Add new coverage to corpus

This branch is going to be 0.1.x now

.gitea/workflows/ci.yml

@@ -0,0 +1,306 @@
+name: CI
+
+on: [push, pull_request]
+
+env:
+  CC: clang
+  CXX: clang++
+
+jobs:
+  pre-commit:
+    runs-on: ubuntu-latest-amd64
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/cache@v4
+        with:
+          path: /var/cache/apt/archives
+          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
+
+      - name: Install dependencies
+        run: |
+          sudo apt-get update -qq
+          sudo apt-get install -y git nodejs pre-commit
+
+      - uses: actions/cache@v4
+        with:
+          path: ~/.cache/pre-commit
+          key: pre-commit-${{ hashFiles('.pre-commit-config.yaml') }}
+
+      - name: Run pre-commit
+        run: |
+          git config --global --add safe.directory "$PWD"
+          pre-commit run --all-files --show-diff-on-failure
+
+  test:
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - name: 64-bit-versions
+            cmake_args: -DCMAKE_CXX_FLAGS=-DUSE_64_BIT=1
+          - name: debug
+            cmake_args: -DCMAKE_BUILD_TYPE=Debug
+          - name: simd-fallback
+            cmake_args: -DUSE_SIMD_FALLBACK=ON
+          - name: gcc
+            cmake_args: -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++
+    runs-on: ubuntu-latest-amd64
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/cache@v4
+        with:
+          path: /var/cache/apt/archives
+          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
+
+      - name: Install 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: |
+          export CCACHE_DIR="$GITHUB_WORKSPACE/.ccache"
+          rm -rf build
+          cmake -S . -B build -G Ninja -DCMAKE_CXX_COMPILER_LAUNCHER=ccache ${{ matrix.cmake_args }}
+          ninja -C build
+          ccache -s
+
+      - name: Test
+        run: |
+          cd build
+          ctest --no-compress-output --test-output-size-passed 100000 --test-output-size-failed 100000 -T Test -j "$(nproc)" --timeout 90 > /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/${{ gitea.run_number }}/${{ matrix.name }}/"
+
+      - name: Test summary
+        if: always()
+        run: |
+          python3 ctest_summary.py build/Testing/*/Test.xml \
+            --link "https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/${{ matrix.name }}/Test.xml.zst" \
+            | tee -a "$GITHUB_STEP_SUMMARY"
+
+  release:
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - runner: ubuntu-latest-amd64
+            arch: amd64
+          - runner: ubuntu-latest-arm64
+            arch: arm64
+    runs-on: ${{ matrix.runner }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/cache@v4
+        with:
+          path: /var/cache/apt/archives
+          key: apt-${{ 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
+          key: ccache-release-${{ matrix.arch }}-${{ gitea.sha }}
+          restore-keys: |
+            ccache-release-${{ matrix.arch }}-
+
+      - name: Build
+        run: |
+          export CCACHE_DIR="$GITHUB_WORKSPACE/.ccache"
+          rm -rf build
+          cmake -S . -B build -G Ninja -DCMAKE_CXX_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_FLAGS=-DNVALGRIND
+          ninja -C build
+          ccache -s
+
+      - name: Test
+        run: |
+          cd build
+          # On arm64, valgrind needs the MAKE_MEM_DEFINED client requests for
+          # https://git.weaselab.dev/weaselab/conflict-set/issues/39, but this
+          # build has -DNVALGRIND, which compiles them out. Skip valgrind
+          # tests here; they run annotated in the test job.
+          ctest --no-compress-output --test-output-size-passed 100000 --test-output-size-failed 100000 ${{ matrix.arch == 'arm64' && '-E valgrind' || '' }} -T Test -j "$(nproc)" --timeout 90 > /dev/null
+
+      - name: Package
+        run: |
+          cd build
+          cpack -G DEB
+          cpack -G RPM
+
+      - name: Build paper
+        if: matrix.arch == 'amd64'
+        run: |
+          cd paper
+          make
+
+      - name: Upload artifacts to MinIO
+        if: always()
+        shell: bash
+        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
+          dest="minio/jenkins/conflict-set/${{ gitea.run_number }}/release-${{ matrix.arch }}/"
+          zstd build/Testing/*/Test.xml
+          mc cp build/Testing/*/Test.xml.zst "$dest"
+          if compgen -G "build/*.deb" > /dev/null; then
+            mc cp build/*.deb "$dest"
+          fi
+          if compgen -G "build/*.rpm" > /dev/null; then
+            mc cp build/*.rpm "$dest"
+          fi
+          if compgen -G "paper/*.pdf" > /dev/null; then
+            mc cp paper/*.pdf "$dest"
+          fi
+
+      - name: Test summary
+        if: always()
+        run: |
+          python3 ctest_summary.py build/Testing/*/Test.xml \
+            --link "https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/release-${{ matrix.arch }}/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
+          key: ccache-coverage-${{ gitea.sha }}
+          restore-keys: |
+            ccache-coverage-
+
+      - name: Build
+        run: |
+          export CCACHE_DIR="$GITHUB_WORKSPACE/.ccache"
+          rm -rf build
+          cmake -S . -B build -G Ninja -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+            -DCMAKE_C_FLAGS=--coverage -DCMAKE_CXX_FLAGS=--coverage \
+            -DCMAKE_BUILD_TYPE=Debug -DDISABLE_TSAN=ON
+          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 "$(nproc)" --timeout 90 > /dev/null
+
+      - name: Coverage report
+        shell: bash
+        run: |
+          gcov_args=(-f ConflictSet.cpp -f LongestCommonPrefix.h -f Metrics.h
+                     --gcov-executable "llvm-cov gcov" --exclude-noncode-lines)
+          gcovr "${gcov_args[@]}" --cobertura > build/coverage.xml
+          gcovr "${gcov_args[@]}"
+          mkdir -p build/coverage_html
+          gcovr "${gcov_args[@]}" --html-details build/coverage_html/index.html
+          gcovr "${gcov_args[@]}" --fail-under-line 100 > /dev/null
+
+      - name: Upload 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
+          dest="minio/jenkins/conflict-set/${{ gitea.run_number }}/coverage/"
+          zstd build/Testing/*/Test.xml
+          mc cp build/Testing/*/Test.xml.zst "$dest"
+          if [ -e build/coverage.xml ]; then
+            mc cp build/coverage.xml "$dest"
+          fi
+          if [ -d build/coverage_html ]; then
+            mc cp -r build/coverage_html "$dest"
+          fi
+
+      - name: Test summary
+        if: always()
+        run: |
+          python3 ctest_summary.py build/Testing/*/Test.xml \
+            --link "https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/coverage/Test.xml.zst" \
+            | tee -a "$GITHUB_STEP_SUMMARY"
+          echo "" | tee -a "$GITHUB_STEP_SUMMARY"
+          echo "📊 [Coverage report](https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/coverage/coverage_html/index.html)" | tee -a "$GITHUB_STEP_SUMMARY"

