Install mc in the msan CI job

The msan job uploads test results to MinIO with `mc cp`, but unlike the
test/release/coverage jobs it never installed the mc binary.  When
MinIO credentials are configured (i.e. after merge) the upload step
would fail.  Install mc to match the other jobs.

.gitea/workflows/ci.yml

@@ -2,62 +2,37 @@ name: CI
 
 on: [push, pull_request]
 
+env:
+  CC: clang
+  CXX: clang++
+
 jobs:
-  build-image:
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - runner: ubuntu-latest-amd64
-            arch: amd64
-          - runner: ubuntu-latest-arm64
-            arch: arm64
-    runs-on: ${{ matrix.runner }}
-    steps:
-      - uses: actions/checkout@v4
-
-      - name: Log in to registry
-        env:
-          REGISTRY_USER: ${{ secrets.REGISTRY_USER }}
-          REGISTRY_TOKEN: ${{ secrets.REGISTRY_TOKEN }}
-        run: |
-          echo "$REGISTRY_TOKEN" \
-            | docker login -u "$REGISTRY_USER" --password-stdin git.weaselab.dev
-
-      - name: Build and push image if changed
-        run: |
-          image=git.weaselab.dev/weaselab/conflict-set-ci
-          hash="$(sha256sum Dockerfile .pre-commit-config.yaml | sha256sum | cut -c 1-16)"
-          latest="$image:latest-${{ matrix.arch }}"
-          current="$(docker buildx imagetools inspect "$latest" \
-            --format '{{index .Image.Config.Labels "dev.weaselab.ci-hash"}}' 2> /dev/null || true)"
-          if [ "$current" = "$hash" ]; then
-            echo "$latest is up to date"
-          else
-            docker build --push --label "dev.weaselab.ci-hash=$hash" -t "$latest" .
-          fi
-
   pre-commit:
-    needs: build-image
     runs-on: ubuntu-latest-amd64
-    container:
-      image: git.weaselab.dev/weaselab/conflict-set-ci:latest-amd64
-      credentials:
-        username: ${{ secrets.REGISTRY_USER }}
-        password: ${{ secrets.REGISTRY_TOKEN }}
     steps:
       - uses: actions/checkout@v4
 
+      - uses: actions/cache@v4
+        with:
+          path: /var/cache/apt/archives
+          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
+
+      - name: Install dependencies
+        run: |
+          sudo apt-get update -qq
+          sudo apt-get install -y git nodejs pre-commit
+
+      - uses: actions/cache@v4
+        with:
+          path: ~/.cache/pre-commit
+          key: pre-commit-${{ hashFiles('.pre-commit-config.yaml') }}
+
       - name: Run pre-commit
-        env:
-          # use the hooks pre-installed in the image
-          HOME: /tmp
         run: |
           git config --global --add safe.directory "$PWD"
           pre-commit run --all-files --show-diff-on-failure
 
   test:
-    needs: build-image
     strategy:
       fail-fast: false
       matrix:
@@ -71,20 +46,35 @@ jobs:
           - name: gcc
             cmake_args: -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++
     runs-on: ubuntu-latest-amd64
-    container:
-      image: git.weaselab.dev/weaselab/conflict-set-ci:latest-amd64
-      credentials:
-        username: ${{ secrets.REGISTRY_USER }}
-        password: ${{ secrets.REGISTRY_TOKEN }}
     steps:
       - uses: actions/checkout@v4
 
       - uses: actions/cache@v4
         with:
-          path: .ccache
+          path: /var/cache/apt/archives
-          key: ccache-${{ matrix.name }}-${{ gitea.sha }}
+          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
-          restore-keys: |
+
-            ccache-${{ matrix.name }}-
+      - name: Install common dependencies
+        run: |
+          sudo apt-get update -qq
+          sudo apt-get install -y \
+            build-essential ccache cmake gcc g++ libc6-dbg \
+            ninja-build python3 valgrind zstd
+          sudo curl -Ls "https://minio.weaselab.dev/public/$(uname -m)/mc.RELEASE.2025-08-13T08-35-41Z" \
+            -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
+
+      - name: Install LLVM toolchain
+        if: matrix.name != 'gcc'
+        run: |
+          . /etc/os-release
+          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
+          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
+          sudo apt-get update -qq
+          sudo apt-get install -y \
+            clang-21 llvm-21 lld-21 mold
+          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
+            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
+          done
 
       - name: Build
         run: |
@@ -120,7 +110,6 @@ jobs:
             | tee -a "$GITHUB_STEP_SUMMARY"
 
   release:
-    needs: build-image
     strategy:
       fail-fast: false
       matrix:
@@ -130,14 +119,38 @@ jobs:
           - runner: ubuntu-latest-arm64
             arch: arm64
     runs-on: ${{ matrix.runner }}
-    container:
-      image: git.weaselab.dev/weaselab/conflict-set-ci:latest-${{ matrix.arch }}
-      credentials:
-        username: ${{ secrets.REGISTRY_USER }}
-        password: ${{ secrets.REGISTRY_TOKEN }}
     steps:
       - uses: actions/checkout@v4
 
