& k, const Snapshot* snapshot) {
  ReadOptions options;
  options.verify_checksums = true;
@@ -477,7 +507,7 @@ std::string DBTestBase::Get(const std::string& k, const Snapshot* snapshot) {
 }

 std::string DBTestBase::Get(int cf, const std::string& k,
-                const Snapshot* snapshot) {
+                            const Snapshot* snapshot) {
  ReadOptions options;
  options.verify_checksums = true;
  options.snapshot = snapshot;
@@ -536,9 +566,9 @@ std::string DBTestBase::AllEntriesFor(const Slice& user_key, int cf) {
  Arena arena;
  ScopedArenaIterator iter;
  if (cf == 0) {
-    iter.set(dbfull()->TEST_NewInternalIterator(&arena));
+    iter.set(dbfull()->NewInternalIterator(&arena));
  } else {
-    iter.set(dbfull()->TEST_NewInternalIterator(&arena, handles_[cf]));
+    iter.set(dbfull()->NewInternalIterator(&arena, handles_[cf]));
  }
  InternalKey target(user_key, kMaxSequenceNumber, kTypeValue);
  iter->Seek(target.Encode());
@@ -571,6 +601,9 @@ std::string DBTestBase::AllEntriesFor(const Slice& user_key, int cf) {
          case kTypeDeletion:
            result += "DEL";
            break;
+          case kTypeSingleDeletion:
+            result += "SDEL";
+            break;
          default:
            assert(false);
            break;
@@ -586,6 +619,7 @@ std::string DBTestBase::AllEntriesFor(const Slice& user_key, int cf) {
  return result;
 }

+#ifndef ROCKSDB_LITE
 int DBTestBase::NumSortedRuns(int cf) {
  ColumnFamilyMetaData cf_meta;
  if (cf == 0) {
@@ -612,20 +646,6 @@ uint64_t DBTestBase::TotalSize(int cf) {
  return cf_meta.size;
 }

-int DBTestBase::NumTableFilesAtLevel(int level, int cf) {
-  std::string property;
-  if (cf == 0) {
-    // default cfd
-    EXPECT_TRUE(db_->GetProperty(
-        "rocksdb.num-files-at-level" + NumberToString(level), &property));
-  } else {
-    EXPECT_TRUE(db_->GetProperty(
-        handles_[cf], "rocksdb.num-files-at-level" + NumberToString(level),
-        &property));
-  }
-  return atoi(property.c_str());
-}
-
 uint64_t DBTestBase::SizeAtLevel(int level) {
  std::vector<LiveFileMetaData> metadata;
  db_->GetLiveFilesMetaData(&metadata);
@@ -652,6 +672,27 @@ int DBTestBase::TotalLiveFiles(int cf) {
  return num_files;
 }

+size_t DBTestBase::CountLiveFiles() {
+  std::vector<LiveFileMetaData> metadata;
+  db_->GetLiveFilesMetaData(&metadata);
+  return metadata.size();
+}
+#endif  // ROCKSDB_LITE
+
+int DBTestBase::NumTableFilesAtLevel(int level, int cf) {
+  std::string property;
+  if (cf == 0) {
+    // default cfd
+    EXPECT_TRUE(db_->GetProperty(
+        "rocksdb.num-files-at-level" + NumberToString(level), &property));
+  } else {
+    EXPECT_TRUE(db_->GetProperty(
+        handles_[cf], "rocksdb.num-files-at-level" + NumberToString(level),
+        &property));
+  }
+  return atoi(property.c_str());
+}
+
 int DBTestBase::TotalTableFiles(int cf, int levels) {
  if (levels == -1) {
    levels = CurrentOptions().num_levels;
@@ -694,12 +735,6 @@ size_t DBTestBase::CountFiles() {
  return files.size() + logfiles.size();
 }

-size_t DBTestBase::CountLiveFiles() {
-  std::vector<LiveFileMetaData> metadata;
-  db_->GetLiveFilesMetaData(&metadata);
-  return metadata.size();
-}
-
 uint64_t DBTestBase::Size(const Slice& start, const Slice& limit, int cf) {
  Range r(start, limit);
  uint64_t size;
@@ -712,7 +747,7 @@ uint64_t DBTestBase::Size(const Slice& start, const Slice& limit, int cf) {
 }

 void DBTestBase::Compact(int cf, const Slice& start, const Slice& limit,
-             uint32_t target_path_id) {
+                         uint32_t target_path_id) {
  CompactRangeOptions compact_options;
  compact_options.target_path_id = target_path_id;
  ASSERT_OK(db_->CompactRange(compact_options, handles_[cf], &start, &limit));
@@ -729,9 +764,8 @@ void DBTestBase::Compact(const Slice& start, const Slice& limit) {

 // Do n memtable compactions, each of which produces an sstable
 // covering the range [small,large].
-void DBTestBase::MakeTables(
-    int n, const std::string& small,
-    const std::string& large, int cf) {
+void DBTestBase::MakeTables(int n, const std::string& small,
+                            const std::string& large, int cf) {
  for (int i = 0; i < n; i++) {
    ASSERT_OK(Put(cf, small, "begin"));
    ASSERT_OK(Put(cf, large, "end"));
@@ -742,8 +776,8 @@ void DBTestBase::MakeTables(

 // Prevent pushing of new sstables into deeper levels by adding
 // tables that cover a specified range to all levels.
-void DBTestBase::FillLevels(
-    const std::string& smallest, const std::string& largest, int cf) {
+void DBTestBase::FillLevels(const std::string& smallest,
+                            const std::string& largest, int cf) {
  MakeTables(db_->NumberLevels(handles_[cf]), smallest, largest, cf);
 }

@@ -760,7 +794,7 @@ void DBTestBase::MoveFilesToLevel(int level, int cf) {
 void DBTestBase::DumpFileCounts(const char* label) {
  fprintf(stderr, "---\n%s:\n", label);
  fprintf(stderr, "maxoverlap: %" PRIu64 "\n",
-      dbfull()->TEST_MaxNextLevelOverlappingBytes());
+          dbfull()->TEST_MaxNextLevelOverlappingBytes());
  for (int level = 0; level < db_->NumberLevels(); level++) {
    int num = NumTableFilesAtLevel(level);
    if (num > 0) {
@@ -775,19 +809,22 @@ std::string DBTestBase::DumpSSTableList() {
  return property;
 }

+void DBTestBase::GetSstFiles(std::string path,
+                             std::vector<std::string>* files) {
+  env_->GetChildren(path, files);
+
+  files->erase(
+      std::remove_if(files->begin(), files->end(), [](std::string name) {
+        uint64_t number;
+        FileType type;
+        return !(ParseFileName(name, &number, &type) && type == kTableFile);
+      }), files->end());
+}
+
 int DBTestBase::GetSstFileCount(std::string path) {
  std::vector<std::string> files;
-  env_->GetChildren(path, &files);
-
-  int sst_count = 0;
-  uint64_t number;
-  FileType type;
-  for (size_t i = 0; i < files.size(); i++) {
-    if (ParseFileName(files[i], &number, &type) && type == kTableFile) {
-      sst_count++;
-    }
-  }
-  return sst_count;
+  GetSstFiles(path, &files);
+  return static_cast<int>(files.size());
 }

 // this will generate non-overlapping files since it keeps increasing key_idx
@@ -866,9 +903,10 @@ void DBTestBase::VerifyIterLast(std::string expected_key, int cf) {
 //   sets newValue with delta
 // If previous value is not empty,
 //   updates previous value with 'b' string of previous value size - 1.
-UpdateStatus DBTestBase::updateInPlaceSmallerSize(
-    char* prevValue, uint32_t* prevSize,
-    Slice delta, std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceSmallerSize(char* prevValue,
+                                                  uint32_t* prevSize,
+                                                  Slice delta,
+                                                  std::string* newValue) {
  if (prevValue == nullptr) {
    *newValue = std::string(delta.size(), 'c');
    return UpdateStatus::UPDATED;
@@ -880,9 +918,10 @@ UpdateStatus DBTestBase::updateInPlaceSmallerSize(
  }
 }

-UpdateStatus DBTestBase::updateInPlaceSmallerVarintSize(
-    char* prevValue, uint32_t* prevSize,
-    Slice delta, std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceSmallerVarintSize(char* prevValue,
+                                                        uint32_t* prevSize,
+                                                        Slice delta,
+                                                        std::string* newValue) {
  if (prevValue == nullptr) {
    *newValue = std::string(delta.size(), 'c');
    return UpdateStatus::UPDATED;
@@ -894,16 +933,17 @@ UpdateStatus DBTestBase::updateInPlaceSmallerVarintSize(
  }
 }

-UpdateStatus DBTestBase::updateInPlaceLargerSize(
-    char* prevValue, uint32_t* prevSize,
-    Slice delta, std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceLargerSize(char* prevValue,
+                                                 uint32_t* prevSize,
+                                                 Slice delta,
+                                                 std::string* newValue) {
  *newValue = std::string(delta.size(), 'c');
  return UpdateStatus::UPDATED;
 }

-UpdateStatus DBTestBase::updateInPlaceNoAction(
-    char* prevValue, uint32_t* prevSize,
-    Slice delta, std::string* newValue) {
+UpdateStatus DBTestBase::updateInPlaceNoAction(char* prevValue,
+                                               uint32_t* prevSize, Slice delta,
+                                               std::string* newValue) {
  return UpdateStatus::UPDATE_FAILED;
 }

@@ -912,9 +952,9 @@ void DBTestBase::validateNumberOfEntries(int numValues, int cf) {
  ScopedArenaIterator iter;
  Arena arena;
  if (cf != 0) {
-    iter.set(dbfull()->TEST_NewInternalIterator(&arena, handles_[cf]));
+    iter.set(dbfull()->NewInternalIterator(&arena, handles_[cf]));
  } else {
-    iter.set(dbfull()->TEST_NewInternalIterator(&arena));
+    iter.set(dbfull()->NewInternalIterator(&arena));
  }
  iter->SeekToFirst();
  ASSERT_EQ(iter->status().ok(), true);
@@ -931,9 +971,8 @@ void DBTestBase::validateNumberOfEntries(int numValues, int cf) {
  ASSERT_EQ(0, seq);
 }

-void DBTestBase::CopyFile(
-    const std::string& source, const std::string& destination,
-    uint64_t size) {
+void DBTestBase::CopyFile(const std::string& source,
+                          const std::string& destination, uint64_t size) {
  const EnvOptions soptions;
  unique_ptr<SequentialFile> srcfile;
  ASSERT_OK(env_->NewSequentialFile(source, &srcfile, soptions));
@@ -27,8 +27,10 @@
 #include <vector>

 #include "db/db_impl.h"
+#include "db/db_test_util.h"
 #include "db/dbformat.h"
 #include "db/filename.h"
+#include "memtable/hash_linklist_rep.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/convenience.h"
@@ -42,12 +44,11 @@
 #include "table/block_based_table_factory.h"
 #include "table/mock_table.h"
 #include "table/plain_table_factory.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/compression.h"
-#include "util/db_test_util.h"
-#include "util/hash_linklist_rep.h"
 #include "util/mock_env.h"
 #include "util/mutexlock.h"
-#include "util/scoped_arena_iterator.h"
+
 #include "util/string_util.h"
 // SyncPoint is not supported in Released Windows Mode.
 #if !(defined NDEBUG) || !defined(OS_WIN)
@@ -83,8 +84,8 @@ class AtomicCounter {
    uint64_t start = env_->NowMicros();
    while (count_ < count) {
      uint64_t now = env_->NowMicros();
-      cond_count_.TimedWait(now + /*1s*/ 1 * 000 * 000);
-      if (env_->NowMicros() - start > /*10s*/ 10 * 000 * 000) {
+      cond_count_.TimedWait(now + /*1s*/ 1 * 1000 * 1000);
+      if (env_->NowMicros() - start > /*10s*/ 10 * 1000 * 1000) {
        return false;
      }
      if (count_ < count) {
@@ -131,9 +132,7 @@ class SpecialEnv : public EnvWrapper {

     public:
      SSTableFile(SpecialEnv* env, unique_ptr<WritableFile>&& base)
-          : env_(env),
-            base_(std::move(base)) {
-      }
+          : env_(env), base_(std::move(base)) {}
      Status Append(const Slice& data) override {
        if (env_->table_write_callback_) {
          (*env_->table_write_callback_)();
@@ -148,6 +147,7 @@ class SpecialEnv : public EnvWrapper {
          return base_->Append(data);
        }
      }
+      Status Truncate(uint64_t size) override { return base_->Truncate(size); }
      Status Close() override {
 // SyncPoint is not supported in Released Windows Mode.
 #if !(defined NDEBUG) || !defined(OS_WIN)
@@ -177,7 +177,7 @@ class SpecialEnv : public EnvWrapper {
    class ManifestFile : public WritableFile {
     public:
      ManifestFile(SpecialEnv* env, unique_ptr<WritableFile>&& b)
-          : env_(env), base_(std::move(b)) { }
+          : env_(env), base_(std::move(b)) {}
      Status Append(const Slice& data) override {
        if (env_->manifest_write_error_.load(std::memory_order_acquire)) {
          return Status::IOError("simulated writer error");
@@ -185,6 +185,7 @@ class SpecialEnv : public EnvWrapper {
          return base_->Append(data);
        }
      }
+      Status Truncate(uint64_t size) override { return base_->Truncate(size); }
      Status Close() override { return base_->Close(); }
      Status Flush() override { return base_->Flush(); }
      Status Sync() override {
@@ -225,6 +226,7 @@ class SpecialEnv : public EnvWrapper {
 #endif
        return s;
      }
+      Status Truncate(uint64_t size) override { return base_->Truncate(size); }
      Status Close() override { return base_->Close(); }
      Status Flush() override { return base_->Flush(); }
      Status Sync() override {
@@ -278,8 +280,7 @@ class SpecialEnv : public EnvWrapper {
     public:
      CountingFile(unique_ptr<RandomAccessFile>&& target,
                   anon::AtomicCounter* counter)
-          : target_(std::move(target)), counter_(counter) {
-      }
+          : target_(std::move(target)), counter_(counter) {}
      virtual Status Read(uint64_t offset, size_t n, Slice* result,
                          char* scratch) const override {
        counter_->Increment();
@@ -324,7 +325,6 @@ class SpecialEnv : public EnvWrapper {
    return s;
  }

-
  virtual void SleepForMicroseconds(int micros) override {
    sleep_counter_.Increment();
    if (no_sleep_) {
@@ -401,7 +401,7 @@ class SpecialEnv : public EnvWrapper {
  std::atomic<int64_t> addon_time_;
  bool no_sleep_;

-  std::atomic<bool> is_wal_sync_thread_safe_ {true};
+  std::atomic<bool> is_wal_sync_thread_safe_{true};
 };

 class DBTestBase : public testing::Test {
@@ -413,32 +413,34 @@ class DBTestBase : public testing::Test {
    kBlockBasedTableWithWholeKeyHashIndex = 2,
    kPlainTableFirstBytePrefix = 3,
    kPlainTableCappedPrefix = 4,
-    kPlainTableAllBytesPrefix = 5,
-    kVectorRep = 6,
-    kHashLinkList = 7,
-    kHashCuckoo = 8,
-    kMergePut = 9,
-    kFilter = 10,
-    kFullFilterWithNewTableReaderForCompactions = 11,
-    kUncompressed = 12,
-    kNumLevel_3 = 13,
-    kDBLogDir = 14,
-    kWalDirAndMmapReads = 15,
-    kManifestFileSize = 16,
-    kPerfOptions = 17,
-    kDeletesFilterFirst = 18,
-    kHashSkipList = 19,
-    kUniversalCompaction = 20,
-    kUniversalCompactionMultiLevel = 21,
-    kCompressedBlockCache = 22,
-    kInfiniteMaxOpenFiles = 23,
-    kxxHashChecksum = 24,
-    kFIFOCompaction = 25,
-    kOptimizeFiltersForHits = 26,
-    kRowCache = 27,
-    kLevelSubcompactions = 28,
-    kUniversalSubcompactions = 29,
-    kEnd = 28
+    kPlainTableCappedPrefixNonMmap = 5,
+    kPlainTableAllBytesPrefix = 6,
+    kVectorRep = 7,
+    kHashLinkList = 8,
+    kHashCuckoo = 9,
+    kMergePut = 10,
+    kFilter = 11,
+    kFullFilterWithNewTableReaderForCompactions = 12,
+    kUncompressed = 13,
+    kNumLevel_3 = 14,
+    kDBLogDir = 15,
+    kWalDirAndMmapReads = 16,
+    kManifestFileSize = 17,
+    kPerfOptions = 18,
+    kDeletesFilterFirst = 19,
+    kHashSkipList = 20,
+    kUniversalCompaction = 21,
+    kUniversalCompactionMultiLevel = 22,
+    kCompressedBlockCache = 23,
+    kInfiniteMaxOpenFiles = 24,
+    kxxHashChecksum = 25,
+    kFIFOCompaction = 26,
+    kOptimizeFiltersForHits = 27,
+    kRowCache = 28,
+    kRecycleLogFiles = 29,
+    kLevelSubcompactions = 30,
+    kUniversalSubcompactions = 31,
+    kEnd = 30
  };
  int option_config_;

@@ -503,9 +505,7 @@ class DBTestBase : public testing::Test {
      const Options& defaultOptions,
      const anon::OptionsOverride& options_override = anon::OptionsOverride());

-  DBImpl* dbfull() {
-    return reinterpret_cast<DBImpl*>(db_);
-  }
+  DBImpl* dbfull() { return reinterpret_cast<DBImpl*>(db_); }

  void CreateColumnFamilies(const std::vector<std::string>& cfs,
                            const Options& options);
@@ -519,9 +519,8 @@ class DBTestBase : public testing::Test {
  void ReopenWithColumnFamilies(const std::vector<std::string>& cfs,
                                const Options& options);

-  Status TryReopenWithColumnFamilies(
-      const std::vector<std::string>& cfs,
-      const std::vector<Options>& options);
+  Status TryReopenWithColumnFamilies(const std::vector<std::string>& cfs,
+                                     const std::vector<Options>& options);

  Status TryReopenWithColumnFamilies(const std::vector<std::string>& cfs,
                                     const Options& options);
@@ -549,6 +548,10 @@ class DBTestBase : public testing::Test {

  Status Delete(int cf, const std::string& k);

+  Status SingleDelete(const std::string& k);
+
+  Status SingleDelete(int cf, const std::string& k);
+
  std::string Get(const std::string& k, const Snapshot* snapshot = nullptr);

  std::string Get(int cf, const std::string& k,
@@ -564,16 +567,20 @@ class DBTestBase : public testing::Test {

  std::string AllEntriesFor(const Slice& user_key, int cf = 0);

+#ifndef ROCKSDB_LITE
  int NumSortedRuns(int cf = 0);

  uint64_t TotalSize(int cf = 0);

-  int NumTableFilesAtLevel(int level, int cf = 0);
-
  uint64_t SizeAtLevel(int level);

  int TotalLiveFiles(int cf = 0);

+  size_t CountLiveFiles();
+#endif  // ROCKSDB_LITE
+
+  int NumTableFilesAtLevel(int level, int cf = 0);
+
  int TotalTableFiles(int cf = 0, int levels = -1);

  // Return spread of files per level
@@ -581,8 +588,6 @@ class DBTestBase : public testing::Test {

  size_t CountFiles();

-  size_t CountLiveFiles();
-
  uint64_t Size(const Slice& start, const Slice& limit, int cf = 0);

  void Compact(int cf, const Slice& start, const Slice& limit,
@@ -608,6 +613,8 @@ class DBTestBase : public testing::Test {

  std::string DumpSSTableList();

+  void GetSstFiles(std::string path, std::vector<std::string>* files);
+
  int GetSstFileCount(std::string path);

  // this will generate non-overlapping files since it keeps increasing key_idx
@@ -631,21 +638,21 @@ class DBTestBase : public testing::Test {
  //   sets newValue with delta
  // If previous value is not empty,
  //   updates previous value with 'b' string of previous value size - 1.
-  static UpdateStatus updateInPlaceSmallerSize(
-      char* prevValue, uint32_t* prevSize,
-      Slice delta, std::string* newValue);
+  static UpdateStatus updateInPlaceSmallerSize(char* prevValue,
+                                               uint32_t* prevSize, Slice delta,
+                                               std::string* newValue);

-  static UpdateStatus updateInPlaceSmallerVarintSize(
-      char* prevValue, uint32_t* prevSize,
-      Slice delta, std::string* newValue);
+  static UpdateStatus updateInPlaceSmallerVarintSize(char* prevValue,
+                                                     uint32_t* prevSize,
+                                                     Slice delta,
+                                                     std::string* newValue);

-  static UpdateStatus updateInPlaceLargerSize(
-      char* prevValue, uint32_t* prevSize,
-      Slice delta, std::string* newValue);
+  static UpdateStatus updateInPlaceLargerSize(char* prevValue,
+                                              uint32_t* prevSize, Slice delta,
+                                              std::string* newValue);

-  static UpdateStatus updateInPlaceNoAction(
-      char* prevValue, uint32_t* prevSize,
-      Slice delta, std::string* newValue);
+  static UpdateStatus updateInPlaceNoAction(char* prevValue, uint32_t* prevSize,
+                                            Slice delta, std::string* newValue);

  // Utility method to test InplaceUpdate
  void validateNumberOfEntries(int numValues, int cf = 0);
@@ -7,9 +7,9 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
 #include "util/sync_point.h"

 namespace rocksdb {
@@ -1210,10 +1210,10 @@ INSTANTIATE_TEST_CASE_P(DBTestUniversalManualCompactionOutputPathId,

 }  // namespace rocksdb

-#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+#endif  // !defined(ROCKSDB_LITE)

 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
+#if !defined(ROCKSDB_LITE)
  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
@@ -7,11 +7,9 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

+#include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/db_test_util.h"
-#if !(defined NDEBUG) || !defined(OS_WIN)
 #include "util/sync_point.h"
-#endif

 namespace rocksdb {
 class DBWALTest : public DBTestBase {
@@ -70,7 +68,6 @@ TEST_F(DBWALTest, RollLog) {
  } while (ChangeOptions());
 }

-#if !(defined NDEBUG) || !defined(OS_WIN)
 TEST_F(DBWALTest, SyncWALNotBlockWrite) {
  Options options = CurrentOptions();
  options.max_write_buffer_number = 4;
@@ -130,15 +127,10 @@ TEST_F(DBWALTest, SyncWALNotWaitWrite) {
  ASSERT_EQ(Get("foo2"), "bar2");
  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
-#endif
 }  // namespace rocksdb

 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
-#else
-  return 0;
-#endif
 }
@@ -22,7 +22,7 @@ namespace rocksdb {

 uint64_t PackSequenceAndType(uint64_t seq, ValueType t) {
  assert(seq <= kMaxSequenceNumber);
-  assert(t <= kValueTypeForSeek);
+  assert(IsValueType(t));
  return (seq << 8) | t;
 }

