debug

build.sh

@@ -1,17 +1,3 @@
-g++ -shared -fPIC \
-    -I/usr/include/python3.10 -I/usr/include/python3.10/numpy -I./include \
-    -L./lib -L/usr/lib/x86_64-linux-gnu \
-    -DRTC_NUMPY_IMPL \
-    rtc_plugins.cpp util/RTCContext.cpp \
-    -lMRTCEngine -lboost_python310 -lboost_numpy310 -lpython3.10 \
-    -Wl,-rpath='$ORIGIN/lib' \
-    -o rtc_plugins.so
-#g++ -shared -fPIC \
-#    -I/usr/include/python3.10 -I/usr/include/python3.10/numpy -I./include \
-#    -I/usr/include/python3.10 -I/usr/include/numpy \
-#    -Wl,-rpath='$ORIGIN/lib' \
-#    -lboost_python310 -lpython3.10 \
-#    -lMRTCEngine -lboost_python310 -lpython3.10 \
-#    -L$(python3 -c "import numpy; print(numpy.get_include())") \
-#    rtc_plugins.cpp util/RTCContext.cpp \
-#    -o rtc_plugins.so
+#g++ -shared -fPIC -std=c++17 $(python3 -m pybind11 --includes) -I./include -L./lib -L/usr/lib/x86_64-linux-gnu -lMRTCEngine -lpython3.10 -Wl,-rpath='$ORIGIN/lib' -o rtc_plugins$(python3-config --extension-suffix) rtc_plugins.cpp util/RTCContext.cpp
+
+g++ -shared -fPIC -std=c++17 $(python3 -m pybind11 --includes) -I./include -L./lib -L/usr/lib/x86_64-linux-gnu -Wl,--no-as-needed -lMRTCEngine -Wl,--as-needed -lpython3.10 -Wl,-rpath='$ORIGIN/lib' -o rtc_plugins$(python3-config --extension-suffix) rtc_plugins.cpp util/RTCContext.cpp

rtc_plugins.cpp

@@ -1,20 +1,11 @@
-// rtc_plugins.cpp
+#include "util/RTCContext.h"
+#include <pybind11/pybind11.h>
+#include <pybind11/numpy.h>  // pybind11 的 NumPy 支持
 
-#define IMPLEMENT_NUMPY_API  // 标记这是实现文件
-#include "util/numpyStub.h"
+namespace py = pybind11;
 #include "util/RTCContext.h"
 
-// 提供转换接口
-void** get_numpy_api() {
-    return (void**)RTC_PLUGINS_ARRAY_API;
-}
-
-
-namespace py = boost::python;
-
-int init(const char* selfUserId, const char* selfDisplayName, const char* selfRoomId, boost::python::object callback) {
-
-    RTCContext::instance().setPyCallback(callback);
+int init(const char* selfUserId, const char* selfDisplayName, const char* selfRoomId) {
     bool res = RTCContext::instance().init(selfUserId, selfDisplayName, selfRoomId);
     if (res) {
         return 0;
@@ -23,7 +14,6 @@ int init(const char* selfUserId, const char* selfDisplayName, const char* selfRo
     }
 }
 int initRecv(const char* destRoomId, const char* srcUserId, const int16_t destChannelIndex) {
-
     bool res = RTCContext::instance().initRecv(destRoomId, srcUserId, destChannelIndex);
     if (res) {
         return 0;
@@ -31,7 +21,7 @@ int initRecv(const char* destRoomId, const char* srcUserId, const int16_t destCh
         return -1;
     }
 }
-int initSend(const char* srcRoomId, const char* destRoomId, const int16_t destChannelIndex, const uint8_t channelNum) {
+int initSend(const char* srcRoomId, const char* destRoomId, const int16_t destChannelIndex, const int16_t channelNum) {
     bool res = RTCContext::instance().initSend(srcRoomId, destRoomId, destChannelIndex, channelNum);
     if (res) {
         return 0;
@@ -39,92 +29,84 @@ int initSend(const char* srcRoomId, const char* destRoomId, const int16_t destCh
         return -1;
     }
 }
-
-py::object create_int16_array() {
-    // 1. 定义数组维度（1维，长度为 4）
-    npy_intp dims[1] = {4};
-
-    // 2. 创建原生 C 数组（int16_t 数据）
-    int16_t data[4] = {1, 2, -3, 4};  // 示例数据
-
-    // 3. 通过 NumPy C API 创建 PyObject*
-    PyObject* py_array = PyArray_SimpleNewFromData(
-            1,              // 维度数
-            dims,           // 各维度大小
-            NPY_INT16,      // 数据类型（np.int16）
-            data            // 数据指针
+// NumPy 数据交互（关键修改）
+py::array_t<int16_t> getNumpyData() {
+    return py::array_t<int16_t>(
+            RTCContext::instance().getNumpyData()  // 假设返回的是已有 NumPy 数组
     );
-
-    if (!py_array) {
-        throw std::runtime_error("Failed to create NumPy array");
-    }
-
-    // 4. 转换为 py::object（自动管理引用计数）
-    return py::object(py::handle<>(py_array));
 }
 