+      - uses: actions/cache@v4
+        with:
+          path: /var/cache/apt/archives
+          key: apt-${{ matrix.arch }}-${{ hashFiles('.gitea/workflows/ci.yml') }}
+
+      - name: Install dependencies
+        run: |
+          . /etc/os-release
+          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
+          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
+          sudo apt-get update -qq
+          sudo apt-get install -y \
+            build-essential ccache clang-21 cmake devscripts \
+            libc6-dbg llvm-21 lld-21 mold ninja-build rpm \
+            valgrind zstd
+          sudo curl -Ls "https://minio.weaselab.dev/public/$(uname -m)/mc.RELEASE.2025-08-13T08-35-41Z" \
+            -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
+          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
+            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
+          done
+
+      - name: Install paper build dependencies
+        if: matrix.arch == 'amd64'
+        run: |
+          sudo apt-get install -y \
+            biber latexmk \
+            texlive-bibtex-extra texlive-fonts-recommended \
+            texlive-latex-extra texlive-pictures
+
       - uses: actions/cache@v4
         with:
           path: .ccache
@@ -188,9 +201,6 @@ jobs:
           dest="minio/jenkins/conflict-set/${{ gitea.run_number }}/release-${{ matrix.arch }}/"
           zstd build/Testing/*/Test.xml
           mc cp build/Testing/*/Test.xml.zst "$dest"
-          # This step runs even when a previous step failed, to upload test
-          # results. The packages may never have been built though, so skip
-          # them if they're missing.
           if compgen -G "build/*.deb" > /dev/null; then
             mc cp build/*.deb "$dest"
           fi
@@ -208,17 +218,114 @@ jobs:
             --link "https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/release-${{ matrix.arch }}/Test.xml.zst" \
             | tee -a "$GITHUB_STEP_SUMMARY"
 
-  coverage:
+  msan:
-    needs: build-image
     runs-on: ubuntu-latest-amd64
-    container:
+    env:
-      image: git.weaselab.dev/weaselab/conflict-set-ci:latest-amd64
+      MSAN_VERSION: "21.1.8"
-      credentials:
+      MSAN_URL: "https://minio.weaselab.dev/public/x86_64/msan-toolchain-21.1.8.tar.zst"
-        username: ${{ secrets.REGISTRY_USER }}
-        password: ${{ secrets.REGISTRY_TOKEN }}
     steps:
       - uses: actions/checkout@v4
 
+      - uses: actions/cache@v4
+        with:
+          path: /var/cache/apt/archives
+          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
+
+      - name: Install dependencies
+        run: |
+          . /etc/os-release
+          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
+          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
+          sudo apt-get update -qq
+          sudo apt-get install -y \
+            build-essential ccache clang-21 cmake libc6-dbg \
+            llvm-21 lld-21 mold ninja-build python3 zstd
+          sudo curl -Ls "https://minio.weaselab.dev/public/$(uname -m)/mc.RELEASE.2025-08-13T08-35-41Z" \
+            -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
+          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
+            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
+          done
+
+      - name: Download MSan toolchain
+        run: |
+          curl -Ls "${MSAN_URL}" -o /tmp/msan-toolchain.tar.zst
+          sudo mkdir -p /opt/msan
+          sudo tar --zstd -xf /tmp/msan-toolchain.tar.zst -C /opt/msan
+
+      - uses: actions/cache@v4
+        with:
+          path: .ccache
+          key: ccache-msan-${{ gitea.sha }}
+          restore-keys: |
+            ccache-msan-
+
+      - name: Build
+        run: |
+          export CCACHE_DIR="$GITHUB_WORKSPACE/.ccache"
+          rm -rf build
+          cmake -S . -B build -G Ninja \
+            -DCMAKE_C_COMPILER=clang \
+            -DCMAKE_CXX_COMPILER=clang++ \
+            -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+            -DCMAKE_BUILD_TYPE=Debug \
+            -DDISABLE_TSAN=ON \
+            -DUSE_MSAN=ON \
+            -DCMAKE_CXX_FLAGS="-fsanitize=memory -stdlib=libc++ -I/opt/msan/include/c++/v1 -L/opt/msan/lib -UNDEBUG" \
+            -DCMAKE_EXE_LINKER_FLAGS="-stdlib=libc++ -Wl,-rpath,/opt/msan/lib" \
+            -DCMAKE_SHARED_LINKER_FLAGS="-stdlib=libc++ -Wl,-rpath,/opt/msan/lib"
+          ninja -C build
+          ccache -s
+
+      - name: Test
+        run: |
+          cd build
+          ctest --no-compress-output --test-output-size-passed 100000 --test-output-size-failed 100000 -T Test -j "$(nproc)" --timeout 300 > /dev/null
+
+      - name: Upload test results to MinIO
+        if: always()
+        env:
+          MINIO_ACCESS_KEY: ${{ secrets.MINIO_ACCESS_KEY }}
+          MC_HOST_minio: https://${{ secrets.MINIO_ACCESS_KEY }}:${{ secrets.MINIO_SECRET_KEY }}@minio.weaselab.dev
+        run: |
+          if [ -z "$MINIO_ACCESS_KEY" ]; then
+            echo "MinIO credentials not configured; skipping upload"
+            exit 0
+          fi
+          zstd build/Testing/*/Test.xml
+          mc cp build/Testing/*/Test.xml.zst "minio/jenkins/conflict-set/${{ gitea.run_number }}/msan/"
+
+      - name: Test summary
+        if: always()
+        run: |
+          python3 ctest_summary.py build/Testing/*/Test.xml \
+            --link "https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/msan/Test.xml.zst" \
+            | tee -a "$GITHUB_STEP_SUMMARY"
+
+  coverage:
+    runs-on: ubuntu-latest-amd64
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/cache@v4
+        with:
+          path: /var/cache/apt/archives
+          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
+
+      - name: Install dependencies
+        run: |
+          . /etc/os-release
+          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
+          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
+          sudo apt-get update -qq
+          sudo apt-get install -y \
+            build-essential ccache clang-21 cmake gcovr \
+            libc6-dbg llvm-21 lld-21 mold ninja-build python3 valgrind zstd
+          sudo curl -Ls "https://minio.weaselab.dev/public/$(uname -m)/mc.RELEASE.2025-08-13T08-35-41Z" \
+            -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
+          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
+            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
+          done
+
       - uses: actions/cache@v4
         with:
           path: .ccache