@@ -31,7 +31,7 @@ void UnPackSequenceAndType(uint64_t packed, uint64_t* seq, ValueType* t) {
  *t = static_cast<ValueType>(packed & 0xff);

  assert(*seq <= kMaxSequenceNumber);
-  assert(*t <= kValueTypeForSeek);
+  assert(IsValueType(*t));
 }

 void AppendInternalKey(std::string* result, const ParsedInternalKey& key) {
@@ -33,13 +33,13 @@ enum ValueType : unsigned char {
  kTypeDeletion = 0x0,
  kTypeValue = 0x1,
  kTypeMerge = 0x2,
-  // Following types are used only in write ahead logs. They are not used in
-  // memtables or sst files:
-  kTypeLogData = 0x3,
-  kTypeColumnFamilyDeletion = 0x4,
-  kTypeColumnFamilyValue = 0x5,
-  kTypeColumnFamilyMerge = 0x6,
-  kMaxValue = 0x7F
+  kTypeLogData = 0x3,               // WAL only.
+  kTypeColumnFamilyDeletion = 0x4,  // WAL only.
+  kTypeColumnFamilyValue = 0x5,     // WAL only.
+  kTypeColumnFamilyMerge = 0x6,     // WAL only.
+  kTypeSingleDeletion = 0x7,
+  kTypeColumnFamilySingleDeletion = 0x8,  // WAL only.
+  kMaxValue = 0x7F                        // Not used for storing records.
 };

 // kValueTypeForSeek defines the ValueType that should be passed when
@@ -48,7 +48,13 @@ enum ValueType : unsigned char {
 // and the value type is embedded as the low 8 bits in the sequence
 // number in internal keys, we need to use the highest-numbered
 // ValueType, not the lowest).
-static const ValueType kValueTypeForSeek = kTypeMerge;
+static const ValueType kValueTypeForSeek = kTypeSingleDeletion;
+
+// Checks whether a type is a value type (i.e. a type used in memtables and sst
+// files).
+inline bool IsValueType(ValueType t) {
+  return t <= kTypeMerge || t == kTypeSingleDeletion;
+}

 // We leave eight bits empty at the bottom so a type and sequence#
 // can be packed together into 64-bits.
@@ -193,13 +199,12 @@ inline bool ParseInternalKey(const Slice& internal_key,
  result->type = static_cast<ValueType>(c);
  assert(result->type <= ValueType::kMaxValue);
  result->user_key = Slice(internal_key.data(), n - 8);
-  return (c <= static_cast<unsigned char>(kValueTypeForSeek));
+  return IsValueType(result->type);
 }

 // Update the sequence number in the internal key.
 // Guarantees not to invalidate ikey.data().
-inline void UpdateInternalKey(std::string* ikey,
-                              uint64_t seq, ValueType t) {
+inline void UpdateInternalKey(std::string* ikey, uint64_t seq, ValueType t) {
  size_t ikey_sz = ikey->size();
  assert(ikey_sz >= 8);
  uint64_t newval = (seq << 8) | t;
@@ -272,7 +277,12 @@ class IterKey {

  Slice GetKey() const { return Slice(key_, key_size_); }

-  size_t Size() { return key_size_; }
+  Slice GetUserKey() const {
+    assert(key_size_ >= 8);
+    return Slice(key_, key_size_ - 8);
+  }
+
+  size_t Size() const { return key_size_; }

  void Clear() { key_size_ = 0; }

@@ -292,7 +302,7 @@ class IterKey {
      char* p = new char[total_size];
      memcpy(p, key_, shared_len);

-      if (key_ != nullptr && key_ != space_) {
+      if (key_ != space_) {
        delete[] key_;
      }

@@ -304,11 +314,30 @@ class IterKey {
    memcpy(key_ + shared_len, non_shared_data, non_shared_len);
  }

-  void SetKey(const Slice& key) {
+  Slice SetKey(const Slice& key) {
    size_t size = key.size();
    EnlargeBufferIfNeeded(size);
    memcpy(key_, key.data(), size);
    key_size_ = size;
+    return Slice(key_, key_size_);
+  }
+
+  // Copies the content of key, updates the reference to the user key in ikey
+  // and returns a Slice referencing the new copy.
+  Slice SetKey(const Slice& key, ParsedInternalKey* ikey) {
+    size_t key_n = key.size();
+    assert(key_n >= 8);
+    SetKey(key);
+    ikey->user_key = Slice(key_, key_n - 8);
+    return Slice(key_, key_n);
+  }
+
+  // Update the sequence number in the internal key.  Guarantees not to
+  // invalidate slices to the key (and the user key).
+  void UpdateInternalKey(uint64_t seq, ValueType t) {
+    assert(key_size_ >= 8);
+    uint64_t newval = (seq << 8) | t;
+    EncodeFixed64(&key_[key_size_ - 8], newval);
  }

  void SetInternalKey(const Slice& key_prefix, const Slice& user_key,
@@ -359,10 +388,10 @@ class IterKey {
  char space_[32];  // Avoid allocation for short keys

  void ResetBuffer() {
-    if (key_ != nullptr && key_ != space_) {
+    if (key_ != space_) {
      delete[] key_;
+      key_ = space_;
    }
-    key_ = space_;
    buf_size_ = sizeof(space_);
    key_size_ = 0;
  }
@@ -53,7 +53,7 @@ void EventHelpers::LogAndNotifyTableFileCreation(
                info.table_properties.filter_policy_name;

    // user collected properties
-    for (const auto& prop : info.table_properties.user_collected_properties) {
+    for (const auto& prop : info.table_properties.readable_properties) {
      jwriter << prop.first << prop.second;
    }
    jwriter.EndObject();
@@ -11,8 +11,6 @@
 // the last "sync". It then checks for data loss errors by purposely dropping
 // file data (or entire files) not protected by a "sync".

-#if !(defined NDEBUG) || !defined(OS_WIN)
-
 #include <map>
 #include <set>
 #include "db/db_impl.h"
@@ -141,6 +139,7 @@ class TestWritableFile : public WritableFile {
                            FaultInjectionTestEnv* env);
  virtual ~TestWritableFile();
  virtual Status Append(const Slice& data) override;
+  virtual Status Truncate(uint64_t size) override { return target_->Truncate(size); }
  virtual Status Close() override;
  virtual Status Flush() override;
  virtual Status Sync() override;
@@ -778,44 +777,12 @@ TEST_P(FaultInjectionTest, FaultTest) {
  } while (ChangeOptions());
 }

-class SleepingBackgroundTask {
- public:
-  SleepingBackgroundTask()
-      : bg_cv_(&mutex_), should_sleep_(true), done_with_sleep_(false) {}
-  void DoSleep() {
-    MutexLock l(&mutex_);
-    while (should_sleep_) {
-      bg_cv_.Wait();
-    }
-    done_with_sleep_ = true;
-    bg_cv_.SignalAll();
-  }
-  void WakeUp() {
-    MutexLock l(&mutex_);
-    should_sleep_ = false;
-    bg_cv_.SignalAll();
-    while (!done_with_sleep_) {
-      bg_cv_.Wait();
-    }
-  }
-
-  static void DoSleepTask(void* arg) {
-    reinterpret_cast<SleepingBackgroundTask*>(arg)->DoSleep();
-  }
-
- private:
-  port::Mutex mutex_;
-  port::CondVar bg_cv_;  // Signalled when background work finishes
-  bool should_sleep_;
-  bool done_with_sleep_;
-};
-
 // Previous log file is not fsynced if sync is forced after log rolling.
 TEST_P(FaultInjectionTest, WriteOptionSyncTest) {
-  SleepingBackgroundTask sleeping_task_low;
+  test::SleepingBackgroundTask sleeping_task_low;
  env_->SetBackgroundThreads(1, Env::HIGH);
  // Block the job queue to prevent flush job from running.
-  env_->Schedule(&SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
                 Env::Priority::HIGH);

  WriteOptions write_options;
@@ -895,10 +862,10 @@ TEST_P(FaultInjectionTest, UninstalledCompaction) {
 }

 TEST_P(FaultInjectionTest, ManualLogSyncTest) {
-  SleepingBackgroundTask sleeping_task_low;
+  test::SleepingBackgroundTask sleeping_task_low;
  env_->SetBackgroundThreads(1, Env::HIGH);
  // Block the job queue to prevent flush job from running.
-  env_->Schedule(&SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
                 Env::Priority::HIGH);

  WriteOptions write_options;
@@ -933,13 +900,7 @@ INSTANTIATE_TEST_CASE_P(FaultTest, FaultInjectionTest, ::testing::Bool());

 }  // namespace rocksdb

-#endif // #if !(defined NDEBUG) || !defined(OS_WIN)
-
 int main(int argc, char** argv) {
-#if !(defined NDEBUG) || !defined(OS_WIN)
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
-#else
-  return 0;
-#endif
 }
@@ -21,9 +21,13 @@
 #include "util/file_reader_writer.h"
 #include "util/logging.h"
 #include "util/stop_watch.h"
+#include "util/sync_point.h"

 namespace rocksdb {

+static const std::string kRocksDbTFileExt = "sst";
+static const std::string kLevelDbTFileExt = "ldb";
+
 // Given a path, flatten the path name by replacing all chars not in
 // {[0-9,a-z,A-Z,-,_,.]} with _. And append '_LOG\0' at the end.
 // Return the number of chars stored in dest not including the trailing '\0'.
@@ -78,7 +82,16 @@ std::string ArchivedLogFileName(const std::string& name, uint64_t number) {
 }

 std::string MakeTableFileName(const std::string& path, uint64_t number) {
-  return MakeFileName(path, number, "sst");
+  return MakeFileName(path, number, kRocksDbTFileExt.c_str());
+}
+
+std::string Rocks2LevelTableFileName(const std::string& fullname) {
+  assert(fullname.size() > kRocksDbTFileExt.size() + 1);
+  if (fullname.size() <= kRocksDbTFileExt.size() + 1) {
+    return "";
+  }
+  return fullname.substr(0, fullname.size() - kRocksDbTFileExt.size()) +
+         kLevelDbTFileExt;
 }

 uint64_t TableFileNameToNumber(const std::string& name) {
@@ -151,8 +164,9 @@ InfoLogPrefix::InfoLogPrefix(bool has_log_dir,

 std::string InfoLogFileName(const std::string& dbname,
    const std::string& db_path, const std::string& log_dir) {
-  if (log_dir.empty())
+  if (log_dir.empty()) {
    return dbname + "/LOG";
+  }

  InfoLogPrefix info_log_prefix(true, db_path);
  return log_dir + "/" + info_log_prefix.buf;
@@ -164,8 +178,9 @@ std::string OldInfoLogFileName(const std::string& dbname, uint64_t ts,
  char buf[50];
  snprintf(buf, sizeof(buf), "%llu", static_cast<unsigned long long>(ts));

-  if (log_dir.empty())
+  if (log_dir.empty()) {
    return dbname + "/LOG.old." + buf;
+  }

  InfoLogPrefix info_log_prefix(true, db_path);
  return log_dir + "/" + info_log_prefix.buf + ".old." + buf;
@@ -271,17 +286,23 @@ bool ParseFileName(const std::string& fname, uint64_t* number,
    if (!ConsumeDecimalNumber(&rest, &num)) {
      return false;
    }
+    if (rest.size() <= 1 || rest[0] != '.') {
+      return false;
+    }
+    rest.remove_prefix(1);
+
    Slice suffix = rest;
-    if (suffix == Slice(".log")) {
+    if (suffix == Slice("log")) {
      *type = kLogFile;
      if (log_type && !archive_dir_found) {
        *log_type = kAliveLogFile;
      }
    } else if (archive_dir_found) {
      return false; // Archive dir can contain only log files
-    } else if (suffix == Slice(".sst")) {
+    } else if (suffix == Slice(kRocksDbTFileExt) ||
+               suffix == Slice(kLevelDbTFileExt)) {
      *type = kTableFile;
-    } else if (suffix == Slice(".dbtmp")) {
+    } else if (suffix == Slice("dbtmp")) {
      *type = kTempFile;
    } else {
      return false;
@@ -302,7 +323,9 @@ Status SetCurrentFile(Env* env, const std::string& dbname,
  std::string tmp = TempFileName(dbname, descriptor_number);
  Status s = WriteStringToFile(env, contents.ToString() + "\n", tmp, true);
  if (s.ok()) {
+    TEST_KILL_RANDOM("SetCurrentFile:0", rocksdb_kill_odds * REDUCE_ODDS2);
    s = env->RenameFile(tmp, CurrentFileName(dbname));
+    TEST_KILL_RANDOM("SetCurrentFile:1", rocksdb_kill_odds * REDUCE_ODDS2);
  }
  if (s.ok()) {
    if (directory_to_fsync != nullptr) {
@@ -331,6 +354,7 @@ Status SetIdentityFile(Env* env, const std::string& dbname) {

 Status SyncManifest(Env* env, const DBOptions* db_options,
                    WritableFileWriter* file) {
+  TEST_KILL_RANDOM("SyncManifest:0", rocksdb_kill_odds * REDUCE_ODDS2);
  if (db_options->disableDataSync) {
    return Status::OK();
  } else {
@@ -55,6 +55,10 @@ extern std::string ArchivedLogFileName(const std::string& dbname,

 extern std::string MakeTableFileName(const std::string& name, uint64_t number);

+// Return the name of sstable with LevelDB suffix
+// created from RocksDB sstable suffixed name
+extern std::string Rocks2LevelTableFileName(const std::string& fullname);
+
 // the reverse function of MakeTableFileName
 // TODO(yhchiang): could merge this function with ParseFileName()
 extern uint64_t TableFileNameToNumber(const std::string& name);
@@ -102,11 +106,13 @@ struct InfoLogPrefix {

 // Return the name of the info log file for "dbname".
 extern std::string InfoLogFileName(const std::string& dbname,
-    const std::string& db_path="", const std::string& log_dir="");
+                                   const std::string& db_path = "",
+                                   const std::string& log_dir = "");

 // Return the name of the old info log file for "dbname".
 extern std::string OldInfoLogFileName(const std::string& dbname, uint64_t ts,
-    const std::string& db_path="", const std::string& log_dir="");
+                                      const std::string& db_path = "",
+                                      const std::string& log_dir = "");

 // Return the name to use for a metadatabase. The result will be prefixed with
 // "dbname".
@@ -111,9 +111,8 @@ void FlushJob::ReportFlushInputSize(const autovector<MemTable*>& mems) {
 }

 void FlushJob::RecordFlushIOStats() {
-  ThreadStatusUtil::IncreaseThreadOperationProperty(
+  ThreadStatusUtil::SetThreadOperationProperty(
      ThreadStatus::FLUSH_BYTES_WRITTEN, IOSTATS(bytes_written));
-  IOSTATS_RESET(bytes_written);
 }

 Status FlushJob::Run(FileMetaData* file_meta) {
@@ -154,6 +153,7 @@ Status FlushJob::Run(FileMetaData* file_meta) {
  if (!s.ok()) {
    cfd_->imm()->RollbackMemtableFlush(mems, meta.fd.GetNumber());
  } else {
+    TEST_SYNC_POINT("FlushJob::InstallResults");
    // Replace immutable memtable with the generated Table
    s = cfd_->imm()->InstallMemtableFlushResults(
        cfd_, mutable_cf_options_, mems, versions_, db_mutex_,
@@ -197,7 +197,7 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
    if (log_buffer_) {
      log_buffer_->FlushBufferToLog();
    }
-    std::vector<Iterator*> memtables;
+    std::vector<InternalIterator*> memtables;
    ReadOptions ro;
    ro.total_order_seek = true;
    Arena arena;
@@ -231,13 +231,15 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,

      TEST_SYNC_POINT_CALLBACK("FlushJob::WriteLevel0Table:output_compression",
                               &output_compression_);
-      s = BuildTable(
-          dbname_, db_options_.env, *cfd_->ioptions(), env_options_,
-          cfd_->table_cache(), iter.get(), meta, cfd_->internal_comparator(),
-          cfd_->int_tbl_prop_collector_factories(), existing_snapshots_,
-          output_compression_, cfd_->ioptions()->compression_opts,
-          mutable_cf_options_.paranoid_file_checks, cfd_->internal_stats(),
-          Env::IO_HIGH, &info.table_properties);
+      s = BuildTable(dbname_, db_options_.env, *cfd_->ioptions(), env_options_,
+                     cfd_->table_cache(), iter.get(), meta,
+                     cfd_->internal_comparator(),
+                     cfd_->int_tbl_prop_collector_factories(), cfd_->GetID(),
+                     existing_snapshots_, output_compression_,
+                     cfd_->ioptions()->compression_opts,
+                     mutable_cf_options_.paranoid_file_checks,
+                     cfd_->internal_stats(), Env::IO_HIGH, &table_properties_);
+      info.table_properties = table_properties_;
      LogFlush(db_options_.info_log);
    }
    Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
@@ -27,12 +27,12 @@
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/transaction_log.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/autovector.h"
 #include "util/event_logger.h"
 #include "util/instrumented_mutex.h"
 #include "util/stop_watch.h"
 #include "util/thread_local.h"
-#include "util/scoped_arena_iterator.h"
 #include "db/internal_stats.h"
 #include "db/write_controller.h"
 #include "db/flush_scheduler.h"
@@ -66,6 +66,7 @@ class FlushJob {
  ~FlushJob();

  Status Run(FileMetaData* file_meta = nullptr);
+  TableProperties GetTableProperties() const { return table_properties_; }

 private:
  void ReportStartedFlush();
@@ -89,6 +90,7 @@ class FlushJob {
  CompressionType output_compression_;
  Statistics* stats_;
  EventLogger* event_logger_;
+  TableProperties table_properties_;
 };

 }  // namespace rocksdb
@@ -61,7 +61,7 @@ class FlushJobTest : public testing::Test {
    unique_ptr<WritableFileWriter> file_writer(
        new WritableFileWriter(std::move(file), EnvOptions()));
    {
-      log::Writer log(std::move(file_writer));
+      log::Writer log(std::move(file_writer), 0, false);
      std::string record;
      new_db.EncodeTo(&record);
      s = log.AddRecord(record);
@@ -28,7 +28,7 @@ namespace rocksdb {
 //     iter.SetFileIndex(file_index);
 //     iter.Seek(target);
 //     iter.Next()
-class LevelIterator : public Iterator {
+class LevelIterator : public InternalIterator {
 public:
  LevelIterator(const ColumnFamilyData* const cfd,
      const ReadOptions& read_options,
@@ -113,7 +113,7 @@ class LevelIterator : public Iterator {
  bool valid_;
  uint32_t file_index_;
  Status status_;
-  std::unique_ptr<Iterator> file_iter_;
+  std::unique_ptr<InternalIterator> file_iter_;
 };

 ForwardIterator::ForwardIterator(DBImpl* db, const ReadOptions& read_options,
@@ -146,10 +146,10 @@ ForwardIterator::~ForwardIterator() {

 void ForwardIterator::Cleanup(bool release_sv) {
  if (mutable_iter_ != nullptr) {
-    mutable_iter_->~Iterator();
+    mutable_iter_->~InternalIterator();
  }
  for (auto* m : imm_iters_) {
-    m->~Iterator();
+    m->~InternalIterator();
  }
  imm_iters_.clear();
  for (auto* f : l0_iters_) {
@@ -14,6 +14,7 @@
 #include "rocksdb/iterator.h"
 #include "rocksdb/options.h"
 #include "db/dbformat.h"
+#include "table/internal_iterator.h"
 #include "util/arena.h"

 namespace rocksdb {
@@ -30,16 +31,15 @@ class MinIterComparator {
  explicit MinIterComparator(const Comparator* comparator) :
    comparator_(comparator) {}

-  bool operator()(Iterator* a, Iterator* b) {
+  bool operator()(InternalIterator* a, InternalIterator* b) {
    return comparator_->Compare(a->key(), b->key()) > 0;
  }
 private:
  const Comparator* comparator_;
 };

-typedef std::priority_queue<Iterator*,
-          std::vector<Iterator*>,
-          MinIterComparator> MinIterHeap;
+typedef std::priority_queue<InternalIterator*, std::vector<InternalIterator*>,
+                            MinIterComparator> MinIterHeap;

 /**
 * ForwardIterator is a special type of iterator that only supports Seek()
@@ -48,7 +48,7 @@ typedef std::priority_queue<Iterator*,
 * the iterator. At the current implementation, snapshot is taken at the
 * time Seek() is called. The Next() followed do not see new values after.
 */
-class ForwardIterator : public Iterator {
+class ForwardIterator : public InternalIterator {
 public:
  ForwardIterator(DBImpl* db, const ReadOptions& read_options,
                  ColumnFamilyData* cfd, SuperVersion* current_sv = nullptr);
@@ -94,11 +94,11 @@ class ForwardIterator : public Iterator {
  MinIterHeap immutable_min_heap_;

  SuperVersion* sv_;
-  Iterator* mutable_iter_;
-  std::vector<Iterator*> imm_iters_;
-  std::vector<Iterator*> l0_iters_;
+  InternalIterator* mutable_iter_;
+  std::vector<InternalIterator*> imm_iters_;
+  std::vector<InternalIterator*> l0_iters_;
  std::vector<LevelIterator*> level_iters_;
-  Iterator* current_;
+  InternalIterator* current_;
  bool valid_;

  // Internal iterator status; set only by one of the unsupported methods.
@@ -57,7 +57,7 @@ void PrintLevelStats(char* buf, size_t len, const std::string& name,
      NumberToHumanString(stats.num_dropped_records);

  snprintf(buf, len,
-           "%4s %6d/%-3d %8.0f %5.1f " /* Level, Files, Size(MB), Score */
+           "%4s %6d/%-3d %8.2f %5.1f " /* Level, Files, Size(MB), Score */
           "%8.1f "                    /* Read(GB) */
           "%7.1f "                    /* Rn(GB) */
           "%8.1f "                    /* Rnp1(GB) */
@@ -130,6 +130,8 @@ static const std::string aggregated_table_properties =
    "aggregated-table-properties";
 static const std::string aggregated_table_properties_at_level =
    aggregated_table_properties + "-at-level";
+static const std::string num_running_compactions = "num-running-compactions";
+static const std::string num_running_flushes = "num-running-flushes";

 const std::string DB::Properties::kNumFilesAtLevelPrefix =
                      rocksdb_prefix + num_files_at_level_prefix;
@@ -143,6 +145,10 @@ const std::string DB::Properties::kMemTableFlushPending =
                      rocksdb_prefix + mem_table_flush_pending;
 const std::string DB::Properties::kCompactionPending =
                      rocksdb_prefix + compaction_pending;
+const std::string DB::Properties::kNumRunningCompactions =
+    rocksdb_prefix + num_running_compactions;
+const std::string DB::Properties::kNumRunningFlushes =
+    rocksdb_prefix + num_running_flushes;
 const std::string DB::Properties::kBackgroundErrors =
                      rocksdb_prefix + background_errors;
 const std::string DB::Properties::kCurSizeActiveMemTable =
@@ -260,6 +266,10 @@ DBPropertyType GetPropertyType(const Slice& property, bool* is_int_property,
    return kTotalSstFilesSize;
  } else if (in == estimate_pending_comp_bytes) {
    return kEstimatePendingCompactionBytes;
+  } else if (in == num_running_flushes) {
+    return kNumRunningFlushes;
+  } else if (in == num_running_compactions) {
+    return kNumRunningCompactions;
  }
  return kUnknown;
 }
@@ -388,11 +398,17 @@ bool InternalStats::GetIntProperty(DBPropertyType property_type,
      // Return number of mem tables that are ready to flush (made immutable)
      *value = (cfd_->imm()->IsFlushPending() ? 1 : 0);
      return true;
+    case kNumRunningFlushes:
+      *value = db->num_running_flushes();
+      return true;
    case kCompactionPending:
      // 1 if the system already determines at least one compaction is needed.
      // 0 otherwise,
      *value = (cfd_->compaction_picker()->NeedsCompaction(vstorage) ? 1 : 0);
      return true;
+    case kNumRunningCompactions:
+      *value = db->num_running_compactions_;
+      return true;
    case kBackgroundErrors:
      // Accumulated number of  errors in background flushes or compactions.
      *value = GetBackgroundErrorCount();
@@ -669,12 +685,13 @@ void InternalStats::DumpCFStats(std::string* value) {
    total_files += files;
    total_files_being_compacted += files_being_compacted[level];
    if (comp_stats_[level].micros > 0 || files > 0) {
-      uint64_t stalls = level == 0 ?
-        (cf_stats_count_[LEVEL0_SLOWDOWN] +
-         cf_stats_count_[LEVEL0_NUM_FILES] +
-         cf_stats_count_[MEMTABLE_COMPACTION])
-        : (stall_leveln_slowdown_count_soft_[level] +
-           stall_leveln_slowdown_count_hard_[level]);
+      uint64_t stalls =
+          level == 0 ? (cf_stats_count_[LEVEL0_SLOWDOWN_TOTAL] +
+                        cf_stats_count_[LEVEL0_NUM_FILES_TOTAL] +
+                        cf_stats_count_[HARD_PENDING_COMPACTION_BYTES_LIMIT] +
+                        cf_stats_count_[MEMTABLE_COMPACTION])
+                     : (stall_leveln_slowdown_count_soft_[level] +
+                        stall_leveln_slowdown_count_hard_[level]);

      stats_sum.Add(comp_stats_[level]);
      total_file_size += vstorage->NumLevelBytes(level);
@@ -716,15 +733,28 @@ void InternalStats::DumpCFStats(std::string* value) {
           curr_ingest / kGB, interval_ingest / kGB);
  value->append(buf);