-int sendCustomAudioData(int16_t destChannelIndex, py::object pD,
+int sendCustomAudioData(int16_t destChannelIndex, py::array_t<int16_t> inputArray,
                         int32_t sampleRate, uint64_t channelNum, uint64_t dataLen) {
-    try {
-        // 强制转换为 int16 连续数组
-        PyObject* py_array = PyArray_FROM_OTF(
-                pD.ptr(),
-                NPY_INT16,
-                NPY_ARRAY_IN_ARRAY | NPY_ARRAY_FORCECAST
-        );
-        if (!py_array) {
-            throw std::runtime_error("Failed to convert input to int16 array");
-        }
+    //py::gil_scoped_release release;
 
-        // 修复点：使用花括号初始化
-        py::object arr{py::handle<>(py_array)};
-
-        // 检查数据长度
-        PyArrayObject* npArray = reinterpret_cast<PyArrayObject*>(arr.ptr());
-        if (PyArray_SIZE(npArray) != static_cast<npy_intp>(dataLen)) {
-            Py_DECREF(py_array);
-            throw std::runtime_error("Array length does not match dataLen");
-        }
-
-        // 处理数据...
-        void* dataPtr = PyArray_DATA(npArray);
-        int ret = RTCContext::instance().sendCustomAudioData(
-                destChannelIndex, dataPtr, sampleRate, channelNum, dataLen
-        );
-        Py_DECREF(py_array);  // 释放临时数组
-        return ret;
-    } catch (...) {
-        PyErr_SetString(PyExc_RuntimeError, "Invalid audio data");
-        return -1;
+    py::array_t<int16_t> contiguous = py::array::ensure(inputArray);
+    if (!contiguous) throw py::value_error("Array conversion failed");
+    auto buf = contiguous.request();
+    if (buf.size != dataLen) {
+        throw py::value_error("Array length does not match dataLen");
     }
+    std::vector<int16_t> localCopy(static_cast<int16_t*>(buf.ptr),
+                                    static_cast<int16_t*>(buf.ptr) + buf.size);
+
+    return RTCContext::instance().sendCustomAudioData(
+            destChannelIndex, localCopy.data(), sampleRate, channelNum, dataLen
+    );
+    //return RTCContext::instance().sendCustomAudioData(
+    //        destChannelIndex, buf.ptr, sampleRate, channelNum, dataLen
+    //);
+}
+py::list getListData() {
+    return RTCContext::instance().getListData();
+}
+int getSize() {
+    return RTCContext::instance().getSize();
+}
+RetAudioFrame getData() {
+    return RTCContext::instance().getData();
+}
+int16_t getDataCount() {
+    return RTCContext::instance().getDataCount();
 }
 
