Bump version

This is the default now

.gitignore

@@ -1,2 +1,3 @@
 .cache
+__pycache__
 build

.pre-commit-config.yaml

@@ -1,11 +1,11 @@
 repos:
   - repo: https://github.com/pre-commit/mirrors-clang-format
-    rev: b111689e7b5cba60be3c62d5db2bd1357f4d36ca
+    rev: 6d365699efc33b1b432eab5b4ae331a19e1857de # frozen: v18.1.2
     hooks:
       - id: clang-format
         exclude: ".*third_party/.*"
   - repo: https://github.com/cheshirekow/cmake-format-precommit
-    rev: e2c2116d86a80e72e7146a06e68b7c228afc6319
+    rev: e2c2116d86a80e72e7146a06e68b7c228afc6319 # frozen: v0.6.13
     hooks:
       - id: cmake-format
   - repo: local
@@ -13,7 +13,7 @@ repos:
       - id: debug verbose check
         name: disallow checking in DEBUG_VERBOSE=1
         description: disallow checking in DEBUG_VERBOSE=1
-        entry: '^#define DEBUG_VERBOSE 1$'
+        entry: "^#define DEBUG_VERBOSE 1$"
         language: pygrep
         types: [c++]
   - repo: local
@@ -21,6 +21,14 @@ repos:
       - id: debug verbose check
         name: disallow checking in SHOW_MEMORY=1
         description: disallow checking in SHOW_MEMORY=1
-        entry: '^#define SHOW_MEMORY 1$'
+        entry: "^#define SHOW_MEMORY 1$"
         language: pygrep
         types: [c++]
+  - repo: https://github.com/shellcheck-py/shellcheck-py
+    rev: a23f6b85d0fdd5bb9d564e2579e678033debbdff # frozen: v0.10.0.1
+    hooks:
+      - id: shellcheck
+  - repo: https://github.com/psf/black
+    rev: 552baf822992936134cbd31a38f69c8cfe7c0f05 # frozen: 24.3.0
+    hooks:
+      - id: black

Bench.cpp

@@ -258,4 +258,107 @@ void benchConflictSet() {
   }
 }
 
-int main(void) { benchConflictSet(); }
+constexpr int kKeyLenForWorstCase = 50;
+
+ConflictSet worstCaseConflictSetForRadixRangeRead(int cardinality) {
+  ConflictSet cs{0};
+
+  for (int i = 0; i < kKeyLenForWorstCase; ++i) {
+    for (int j = 0; j < cardinality; ++j) {
+      auto b = std::vector<uint8_t>(i, 0);
+      b.push_back(j);
+      auto e = std::vector<uint8_t>(i, 255);
+      e.push_back(255 - j);
+      weaselab::ConflictSet::WriteRange w[] = {{
+                                                   {b.data(), int(b.size())},
+                                                   {nullptr, 0},
+                                               },
+                                               {
+                                                   {e.data(), int(e.size())},
+                                                   {nullptr, 0},
+                                               }};
+      std::sort(std::begin(w), std::end(w),
+                [](const auto &lhs, const auto &rhs) {
+                  int cl = std::min(lhs.begin.len, rhs.begin.len);
+                  if (cl > 0) {
+                    int c = memcmp(lhs.begin.p, rhs.begin.p, cl);
+                    if (c != 0) {
+                      return c < 0;
+                    }
+                  }
+                  return lhs.begin.len < rhs.begin.len;
+                });
+      cs.addWrites(w, sizeof(w) / sizeof(w[0]), 0);
+    }
+  }
+
+  // Defeat short-circuiting on the left
+  {
+    auto k = std::vector<uint8_t>(kKeyLenForWorstCase, 0);
+    weaselab::ConflictSet::WriteRange w[] = {
+        {
+            {k.data(), int(k.size())},
+            {nullptr, 0},
+        },
+    };
+    cs.addWrites(w, sizeof(w) / sizeof(w[0]), 1);
+  }
+
+  // Defeat short-circuiting on the right
+  {
+    auto k = std::vector<uint8_t>(kKeyLenForWorstCase, 255);
+    weaselab::ConflictSet::WriteRange w[] = {
+        {
+            {k.data(), int(k.size())},
+            {nullptr, 0},
+        },
+    };
+    cs.addWrites(w, sizeof(w) / sizeof(w[0]), 1);
+  }
+
+  return cs;
+}
+
+void benchWorstCaseForRadixRangeRead() {
+  ankerl::nanobench::Bench bench;
+
+  std::unique_ptr<ConflictSet> cs[256];
+  for (int i = 0; i < 256; ++i) {
+    cs[i] =
+        std::make_unique<ConflictSet>(worstCaseConflictSetForRadixRangeRead(i));
+  }
+
+  auto begin = std::vector<uint8_t>(kKeyLenForWorstCase - 1, 0);
+  begin.push_back(1);
+  auto end = std::vector<uint8_t>(kKeyLenForWorstCase - 1, 255);
+  end.push_back(254);
+
+  weaselab::ConflictSet::Result result;
+  weaselab::ConflictSet::ReadRange r{
+      {begin.data(), int(begin.size())}, {end.data(), int(end.size())}, 0};
+
+  bench.run("worst case for radix tree", [&]() {
+    for (int i = 0; i < 256; ++i) {
+      result = weaselab::ConflictSet::TooOld;
+      cs[i]->check(&r, &result, 1);
+      if (result != weaselab::ConflictSet::Commit) {
+        abort();
+      }
+    }
+  });
+
+  // for (int i = 0; i < 256; ++i) {
+  //   bench.run("worst case for radix tree, span " + std::to_string(i), [&]() {
+  //     result = weaselab::ConflictSet::TooOld;
+  //     cs[i]->check(&r, &result, 1);
+  //     if (result != weaselab::ConflictSet::Commit) {
+  //       abort();
+  //     }
+  //   });
+  // }
+}
+
+int main(void) {
+  benchConflictSet();
+  benchWorstCaseForRadixRangeRead();
+}

CMakeLists.txt

@@ -1,15 +1,30 @@
 cmake_minimum_required(VERSION 3.18)
 project(
   conflict-set
-  VERSION 0.0.1
+  VERSION 0.0.8
   DESCRIPTION
     "A data structure for optimistic concurrency control on ranges of bitwise-lexicographically-ordered keys."
   HOMEPAGE_URL "https://git.weaselab.dev/weaselab/conflict-set"
   LANGUAGES C CXX)
 set(CMAKE_CXX_STANDARD 20)
 
+file(WRITE ${CMAKE_CURRENT_BINARY_DIR}/version.txt ${PROJECT_VERSION})
+configure_file(${CMAKE_CURRENT_SOURCE_DIR}/version.txt.in
+               ${CMAKE_CURRENT_SOURCE_DIR}/paper/version.txt)
+
+include(CMakePushCheckState)
+include(CheckCXXCompilerFlag)
+include(CheckIncludeFileCXX)
+include(CheckCXXSourceCompiles)
+
 set(DEFAULT_BUILD_TYPE "Release")
 
+if(EMSCRIPTEN OR CMAKE_SYSTEM_NAME STREQUAL WASI)
+  set(WASM ON)
+else()
+  set(WASM OFF)
+endif()
+
 if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
   message(
     STATUS
@@ -23,19 +38,33 @@ if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
 endif()
 
 add_compile_options(-fdata-sections -ffunction-sections -Wswitch-enum
-                    -Werror=switch-enum)
+                    -Werror=switch-enum -fPIC)
+
+set(full_relro_flags "-pie;LINKER:-z,relro,-z,now,-z,noexecstack")
+cmake_push_check_state()
+list(APPEND CMAKE_REQUIRED_LINK_OPTIONS ${full_relro_flags})
+check_cxx_source_compiles("int main(){}" HAS_FULL_RELRO FAIL_REGEX "warning:")
+if(HAS_FULL_RELRO)
+  add_link_options(${full_relro_flags})
+endif()
+cmake_pop_check_state()
+
+set(version_script_flags
+    LINKER:--version-script=${CMAKE_CURRENT_SOURCE_DIR}/linker.map)
+cmake_push_check_state()
+list(APPEND CMAKE_REQUIRED_LINK_OPTIONS ${version_script_flags})
+check_cxx_source_compiles("int main(){}" HAS_VERSION_SCRIPT FAIL_REGEX
+                          "warning:")
+cmake_pop_check_state()
 
 option(USE_SIMD_FALLBACK
        "Use fallback implementations of functions that use SIMD" OFF)
 