-  snprintf(buf, sizeof(buf),
-           "Stalls(count): %" PRIu64 " level0_slowdown, "
-           "%" PRIu64 " level0_numfiles, %" PRIu64 " memtable_compaction, "
-           "%" PRIu64 " leveln_slowdown_soft, "
-           "%" PRIu64 " leveln_slowdown_hard\n",
-           cf_stats_count_[LEVEL0_SLOWDOWN],
-           cf_stats_count_[LEVEL0_NUM_FILES],
-           cf_stats_count_[MEMTABLE_COMPACTION],
-           total_slowdown_count_soft, total_slowdown_count_hard);
+  snprintf(buf, sizeof(buf), "Stalls(count): %" PRIu64
+                             " level0_slowdown, "
+                             "%" PRIu64
+                             " level0_slowdown_with_compaction, "
+                             "%" PRIu64
+                             " level0_numfiles, "
+                             "%" PRIu64
+                             " level0_numfiles_with_compaction, "
+                             "%" PRIu64
+                             " pending_compaction_bytes, "
+                             "%" PRIu64
+                             " memtable_compaction, "
+                             "%" PRIu64
+                             " leveln_slowdown_soft, "
+                             "%" PRIu64 " leveln_slowdown_hard\n",
+           cf_stats_count_[LEVEL0_SLOWDOWN_TOTAL],
+           cf_stats_count_[LEVEL0_SLOWDOWN_WITH_COMPACTION],
+           cf_stats_count_[LEVEL0_NUM_FILES_TOTAL],
+           cf_stats_count_[LEVEL0_NUM_FILES_WITH_COMPACTION],
+           cf_stats_count_[HARD_PENDING_COMPACTION_BYTES_LIMIT],
+           cf_stats_count_[MEMTABLE_COMPACTION], total_slowdown_count_soft,
+           total_slowdown_count_hard);
  value->append(buf);

  cf_stats_snapshot_.ingest_bytes = curr_ingest;
@@ -38,7 +38,10 @@ enum DBPropertyType : uint32_t {
                                 // in memory that have already been flushed
  kMemtableFlushPending,         // Return 1 if mem table flushing is pending,
                                 // otherwise 0.
+  kNumRunningFlushes,      // Return the number of currently running flushes.
  kCompactionPending,      // Return 1 if a compaction is pending. Otherwise 0.
+  kNumRunningCompactions,  // Return the number of currently running
+                           // compactions.
  kBackgroundErrors,       // Return accumulated background errors encountered.
  kCurSizeActiveMemTable,  // Return current size of the active memtable
  kCurSizeAllMemTables,    // Return current size of unflushed
@@ -49,10 +52,10 @@ enum DBPropertyType : uint32_t {
                                   // memtable.
  kNumEntriesInImmutableMemtable,  // Return sum of number of entries in all
                                   // the immutable mem tables.
-  kNumDeletesInMutableMemtable,    // Return number of entries in the mutable
-                                   // memtable.
-  kNumDeletesInImmutableMemtable,  // Return sum of number of deletes in all
-                                   // the immutable mem tables.
+  kNumDeletesInMutableMemtable,    // Return number of deletion entries in the
+                                   // mutable memtable.
+  kNumDeletesInImmutableMemtable,  // Return the total number of deletion
+                                   // entries in all the immutable mem tables.
  kEstimatedNumKeys,  // Estimated total number of keys in the database.
  kEstimatedUsageByTableReaders,  // Estimated memory by table readers.
  kIsFileDeletionEnabled,         // Equals disable_delete_obsolete_files_,
@@ -80,9 +83,12 @@ extern DBPropertyType GetPropertyType(const Slice& property,
 class InternalStats {
 public:
  enum InternalCFStatsType {
-    LEVEL0_SLOWDOWN,
+    LEVEL0_SLOWDOWN_TOTAL,
+    LEVEL0_SLOWDOWN_WITH_COMPACTION,
    MEMTABLE_COMPACTION,
-    LEVEL0_NUM_FILES,
+    LEVEL0_NUM_FILES_TOTAL,
+    LEVEL0_NUM_FILES_WITH_COMPACTION,
+    HARD_PENDING_COMPACTION_BYTES_LIMIT,
    WRITE_STALLS_ENUM_MAX,
    BYTES_FLUSHED,
    INTERNAL_CF_STATS_ENUM_MAX,
@@ -350,9 +356,12 @@ class InternalStats {
 class InternalStats {
 public:
  enum InternalCFStatsType {
-    LEVEL0_SLOWDOWN,
+    LEVEL0_SLOWDOWN_TOTAL,
+    LEVEL0_SLOWDOWN_WITH_COMPACTION,
    MEMTABLE_COMPACTION,
-    LEVEL0_NUM_FILES,
+    LEVEL0_NUM_FILES_TOTAL,
+    LEVEL0_NUM_FILES_WITH_COMPACTION,
+    HARD_PENDING_COMPACTION_BYTES_LIMIT,
    WRITE_STALLS_ENUM_MAX,
    BYTES_FLUSHED,
    INTERNAL_CF_STATS_ENUM_MAX,
@@ -8,6 +8,7 @@
 #include "db/filename.h"
 #include "db/version_set.h"
 #include "db/write_batch_internal.h"
+#include "memtable/hash_linklist_rep.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/db.h"
@@ -22,7 +23,6 @@
 #include "table/block_based_table_factory.h"
 #include "table/plain_table_factory.h"
 #include "util/hash.h"
-#include "util/hash_linklist_rep.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
 #include "util/rate_limiter.h"
@@ -152,6 +152,40 @@ class EventListenerTest : public testing::Test {
  std::vector<ColumnFamilyHandle*> handles_;
 };

+struct TestPropertiesCollector : public rocksdb::TablePropertiesCollector {
+  virtual rocksdb::Status AddUserKey(const rocksdb::Slice& key,
+                                     const rocksdb::Slice& value,
+                                     rocksdb::EntryType type,
+                                     rocksdb::SequenceNumber seq,
+                                     uint64_t file_size) override {
+    return Status::OK();
+  }
+  virtual rocksdb::Status Finish(
+      rocksdb::UserCollectedProperties* properties) override {
+    properties->insert({"0", "1"});
+    return Status::OK();
+  }
+
+  virtual const char* Name() const override {
+    return "TestTablePropertiesCollector";
+  }
+
+  rocksdb::UserCollectedProperties GetReadableProperties() const override {
+    rocksdb::UserCollectedProperties ret;
+    ret["2"] = "3";
+    return ret;
+  }
+};
+
+class TestPropertiesCollectorFactory : public TablePropertiesCollectorFactory {
+ public:
+  virtual TablePropertiesCollector* CreateTablePropertiesCollector(
+      TablePropertiesCollectorFactory::Context context) override {
+    return new TestPropertiesCollector;
+  }
+  const char* Name() const override { return "TestTablePropertiesCollector"; }
+};
+
 class TestCompactionListener : public EventListener {
 public:
  void OnCompactionCompleted(DB *db, const CompactionJobInfo& ci) override {
@@ -161,6 +195,16 @@ class TestCompactionListener : public EventListener {
    ASSERT_GT(ci.output_files.size(), 0U);
    ASSERT_EQ(db->GetEnv()->GetThreadID(), ci.thread_id);
    ASSERT_GT(ci.thread_id, 0U);
+
+    for (auto fl : {ci.input_files, ci.output_files}) {
+      for (auto fn : fl) {
+        auto it = ci.table_properties.find(fn);
+        ASSERT_NE(it, ci.table_properties.end());
+        auto tp = it->second;
+        ASSERT_TRUE(tp != nullptr);
+        ASSERT_EQ(tp->user_collected_properties.find("0")->second, "1");
+      }
+    }
  }

  std::vector<DB*> compacted_dbs_;
@@ -186,6 +230,8 @@ TEST_F(EventListenerTest, OnSingleDBCompactionTest) {
  options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
  options.level0_file_num_compaction_trigger = kNumL0Files;
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());

  TestCompactionListener* listener = new TestCompactionListener();
  options.listeners.emplace_back(listener);
@@ -274,6 +320,8 @@ class TestFlushListener : public EventListener {
    ASSERT_EQ(prev_fc_info_.file_path, info.file_path);
    ASSERT_EQ(db->GetEnv()->GetThreadID(), info.thread_id);
    ASSERT_GT(info.thread_id, 0U);
+    ASSERT_EQ(info.table_properties.user_collected_properties.find("0")->second,
+              "1");
  }

  std::vector<std::string> flushed_column_family_names_;
@@ -299,6 +347,8 @@ TEST_F(EventListenerTest, OnSingleDBFlushTest) {
  std::vector<std::string> cf_names = {
      "pikachu", "ilya", "muromec", "dobrynia",
      "nikitich", "alyosha", "popovich"};
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());
  CreateAndReopenWithCF(cf_names, &options);

  ASSERT_OK(Put(1, "pikachu", std::string(90000, 'p')));
@@ -330,6 +380,8 @@ TEST_F(EventListenerTest, MultiCF) {
 #endif  // ROCKSDB_USING_THREAD_STATUS
  TestFlushListener* listener = new TestFlushListener(options.env);
  options.listeners.emplace_back(listener);
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());
  std::vector<std::string> cf_names = {
      "pikachu", "ilya", "muromec", "dobrynia",
      "nikitich", "alyosha", "popovich"};
@@ -360,6 +412,8 @@ TEST_F(EventListenerTest, MultiDBMultiListeners) {
 #if ROCKSDB_USING_THREAD_STATUS
  options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());
  std::vector<TestFlushListener*> listeners;
  const int kNumDBs = 5;
  const int kNumListeners = 10;
@@ -454,6 +508,8 @@ TEST_F(EventListenerTest, DisableBGCompaction) {
  options.compaction_style = kCompactionStyleNone;
  options.compression = kNoCompression;
  options.write_buffer_size = 100000;  // Small write buffer
+  options.table_properties_collector_factories.push_back(
+      std::make_shared<TestPropertiesCollectorFactory>());

  CreateAndReopenWithCF({"pikachu"}, &options);
  ColumnFamilyMetaData cf_meta;
@@ -22,14 +22,24 @@ enum RecordType {
  // For fragments
  kFirstType = 2,
  kMiddleType = 3,
-  kLastType = 4
+  kLastType = 4,
+
+  // For recycled log files
+  kRecyclableFullType = 5,
+  kRecyclableFirstType = 6,
+  kRecyclableMiddleType = 7,
+  kRecyclableLastType = 8,
 };
-static const int kMaxRecordType = kLastType;
+static const int kMaxRecordType = kRecyclableLastType;

 static const unsigned int kBlockSize = 32768;

 // Header is checksum (4 bytes), type (1 byte), length (2 bytes).
 static const int kHeaderSize = 4 + 1 + 2;

+// Recyclable header is checksum (4 bytes), type (1 byte), log number
+// (4 bytes), length (2 bytes).
+static const int kRecyclableHeaderSize = 4 + 1 + 4 + 2;
+
 }  // namespace log
 }  // namespace rocksdb
@@ -21,9 +21,12 @@ namespace log {
 Reader::Reporter::~Reporter() {
 }

-Reader::Reader(unique_ptr<SequentialFileReader>&& _file, Reporter* reporter,
-               bool checksum, uint64_t initial_offset)
-    : file_(std::move(_file)),
+Reader::Reader(std::shared_ptr<Logger> info_log,
+	       unique_ptr<SequentialFileReader>&& _file,
+               Reporter* reporter, bool checksum, uint64_t initial_offset,
+               uint64_t log_num)
+    : info_log_(info_log),
+      file_(std::move(_file)),
      reporter_(reporter),
      checksum_(checksum),
      backing_store_(new char[kBlockSize]),
@@ -33,7 +36,8 @@ Reader::Reader(unique_ptr<SequentialFileReader>&& _file, Reporter* reporter,
      eof_offset_(0),
      last_record_offset_(0),
      end_of_buffer_offset_(0),
-      initial_offset_(initial_offset) {}
+      initial_offset_(initial_offset),
+      log_number_(log_num) {}

 Reader::~Reader() {
  delete[] backing_store_;
@@ -62,8 +66,15 @@ bool Reader::SkipToInitialBlock() {
  return true;
 }

+// For kAbsoluteConsistency, on clean shutdown we don't expect any error
+// in the log files.  For other modes, we can ignore only incomplete records
+// in the last log file, which are presumably due to a write in progress
+// during restart (or from log recycling).
+//
+// TODO krad: Evaluate if we need to move to a more strict mode where we
+// restrict the inconsistency to only the last log
 bool Reader::ReadRecord(Slice* record, std::string* scratch,
-                        const bool report_eof_inconsistency) {
+                        WALRecoveryMode wal_recovery_mode) {
  if (last_record_offset_ < initial_offset_) {
    if (!SkipToInitialBlock()) {
      return false;
@@ -80,10 +91,11 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
  Slice fragment;
  while (true) {
    uint64_t physical_record_offset = end_of_buffer_offset_ - buffer_.size();
-    const unsigned int record_type =
-        ReadPhysicalRecord(&fragment, report_eof_inconsistency);
+    size_t drop_size;
+    const unsigned int record_type = ReadPhysicalRecord(&fragment, &drop_size);
    switch (record_type) {
      case kFullType:
+      case kRecyclableFullType:
        if (in_fragmented_record && !scratch->empty()) {
          // Handle bug in earlier versions of log::Writer where
          // it could emit an empty kFirstType record at the tail end
@@ -98,6 +110,7 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
        return true;

      case kFirstType:
+      case kRecyclableFirstType:
        if (in_fragmented_record && !scratch->empty()) {
          // Handle bug in earlier versions of log::Writer where
          // it could emit an empty kFirstType record at the tail end
@@ -111,6 +124,7 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
        break;

      case kMiddleType:
+      case kRecyclableMiddleType:
        if (!in_fragmented_record) {
          ReportCorruption(fragment.size(),
                           "missing start of fragmented record(1)");
@@ -120,6 +134,7 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
        break;

      case kLastType:
+      case kRecyclableLastType:
        if (!in_fragmented_record) {
          ReportCorruption(fragment.size(),
                           "missing start of fragmented record(2)");
@@ -131,9 +146,17 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
        }
        break;

+      case kBadHeader:
+        if (wal_recovery_mode == WALRecoveryMode::kAbsoluteConsistency) {
+          // in clean shutdown we don't expect any error in the log files
+          ReportCorruption(drop_size, "truncated header");
+        }
+      // fall-thru
+
      case kEof:
        if (in_fragmented_record) {
-          if (report_eof_inconsistency) {
+          if (wal_recovery_mode == WALRecoveryMode::kAbsoluteConsistency) {
+            // in clean shutdown we don't expect any error in the log files
            ReportCorruption(scratch->size(), "error reading trailing data");
          }
          // This can be caused by the writer dying immediately after
@@ -143,6 +166,23 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch,
        }
        return false;

+      case kOldRecord:
+        if (wal_recovery_mode != WALRecoveryMode::kSkipAnyCorruptedRecords) {
+          // Treat a record from a previous instance of the log as EOF.
+          if (in_fragmented_record) {
+            if (wal_recovery_mode == WALRecoveryMode::kAbsoluteConsistency) {
+              // in clean shutdown we don't expect any error in the log files
+              ReportCorruption(scratch->size(), "error reading trailing data");
+            }
+            // This can be caused by the writer dying immediately after
+            //  writing a physical record but before completing the next; don't
+            //  treat it as a corruption, just ignore the entire logical record.
+            scratch->clear();
+          }
+          return false;
+        }
+      // fall-thru
+
      case kBadRecord:
        if (in_fragmented_record) {
          ReportCorruption(scratch->size(), "error in middle of record");
@@ -244,36 +284,49 @@ void Reader::ReportDrop(size_t bytes, const Status& reason) {
  }
 }

-unsigned int Reader::ReadPhysicalRecord(Slice* result,
-                                        const bool report_eof_inconsistency) {
+bool Reader::ReadMore(size_t* drop_size, int *error) {
+  if (!eof_ && !read_error_) {
+    // Last read was a full read, so this is a trailer to skip
+    buffer_.clear();
+    Status status = file_->Read(kBlockSize, &buffer_, backing_store_);
+    end_of_buffer_offset_ += buffer_.size();
+    if (!status.ok()) {
+      buffer_.clear();
+      ReportDrop(kBlockSize, status);
+      read_error_ = true;
+      *error = kEof;
+      return false;
+    } else if (buffer_.size() < (size_t)kBlockSize) {
+      eof_ = true;
+      eof_offset_ = buffer_.size();
+    }
+    return true;
+  } else {
+    // Note that if buffer_ is non-empty, we have a truncated header at the
+    //  end of the file, which can be caused by the writer crashing in the
+    //  middle of writing the header. Unless explicitly requested we don't
+    //  considering this an error, just report EOF.
+    if (buffer_.size()) {
+      *drop_size = buffer_.size();
+      buffer_.clear();
+      *error = kBadHeader;
+      return false;
+    }
+    buffer_.clear();
+    *error = kEof;
+    return false;
+  }
+}
+
+unsigned int Reader::ReadPhysicalRecord(Slice* result, size_t* drop_size) {
  while (true) {
+    // We need at least the minimum header size
    if (buffer_.size() < (size_t)kHeaderSize) {
-      if (!eof_ && !read_error_) {
-        // Last read was a full read, so this is a trailer to skip
-        buffer_.clear();
-        Status status = file_->Read(kBlockSize, &buffer_, backing_store_);
-        end_of_buffer_offset_ += buffer_.size();
-        if (!status.ok()) {
-          buffer_.clear();
-          ReportDrop(kBlockSize, status);
-          read_error_ = true;
-          return kEof;
-        } else if (buffer_.size() < (size_t)kBlockSize) {
-          eof_ = true;
-          eof_offset_ = buffer_.size();
-        }
-        continue;
-      } else {
-        // Note that if buffer_ is non-empty, we have a truncated header at the
-        //  end of the file, which can be caused by the writer crashing in the
-        //  middle of writing the header. Unless explicitly requested we don't
-        //  considering this an error, just report EOF.
-        if (buffer_.size() && report_eof_inconsistency) {
-          ReportCorruption(buffer_.size(), "truncated header");
-        }
-        buffer_.clear();
-        return kEof;
+      int r;
+      if (!ReadMore(drop_size, &r)) {
+	return r;
      }
+      continue;
    }

    // Parse the header
@@ -282,18 +335,34 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result,
    const uint32_t b = static_cast<uint32_t>(header[5]) & 0xff;
    const unsigned int type = header[6];
    const uint32_t length = a | (b << 8);
-    if (kHeaderSize + length > buffer_.size()) {
-      size_t drop_size = buffer_.size();
+    int header_size = kHeaderSize;
+    if (type >= kRecyclableFullType && type <= kRecyclableLastType) {
+      header_size = kRecyclableHeaderSize;
+      // We need enough for the larger header
+      if (buffer_.size() < (size_t)kRecyclableHeaderSize) {
+	int r;
+	if (!ReadMore(drop_size, &r)) {
+          return r;
+        }
+	continue;
+      }
+      const uint32_t log_num = DecodeFixed32(header + 7);
+      if (log_num != log_number_) {
+        return kOldRecord;
+      }
+    }
+    if (header_size + length > buffer_.size()) {
+      *drop_size = buffer_.size();
      buffer_.clear();
      if (!eof_) {
-        ReportCorruption(drop_size, "bad record length");
+        ReportCorruption(*drop_size, "bad record length");
        return kBadRecord;
      }
      // If the end of the file has been reached without reading |length| bytes
      // of payload, assume the writer died in the middle of writing the record.
      // Don't report a corruption unless requested.
-      if (drop_size && report_eof_inconsistency) {
-        ReportCorruption(drop_size, "truncated header");
+      if (*drop_size) {
+        return kBadHeader;
      }
      return kEof;
    }
@@ -311,29 +380,29 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result,
    // Check crc
    if (checksum_) {
      uint32_t expected_crc = crc32c::Unmask(DecodeFixed32(header));
-      uint32_t actual_crc = crc32c::Value(header + 6, 1 + length);
+      uint32_t actual_crc = crc32c::Value(header + 6, length + header_size - 6);
      if (actual_crc != expected_crc) {
        // Drop the rest of the buffer since "length" itself may have
        // been corrupted and if we trust it, we could find some
        // fragment of a real log record that just happens to look
        // like a valid log record.
-        size_t drop_size = buffer_.size();
+        *drop_size = buffer_.size();
        buffer_.clear();
-        ReportCorruption(drop_size, "checksum mismatch");
+        ReportCorruption(*drop_size, "checksum mismatch");
        return kBadRecord;
      }
    }

-    buffer_.remove_prefix(kHeaderSize + length);
+    buffer_.remove_prefix(header_size + length);

    // Skip physical record that started before initial_offset_
-    if (end_of_buffer_offset_ - buffer_.size() - kHeaderSize - length <
+    if (end_of_buffer_offset_ - buffer_.size() - header_size - length <
        initial_offset_) {
      result->clear();
      return kBadRecord;
    }

-    *result = Slice(header + kHeaderSize, length);
+    *result = Slice(header + header_size, length);
    return type;
  }
 }
@@ -14,10 +14,12 @@
 #include "db/log_format.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
+#include "rocksdb/options.h"

 namespace rocksdb {

 class SequentialFileReader;
+class Logger;
 using std::unique_ptr;

 namespace log {
@@ -51,8 +53,10 @@ class Reader {
  //
  // The Reader will start reading at the first record located at physical
  // position >= initial_offset within the file.
-  Reader(unique_ptr<SequentialFileReader>&& file, Reporter* reporter,
-         bool checksum, uint64_t initial_offset);
+  Reader(std::shared_ptr<Logger> info_log,
+	 unique_ptr<SequentialFileReader>&& file,
+         Reporter* reporter, bool checksum, uint64_t initial_offset,
+         uint64_t log_num);

  ~Reader();

@@ -62,7 +66,8 @@ class Reader {
  // will only be valid until the next mutating operation on this
  // reader or the next mutation to *scratch.
  bool ReadRecord(Slice* record, std::string* scratch,
-                  bool report_eof_inconsistency = false);
+                  WALRecoveryMode wal_recovery_mode =
+                      WALRecoveryMode::kTolerateCorruptedTailRecords);