-void init_numpy() {
-    // 直接调用底层函数，绕过宏的问题
-    if (_import_array() < 0) {
-        PyErr_SetString(PyExc_ImportError, "numpy.core.multiarray failed to import");
-        throw std::runtime_error("NumPy initialization failed");
-    }
-    std::cout << "NumPy API addr: " << PyArray_API << std::endl;
-}
-BOOST_PYTHON_MODULE(rtc_plugins) {
-    try {
-        init_numpy();
-        void** numpyApi = (void**)PyArray_API;
-        if (!numpyApi || !numpyApi[93]) { // 93是PyArray_SimpleNew的偏移量
-            std::cout << "NumPy API corrupt! Key functions missing." << std::endl;
-            PyErr_Print();
-            throw std::runtime_error("Invalid NumPy API state");
-        } else {
-            RTCContext::instance().setNumpyApi(numpyApi);
-            std::cout << "set numpyApi succ:" << numpyApi[93] << std::endl;
-        }
+PYBIND11_MODULE(rtc_plugins, m) {
+    // 可选：暴露 RetAudioFrame 类（需额外绑定）
+    py::class_<RetAudioFrame>(m, "RetAudioFrame")
+        // 定义 data 属性（包含 getter 和 setter）
+        .def_property("data",
+            // Getter：返回 NumPy 数组
+            [](RetAudioFrame& self) {
+                return py::array_t<short>(
+                    {self.dataCount},    // 数组形状
+                    {sizeof(short)},     // 步长
+                    self.data.get()      // 数据指针（原始内存）
+                );
+            },
+            // Setter：从 NumPy 数组复制数据
+            [](RetAudioFrame& self, py::array_t<short> arr) {
+                auto buf = arr.request();  // 获取数组信息
+                self.data.reset(new short[buf.size]);  // 重新分配内存
+                std::memcpy(
+                        self.data.get(),       // 目标指针
+                        buf.ptr,               // 源数据指针
+                        buf.size * sizeof(short)  // 数据大小
+                );
+                self.dataCount = buf.size;  // 更新数据长度
+           }
+        )
+        .def_readwrite("dataCount", &RetAudioFrame::dataCount)
+        .def_readwrite("sampleRate", &RetAudioFrame::sampleRate)
+        .def_readwrite("numChannels", &RetAudioFrame::numChannels)
+        .def_readwrite("channelIndex", &RetAudioFrame::channelIndex);
+    m.def("init", &init);
+    m.def("initRecv", &initRecv);
+    m.def("initSend", &initSend);
+    m.def("sendCustomAudioData", &sendCustomAudioData);
+    m.def("getSize", &getSize);
+    m.def("getData", &getData);
+    m.def("getNumpyData", &getNumpyData);
+    m.def("getListData", &getListData);
+    m.def("getDataCount", &getDataCount);
+
 
-        py::def("init", &init);
-        py::def("initRecv", &initRecv);
-        py::def("initSend", &initSend);
-        py::def("sendCustomAudioData", &sendCustomAudioData);
-    } catch (...) {
-        PyErr_SetString(PyExc_RuntimeError, "Module initialization failed");
-    }
 }

test_audio_r.py

@@ -0,0 +1,45 @@
+import rtc_plugins
+import time
+import sounddevice as sd
+import numpy as np
+import mmap
+import os
+from ctypes import c_int16
+import struct
+
+srcUserId = "srcUser12"
+destUserId = "destUser12"
+
+srcDisplayName = "srcDisplayName12"
+destDisplayName = "destDisplayName12"
+srcRoomId = "srcRoom12"
+#destRoomId = "destRoomId12"
+destRoomId = srcRoomId
+srcChannelIndex = 46
+destChannelIndex = 47
+def my_callback_r(shmName, dataSize, dataCount, sampleRate, numChannels, channelIndex):
+    print(f"my_callback_r, dataSize:{dataSize}, dataCount:{dataCount}, sampleRate:{sampleRate}, numChannels:{numChannels}, channelIndex:{channelIndex}")
+    print(f"data:{shmName}")
+    print("after my_callback_r")
+
+ret = rtc_plugins.init(destUserId, destDisplayName, destRoomId, my_callback_r)
+if ret != 0:
+    print(f"init fail, ret:{ret}")
+    exit(1)
+ret = rtc_plugins.initRecv(destRoomId, srcUserId, destChannelIndex)
+if ret != 0:
+    print(f"initRecv fail, ret:{ret}")
+    exit(1)
+#ret = rtc_plugins.initSend(destRoomId, srcRoomId, srcChannelIndex, 1)
+#if ret != 0:
+#    print(f"initSend fail, ret:{ret}")
+#    exit(1)
+
+sampleRate = 16000
+while True:
+    frame = rtc_plugins.getListData()
+    sd.play(frame, sampleRate)
+    sd.wait()
+    print(f"get frame:{frame}")
+
+    time.sleep(0.005)