 # This is encouraged according to
 # https://valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.clientreq
-include_directories(SYSTEM ${CMAKE_SOURCE_DIR}/third_party/valgrind)
+include_directories(SYSTEM ${CMAKE_CURRENT_SOURCE_DIR}/third_party/valgrind)
 
-if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
-  add_compile_options(-Wno-maybe-uninitialized
-                      $<$<COMPILE_LANGUAGE:CXX>:-Wno-invalid-offsetof>)
-endif()
+add_compile_options($<$<COMPILE_LANGUAGE:CXX>:-Wno-invalid-offsetof>)
 
 if(APPLE)
   add_link_options(-Wl,-dead_strip)
@@ -43,10 +72,22 @@ else()
   add_link_options(-Wl,--gc-sections)
 endif()
 
-include(CheckIncludeFileCXX)
-include(CMakePushCheckState)
+if(EMSCRIPTEN)
+  # https://github.com/emscripten-core/emscripten/issues/15377#issuecomment-1285167486
+  add_link_options(-lnodefs.js -lnoderawfs.js)
+  add_link_options(-s ALLOW_MEMORY_GROWTH)
+endif()
 
 if(NOT USE_SIMD_FALLBACK)
+  cmake_push_check_state()
+  list(APPEND CMAKE_REQUIRED_FLAGS -msimd128)
+  check_include_file_cxx("wasm_simd128.h" HAS_WASM_SIMD)
+  if(HAS_WASM_SIMD)
+    add_compile_options(-msimd128)
+    add_compile_definitions(HAS_WASM_SIMD)
+  endif()
+  cmake_pop_check_state()
+
   cmake_push_check_state()
   list(APPEND CMAKE_REQUIRED_FLAGS -mavx)
   check_include_file_cxx("immintrin.h" HAS_AVX)
@@ -64,70 +105,104 @@ endif()
 
 set(CMAKE_CXX_IMPLICIT_LINK_LIBRARIES "")
 
-add_library(${PROJECT_NAME}_object OBJECT ConflictSet.cpp)
-target_compile_options(${PROJECT_NAME}_object PRIVATE -fPIC -fno-exceptions
+add_library(${PROJECT_NAME}-object OBJECT ConflictSet.cpp)
+target_compile_options(${PROJECT_NAME}-object PRIVATE -fno-exceptions
                                                       -fvisibility=hidden)
-target_include_directories(${PROJECT_NAME}_object
-                           PRIVATE ${CMAKE_SOURCE_DIR}/include)
+target_include_directories(${PROJECT_NAME}-object
+                           PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/include)
 
-add_library(${PROJECT_NAME} SHARED $<TARGET_OBJECTS:${PROJECT_NAME}_object>)
+add_library(${PROJECT_NAME} SHARED $<TARGET_OBJECTS:${PROJECT_NAME}-object>)
 set_target_properties(
   ${PROJECT_NAME} PROPERTIES LIBRARY_OUTPUT_DIRECTORY
-                             "${CMAKE_BINARY_DIR}/radix_tree")
+                             "${CMAKE_CURRENT_BINARY_DIR}/radix_tree")
 if(NOT CMAKE_BUILD_TYPE STREQUAL Debug)
   set_target_properties(${PROJECT_NAME} PROPERTIES LINKER_LANGUAGE C)
 endif()
 
-if(NOT APPLE)
-  target_link_options(${PROJECT_NAME} PRIVATE
-                      LINKER:--version-script=${CMAKE_SOURCE_DIR}/linker.map)
+if(HAS_VERSION_SCRIPT)
+  target_link_options(
+    ${PROJECT_NAME} PRIVATE
+    LINKER:--version-script=${CMAKE_CURRENT_SOURCE_DIR}/linker.map)
 endif()
 
 add_library(${PROJECT_NAME}-static STATIC
-            $<TARGET_OBJECTS:${PROJECT_NAME}_object>)
+            $<TARGET_OBJECTS:${PROJECT_NAME}-object>)
 if(NOT CMAKE_BUILD_TYPE STREQUAL Debug)
   set_target_properties(${PROJECT_NAME}-static PROPERTIES LINKER_LANGUAGE C)
 endif()
 
-if(NOT APPLE AND CMAKE_OBJCOPY)
+if(APPLE)
   add_custom_command(
-    TARGET conflict-set-static
+    TARGET ${PROJECT_NAME}-static
+    PRE_LINK
+    COMMAND ${CMAKE_CURRENT_SOURCE_DIR}/privatize_symbols_macos.sh
+            $<TARGET_OBJECTS:${PROJECT_NAME}-object>)
+else()
+  add_custom_command(
+    TARGET ${PROJECT_NAME}-static
     POST_BUILD
     COMMAND
-      ${CMAKE_OBJCOPY} --keep-global-symbols=${CMAKE_SOURCE_DIR}/symbols.txt
-      $<TARGET_FILE:${PROJECT_NAME}-static>)
+      ${CMAKE_OBJCOPY}
+      --keep-global-symbols=${CMAKE_CURRENT_SOURCE_DIR}/symbol-exports.txt
+      $<TARGET_FILE:${PROJECT_NAME}-static> || echo
+      "Proceeding with all symbols global in static library")
 endif()
 
-set(TEST_FLAGS -Wall -Wextra -Wunreachable-code -Wpedantic -UNDEBUG)
-
 include(CTest)
 
-if(BUILD_TESTING)
+# disable tests if this is being used through e.g. FetchContent
+if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
 
-  # Shared library version of FoundationDB's skip list implementation
-  add_library(skip_list SHARED SkipList.cpp)
-  target_compile_options(skip_list PRIVATE -fPIC -fno-exceptions
-                                           -fvisibility=hidden)
-  target_include_directories(skip_list PUBLIC ${CMAKE_SOURCE_DIR}/include)
-  set_target_properties(skip_list PROPERTIES LIBRARY_OUTPUT_DIRECTORY
-                                             "${CMAKE_BINARY_DIR}/skip_list")
-  set_target_properties(skip_list PROPERTIES OUTPUT_NAME ${PROJECT_NAME})
-  set_target_properties(skip_list PROPERTIES VERSION ${PROJECT_VERSION}
-                                             SOVERSION ${PROJECT_VERSION_MAJOR})
+  set(TEST_FLAGS -Wall -Wextra -Wunreachable-code -Wpedantic -UNDEBUG)
 
-  # Shared library version of a std::unordered_map-based conflict set (point
-  # queries only)
-  add_library(hash_table SHARED HashTable.cpp)
-  target_compile_options(hash_table PRIVATE -fPIC -fno-exceptions
-                                            -fvisibility=hidden)
-  target_include_directories(hash_table PUBLIC ${CMAKE_SOURCE_DIR}/include)
-  set_target_properties(hash_table PROPERTIES LIBRARY_OUTPUT_DIRECTORY
-                                              "${CMAKE_BINARY_DIR}/hash_table")
-  set_target_properties(hash_table PROPERTIES OUTPUT_NAME ${PROJECT_NAME})
-  set_target_properties(
-    hash_table PROPERTIES VERSION ${PROJECT_VERSION} SOVERSION
-                                                     ${PROJECT_VERSION_MAJOR})
+  # corpus tests, which are tests curated by libfuzzer. The goal is to get broad
+  # coverage with a small number of tests.
 