  // Returns the physical offset of the last record returned by ReadRecord.
  //
@@ -84,6 +89,7 @@ class Reader {
  SequentialFileReader* file() { return file_.get(); }

 private:
+  std::shared_ptr<Logger> info_log_;
  const unique_ptr<SequentialFileReader> file_;
  Reporter* const reporter_;
  bool const checksum_;
@@ -104,6 +110,9 @@ class Reader {
  // Offset at which to start looking for the first record to return
  uint64_t const initial_offset_;

+  // which log number this is
+  uint64_t const log_number_;
+
  // Extend record types with the following special values
  enum {
    kEof = kMaxRecordType + 1,
@@ -112,7 +121,11 @@ class Reader {
    // * The record has an invalid CRC (ReadPhysicalRecord reports a drop)
    // * The record is a 0-length record (No drop is reported)
    // * The record is below constructor's initial_offset (No drop is reported)
-    kBadRecord = kMaxRecordType + 2
+    kBadRecord = kMaxRecordType + 2,
+    // Returned when we fail to read a valid header.
+    kBadHeader = kMaxRecordType + 3,
+    // Returned when we read an old record from a previous user of the log.
+    kOldRecord = kMaxRecordType + 4,
  };

  // Skips all blocks that are completely before "initial_offset_".
@@ -121,8 +134,10 @@ class Reader {
  bool SkipToInitialBlock();

  // Return type, or one of the preceding special values
-  unsigned int ReadPhysicalRecord(Slice* result,
-                                  bool report_eof_inconsistency = false);
+  unsigned int ReadPhysicalRecord(Slice* result, size_t* drop_size);
+
+  // Read some more
+  bool ReadMore(size_t* drop_size, int *error);

  // Reports dropped bytes to the reporter.
  // buffer_ must be updated to remove the dropped bytes prior to invocation.
@@ -43,7 +43,7 @@ static std::string RandomSkewedString(int i, Random* rnd) {
  return BigString(NumberString(i), rnd->Skewed(17));
 }

-class LogTest : public testing::Test {
+class LogTest : public ::testing::TestWithParam<int> {
 private:
  class StringSource : public SequentialFile {
   public:
@@ -153,19 +153,26 @@ class LogTest : public testing::Test {

  // Record metadata for testing initial offset functionality
  static size_t initial_offset_record_sizes_[];
-  static uint64_t initial_offset_last_record_offsets_[];
+  uint64_t initial_offset_last_record_offsets_[4];

 public:
  LogTest()
      : reader_contents_(),
-        dest_holder_(
-            test::GetWritableFileWriter(
-              new test::StringSink(&reader_contents_))),
+        dest_holder_(test::GetWritableFileWriter(
+            new test::StringSink(&reader_contents_))),
        source_holder_(
            test::GetSequentialFileReader(new StringSource(reader_contents_))),
-        writer_(std::move(dest_holder_)),
-        reader_(std::move(source_holder_), &report_, true /*checksum*/,
-                0 /*initial_offset*/) {}
+        writer_(std::move(dest_holder_), 123, GetParam()),
+        reader_(NULL, std::move(source_holder_), &report_, true /*checksum*/,
+                0 /*initial_offset*/, 123) {
+    int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+    initial_offset_last_record_offsets_[0] = 0;
+    initial_offset_last_record_offsets_[1] = header_size + 10000;
+    initial_offset_last_record_offsets_[2] = 2 * (header_size + 10000);
+    initial_offset_last_record_offsets_[3] = 2 * (header_size + 10000) +
+                                             (2 * log::kBlockSize - 1000) +
+                                             3 * header_size;
+  }

  void Write(const std::string& msg) {
    writer_.AddRecord(Slice(msg));
@@ -175,10 +182,11 @@ class LogTest : public testing::Test {
    return dest_contents().size();
  }

-  std::string Read(const bool report_eof_inconsistency = false) {
+  std::string Read(const WALRecoveryMode wal_recovery_mode =
+                       WALRecoveryMode::kTolerateCorruptedTailRecords) {
    std::string scratch;
    Slice record;
-    if (reader_.ReadRecord(&record, &scratch, report_eof_inconsistency)) {
+    if (reader_.ReadRecord(&record, &scratch, wal_recovery_mode)) {
      return record.ToString();
    } else {
      return "EOF";
@@ -200,9 +208,11 @@ class LogTest : public testing::Test {
    dest->Drop(bytes);
  }

-  void FixChecksum(int header_offset, int len) {
+  void FixChecksum(int header_offset, int len, bool recyclable) {
    // Compute crc of type/len/data
-    uint32_t crc = crc32c::Value(&dest_contents()[header_offset+6], 1 + len);
+    int header_size = recyclable ? kRecyclableHeaderSize : kHeaderSize;
+    uint32_t crc = crc32c::Value(&dest_contents()[header_offset + 6],
+                                 header_size - 6 + len);
    crc = crc32c::Mask(crc);
    EncodeFixed32(&dest_contents()[header_offset], crc);
  }
@@ -259,8 +269,8 @@ class LogTest : public testing::Test {
    unique_ptr<SequentialFileReader> file_reader(
        test::GetSequentialFileReader(new StringSource(reader_contents_)));
    unique_ptr<Reader> offset_reader(
-        new Reader(std::move(file_reader), &report_, true /*checksum*/,
-                   WrittenBytes() + offset_past_end));
+        new Reader(NULL, std::move(file_reader), &report_,
+                   true /*checksum*/, WrittenBytes() + offset_past_end, 123));
    Slice record;
    std::string scratch;
    ASSERT_TRUE(!offset_reader->ReadRecord(&record, &scratch));
@@ -271,8 +281,9 @@ class LogTest : public testing::Test {
    WriteInitialOffsetLog();
    unique_ptr<SequentialFileReader> file_reader(
        test::GetSequentialFileReader(new StringSource(reader_contents_)));
-    unique_ptr<Reader> offset_reader(new Reader(
-        std::move(file_reader), &report_, true /*checksum*/, initial_offset));
+    unique_ptr<Reader> offset_reader(
+        new Reader(NULL, std::move(file_reader), &report_,
+                   true /*checksum*/, initial_offset, 123));
    Slice record;
    std::string scratch;
    ASSERT_TRUE(offset_reader->ReadRecord(&record, &scratch));
@@ -291,16 +302,9 @@ size_t LogTest::initial_offset_record_sizes_[] =
     2 * log::kBlockSize - 1000,  // Span three blocks
     1};

-uint64_t LogTest::initial_offset_last_record_offsets_[] =
-    {0,
-     kHeaderSize + 10000,
-     2 * (kHeaderSize + 10000),
-     2 * (kHeaderSize + 10000) +
-         (2 * log::kBlockSize - 1000) + 3 * kHeaderSize};
+TEST_P(LogTest, Empty) { ASSERT_EQ("EOF", Read()); }

-TEST_F(LogTest, Empty) { ASSERT_EQ("EOF", Read()); }
-
-TEST_F(LogTest, ReadWrite) {
+TEST_P(LogTest, ReadWrite) {
  Write("foo");
  Write("bar");
  Write("");
@@ -313,7 +317,7 @@ TEST_F(LogTest, ReadWrite) {
  ASSERT_EQ("EOF", Read());  // Make sure reads at eof work
 }

-TEST_F(LogTest, ManyBlocks) {
+TEST_P(LogTest, ManyBlocks) {
  for (int i = 0; i < 100000; i++) {
    Write(NumberString(i));
  }
@@ -323,7 +327,7 @@ TEST_F(LogTest, ManyBlocks) {
  ASSERT_EQ("EOF", Read());
 }

-TEST_F(LogTest, Fragmentation) {
+TEST_P(LogTest, Fragmentation) {
  Write("small");
  Write(BigString("medium", 50000));
  Write(BigString("large", 100000));
@@ -333,11 +337,12 @@ TEST_F(LogTest, Fragmentation) {
  ASSERT_EQ("EOF", Read());
 }

-TEST_F(LogTest, MarginalTrailer) {
+TEST_P(LogTest, MarginalTrailer) {
  // Make a trailer that is exactly the same length as an empty record.
-  const int n = kBlockSize - 2*kHeaderSize;
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int n = kBlockSize - 2 * header_size;
  Write(BigString("foo", n));
-  ASSERT_EQ((unsigned int)(kBlockSize - kHeaderSize), WrittenBytes());
+  ASSERT_EQ((unsigned int)(kBlockSize - header_size), WrittenBytes());
  Write("");
  Write("bar");
  ASSERT_EQ(BigString("foo", n), Read());
@@ -346,11 +351,12 @@ TEST_F(LogTest, MarginalTrailer) {
  ASSERT_EQ("EOF", Read());
 }

-TEST_F(LogTest, MarginalTrailer2) {
+TEST_P(LogTest, MarginalTrailer2) {
  // Make a trailer that is exactly the same length as an empty record.
-  const int n = kBlockSize - 2*kHeaderSize;
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int n = kBlockSize - 2 * header_size;
  Write(BigString("foo", n));
-  ASSERT_EQ((unsigned int)(kBlockSize - kHeaderSize), WrittenBytes());
+  ASSERT_EQ((unsigned int)(kBlockSize - header_size), WrittenBytes());
  Write("bar");
  ASSERT_EQ(BigString("foo", n), Read());
  ASSERT_EQ("bar", Read());
@@ -359,10 +365,11 @@ TEST_F(LogTest, MarginalTrailer2) {
  ASSERT_EQ("", ReportMessage());
 }

-TEST_F(LogTest, ShortTrailer) {
-  const int n = kBlockSize - 2*kHeaderSize + 4;
+TEST_P(LogTest, ShortTrailer) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int n = kBlockSize - 2 * header_size + 4;
  Write(BigString("foo", n));
-  ASSERT_EQ((unsigned int)(kBlockSize - kHeaderSize + 4), WrittenBytes());
+  ASSERT_EQ((unsigned int)(kBlockSize - header_size + 4), WrittenBytes());
  Write("");
  Write("bar");
  ASSERT_EQ(BigString("foo", n), Read());
@@ -371,15 +378,16 @@ TEST_F(LogTest, ShortTrailer) {
  ASSERT_EQ("EOF", Read());
 }

-TEST_F(LogTest, AlignedEof) {
-  const int n = kBlockSize - 2*kHeaderSize + 4;
+TEST_P(LogTest, AlignedEof) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int n = kBlockSize - 2 * header_size + 4;
  Write(BigString("foo", n));
-  ASSERT_EQ((unsigned int)(kBlockSize - kHeaderSize + 4), WrittenBytes());
+  ASSERT_EQ((unsigned int)(kBlockSize - header_size + 4), WrittenBytes());
  ASSERT_EQ(BigString("foo", n), Read());
  ASSERT_EQ("EOF", Read());
 }

-TEST_F(LogTest, RandomRead) {
+TEST_P(LogTest, RandomRead) {
  const int N = 500;
  Random write_rnd(301);
  for (int i = 0; i < N; i++) {
@@ -394,7 +402,7 @@ TEST_F(LogTest, RandomRead) {

 // Tests of all the error paths in log_reader.cc follow:

-TEST_F(LogTest, ReadError) {
+TEST_P(LogTest, ReadError) {
  Write("foo");
  ForceError();
  ASSERT_EQ("EOF", Read());
@@ -402,17 +410,17 @@ TEST_F(LogTest, ReadError) {
  ASSERT_EQ("OK", MatchError("read error"));
 }

-TEST_F(LogTest, BadRecordType) {
+TEST_P(LogTest, BadRecordType) {
  Write("foo");
  // Type is stored in header[6]
  IncrementByte(6, 100);
-  FixChecksum(0, 3);
+  FixChecksum(0, 3, false);
  ASSERT_EQ("EOF", Read());
  ASSERT_EQ(3U, DroppedBytes());
  ASSERT_EQ("OK", MatchError("unknown record type"));
 }

-TEST_F(LogTest, TruncatedTrailingRecordIsIgnored) {
+TEST_P(LogTest, TruncatedTrailingRecordIsIgnored) {
  Write("foo");
  ShrinkSize(4);   // Drop all payload as well as a header byte
  ASSERT_EQ("EOF", Read());
@@ -421,17 +429,18 @@ TEST_F(LogTest, TruncatedTrailingRecordIsIgnored) {
  ASSERT_EQ("", ReportMessage());
 }

-TEST_F(LogTest, TruncatedTrailingRecordIsNotIgnored) {
+TEST_P(LogTest, TruncatedTrailingRecordIsNotIgnored) {
  Write("foo");
  ShrinkSize(4);  // Drop all payload as well as a header byte
-  ASSERT_EQ("EOF", Read(/*report_eof_inconsistency*/ true));
+  ASSERT_EQ("EOF", Read(WALRecoveryMode::kAbsoluteConsistency));
  // Truncated last record is ignored, not treated as an error
  ASSERT_GT(DroppedBytes(), 0U);
  ASSERT_EQ("OK", MatchError("Corruption: truncated header"));
 }

-TEST_F(LogTest, BadLength) {
-  const int kPayloadSize = kBlockSize - kHeaderSize;
+TEST_P(LogTest, BadLength) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  const int kPayloadSize = kBlockSize - header_size;
  Write(BigString("bar", kPayloadSize));
  Write("foo");
  // Least significant size byte is stored in header[4].
@@ -441,7 +450,7 @@ TEST_F(LogTest, BadLength) {
  ASSERT_EQ("OK", MatchError("bad record length"));
 }

-TEST_F(LogTest, BadLengthAtEndIsIgnored) {
+TEST_P(LogTest, BadLengthAtEndIsIgnored) {
  Write("foo");
  ShrinkSize(1);
  ASSERT_EQ("EOF", Read());
@@ -449,63 +458,63 @@ TEST_F(LogTest, BadLengthAtEndIsIgnored) {
  ASSERT_EQ("", ReportMessage());
 }

-TEST_F(LogTest, BadLengthAtEndIsNotIgnored) {
+TEST_P(LogTest, BadLengthAtEndIsNotIgnored) {
  Write("foo");
  ShrinkSize(1);
-  ASSERT_EQ("EOF", Read(/*report_eof_inconsistency=*/true));
+  ASSERT_EQ("EOF", Read(WALRecoveryMode::kAbsoluteConsistency));
  ASSERT_GT(DroppedBytes(), 0U);
  ASSERT_EQ("OK", MatchError("Corruption: truncated header"));
 }

-TEST_F(LogTest, ChecksumMismatch) {
-  Write("foo");
-  IncrementByte(0, 10);
+TEST_P(LogTest, ChecksumMismatch) {
+  Write("foooooo");
+  IncrementByte(0, 14);
  ASSERT_EQ("EOF", Read());
-  ASSERT_EQ(10U, DroppedBytes());
+  ASSERT_EQ(14U + 4 * !!GetParam(), DroppedBytes());
  ASSERT_EQ("OK", MatchError("checksum mismatch"));
 }

-TEST_F(LogTest, UnexpectedMiddleType) {
+TEST_P(LogTest, UnexpectedMiddleType) {
  Write("foo");
-  SetByte(6, kMiddleType);
-  FixChecksum(0, 3);
+  SetByte(6, GetParam() ? kRecyclableMiddleType : kMiddleType);
+  FixChecksum(0, 3, !!GetParam());
  ASSERT_EQ("EOF", Read());
  ASSERT_EQ(3U, DroppedBytes());
  ASSERT_EQ("OK", MatchError("missing start"));
 }

-TEST_F(LogTest, UnexpectedLastType) {
+TEST_P(LogTest, UnexpectedLastType) {
  Write("foo");
-  SetByte(6, kLastType);
-  FixChecksum(0, 3);
+  SetByte(6, GetParam() ? kRecyclableLastType : kLastType);
+  FixChecksum(0, 3, !!GetParam());
  ASSERT_EQ("EOF", Read());
  ASSERT_EQ(3U, DroppedBytes());
  ASSERT_EQ("OK", MatchError("missing start"));
 }

-TEST_F(LogTest, UnexpectedFullType) {
+TEST_P(LogTest, UnexpectedFullType) {
  Write("foo");
  Write("bar");
-  SetByte(6, kFirstType);
-  FixChecksum(0, 3);
+  SetByte(6, GetParam() ? kRecyclableFirstType : kFirstType);
+  FixChecksum(0, 3, !!GetParam());
  ASSERT_EQ("bar", Read());
  ASSERT_EQ("EOF", Read());
  ASSERT_EQ(3U, DroppedBytes());
  ASSERT_EQ("OK", MatchError("partial record without end"));
 }

-TEST_F(LogTest, UnexpectedFirstType) {
+TEST_P(LogTest, UnexpectedFirstType) {
  Write("foo");
  Write(BigString("bar", 100000));
-  SetByte(6, kFirstType);
-  FixChecksum(0, 3);
+  SetByte(6, GetParam() ? kRecyclableFirstType : kFirstType);
+  FixChecksum(0, 3, !!GetParam());
  ASSERT_EQ(BigString("bar", 100000), Read());
  ASSERT_EQ("EOF", Read());
  ASSERT_EQ(3U, DroppedBytes());
  ASSERT_EQ("OK", MatchError("partial record without end"));
 }

-TEST_F(LogTest, MissingLastIsIgnored) {
+TEST_P(LogTest, MissingLastIsIgnored) {
  Write(BigString("bar", kBlockSize));
  // Remove the LAST block, including header.
  ShrinkSize(14);
@@ -514,16 +523,16 @@ TEST_F(LogTest, MissingLastIsIgnored) {
  ASSERT_EQ(0U, DroppedBytes());
 }

-TEST_F(LogTest, MissingLastIsNotIgnored) {
+TEST_P(LogTest, MissingLastIsNotIgnored) {
  Write(BigString("bar", kBlockSize));
  // Remove the LAST block, including header.
  ShrinkSize(14);
-  ASSERT_EQ("EOF", Read(/*report_eof_inconsistency=*/true));
+  ASSERT_EQ("EOF", Read(WALRecoveryMode::kAbsoluteConsistency));
  ASSERT_GT(DroppedBytes(), 0U);
  ASSERT_EQ("OK", MatchError("Corruption: error reading trailing data"));
 }

-TEST_F(LogTest, PartialLastIsIgnored) {
+TEST_P(LogTest, PartialLastIsIgnored) {
  Write(BigString("bar", kBlockSize));
  // Cause a bad record length in the LAST block.
  ShrinkSize(1);
@@ -532,18 +541,18 @@ TEST_F(LogTest, PartialLastIsIgnored) {
  ASSERT_EQ(0U, DroppedBytes());
 }

-TEST_F(LogTest, PartialLastIsNotIgnored) {
+TEST_P(LogTest, PartialLastIsNotIgnored) {
  Write(BigString("bar", kBlockSize));
  // Cause a bad record length in the LAST block.
  ShrinkSize(1);
-  ASSERT_EQ("EOF", Read(/*report_eof_inconsistency=*/true));
+  ASSERT_EQ("EOF", Read(WALRecoveryMode::kAbsoluteConsistency));
  ASSERT_GT(DroppedBytes(), 0U);
  ASSERT_EQ("OK", MatchError(
                      "Corruption: truncated headerCorruption: "
                      "error reading trailing data"));
 }

-TEST_F(LogTest, ErrorJoinsRecords) {
+TEST_P(LogTest, ErrorJoinsRecords) {
  // Consider two fragmented records:
  //    first(R1) last(R1) first(R2) last(R2)
  // where the middle two fragments disappear.  We do not want
@@ -566,46 +575,60 @@ TEST_F(LogTest, ErrorJoinsRecords) {
  ASSERT_GE(dropped, 2 * kBlockSize);
 }

-TEST_F(LogTest, ReadStart) { CheckInitialOffsetRecord(0, 0); }
+TEST_P(LogTest, ReadStart) { CheckInitialOffsetRecord(0, 0); }

-TEST_F(LogTest, ReadSecondOneOff) { CheckInitialOffsetRecord(1, 1); }
+TEST_P(LogTest, ReadSecondOneOff) { CheckInitialOffsetRecord(1, 1); }

-TEST_F(LogTest, ReadSecondTenThousand) { CheckInitialOffsetRecord(10000, 1); }
+TEST_P(LogTest, ReadSecondTenThousand) { CheckInitialOffsetRecord(10000, 1); }

-TEST_F(LogTest, ReadSecondStart) { CheckInitialOffsetRecord(10007, 1); }
+TEST_P(LogTest, ReadSecondStart) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  CheckInitialOffsetRecord(10000 + header_size, 1);
+}

-TEST_F(LogTest, ReadThirdOneOff) { CheckInitialOffsetRecord(10008, 2); }
+TEST_P(LogTest, ReadThirdOneOff) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  CheckInitialOffsetRecord(10000 + header_size + 1, 2);
+}

-TEST_F(LogTest, ReadThirdStart) { CheckInitialOffsetRecord(20014, 2); }
+TEST_P(LogTest, ReadThirdStart) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  CheckInitialOffsetRecord(20000 + 2 * header_size, 2);
+}

-TEST_F(LogTest, ReadFourthOneOff) { CheckInitialOffsetRecord(20015, 3); }
+TEST_P(LogTest, ReadFourthOneOff) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  CheckInitialOffsetRecord(20000 + 2 * header_size + 1, 3);
+}

-TEST_F(LogTest, ReadFourthFirstBlockTrailer) {
+TEST_P(LogTest, ReadFourthFirstBlockTrailer) {
  CheckInitialOffsetRecord(log::kBlockSize - 4, 3);
 }

-TEST_F(LogTest, ReadFourthMiddleBlock) {
+TEST_P(LogTest, ReadFourthMiddleBlock) {
  CheckInitialOffsetRecord(log::kBlockSize + 1, 3);
 }

-TEST_F(LogTest, ReadFourthLastBlock) {
+TEST_P(LogTest, ReadFourthLastBlock) {
  CheckInitialOffsetRecord(2 * log::kBlockSize + 1, 3);
 }

-TEST_F(LogTest, ReadFourthStart) {
+TEST_P(LogTest, ReadFourthStart) {
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
  CheckInitialOffsetRecord(
-      2 * (kHeaderSize + 1000) + (2 * log::kBlockSize - 1000) + 3 * kHeaderSize,
+      2 * (header_size + 1000) + (2 * log::kBlockSize - 1000) + 3 * header_size,
      3);
 }

-TEST_F(LogTest, ReadEnd) { CheckOffsetPastEndReturnsNoRecords(0); }
+TEST_P(LogTest, ReadEnd) { CheckOffsetPastEndReturnsNoRecords(0); }

-TEST_F(LogTest, ReadPastEnd) { CheckOffsetPastEndReturnsNoRecords(5); }
+TEST_P(LogTest, ReadPastEnd) { CheckOffsetPastEndReturnsNoRecords(5); }

-TEST_F(LogTest, ClearEofSingleBlock) {
+TEST_P(LogTest, ClearEofSingleBlock) {
  Write("foo");
  Write("bar");
-  ForceEOF(3 + kHeaderSize + 2);
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  ForceEOF(3 + header_size + 2);
  ASSERT_EQ("foo", Read());
  UnmarkEOF();
  ASSERT_EQ("bar", Read());
@@ -617,12 +640,13 @@ TEST_F(LogTest, ClearEofSingleBlock) {
  ASSERT_TRUE(IsEOF());
 }

-TEST_F(LogTest, ClearEofMultiBlock) {
+TEST_P(LogTest, ClearEofMultiBlock) {
  size_t num_full_blocks = 5;
-  size_t n = (kBlockSize - kHeaderSize) * num_full_blocks + 25;
+  int header_size = GetParam() ? kRecyclableHeaderSize : kHeaderSize;
+  size_t n = (kBlockSize - header_size) * num_full_blocks + 25;
  Write(BigString("foo", n));
  Write(BigString("bar", n));
-  ForceEOF(n + num_full_blocks * kHeaderSize + 10);
+  ForceEOF(n + num_full_blocks * header_size + header_size + 3);
  ASSERT_EQ(BigString("foo", n), Read());
  ASSERT_TRUE(IsEOF());
  UnmarkEOF();
@@ -634,7 +658,7 @@ TEST_F(LogTest, ClearEofMultiBlock) {
  ASSERT_TRUE(IsEOF());
 }

