Add split buffer facility to support QNX PMEM and other buffering systems

.bazelignore

@@ -1 +1,2 @@
 bazel-ipc
+build

.bazelrc

@@ -9,6 +9,13 @@ build:asan --copt -g
 build:asan --copt -fno-omit-frame-pointer
 build:asan --linkopt -fsanitize=address
 
+build:tsan --strip=never
+build:tsan --copt -fsanitize=thread
+build:tsan --copt -DTHREAD_SANITIZER
+build:tsan --copt -g
+build:tsan --copt -fno-omit-frame-pointer
+build:tsan --linkopt -fsanitize=thread
+
 # Enable the built-in platform-specific configuration feature
 common --enable_platform_specific_config
 

.cursor/rules/test-runner.mdc

@@ -0,0 +1,11 @@
+---
+description: Use Bazelisk for repository test commands
+alwaysApply: true
+---
+
+# Test Runner
+
+Use `bazelisk` rather than `bazel` for this repository.
+
+- Run the full test suite with `bazelisk test //...`.
+- Prefer `bazelisk test <target>` for focused Bazel test runs.

.github/workflows/ci.yml

@@ -2,18 +2,36 @@ name: CI
 
 on: [push, pull_request]
 
+permissions:
+  contents: read
+
 jobs:
   test:
+    name: test (${{ matrix.os }}, ${{ matrix.mode }})
     runs-on: ${{ matrix.os }}
     timeout-minutes: 60 # liberal upper bound to protect against tests stalling
     strategy:
       fail-fast: false
       matrix:
         include:
           - os: ubuntu-latest
+            mode: fastbuild
+            bazel_flags:
+          - os: ubuntu-latest
+            mode: opt
+            bazel_flags: -c opt
+          - os: ubuntu-24.04-arm
+            mode: fastbuild
+            bazel_flags:
           - os: ubuntu-24.04-arm
+            mode: opt
+            bazel_flags: -c opt
           - os: macos-latest
+            mode: fastbuild
             bazel_flags: --config=macos_arm64
+          - os: macos-latest
+            mode: opt
+            bazel_flags: --config=macos_arm64 -c opt
 
     steps:
       - name: Checkout code
@@ -25,7 +43,7 @@ jobs:
           # Avoid downloading Bazel every time.
           bazelisk-cache: true
           # Store build cache per workflow.
-          disk-cache: ${{ github.workflow }}-${{ matrix.os }}
+          disk-cache: ${{ github.workflow }}-${{ matrix.os }}-${{ matrix.mode }}
           # Share repository cache between workflows.
           repository-cache: true
 
@@ -37,21 +55,174 @@ jobs:
 
       - name: Run tests
         run: |
-          bazel test //... \
+          bazel test \
+            //... \
+            --verbose_failures \
+            --test_output=errors \
+            ${{ matrix.bazel_flags }}
+
+      - name: Run split-buffer client tests
+        run: |
+          bazel test \
+            //client:client_test \
+            //c_client:client_test \
             --verbose_failures \
             --test_output=errors \
+            --test_arg=--use_split_buffers \
             ${{ matrix.bazel_flags }}
 
       - name: Upload Bazel test logs
         uses: actions/upload-artifact@v7
         if: failure() # Upload artifacts only if a step failed
         with:
-          name: bazel-test-logs-${{ matrix.os }}
+          name: bazel-test-logs-${{ matrix.os }}-${{ matrix.mode }}
           path: bazel-testlogs
 
+  diagnostics:
+    name: diagnostics (${{ matrix.os }}, ${{ matrix.mode }})
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 60
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: ubuntu-latest
+            mode: asan
+            experimental: false
+            bazel_flags: --config=asan
+            run_under:
+          - os: macos-latest
+            mode: asan
+            experimental: false
+            bazel_flags: --config=macos_arm64 --config=asan
+            run_under:
+          - os: ubuntu-latest
+            mode: tsan
+            experimental: true
+            bazel_flags: --config=tsan
+            run_under:
+          - os: macos-latest
+            mode: tsan
+            experimental: true
+            bazel_flags: --config=macos_arm64 --config=tsan
+            run_under:
+          - os: ubuntu-latest
+            mode: valgrind
+            experimental: false
+            bazel_flags:
+            run_under: --run_under='valgrind --tool=memcheck --leak-check=full --show-leak-kinds=definite,indirect --errors-for-leak-kinds=definite,indirect --error-exitcode=1'
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v6
+
+      - name: Install Bazel
+        uses: bazel-contrib/setup-bazel@0.18.0
+        with:
+          bazelisk-cache: true
+          disk-cache: ${{ github.workflow }}-${{ matrix.os }}-${{ matrix.mode }}
+          repository-cache: true
+
+      - name: Install Valgrind
+        if: matrix.mode == 'valgrind'
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y valgrind
+
+      - name: Run ${{ matrix.mode }} C client test
+        run: |
+          set +e
+          bazel test \
+            //c_client:client_test \
+            --verbose_failures \
+            --test_output=errors \
+            ${{ matrix.bazel_flags }} \
+            ${{ matrix.run_under }}
+          status=$?
+          if [ "$status" -ne 0 ] && [ "${{ matrix.experimental }}" = "true" ]; then
+            echo "::warning::Experimental ${{ matrix.mode }} diagnostics failed with exit code $status"
+            exit 0
+          fi
+          exit "$status"
+
+      - name: Run ${{ matrix.mode }} split-buffer client tests
+        run: |
+          set +e
+          bazel test \
+            //client:client_test \
+            //c_client:client_test \
+            --verbose_failures \
+            --test_output=errors \
+            --test_arg=--use_split_buffers \
+            ${{ matrix.bazel_flags }} \
+            ${{ matrix.run_under }}
+          status=$?
+          if [ "$status" -ne 0 ] && [ "${{ matrix.experimental }}" = "true" ]; then
+            echo "::warning::Experimental ${{ matrix.mode }} diagnostics failed with exit code $status"
+            exit 0
+          fi
+          exit "$status"
+
+      - name: Upload diagnostic Bazel test logs
+        uses: actions/upload-artifact@v7
+        if: ${{ failure() || matrix.experimental }}
+        with:
+          name: bazel-diagnostic-test-logs-${{ matrix.os }}-${{ matrix.mode }}
+          path: bazel-testlogs
+
+  latency:
+    name: latency (${{ matrix.platform }}, opt)
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 60
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: ubuntu-latest
+            platform: linux
+            bazel_flags: -c opt
+          - os: macos-latest
+            platform: macos
+            bazel_flags: --config=macos_arm64 -c opt
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v6
+
+      - name: Install Bazel
+        uses: bazel-contrib/setup-bazel@0.18.0
+        with:
+          bazelisk-cache: true
+          disk-cache: ${{ github.workflow }}-${{ matrix.os }}-latency-opt
+          repository-cache: true
+
+      - name: Run optimized latency tests
+        run: |
+          bazel test \
+            //client:latency_test \
+            --test_filter=LatencyTest.PublisherLatencyHistogram:LatencyTest.SubscriberLatency:LatencyTest.PubSubLatency \
+            --verbose_failures \
+            --test_output=all \
+            --test_env=SUBSPACE_LATENCY_OS=${{ matrix.platform }} \
+            --test_env=SUBSPACE_LATENCY_MODE=opt \
+            --test_env=SUBSPACE_LATENCY_REVISION=current \
+            ${{ matrix.bazel_flags }}
+
+      - name: Upload latency Bazel test logs
+        uses: actions/upload-artifact@v7
+        if: failure()
+        with:
+          name: bazel-latency-test-logs-${{ matrix.platform }}
+          path: bazel-testlogs/client/latency_test
+
   cmake:
+    name: cmake (${{ matrix.build_type }})
     runs-on: ubuntu-latest
     timeout-minutes: 60
+    strategy:
+      fail-fast: false
+      matrix:
+        build_type: [Debug, Release]
 