+  file(GLOB CORPUS_TESTS ${CMAKE_CURRENT_SOURCE_DIR}/corpus/*)
+
+  # extra testing that relies on shared libraries, which aren't available with
+  # wasm
+  if(NOT WASM)
+    # Shared library version of FoundationDB's skip list implementation
+    add_library(skip_list SHARED SkipList.cpp)
+    target_compile_options(skip_list PRIVATE -fno-exceptions
+                                             -fvisibility=hidden)
+    target_include_directories(skip_list
+                               PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include)
+    set_target_properties(
+      skip_list PROPERTIES LIBRARY_OUTPUT_DIRECTORY
+                           "${CMAKE_CURRENT_BINARY_DIR}/skip_list")
+    set_target_properties(skip_list PROPERTIES OUTPUT_NAME ${PROJECT_NAME})
+    set_target_properties(
+      skip_list PROPERTIES VERSION ${PROJECT_VERSION} SOVERSION
+                                                      ${PROJECT_VERSION_MAJOR})
+
+    # Shared library version of a std::unordered_map-based conflict set (point
+    # queries only)
+    add_library(hash_table SHARED HashTable.cpp)
+    target_compile_options(hash_table PRIVATE -fno-exceptions
+                                              -fvisibility=hidden)
+    target_include_directories(hash_table
+                               PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include)
+    set_target_properties(
+      hash_table PROPERTIES LIBRARY_OUTPUT_DIRECTORY
+                            "${CMAKE_CURRENT_BINARY_DIR}/hash_table")
+    set_target_properties(hash_table PROPERTIES OUTPUT_NAME ${PROJECT_NAME})
+    set_target_properties(
+      hash_table PROPERTIES VERSION ${PROJECT_VERSION} SOVERSION
+                                                       ${PROJECT_VERSION_MAJOR})
+
+    add_executable(driver_skip_list TestDriver.cpp)
+    target_compile_options(driver_skip_list PRIVATE ${TEST_FLAGS})
+    target_link_libraries(driver_skip_list PRIVATE skip_list)
+
+    foreach(TEST ${CORPUS_TESTS})
+      get_filename_component(hash ${TEST} NAME)
+      add_test(NAME skip_list_${hash} COMMAND driver_skip_list ${TEST})
+    endforeach()
+  endif()
+
+  # ad hoc testing
   add_executable(conflict_set_main ConflictSet.cpp)
   target_include_directories(conflict_set_main
                              PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/include)
@@ -154,10 +229,7 @@ if(BUILD_TESTING)
     endif()
   endif()
 
-  # corpus tests
-
-  file(GLOB CORPUS_TESTS ${CMAKE_SOURCE_DIR}/corpus/*)
-
+  # whitebox tests
   add_executable(fuzz_driver ConflictSet.cpp FuzzTestDriver.cpp)
   target_compile_options(fuzz_driver PRIVATE ${TEST_FLAGS})
   if(NOT CMAKE_CROSSCOMPILING)
@@ -172,8 +244,7 @@ if(BUILD_TESTING)
     add_test(NAME conflict_set_fuzz_${hash} COMMAND fuzz_driver ${TEST})
   endforeach()
 
-  # tsan
-
+  # tsan tests
   if(NOT CMAKE_CROSSCOMPILING AND NOT CMAKE_BUILD_TYPE STREQUAL Debug)
     add_executable(tsan_driver ConflictSet.cpp FuzzTestDriver.cpp)
     target_compile_options(tsan_driver PRIVATE ${TEST_FLAGS} -fsanitize=thread)
@@ -187,23 +258,37 @@ if(BUILD_TESTING)
     endforeach()
   endif()
 
+  # blackbox tests
   add_executable(driver TestDriver.cpp)
   target_compile_options(driver PRIVATE ${TEST_FLAGS})
   target_link_libraries(driver PRIVATE ${PROJECT_NAME})
-
-  add_executable(script_test ScriptTest.cpp)
-  target_compile_options(script_test PRIVATE ${TEST_FLAGS})
-  target_link_libraries(script_test PRIVATE ${PROJECT_NAME})
-
-  file(GLOB SCRIPT_TESTS ${CMAKE_SOURCE_DIR}/script_tests/*)
-  foreach(TEST ${SCRIPT_TESTS})
-    get_filename_component(name ${TEST} NAME)
-    add_test(NAME conflict_set_script_${name} COMMAND script_test ${TEST})
+  foreach(TEST ${CORPUS_TESTS})
+    get_filename_component(hash ${TEST} NAME)
+    add_test(NAME conflict_set_blackbox_${hash} COMMAND driver ${TEST})
   endforeach()
 
-  add_executable(driver_skip_list TestDriver.cpp)
-  target_compile_options(driver_skip_list PRIVATE ${TEST_FLAGS})
-  target_link_libraries(driver_skip_list PRIVATE skip_list)
+  # scripted tests. Written manually to fill in anything libfuzzer couldn't
+  # find.
+  if(NOT CMAKE_CROSSCOMPILING)
+    find_package(Python3 REQUIRED COMPONENTS Interpreter)
+    set_property(
+      DIRECTORY
+      APPEND
+      PROPERTY CMAKE_CONFIGURE_DEPENDS
+               ${CMAKE_CURRENT_SOURCE_DIR}/test_conflict_set.py)
+    execute_process(
+      COMMAND ${Python3_EXECUTABLE}
+              ${CMAKE_CURRENT_SOURCE_DIR}/test_conflict_set.py list
+      OUTPUT_VARIABLE SCRIPT_TESTS)
+    foreach(TEST ${SCRIPT_TESTS})
+      add_test(
+        NAME script_test_${TEST}
+        COMMAND
+          ${Python3_EXECUTABLE}
+          ${CMAKE_CURRENT_SOURCE_DIR}/test_conflict_set.py test ${TEST}
+          --build-dir ${CMAKE_CURRENT_BINARY_DIR})
+    endforeach()
+  endif()
 
   find_program(VALGRIND_EXE valgrind)
   if(VALGRIND_EXE AND NOT CMAKE_CROSSCOMPILING)
@@ -212,12 +297,6 @@ if(BUILD_TESTING)
                      $<TARGET_FILE:driver> ${CORPUS_TESTS})
   endif()
 
-  foreach(TEST ${CORPUS_TESTS})
-    get_filename_component(hash ${TEST} NAME)
-    add_test(NAME conflict_set_blackbox_${hash} COMMAND driver ${TEST})
-    add_test(NAME skip_list_${hash} COMMAND driver_skip_list ${TEST})
-  endforeach()
-
   # api smoke tests
 
   # c90
@@ -238,31 +317,46 @@ if(BUILD_TESTING)
                         PROPERTIES CXX_STANDARD_REQUIRED ON)
   add_test(NAME conflict_set_cxx_api_test COMMAND conflict_set_cxx_api_test)
 
-  if(NOT APPLE AND NOT CMAKE_BUILD_TYPE STREQUAL Debug)
+  # symbol visibility tests
+  if(NOT WASM AND NOT CMAKE_BUILD_TYPE STREQUAL Debug)
+    if(APPLE)
+      set(symbol_exports ${CMAKE_CURRENT_SOURCE_DIR}/apple-symbol-exports.txt)
+      set(symbol_imports ${CMAKE_CURRENT_SOURCE_DIR}/apple-symbol-imports.txt)
+    else()
+      set(symbol_exports ${CMAKE_CURRENT_SOURCE_DIR}/symbol-exports.txt)
+      if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64)
+        set(symbol_imports
+            ${CMAKE_CURRENT_SOURCE_DIR}/aarch64-symbol-imports.txt)
+      else()
+        set(symbol_imports ${CMAKE_CURRENT_SOURCE_DIR}/symbol-imports.txt)
+      endif()
+    endif()
     add_test(
       NAME conflict_set_shared_symbols
-      COMMAND ${CMAKE_SOURCE_DIR}/test_symbols.sh
-              $<TARGET_FILE:${PROJECT_NAME}> ${CMAKE_SOURCE_DIR}/symbols.txt)
+      COMMAND
+        ${CMAKE_CURRENT_SOURCE_DIR}/test_symbols.sh
+        $<TARGET_FILE:${PROJECT_NAME}> ${symbol_exports} ${symbol_imports})
     add_test(
       NAME conflict_set_static_symbols
       COMMAND
-        ${CMAKE_SOURCE_DIR}/test_symbols.sh
-        $<TARGET_FILE:${PROJECT_NAME}-static> ${CMAKE_SOURCE_DIR}/symbols.txt)
+        ${CMAKE_CURRENT_SOURCE_DIR}/test_symbols.sh
+        $<TARGET_FILE:${PROJECT_NAME}-static> ${symbol_exports}
+        ${symbol_imports})
   endif()
 