Bench.cpp

@@ -37,15 +37,16 @@ ConflictSet::ReadRange singleton(Arena &arena, TrivialSpan key) {
 }
 
 ConflictSet::ReadRange prefixRange(Arena &arena, TrivialSpan key) {
-  int index;
-  for (index = key.size() - 1; index >= 0; index--)
-    if ((key[index]) != 255)
+  int index = key.size() - 1;
+  for (; index >= 0; index--)
+    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"
@@ -53,6 +53,7 @@ endif()
 
 if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
   add_compile_options("-Wno-maybe-uninitialized")
+  add_compile_options("-Wno-maybe-musttail-local-addr")
 endif()
 
 if(NOT APPLE)
@@ -60,12 +61,16 @@ if(NOT APPLE)
   add_compile_options(-g -fno-omit-frame-pointer)
 endif()
 
-set(full_relro_flags "-pie;LINKER:-z,relro,-z,now,-z,noexecstack")
+set(relro_flags "LINKER:-z,relro,-z,now,-z,noexecstack")
+set(full_relro_flags "-pie;${relro_flags}")
 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})
+  # -pie only applies to executables; passing it when linking a shared library
+  # makes the driver pull in Scrt1.o, which requires main.
+  add_link_options("$<$<STREQUAL:$<TARGET_PROPERTY:TYPE>,EXECUTABLE>:-pie>"
+                   ${relro_flags})
 endif()
 cmake_pop_check_state()
 