Bench.cpp

@@ -37,15 +37,16 @@ ConflictSet::ReadRange singleton(Arena &arena, TrivialSpan key) {
 }
 
 ConflictSet::ReadRange prefixRange(Arena &arena, TrivialSpan key) {
-  int index;
+  int index = key.size() - 1;
-  for (index = key.size() - 1; index >= 0; index--)
+  for (; index >= 0; index--)
-    if ((key[index]) != 255)
+    if (key[index] != 255)
       break;
 
   // Must not be called with a string that consists only of zero or more '\xff'
-  // bytes.
+  // bytes, or with an empty string (which has no finite upper bound).
   if (index < 0) {
     assert(false);
+    std::abort();
   }
 
   uint8_t *buf = new (arena) uint8_t[index + 1];

CMakeLists.txt

@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.18)
 project(
   conflict-set
-  VERSION 0.0.14
+  VERSION 0.1.0
   DESCRIPTION
     "A data structure for optimistic concurrency control on ranges of bitwise-lexicographically-ordered keys."
   HOMEPAGE_URL "https://git.weaselab.dev/weaselab/conflict-set"
@@ -49,6 +49,18 @@ if(CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
         ${LLVM_OBJCOPY}
         CACHE FILEPATH "path to objcopy binary" FORCE)
   endif()
+  if(USE_MSAN)
+    find_program(LLD_LINKER lld)
+    if(LLD_LINKER)
+      set(CMAKE_LINKER_TYPE
+          "LLD"
+          CACHE STRING "Use LLD linker" FORCE)
+      set(CMAKE_LINKER
+          ${LLD_LINKER}
+          CACHE FILEPATH "path to linker binary" FORCE)
+      add_link_options("-fuse-ld=lld")
+    endif()
+  endif()
 endif()
 
 if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
@@ -103,6 +115,8 @@ option(USE_SIMD_FALLBACK
 
 option(DISABLE_TSAN "Disable TSAN" OFF)
 
+option(USE_MSAN "Build with MemorySanitizer (disables ASan/UBSan)" OFF)
+
 # This is encouraged according to
 # https://valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.clientreq
 include_directories(SYSTEM ${CMAKE_CURRENT_SOURCE_DIR}/third_party/valgrind)
@@ -139,6 +153,7 @@ add_custom_command(
   COMMAND_EXPAND_LISTS)
 
 add_library(${PROJECT_NAME} SHARED ${CMAKE_BINARY_DIR}/${PROJECT_NAME}.o)
+add_dependencies(${PROJECT_NAME} ${PROJECT_NAME}-object)
 set_target_properties(
   ${PROJECT_NAME} PROPERTIES LIBRARY_OUTPUT_DIRECTORY
                              "${CMAKE_CURRENT_BINARY_DIR}/radix_tree")
@@ -155,6 +170,7 @@ if(HAS_VERSION_SCRIPT)
 endif()
 
 add_library(${PROJECT_NAME}-static STATIC ${CMAKE_BINARY_DIR}/${PROJECT_NAME}.o)
+add_dependencies(${PROJECT_NAME}-static ${PROJECT_NAME}-object)
 if(CMAKE_BUILD_TYPE STREQUAL Debug)
   set_target_properties(${PROJECT_NAME}-static PROPERTIES LINKER_LANGUAGE CXX)
 else()
@@ -230,8 +246,10 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   target_compile_definitions(conflict_set_main PRIVATE ENABLE_MAIN)
   target_link_libraries(conflict_set_main PRIVATE nanobench)
 
