Set version to 2.5.0-b1

Using std::barrier performs extremely poorly (~1 hour vs ~1 minute to run the test suite) in certain macOS environments.
To unblock our macOS CI pipeline, std::barrier has been replaced with a custom mutex-based barrier (Barrier) that significantly improves performance without compromising correctness.

.github/workflows/macos.yml

@@ -71,6 +71,9 @@ jobs:
           nproc --version
           echo -n "nproc returns: "
           nproc
+          system_profiler SPHardwareDataType
+          sysctl -n hw.logicalcpu
+          clang --version
       - name: configure Conan
         run : |
           conan profile new default --detect || true
@@ -89,9 +92,8 @@ jobs:
           generator: ${{ matrix.generator }}
           configuration: ${{ matrix.configuration }}
           cmake-args: "-Dassert=TRUE -Dwerr=TRUE ${{ matrix.cmake-args }}"
-      # TODO: Temporary disabled tests
+      - name: test
-      # - name: test
+        run: |
-      #   run: |
+          n=$(nproc)
-      #     n=$(nproc)
+          echo "Using $n test jobs"
-      #     echo "Using $n test jobs"
+          ${build_dir}/rippled --unittest --unittest-jobs $n
-      #     ${build_dir}/rippled --unittest --unittest-jobs $n

.github/workflows/nix.yml

@@ -247,7 +247,7 @@ jobs:
           mkdir -p ~/.conan
           tar -xzf conan.tar -C ~/.conan
       - name: install gcovr
-        run: pip install "gcovr>=7,<8"
+        run: pip install "gcovr>=7,<9"
       - name: check environment
         run: |
           echo ${PATH} | tr ':' '\n'

cfg/validators-example.txt

@@ -26,7 +26,7 @@
 #
 #   Examples:
 #    https://vl.ripple.com
-#    https://vl.xrplf.org
+#    https://unl.xrplf.org
 #    http://127.0.0.1:8000
 #    file:///etc/opt/ripple/vl.txt
 #

cmake/CodeCoverage.cmake

@@ -98,6 +98,9 @@
 # 2024-04-03, Bronek Kozicki
 #     - add support for output formats: jacoco, clover, lcov
 #
+# 2025-05-12, Jingchen Wu
+#     - add -fprofile-update=atomic to ensure atomic profile generation
+#
 # USAGE:
 #
 # 1. Copy this file into your cmake modules path.
@@ -200,15 +203,27 @@ set(COVERAGE_COMPILER_FLAGS "-g --coverage"
     CACHE INTERNAL "")
 if(CMAKE_CXX_COMPILER_ID MATCHES "(GNU|Clang)")
     include(CheckCXXCompilerFlag)
+    include(CheckCCompilerFlag)
+
     check_cxx_compiler_flag(-fprofile-abs-path HAVE_cxx_fprofile_abs_path)
     if(HAVE_cxx_fprofile_abs_path)
         set(COVERAGE_CXX_COMPILER_FLAGS "${COVERAGE_COMPILER_FLAGS} -fprofile-abs-path")
     endif()
-    include(CheckCCompilerFlag)
+
     check_c_compiler_flag(-fprofile-abs-path HAVE_c_fprofile_abs_path)
     if(HAVE_c_fprofile_abs_path)
         set(COVERAGE_C_COMPILER_FLAGS "${COVERAGE_COMPILER_FLAGS} -fprofile-abs-path")
     endif()
+
+    check_cxx_compiler_flag(-fprofile-update HAVE_cxx_fprofile_update)
+    if(HAVE_cxx_fprofile_update)
+        set(COVERAGE_CXX_COMPILER_FLAGS "${COVERAGE_COMPILER_FLAGS}  -fprofile-update=atomic")
+    endif()
+
+    check_c_compiler_flag(-fprofile-update HAVE_c_fprofile_update)
+    if(HAVE_c_fprofile_update)
+        set(COVERAGE_C_COMPILER_FLAGS "${COVERAGE_COMPILER_FLAGS} -fprofile-update=atomic")
+    endif()
 endif()
 
 set(CMAKE_Fortran_FLAGS_COVERAGE

src/libxrpl/protocol/BuildInfo.cpp

@@ -36,7 +36,7 @@ namespace BuildInfo {
 //  and follow the format described at http://semver.org/
 //------------------------------------------------------------------------------
 // clang-format off
-char const* const versionString = "2.4.0"
+char const* const versionString = "2.5.0-b1"
 // clang-format on
 