@@ -134,6 +139,7 @@ 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")
@@ -150,6 +156,7 @@ 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()
@@ -378,9 +385,30 @@ 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)
-      add_test(NAME hardening_check
-               COMMAND ${HARDENING_CHECK} $<TARGET_FILE:${PROJECT_NAME}>
-                       --nofortify --nostackprotector)
+      # 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)
+        if(hardening_check_help MATCHES "nocfprotection")
+          list(APPEND hardening_check_arch_flags --nocfprotection)
+        endif()
+      else()
+        if(hardening_check_help MATCHES "nobranchprotection")
+          list(APPEND hardening_check_arch_flags --nobranchprotection)
+        endif()
+      endif()
+      add_test(
+        NAME hardening_check
+        COMMAND ${HARDENING_CHECK} $<TARGET_FILE:${PROJECT_NAME}> --nofortify
+                --nostackprotector ${hardening_check_arch_flags})
     endif()
   endif()
 

ConflictSet.cpp

@@ -52,6 +52,14 @@ limitations under the License.
 #endif
 #endif
 
+#ifndef __SANITIZE_ADDRESS__
+#if defined(__has_feature)
+#if __has_feature(address_sanitizer)
+#define __SANITIZE_ADDRESS__
+#endif
+#endif
+#endif
+
 #include <memcheck.h>
 
 using namespace weaselab;
@@ -2246,6 +2254,13 @@ bool checkMaxBetweenExclusiveImpl(Node3 *n, int begin, int end,
     mask |= inBounds(self->index[i]) << i;
   }
   mask &= (1 << self->numChildren) - 1;
+#ifdef __aarch64__
+  // The bits surviving the mask above don't derive from uninitialized slots,
+  // but clang 21+ on aarch64 lowers inBounds through flags+csel, which
+  // memcheck models imprecisely, tainting bits the mask provably clears.
+  // https://git.weaselab.dev/weaselab/conflict-set/issues/39
+  VALGRIND_MAKE_MEM_DEFINED(&mask, sizeof(mask));
+#endif
   if (!mask) {
     return true;
   }
@@ -2257,7 +2272,13 @@ bool checkMaxBetweenExclusiveImpl(Node3 *n, int begin, int end,
     compared |= (self->childMaxVersion[i] > readVersion) << i;
   }
 
-  return !(compared & mask) && firstRangeOk;
+  uint32_t compared_masked = compared & mask;
+#ifdef __aarch64__
+  // Same imprecise csel modeling as above.
+  // https://git.weaselab.dev/weaselab/conflict-set/issues/39
+  VALGRIND_MAKE_MEM_DEFINED(&compared_masked, sizeof(compared_masked));
+#endif
+  return !compared_masked && firstRangeOk;
 }
 
 template <bool kAVX512>
@@ -3048,7 +3069,12 @@ Node *firstGeqPhysical(Node *n, const TrivialSpan key) {
 #define MUSTTAIL
 #endif
 
-#if __has_attribute(preserve_none)
+// ASan + preserve_none miscompiles the continuation chains on aarch64 with
+// every clang tested (20, 21, trunk 22). The chains work with the default
+// calling convention, so use that under ASan on aarch64.
+// https://git.weaselab.dev/weaselab/conflict-set/issues/38
+#if __has_attribute(preserve_none) &&                                          \
+    !(defined(__aarch64__) && defined(__SANITIZE_ADDRESS__))
 #define PRESERVE_NONE __attribute__((preserve_none))
 #else
 #define PRESERVE_NONE
@@ -4984,7 +5010,7 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
     // encounter invalidated nodes.
     for (auto *iter = firstRangeWrite; iter != nullptr;
          iter = iter->nextRangeWrite) {
-      if (iter->endInsertionPoint != nullptr) {
+      assert(iter->endInsertionPoint != nullptr);
       while (iter->insertionPoint->releaseDeferred) {
         iter->insertionPoint = iter->insertionPoint->forwardTo;
       }
@@ -4994,7 +5020,6 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
       eraseInRange(iter->insertionPoint, iter->endInsertionPoint,
                    &writeContext);
     }
-    }
 
     if (count > kStackResultMax) [[unlikely]] {
       safe_free(context.results, count * sizeof(interleaved_insert::Result));
@@ -5564,7 +5589,12 @@ ConflictSet::ConflictSet(ConflictSet &&other) noexcept
     : impl(std::exchange(other.impl, nullptr)) {}
 