-TEST_F(LogTest, ClearEofError) {
+TEST_P(LogTest, ClearEofError) {
  // If an error occurs during Read() in UnmarkEOF(), the records contained
  // in the buffer should be returned on subsequent calls of ReadRecord()
  // until no more full records are left, whereafter ReadRecord() should return
@@ -652,7 +676,7 @@ TEST_F(LogTest, ClearEofError) {
  ASSERT_EQ("EOF", Read());
 }

-TEST_F(LogTest, ClearEofError2) {
+TEST_P(LogTest, ClearEofError2) {
  Write("foo");
  Write("bar");
  UnmarkEOF();
@@ -666,6 +690,8 @@ TEST_F(LogTest, ClearEofError2) {
  ASSERT_EQ("OK", MatchError("read error"));
 }

+INSTANTIATE_TEST_CASE_P(bool, LogTest, ::testing::Values(0, 2));
+
 }  // namespace log
 }  // namespace rocksdb

@@ -18,8 +18,12 @@
 namespace rocksdb {
 namespace log {

-Writer::Writer(unique_ptr<WritableFileWriter>&& dest)
-    : dest_(std::move(dest)), block_offset_(0) {
+Writer::Writer(unique_ptr<WritableFileWriter>&& dest,
+               uint64_t log_number, bool recycle_log_files)
+    : dest_(std::move(dest)),
+      block_offset_(0),
+      log_number_(log_number),
+      recycle_log_files_(recycle_log_files) {
  for (int i = 0; i <= kMaxRecordType; i++) {
    char t = static_cast<char>(i);
    type_crc_[i] = crc32c::Value(&t, 1);
@@ -33,6 +37,10 @@ Status Writer::AddRecord(const Slice& slice) {
  const char* ptr = slice.data();
  size_t left = slice.size();

+  // Header size varies depending on whether we are recycling or not.
+  const int header_size =
+      recycle_log_files_ ? kRecyclableHeaderSize : kHeaderSize;
+
  // Fragment the record if necessary and emit it.  Note that if slice
  // is empty, we still want to iterate once to emit a single
  // zero-length record
@@ -41,32 +49,34 @@ Status Writer::AddRecord(const Slice& slice) {
  do {
    const int leftover = kBlockSize - block_offset_;
    assert(leftover >= 0);
-    if (leftover < kHeaderSize) {
+    if (leftover < header_size) {
      // Switch to a new block
      if (leftover > 0) {
-        // Fill the trailer (literal below relies on kHeaderSize being 7)
-        assert(kHeaderSize == 7);
-        dest_->Append(Slice("\x00\x00\x00\x00\x00\x00", leftover));
+        // Fill the trailer (literal below relies on kHeaderSize and
+        // kRecyclableHeaderSize being <= 11)
+        assert(header_size <= 11);
+        dest_->Append(
+            Slice("\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00", leftover));
      }
      block_offset_ = 0;
    }

-    // Invariant: we never leave < kHeaderSize bytes in a block.
-    assert(static_cast<int>(kBlockSize) - block_offset_ >= kHeaderSize);
+    // Invariant: we never leave < header_size bytes in a block.
+    assert(static_cast<int>(kBlockSize) - block_offset_ >= header_size);

-    const size_t avail = kBlockSize - block_offset_ - kHeaderSize;
+    const size_t avail = kBlockSize - block_offset_ - header_size;
    const size_t fragment_length = (left < avail) ? left : avail;

    RecordType type;
    const bool end = (left == fragment_length);
    if (begin && end) {
-      type = kFullType;
+      type = recycle_log_files_ ? kRecyclableFullType : kFullType;
    } else if (begin) {
-      type = kFirstType;
+      type = recycle_log_files_ ? kRecyclableFirstType : kFirstType;
    } else if (end) {
-      type = kLastType;
+      type = recycle_log_files_ ? kRecyclableLastType : kLastType;
    } else {
-      type = kMiddleType;
+      type = recycle_log_files_ ? kRecyclableMiddleType : kMiddleType;
    }

    s = EmitPhysicalRecord(type, ptr, fragment_length);
@@ -79,28 +89,48 @@ Status Writer::AddRecord(const Slice& slice) {

 Status Writer::EmitPhysicalRecord(RecordType t, const char* ptr, size_t n) {
  assert(n <= 0xffff);  // Must fit in two bytes
-  assert(block_offset_ + kHeaderSize + n <= kBlockSize);
+
+  size_t header_size;
+  char buf[kRecyclableHeaderSize];

  // Format the header
-  char buf[kHeaderSize];
  buf[4] = static_cast<char>(n & 0xff);
  buf[5] = static_cast<char>(n >> 8);
  buf[6] = static_cast<char>(t);

+  uint32_t crc = type_crc_[t];
+  if (t < kRecyclableFullType) {
+    // Legacy record format
+    assert(block_offset_ + kHeaderSize + n <= kBlockSize);
+    header_size = kHeaderSize;
+  } else {
+    // Recyclable record format
+    assert(block_offset_ + kRecyclableHeaderSize + n <= kBlockSize);
+    header_size = kRecyclableHeaderSize;
+
+    // Only encode low 32-bits of the 64-bit log number.  This means
+    // we will fail to detect an old record if we recycled a log from
+    // ~4 billion logs ago, but that is effectively impossible, and
+    // even if it were we'dbe far more likely to see a false positive
+    // on the 32-bit CRC.
+    EncodeFixed32(buf + 7, static_cast<uint32_t>(log_number_));
+    crc = crc32c::Extend(crc, buf + 7, 4);
+  }
+
  // Compute the crc of the record type and the payload.
-  uint32_t crc = crc32c::Extend(type_crc_[t], ptr, n);
-  crc = crc32c::Mask(crc);                 // Adjust for storage
+  crc = crc32c::Extend(crc, ptr, n);
+  crc = crc32c::Mask(crc);  // Adjust for storage
  EncodeFixed32(buf, crc);

  // Write the header and the payload
-  Status s = dest_->Append(Slice(buf, kHeaderSize));
+  Status s = dest_->Append(Slice(buf, header_size));
  if (s.ok()) {
    s = dest_->Append(Slice(ptr, n));
    if (s.ok()) {
      s = dest_->Flush();
    }
  }
-  block_offset_ += kHeaderSize + n;
+  block_offset_ += header_size + n;
  return s;
 }

@@ -43,7 +43,7 @@ namespace log {
 * Data is written out in kBlockSize chunks. If next record does not fit
 * into the space left, the leftover space will be padded with \0.
 *
- * Record format:
+ * Legacy record format:
 *
 * +---------+-----------+-----------+--- ... ---+
 * |CRC (4B) | Size (2B) | Type (1B) | Payload   |
@@ -57,13 +57,23 @@ namespace log {
 *        blocks that are larger than kBlockSize
 * Payload = Byte stream as long as specified by the payload size
 *
+ * Recyclable record format:
+ *
+ * +---------+-----------+-----------+----------------+--- ... ---+
+ * |CRC (4B) | Size (2B) | Type (1B) | Log number (4B)| Payload   |
+ * +---------+-----------+-----------+----------------+--- ... ---+
+ *
+ * Same as above, with the addition of
+ * Log number = 32bit log file number, so that we can distinguish between
+ * records written by the most recent log writer vs a previous one.
 */
 class Writer {
 public:
  // Create a writer that will append data to "*dest".
  // "*dest" must be initially empty.
  // "*dest" must remain live while this Writer is in use.
-  explicit Writer(unique_ptr<WritableFileWriter>&& dest);
+  explicit Writer(unique_ptr<WritableFileWriter>&& dest,
+                  uint64_t log_number, bool recycle_log_files);
  ~Writer();

  Status AddRecord(const Slice& slice);
@@ -74,6 +84,8 @@ class Writer {
 private:
  unique_ptr<WritableFileWriter> dest_;
  int block_offset_;       // Current offset in block
+  uint64_t log_number_;
+  bool recycle_log_files_;

  // crc32c values for all supported record types.  These are
  // pre-computed to reduce the overhead of computing the crc of the
@@ -21,6 +21,7 @@
 #include "rocksdb/iterator.h"
 #include "rocksdb/merge_operator.h"
 #include "rocksdb/slice_transform.h"
+#include "table/internal_iterator.h"
 #include "table/merger.h"
 #include "util/arena.h"
 #include "util/coding.h"
@@ -202,7 +203,7 @@ const char* EncodeKey(std::string* scratch, const Slice& target) {
  return scratch->data();
 }

-class MemTableIterator: public Iterator {
+class MemTableIterator : public InternalIterator {
 public:
  MemTableIterator(
      const MemTable& mem, const ReadOptions& read_options, Arena* arena)
@@ -230,10 +231,15 @@ class MemTableIterator: public Iterator {
  virtual void Seek(const Slice& k) override {
    PERF_TIMER_GUARD(seek_on_memtable_time);
    PERF_COUNTER_ADD(seek_on_memtable_count, 1);
-    if (bloom_ != nullptr &&
-        !bloom_->MayContain(prefix_extractor_->Transform(ExtractUserKey(k)))) {
-      valid_ = false;
-      return;
+    if (bloom_ != nullptr) {
+      if (!bloom_->MayContain(
+              prefix_extractor_->Transform(ExtractUserKey(k)))) {
+        PERF_COUNTER_ADD(bloom_memtable_miss_count, 1);
+        valid_ = false;
+        return;
+      } else {
+        PERF_COUNTER_ADD(bloom_memtable_hit_count, 1);
+      }
    }
    iter_->Seek(k, nullptr);
    valid_ = iter_->Valid();
@@ -280,7 +286,8 @@ class MemTableIterator: public Iterator {
  void operator=(const MemTableIterator&);
 };

-Iterator* MemTable::NewIterator(const ReadOptions& read_options, Arena* arena) {
+InternalIterator* MemTable::NewIterator(const ReadOptions& read_options,
+                                        Arena* arena) {
  assert(arena != nullptr);
  auto mem = arena->AllocateAligned(sizeof(MemTableIterator));
  return new (mem) MemTableIterator(*this, read_options, arena);
@@ -444,9 +451,10 @@ static bool SaveValue(void* arg, const char* entry) {
        *(s->found_final_value) = true;
        return false;
      }
-      case kTypeDeletion: {
+      case kTypeDeletion:
+      case kTypeSingleDeletion: {
        if (*(s->merge_in_progress)) {
-          assert(merge_operator);
+          assert(merge_operator != nullptr);
          *(s->status) = Status::OK();
          bool merge_success = false;
          {
@@ -507,12 +515,18 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
  Slice user_key = key.user_key();
  bool found_final_value = false;
  bool merge_in_progress = s->IsMergeInProgress();
-
-  if (prefix_bloom_ &&
-      !prefix_bloom_->MayContain(prefix_extractor_->Transform(user_key))) {
+  bool const may_contain =
+      nullptr == prefix_bloom_
+          ? false
+          : prefix_bloom_->MayContain(prefix_extractor_->Transform(user_key));
+  if (prefix_bloom_ && !may_contain) {
    // iter is null if prefix bloom says the key does not exist
+    PERF_COUNTER_ADD(bloom_memtable_miss_count, 1);
    *seq = kMaxSequenceNumber;
  } else {
+    if (prefix_bloom_) {
+      PERF_COUNTER_ADD(bloom_memtable_hit_count, 1);
+    }
    Saver saver;
    saver.status = s;
    saver.found_final_value = &found_final_value;
@@ -31,6 +31,7 @@ class Mutex;
 class MemTableIterator;
 class MergeContext;
 class WriteBuffer;
+class InternalIterator;

 struct MemTableOptions {
  explicit MemTableOptions(
@@ -140,7 +141,7 @@ class MemTable {
  // arena: If not null, the arena needs to be used to allocate the Iterator.
  //        Calling ~Iterator of the iterator will destroy all the states but
  //        those allocated in arena.
-  Iterator* NewIterator(const ReadOptions& read_options, Arena* arena);
+  InternalIterator* NewIterator(const ReadOptions& read_options, Arena* arena);

  // Add an entry into memtable that maps key to value at the
  // specified sequence number and with the specified type.
@@ -138,9 +138,9 @@ bool MemTableListVersion::GetFromList(std::list<MemTable*>* list,
  return false;
 }

-void MemTableListVersion::AddIterators(const ReadOptions& options,
-                                       std::vector<Iterator*>* iterator_list,
-                                       Arena* arena) {
+void MemTableListVersion::AddIterators(
+    const ReadOptions& options, std::vector<InternalIterator*>* iterator_list,
+    Arena* arena) {
  for (auto& m : memlist_) {
    iterator_list->push_back(m->NewIterator(options, arena));
  }
@@ -74,7 +74,8 @@ class MemTableListVersion {
  }

  void AddIterators(const ReadOptions& options,
-                    std::vector<Iterator*>* iterator_list, Arena* arena);
+                    std::vector<InternalIterator*>* iterator_list,
+                    Arena* arena);

  void AddIterators(const ReadOptions& options,
                    MergeIteratorBuilder* merge_iter_builder);
@@ -132,8 +132,6 @@ DEFINE_int64(seed, 0,
             "Seed base for random number generators. "
             "When 0 it is deterministic.");

-static rocksdb::Env* FLAGS_env = rocksdb::Env::Default();
-
 namespace rocksdb {

 namespace {
@@ -592,6 +590,7 @@ int main(int argc, char** argv) {
  std::unique_ptr<rocksdb::MemTableRepFactory> factory;
  if (FLAGS_memtablerep == "skiplist") {
    factory.reset(new rocksdb::SkipListFactory);
+#ifndef ROCKSDB_LITE
  } else if (FLAGS_memtablerep == "vector") {
    factory.reset(new rocksdb::VectorRepFactory);
  } else if (FLAGS_memtablerep == "hashskiplist") {
@@ -613,6 +612,7 @@ int main(int argc, char** argv) {
        static_cast<uint32_t>(FLAGS_hash_function_count)));
    options.prefix_extractor.reset(
        rocksdb::NewFixedPrefixTransform(FLAGS_prefix_length));
+#endif  // ROCKSDB_LITE
  } else {
    fprintf(stdout, "Unknown memtablerep: %s\n", FLAGS_memtablerep.c_str());
    exit(1);
@@ -12,9 +12,9 @@
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/merge_operator.h"
+#include "table/internal_iterator.h"
 #include "util/perf_context_imp.h"
 #include "util/statistics.h"
-#include "util/stop_watch.h"

 namespace rocksdb {

@@ -41,8 +41,7 @@ Status MergeHelper::TimedFullMerge(const Slice& key, const Slice* value,
  bool success =
      merge_operator->FullMerge(key, value, operands, result, logger);

-  RecordTick(statistics, MERGE_OPERATION_TOTAL_TIME,
-             env != nullptr ? timer.ElapsedNanos() : 0);
+  RecordTick(statistics, MERGE_OPERATION_TOTAL_TIME, timer.ElapsedNanosSafe());

  if (!success) {
    RecordTick(statistics, NUMBER_MERGE_FAILURES);
@@ -58,39 +57,47 @@ Status MergeHelper::TimedFullMerge(const Slice& key, const Slice* value,
 //       keys_ stores the list of keys encountered while merging.
 //       operands_ stores the list of merge operands encountered while merging.
 //       keys_[i] corresponds to operands_[i] for each i.
-Status MergeHelper::MergeUntil(Iterator* iter, const SequenceNumber stop_before,
-                               const bool at_bottom, Statistics* stats,
-                               Env* env_) {
+Status MergeHelper::MergeUntil(InternalIterator* iter,
+                               const SequenceNumber stop_before,
+                               const bool at_bottom) {
  // Get a copy of the internal key, before it's invalidated by iter->Next()
  // Also maintain the list of merge operands seen.
  assert(HasOperator());
  keys_.clear();
  operands_.clear();
-  keys_.push_front(iter->key().ToString());
-  operands_.push_front(iter->value().ToString());
  assert(user_merge_operator_);
+  bool first_key = true;

  // We need to parse the internal key again as the parsed key is
  // backed by the internal key!
  // Assume no internal key corruption as it has been successfully parsed
  // by the caller.
-  // Invariant: keys_.back() will not change. Hence, orig_ikey is always valid.
+  // original_key_is_iter variable is just caching the information:
+  // original_key_is_iter == (iter->key().ToString() == original_key)
+  bool original_key_is_iter = true;
+  std::string original_key = iter->key().ToString();
+  // Important:
+  // orig_ikey is backed by original_key if keys_.empty()
+  // orig_ikey is backed by keys_.back() if !keys_.empty()
  ParsedInternalKey orig_ikey;
-  ParseInternalKey(keys_.back(), &orig_ikey);
+  ParseInternalKey(original_key, &orig_ikey);

  Status s;
  bool hit_the_next_user_key = false;
-  for (iter->Next(); iter->Valid(); iter->Next()) {
+  for (; iter->Valid(); iter->Next(), original_key_is_iter = false) {
    ParsedInternalKey ikey;
-    assert(operands_.size() >= 1);        // Should be invariants!
    assert(keys_.size() == operands_.size());

    if (!ParseInternalKey(iter->key(), &ikey)) {
      // stop at corrupted key
      if (assert_valid_internal_key_) {
-        assert(!"corrupted internal key is not expected");
+        assert(!"Corrupted internal key not expected.");
+        return Status::Corruption("Corrupted internal key not expected.");
      }
      break;
+    } else if (first_key) {
+      assert(user_comparator_->Equal(ikey.user_key, orig_ikey.user_key));
+      first_key = false;
    } else if (!user_comparator_->Equal(ikey.user_key, orig_ikey.user_key)) {
      // hit a different user key, stop right here
      hit_the_next_user_key = true;
@@ -102,14 +109,31 @@ Status MergeHelper::MergeUntil(Iterator* iter, const SequenceNumber stop_before,

    // At this point we are guaranteed that we need to process this key.

-    assert(ikey.type <= kValueTypeForSeek);
+    assert(IsValueType(ikey.type));
    if (ikey.type != kTypeMerge) {
+      if (ikey.type != kTypeValue && ikey.type != kTypeDeletion) {
+        // Merges operands can only be used with puts and deletions, single
+        // deletions are not supported.
+        assert(false);
+        // release build doesn't have asserts, so we return error status
+        return Status::InvalidArgument(
+            " Merges operands can only be used with puts and deletions, single "
+            "deletions are not supported.");
+      }
+
      // hit a put/delete
      //   => merge the put value or a nullptr with operands_
      //   => store result in operands_.back() (and update keys_.back())
      //   => change the entry type to kTypeValue for keys_.back()
      // We are done! Success!
-      //
+
+      // If there are no operands, just return the Status::OK(). That will cause
+      // the compaction iterator to write out the key we're currently at, which
+      // is the put/delete we just encountered.
+      if (keys_.empty()) {
+        return Status::OK();
+      }
+
      // TODO(noetzli) If the merge operator returns false, we are currently
      // (almost) silently dropping the put/delete. That's probably not what we
      // want.
@@ -117,14 +141,14 @@ Status MergeHelper::MergeUntil(Iterator* iter, const SequenceNumber stop_before,
      const Slice* val_ptr = (kTypeValue == ikey.type) ? &val : nullptr;
      std::string merge_result;
      s = TimedFullMerge(ikey.user_key, val_ptr, operands_,
-                         user_merge_operator_, stats, env_, logger_,
+                         user_merge_operator_, stats_, env_, logger_,
                         &merge_result);

      // We store the result in keys_.back() and operands_.back()
      // if nothing went wrong (i.e.: no operand corruption on disk)
      if (s.ok()) {
        // The original key encountered
-        std::string original_key = std::move(keys_.back());
+        original_key = std::move(keys_.back());
        orig_ikey.type = kTypeValue;
        UpdateInternalKey(&original_key, orig_ikey.sequence, orig_ikey.type);
        keys_.clear();
@@ -138,18 +162,42 @@ Status MergeHelper::MergeUntil(Iterator* iter, const SequenceNumber stop_before,
      return s;
    } else {
      // hit a merge
+      //   => if there is a compaction filter, apply it.
      //   => merge the operand into the front of the operands_ list
-      //   => use the user's associative merge function to determine how.
+      //      if not filtered
      //   => then continue because we haven't yet seen a Put/Delete.
-      assert(!operands_.empty()); // Should have at least one element in it
-
-      // keep queuing keys and operands until we either meet a put / delete
+      //
+      // Keep queuing keys and operands until we either meet a put / delete
      // request or later did a partial merge.
-      keys_.push_front(iter->key().ToString());
-      operands_.push_front(iter->value().ToString());
+
+      Slice value_slice = iter->value();
+      // add an operand to the list if:
+      // 1) it's included in one of the snapshots. in that case we *must* write
+      // it out, no matter what compaction filter says
+      // 2) it's not filtered by a compaction filter
+      if (ikey.sequence <= latest_snapshot_ ||
+          !FilterMerge(orig_ikey.user_key, value_slice)) {
+        if (original_key_is_iter) {
+          // this is just an optimization that saves us one memcpy
+          keys_.push_front(std::move(original_key));
+        } else {
+          keys_.push_front(iter->key().ToString());
+        }
+        if (keys_.size() == 1) {
+          // we need to re-anchor the orig_ikey because it was anchored by
+          // original_key before
+          ParseInternalKey(keys_.back(), &orig_ikey);
+        }
+        operands_.push_front(value_slice.ToString());
+      }
    }
  }

+  if (operands_.size() == 0) {
+    // we filtered out all the merge operands
+    return Status::OK();
+  }
+
  // We are sure we have seen this key's entire history if we are at the
  // last level and exhausted all internal keys of this user key.
  // NOTE: !iter->Valid() does not necessarily mean we hit the
@@ -174,11 +222,13 @@ Status MergeHelper::MergeUntil(Iterator* iter, const SequenceNumber stop_before,
    assert(operands_.size() == keys_.size());
    std::string merge_result;
    s = TimedFullMerge(orig_ikey.user_key, nullptr, operands_,
-                       user_merge_operator_, stats, env_, logger_,
+                       user_merge_operator_, stats_, env_, logger_,
                       &merge_result);
    if (s.ok()) {
      // The original key encountered
-      std::string original_key = std::move(keys_.back());
+      // We are certain that keys_ is not empty here (see assertions couple of
+      // lines before).
+      original_key = std::move(keys_.back());
      orig_ikey.type = kTypeValue;
      UpdateInternalKey(&original_key, orig_ikey.sequence, orig_ikey.type);
      keys_.clear();
@@ -200,14 +250,14 @@ Status MergeHelper::MergeUntil(Iterator* iter, const SequenceNumber stop_before,
      bool merge_success = false;
      std::string merge_result;
      {
-        StopWatchNano timer(env_, stats != nullptr);
+        StopWatchNano timer(env_, stats_ != nullptr);
        PERF_TIMER_GUARD(merge_operator_time_nanos);
        merge_success = user_merge_operator_->PartialMergeMulti(
            orig_ikey.user_key,
            std::deque<Slice>(operands_.begin(), operands_.end()),
            &merge_result, logger_);
-        RecordTick(stats, MERGE_OPERATION_TOTAL_TIME,
-                   env_ != nullptr ? timer.ElapsedNanos() : 0);
+        RecordTick(stats_, MERGE_OPERATION_TOTAL_TIME,
+                   timer.ElapsedNanosSafe());
      }
      if (merge_success) {
        // Merging of operands (associative merge) was successful.
@@ -231,7 +281,6 @@ MergeOutputIterator::MergeOutputIterator(const MergeHelper* merge_helper)
 void MergeOutputIterator::SeekToFirst() {
  const auto& keys = merge_helper_->keys();
  const auto& values = merge_helper_->values();
-  assert(keys.size() > 0);
  assert(keys.size() == values.size());
  it_keys_ = keys.rbegin();
  it_values_ = values.rbegin();
@@ -242,4 +291,17 @@ void MergeOutputIterator::Next() {
  ++it_values_;
 }