   # bench
-
   add_executable(conflict_set_bench Bench.cpp)
   target_link_libraries(conflict_set_bench PRIVATE ${PROJECT_NAME})
   set_target_properties(conflict_set_bench PROPERTIES SKIP_BUILD_RPATH ON)
-
   add_executable(real_data_bench RealDataBench.cpp)
   target_link_libraries(real_data_bench PRIVATE ${PROJECT_NAME})
   set_target_properties(real_data_bench PROPERTIES SKIP_BUILD_RPATH ON)
 endif()
+
 # packaging
 
 set(CPACK_PACKAGE_CONTACT andrew@weaselab.dev)
+set(CMAKE_INSTALL_DEFAULT_COMPONENT_NAME all)
 
 set(CPACK_PACKAGE_VENDOR "Weaselab")
 set(CPACK_RESOURCE_FILE_LICENSE "${CMAKE_CURRENT_SOURCE_DIR}/LICENSE")
@@ -276,6 +370,26 @@ set(CPACK_RPM_FILE_NAME RPM-DEFAULT)
 
 # deb
 set(CPACK_DEBIAN_FILE_NAME DEB-DEFAULT)
+# see *-imports.txt - dependency versions need to be synced with symbol versions
+if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64)
+  set(CPACK_DEBIAN_PACKAGE_DEPENDS "libc6 (>= 2.17)")
+else()
+  set(CPACK_DEBIAN_PACKAGE_DEPENDS "libc6 (>= 2.14)")
+endif()
+
+# macos
+set(CMAKE_OSX_DEPLOYMENT_TARGET 11.0)
+if(APPLE)
+  find_program(PANDOC_EXE pandoc)
+  if(PANDOC_EXE)
+    execute_process(COMMAND ${PANDOC_EXE} ${CMAKE_CURRENT_SOURCE_DIR}/README.md
+                            -o ${CMAKE_CURRENT_BINARY_DIR}/README.txt)
+    set(CPACK_RESOURCE_FILE_README ${CMAKE_CURRENT_BINARY_DIR}/README.txt)
+  endif()
+  configure_file(${CMAKE_CURRENT_SOURCE_DIR}/LICENSE
+                 ${CMAKE_CURRENT_BINARY_DIR}/LICENSE.txt COPYONLY)
+  set(CPACK_RESOURCE_FILE_LICENSE ${CMAKE_CURRENT_BINARY_DIR}/LICENSE.txt)
+endif()
 
 include(CPack)
 
@@ -297,12 +411,15 @@ set_target_properties(
 
 install(
   TARGETS ${PROJECT_NAME} ${PROJECT_NAME}-static
-  EXPORT conflict-setConfig
+  EXPORT ${PROJECT_NAME}Config
   ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
   LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
   RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+
 install(DIRECTORY include/
         DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/${PROJECT_NAME})
 
-install(EXPORT conflict-setConfig
-        DESTINATION ${CMAKE_INSTALL_DATAROOTDIR}/conflict-set/cmake)
+install(EXPORT ${PROJECT_NAME}Config
+        DESTINATION ${CMAKE_INSTALL_DATAROOTDIR}/${PROJECT_NAME}/cmake)
+
+cpack_add_component(all)

HashTable.cpp

@@ -99,7 +99,7 @@ void ConflictSet::setOldestVersion(int64_t oldestVersion) {
 int64_t ConflictSet::getBytes() const { return -1; }
 
 ConflictSet::ConflictSet(int64_t oldestVersion)
-    : impl(new (safe_malloc(sizeof(Impl))) Impl{oldestVersion}) {}
+    : impl(new(safe_malloc(sizeof(Impl))) Impl{oldestVersion}) {}
 
 ConflictSet::~ConflictSet() {
   if (impl) {

Internal.h

@@ -99,8 +99,7 @@ __attribute__((always_inline)) inline void safe_free(void *p, size_t s) {
   mallocBytesDelta -= s;
 #if SHOW_MEMORY
   mallocBytes -= s;
-  free(p);
-#else
+#endif
 #ifndef NDEBUG
   (char *&)p -= kMallocHeaderSize;
   size_t expected;
@@ -108,7 +107,6 @@ __attribute__((always_inline)) inline void safe_free(void *p, size_t s) {
   assert(s == expected);
 #endif
   free(p);
-#endif
 }
 
 // ==================== BEGIN ARENA IMPL ====================
@@ -154,6 +152,7 @@ inline void *operator new[](size_t size, std::align_val_t align,
 
 /// align must be a power of two
 template <class T> T *align_up(T *t, size_t align) {
+  assert(std::popcount(align) == 1);
   auto unaligned = uintptr_t(t);
   auto aligned = (unaligned + align - 1) & ~(align - 1);
   return reinterpret_cast<T *>(reinterpret_cast<char *>(t) + aligned -
@@ -162,6 +161,7 @@ template <class T> T *align_up(T *t, size_t align) {
 
 /// align must be a power of two
 constexpr inline int align_up(uint32_t unaligned, uint32_t align) {
+  assert(std::popcount(align) == 1);
   return (unaligned + align - 1) & ~(align - 1);
 }
 
@@ -177,12 +177,10 @@ struct Arena::ArenaImpl {
   uint8_t *begin() { return reinterpret_cast<uint8_t *>(this + 1); }
 };
 
-static_assert(sizeof(Arena::ArenaImpl) == 16);
-static_assert(alignof(Arena::ArenaImpl) == 8);
-
 inline Arena::Arena(int initialSize) : impl(nullptr) {
   if (initialSize > 0) {
-    auto allocationSize = align_up(initialSize + sizeof(ArenaImpl), 16);
+    auto allocationSize =
+        align_up(initialSize + sizeof(ArenaImpl), alignof(ArenaImpl));
     impl = (Arena::ArenaImpl *)safe_malloc(allocationSize);
     impl->prev = nullptr;
     impl->capacity = allocationSize - sizeof(ArenaImpl);
@@ -218,7 +216,7 @@ inline void *operator new(size_t size, std::align_val_t align, Arena &arena) {
                           (arena.impl ? std::max<int>(sizeof(Arena::ArenaImpl),
                                                       arena.impl->capacity * 2)
                                       : 0)),
-        16);
+        alignof(Arena::ArenaImpl));
     auto *impl = (Arena::ArenaImpl *)safe_malloc(allocationSize);
     impl->prev = arena.impl;
     impl->capacity = allocationSize - sizeof(Arena::ArenaImpl);
@@ -257,10 +255,65 @@ template <class T> struct ArenaAlloc {
   void deallocate(T *, size_t) noexcept {}
 };
 
-template <class T> using Vector = std::vector<T, ArenaAlloc<T>>;
-template <class T> auto vector(Arena &arena) {
-  return Vector<T>(ArenaAlloc<T>(&arena));
-}
+template <class T> struct Vector {
+  static_assert(std::is_trivially_destructible_v<T>);
+  static_assert(std::is_trivially_copyable_v<T>);
+
+  explicit Vector(Arena *arena)
+      : arena(arena), t(nullptr), size_(0), capacity(0) {}
+
+  void append(std::span<const T> slice) {
+    if (size_ + int(slice.size()) > capacity) {
+      grow(std::max<int>(size_ + slice.size(), capacity * 2));
+    }
+    if (slice.size() > 0) {
+      memcpy(const_cast<std::remove_const_t<T> *>(t) + size_, slice.data(),
+             slice.size() * sizeof(T));
+    }
+    size_ += slice.size();
+  }
+
+  void push_back(const T &t) { append(std::span<const T>(&t, 1)); }
+
+  T *begin() { return t; }
+  T *end() { return t + size_; }
+  T *data() { return t; }
+  T &back() {
+    assert(size_ > 0);
+    return t[size_ - 1];
+  }
+  T &operator[](int i) {
+    assert(i >= 0 && i < size_);
+    return t[i];
+  }
+
+  void pop_back() {
+    assert(size_ > 0);
+    --size_;
+  }
+
+  int size() const { return size_; }
+
+  operator std::span<const T>() const { return std::span(t, size_); }
+
+private:
+  void grow(int newCapacity) {
+    capacity = newCapacity;
+    auto old = std::span<const T>(*this);
+    t = (T *)new (std::align_val_t(alignof(T)), *arena)
+        uint8_t[capacity * sizeof(T)];
+    size_ = 0;
+    append(old);
+  }
+
+  Arena *arena;
+  T *t;
+  int size_;
+  int capacity;
+};
+
+template <class T> auto vector(Arena &arena) { return Vector<T>(&arena); }
+
 template <class T, class C> using Set = std::set<T, C, ArenaAlloc<T>>;
 template <class T, class C = std::less<T>> auto set(Arena &arena) {
   return Set<T, C>(ArenaAlloc<T>(&arena));
@@ -414,13 +467,15 @@ inline uint32_t Arbitrary::bounded(uint32_t s) {
 // ==================== END ARBITRARY IMPL ====================
 