test_audio_s.py

@@ -0,0 +1,51 @@
+import rtc_plugins
+import time
+import numpy as np
+from scipy.io import wavfile
+
+srcUserId = "srcUser12"
+destUserId = "destUser12"
+
+srcDisplayName = "srcDisplayName12"
+destDisplayName = "destDisplayName12"
+srcRoomId = "srcRoom12"
+#destRoomId = "destRoomId12"
+destRoomId = srcRoomId
+srcChannelIndex = 46
+destChannelIndex = 47
+
+ret = rtc_plugins.init(srcUserId, srcDisplayName, srcRoomId)
+if ret != 0:
+    print(f"init fail, ret:{ret}")
+    exit(1)
+ret = rtc_plugins.initSend(srcRoomId, destRoomId, destChannelIndex, 1)
+if ret != 0:
+    print(f"initSend fail, ret:{ret}")
+    exit(1)
+#audioData = np.array([0, 1, -1, 0], dtype=np.int16)
+sampleRate, audioData = wavfile.read("xusample1.wav")
+print(f"sampleRate:{sampleRate} HZ")
+print(f"shape:{audioData.shape}")
+print(f"type:{audioData.dtype}")
+if audioData.dtype != np.int16:
+    audioData = (audioData * 32767).astype(np.int16)
+ret = rtc_plugins.sendCustomAudioData(destChannelIndex, audioData, sampleRate, 1, len(audioData))
+if ret != 0:
+    print(f"send fail, ret:{ret}")
+print("send succ")
+ret = rtc_plugins.initRecv(srcRoomId, srcUserId, srcChannelIndex)
+if ret != 0:
+    print(f"initRecv fail, ret:{ret}")
+    exit(1)
+for i in range(100):
+    ret = rtc_plugins.sendCustomAudioData(destChannelIndex, audioData, sampleRate, 1, len(audioData))
+    if ret != 0:
+        print(f"send fail, ret:{ret}")
+
+    #size = rtc_plugins.getSize()
+    #print(f"data size:{size}")
+    #frame = rtc_plugins.getListData()
+    #print(f"get frame:{frame}")
+    #dataCount = rtc_plugins.getDataCount()
+    #print(f"data count:{dataCount}")
+    time.sleep(3)

test_send.py

@@ -17,7 +17,7 @@ ret = rtc_plugins.init(srcUserId, srcDisplayName, srcRoomId, my_callback)
 if ret != 0:
     print(f"init fail, ret:{ret}")
     exit(1)
-ret = rtc_plugins.initSend(srcRoomId, destRoomId, destChannelIndex, 1)
+ret = rtc_plugins.initSend(srcRoomId, destRoomId, destChannelIndex)
 if ret != 0:
     print(f"initSend fail, ret:{ret}")
     exit(1)

test_time_r.py

@@ -4,6 +4,7 @@ import numpy as np
 import mmap
 import os
 from ctypes import c_int16
+import struct
 
 srcUserId = "srcUser12"
 destUserId = "destUser12"
@@ -15,11 +16,12 @@ srcRoomId = "srcRoom12"
 destRoomId = srcRoomId
 srcChannelIndex = 46
 destChannelIndex = 47
-def my_callback(shmName, dataSize, dataCount, sampleRate, numChannels, channelIndex):
-    print(f"dataSize:{dataSize}, dataCount:{dataCount}, sampleRate:{sampleRate}, numChannels:{numChannels}, channelIndex:{channelIndex}")
+def my_callback_r(shmName, dataSize, dataCount, sampleRate, numChannels, channelIndex):
+    print(f"my_callback_r, dataSize:{dataSize}, dataCount:{dataCount}, sampleRate:{sampleRate}, numChannels:{numChannels}, channelIndex:{channelIndex}")
     print(f"data:{shmName}")
+    print("after my_callback_r")
 