+bool MergeHelper::FilterMerge(const Slice& user_key, const Slice& value_slice) {
+  if (compaction_filter_ == nullptr) {
+    return false;
+  }
+  if (stats_ != nullptr) {
+    filter_timer_.Start();
+  }
+  bool to_delete =
+      compaction_filter_->FilterMergeOperand(level_, user_key, value_slice);
+  total_filter_time_ += filter_timer_.ElapsedNanosSafe();
+  return to_delete;
+}
+
 } // namespace rocksdb
@@ -10,8 +10,10 @@
 #include <string>

 #include "db/dbformat.h"
+#include "rocksdb/compaction_filter.h"
 #include "rocksdb/env.h"
 #include "rocksdb/slice.h"
+#include "util/stop_watch.h"

 namespace rocksdb {

@@ -20,20 +22,30 @@ class Iterator;
 class Logger;
 class MergeOperator;
 class Statistics;
+class InternalIterator;

 class MergeHelper {
 public:
-  MergeHelper(const Comparator* user_comparator,
-              const MergeOperator* user_merge_operator, Logger* logger,
+  MergeHelper(Env* env, const Comparator* user_comparator,
+              const MergeOperator* user_merge_operator,
+              const CompactionFilter* compaction_filter, Logger* logger,
              unsigned min_partial_merge_operands,
-              bool assert_valid_internal_key)
-      : user_comparator_(user_comparator),
+              bool assert_valid_internal_key, SequenceNumber latest_snapshot,
+              int level = 0, Statistics* stats = nullptr)
+      : env_(env),
+        user_comparator_(user_comparator),
        user_merge_operator_(user_merge_operator),
+        compaction_filter_(compaction_filter),
        logger_(logger),
        min_partial_merge_operands_(min_partial_merge_operands),
        assert_valid_internal_key_(assert_valid_internal_key),
+        latest_snapshot_(latest_snapshot),
+        level_(level),
        keys_(),
-        operands_() {
+        operands_(),
+        filter_timer_(env_),
+        total_filter_time_(0U),
+        stats_(stats) {
    assert(user_comparator_ != nullptr);
  }

@@ -62,15 +74,22 @@ class MergeHelper {
  //                   0 means no restriction
  // at_bottom:   (IN) true if the iterator covers the bottem level, which means
  //                   we could reach the start of the history of this user key.
+  //
  // Returns one of the following statuses:
  // - OK: Entries were successfully merged.
  // - MergeInProgress: Put/Delete not encountered and unable to merge operands.
-  // - Corruption: Merge operator reported unsuccessful merge.
+  // - Corruption: Merge operator reported unsuccessful merge or a corrupted
+  //   key has been encountered and not expected (applies only when compiling
+  //   with asserts removed).
  //
  // REQUIRED: The first key in the input is not corrupted.
-  Status MergeUntil(Iterator* iter, const SequenceNumber stop_before = 0,
-                    const bool at_bottom = false, Statistics* stats = nullptr,
-                    Env* env_ = nullptr);
+  Status MergeUntil(InternalIterator* iter,
+                    const SequenceNumber stop_before = 0,
+                    const bool at_bottom = false);
+
+  // Filters a merge operand using the compaction filter specified
+  // in the constructor. Returns true if the operand should be filtered out.
+  bool FilterMerge(const Slice& user_key, const Slice& value_slice);

  // Query the merge result
  // These are valid until the next MergeUntil call
@@ -99,19 +118,28 @@ class MergeHelper {
  // TODO: Re-style this comment to be like the first one
  const std::deque<std::string>& keys() const { return keys_; }
  const std::deque<std::string>& values() const { return operands_; }
+  uint64_t TotalFilterTime() const { return total_filter_time_; }
  bool HasOperator() const { return user_merge_operator_ != nullptr; }

 private:
+  Env* env_;
  const Comparator* user_comparator_;
  const MergeOperator* user_merge_operator_;
+  const CompactionFilter* compaction_filter_;
  Logger* logger_;
  unsigned min_partial_merge_operands_;
  bool assert_valid_internal_key_; // enforce no internal key corruption?
+  SequenceNumber latest_snapshot_;
+  int level_;

  // the scratch area that holds the result of MergeUntil
  // valid up to the next MergeUntil call
  std::deque<std::string> keys_;    // Keeps track of the sequence of keys seen
  std::deque<std::string> operands_;  // Parallel with keys_; stores the values
+
+  StopWatchNano filter_timer_;
+  uint64_t total_filter_time_;
+  Statistics* stats_;
 };

 // MergeOutputIterator can be used to iterate over the result of a merge.
@@ -18,37 +18,24 @@ namespace rocksdb {

 class MergeHelperTest : public testing::Test {
 public:
-  MergeHelperTest() = default;
+  MergeHelperTest() { env_ = Env::Default(); }
+
  ~MergeHelperTest() = default;

-  Status RunUInt64MergeHelper(SequenceNumber stop_before, bool at_bottom) {
-    InitIterator();
-    merge_op_ = MergeOperators::CreateUInt64AddOperator();
-    merge_helper_.reset(new MergeHelper(BytewiseComparator(), merge_op_.get(),
-                                        nullptr, 2U, false));
-    return merge_helper_->MergeUntil(iter_.get(), stop_before, at_bottom,
-                                     nullptr, Env::Default());
-  }
-
-  Status RunStringAppendMergeHelper(SequenceNumber stop_before,
-                                    bool at_bottom) {
-    InitIterator();
-    merge_op_ = MergeOperators::CreateStringAppendTESTOperator();
-    merge_helper_.reset(new MergeHelper(BytewiseComparator(), merge_op_.get(),
-                                        nullptr, 2U, false));
-    return merge_helper_->MergeUntil(iter_.get(), stop_before, at_bottom,
-                                     nullptr, Env::Default());
-  }
-
-  std::string Key(const std::string& user_key, const SequenceNumber& seq,
-      const ValueType& t) {
-    return InternalKey(user_key, seq, t).Encode().ToString();
+  Status Run(SequenceNumber stop_before, bool at_bottom,
+             SequenceNumber latest_snapshot = 0) {
+    iter_.reset(new test::VectorIterator(ks_, vs_));
+    iter_->SeekToFirst();
+    merge_helper_.reset(new MergeHelper(env_, BytewiseComparator(),
+                                        merge_op_.get(), filter_.get(), nullptr,
+                                        2U, false, latest_snapshot));
+    return merge_helper_->MergeUntil(iter_.get(), stop_before, at_bottom);
  }

  void AddKeyVal(const std::string& user_key, const SequenceNumber& seq,
                 const ValueType& t, const std::string& val,
                 bool corrupt = false) {
-    InternalKey ikey = InternalKey(user_key, seq, t);
+    InternalKey ikey(user_key, seq, t);
    if (corrupt) {
      test::CorruptKeyType(&ikey);
    }
@@ -56,66 +43,63 @@ class MergeHelperTest : public testing::Test {
    vs_.push_back(val);
  }

-  void InitIterator() {
-    iter_.reset(new test::VectorIterator(ks_, vs_));
-    iter_->SeekToFirst();
-  }
-
-  std::string EncodeInt(uint64_t x) {
-    std::string result;
-    PutFixed64(&result, x);
-    return result;
-  }
-
+  Env* env_;
  std::unique_ptr<test::VectorIterator> iter_;
  std::shared_ptr<MergeOperator> merge_op_;
  std::unique_ptr<MergeHelper> merge_helper_;
  std::vector<std::string> ks_;
  std::vector<std::string> vs_;
+  std::unique_ptr<test::FilterNumber> filter_;
 };

 // If MergeHelper encounters a new key on the last level, we know that
 // the key has no more history and it can merge keys.
 TEST_F(MergeHelperTest, MergeAtBottomSuccess) {
-  AddKeyVal("a", 20, kTypeMerge, EncodeInt(1U));
-  AddKeyVal("a", 10, kTypeMerge, EncodeInt(3U));
-  AddKeyVal("b", 10, kTypeMerge, EncodeInt(4U));  // <- Iterator after merge
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();

-  ASSERT_TRUE(RunUInt64MergeHelper(0, true).ok());
+  AddKeyVal("a", 20, kTypeMerge, test::EncodeInt(1U));
+  AddKeyVal("a", 10, kTypeMerge, test::EncodeInt(3U));
+  AddKeyVal("b", 10, kTypeMerge, test::EncodeInt(4U));  // <- iter_ after merge
+
+  ASSERT_TRUE(Run(0, true).ok());
  ASSERT_EQ(ks_[2], iter_->key());
-  ASSERT_EQ(Key("a", 20, kTypeValue), merge_helper_->keys()[0]);
-  ASSERT_EQ(EncodeInt(4U), merge_helper_->values()[0]);
+  ASSERT_EQ(test::KeyStr("a", 20, kTypeValue), merge_helper_->keys()[0]);
+  ASSERT_EQ(test::EncodeInt(4U), merge_helper_->values()[0]);
  ASSERT_EQ(1U, merge_helper_->keys().size());
  ASSERT_EQ(1U, merge_helper_->values().size());
 }

 // Merging with a value results in a successful merge.
 TEST_F(MergeHelperTest, MergeValue) {
-  AddKeyVal("a", 40, kTypeMerge, EncodeInt(1U));
-  AddKeyVal("a", 30, kTypeMerge, EncodeInt(3U));
-  AddKeyVal("a", 20, kTypeValue, EncodeInt(4U));  // <- Iterator after merge
-  AddKeyVal("a", 10, kTypeMerge, EncodeInt(1U));
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();

-  ASSERT_TRUE(RunUInt64MergeHelper(0, false).ok());
+  AddKeyVal("a", 40, kTypeMerge, test::EncodeInt(1U));
+  AddKeyVal("a", 30, kTypeMerge, test::EncodeInt(3U));
+  AddKeyVal("a", 20, kTypeValue, test::EncodeInt(4U));  // <- iter_ after merge
+  AddKeyVal("a", 10, kTypeMerge, test::EncodeInt(1U));
+
+  ASSERT_TRUE(Run(0, false).ok());
  ASSERT_EQ(ks_[3], iter_->key());
-  ASSERT_EQ(Key("a", 40, kTypeValue), merge_helper_->keys()[0]);
-  ASSERT_EQ(EncodeInt(8U), merge_helper_->values()[0]);
+  ASSERT_EQ(test::KeyStr("a", 40, kTypeValue), merge_helper_->keys()[0]);
+  ASSERT_EQ(test::EncodeInt(8U), merge_helper_->values()[0]);
  ASSERT_EQ(1U, merge_helper_->keys().size());
  ASSERT_EQ(1U, merge_helper_->values().size());
 }

 // Merging stops before a snapshot.
 TEST_F(MergeHelperTest, SnapshotBeforeValue) {
-  AddKeyVal("a", 50, kTypeMerge, EncodeInt(1U));
-  AddKeyVal("a", 40, kTypeMerge, EncodeInt(3U));  // <- Iterator after merge
-  AddKeyVal("a", 30, kTypeMerge, EncodeInt(1U));
-  AddKeyVal("a", 20, kTypeValue, EncodeInt(4U));
-  AddKeyVal("a", 10, kTypeMerge, EncodeInt(1U));
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();

-  ASSERT_TRUE(RunUInt64MergeHelper(31, true).IsMergeInProgress());
+  AddKeyVal("a", 50, kTypeMerge, test::EncodeInt(1U));
+  AddKeyVal("a", 40, kTypeMerge, test::EncodeInt(3U));  // <- iter_ after merge
+  AddKeyVal("a", 30, kTypeMerge, test::EncodeInt(1U));
+  AddKeyVal("a", 20, kTypeValue, test::EncodeInt(4U));
+  AddKeyVal("a", 10, kTypeMerge, test::EncodeInt(1U));
+
+  ASSERT_TRUE(Run(31, true).IsMergeInProgress());
  ASSERT_EQ(ks_[2], iter_->key());
-  ASSERT_EQ(Key("a", 50, kTypeMerge), merge_helper_->keys()[0]);
-  ASSERT_EQ(EncodeInt(4U), merge_helper_->values()[0]);
+  ASSERT_EQ(test::KeyStr("a", 50, kTypeMerge), merge_helper_->keys()[0]);
+  ASSERT_EQ(test::EncodeInt(4U), merge_helper_->values()[0]);
  ASSERT_EQ(1U, merge_helper_->keys().size());
  ASSERT_EQ(1U, merge_helper_->values().size());
 }
@@ -123,15 +107,17 @@ TEST_F(MergeHelperTest, SnapshotBeforeValue) {
 // MergeHelper preserves the operand stack for merge operators that
 // cannot do a partial merge.
 TEST_F(MergeHelperTest, NoPartialMerge) {
+  merge_op_ = MergeOperators::CreateStringAppendTESTOperator();
+
  AddKeyVal("a", 50, kTypeMerge, "v2");
-  AddKeyVal("a", 40, kTypeMerge, "v");  // <- Iterator after merge
+  AddKeyVal("a", 40, kTypeMerge, "v");  // <- iter_ after merge
  AddKeyVal("a", 30, kTypeMerge, "v");

-  ASSERT_TRUE(RunStringAppendMergeHelper(31, true).IsMergeInProgress());
+  ASSERT_TRUE(Run(31, true).IsMergeInProgress());
  ASSERT_EQ(ks_[2], iter_->key());
-  ASSERT_EQ(Key("a", 40, kTypeMerge), merge_helper_->keys()[0]);
+  ASSERT_EQ(test::KeyStr("a", 40, kTypeMerge), merge_helper_->keys()[0]);
  ASSERT_EQ("v", merge_helper_->values()[0]);
-  ASSERT_EQ(Key("a", 50, kTypeMerge), merge_helper_->keys()[1]);
+  ASSERT_EQ(test::KeyStr("a", 50, kTypeMerge), merge_helper_->keys()[1]);
  ASSERT_EQ("v2", merge_helper_->values()[1]);
  ASSERT_EQ(2U, merge_helper_->keys().size());
  ASSERT_EQ(2U, merge_helper_->values().size());
@@ -139,44 +125,162 @@ TEST_F(MergeHelperTest, NoPartialMerge) {

 // A single operand can not be merged.
 TEST_F(MergeHelperTest, SingleOperand) {
-  AddKeyVal("a", 50, kTypeMerge, EncodeInt(1U));
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();

-  ASSERT_TRUE(RunUInt64MergeHelper(31, true).IsMergeInProgress());
+  AddKeyVal("a", 50, kTypeMerge, test::EncodeInt(1U));
+
+  ASSERT_TRUE(Run(31, true).IsMergeInProgress());
  ASSERT_FALSE(iter_->Valid());
-  ASSERT_EQ(Key("a", 50, kTypeMerge), merge_helper_->keys()[0]);
-  ASSERT_EQ(EncodeInt(1U), merge_helper_->values()[0]);
+  ASSERT_EQ(test::KeyStr("a", 50, kTypeMerge), merge_helper_->keys()[0]);
+  ASSERT_EQ(test::EncodeInt(1U), merge_helper_->values()[0]);
  ASSERT_EQ(1U, merge_helper_->keys().size());
  ASSERT_EQ(1U, merge_helper_->values().size());
 }

 // Merging with a deletion turns the deletion into a value
 TEST_F(MergeHelperTest, MergeDeletion) {
-  AddKeyVal("a", 30, kTypeMerge, EncodeInt(3U));
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();
+
+  AddKeyVal("a", 30, kTypeMerge, test::EncodeInt(3U));
  AddKeyVal("a", 20, kTypeDeletion, "");

-  ASSERT_TRUE(RunUInt64MergeHelper(15, false).ok());
+  ASSERT_TRUE(Run(15, false).ok());
  ASSERT_FALSE(iter_->Valid());
-  ASSERT_EQ(Key("a", 30, kTypeValue), merge_helper_->keys()[0]);
-  ASSERT_EQ(EncodeInt(3U), merge_helper_->values()[0]);
+  ASSERT_EQ(test::KeyStr("a", 30, kTypeValue), merge_helper_->keys()[0]);
+  ASSERT_EQ(test::EncodeInt(3U), merge_helper_->values()[0]);
  ASSERT_EQ(1U, merge_helper_->keys().size());
  ASSERT_EQ(1U, merge_helper_->values().size());
 }

 // The merge helper stops upon encountering a corrupt key
 TEST_F(MergeHelperTest, CorruptKey) {
-  AddKeyVal("a", 30, kTypeMerge, EncodeInt(3U));
-  AddKeyVal("a", 25, kTypeMerge, EncodeInt(1U));
-  // Corrupt key
-  AddKeyVal("a", 20, kTypeDeletion, "", true);  // <- Iterator after merge
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();

-  ASSERT_TRUE(RunUInt64MergeHelper(15, false).IsMergeInProgress());
+  AddKeyVal("a", 30, kTypeMerge, test::EncodeInt(3U));
+  AddKeyVal("a", 25, kTypeMerge, test::EncodeInt(1U));
+  // Corrupt key
+  AddKeyVal("a", 20, kTypeDeletion, "", true);  // <- iter_ after merge
+
+  ASSERT_TRUE(Run(15, false).IsMergeInProgress());
  ASSERT_EQ(ks_[2], iter_->key());
-  ASSERT_EQ(Key("a", 30, kTypeMerge), merge_helper_->keys()[0]);
-  ASSERT_EQ(EncodeInt(4U), merge_helper_->values()[0]);
+  ASSERT_EQ(test::KeyStr("a", 30, kTypeMerge), merge_helper_->keys()[0]);
+  ASSERT_EQ(test::EncodeInt(4U), merge_helper_->values()[0]);
  ASSERT_EQ(1U, merge_helper_->keys().size());
  ASSERT_EQ(1U, merge_helper_->values().size());
 }

+// The compaction filter is called on every merge operand
+TEST_F(MergeHelperTest, FilterMergeOperands) {
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();
+  filter_.reset(new test::FilterNumber(5U));
+
+  AddKeyVal("a", 30, kTypeMerge, test::EncodeInt(3U));
+  AddKeyVal("a", 29, kTypeMerge, test::EncodeInt(5U));  // Filtered
+  AddKeyVal("a", 28, kTypeMerge, test::EncodeInt(3U));
+  AddKeyVal("a", 27, kTypeMerge, test::EncodeInt(1U));
+  AddKeyVal("a", 26, kTypeMerge, test::EncodeInt(5U));  // Filtered
+  AddKeyVal("a", 25, kTypeValue, test::EncodeInt(1U));
+
+  ASSERT_TRUE(Run(15, false).ok());
+  ASSERT_FALSE(iter_->Valid());
+  MergeOutputIterator merge_output_iter(merge_helper_.get());
+  merge_output_iter.SeekToFirst();
+  ASSERT_EQ(test::KeyStr("a", 30, kTypeValue),
+            merge_output_iter.key().ToString());
+  ASSERT_EQ(test::EncodeInt(8U), merge_output_iter.value().ToString());
+  merge_output_iter.Next();
+  ASSERT_FALSE(merge_output_iter.Valid());
+}
+
+TEST_F(MergeHelperTest, FilterAllMergeOperands) {
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();
+  filter_.reset(new test::FilterNumber(5U));
+
+  AddKeyVal("a", 30, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("a", 29, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("a", 28, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("a", 27, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("a", 26, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("a", 25, kTypeMerge, test::EncodeInt(5U));
+
+  // filtered out all
+  ASSERT_TRUE(Run(15, false).ok());
+  ASSERT_FALSE(iter_->Valid());
+  MergeOutputIterator merge_output_iter(merge_helper_.get());
+  merge_output_iter.SeekToFirst();
+  ASSERT_FALSE(merge_output_iter.Valid());
+
+  // we have one operand that will survive because it's a delete
+  AddKeyVal("a", 24, kTypeDeletion, test::EncodeInt(5U));
+  AddKeyVal("b", 23, kTypeValue, test::EncodeInt(5U));
+  ASSERT_TRUE(Run(15, true).ok());
+  merge_output_iter = MergeOutputIterator(merge_helper_.get());
+  ASSERT_TRUE(iter_->Valid());
+  merge_output_iter.SeekToFirst();
+  ASSERT_FALSE(merge_output_iter.Valid());
+
+  // when all merge operands are filtered out, we leave the iterator pointing to
+  // the Put/Delete that survived
+  ASSERT_EQ(test::KeyStr("a", 24, kTypeDeletion), iter_->key().ToString());
+  ASSERT_EQ(test::EncodeInt(5U), iter_->value().ToString());
+}
+
+// Make sure that merge operands are filtered at the beginning
+TEST_F(MergeHelperTest, FilterFirstMergeOperand) {
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();
+  filter_.reset(new test::FilterNumber(5U));
+
+  AddKeyVal("a", 31, kTypeMerge, test::EncodeInt(5U));  // Filtered
+  AddKeyVal("a", 30, kTypeMerge, test::EncodeInt(5U));  // Filtered
+  AddKeyVal("a", 29, kTypeMerge, test::EncodeInt(2U));
+  AddKeyVal("a", 28, kTypeMerge, test::EncodeInt(1U));
+  AddKeyVal("a", 27, kTypeMerge, test::EncodeInt(3U));
+  AddKeyVal("a", 26, kTypeMerge, test::EncodeInt(5U));  // Filtered
+  AddKeyVal("a", 25, kTypeMerge, test::EncodeInt(5U));  // Filtered
+  AddKeyVal("b", 24, kTypeValue, test::EncodeInt(5U));  // next user key
+
+  ASSERT_OK(Run(15, true));
+  ASSERT_TRUE(iter_->Valid());
+  MergeOutputIterator merge_output_iter(merge_helper_.get());
+  merge_output_iter.SeekToFirst();
+  // sequence number is 29 here, because the first merge operand got filtered
+  // out
+  ASSERT_EQ(test::KeyStr("a", 29, kTypeValue),
+            merge_output_iter.key().ToString());
+  ASSERT_EQ(test::EncodeInt(6U), merge_output_iter.value().ToString());
+  merge_output_iter.Next();
+  ASSERT_FALSE(merge_output_iter.Valid());
+
+  // make sure that we're passing user keys into the filter
+  ASSERT_EQ("a", filter_->last_merge_operand_key());
+}
+
+// Make sure that merge operands are not filtered out if there's a snapshot
+// pointing at them
+TEST_F(MergeHelperTest, DontFilterMergeOperandsBeforeSnapshotTest) {
+  merge_op_ = MergeOperators::CreateUInt64AddOperator();
+  filter_.reset(new test::FilterNumber(5U));
+
+  AddKeyVal("a", 31, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("a", 30, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("a", 29, kTypeMerge, test::EncodeInt(2U));
+  AddKeyVal("a", 28, kTypeMerge, test::EncodeInt(1U));
+  AddKeyVal("a", 27, kTypeMerge, test::EncodeInt(3U));
+  AddKeyVal("a", 26, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("a", 25, kTypeMerge, test::EncodeInt(5U));
+  AddKeyVal("b", 24, kTypeValue, test::EncodeInt(5U));
+
+  ASSERT_OK(Run(15, true, 32));
+  ASSERT_TRUE(iter_->Valid());
+  MergeOutputIterator merge_output_iter(merge_helper_.get());
+  merge_output_iter.SeekToFirst();
+  ASSERT_EQ(test::KeyStr("a", 31, kTypeValue),
+            merge_output_iter.key().ToString());
+  ASSERT_EQ(test::EncodeInt(26U), merge_output_iter.value().ToString());
+  merge_output_iter.Next();
+  ASSERT_FALSE(merge_output_iter.Valid());
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -5,6 +5,7 @@
 //
 #include <algorithm>
 #include <iostream>
+#include <thread>
 #include <vector>