 struct ReferenceImpl {
-  explicit ReferenceImpl(int64_t oldestVersion) : oldestVersion(oldestVersion) {
+  explicit ReferenceImpl(int64_t oldestVersion)
+      : oldestVersion(oldestVersion), newestVersion(oldestVersion) {
     writeVersionMap[""] = oldestVersion;
   }
   void check(const ConflictSet::ReadRange *reads, ConflictSet::Result *results,
              int count) const {
     for (int i = 0; i < count; ++i) {
-      if (reads[i].readVersion < oldestVersion) {
+      if (reads[i].readVersion < oldestVersion ||
+          reads[i].readVersion < newestVersion - 2e9) {
         results[i] = ConflictSet::TooOld;
         continue;
       }
@@ -442,6 +497,8 @@ struct ReferenceImpl {
   }
   void addWrites(const ConflictSet::WriteRange *writes, int count,
                  int64_t writeVersion) {
+    assert(writeVersion >= newestVersion);
+    newestVersion = writeVersion;
     for (int i = 0; i < count; ++i) {
       auto begin =
           std::string((const char *)writes[i].begin.p, writes[i].begin.len);
@@ -461,11 +518,12 @@ struct ReferenceImpl {
   }
 
   void setOldestVersion(int64_t oldestVersion) {
-    assert(oldestVersion >= oldestVersion);
+    assert(oldestVersion >= this->oldestVersion);
     this->oldestVersion = oldestVersion;
   }
 
   int64_t oldestVersion;
+  int64_t newestVersion;
   std::map<std::string, int64_t> writeVersionMap;
 };
 
@@ -525,18 +583,22 @@ template <class ConflictSetImpl> struct TestDriver {
   explicit TestDriver(const uint8_t *data, size_t size)
       : arbitrary({data, size}) {}
 
-  int64_t writeVersion = 0;
-  int64_t oldestVersion = 0;
+  int64_t oldestVersion = arbitrary.next();
+  int64_t writeVersion = oldestVersion;
   ConflictSetImpl cs{oldestVersion};
   ReferenceImpl refImpl{oldestVersion};
 
-  constexpr static auto kMaxKeyLen = 8;
+  constexpr static auto kMaxKeySuffixLen = 8;
 
   bool ok = true;
 
+  const int prefixLen = arbitrary.bounded(512);
+  const int prefixByte = arbitrary.randT<uint8_t>();
+
   // Call until it returns true, for "done". Check internal invariants etc
   // between calls to next.
   bool next() {
+    assert(cs.getBytes() >= 0);
     if (!arbitrary.hasEntropy()) {
       return true;
     }
@@ -544,7 +606,8 @@ template <class ConflictSetImpl> struct TestDriver {
     {
       int numPointWrites = arbitrary.bounded(100);
       int numRangeWrites = arbitrary.bounded(100);
-      int64_t v = ++writeVersion;
+      int64_t v = (writeVersion += arbitrary.bounded(10) ? arbitrary.bounded(10)
+                                                         : arbitrary.next());
       auto *writes =
           new (arena) ConflictSet::WriteRange[numPointWrites + numRangeWrites];
       auto keys = set<std::string_view>(arena);
@@ -552,9 +615,10 @@ template <class ConflictSetImpl> struct TestDriver {
         if (!arbitrary.hasEntropy()) {
           return true;
         }
-        int keyLen = arbitrary.bounded(kMaxKeyLen);
+        int keyLen = prefixLen + arbitrary.bounded(kMaxKeySuffixLen);
         auto *begin = new (arena) uint8_t[keyLen];
-        arbitrary.randomBytes(begin, keyLen);
+        memset(begin, prefixByte, prefixLen);
+        arbitrary.randomBytes(begin + prefixLen, keyLen - prefixLen);
         keys.insert(std::string_view((const char *)begin, keyLen));
       }
 
@@ -585,33 +649,41 @@ template <class ConflictSetImpl> struct TestDriver {
         }
 #if DEBUG_VERBOSE && !defined(NDEBUG)
         if (writes[i].end.len == 0) {
-          fprintf(stderr, "Write: {%s} -> %" PRId64 "\n",
-                  printable(writes[i].begin).c_str(), writeVersion);
+          fprintf(stderr, "Write: {%s}\n", printable(writes[i].begin).c_str());
         } else {
-          fprintf(stderr, "Write: [%s, %s) -> %" PRId64 "\n",
+          fprintf(stderr, "Write: [%s, %s)\n",
                   printable(writes[i].begin).c_str(),
-                  printable(writes[i].end).c_str(), writeVersion);
+                  printable(writes[i].end).c_str());
         }
 #endif
       }
       assert(iter == keys.end());
       assert(i == numPointWrites + numRangeWrites);
 
+#if DEBUG_VERBOSE && !defined(NDEBUG)
+      fprintf(stderr, "Write @ %" PRId64 "\n", v);
+#endif
+
       CALLGRIND_START_INSTRUMENTATION;
       cs.addWrites(writes, numPointWrites + numRangeWrites, v);
       CALLGRIND_STOP_INSTRUMENTATION;
 
       refImpl.addWrites(writes, numPointWrites + numRangeWrites, v);
 
-      oldestVersion = std::max<int64_t>(writeVersion - arbitrary.bounded(10),
-                                        oldestVersion);
+      oldestVersion +=
+          arbitrary.bounded(10) ? arbitrary.bounded(10) : arbitrary.next();
+      oldestVersion = std::min(oldestVersion, writeVersion);
       cs.setOldestVersion(oldestVersion);
       refImpl.setOldestVersion(oldestVersion);
     }
     {
       int numPointReads = arbitrary.bounded(100);
       int numRangeReads = arbitrary.bounded(100);
-      int64_t v = std::max<int64_t>(writeVersion - arbitrary.bounded(10), 0);
+
+      int64_t v = std::max<int64_t>(writeVersion - (arbitrary.bounded(10)
+                                                        ? arbitrary.bounded(10)
+                                                        : arbitrary.next()),
+                                    0);
       auto *reads =
           new (arena) ConflictSet::ReadRange[numPointReads + numRangeReads];
       auto keys = set<std::string_view>(arena);
@@ -619,9 +691,10 @@ template <class ConflictSetImpl> struct TestDriver {
         if (!arbitrary.hasEntropy()) {
           return true;
         }
-        int keyLen = arbitrary.bounded(kMaxKeyLen);
+        int keyLen = prefixLen + arbitrary.bounded(kMaxKeySuffixLen);
         auto *begin = new (arena) uint8_t[keyLen];
-        arbitrary.randomBytes(begin, keyLen);
+        memset(begin, prefixByte, prefixLen);
+        arbitrary.randomBytes(begin + prefixLen, keyLen - prefixLen);
         keys.insert(std::string_view((const char *)begin, keyLen));
       }
 
@@ -652,12 +725,12 @@ template <class ConflictSetImpl> struct TestDriver {
         reads[i].readVersion = v;
 #if DEBUG_VERBOSE && !defined(NDEBUG)
         if (reads[i].end.len == 0) {
-          fprintf(stderr, "Read: {%s} @ %d\n",
-                  printable(reads[i].begin).c_str(), int(reads[i].readVersion));
+          fprintf(stderr, "Read: {%s} @ %" PRId64 "\n",
+                  printable(reads[i].begin).c_str(), reads[i].readVersion);
         } else {
-          fprintf(stderr, "Read: [%s, %s) @ %d\n",
+          fprintf(stderr, "Read: [%s, %s) @ %" PRId64 "\n",
                   printable(reads[i].begin).c_str(),
-                  printable(reads[i].end).c_str(), int(reads[i].readVersion));
+                  printable(reads[i].end).c_str(), reads[i].readVersion);
         }
 #endif
       }