 #if defined(DEBUG) || defined(SANITIZER)

src/test/basics/IntrusiveShared_test.cpp

@@ -9,6 +9,7 @@
 #include <atomic>
 #include <barrier>
 #include <chrono>
+#include <condition_variable>
 #include <latch>
 #include <optional>
 #include <random>
@@ -19,6 +20,55 @@
 namespace ripple {
 namespace tests {
 
+/**
+Experimentally, we discovered that using std::barrier performs extremely
+poorly (~1 hour vs ~1 minute to run the test suite) in certain macOS
+environments. To unblock our macOS CI pipeline, we replaced std::barrier with a
+custom mutex-based barrier (Barrier) that significantly improves performance
+without compromising correctness. For future reference, if we ever consider
+reintroducing std::barrier, the following configuration is known to exhibit the
+problem:
+
+    Model Name: Mac mini
+    Model Identifier: Mac14,3
+    Model Number: Z16K000R4LL/A
+    Chip: Apple M2
+    Total Number of Cores: 8 (4 performance and 4 efficiency)
+    Memory: 24 GB
+    System Firmware Version: 11881.41.5
+    OS Loader Version: 11881.1.1
+    Apple clang version 16.0.0 (clang-1600.0.26.3)
+    Target: arm64-apple-darwin24.0.0
+    Thread model: posix
+
+ */
+struct Barrier
+{
+    std::mutex mtx;
+    std::condition_variable cv;
+    int count;
+    int const initial;
+
+    Barrier(int n) : count(n), initial(n)
+    {
+    }
+
+    void
+    arrive_and_wait()
+    {
+        std::unique_lock lock(mtx);
+        if (--count == 0)
+        {
+            count = initial;
+            cv.notify_all();
+        }
+        else
+        {
+            cv.wait(lock, [&] { return count == initial; });
+        }
+    }
+};
+
 namespace {
 enum class TrackedState : std::uint8_t {
     uninitialized,
@@ -500,9 +550,9 @@ public:
         constexpr int loopIters = 2 * 1024;
         constexpr int numThreads = 16;
         std::vector<SharedIntrusive<TIBase>> toClone;
-        std::barrier loopStartSyncPoint{numThreads};
+        Barrier loopStartSyncPoint{numThreads};
-        std::barrier postCreateToCloneSyncPoint{numThreads};
+        Barrier postCreateToCloneSyncPoint{numThreads};
-        std::barrier postCreateVecOfPointersSyncPoint{numThreads};
+        Barrier postCreateVecOfPointersSyncPoint{numThreads};
         auto engines = [&]() -> std::vector<std::default_random_engine> {
             std::random_device rd;
             std::vector<std::default_random_engine> result;
@@ -628,10 +678,10 @@ public:
         constexpr int flipPointersLoopIters = 256;
         constexpr int numThreads = 16;
         std::vector<SharedIntrusive<TIBase>> toClone;
-        std::barrier loopStartSyncPoint{numThreads};
+        Barrier loopStartSyncPoint{numThreads};
-        std::barrier postCreateToCloneSyncPoint{numThreads};
+        Barrier postCreateToCloneSyncPoint{numThreads};
-        std::barrier postCreateVecOfPointersSyncPoint{numThreads};
+        Barrier postCreateVecOfPointersSyncPoint{numThreads};
-        std::barrier postFlipPointersLoopSyncPoint{numThreads};
+        Barrier postFlipPointersLoopSyncPoint{numThreads};
         auto engines = [&]() -> std::vector<std::default_random_engine> {
             std::random_device rd;
             std::vector<std::default_random_engine> result;
@@ -761,9 +811,9 @@ public:
         constexpr int lockWeakLoopIters = 256;
         constexpr int numThreads = 16;
         std::vector<SharedIntrusive<TIBase>> toLock;
-        std::barrier loopStartSyncPoint{numThreads};
+        Barrier loopStartSyncPoint{numThreads};
-        std::barrier postCreateToLockSyncPoint{numThreads};
+        Barrier postCreateToLockSyncPoint{numThreads};
-        std::barrier postLockWeakLoopSyncPoint{numThreads};
+        Barrier postLockWeakLoopSyncPoint{numThreads};
 
         // lockAndDestroy creates weak pointers from the strong pointer
         // and runs a loop that locks the weak pointer. At the end of the loop