 #include "rocksdb/db.h"
@@ -444,6 +445,7 @@ void ProfileQueries(bool enabled_time = false) {
  }
 }

+#ifndef ROCKSDB_LITE
 TEST_F(PerfContextTest, KeyComparisonCount) {
  SetPerfLevel(kEnableCount);
  ProfileQueries();
@@ -454,6 +456,7 @@ TEST_F(PerfContextTest, KeyComparisonCount) {
  SetPerfLevel(kEnableTime);
  ProfileQueries(true);
 }
+#endif  // ROCKSDB_LITE

 // make perf_context_test
 // export ROCKSDB_TESTS=PerfContextTest.SeekKeyComparison
@@ -539,6 +542,49 @@ TEST_F(PerfContextTest, SeekKeyComparison) {
  }
 }

+TEST_F(PerfContextTest, DBMutexLockCounter) {
+  SetPerfLevel(kEnableTime);
+  int stats_code[] = {0, static_cast<int>(DB_MUTEX_WAIT_MICROS)};
+  for (int c = 0; c < 2; ++c) {
+    InstrumentedMutex mutex(nullptr, Env::Default(), stats_code[c]);
+    mutex.Lock();
+    std::thread child_thread([&] {
+      SetPerfLevel(kEnableTime);
+      perf_context.Reset();
+      ASSERT_EQ(perf_context.db_mutex_lock_nanos, 0);
+      mutex.Lock();
+      mutex.Unlock();
+      if (stats_code[c] == DB_MUTEX_WAIT_MICROS) {
+        // increment the counter only when it's a DB Mutex
+        ASSERT_GT(perf_context.db_mutex_lock_nanos, 0);
+      } else {
+        ASSERT_EQ(perf_context.db_mutex_lock_nanos, 0);
+      }
+    });
+    Env::Default()->SleepForMicroseconds(100);
+    mutex.Unlock();
+    child_thread.join();
+  }
+}
+
+TEST_F(PerfContextTest, FalseDBMutexWait) {
+  SetPerfLevel(kEnableTime);
+  int stats_code[] = {0, static_cast<int>(DB_MUTEX_WAIT_MICROS)};
+  for (int c = 0; c < 2; ++c) {
+    InstrumentedMutex mutex(nullptr, Env::Default(), stats_code[c]);
+    InstrumentedCondVar lock(&mutex);
+    perf_context.Reset();
+    mutex.Lock();
+    lock.TimedWait(100);
+    mutex.Unlock();
+    if (stats_code[c] == static_cast<int>(DB_MUTEX_WAIT_MICROS)) {
+      // increment the counter only when it's a DB Mutex
+      ASSERT_GT(perf_context.db_condition_wait_nanos, 0);
+    } else {
+      ASSERT_EQ(perf_context.db_condition_wait_nanos, 0);
+    }
+  }
+}
 }

 int main(int argc, char** argv) {
@@ -27,6 +27,7 @@
 #include "rocksdb/table.h"
 #include "table/meta_blocks.h"
 #include "table/bloom_block.h"
+#include "table/table_builder.h"
 #include "table/plain_table_factory.h"
 #include "table/plain_table_reader.h"
 #include "util/hash.h"
@@ -41,28 +42,33 @@ using std::unique_ptr;

 namespace rocksdb {

-class PlainTableDBTest : public testing::Test {
+class PlainTableDBTest : public testing::Test,
+                         public testing::WithParamInterface<bool> {
 protected:
 private:
  std::string dbname_;
  Env* env_;
  DB* db_;

+  bool mmap_mode_;
  Options last_options_;

 public:
-  PlainTableDBTest() : env_(Env::Default()) {
-    dbname_ = test::TmpDir() + "/plain_table_db_test";
-    EXPECT_OK(DestroyDB(dbname_, Options()));
-    db_ = nullptr;
-    Reopen();
-  }
+  PlainTableDBTest() : env_(Env::Default()) {}

  ~PlainTableDBTest() {
    delete db_;
    EXPECT_OK(DestroyDB(dbname_, Options()));
  }

+  void SetUp() override {
+    mmap_mode_ = GetParam();
+    dbname_ = test::TmpDir() + "/plain_table_db_test";
+    EXPECT_OK(DestroyDB(dbname_, Options()));
+    db_ = nullptr;
+    Reopen();
+  }
+
  // Return the current option configuration.
  Options CurrentOptions() {
    Options options;
@@ -81,7 +87,7 @@ class PlainTableDBTest : public testing::Test {
    options.memtable_factory.reset(NewHashLinkListRepFactory(4, 0, 3, true));

    options.prefix_extractor.reset(NewFixedPrefixTransform(8));
-    options.allow_mmap_reads = true;
+    options.allow_mmap_reads = mmap_mode_;
    return options;
  }

@@ -186,7 +192,7 @@ class PlainTableDBTest : public testing::Test {
  }
 };

-TEST_F(PlainTableDBTest, Empty) {
+TEST_P(PlainTableDBTest, Empty) {
  ASSERT_TRUE(dbfull() != nullptr);
  ASSERT_EQ("NOT_FOUND", Get("0000000000000foo"));
 }
@@ -207,7 +213,7 @@ class TestPlainTableReader : public PlainTableReader {
      : PlainTableReader(ioptions, std::move(file), env_options, icomparator,
                         encoding_type, file_size, table_properties),
        expect_bloom_not_match_(expect_bloom_not_match) {
-    Status s = MmapDataFile();
+    Status s = MmapDataIfNeeded();
    EXPECT_TRUE(s.ok());

    s = PopulateIndex(const_cast<TableProperties*>(table_properties),
@@ -256,28 +262,29 @@ class TestPlainTableFactory : public PlainTableFactory {
        store_index_in_file_(options.store_index_in_file),
        expect_bloom_not_match_(expect_bloom_not_match) {}

-  Status NewTableReader(const ImmutableCFOptions& ioptions,
-                        const EnvOptions& env_options,
-                        const InternalKeyComparator& internal_comparator,
+  Status NewTableReader(const TableReaderOptions& table_reader_options,
                        unique_ptr<RandomAccessFileReader>&& file,
                        uint64_t file_size,
                        unique_ptr<TableReader>* table) const override {
    TableProperties* props = nullptr;
-    auto s = ReadTableProperties(file.get(), file_size, kPlainTableMagicNumber,
-                                 ioptions.env, ioptions.info_log, &props);
+    auto s =
+        ReadTableProperties(file.get(), file_size, kPlainTableMagicNumber,
+                            table_reader_options.ioptions.env,
+                            table_reader_options.ioptions.info_log, &props);
    EXPECT_TRUE(s.ok());

    if (store_index_in_file_) {
      BlockHandle bloom_block_handle;
      s = FindMetaBlock(file.get(), file_size, kPlainTableMagicNumber,
-                        ioptions.env, BloomBlockBuilder::kBloomBlock,
-                        &bloom_block_handle);
+                        table_reader_options.ioptions.env,
+                        BloomBlockBuilder::kBloomBlock, &bloom_block_handle);
      EXPECT_TRUE(s.ok());

      BlockHandle index_block_handle;
-      s = FindMetaBlock(
-          file.get(), file_size, kPlainTableMagicNumber, ioptions.env,
-          PlainTableIndexBuilder::kPlainTableIndexBlock, &index_block_handle);
+      s = FindMetaBlock(file.get(), file_size, kPlainTableMagicNumber,
+                        table_reader_options.ioptions.env,
+                        PlainTableIndexBuilder::kPlainTableIndexBlock,
+                        &index_block_handle);
      EXPECT_TRUE(s.ok());
    }

@@ -289,9 +296,10 @@ class TestPlainTableFactory : public PlainTableFactory {
        DecodeFixed32(encoding_type_prop->second.c_str()));

    std::unique_ptr<PlainTableReader> new_reader(new TestPlainTableReader(
-        env_options, internal_comparator, encoding_type, file_size,
+        table_reader_options.env_options,
+        table_reader_options.internal_comparator, encoding_type, file_size,
        bloom_bits_per_key_, hash_table_ratio_, index_sparseness_, props,
-        std::move(file), ioptions, expect_bloom_not_match_,
+        std::move(file), table_reader_options.ioptions, expect_bloom_not_match_,
        store_index_in_file_));

    *table = std::move(new_reader);
@@ -306,7 +314,7 @@ class TestPlainTableFactory : public PlainTableFactory {
  bool* expect_bloom_not_match_;
 };

-TEST_F(PlainTableDBTest, Flush) {
+TEST_P(PlainTableDBTest, Flush) {
  for (size_t huge_page_tlb_size = 0; huge_page_tlb_size <= 2 * 1024 * 1024;
       huge_page_tlb_size += 2 * 1024 * 1024) {
    for (EncodingType encoding_type : {kPlain, kPrefix}) {
@@ -393,7 +401,7 @@ TEST_F(PlainTableDBTest, Flush) {
  }
 }

-TEST_F(PlainTableDBTest, Flush2) {
+TEST_P(PlainTableDBTest, Flush2) {
  for (size_t huge_page_tlb_size = 0; huge_page_tlb_size <= 2 * 1024 * 1024;
       huge_page_tlb_size += 2 * 1024 * 1024) {
    for (EncodingType encoding_type : {kPlain, kPrefix}) {
@@ -473,7 +481,7 @@ TEST_F(PlainTableDBTest, Flush2) {
  }
 }

-TEST_F(PlainTableDBTest, Iterator) {
+TEST_P(PlainTableDBTest, Iterator) {
  for (size_t huge_page_tlb_size = 0; huge_page_tlb_size <= 2 * 1024 * 1024;
       huge_page_tlb_size += 2 * 1024 * 1024) {
    for (EncodingType encoding_type : {kPlain, kPrefix}) {
@@ -607,7 +615,7 @@ std::string MakeLongKey(size_t length, char c) {
 }
 }  // namespace

-TEST_F(PlainTableDBTest, IteratorLargeKeys) {
+TEST_P(PlainTableDBTest, IteratorLargeKeys) {
  Options options = CurrentOptions();

  PlainTableOptions plain_table_options;
@@ -657,7 +665,7 @@ std::string MakeLongKeyWithPrefix(size_t length, char c) {
 }
 }  // namespace

-TEST_F(PlainTableDBTest, IteratorLargeKeysWithPrefix) {
+TEST_P(PlainTableDBTest, IteratorLargeKeysWithPrefix) {
  Options options = CurrentOptions();

  PlainTableOptions plain_table_options;
@@ -699,7 +707,7 @@ TEST_F(PlainTableDBTest, IteratorLargeKeysWithPrefix) {
  delete iter;
 }

-TEST_F(PlainTableDBTest, IteratorReverseSuffixComparator) {
+TEST_P(PlainTableDBTest, IteratorReverseSuffixComparator) {
  Options options = CurrentOptions();
  options.create_if_missing = true;
  // Set only one bucket to force bucket conflict.
@@ -768,7 +776,7 @@ TEST_F(PlainTableDBTest, IteratorReverseSuffixComparator) {
  delete iter;
 }

-TEST_F(PlainTableDBTest, HashBucketConflict) {
+TEST_P(PlainTableDBTest, HashBucketConflict) {
  for (size_t huge_page_tlb_size = 0; huge_page_tlb_size <= 2 * 1024 * 1024;
       huge_page_tlb_size += 2 * 1024 * 1024) {
    for (unsigned char i = 1; i <= 3; i++) {
@@ -861,7 +869,7 @@ TEST_F(PlainTableDBTest, HashBucketConflict) {
  }
 }

-TEST_F(PlainTableDBTest, HashBucketConflictReverseSuffixComparator) {
+TEST_P(PlainTableDBTest, HashBucketConflictReverseSuffixComparator) {
  for (size_t huge_page_tlb_size = 0; huge_page_tlb_size <= 2 * 1024 * 1024;
       huge_page_tlb_size += 2 * 1024 * 1024) {
    for (unsigned char i = 1; i <= 3; i++) {
@@ -954,7 +962,7 @@ TEST_F(PlainTableDBTest, HashBucketConflictReverseSuffixComparator) {
  }
 }

-TEST_F(PlainTableDBTest, NonExistingKeyToNonEmptyBucket) {
+TEST_P(PlainTableDBTest, NonExistingKeyToNonEmptyBucket) {
  Options options = CurrentOptions();
  options.create_if_missing = true;
  // Set only one bucket to force bucket conflict.
@@ -1010,7 +1018,7 @@ static std::string RandomString(Random* rnd, int len) {
  return r;
 }

-TEST_F(PlainTableDBTest, CompactionTrigger) {
+TEST_P(PlainTableDBTest, CompactionTrigger) {
  Options options = CurrentOptions();
  options.write_buffer_size = 120 << 10;  // 100KB
  options.num_levels = 3;
@@ -1045,7 +1053,7 @@ TEST_F(PlainTableDBTest, CompactionTrigger) {
  ASSERT_EQ(NumTableFilesAtLevel(1), 1);
 }

-TEST_F(PlainTableDBTest, AdaptiveTable) {
+TEST_P(PlainTableDBTest, AdaptiveTable) {
  Options options = CurrentOptions();
  options.create_if_missing = true;

@@ -1088,6 +1096,8 @@ TEST_F(PlainTableDBTest, AdaptiveTable) {
  ASSERT_NE("v5", Get("3000000000000bar"));
 }

+INSTANTIATE_TEST_CASE_P(PlainTableDBTest, PlainTableDBTest, ::testing::Bool());
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -8,8 +8,8 @@
 #ifndef GFLAGS
 #include <cstdio>
 int main() {
-  fprintf(stderr, "Please install gflags to run rocksdb tools\n");
-  return 1;
+  fprintf(stderr, "Please install gflags to run this test... Skipping...\n");
+  return 0;
 }
 #else

@@ -20,9 +20,11 @@ int main() {
 #include <gflags/gflags.h>
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
+#include "rocksdb/filter_policy.h"
 #include "rocksdb/perf_context.h"
 #include "rocksdb/slice_transform.h"
 #include "rocksdb/memtablerep.h"
+#include "rocksdb/table.h"
 #include "util/histogram.h"
 #include "util/stop_watch.h"
 #include "util/string_util.h"
@@ -163,6 +165,12 @@ class PrefixTest : public testing::Test {
    options.memtable_prefix_bloom_huge_page_tlb_size =
        FLAGS_memtable_prefix_bloom_huge_page_tlb_size;

+    options.prefix_extractor.reset(NewFixedPrefixTransform(8));
+    BlockBasedTableOptions bbto;
+    bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+    bbto.whole_key_filtering = false;
+    options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+
    Status s = DB::Open(options, kDbName,  &db);
    EXPECT_OK(s);
    return std::shared_ptr<DB>(db);
@@ -393,6 +401,55 @@ TEST_F(PrefixTest, TestResult) {
  }
 }

+// Show results in prefix
+TEST_F(PrefixTest, PrefixValid) {
+  for (int num_buckets = 1; num_buckets <= 2; num_buckets++) {
+    FirstOption();
+    while (NextOptions(num_buckets)) {
+      std::cout << "*** Mem table: " << options.memtable_factory->Name()
+                << " number of buckets: " << num_buckets << std::endl;
+      DestroyDB(kDbName, Options());
+      auto db = OpenDb();
+      WriteOptions write_options;
+      ReadOptions read_options;
+
+      // Insert keys with common prefix and one key with different
+      Slice v16("v16");
+      Slice v17("v17");
+      Slice v18("v18");
+      Slice v19("v19");
+      PutKey(db.get(), write_options, 12345, 6, v16);
+      PutKey(db.get(), write_options, 12345, 7, v17);
+      PutKey(db.get(), write_options, 12345, 8, v18);
+      PutKey(db.get(), write_options, 12345, 9, v19);
+      PutKey(db.get(), write_options, 12346, 8, v16);
+      db->Flush(FlushOptions());
+      db->Delete(write_options, TestKeyToSlice(TestKey(12346, 8)));
+      db->Flush(FlushOptions());
+      read_options.prefix_same_as_start = true;
+      std::unique_ptr<Iterator> iter(db->NewIterator(read_options));
+      SeekIterator(iter.get(), 12345, 6);
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_TRUE(v16 == iter->value());
+
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_TRUE(v17 == iter->value());
+
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_TRUE(v18 == iter->value());
+
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_TRUE(v19 == iter->value());
+      iter->Next();
+      ASSERT_FALSE(iter->Valid());
+      ASSERT_EQ(kNotFoundResult, Get(db.get(), read_options, 12346, 8));
+    }
+  }
+}
+
 TEST_F(PrefixTest, DynamicPrefixIterator) {
  while (NextOptions(FLAGS_bucket_count)) {
    std::cout << "*** Mem table: " << options.memtable_factory->Name()
@@ -81,8 +81,8 @@
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
 #include "rocksdb/immutable_options.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/file_reader_writer.h"
-#include "util/scoped_arena_iterator.h"

 namespace rocksdb {

@@ -249,8 +249,8 @@ class Repairer {
    // corruptions cause entire commits to be skipped instead of
    // propagating bad information (like overly large sequence
    // numbers).
-    log::Reader reader(std::move(lfile_reader), &reporter,
-                       true /*enable checksum*/, 0 /*initial_offset*/);
+    log::Reader reader(options_.info_log, std::move(lfile_reader), &reporter,
+                       true /*enable checksum*/, 0 /*initial_offset*/, log);

    // Read all the records and add to a memtable
    std::string scratch;
@@ -290,10 +290,11 @@ class Repairer {
      ro.total_order_seek = true;
      Arena arena;
      ScopedArenaIterator iter(mem->NewIterator(ro, &arena));
-      status = BuildTable(dbname_, env_, ioptions_, env_options_, table_cache_,
-                          iter.get(), &meta, icmp_,
-                          &int_tbl_prop_collector_factories_, {},
-                          kNoCompression, CompressionOptions(), false, nullptr);
+      status = BuildTable(
+          dbname_, env_, ioptions_, env_options_, table_cache_, iter.get(),
+          &meta, icmp_, &int_tbl_prop_collector_factories_,
+          TablePropertiesCollectorFactory::Context::kUnknownColumnFamily, {},
+          kNoCompression, CompressionOptions(), false, nullptr);
    }
    delete mem->Unref();
    delete cf_mems_default;
@@ -339,7 +340,7 @@ class Repairer {
    t->meta.fd = FileDescriptor(t->meta.fd.GetNumber(), t->meta.fd.GetPathId(),
                                file_size);
    if (status.ok()) {
-      Iterator* iter = table_cache_->NewIterator(
+      InternalIterator* iter = table_cache_->NewIterator(
          ReadOptions(), env_options_, icmp_, t->meta.fd);
      bool empty = true;
      ParsedInternalKey parsed;
@@ -412,7 +413,7 @@ class Repairer {
    {
      unique_ptr<WritableFileWriter> file_writer(
          new WritableFileWriter(std::move(file), env_options));
-      log::Writer log(std::move(file_writer));
+      log::Writer log(std::move(file_writer), 0, false);
      std::string record;
      edit_->EncodeTo(&record);
      status = log.AddRecord(record);
@@ -14,7 +14,9 @@
 #include "db/version_edit.h"

 #include "rocksdb/statistics.h"
+#include "table/internal_iterator.h"
 #include "table/iterator_wrapper.h"
+#include "table/table_builder.h"
 #include "table/table_reader.h"
 #include "table/get_context.h"
 #include "util/coding.h"
@@ -106,8 +108,8 @@ Status TableCache::GetTableReader(
                                   ioptions_.statistics, record_read_stats,
                                   file_read_hist));
    s = ioptions_.table_factory->NewTableReader(
-        ioptions_, env_options, internal_comparator, std::move(file_reader),
-        fd.GetFileSize(), table_reader);
+        TableReaderOptions(ioptions_, env_options, internal_comparator),
+        std::move(file_reader), fd.GetFileSize(), table_reader);
    TEST_SYNC_POINT("TableCache::GetTableReader:0");
  }
  return s;
@@ -147,13 +149,11 @@ Status TableCache::FindTable(const EnvOptions& env_options,
  return s;
 }

-Iterator* TableCache::NewIterator(const ReadOptions& options,
-                                  const EnvOptions& env_options,
-                                  const InternalKeyComparator& icomparator,
-                                  const FileDescriptor& fd,
-                                  TableReader** table_reader_ptr,
-                                  HistogramImpl* file_read_hist,
-                                  bool for_compaction, Arena* arena) {
+InternalIterator* TableCache::NewIterator(
+    const ReadOptions& options, const EnvOptions& env_options,
+    const InternalKeyComparator& icomparator, const FileDescriptor& fd,
+    TableReader** table_reader_ptr, HistogramImpl* file_read_hist,
+    bool for_compaction, Arena* arena) {
  PERF_TIMER_GUARD(new_table_iterator_nanos);

  if (table_reader_ptr != nullptr) {
@@ -170,7 +170,7 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,
        env_options, icomparator, fd, /* sequential mode */ true,
        /* record stats */ false, nullptr, &table_reader_unique_ptr);
    if (!s.ok()) {
-      return NewErrorIterator(s, arena);
+      return NewErrorInternalIterator(s, arena);
    }
    table_reader = table_reader_unique_ptr.release();
  } else {
@@ -181,13 +181,13 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,
                    options.read_tier == kBlockCacheTier /* no_io */,
                    !for_compaction /* record read_stats */, file_read_hist);
      if (!s.ok()) {
-        return NewErrorIterator(s, arena);
+        return NewErrorInternalIterator(s, arena);
      }
      table_reader = GetTableReaderFromHandle(handle);
    }
  }

-  Iterator* result = table_reader->NewIterator(options, arena);
+  InternalIterator* result = table_reader->NewIterator(options, arena);

  if (create_new_table_reader) {
    assert(handle == nullptr);
@@ -29,6 +29,7 @@ class Arena;
 struct FileDescriptor;
 class GetContext;
 class HistogramImpl;
+class InternalIterator;

 class TableCache {
 public:
@@ -43,12 +44,12 @@ class TableCache {
  // the returned iterator.  The returned "*tableptr" object is owned by
  // the cache and should not be deleted, and is valid for as long as the
  // returned iterator is live.
-  Iterator* NewIterator(const ReadOptions& options, const EnvOptions& toptions,
-                        const InternalKeyComparator& internal_comparator,
-                        const FileDescriptor& file_fd,
-                        TableReader** table_reader_ptr = nullptr,
-                        HistogramImpl* file_read_hist = nullptr,
-                        bool for_compaction = false, Arena* arena = nullptr);
+  InternalIterator* NewIterator(
+      const ReadOptions& options, const EnvOptions& toptions,
+      const InternalKeyComparator& internal_comparator,
+      const FileDescriptor& file_fd, TableReader** table_reader_ptr = nullptr,
+      HistogramImpl* file_read_hist = nullptr, bool for_compaction = false,
+      Arena* arena = nullptr);

  // If a seek to internal key "k" in specified file finds an entry,
  // call (*handle_result)(arg, found_key, found_value) repeatedly until
@@ -19,7 +19,9 @@ Status InternalKeyPropertiesCollector::InternalAdd(const Slice& key,
    return Status::InvalidArgument("Invalid internal key");
  }

-  if (ikey.type == ValueType::kTypeDeletion) {
+  // Note: We count both, deletions and single deletions here.
+  if (ikey.type == ValueType::kTypeDeletion ||
+      ikey.type == ValueType::kTypeSingleDeletion) {
    ++deleted_keys_;
  }