 ConflictSet &ConflictSet::operator=(ConflictSet &&other) noexcept {
+  if (this != &other) {
+    if (impl) {
+      internal_destroy(impl);
+    }
     impl = std::exchange(other.impl, nullptr);
+  }
   return *this;
 }
 
@@ -5649,13 +5679,13 @@ std::string getPartialKeyPrintable(Node *n) {
 }
 
 std::string strinc(std::string_view str, bool &ok) {
-  int index;
-  for (index = str.size() - 1; index >= 0; index--)
-    if ((uint8_t &)(str[index]) != 255)
+  int index = static_cast<int>(str.size()) - 1;
+  for (; index >= 0; index--)
+    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 {};
@@ -5663,7 +5693,8 @@ 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

@@ -1,58 +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 \
-    binutils-aarch64-linux-gnu \
-    build-essential \
-    ccache \
-    cmake \
-    curl \
-    devscripts \
-    g++-aarch64-linux-gnu \
-    gcovr \
-    git \
-    gnupg \
-    libc6-dbg \
-    lsb-release \
-    mold \
-    ninja-build \
-    pre-commit \
-    python3-requests \
-    qemu-user \
-    rpm \
-    software-properties-common \
-    texlive-full \
-    wget \
-    zstd
-
-# Install recent valgrind from source
-RUN curl -Ls https://sourceware.org/pub/valgrind/valgrind-3.22.0.tar.bz2 -o valgrind.tar.bz2 && \
-    echo "c811db5add2c5f729944caf47c4e7a65dcaabb9461e472b578765dd7bf6d2d4c  valgrind.tar.bz2" > valgrind-sha.txt && \
-    sha256sum --quiet -c valgrind-sha.txt && \
-    mkdir valgrind && \
-    tar --strip-components 1 --no-same-owner --no-same-permissions --directory valgrind -xjf valgrind.tar.bz2 && \
-    cd valgrind && \
-    ./configure --enable-only64bit --enable-lto && \
-    make -j`nproc` && \
-    make install && \
-    cd .. && \
-    rm -rf /tmp/*
-
-# Recent clang
-RUN wget https://apt.llvm.org/llvm.sh && chmod +x ./llvm.sh && ./llvm.sh 20
-
-RUN apt-get -y install clang llvm
-
-# Set after building valgrind, which doesn't build with clang for some reason
-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 {
+  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;
-  return -1;
+  (void)cs;
+  // 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;
 }
 }