-  if(NOT APPLE)
+  if(NOT APPLE AND NOT USE_MSAN)
-    # libfuzzer target, to generate/manage corpus
+    # libfuzzer target, to generate/manage corpus. MSan requires an instrumented
+    # libfuzzer runtime, which is not shipped with the compiler, so skip this
+    # target when building with MSan.
     set(FUZZ_FLAGS "-fsanitize=fuzzer-no-link,address,undefined")
     include(CheckCXXCompilerFlag)
     cmake_push_check_state()
@@ -255,8 +273,13 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   add_executable(fuzz_driver ConflictSet.cpp FuzzTestDriver.cpp)
   target_compile_options(fuzz_driver PRIVATE ${TEST_FLAGS})
   if(NOT CMAKE_CROSSCOMPILING)
-    target_compile_options(fuzz_driver PRIVATE -fsanitize=address,undefined)
+    if(USE_MSAN)
-    target_link_options(fuzz_driver PRIVATE -fsanitize=address,undefined)
+      target_compile_options(fuzz_driver PRIVATE -fsanitize=memory)
+      target_link_options(fuzz_driver PRIVATE -fsanitize=memory)
+    else()
+      target_compile_options(fuzz_driver PRIVATE -fsanitize=address,undefined)
+      target_link_options(fuzz_driver PRIVATE -fsanitize=address,undefined)
+    endif()
   endif()
   target_compile_definitions(fuzz_driver PRIVATE ENABLE_FUZZ)
   target_include_directories(fuzz_driver
@@ -267,7 +290,9 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   endforeach()
 
   # tsan tests
-  if(NOT CMAKE_CROSSCOMPILING AND NOT DISABLE_TSAN)
+  if(NOT CMAKE_CROSSCOMPILING
+     AND NOT DISABLE_TSAN
+     AND NOT USE_MSAN)
     add_executable(tsan_driver ConflictSet.cpp FuzzTestDriver.cpp)
     target_compile_options(tsan_driver PRIVATE ${TEST_FLAGS} -fsanitize=thread)
     target_link_options(tsan_driver PRIVATE -fsanitize=thread)
@@ -383,13 +408,25 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   if(NOT CMAKE_CROSSCOMPILING)
     find_program(HARDENING_CHECK hardening-check)
     if(HARDENING_CHECK)
+      # Not all versions of hardening-check support the same options, so query
+      # the help output before using architecture-specific skips.
+      execute_process(
+        COMMAND ${HARDENING_CHECK} --help
+        OUTPUT_VARIABLE hardening_check_help
+        ERROR_VARIABLE hardening_check_help
+        OUTPUT_STRIP_TRAILING_WHITESPACE ERROR_STRIP_TRAILING_WHITESPACE)
+      set(hardening_check_arch_flags "")
       # Control flow integrity (CET) is x86-only and branch protection (PAC/BTI)
       # is arm64-only, so ignore whichever doesn't apply.
       if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64 OR CMAKE_SYSTEM_PROCESSOR
                                                     STREQUAL arm64)
-        set(hardening_check_arch_flags --nocfprotection)
+        if(hardening_check_help MATCHES "nocfprotection")
+          list(APPEND hardening_check_arch_flags --nocfprotection)
+        endif()
       else()
-        set(hardening_check_arch_flags --nobranchprotection)
+        if(hardening_check_help MATCHES "nobranchprotection")
+          list(APPEND hardening_check_arch_flags --nobranchprotection)
+        endif()
       endif()
       add_test(
         NAME hardening_check
@@ -477,6 +514,11 @@ target_include_directories(
   PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>
          $<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}/${PROJECT_NAME}>)
 
+if(USE_MSAN)
+  target_compile_options(${PROJECT_NAME} PUBLIC -fsanitize=memory)
+  target_link_options(${PROJECT_NAME} PUBLIC -fsanitize=memory)
+endif()
+
 set_target_properties(
   ${PROJECT_NAME} PROPERTIES VERSION ${PROJECT_VERSION}
                              SOVERSION ${PROJECT_VERSION_MAJOR})