@@ -47,18 +49,22 @@ InternalKeyPropertiesCollector::GetReadableProperties() const {
 }

 namespace {
+
 EntryType GetEntryType(ValueType value_type) {
  switch (value_type) {
    case kTypeValue:
      return kEntryPut;
    case kTypeDeletion:
      return kEntryDelete;
+    case kTypeSingleDeletion:
+      return kEntrySingleDelete;
    case kTypeMerge:
      return kEntryMerge;
    default:
      return kEntryOther;
  }
 }
+
 }  // namespace

 Status UserKeyTablePropertiesCollector::InternalAdd(const Slice& key,
@@ -36,12 +36,13 @@ class IntTblPropCollector {
  virtual bool NeedCompact() const { return false; }
 };

-// Facrtory for internal table properties collector.
+// Factory for internal table properties collector.
 class IntTblPropCollectorFactory {
 public:
  virtual ~IntTblPropCollectorFactory() {}
  // has to be thread-safe
-  virtual IntTblPropCollector* CreateIntTblPropCollector() = 0;
+  virtual IntTblPropCollector* CreateIntTblPropCollector(
+      uint32_t column_family_id) = 0;

  // The name of the properties collector can be used for debugging purpose.
  virtual const char* Name() const = 0;
@@ -69,7 +70,8 @@ class InternalKeyPropertiesCollector : public IntTblPropCollector {
 class InternalKeyPropertiesCollectorFactory
    : public IntTblPropCollectorFactory {
 public:
-  virtual IntTblPropCollector* CreateIntTblPropCollector() override {
+  virtual IntTblPropCollector* CreateIntTblPropCollector(
+      uint32_t column_family_id) override {
    return new InternalKeyPropertiesCollector();
  }

@@ -114,9 +116,12 @@ class UserKeyTablePropertiesCollectorFactory
  explicit UserKeyTablePropertiesCollectorFactory(
      std::shared_ptr<TablePropertiesCollectorFactory> user_collector_factory)
      : user_collector_factory_(user_collector_factory) {}
-  virtual IntTblPropCollector* CreateIntTblPropCollector() override {
+  virtual IntTblPropCollector* CreateIntTblPropCollector(
+      uint32_t column_family_id) override {
+    TablePropertiesCollectorFactory::Context context;
+    context.column_family_id = column_family_id;
    return new UserKeyTablePropertiesCollector(
-        user_collector_factory_->CreateTablePropertiesCollector());
+        user_collector_factory_->CreateTablePropertiesCollector(context));
  }

  virtual const char* Name() const override {
@@ -6,6 +6,7 @@
 #include <map>
 #include <memory>
 #include <string>
+#include <utility>
 #include <vector>

 #include "db/db_impl.h"
@@ -34,6 +35,8 @@ class TablePropertiesTest : public testing::Test,

 // Utilities test functions
 namespace {
+static const uint32_t kTestColumnFamilyId = 66;
+
 void MakeBuilder(const Options& options, const ImmutableCFOptions& ioptions,
                 const InternalKeyComparator& internal_comparator,
                 const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
@@ -45,7 +48,8 @@ void MakeBuilder(const Options& options, const ImmutableCFOptions& ioptions,

  builder->reset(NewTableBuilder(
      ioptions, internal_comparator, int_tbl_prop_collector_factories,
-      writable->get(), options.compression, options.compression_opts));
+      kTestColumnFamilyId /* column_family_id */, writable->get(),
+      options.compression, options.compression_opts));
 }
 }  // namespace

@@ -58,16 +62,19 @@ class RegularKeysStartWithA: public TablePropertiesCollector {
     std::string encoded;
     std::string encoded_num_puts;
     std::string encoded_num_deletes;
+     std::string encoded_num_single_deletes;
     std::string encoded_num_size_changes;
     PutVarint32(&encoded, count_);
     PutVarint32(&encoded_num_puts, num_puts_);
     PutVarint32(&encoded_num_deletes, num_deletes_);
+     PutVarint32(&encoded_num_single_deletes, num_single_deletes_);
     PutVarint32(&encoded_num_size_changes, num_size_changes_);
     *properties = UserCollectedProperties{
         {"TablePropertiesTest", message_},
         {"Count", encoded},
         {"NumPuts", encoded_num_puts},
         {"NumDeletes", encoded_num_deletes},
+         {"NumSingleDeletes", encoded_num_single_deletes},
         {"NumSizeChanges", encoded_num_size_changes},
     };
     return Status::OK();
@@ -83,6 +90,8 @@ class RegularKeysStartWithA: public TablePropertiesCollector {
      num_puts_++;
    } else if (type == kEntryDelete) {
      num_deletes_++;
+    } else if (type == kEntrySingleDelete) {
+      num_single_deletes_++;
    }
    if (file_size < file_size_) {
      message_ = "File size should not decrease.";
@@ -102,6 +111,7 @@ class RegularKeysStartWithA: public TablePropertiesCollector {
  uint32_t count_ = 0;
  uint32_t num_puts_ = 0;
  uint32_t num_deletes_ = 0;
+  uint32_t num_single_deletes_ = 0;
  uint32_t num_size_changes_ = 0;
  uint64_t file_size_ = 0;
 };
@@ -171,14 +181,17 @@ class RegularKeysStartWithAFactory : public IntTblPropCollectorFactory,
 public:
  explicit RegularKeysStartWithAFactory(bool backward_mode)
      : backward_mode_(backward_mode) {}
-  virtual TablePropertiesCollector* CreateTablePropertiesCollector() override {
+  virtual TablePropertiesCollector* CreateTablePropertiesCollector(
+      TablePropertiesCollectorFactory::Context context) override {
+    EXPECT_EQ(kTestColumnFamilyId, context.column_family_id);
    if (!backward_mode_) {
      return new RegularKeysStartWithA();
    } else {
      return new RegularKeysStartWithABackwardCompatible();
    }
  }
-  virtual IntTblPropCollector* CreateIntTblPropCollector() override {
+  virtual IntTblPropCollector* CreateIntTblPropCollector(
+      uint32_t column_family_id) override {
    return new RegularKeysStartWithAInternal();
  }
  const char* Name() const override { return "RegularKeysStartWithA"; }
@@ -217,18 +230,18 @@ namespace {
 void TestCustomizedTablePropertiesCollector(
    bool backward_mode, uint64_t magic_number, bool test_int_tbl_prop_collector,
    const Options& options, const InternalKeyComparator& internal_comparator) {
-  const std::string kDeleteFlag = "D";
  // make sure the entries will be inserted with order.
-  std::map<std::string, std::string> kvs = {
-      {"About   ", "val5"},  // starts with 'A'
-      {"Abstract", "val2"},  // starts with 'A'
-      {"Around  ", "val7"},  // starts with 'A'
-      {"Beyond  ", "val3"},
-      {"Builder ", "val1"},
-      {"Love    ", kDeleteFlag},
-      {"Cancel  ", "val4"},
-      {"Find    ", "val6"},
-      {"Rocks   ", kDeleteFlag},
+  std::map<std::pair<std::string, ValueType>, std::string> kvs = {
+      {{"About   ", kTypeValue}, "val5"},  // starts with 'A'
+      {{"Abstract", kTypeValue}, "val2"},  // starts with 'A'
+      {{"Around  ", kTypeValue}, "val7"},  // starts with 'A'
+      {{"Beyond  ", kTypeValue}, "val3"},
+      {{"Builder ", kTypeValue}, "val1"},
+      {{"Love    ", kTypeDeletion}, ""},
+      {{"Cancel  ", kTypeValue}, "val4"},
+      {{"Find    ", kTypeValue}, "val6"},
+      {{"Rocks   ", kTypeDeletion}, ""},
+      {{"Foo     ", kTypeSingleDeletion}, ""},
  };

  // -- Step 1: build table
@@ -248,9 +261,7 @@ void TestCustomizedTablePropertiesCollector(

  SequenceNumber seqNum = 0U;
  for (const auto& kv : kvs) {
-    InternalKey ikey(kv.first, seqNum++, (kv.second != kDeleteFlag)
-                                             ? ValueType::kTypeValue
-                                             : ValueType::kTypeDeletion);
+    InternalKey ikey(kv.first.first, seqNum++, kv.first.second);
    builder->Add(ikey.Encode(), kv.second);
  }
  ASSERT_OK(builder->Finish());
@@ -270,31 +281,36 @@ void TestCustomizedTablePropertiesCollector(

  auto user_collected = props->user_collected_properties;

-  ASSERT_TRUE(user_collected.find("TablePropertiesTest") !=
-              user_collected.end());
+  ASSERT_NE(user_collected.find("TablePropertiesTest"), user_collected.end());
  ASSERT_EQ("Rocksdb", user_collected.at("TablePropertiesTest"));

  uint32_t starts_with_A = 0;
-  ASSERT_TRUE(user_collected.find("Count") != user_collected.end());
+  ASSERT_NE(user_collected.find("Count"), user_collected.end());
  Slice key(user_collected.at("Count"));
  ASSERT_TRUE(GetVarint32(&key, &starts_with_A));
  ASSERT_EQ(3u, starts_with_A);

  if (!backward_mode && !test_int_tbl_prop_collector) {
-    uint32_t num_deletes;
-    ASSERT_TRUE(user_collected.find("NumDeletes") != user_collected.end());
-    Slice key_deletes(user_collected.at("NumDeletes"));
-    ASSERT_TRUE(GetVarint32(&key_deletes, &num_deletes));
-    ASSERT_EQ(2u, num_deletes);
-
    uint32_t num_puts;
-    ASSERT_TRUE(user_collected.find("NumPuts") != user_collected.end());
+    ASSERT_NE(user_collected.find("NumPuts"), user_collected.end());
    Slice key_puts(user_collected.at("NumPuts"));
    ASSERT_TRUE(GetVarint32(&key_puts, &num_puts));
    ASSERT_EQ(7u, num_puts);

+    uint32_t num_deletes;
+    ASSERT_NE(user_collected.find("NumDeletes"), user_collected.end());
+    Slice key_deletes(user_collected.at("NumDeletes"));
+    ASSERT_TRUE(GetVarint32(&key_deletes, &num_deletes));
+    ASSERT_EQ(2u, num_deletes);
+
+    uint32_t num_single_deletes;
+    ASSERT_NE(user_collected.find("NumSingleDeletes"), user_collected.end());
+    Slice key_single_deletes(user_collected.at("NumSingleDeletes"));
+    ASSERT_TRUE(GetVarint32(&key_single_deletes, &num_single_deletes));
+    ASSERT_EQ(1u, num_single_deletes);
+
    uint32_t num_size_changes;
-    ASSERT_TRUE(user_collected.find("NumSizeChanges") != user_collected.end());
+    ASSERT_NE(user_collected.find("NumSizeChanges"), user_collected.end());
    Slice key_size_changes(user_collected.at("NumSizeChanges"));
    ASSERT_TRUE(GetVarint32(&key_size_changes, &num_size_changes));
    ASSERT_GE(num_size_changes, 2u);
@@ -350,6 +366,7 @@ void TestInternalKeyPropertiesCollector(
      InternalKey("X       ", 4, ValueType::kTypeDeletion),
      InternalKey("Y       ", 5, ValueType::kTypeDeletion),
      InternalKey("Z       ", 6, ValueType::kTypeDeletion),
+      InternalKey("a       ", 7, ValueType::kTypeSingleDeletion),
  };

  std::unique_ptr<TableBuilder> builder;
@@ -403,27 +420,34 @@ void TestInternalKeyPropertiesCollector(
    std::unique_ptr<TableProperties> props_guard(props);
    auto user_collected = props->user_collected_properties;
    uint64_t deleted = GetDeletedKeys(user_collected);
-    ASSERT_EQ(4u, deleted);
+    ASSERT_EQ(5u, deleted);  // deletes + single-deletes

    if (sanitized) {
      uint32_t starts_with_A = 0;
-      ASSERT_TRUE(user_collected.find("Count") != user_collected.end());
+      ASSERT_NE(user_collected.find("Count"), user_collected.end());
      Slice key(user_collected.at("Count"));
      ASSERT_TRUE(GetVarint32(&key, &starts_with_A));
      ASSERT_EQ(1u, starts_with_A);

      if (!backward_mode) {
+        uint32_t num_puts;
+        ASSERT_NE(user_collected.find("NumPuts"), user_collected.end());
+        Slice key_puts(user_collected.at("NumPuts"));
+        ASSERT_TRUE(GetVarint32(&key_puts, &num_puts));
+        ASSERT_EQ(3u, num_puts);
+
        uint32_t num_deletes;
-        ASSERT_TRUE(user_collected.find("NumDeletes") != user_collected.end());
+        ASSERT_NE(user_collected.find("NumDeletes"), user_collected.end());
        Slice key_deletes(user_collected.at("NumDeletes"));
        ASSERT_TRUE(GetVarint32(&key_deletes, &num_deletes));
        ASSERT_EQ(4u, num_deletes);

-        uint32_t num_puts;
-        ASSERT_TRUE(user_collected.find("NumPuts") != user_collected.end());
-        Slice key_puts(user_collected.at("NumPuts"));
-        ASSERT_TRUE(GetVarint32(&key_puts, &num_puts));
-        ASSERT_EQ(3u, num_puts);
+        uint32_t num_single_deletes;
+        ASSERT_NE(user_collected.find("NumSingleDeletes"),
+                  user_collected.end());
+        Slice key_single_deletes(user_collected.at("NumSingleDeletes"));
+        ASSERT_TRUE(GetVarint32(&key_single_deletes, &num_single_deletes));
+        ASSERT_EQ(1u, num_single_deletes);
      }
    }
  }
@@ -262,8 +262,10 @@ Status TransactionLogIteratorImpl::OpenLogReader(const LogFile* logFile) {
    return s;
  }
  assert(file);
-  currentLogReader_.reset(new log::Reader(std::move(file), &reporter_,
-                                          read_options_.verify_checksums_, 0));
+  currentLogReader_.reset(new log::Reader(options_->info_log,
+					  std::move(file), &reporter_,
+                                          read_options_.verify_checksums_, 0,
+                                          logFile->LogNumber()));
  return Status::OK();
 }
 }  //  namespace rocksdb
@@ -82,6 +82,7 @@ class VersionBuilder::Rep {
  };

  const EnvOptions& env_options_;
+  Logger* info_log_;
  TableCache* table_cache_;
  VersionStorageInfo* base_vstorage_;
  LevelState* levels_;
@@ -89,9 +90,10 @@ class VersionBuilder::Rep {
  FileComparator level_nonzero_cmp_;

 public:
-  Rep(const EnvOptions& env_options, TableCache* table_cache,
+  Rep(const EnvOptions& env_options, Logger* info_log, TableCache* table_cache,
      VersionStorageInfo* base_vstorage)
      : env_options_(env_options),
+        info_log_(info_log),
        table_cache_(table_cache),
        base_vstorage_(base_vstorage) {
    levels_ = new LevelState[base_vstorage_->num_levels()];
@@ -134,7 +136,10 @@ class VersionBuilder::Rep {
        auto f2 = level_files[i];
        if (level == 0) {
          assert(level_zero_cmp_(f1, f2));
-          assert(f1->largest_seqno > f2->largest_seqno);
+          assert(f1->largest_seqno > f2->largest_seqno ||
+                 // We can have multiple files with seqno = 0 as a result of
+                 // using DB::AddFile()
+                 (f1->largest_seqno == 0 && f2->largest_seqno == 0));
        } else {
          assert(level_nonzero_cmp_(f1, f2));

@@ -335,15 +340,16 @@ class VersionBuilder::Rep {
    if (levels_[level].deleted_files.count(f->fd.GetNumber()) > 0) {
      // File is deleted: do nothing
    } else {
-      vstorage->AddFile(level, f);
+      vstorage->AddFile(level, f, info_log_);
    }
  }
 };

 VersionBuilder::VersionBuilder(const EnvOptions& env_options,
                               TableCache* table_cache,
-                               VersionStorageInfo* base_vstorage)
-    : rep_(new Rep(env_options, table_cache, base_vstorage)) {}
+                               VersionStorageInfo* base_vstorage,
+                               Logger* info_log)
+    : rep_(new Rep(env_options, info_log, table_cache, base_vstorage)) {}
 VersionBuilder::~VersionBuilder() { delete rep_; }
 void VersionBuilder::CheckConsistency(VersionStorageInfo* vstorage) {
  rep_->CheckConsistency(vstorage);
@@ -24,7 +24,7 @@ class InternalStats;
 class VersionBuilder {
 public:
  VersionBuilder(const EnvOptions& env_options, TableCache* table_cache,
-                 VersionStorageInfo* base_vstorage);
+                 VersionStorageInfo* base_vstorage, Logger* info_log = nullptr);
  ~VersionBuilder();
  void CheckConsistency(VersionStorageInfo* vstorage);
  void CheckConsistencyForDeletes(VersionEdit* edit, uint64_t number,
@@ -77,7 +77,7 @@ class VersionBuilderTest : public testing::Test {
  }

  void UpdateVersionStorageInfo() {
-    vstorage_.UpdateFilesBySize();
+    vstorage_.UpdateFilesByCompactionPri(mutable_cf_options_);
    vstorage_.UpdateNumNonEmptyLevels();
    vstorage_.GenerateFileIndexer();
    vstorage_.GenerateLevelFilesBrief();
@@ -12,6 +12,7 @@
 #include "db/version_set.h"
 #include "util/coding.h"
 #include "util/event_logger.h"
+#include "util/sync_point.h"
 #include "rocksdb/slice.h"

 namespace rocksdb {
@@ -32,12 +33,22 @@ enum Tag {
  // these are new formats divergent from open source leveldb
  kNewFile2 = 100,
  kNewFile3 = 102,
+  kNewFile4 = 103,      // 4th (the latest) format version of adding files
  kColumnFamily = 200,  // specify column family for version edit
  kColumnFamilyAdd = 201,
  kColumnFamilyDrop = 202,
  kMaxColumnFamily = 203,
 };

+enum CustomTag {
+  kTerminate = 1,  // The end of customized fields
+  kNeedCompaction = 2,
+  kPathId = 65,
+};
+// If this bit for the custom tag is set, opening DB should fail if
+// we don't know this field.
+uint32_t kCustomTagNonSafeIgnoreMask = 1 << 6;
+
 uint64_t PackFileNumberAndPathId(uint64_t number, uint64_t path_id) {
  assert(number <= kFileNumberMask);
  return number | (path_id * (kFileNumberMask + 1));
@@ -102,7 +113,11 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
    if (!f.smallest.Valid() || !f.largest.Valid()) {
      return false;
    }
-    if (f.fd.GetPathId() == 0) {
+    bool has_customized_fields = false;
+    if (f.marked_for_compaction) {
+      PutVarint32(dst, kNewFile4);
+      has_customized_fields = true;
+    } else if (f.fd.GetPathId() == 0) {
      // Use older format to make sure user can roll back the build if they
      // don't config multiple DB paths.
      PutVarint32(dst, kNewFile2);
@@ -111,7 +126,8 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
    }
    PutVarint32(dst, new_files_[i].first);  // level
    PutVarint64(dst, f.fd.GetNumber());
-    if (f.fd.GetPathId() != 0) {
+    if (f.fd.GetPathId() != 0 && !has_customized_fields) {
+      // kNewFile3
      PutVarint32(dst, f.fd.GetPathId());
    }
    PutVarint64(dst, f.fd.GetFileSize());
@@ -119,6 +135,48 @@ bool VersionEdit::EncodeTo(std::string* dst) const {
    PutLengthPrefixedSlice(dst, f.largest.Encode());
    PutVarint64(dst, f.smallest_seqno);
    PutVarint64(dst, f.largest_seqno);
+    if (has_customized_fields) {
+      // Customized fields' format:
+      // +-----------------------------+
+      // | 1st field's tag (varint32)  |
+      // +-----------------------------+
+      // | 1st field's size (varint32) |
+      // +-----------------------------+
+      // |    bytes for 1st field      |
+      // |  (based on size decoded)    |
+      // +-----------------------------+
+      // |                             |
+      // |          ......             |
+      // |                             |
+      // +-----------------------------+
+      // | last field's size (varint32)|
+      // +-----------------------------+
+      // |    bytes for last field     |
+      // |  (based on size decoded)    |
+      // +-----------------------------+
+      // | terminating tag (varint32)  |
+      // +-----------------------------+
+      //
+      // Customized encoding for fields:
+      //   tag kPathId: 1 byte as path_id
+      //   tag kNeedCompaction:
+      //        now only can take one char value 1 indicating need-compaction
+      //
+      if (f.fd.GetPathId() != 0) {
+        PutVarint32(dst, CustomTag::kPathId);
+        char p = static_cast<char>(f.fd.GetPathId());
+        PutLengthPrefixedSlice(dst, Slice(&p, 1));
+      }
+      if (f.marked_for_compaction) {
+        PutVarint32(dst, CustomTag::kNeedCompaction);
+        char p = static_cast<char>(1);
+        PutLengthPrefixedSlice(dst, Slice(&p, 1));
+      }
+      TEST_SYNC_POINT_CALLBACK("VersionEdit::EncodeTo:NewFile4:CustomizeFields",
+                               dst);
+
+      PutVarint32(dst, CustomTag::kTerminate);
+    }
  }

  // 0 is default and does not need to be explicitly written
@@ -161,6 +219,63 @@ bool VersionEdit::GetLevel(Slice* input, int* level, const char** msg) {
  }
 }

+const char* VersionEdit::DecodeNewFile4From(Slice* input) {
+  const char* msg = nullptr;
+  int level;
+  FileMetaData f;
+  uint64_t number;
+  uint32_t path_id = 0;
+  uint64_t file_size;
+  if (GetLevel(input, &level, &msg) && GetVarint64(input, &number) &&
+      GetVarint64(input, &file_size) && GetInternalKey(input, &f.smallest) &&
+      GetInternalKey(input, &f.largest) &&
+      GetVarint64(input, &f.smallest_seqno) &&
+      GetVarint64(input, &f.largest_seqno)) {
+    // See comments in VersionEdit::EncodeTo() for format of customized fields
+    while (true) {
+      uint32_t custom_tag;
+      Slice field;
+      if (!GetVarint32(input, &custom_tag)) {
+        return "new-file4 custom field";
+      }
+      if (custom_tag == kTerminate) {
+        break;
+      }
+      if (!GetLengthPrefixedSlice(input, &field)) {
+        return "new-file4 custom field lenth prefixed slice error";
+      }
+      switch (custom_tag) {
+        case kPathId:
+          if (field.size() != 1) {
+            return "path_id field wrong size";
+          }
+          path_id = field[0];
+          if (path_id > 3) {
+            return "path_id wrong vaue";
+          }
+          break;
+        case kNeedCompaction:
+          if (field.size() != 1) {
+            return "need_compaction field wrong size";
+          }
+          f.marked_for_compaction = (field[0] == 1);
+          break;
+        default:
+          if ((custom_tag & kCustomTagNonSafeIgnoreMask) != 0) {
+            // Should not proceed if cannot understand it
+            return "new-file4 custom field not supported";
+          }
+          break;
+      }
+    }
+  } else {
+    return "new-file4 entry";
+  }
+  f.fd = FileDescriptor(number, path_id, file_size);
+  new_files_.push_back(std::make_pair(level, f));
+  return nullptr;
+}
+
 Status VersionEdit::DecodeFrom(const Slice& src) {
  Clear();
  Slice input = src;
@@ -304,6 +419,11 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
        break;
      }

+      case kNewFile4: {
+        msg = DecodeNewFile4From(&input);
+        break;
+      }
+
      case kColumnFamily:
        if (!GetVarint32(&input, &column_family_)) {
          if (!msg) {
--- a/Show More
+++ b/Show More