     steps:
       - name: Checkout code
@@ -63,10 +234,10 @@ jobs:
           sudo apt-get install -y protobuf-compiler
 
       - name: Configure CMake
-        run: cmake -S . -B build/cmake -DCMAKE_BUILD_TYPE=Debug
+        run: cmake -S . -B build/cmake-${{ matrix.build_type }} -DCMAKE_BUILD_TYPE=${{ matrix.build_type }}
 
       - name: Build CMake targets
-        run: cmake --build build/cmake --parallel $(nproc)
+        run: cmake --build build/cmake-${{ matrix.build_type }} --parallel $(nproc)
 
       - name: Run CMake tests
-        run: ctest --test-dir build/cmake --output-on-failure
+        run: ctest --test-dir build/cmake-${{ matrix.build_type }} --output-on-failure

CMakeLists.txt

@@ -144,6 +144,14 @@ add_subdirectory(shadow)
 add_subdirectory(c_client)
 add_subdirectory(rust_client)
 
+if(TARGET client_test AND TARGET nop_plugin)
+    add_dependencies(client_test nop_plugin)
+endif()
+
+if(TARGET client_test AND TARGET split_buffer_free_test_plugin)
+    add_dependencies(client_test split_buffer_free_test_plugin)
+endif()
+
 # --- Python bindings and tests ---
 if(SUBSPACE_PYTHON)
     add_subdirectory(client/python)

MODULE.bazel

@@ -1,6 +1,6 @@
 module(
     name = "subspace",
-    version = "2.6.3",
+    version = "2.7.0",
 )
 
 bazel_dep(name = "bazel_skylib", version = "1.9.0")

asio_rpc/client/rpc_client.cc

@@ -351,10 +351,6 @@ RpcClient::InvokeMethod(int method_id, const google::protobuf::Any &request,
 
   absl::StatusOr<void *> buffer;
   for (;;) {
-    if (method->request_publisher->NumSubscribers() == 0) {
-      Destroy();
-      return absl::InternalError("No subscribers; is the RPC server running?");
-    }
     buffer = method->request_publisher->GetMessageBuffer(
         int32_t(req.ByteSizeLong()));
     if (!buffer.ok()) {
@@ -451,10 +447,6 @@ absl::Status RpcClient::InvokeMethod(
 
   absl::StatusOr<void *> buffer;
   for (;;) {
-    if (method->request_publisher->NumSubscribers() == 0) {
-      Destroy();
-      return absl::InternalError("No subscribers; is the RPC server running?");
-    }
     buffer = method->request_publisher->GetMessageBuffer(
         int32_t(req.ByteSizeLong()));
     if (!buffer.ok()) {

c_client/BUILD.bazel

@@ -30,7 +30,7 @@ cc_test(
         "@abseil-cpp//absl/hash:hash_testing",
         "@abseil-cpp//absl/status",
         "@abseil-cpp//absl/status:statusor",
-        "@googletest//:gtest_main",
+        "@googletest//:gtest",
         "@coroutines//:co",
     ],
 )

c_client/CMakeLists.txt

@@ -21,4 +21,5 @@ target_link_libraries(subspace_c_client PUBLIC
 
 add_executable(c_client_test client_test.cc)
 add_test(NAME c_client_test COMMAND c_client_test)
-target_link_libraries(c_client_test PUBLIC subspace_c_client libserver gtest gtest_main)
+add_test(NAME c_client_split_buffer_test COMMAND c_client_test --use_split_buffers)
+target_link_libraries(c_client_test PUBLIC subspace_c_client libserver gtest absl::flags_parse)