Add address sanitizer and undefined behavior sanitizer

.github/workflows/android.yaml

@@ -137,7 +137,7 @@ jobs:
             git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/sherpa-onnx-libs main
 
       - name: Release android libs
-        if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa' && github.event_name == 'push' && contains(github.ref, 'refs/tags/')
+        if: (github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa') && github.event_name == 'push' && contains(github.ref, 'refs/tags/')
         uses: svenstaro/upload-release-action@v2
         with:
           file_glob: true

.github/workflows/flutter-linux.yaml

@@ -217,3 +217,11 @@ jobs:
         with:
           name: flutter-sherpa-onnx-linux-x64
           path: ./*.tar.bz2
+
+      # - name: Release android libs
+      #   if: (github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa') && github.event_name == 'push' && contains(github.ref, 'refs/tags/')
+      #   uses: svenstaro/upload-release-action@v2
+      #   with:
+      #     file_glob: true
+      #     overwrite: true
+      #     file: flutter*.tar.bz2

.github/workflows/flutter-macos.yaml

@@ -180,3 +180,11 @@ jobs:
         with:
           name: flutter-sherpa-onnx-app-macos-${{ matrix.arch }}
           path: ./*.tar.bz2
+
+      - name: Release android libs
+        if: (github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa') && github.event_name == 'push' && contains(github.ref, 'refs/tags/')
+        uses: svenstaro/upload-release-action@v2
+        with:
+          file_glob: true
+          overwrite: true
+          file: flutter*.tar.bz2

.github/workflows/flutter-windows-x64.yaml

@@ -157,3 +157,11 @@ jobs:
         with:
           name: flutter-sherpa-onnx-windows-x64
           path: ./*.tar.bz2
+
+      - name: Release android libs
+        if: (github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa') && github.event_name == 'push' && contains(github.ref, 'refs/tags/')
+        uses: svenstaro/upload-release-action@v2
+        with:
+          file_glob: true
+          overwrite: true
+          file: flutter*.tar.bz2

.github/workflows/sanitizer.yaml

@@ -0,0 +1,186 @@
+name: sanitizer
+
+on:
+  workflow_dispatch:
+
+  schedule:
+    # minute (0-59)
+    # hour (0-23)
+    # day of the month (1-31)
+    # month (1-12)
+    # day of the week (0-6)
+    # nightly build at 22:50 UTC time every day
+    - cron: "50 22 * * *"
+
+concurrency:
+  group: sanitizer-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  sanitizer:
+    runs-on: ${{ matrix.os }}
+    name: sanitizer
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [macos-latest]
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: ccache
+        uses: hendrikmuhs/ccache-action@v1.2
+        with:
+          key: ${{ matrix.os }}-sanitizer
+
+      - name: Configure CMake
+        shell: bash
+        run: |
+          export CMAKE_CXX_COMPILER_LAUNCHER=ccache
+          export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
+          cmake --version
+
+          mkdir build
+          cd build
+
+          cmake \
+            -DSHERPA_ONNX_ENABLE_PYTHON=ON \
+            -DSHERPA_ONNX_ENABLE_TESTS=ON \
+            -DSHERPA_ONNX_ENABLE_JNI=ON \
+            -DSHERPA_ONNX_ENABLE_SANITIZER=ON \
+            -D BUILD_SHARED_LIBS=ON \
+            -D CMAKE_BUILD_TYPE=Release \
+            -DCMAKE_INSTALL_PREFIX=./install \
+            ..
+
+      - name: Build sherpa-onnx
+        shell: bash
+        run: |
+          export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
+
+          cd build
+          make -j2
+          make install
+
+          ls -lh lib
+          ls -lh bin
+
+          file ./bin/sherpa-onnx
+
+      - name: Display dependencies of sherpa-onnx for macos
+        shell: bash
+        run: |
+          file bin/sherpa-onnx
+          otool -L build/bin/sherpa-onnx
+          otool -l build/bin/sherpa-onnx
+
+      - name: Test offline transducer
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx-offline
+
+          .github/scripts/test-offline-transducer.sh
+
+      - name: Test online CTC
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx
+
+          .github/scripts/test-online-ctc.sh
+
+      - name: Test offline punctuation
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx-offline-punctuation
+
+          .github/scripts/test-offline-punctuation.sh
+
+      - name: Test C API
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export SLID_EXE=spoken-language-identification-c-api
+          export SID_EXE=speaker-identification-c-api
+          export AT_EXE=audio-tagging-c-api
+          export PUNCT_EXE=add-punctuation-c-api
+
+          .github/scripts/test-c-api.sh
+
+      - name: Test Audio tagging
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx-offline-audio-tagging
+
+          .github/scripts/test-audio-tagging.sh
+
+      - name: Test spoken language identification (C++ API)
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx-offline-language-identification
+
+          .github/scripts/test-spoken-language-identification.sh
+
+      - name: Test transducer kws
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx-keyword-spotter
+
+          .github/scripts/test-kws.sh
+
+      - name: Test offline TTS
+        if: matrix.with_tts == 'ON'
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx-offline-tts
+
+          .github/scripts/test-offline-tts.sh
+
+      - name: Test online paraformer
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx
+
+          .github/scripts/test-online-paraformer.sh
+
+      - name: Test offline Whisper
+        if: matrix.build_type != 'Debug'
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx-offline
+
+          .github/scripts/test-offline-whisper.sh
+
+      - name: Test offline CTC
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx-offline
+
+          .github/scripts/test-offline-ctc.sh
+
+      - name: Test online transducer
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=sherpa-onnx
+
+          .github/scripts/test-online-transducer.sh
+
+      - name: Test online transducer (C API)
+        shell: bash
+        run: |
+          export PATH=$PWD/build/bin:$PATH
+          export EXE=decode-file-c-api
+
+          .github/scripts/test-online-transducer.sh