-ret = rtc_plugins.init(destUserId, destDisplayName, destRoomId, my_callback)
+ret = rtc_plugins.init(destUserId, destDisplayName, destRoomId, my_callback_r)
 if ret != 0:
     print(f"init fail, ret:{ret}")
     exit(1)
@@ -34,10 +36,17 @@ if ret != 0:
 
 while True:
     print("recv")
-    #audioData = np.array([0, 1, -1, 0], dtype=np.int16)
-    #ret = rtc_plugins.sendCustomAudioData(srcChannelIndex, audioData, 48000, 1, len(audioData))
-    #if ret != 0:
-    #    print(f"resend fail, ret:{ret}")
-    #else:
-    #    print("resend succ")
-    time.sleep(30)
+    audioData = np.array([0, 1, -1, 0], dtype=np.int16)
+    ret = rtc_plugins.sendCustomAudioData(srcChannelIndex, audioData, 48000, 1, len(audioData))
+    if ret != 0:
+        print(f"resend fail, ret:{ret}")
+    else:
+        print("resend succ")
+    size = rtc_plugins.getSize()
+    print(f"data size:{size}")
+    #frame = rtc_plugins.getNumpyData()
+    frame = rtc_plugins.getListData()
+    print(f"get frame:{frame}")
+    dataCount = rtc_plugins.getDataCount()
+    print(f"data count:{dataCount}")
+    time.sleep(0.005)

test_time_s.py

@@ -12,8 +12,10 @@ srcRoomId = "srcRoom12"
 destRoomId = srcRoomId
 srcChannelIndex = 46
 destChannelIndex = 47
-def my_callback(npData, dataCount, sampleRate, numChannels, channelIndex):
-    print(f"dataCount:{dataCount}, sampleRate:{sampleRate}, numChannels:{numChannels}, channelIndex:{channelIndex}")
+def my_callback(shmName, dataSize, dataCount, sampleRate, numChannels, channelIndex):
+    print(f"my_callback, dataSize:{dataSize}, dataCount:{dataCount}, sampleRate:{sampleRate}, numChannels:{numChannels}, channelIndex:{channelIndex}")
+    print(f"data:{shmName}")
+    print("after my_callback_r")
 ret = rtc_plugins.init(srcUserId, srcDisplayName, srcRoomId, my_callback)
 if ret != 0:
     print(f"init fail, ret:{ret}")
@@ -34,4 +36,11 @@ for i in range(100):
     ret = rtc_plugins.sendCustomAudioData(destChannelIndex, audioData, 48000, 1, len(audioData))
     if ret != 0:
         print(f"send fail, ret:{ret}")
-    time.sleep(30)
+
+    size = rtc_plugins.getSize()
+    print(f"data size:{size}")
+    frame = rtc_plugins.getListData()
+    print(f"get frame:{frame}")
+    dataCount = rtc_plugins.getDataCount()
+    print(f"data count:{dataCount}")
+    time.sleep(3)