ConflictSet.cpp

@@ -3864,11 +3864,10 @@ PRESERVE_NONE void right_side_iter(Job *job, Context *context) {
 void Job::init(const ConflictSet::ReadRange *read, ConflictSet::Result *result,
                Node *root, int64_t oldestVersionFullPrecision) {
   auto begin = TrivialSpan(read->begin.p, read->begin.len);
-  auto end = TrivialSpan(read->end.p, read->end.len);
   if (read->readVersion < oldestVersionFullPrecision) [[unlikely]] {
     *result = ConflictSet::TooOld;
     continuation = complete;
-  } else if (end.size() == 0) {
+  } else if (read->end.len == 0) {
     this->begin = begin;
     this->n = root;
     this->readVersion = InternalVersionT(read->readVersion);
@@ -3876,7 +3875,7 @@ void Job::init(const ConflictSet::ReadRange *read, ConflictSet::Result *result,
     continuation = check::point_read_state_machine::begin;
   } else {
     this->begin = begin;
-    this->end = end;
+    this->end = TrivialSpan(read->end.p, read->end.len);
     this->n = root;
     this->readVersion = InternalVersionT(read->readVersion);
     this->result = result;
@@ -5046,8 +5045,8 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
       for (int i = 0; i < count; ++i) {
         const auto &w = writes[i];
         auto begin = TrivialSpan(w.begin.p, w.begin.len);
-        auto end = TrivialSpan(w.end.p, w.end.len);
         if (w.end.len > 0) {
+          auto end = TrivialSpan(w.end.p, w.end.len);
           addWriteRange(rootParent->children[0], begin, end,
                         InternalVersionT(writeVersion), &writeContext);
         } else {
@@ -5589,7 +5588,12 @@ ConflictSet::ConflictSet(ConflictSet &&other) noexcept
     : impl(std::exchange(other.impl, nullptr)) {}
 
 ConflictSet &ConflictSet::operator=(ConflictSet &&other) noexcept {
-  impl = std::exchange(other.impl, nullptr);
+  if (this != &other) {
+    if (impl) {
+      internal_destroy(impl);
+    }
+    impl = std::exchange(other.impl, nullptr);
+  }
   return *this;
 }
 
@@ -5674,13 +5678,13 @@ std::string getPartialKeyPrintable(Node *n) {
 }
 
 std::string strinc(std::string_view str, bool &ok) {
-  int index;
+  int index = static_cast<int>(str.size()) - 1;
-  for (index = str.size() - 1; index >= 0; index--)
+  for (; index >= 0; index--)
-    if ((uint8_t &)(str[index]) != 255)
+    if (static_cast<uint8_t>(str[index]) != 255)
       break;
 
   // Must not be called with a string that consists only of zero or more
-  // '\xff' bytes.
+  // '\xff' bytes, and the empty string has no successor.
   if (index < 0) {
     ok = false;
     return {};
@@ -5688,7 +5692,8 @@ std::string strinc(std::string_view str, bool &ok) {
   ok = true;
 
   auto r = std::string(str.substr(0, index + 1));
-  ((uint8_t &)r[r.size() - 1])++;
+  auto &last = r[r.size() - 1];
+  last = static_cast<char>(static_cast<uint8_t>(last) + 1);
   return r;
 }
 

Dockerfile

@@ -1,46 +0,0 @@
-FROM ubuntu:rolling
-
-WORKDIR /tmp
-ENV HOME=/tmp
-RUN chmod -R 777 /tmp
-
-# Install apt dependencies
-RUN apt-get update
-RUN apt-get upgrade -y
-RUN TZ=America/Los_Angeles DEBIAN_FRONTEND=noninteractive apt-get install -y \
-    biber \
-    build-essential \
-    ccache \
-    clang \
-    cmake \
-    curl \
-    devscripts \
-    gcovr \
-    git \
-    latexmk \
-    libc6-dbg \
-    llvm \
-    mold \
-    ninja-build \
-    nodejs \
-    pre-commit \
-    python3-requests \
-    rpm \
-    texlive-bibtex-extra \
-    texlive-fonts-recommended \
-    texlive-latex-extra \
-    texlive-pictures \
-    valgrind \
-    wget \
-    zstd
-
-# MinIO client, for uploading build artifacts
-RUN curl -Ls "https://dl.min.io/client/mc/release/linux-$(dpkg --print-architecture)/mc" \
-    -o /usr/local/bin/mc && chmod +x /usr/local/bin/mc
-
-ENV CC=clang
-ENV CXX=clang++
-
-# Try to have all the pre-commit hooks we'll need already initialized
-COPY .pre-commit-config.yaml /tmp/
-RUN git init && pre-commit install-hooks

HashTable.cpp

@@ -96,7 +96,9 @@ void ConflictSet::setOldestVersion(int64_t oldestVersion) {
   return impl->setOldestVersion(oldestVersion);
 }
 
-int64_t ConflictSet::getBytes() const { return -1; }
+// The hash_table implementation does not track memory usage, so return 0 to
+// satisfy the API contract that getBytes() returns a non-negative value.
+int64_t ConflictSet::getBytes() const { return 0; }
 
 void ConflictSet::getMetricsV1(MetricsV1 **metrics, int *count) const {
   *metrics = nullptr;
@@ -119,7 +121,13 @@ ConflictSet::ConflictSet(ConflictSet &&other) noexcept
     : impl(std::exchange(other.impl, nullptr)) {}
 