CMakeLists.txt

@@ -2,6 +2,8 @@ cmake_minimum_required(VERSION 3.13 FATAL_ERROR)
 
 set(CMAKE_OSX_DEPLOYMENT_TARGET "10.14" CACHE STRING "Minimum OS X deployment version. Used only for macOS")
 
+
+
 project(sherpa-onnx)
 
 set(SHERPA_ONNX_VERSION "1.9.26")
@@ -32,6 +34,7 @@ option(SHERPA_ONNX_ENABLE_BINARY "Whether to build binaries" ON)
 option(SHERPA_ONNX_ENABLE_TTS "Whether to build TTS related code" ON)
 option(SHERPA_ONNX_LINK_LIBSTDCPP_STATICALLY "True to link libstdc++ statically. Used only when BUILD_SHARED_LIBS is OFF on Linux" ON)
 option(SHERPA_ONNX_USE_PRE_INSTALLED_ONNXRUNTIME_IF_AVAILABLE "True to use pre-installed onnxruntime if available" ON)
+option(SHERPA_ONNX_ENABLE_SANITIZER "Whether to enable ubsan and asan" OFF)
 
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/lib")
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/lib")
@@ -121,6 +124,7 @@ message(STATUS "SHERPA_ONNX_ENABLE_BINARY ${SHERPA_ONNX_ENABLE_BINARY}")
 message(STATUS "SHERPA_ONNX_ENABLE_TTS ${SHERPA_ONNX_ENABLE_TTS}")
 message(STATUS "SHERPA_ONNX_LINK_LIBSTDCPP_STATICALLY ${SHERPA_ONNX_LINK_LIBSTDCPP_STATICALLY}")
 message(STATUS "SHERPA_ONNX_USE_PRE_INSTALLED_ONNXRUNTIME_IF_AVAILABLE ${SHERPA_ONNX_USE_PRE_INSTALLED_ONNXRUNTIME_IF_AVAILABLE}")
+message(STATUS "SHERPA_ONNX_ENABLE_SANITIZER: ${SHERPA_ONNX_ENABLE_SANITIZER}")
 
 if(SHERPA_ONNX_ENABLE_TTS)
   message(STATUS "TTS is enabled")
@@ -267,6 +271,33 @@ if(SHERPA_ONNX_ENABLE_TTS)
   include(cppjieba) # For Chinese TTS. It is a header-only C++ library
 endif()
 
+# if(NOT MSVC AND CMAKE_BUILD_TYPE STREQUAL Debug AND (CMAKE_CXX_COMPILER_ID STREQUAL "Clang" OR CMAKE_CXX_COMPILER_ID STREQUAL "AppleClang"))
+if(SHERPA_ONNX_ENABLE_SANITIZER)
+  message(WARNING "enable ubsan and asan")
+  set(CMAKE_REQUIRED_LIBRARIES -lubsan -lasan)
+  include(CheckCCompilerFlag)
+
+  set(flags -fsanitize=undefined )
+  string(APPEND flags " -fno-sanitize-recover=undefined ")
+  string(APPEND flags " -fsanitize=integer ")
+  string(APPEND flags " -fsanitize=nullability ")
+  string(APPEND flags " -fsanitize=implicit-conversion ")
+  string(APPEND flags " -fsanitize=bounds ")
+  string(APPEND flags " -fsanitize=address ")
+
+  if(OFF)
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${flags} -Wall -Wextra")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${flags} -Wall -Wextra")
+  else()
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${flags}")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${flags}")
+  endif()
+
+  set(CMAKE_EXECUTBLE_LINKER_FLAGS "${CMAKE_EXECUTBLE_LINKER_FLAGS} ${flags}")
+
+  add_compile_options(-fno-omit-frame-pointer)
+endif()
+
 add_subdirectory(sherpa-onnx)
 
 if(SHERPA_ONNX_ENABLE_C_API AND SHERPA_ONNX_ENABLE_BINARY)