util/RTCContext.cpp

@@ -1,5 +1,5 @@
 #include "RTCContext.h"
-#define GIL
+#include <fstream>
 
 void RTCContext::onRoom(uint32_t typeId, RTCENGINE_NAMESPACE::MRTCRoomInfo& roomInfo) {
     //std::cout << "RTCContext::onRoom():" << roomInfo.roomId << "," << roomInfo.displayName << "," << roomInfo.userId << "," << roomInfo.message;
@@ -7,14 +7,16 @@ void RTCContext::onRoom(uint32_t typeId, RTCENGINE_NAMESPACE::MRTCRoomInfo& room
     std::lock_guard<std::mutex> lock(mutex_);
     isOnRoom_ = true;
 }
-void RTCContext::onConsumer(uint32_t msgId, const char* roomId, const char* peerId,
-                            RTCENGINE_NAMESPACE::MRTCConsumerInfo& consumerInfo) {
-    std::cout << "RTCContext::onConsumer():" << consumerInfo.roomId << "," << consumerInfo.displayName << ","
-              << consumerInfo.channelIndex << std::endl;
-    if (isRecv_) {
-        std::lock_guard<std::mutex> lock(mutex_);
+void RTCContext::onConsumer(uint32_t msgId, const char* roomId, const char* peerId, RTCENGINE_NAMESPACE::MRTCConsumerInfo& consumerInfo) {
+    std::cout << "RTCContext::onConsumer()" << std::endl;
+    std::cout << "RTCContext::onConsumer():msgId:" << msgId << ", roomId:" << consumerInfo.roomId << ", displayName:"
+              << consumerInfo.displayName << ", type:" << consumerInfo.audioSourceType << ", channelIndex:" << (int)consumerInfo.channelIndex
+              << std::endl;
+    if (msgId == RTCENGINE_NAMESPACE::MRTCConsumerStatusType::TYPE_STATUS_CONSUMER_START && std::string(consumerInfo.kind) == "audio") {
+        std::lock_guard <std::mutex> lock(mutex_);
+        isOnConsumer_ = true;
         std::cout << "registerSoundLevelListener" << std::endl;
-        int16_t ret1 = rtcEngine_->registerSoundLevelListener(mrtc::TYPE_AUDIO_SOURCE_CUSTOM, consumerInfo.roomId,
+        int16_t ret1 = rtcEngine_->registerSoundLevelListener(consumerInfo.audioSourceType, roomId,
                                                               peerId, consumerInfo.channelIndex, this);
         if (0 != ret1) {
             std::cout << "RTCContext::instance().registerSoundLevelListener() inUser failed, ret:" << ret1;
@@ -22,7 +24,7 @@ void RTCContext::onConsumer(uint32_t msgId, const char* roomId, const char* peer
         }
 
         std::cout << "muteAudio" << std::endl;
-        int16_t ret2 = rtcEngine_->muteAudio(consumerInfo.roomId, peerId, mrtc::TYPE_AUDIO_SOURCE_CUSTOM,
+        int16_t ret2 = rtcEngine_->muteAudio(roomId, peerId, consumerInfo.audioSourceType,
                                              false, consumerInfo.channelIndex);
         if (0 != ret2) {
             std::cout << "RTCContext::instance().muteAudio() failed, ret:" << ret2;
@@ -45,6 +47,7 @@ void RTCContext::onCallBackMessage(uint32_t msgId, const char* msg) {
     }
 
     std::cout << "RTCContext::onCallBackMessage(), msgId:" << msgId << ", msg:" << msg << std::endl;
+    //std::cout << "RTCContext::onCallBackMessage()" << std::endl;
 }
 void RTCContext::onCallBackCustomData(RTCENGINE_NAMESPACE::MRTCCustomDataObject object) {
     //std::cout << "RTCContext::onCallBackCustomData(), obj:" << object.peerId << "," << object.data << "," << object.data_length;
@@ -53,154 +56,43 @@ void RTCContext::onCallBackCustomData(RTCENGINE_NAMESPACE::MRTCCustomDataObject
 void RTCContext::onSoundLevelUpdate(const char* roomId, const char* peerId, uint16_t audioSourceType,
                                     uint8_t channelIndex, uint16_t volume, int32_t vad)
 {
-    std::cout << "RTCContext::onSoundLevelUpdate()" << std::endl;
+    //std::cout << "RTCContext::onSoundLevelUpdate()" << std::endl;
 }
 