 ConflictSet &ConflictSet::operator=(ConflictSet &&other) noexcept {
-  impl = std::exchange(other.impl, nullptr);
+  if (this != &other) {
+    if (impl) {
+      impl->~Impl();
+      safe_free(impl, sizeof(Impl));
+    }
+    impl = std::exchange(other.impl, nullptr);
+  }
   return *this;
 }
 
@@ -155,7 +163,10 @@ __attribute__((__visibility__("default"))) void ConflictSet_destroy(void *cs) {
 }
 __attribute__((__visibility__("default"))) int64_t
 ConflictSet_getBytes(void *cs) {
-  using Impl = ConflictSet::Impl;
+  (void)cs;
-  return -1;
+  // The hash_table implementation does not track memory usage, so return 0 to
+  // satisfy the API contract that ConflictSet_getBytes returns a non-negative
+  // value.
+  return 0;
 }
 }

RealDataBench.cpp

@@ -1,5 +1,6 @@
 #include <ConflictSet.h>
 
+#include <algorithm>
 #include <cerrno>
 #include <chrono>
 #include <cstdio>
@@ -77,10 +78,10 @@ int main(int argc, const char **argv) {
       begin = end + 1;
       end = (uint8_t *)memchr(begin, '\n', size);
 
-      if (line.size() > 0 && line[0] == 'P') {
+      if (line.size() >= 2 && line[0] == 'P') {
-        write = line.subspan(2, line.size());
+        write = line.subspan(2, line.size() - 2);
-      } else if (line.size() > 0 && line[0] == 'L') {
+      } else if (line.size() >= 2 && line[0] == 'L') {
-        reads.push_back(line.subspan(2, line.size()));
+        reads.push_back(line.subspan(2, line.size() - 2));
       } else if (line.empty()) {
         {
           readRanges.resize(reads.size());
@@ -90,7 +91,7 @@ int main(int argc, const char **argv) {
             iter->begin.len = read.size();
             checkBytes += read.size();
             iter->end.len = 0;
-            iter->readVersion = version - 100;
+            iter->readVersion = std::max<int64_t>(0, version - 100);
             ++iter;
           }
         }
@@ -121,7 +122,7 @@ int main(int argc, const char **argv) {
         }
 
         timer = now();
-        cs.setOldestVersion(version - 10000);
+        cs.setOldestVersion(std::max<int64_t>(0, version - 10000));
         gcTime += now() - timer;
       }
     }

ServerBench.cpp

@@ -12,6 +12,7 @@
 #include <sys/ioctl.h>
 #include <sys/resource.h>
 #include <sys/socket.h>
+#include <sys/syscall.h>
 #include <sys/types.h>
 #include <sys/uio.h>
 #include <thread>

SkipList.cpp

@@ -38,6 +38,9 @@ std::span<const uint8_t> keyAfter(Arena &arena, std::span<const uint8_t> key) {
 
 std::span<const uint8_t> copyToArena(Arena &arena,
                                      std::span<const uint8_t> key) {
+  if (key.size() == 0) {
+    return std::span<const uint8_t>();
+  }
   auto result = std::span<uint8_t>(new (arena) uint8_t[key.size()], key.size());
   memcpy(result.data(), key.data(), key.size());
   return result;
@@ -97,7 +100,7 @@ force_inline bool getCharacter(const KeyInfo &ki, int character,
 
 bool operator<(const KeyInfo &lhs, const KeyInfo &rhs) {
   int i = std::min(lhs.key.size(), rhs.key.size());
-  int c = memcmp(lhs.key.data(), rhs.key.data(), i);
+  int c = i > 0 ? memcmp(lhs.key.data(), rhs.key.data(), i) : 0;
   if (c != 0)
     return c < 0;
 
@@ -287,7 +290,8 @@ private:
 
   static force_inline bool less(const uint8_t *a, int aLen, const uint8_t *b,
                                 int bLen) {
-    int c = memcmp(a, b, std::min(aLen, bLen));
+    int n = std::min(aLen, bLen);
+    int c = n > 0 ? memcmp(a, b, n) : 0;
     if (c < 0)
       return true;
     if (c > 0)
@@ -754,7 +758,8 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
 
   void addWrites(const ConflictSet::WriteRange *writes, int count,
                  int64_t writeVersion) {
-    auto points = std::vector<KeyInfo>(count * 2);
+    auto points = std::vector<KeyInfo>();
+    points.reserve(count * 2);
     Arena arena;
 
     for (int r = 0; r < count; r++) {
@@ -981,7 +986,12 @@ ConflictSet::ConflictSet(ConflictSet &&other) noexcept
     : impl(std::exchange(other.impl, nullptr)) {}
 