Jenkinsfile

@@ -1,151 +0,0 @@
-def CleanBuildAndTest(String cmakeArgs) {
-    sh """
-    export CCACHE_DIR=/ccache
-    rm -rf build
-    mkdir build
-    cd build
-    cmake .. -G Ninja -DCMAKE_CXX_COMPILER_LAUNCHER=ccache ${cmakeArgs} 
-    ninja
-    ccache -s
-    """
-    catchError {
-        sh '''
-        cd build
-        ctest --no-compress-output --test-output-size-passed 100000 --test-output-size-failed 100000 -T Test -j `nproc` --timeout 90 > /dev/null
-        zstd Testing/*/Test.xml
-        '''
-    }
-    xunit tools: [CTest(pattern: 'build/Testing/*/Test.xml')], skipPublishingChecks: false
-    minio bucket: 'jenkins', credentialsId: 'jenkins-minio', excludes: '', host: 'minio.weaselab.dev', includes: 'build/Testing/*/Test.xml.zst', targetFolder: '${JOB_NAME}/${BUILD_NUMBER}/${STAGE_NAME}/'
-}
-
-pipeline {
-    agent any
-    stages {
-        stage('Pre-commit') {
-            agent {
-                dockerfile {
-                    args '-v /home/jenkins/ccache:/ccache'
-                    reuseNode true
-                }
-            }
-            steps {
-                script {
-                    env.HOME = env.WORKSPACE
-                }
-                sh 'pre-commit run --all-files --show-diff-on-failure'
-            }
-        }
-        stage('64 bit versions') {
-            agent {
-                dockerfile {
-                    args '-v /home/jenkins/ccache:/ccache'
-                    reuseNode true
-                }
-            }
-            steps {
-                CleanBuildAndTest("-DCMAKE_CXX_FLAGS=-DUSE_64_BIT=1")
-            }
-        }
-        stage('Debug') {
-            agent {
-                dockerfile {
-                    args '-v /home/jenkins/ccache:/ccache'
-                    reuseNode true
-                }
-            }
-            steps {
-                CleanBuildAndTest("-DCMAKE_BUILD_TYPE=Debug")
-            }
-        }
-        stage('SIMD fallback') {
-            agent {
-                dockerfile {
-                    args '-v /home/jenkins/ccache:/ccache'
-                    reuseNode true
-                }
-            }
-            steps {
-                CleanBuildAndTest("-DUSE_SIMD_FALLBACK=ON")
-            }
-        }
-        stage('Release [clang]') {
-            agent {
-                dockerfile {
-                    args '-v /home/jenkins/ccache:/ccache'
-                    reuseNode true
-                }
-            }
-            steps {
-                CleanBuildAndTest("-DCMAKE_CXX_FLAGS=-DNVALGRIND")
-                recordIssues(tools: [clang()])
-                sh '''
-                cd build
-                cpack -G DEB
-                cpack -G RPM
-                '''
-                sh '''
-                cd paper
-                make
-                '''
-                minio bucket: 'jenkins', credentialsId: 'jenkins-minio', excludes: '', host: 'minio.weaselab.dev', includes: 'build/*.deb,build/*.rpm,paper/*.pdf', targetFolder: '${JOB_NAME}/${BUILD_NUMBER}/${STAGE_NAME}/'
-            }
-        }
-        stage('gcc') {
-            agent {
-                dockerfile {
-                    args '-v /home/jenkins/ccache:/ccache'
-                    reuseNode true
-                }
-            }
-            steps {
-                CleanBuildAndTest("-DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++")
-                recordIssues(tools: [gcc()])
-            }
-        }
-        stage('Release [clang,aarch64]') {
-            agent {
-                dockerfile {
-                    args '-v /home/jenkins/ccache:/ccache'
-                    reuseNode true
-                }
-            }
-            steps {
-                CleanBuildAndTest("-DCMAKE_TOOLCHAIN_FILE=../aarch64-toolchain.cmake -DCMAKE_CXX_FLAGS=-DNVALGRIND")
-                sh '''
-                cd build
-                cpack -G DEB
-                cpack -G RPM
-                '''
-                minio bucket: 'jenkins', credentialsId: 'jenkins-minio', excludes: '', host: 'minio.weaselab.dev', includes: 'build/*.deb,build/*.rpm', targetFolder: '${JOB_NAME}/${BUILD_NUMBER}/${STAGE_NAME}'
-            }
-        }
-        stage('Coverage') {
-            agent {
-                dockerfile {
-                    args '-v /home/jenkins/ccache:/ccache'
-                    reuseNode true
-                }
-            }
-            steps {
-                script {
-                    gcov_args = "-f ConflictSet.cpp -f LongestCommonPrefix.h -f Metrics.h --gcov-executable 'llvm-cov gcov' --exclude-noncode-lines"
-                }
-                CleanBuildAndTest("-DCMAKE_C_FLAGS=--coverage -DCMAKE_CXX_FLAGS=--coverage -DCMAKE_BUILD_TYPE=Debug -DDISABLE_TSAN=ON")
-                sh """
-                gcovr ${gcov_args} --cobertura > build/coverage.xml
-                """
-                recordCoverage qualityGates: [[criticality: 'NOTE', metric: 'MODULE']], tools: [[parser: 'COBERTURA', pattern: 'build/coverage.xml']]
-                sh """
-                gcovr ${gcov_args}
-                gcovr ${gcov_args} --fail-under-line 100 > /dev/null
-                """
-            }
-        }
-    }
-    post {
-        always {
-            emailext mimeType: 'text/html', body: '${SCRIPT, template="groovy-html.template"}', subject: "${env.JOB_NAME} - Build# ${env.BUILD_NUMBER} - ${currentBuild.currentResult}", to: 'andrew@weaselab.dev'
-        }
-    }
-}