+void printTimestamp() {
+    // 获取系统当前时间点
+    auto now = std::chrono::system_clock::now();
+
+    // 转换为时间戳（秒 + 毫秒）
+    auto timestamp = std::chrono::duration_cast<std::chrono::seconds>(
+            now.time_since_epoch()).count();
+    auto milliseconds = std::chrono::duration_cast<std::chrono::milliseconds>(
+            now.time_since_epoch()).count() % 1000;
+
+    // 转换为本地时间（可读格式）
+    std::time_t time = std::chrono::system_clock::to_time_t(now);
+    std::cout << "Timestamp: " << timestamp << "." << milliseconds << std::endl;
+}
 void RTCContext::onAudioProcess(const char* roomId, const char* peerId,
                                 mrtc::MRTCAudioFrame& audioFrame,
                                 mrtc::MRTCAudioSourceType audioSourceType)
 {
-    namespace np = boost::python::numpy;
-    namespace py = boost::python;
-    Py_Initialize();          // 初始化 Python
-    np::initialize();
-    std::cout << "=== 开始音频处理 ===" << std::endl;
-    std::cout << "audioFrame:" << audioFrame.dataCount << "," << audioFrame.sampleRate << "," <<
-               audioFrame.numChannels << "," << audioFrame.channelIndex << std::endl;
-
-    // 1. 获取GIL
-    std::cout << "[1] 获取GIL锁..." << std::endl;
-#ifdef GIL
-    PyGILState_STATE gstate = PyGILState_Ensure();
-#endif
-
-    try {
-        // 2. 输入参数校验
-        std::cout << "[2] 检查输入参数..." << std::endl;
-        std::cout << "    dataCount: " << audioFrame.dataCount
-                  << " (max: " << std::numeric_limits<npy_intp>::max() << ")" << std::endl;
-
-        if (!audioFrame.data || audioFrame.dataCount <= 0) {
-            std::cout << "[ERROR] 无效音频数据指针或长度" << std::endl;
-            throw std::invalid_argument("Invalid audio frame data");
-        }
-
-        const size_t data_size = audioFrame.dataCount * sizeof(int16_t);
-
-        std::cout << "step1" << std::endl;
-        namespace py = boost::python;
-        namespace np = boost::python::numpy;
-        npy_intp shape[1] = { static_cast<npy_intp>(audioFrame.dataCount) };
-        std::cout << "step2" << std::endl;
-
-        // 7. 执行回调
-        if (!pyCallback_.is_none()) {
-            std::cout << "[7] 准备执行Python回调..." << std::endl;
-            // 增加引用计数防止提前释放
-            //Py_INCREF(pyCallback_.ptr());
-            try {
-
-                //PyGILState_STATE gstate = PyGILState_Ensure();
-                std::cout << "data:" << audioFrame.data << std::endl;
-                std::cout << "当前线程是否持有 GIL: " << PyGILState_Check() << std::endl;
-                np::dtype dtype = np::dtype::get_builtin<int16_t>();
-                std::cout << "init dtype" << std::endl;
-                if (!Py_IsInitialized()) {
-                    std::cerr << "Python 解释器未初始化！" << std::endl;
-                    return;
-                }
-                try {
-                    py::object str_repr = py::str(dtype);
-                    std::cout << "str_repr" << std::endl;
-                    if(str_repr.ptr() != Py_None) {
-                        std::cout << "str_repr is not null" << std::endl;
-                        std::string dtype_str = py::extract<std::string>(str_repr);
-                        std::cout << "数据类型: " << dtype_str << std::endl;
-                    } else {
-                        std::cout << "数据类型: None" << std::endl;
-                    }
-                } catch (const py::error_already_set&) {
-                    std::cout<< "数据类型转换错误" << std::endl;
-                    PyErr_Clear();
-                }
-                std::cout << "数据形状: " << shape[0] << std::endl;
-                np::ndarray audioArray = np::from_data(
-                        audioFrame.data,                // 数据指针
-                        dtype,                          // 数据类型 (int16)
-                        py::make_tuple(shape[0]),       // 形状 (1D)
-                        py::make_tuple(sizeof(int16_t)), // 步长
-                        py::object()                    // 所有者（Python管理）
-                );
-                std::cout << "    数据拷贝完成" << std::endl;
-                pyCallback_(