 ConflictSet &ConflictSet::operator=(ConflictSet &&other) noexcept {
-  impl = std::exchange(other.impl, nullptr);
+  if (this != &other) {
+    if (impl) {
+      internal_destroy(impl);
+    }
+    impl = std::exchange(other.impl, nullptr);
+  }
   return *this;
 }
 

build_msan_toolchain.sh

@@ -0,0 +1,43 @@
+#!/bin/bash
+
+set -euxo pipefail
+
+LLVM_VERSION="${LLVM_VERSION:-21}"
+MSAN_PREFIX="${MSAN_PREFIX:-$PWD/msan}"
+JOBS="${JOBS:-$(nproc)}"
+
+cd /tmp
+rm -rf libcxx-msan
+mkdir libcxx-msan
+cd libcxx-msan
+
+git clone --depth=1 "https://github.com/llvm/llvm-project.git" -b "release/${LLVM_VERSION}.x"
+
+cmake -S llvm-project/runtimes -B build \
+  -DCMAKE_BUILD_TYPE=Release \
+  -DCMAKE_C_COMPILER="clang-${LLVM_VERSION}" \
+  -DCMAKE_CXX_COMPILER="clang++-${LLVM_VERSION}" \
+  -DLLVM_USE_SANITIZER=MemoryWithOrigins \
+  -DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi;libunwind" \
+  -DLLVM_TARGETS_TO_BUILD=Native \
+  -DCMAKE_INSTALL_PREFIX="${MSAN_PREFIX}" \
+  -DLIBCXXABI_USE_LLVM_UNWINDER=ON \
+  -DLIBCXXABI_ENABLE_STATIC_UNWINDER=ON \
+  -DLIBCXX_USE_COMPILER_RT=ON \
+  -DLIBCXXABI_USE_COMPILER_RT=ON \
+  -DLIBUNWIND_USE_COMPILER_RT=ON \
+  -DLIBCXX_INCLUDE_TESTS=OFF \
+  -DLIBCXXABI_INCLUDE_TESTS=OFF \
+  -DLIBUNWIND_INCLUDE_TESTS=OFF \
+  -DLIBCXX_ENABLE_CLANG_TIDY=OFF
+
+cmake --build build -j"${JOBS}"
+cmake --install build
+
+VERSION="$(clang-"${LLVM_VERSION}" --version | head -n1 | sed -E 's/.*clang version ([0-9.]+).*/\1/')"
+TARBALL="msan-toolchain-${VERSION}.tar.zst"
+
+tar --zstd -cf "${TARBALL}" -C "${MSAN_PREFIX}" .
+
+TARBALL_PATH="$(pwd)/${TARBALL}"
+echo "Created: ${TARBALL_PATH}"

conflict_set.py

@@ -27,23 +27,37 @@ class Result(enum.Enum):
     TOO_OLD = 2
 
 
-def write(begin: bytes, end: Optional[bytes] = None) -> WriteRange:
+def _make_key(buf: bytes) -> tuple[_Key, bytearray]:
-    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
+    """Create a _Key and a backing bytearray that must be kept alive."""
+    backing = bytearray(buf)
+    array = (ctypes.c_ubyte * len(backing)).from_buffer(backing)
+    return _Key(array, len(array)), backing
 
+
+def write(begin: bytes, end: Optional[bytes] = None) -> WriteRange:
+    begin_key, begin_buf = _make_key(begin)
     if end is None:
-        e = (ctypes.c_ubyte * 0)()
+        end_key = _Key((ctypes.c_ubyte * 0)(), 0)
+        end_buf = None
     else:
-        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
+        end_key, end_buf = _make_key(end)
-    return WriteRange(_Key(b, len(b)), _Key(e, len(e)))
+    result = WriteRange(begin_key, end_key)
+    result._begin_buf = begin_buf
+    result._end_buf = end_buf
+    return result
 
 
 def read(version: int, begin: bytes, end: Optional[bytes] = None) -> ReadRange:
-    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
+    begin_key, begin_buf = _make_key(begin)
     if end is None:
-        e = (ctypes.c_ubyte * 0)()
+        end_key = _Key((ctypes.c_ubyte * 0)(), 0)
+        end_buf = None
     else:
-        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
+        end_key, end_buf = _make_key(end)
-    return ReadRange(_Key(b, len(b)), _Key(e, len(e)), version)
+    result = ReadRange(begin_key, end_key, version)
+    result._begin_buf = begin_buf
+    result._end_buf = end_buf
+    return result
 
 
 class ConflictSet:
@@ -88,6 +102,7 @@ class ConflictSet:
             ctypes.POINTER(ctypes.c_int),
             ctypes.c_int,
         )
+        self._lib.ConflictSet_check.restype = None
 
         self._lib.ConflictSet_addWrites.argtypes = (
             ctypes.c_void_p,
@@ -95,13 +110,16 @@ class ConflictSet:
             ctypes.c_int,
             ctypes.c_int64,
         )
+        self._lib.ConflictSet_addWrites.restype = None
 
         self._lib.ConflictSet_setOldestVersion.argtypes = (
             ctypes.c_void_p,
             ctypes.c_int64,
         )
+        self._lib.ConflictSet_setOldestVersion.restype = None
 
         self._lib.ConflictSet_destroy.argtypes = (ctypes.c_void_p,)
+        self._lib.ConflictSet_destroy.restype = None
 
         self._lib.ConflictSet_getBytes.argtypes = (ctypes.c_void_p,)
         self._lib.ConflictSet_getBytes.restype = ctypes.c_int64

include/ConflictSet.h

@@ -88,7 +88,8 @@ struct __attribute__((__visibility__("default"))) ConflictSet {
 