Jenkinsfile

@@ -110,7 +110,11 @@ pipeline {
                 sh '''
                 gcovr -f ConflictSet.cpp --cobertura > build/coverage.xml
                 '''
-                cobertura autoUpdateHealth: false, autoUpdateStability: false, coberturaReportFile: 'build/coverage.xml', conditionalCoverageTargets: '70, 0, 0', failUnhealthy: false, failUnstable: false, lineCoverageTargets: '80, 0, 0', maxNumberOfBuilds: 0, methodCoverageTargets: '80, 0, 0', onlyStable: false, sourceEncoding: 'ASCII', zoomCoverageChart: false
+                recordCoverage qualityGates: [[criticality: 'NOTE', metric: 'MODULE']], tools: [[parser: 'COBERTURA', pattern: 'build/coverage.xml']]
+                sh '''
+                # Suppress again, because we haven't dealt with function multi-versioning for x86 yet
+                # gcovr -f ConflictSet.cpp --fail-under-line 100 > /dev/null
+                '''
             }
         }
     }

README.md

@@ -60,25 +60,27 @@ Performance counters:
 
 |               ns/op |                op/s |    err% |     total | benchmark
 |--------------------:|--------------------:|--------:|----------:|:----------
-|              246.99 |        4,048,700.59 |    0.2% |      0.01 | `point reads`
-|              260.16 |        3,843,784.65 |    0.1% |      0.01 | `prefix reads`
-|              493.35 |        2,026,953.19 |    0.1% |      0.01 | `range reads`
-|              462.05 |        2,164,289.23 |    0.6% |      0.01 | `point writes`
-|              448.19 |        2,231,205.25 |    0.9% |      0.01 | `prefix writes`
-|              255.83 |        3,908,845.72 |    1.5% |      0.02 | `range writes`
-|              582.63 |        1,716,349.02 |    1.3% |      0.01 | `monotonic increasing point writes`
+|              256.89 |        3,892,784.92 |    0.3% |      0.01 | `point reads`
+|              272.90 |        3,664,395.04 |    0.2% |      0.01 | `prefix reads`
+|              507.22 |        1,971,549.50 |    0.7% |      0.01 | `range reads`
+|              452.66 |        2,209,181.91 |    0.5% |      0.01 | `point writes`
+|              438.09 |        2,282,619.96 |    0.4% |      0.01 | `prefix writes`
+|              253.33 |        3,947,420.36 |    2.5% |      0.02 | `range writes`
+|              574.07 |        1,741,936.71 |    0.3% |      0.01 | `monotonic increasing point writes`
+|          151,562.50 |            6,597.94 |    1.5% |      0.01 | `worst case for radix tree`
 
 ## Radix tree (this implementation)
 
 |               ns/op |                op/s |    err% |     total | benchmark
 |--------------------:|--------------------:|--------:|----------:|:----------
-|               19.42 |       51,483,206.67 |    0.3% |      0.01 | `point reads`
-|               58.43 |       17,115,612.57 |    0.1% |      0.01 | `prefix reads`
-|              216.09 |        4,627,766.60 |    0.2% |      0.01 | `range reads`
-|               28.35 |       35,267,567.72 |    0.2% |      0.01 | `point writes`
-|               43.43 |       23,026,226.17 |    0.2% |      0.01 | `prefix writes`
+|               19.83 |       50,420,955.28 |    0.1% |      0.01 | `point reads`
+|               55.95 |       17,872,542.40 |    0.5% |      0.01 | `prefix reads`
+|               88.28 |       11,327,709.50 |    0.4% |      0.01 | `range reads`
+|               29.15 |       34,309,531.64 |    0.5% |      0.01 | `point writes`
+|               42.36 |       23,607,424.27 |    1.1% |      0.01 | `prefix writes`
 |               50.00 |       20,000,000.00 |    0.0% |      0.01 | `range writes`
-|               92.38 |       10,824,863.69 |    4.1% |      0.01 | `monotonic increasing point writes`
+|               93.52 |       10,692,413.79 |    3.3% |      0.01 | `monotonic increasing point writes`
+|        2,388,417.00 |              418.69 |    0.4% |      0.03 | `worst case for radix tree`
 
 # "Real data" test
 

RealDataBench.cpp

@@ -1,8 +1,11 @@
 #include <ConflictSet.h>
 
+#include <cerrno>
 #include <chrono>
+#include <cstdio>
 #include <cstring>
 #include <fcntl.h>
+#include <string_view>
 #include <sys/mman.h>
 #include <sys/stat.h>
 #include <unistd.h>

ScriptTest.cpp

@@ -1,155 +0,0 @@
-#include "Internal.h"
-#include <ConflictSet.h>
-
-#include <chrono>
-#include <cstring>
-#include <fcntl.h>
-#include <sys/mman.h>
-#include <sys/stat.h>
-#include <unistd.h>
-#include <vector>
-
-inline size_t getPageSize() {
-  static size_t kPageSize = sysconf(_SC_PAGESIZE);
-  return kPageSize;
-}
-
-/// Helper for rounding up to page size (or some other alignment)
-constexpr inline size_t rightAlign(size_t offset, size_t alignment) {
-  return offset % alignment == 0 ? offset
-                                 : ((offset / alignment) + 1) * alignment;
-}
-
-using StringView = std::basic_string_view<uint8_t>;
-
-inline StringView operator"" _v(const char *str, size_t size) {
-  return {reinterpret_cast<const uint8_t *>(str), size};
-}
-
-int main(int argc, const char **argv) {
-
-  ConflictSet cs{0};
-  ReferenceImpl ref{0};
-
-  for (int i = 1; i < argc; ++i) {
-    int fd = open(argv[i], O_RDONLY);
-    struct stat st;
-    if (fstat(fd, &st) == -1) {
-      int err = errno;
-      fprintf(stderr, "stat error %s - %s\n", argv[i], strerror(err));
-      fflush(stderr);
-      abort();
-    }
-
-    int64_t size = rightAlign(st.st_size, getPageSize());
-    const uint8_t *begin =
-        (uint8_t *)mmap(0, size, PROT_READ, MAP_PRIVATE, fd, 0);
-    madvise((void *)begin, size, MADV_SEQUENTIAL);
-    auto *const mapOriginal = begin;
-    const auto sizeOriginal = size;
-
-    StringView b;
-    StringView e;
-    int64_t v = 0;
-    int64_t lastWriteVersion = 0;
-    int64_t lastOldestVersion = 0;
-    std::vector<ConflictSet::WriteRange> writeRanges;
-    std::vector<ConflictSet::ReadRange> readRanges;
-    std::vector<ConflictSet::Result> results;
-
-    for (uint8_t *end = (uint8_t *)memchr(begin, '\n', size); end != nullptr;) {
-      StringView line{begin, static_cast<size_t>(end - begin)};
-      size -= end - begin + 1;
-      begin = end + 1;
-      end = (uint8_t *)memchr(begin, '\n', size);
-
-      if (line.starts_with("begin"_v)) {
-        b = line.substr("begin "_v.size(), line.size());
-        printf("b <- %.*s\n", int(b.size()), b.data());
-      } else if (line.starts_with("end"_v)) {
-        e = line.substr("end "_v.size(), line.size());
-        printf("e <- %.*s\n", int(e.size()), e.data());
-      } else if (line.starts_with("version"_v)) {
-        line = line.substr("version "_v.size(), line.size());
-        v = 0;
-        for (auto c : line) {
-          v = v * 10 + int(c) - int('0');
-        }
-        printf("v <- %" PRId64 "\n", v);
-      } else if (line.starts_with("pointread"_v)) {
-        printf("pointread\n");
-        ConflictSet::ReadRange r;
-        r.begin.p = b.data();
-        r.begin.len = b.size();
-        r.end.len = 0;
-        r.readVersion = v;
-        readRanges.push_back(r);
-      } else if (line.starts_with("pointwrite"_v)) {
-        printf("pointwrite\n");
-        assert(writeRanges.empty() ||
-               (writeRanges.back().end.len == 0 ? writeRanges.back().begin
-                                                : writeRanges.back().end) < b);
-        ConflictSet::WriteRange w;
-        w.begin.p = b.data();
-        w.begin.len = b.size();
-        w.end.len = 0;
-        writeRanges.push_back(w);
-      } else if (line.starts_with("rangeread"_v)) {
-        printf("rangeread\n");
-        ConflictSet::ReadRange r;
-        r.begin.p = b.data();
-        r.begin.len = b.size();
-        r.end.p = e.data();
-        r.end.len = e.size();
-        r.readVersion = v;
-        readRanges.push_back(r);
-      } else if (line.starts_with("rangewrite"_v)) {
-        printf("rangewrite\n");
-        assert(b < e);
-        assert(writeRanges.empty() ||
-               (writeRanges.back().end.len == 0 ? writeRanges.back().begin
-                                                : writeRanges.back().end) < b);
-        ConflictSet::WriteRange w;
-        w.begin.p = b.data();
-        w.begin.len = b.size();
-        w.end.p = e.data();
-        w.end.len = e.size();
-        writeRanges.push_back(w);
-      } else if (line.starts_with("check"_v)) {
-        printf("check\n");
-        Arena arena;
-        auto *expected = new (arena) ConflictSet::Result[readRanges.size()];
-        auto *actual = new (arena) ConflictSet::Result[readRanges.size()];
-        ref.check(readRanges.data(), expected, readRanges.size());
-        cs.check(readRanges.data(), actual, readRanges.size());
-        for (int i = 0; i < int(readRanges.size()); ++i) {
-          if (expected[i] != actual[i]) {
-            fprintf(stderr, "Expected %s, got %s at index %d\n",
-                    resultToStr(expected[i]), resultToStr(actual[i]), i);
-            return 1;
-          }
-        }
-        readRanges = {};
-      } else if (line.starts_with("addwrites"_v)) {
-        printf("addwrites\n");
-        assert(v > lastWriteVersion);
-        lastWriteVersion = v;
-        cs.addWrites(writeRanges.data(), writeRanges.size(), v);
-        ref.addWrites(writeRanges.data(), writeRanges.size(), v);
-        writeRanges = {};
-      } else if (line.starts_with("setoldest"_v)) {
-        printf("setoldest\n");
-        assert(v > lastOldestVersion);
-        lastOldestVersion = v;
-        cs.setOldestVersion(v);
-        ref.setOldestVersion(v);
-      } else if (line.empty() || line.starts_with(";"_v)) {
-        // skip
-      } else {
-        printf("Unrecognized line: %.*s\n", int(line.size()), line.data());
-      }
-    }
-    munmap((void *)mapOriginal, sizeOriginal);
-    close(fd);
-  }
-}