c-api-examples/add-punctuation-c-api.c

@@ -43,8 +43,8 @@ int32_t main() {
   const char *texts[] = {
       "这是一个测试你好吗How are you我很好thank you are you ok谢谢你",
       "我们都是木头人不会说话不会动",
-      "The African blogosphere is rapidly expanding bringing more voices "
-      "online in the form of commentaries opinions analyses rants and poetry",
+      ("The African blogosphere is rapidly expanding bringing more voices "
+       "online in the form of commentaries opinions analyses rants and poetry"),
   };
 
   int32_t n = sizeof(texts) / sizeof(const char *);

c-api-examples/decode-file-c-api.c

@@ -179,8 +179,6 @@ int32_t main(int32_t argc, char *argv[]) {
 
 #define N 3200  // 0.2 s. Sample rate is fixed to 16 kHz
 
-  int16_t buffer[N];
-  float samples[N];
   fprintf(stderr, "sample rate: %d, num samples: %d, duration: %.2f s\n",
           wave->sample_rate, wave->num_samples,
           (float)wave->num_samples / wave->sample_rate);

c-api-examples/streaming-hlg-decode-file-c-api.c

@@ -65,8 +65,6 @@ int32_t main() {
 // simulate streaming. You can choose an arbitrary N
 #define N 3200
 
-  int16_t buffer[N];
-  float samples[N];
   fprintf(stderr, "sample rate: %d, num samples: %d, duration: %.2f s\n",
           wave->sample_rate, wave->num_samples,
           (float)wave->num_samples / wave->sample_rate);

cmake/espeak-ng-for-piper.cmake

@@ -14,10 +14,7 @@ function(download_espeak_ng_for_piper)
   set(USE_SPEECHPLAYER OFF CACHE BOOL "" FORCE)
   set(EXTRA_cmn ON CACHE BOOL "" FORCE)
   set(EXTRA_ru ON CACHE BOOL "" FORCE)
-
-  if(SHERPA_ONNX_ENABLE_WASM)
-    set(BUILD_ESPEAK_NG_EXE OFF CACHE BOOL "" FORCE)
-  endif()
+  set(BUILD_ESPEAK_NG_EXE OFF CACHE BOOL "" FORCE)
 
   # If you don't have access to the Internet,
   # please pre-download kaldi-decoder

sherpa-onnx/c-api/c-api.cc

@@ -187,7 +187,7 @@ const SherpaOnnxOnlineRecognizerResult *GetOnlineStreamResult(
   r->text = pText;
 
   // copy json
-  const auto &json = result.AsJsonString();
+  std::string json = result.AsJsonString();
   char *pJson = new char[json.size() + 1];
   std::copy(json.begin(), json.end(), pJson);
   pJson[json.size()] = 0;
@@ -445,7 +445,7 @@ const SherpaOnnxOfflineRecognizerResult *GetOfflineStreamResult(
   r->text = pText;
 
   // copy json
-  const auto &json = result.AsJsonString();
+  std::string json = result.AsJsonString();
   char *pJson = new char[json.size() + 1];
   std::copy(json.begin(), json.end(), pJson);
   pJson[json.size()] = 0;
@@ -643,7 +643,7 @@ const SherpaOnnxKeywordResult *GetKeywordResult(
   r->keyword = pKeyword;
 
   // copy json
-  const auto &json = result.AsJsonString();
+  std::string json = result.AsJsonString();
   char *pJson = new char[json.size() + 1];
   std::copy(json.begin(), json.end(), pJson);
   pJson[json.size()] = 0;

sherpa-onnx/csrc/audio-tagging-ced-impl.h

@@ -73,7 +73,7 @@ class AudioTaggingCEDImpl : public AudioTaggingImpl {
     std::vector<float> f = s->GetFrames();
 
     int32_t num_frames = f.size() / feat_dim;
-    assert(feat_dim * num_frames == f.size());
+    assert(feat_dim * num_frames == static_cast<int32_t>(f.size()));
 
     std::array<int64_t, 3> shape = {1, num_frames, feat_dim};
 

sherpa-onnx/csrc/audio-tagging-label-file.cc

@@ -65,7 +65,7 @@ void AudioTaggingLabels::Init(std::istream &is) {
       exit(-1);
     }
 
-    if (i != names_.size()) {
+    if (i != static_cast<int32_t>(names_.size())) {
       SHERPA_ONNX_LOGE(
           "Index should be sorted and contiguous. Expected index: %d, given: "
           "%d.",

sherpa-onnx/csrc/audio-tagging-zipformer-impl.h

@@ -74,7 +74,7 @@ class AudioTaggingZipformerImpl : public AudioTaggingImpl {
 
     int32_t num_frames = f.size() / feat_dim;
 
-    assert(feat_dim * num_frames == f.size());
+    assert(feat_dim * num_frames == static_cast<int32_t>(f.size()));
 
     std::array<int64_t, 3> shape = {1, num_frames, feat_dim};
 

sherpa-onnx/csrc/cat.cc

@@ -92,7 +92,7 @@ Ort::Value Cat(OrtAllocator *allocator,
     }
   }
 
-  return std::move(ans);
+  return ans;
 }
 
 template Ort::Value Cat<float>(OrtAllocator *allocator,