   ~ConflictSet();
 
-  /** Returns the total bytes in use by this ConflictSet */
+  /** Returns the total bytes in use by this ConflictSet. Implementations that
+   * do not track memory usage return 0. */
   int64_t getBytes() const;
 
   /** Experimental! */
@@ -132,6 +133,13 @@ struct __attribute__((__visibility__("default"))) ConflictSet {
 
 private:
   Impl *impl;
+#if __cplusplus <= 199711L
+  /* Declared private and left undefined to prevent copying in C++98/C++03.
+     The compiler would otherwise implicitly generate public copy operations,
+     which share the opaque Impl* and cause a double-free. */
+  ConflictSet(const ConflictSet &);
+  ConflictSet &operator=(const ConflictSet &);
+#endif
 };
 } /* namespace weaselab */
 
@@ -211,7 +219,8 @@ ConflictSet *ConflictSet_create(int64_t oldestVersion);
 
 void ConflictSet_destroy(ConflictSet *cs);
 
-/** Returns the total bytes in use by this ConflictSet */
+/** Returns the total bytes in use by this ConflictSet. Implementations that
+ * do not track memory usage return 0. */
 int64_t ConflictSet_getBytes(const ConflictSet *cs);
 
 #endif

test_conflict_set.py

@@ -57,6 +57,53 @@ def test_conflict_set():
         assert cs.check(read(0, key), read(1, key)) == [Result.TOO_OLD, Result.COMMIT]
 
 
+def test_hash_table_getBytes():
+    # Regression test for issue #62: the hash_table implementation is
+    # point-query only and does not track memory usage, but getBytes() must
+    # still return a non-negative value rather than -1.
+    with ConflictSet(0, build_dir=build_dir, implementation="hash_table") as cs:
+        assert cs.getBytes() == 0
+        cs.addWrites(1, write(b"key"))
+        assert cs.getBytes() >= 0
+        assert cs.check(read(0, b"key")) == [Result.CONFLICT]
+
+
+def test_write_read_without_outer_reference():
+    # Regression test for issue #42: WriteRange/ReadRange must keep their
+    # backing key buffers alive, because the C library reads the pointer
+    # stored in _Key while addWrites/check run.
+    with DebugConflictSet() as cs:
+        # The bytes literal is not referenced after this expression.
+        cs.addWrites(1, write(b"key"))
+        assert cs.check(read(0, b"key")) == [Result.CONFLICT]
+
+        cs.addWrites(2, write(b"a", b"z"))
+        assert cs.check(read(1, b"a", b"z")) == [Result.CONFLICT]
+        assert cs.check(read(1, b"b")) == [Result.CONFLICT]
+        assert cs.check(read(1, b"0")) == [Result.COMMIT]
+
+
+def test_range_keeps_key_buffers_alive():
+    # Verify the fix for issue #42: returned range objects must retain a
+    # reference to the backing bytearray so the C pointer stays valid after
+    # the helper returns.
+    w = write(b"key")
+    assert w._begin_buf == bytearray(b"key")
+    assert w._end_buf is None
+
+    w2 = write(b"a", b"z")
+    assert w2._begin_buf == bytearray(b"a")
+    assert w2._end_buf == bytearray(b"z")
+
+    r = read(0, b"key")
+    assert r._begin_buf == bytearray(b"key")
+    assert r._end_buf is None
+
+    r2 = read(1, b"a", b"z")
+    assert r2._begin_buf == bytearray(b"a")
+    assert r2._end_buf == bytearray(b"z")
+
+
 def test_update_zero_should_commit():
     with DebugConflictSet() as cs1:
         with DebugConflictSet() as cs2:
@@ -68,7 +115,7 @@ def test_update_zero_should_commit():
             for i in range(256 - 17, 256):
                 cs2.addWrites(int(1), write(bytes([i])))
             # Scan until first point write
-            cs2.check(read(0, b"\x00", bytes([256 - 17])))
+            assert cs2.check(read(0, b"\x00", bytes([256 - 17]))) == [Result.COMMIT]
 
 
 def test_update_zero_should_conflict():
@@ -81,7 +128,7 @@ def test_update_zero_should_conflict():
             # "zero" is now 2**31 + 100
             cs1.addWrites(2**32 + 101, write(b"", b"\x02"), write(b"\x01"))
             # rangeVersion of \x01 is now 2**31 + 100 ("max" of (2**31 + 100, 2**32 + 101))
-            cs1.check(read(2**32 + 1, b"\x00"))
+            assert cs1.check(read(2**32 + 1, b"\x00")) == [Result.CONFLICT]
             # but 2**32 + 1 ">" 2**31 + 100 , and it incorrectly commits