SkipList.cpp

@@ -16,6 +16,8 @@
  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  * See the License for the specific language governing permissions and
  * limitations under the License.
+ *
+ * This source code is modified to compile outside of FoundationDB
  */
 
 #include "ConflictSet.h"
@@ -268,13 +270,21 @@ public:
   }
 
   explicit SkipList(Version version = 0) {
+#if DEBUG_VERBOSE
+    fprintf(stderr, "skip_list: create\n");
+#endif
     header = Node::create(StringRef(), MaxLevels - 1);
     for (int l = 0; l < MaxLevels; l++) {
       header->setNext(l, nullptr);
       header->setMaxVersion(l, version);
     }
   }
-  ~SkipList() { destroy(); }
+  ~SkipList() {
+#if DEBUG_VERBOSE
+    fprintf(stderr, "skip_list: destroy\n");
+#endif
+    destroy();
+  }
   SkipList(SkipList &&other) noexcept : header(other.header) {
     other.header = nullptr;
   }
@@ -567,7 +577,8 @@ struct SkipListConflictSet {};
 
 struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
   Impl(int64_t oldestVersion)
-      : oldestVersion(oldestVersion), skipList(oldestVersion) {}
+      : oldestVersion(oldestVersion), newestVersion(oldestVersion),
+        skipList(oldestVersion) {}
   void check(const ConflictSet::ReadRange *reads, ConflictSet::Result *results,
              int count) const {
     Arena arena;
@@ -582,7 +593,8 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
     }
     skipList.detectConflicts(ranges, count, results);
     for (int i = 0; i < count; ++i) {
-      if (reads[i].readVersion < oldestVersion) {
+      if (reads[i].readVersion < oldestVersion ||
+          reads[i].readVersion < newestVersion - 2e9) {
         results[i] = TooOld;
       }
     }
@@ -590,6 +602,8 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
 
   void addWrites(const ConflictSet::WriteRange *writes, int count,
                  int64_t writeVersion) {
+    assert(writeVersion >= newestVersion);
+    newestVersion = writeVersion;
     Arena arena;
     const int stringCount = count * 2;
 
@@ -603,6 +617,10 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
     for (int s = stripes - 1; s >= 0; s--) {
       for (int i = 0; i * 2 < ss; ++i) {
         const auto &w = writes[s * stripeSize / 2 + i];
+#if DEBUG_VERBOSE
+        printf("Write begin: %s\n", printable(w.begin).c_str());
+        fflush(stdout);
+#endif
         values[i * 2] = {w.begin.p, size_t(w.begin.len)};
         values[i * 2 + 1] = w.end.len > 0
                                 ? StringRef{w.end.p, size_t(w.end.len)}
@@ -616,6 +634,7 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
   }
 
   void setOldestVersion(int64_t oldestVersion) {
+    assert(oldestVersion >= this->oldestVersion);
     this->oldestVersion = oldestVersion;
     SkipList::Finger finger;
     int temp;
@@ -634,16 +653,20 @@ private:
   Arena removalArena;
   std::span<const uint8_t> removalKey;
   int64_t oldestVersion;
+  int64_t newestVersion;
   SkipList skipList;
 };
 
-void ConflictSet::check(const ReadRange *reads, Result *results,
-                        int count) const {
+// Internal entry points. Public entry points should just delegate to these
+
+void internal_check(ConflictSet::Impl *impl,
+                    const ConflictSet::ReadRange *reads,
+                    ConflictSet::Result *results, int count) {
   impl->check(reads, results, count);
 }
-
-void ConflictSet::addWrites(const WriteRange *writes, int count,
-                            int64_t writeVersion) {
+void internal_addWrites(ConflictSet::Impl *impl,
+                        const ConflictSet::WriteRange *writes, int count,
+                        int64_t writeVersion) {
   mallocBytesDelta = 0;
   impl->addWrites(writes, count, writeVersion);
   impl->totalBytes += mallocBytesDelta;
@@ -654,7 +677,7 @@ void ConflictSet::addWrites(const WriteRange *writes, int count,
 #endif
 }
 
-void ConflictSet::setOldestVersion(int64_t oldestVersion) {
+void internal_setOldestVersion(ConflictSet::Impl *impl, int64_t oldestVersion) {
   mallocBytesDelta = 0;
   impl->setOldestVersion(oldestVersion);
   impl->totalBytes += mallocBytesDelta;
@@ -664,19 +687,43 @@ void ConflictSet::setOldestVersion(int64_t oldestVersion) {
   }
 #endif
 }
+ConflictSet::Impl *internal_create(int64_t oldestVersion) {
+  mallocBytesDelta = 0;
+  auto *result = new (safe_malloc(sizeof(ConflictSet::Impl)))
+      ConflictSet::Impl{oldestVersion};
+  result->totalBytes += mallocBytesDelta;
+  return result;
+}
 
-int64_t ConflictSet::getBytes() const { return impl->totalBytes; }
+void internal_destroy(ConflictSet::Impl *impl) {
+  impl->~Impl();
+  safe_free(impl, sizeof(ConflictSet::Impl));
+}
+
+int64_t internal_getBytes(ConflictSet::Impl *impl) { return impl->totalBytes; }
+
+void ConflictSet::check(const ReadRange *reads, Result *results,
+                        int count) const {
+  internal_check(impl, reads, results, count);
+}
+
+void ConflictSet::addWrites(const WriteRange *writes, int count,
+                            int64_t writeVersion) {
+  internal_addWrites(impl, writes, count, writeVersion);
+}
+
+void ConflictSet::setOldestVersion(int64_t oldestVersion) {
+  internal_setOldestVersion(impl, oldestVersion);
+}
+
+int64_t ConflictSet::getBytes() const { return internal_getBytes(impl); }
 
 ConflictSet::ConflictSet(int64_t oldestVersion)
-    : impl((mallocBytesDelta = 0,
-            new (safe_malloc(sizeof(Impl))) Impl{oldestVersion})) {
-  impl->totalBytes += mallocBytesDelta;
-}
+    : impl(internal_create(oldestVersion)) {}
 
 ConflictSet::~ConflictSet() {
   if (impl) {
-    impl->~Impl();
-    safe_free(impl, sizeof(Impl));
+    internal_destroy(impl);
   }
 }
 