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') {
-        write = line.subspan(2, line.size());
-      } else if (line.size() > 0 && line[0] == 'L') {
-        reads.push_back(line.subspan(2, line.size()));
+      if (line.size() >= 2 && line[0] == 'P') {
+        write = line.subspan(2, line.size() - 2);
+      } else if (line.size() >= 2 && line[0] == 'L') {
+        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 {
+  if (this != &other) {
+    if (impl) {
+      internal_destroy(impl);
+    }
     impl = std::exchange(other.impl, nullptr);
+  }
   return *this;
 }
 

conflict_set.py

@@ -27,23 +27,37 @@ class Result(enum.Enum):
     TOO_OLD = 2
 
 
-def write(begin: bytes, end: Optional[bytes] = None) -> WriteRange:
-    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
+def _make_key(buf: bytes) -> tuple[_Key, bytearray]:
+    """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))
-    return WriteRange(_Key(b, len(b)), _Key(e, len(e)))
+        end_key, end_buf = _make_key(end)
+    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))
-    return ReadRange(_Key(b, len(b)), _Key(e, len(e)), version)
+        end_key, end_buf = _make_key(end)
+    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

ctest_summary.py

@@ -0,0 +1,76 @@
+#!/usr/bin/env python3
+"""Summarize a CTest Test.xml as markdown.
+
+Intended for $GITHUB_STEP_SUMMARY in CI, where only the first few failures
+are shown inline (pass --link to point at the full Test.xml). Also reusable
+locally to print every failure from a downloaded Test.xml with --all.
+"""
+
+import argparse
+import base64
+import gzip
+import xml.etree.ElementTree as ET
+
+# Failure output is truncated to this many trailing characters, which is
+# usually enough to include e.g. an ASan report's summary.
+OUTPUT_TAIL_CHARS = 3000
+
+
+def test_output(test):
+    value = test.find("./Results/Measurement/Value")
+    if value is None or value.text is None:
+        return ""
+    text = value.text
+    if value.get("encoding") == "base64":
+        raw = base64.b64decode(text)
+        if value.get("compression") == "gzip":
+            raw = gzip.decompress(raw)
+        text = raw.decode(errors="replace")
+    return text
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("test_xml", help="path to a ctest Testing/*/Test.xml")
+    parser.add_argument(
+        "--inline",
+        type=int,
+        default=5,
+        help="how many failures to show inline (default 5)",
+    )
+    parser.add_argument("--all", action="store_true", help="show every failure inline")
+    parser.add_argument(
+        "--link", help="URL of the full Test.xml, linked when failures are elided"
+    )
+    args = parser.parse_args()
+
+    testing = ET.parse(args.test_xml).getroot().find("Testing")
+    tests = testing.findall("Test")
+    failed = [t for t in tests if t.get("Status") == "failed"]
+    notrun = sum(1 for t in tests if t.get("Status") == "notrun")
+
+    if not failed:
+        print(f"✅ All {len(tests) - notrun} tests passed")
+    else:
+        print(f"❌ {len(failed)} of {len(tests)} tests failed\n")
+        shown = failed if args.all else failed[: args.inline]
+        for test in shown:
+            name = test.findtext("Name")
+            output = test_output(test)[-OUTPUT_TAIL_CHARS:].strip()
+            print(f"<details><summary><code>{name}</code></summary>\n")
+            print("````")
+            print(output)
+            print("````")
+            print("</details>\n")
+        remaining = len(failed) - len(shown)
+        if remaining > 0:
+            more = f"… and {remaining} more"
+            if args.link:
+                more += f" — full list in [Test.xml]({args.link})"
+            print(more)
+    if notrun:
+        print(f"\n⚠️ {notrun} tests not run")
+
+
+if __name__ == "__main__":
+    main()

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