@@ -697,37 +744,34 @@ extern "C" {
 __attribute__((__visibility__("default"))) void
 ConflictSet_check(void *cs, const ConflictSet_ReadRange *reads,
                   ConflictSet_Result *results, int count) {
-  ((ConflictSet::Impl *)cs)->check(reads, results, count);
+  internal_check((ConflictSet::Impl *)cs, reads, results, count);
 }
 __attribute__((__visibility__("default"))) void
 ConflictSet_addWrites(void *cs, const ConflictSet_WriteRange *writes, int count,
                       int64_t writeVersion) {
-  ((ConflictSet::Impl *)cs)->addWrites(writes, count, writeVersion);
+  internal_addWrites((ConflictSet::Impl *)cs, writes, count, writeVersion);
 }
 __attribute__((__visibility__("default"))) void
 ConflictSet_setOldestVersion(void *cs, int64_t oldestVersion) {
-  ((ConflictSet::Impl *)cs)->setOldestVersion(oldestVersion);
+  internal_setOldestVersion((ConflictSet::Impl *)cs, oldestVersion);
 }
 __attribute__((__visibility__("default"))) void *
 ConflictSet_create(int64_t oldestVersion) {
-  return new (safe_malloc(sizeof(ConflictSet::Impl)))
-      ConflictSet::Impl{oldestVersion};
+  return internal_create(oldestVersion);
 }
 __attribute__((__visibility__("default"))) void ConflictSet_destroy(void *cs) {
-  using Impl = ConflictSet::Impl;
-  ((Impl *)cs)->~Impl();
-  safe_free(cs, sizeof(Impl));
+  internal_destroy((ConflictSet::Impl *)cs);
 }
 __attribute__((__visibility__("default"))) int64_t
 ConflictSet_getBytes(void *cs) {
-  using Impl = ConflictSet::Impl;
-  return ((Impl *)cs)->totalBytes;
+  return internal_getBytes((ConflictSet::Impl *)cs);
 }
 }
 
 #if SHOW_MEMORY
 struct __attribute__((visibility("default"))) PeakPrinter {
   ~PeakPrinter() {
+    printf("--- skip_list ---\n");
     printf("malloc bytes: %g\n", double(mallocBytes));
     printf("Peak malloc bytes: %g\n", double(peakMallocBytes));
   }

aarch64-symbol-imports.txt

@@ -0,0 +1,8 @@
+__stack_chk_fail@GLIBC_2.17
+__stack_chk_guard@GLIBC_2.17
+abort@GLIBC_2.17
+free@GLIBC_2.17
+malloc@GLIBC_2.17
+memcpy@GLIBC_2.17
+memmove@GLIBC_2.17
+memset@GLIBC_2.17

apple-symbol-exports.txt

@@ -0,0 +1,17 @@
+_ConflictSet_addWrites
+_ConflictSet_check
+_ConflictSet_create
+_ConflictSet_destroy
+_ConflictSet_getBytes
+_ConflictSet_setOldestVersion
+__ZN8weaselab11ConflictSet16setOldestVersionEx
+__ZN8weaselab11ConflictSet9addWritesEPKNS0_10WriteRangeEix
+__ZN8weaselab11ConflictSetC1EOS0_
+__ZN8weaselab11ConflictSetC1Ex
+__ZN8weaselab11ConflictSetC2EOS0_
+__ZN8weaselab11ConflictSetC2Ex
+__ZN8weaselab11ConflictSetD1Ev
+__ZN8weaselab11ConflictSetD2Ev
+__ZN8weaselab11ConflictSetaSEOS0_
+__ZNK8weaselab11ConflictSet5checkEPKNS0_9ReadRangeEPNS0_6ResultEi
+__ZNK8weaselab11ConflictSet8getBytesEv

apple-symbol-imports.txt

@@ -0,0 +1,10 @@
+___stack_chk_fail
+___stack_chk_guard
+__tlv_bootstrap
+_abort
+_bzero
+_free
+_malloc
+_memcpy
+_memmove
+dyld_stub_binder

conflict_set.py

@@ -0,0 +1,138 @@
+import ctypes
+import enum
+import os
+
+from typing import Optional
+
+
+class _Key(ctypes.Structure):
+    _fields_ = [("p", ctypes.POINTER(ctypes.c_ubyte)), ("len", ctypes.c_int)]
+
+
+class ReadRange(ctypes.Structure):
+    _fields_ = [
+        ("begin", _Key),
+        ("end", _Key),
+        ("readVersion", ctypes.c_int64),
+    ]
+
+
+class WriteRange(ctypes.Structure):
+    _fields_ = [("begin", _Key), ("end", _Key)]
+
+
+class Result(enum.Enum):
+    COMMIT = 0
+    CONFLICT = 1
+    TOO_OLD = 2
+
+
+def write(begin: bytes, end: Optional[bytes] = None) -> WriteRange:
+    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
+
+    if end is None:
+        e = (ctypes.c_ubyte * 0)()
+    else:
+        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
+    return WriteRange(_Key(b, len(b)), _Key(e, len(e)))
+
+
+def read(version: int, begin: bytes, end: Optional[bytes] = None) -> ReadRange:
+    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
+    if end is None:
+        e = (ctypes.c_ubyte * 0)()
+    else:
+        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
+    return ReadRange(_Key(b, len(b)), _Key(e, len(e)), version)
+
+
+class ConflictSet:
+    def __init__(
+        self,
+        version: int = 0,
+        build_dir: Optional[str] = None,
+        implementation: Optional[str] = None,
+    ) -> None:
+        self._lib = None
+        if build_dir is None:
+            build_dir = os.path.dirname(__file__) + "/build"
+        if implementation is None:
+            implementation = "radix_tree"
+        for f in (
+            build_dir + "/" + implementation + "/libconflict-set.so.0",
+            os.path.dirname(__file__)
+            + "/build/"
+            + implementation
+            + "/libconflict-set.0.dylib",
+        ):
+            try:
+                self._lib = ctypes.cdll.LoadLibrary(f)
+            except:
+                pass
+
+        if self._lib is None:
+            import sys
+
+            print(
+                "Could not find libconflict-set implementation " + implementation,
+                file=sys.stderr,
+            )
+            sys.exit(1)
+
+        self._lib.ConflictSet_create.argtypes = (ctypes.c_int64,)
+        self._lib.ConflictSet_create.restype = ctypes.c_void_p
+
+        self._lib.ConflictSet_check.argtypes = (
+            ctypes.c_void_p,
+            ctypes.POINTER(ReadRange),
+            ctypes.POINTER(ctypes.c_int),
+            ctypes.c_int,
+        )
+
+        self._lib.ConflictSet_addWrites.argtypes = (
+            ctypes.c_void_p,
+            ctypes.POINTER(WriteRange),
+            ctypes.c_int,
+            ctypes.c_int64,
+        )
+
+        self._lib.ConflictSet_setOldestVersion.argtypes = (
+            ctypes.c_void_p,
+            ctypes.c_int64,
+        )
+
+        self._lib.ConflictSet_destroy.argtypes = (ctypes.c_void_p,)
+
+        self._lib.ConflictSet_getBytes.argtypes = (ctypes.c_void_p,)
+        self._lib.ConflictSet_getBytes.restype = ctypes.c_int64
+
+        self.p = self._lib.ConflictSet_create(version)
+
+    def addWrites(self, version: int, *writes: WriteRange):
+        self._lib.ConflictSet_addWrites(
+            self.p, (WriteRange * len(writes))(*writes), len(writes), version
+        )
+
+    def check(self, *reads: ReadRange) -> list[Result]:
+        r = (ctypes.c_int * len(reads))()
+        self._lib.ConflictSet_check(
+            self.p, (ReadRange * len(reads))(*reads), r, len(reads)
+        )
+        return [Result(x) for x in r]
+
+    def setOldestVersion(self, version: int) -> None:
+        self._lib.ConflictSet_setOldestVersion(self.p, version)
+
+    def getBytes(self) -> int:
+        return self._lib.ConflictSet_getBytes(self.p)
+
+    def __enter__(self):
+        return self
+
+    def close(self) -> None:
+        if self.p is not None:
+            self._lib.ConflictSet_destroy(self.p)
+            self.p = None
+
+    def __exit__(self, exception_type, exception_value, exception_traceback):
+        self.close()

corpus/0e410d4488defd877d2c9690a26970f57ab78b0c

@@ -0,0 +1,4 @@
+*
+
+
+

corpus/22c23c2418773edaaff055ed387f82c8ad078f9f

@@ -0,0 +1,2 @@
+*
+