-                        audioArray,                     // numpy 数组
-                        data_size,          // 数据大小
-                        audioFrame.dataCount,
-                        audioFrame.sampleRate,
-                        audioFrame.numChannels,
-                        audioFrame.channelIndex
-                );
-                std::cout << "    after callback" << std::endl;
-                if (PyErr_Occurred()) {
-                    PyObject *type, *value, *traceback;
-                    PyErr_Fetch(&type, &value, &traceback);
-                    if (value) {
-                        PyObject* str = PyObject_Str(value);
-                        if (str) {
-                            std::cerr << "Python Error: " << PyUnicode_AsUTF8(str) << std::endl;
-                            Py_DECREF(str);
-                        }
-                    }
-                    Py_XDECREF(type);
-                    Py_XDECREF(value);
-                    Py_XDECREF(traceback);
-                    //PyErr_Print();
-                    throw std::runtime_error("Python callback error");
-                }
-                std::cout << "    回调执行成功" << std::endl;
-
-            } catch (const py::error_already_set& e) {
-                std::cerr << "[PYTHON ERROR] ";
-                PyErr_Print();  // 自动打印到stderr
-                // 可选：获取更详细的错误信息
-                if (PyErr_Occurred()) {
-                    PyObject *type, *value, *traceback;
-                    PyErr_Fetch(&type, &value, &traceback);
-                    std::cerr << "Details: "
-                              << PyUnicode_AsUTF8(PyObject_Str(value)) << std::endl;
-                    PyErr_Restore(type, value, traceback);
-                }
-                //Py_DECREF(pyCallback_.ptr());
-            } catch (...) {
-                std::cout << "[ERROR] 回调执行失败" << std::endl;
-
-                //Py_DECREF(pyCallback_.ptr());
-                throw;
-            }
-            //Py_DECREF(pyCallback_.ptr());
-        } else {
-            std::cout << "[7] 无回调函数设置" << std::endl;
-        }
-
-        // 8. 释放资源
-        std::cout << "[8] 释放共享内存资源..." << std::endl;
-
-        std::cout << "[9] 释放GIL..." << std::endl;
-#ifdef GIL
-        PyGILState_Release(gstate);
-#endif
-        std::cout << "=== 音频处理完成 ===" << std::endl;
-
-    } catch (const std::exception& e) {
-        std::cout << "[EXCEPTION] 异常捕获: " << e.what() << std::endl;
-#ifdef GIL
-        PyGILState_Release(gstate);
-#endif
-        std::cerr << "Audio process error: " << e.what() << std::endl;
+    //std::cout << "onAudioProcess, roomId:" << roomId << ", peerId:" << peerId << ", type:" << audioSourceType
+    //          << ", datacount:" << audioFrame.dataCount << std::endl;
+    if (audioFrame.data == nullptr || audioFrame.dataCount <= 0) {
+        return;
     }
-#ifdef GIL
-    PyGILState_Release(gstate);
-#endif
+    // 以二进制追加模式打开文件
+    //std::ofstream outFile("audio_data.pcm", std::ios::binary | std::ios::app);
+    //if (!outFile.is_open()) {
+    //    std::cerr << "Failed to open file for writing!" << std::endl;
+    //    return;
+    //}
+
+    //// 写入数据（int16_t格式）
+    //outFile.write(reinterpret_cast<const char*>(audioFrame.data),
+    //              audioFrame.dataCount);
+    setData(audioFrame);
 }
 
 void RTCContext::onProducer(uint32_t msgId, mrtc::MRTCProducerInfo& info)
@@ -210,12 +102,6 @@ void RTCContext::onProducer(uint32_t msgId, mrtc::MRTCProducerInfo& info)
 }
 bool RTCContext::init(const char* selfUserId, const char* selfDisplayName, const char* selfRoomId)
 {
-    std::cout << "init, numpyApi_:" << numpyApi_[93] << std::endl;
-    if (!numpyApi_ || !numpyApi_[93]) { // 93是PyArray_SimpleNew的偏移量
-        std::cout << "numpyApi_ is null in init" << std::endl;
-    } else {
-        std::cout << "numpyApi_ is not null in init" << std::endl;
-    }
     mrtc::IMRTCEngineFactory * rtcFactory = mrtc::getMRTCEngineFactory();
     if (!rtcFactory)
     {
@@ -261,20 +147,21 @@ bool RTCContext::init(const char* selfUserId, const char* selfDisplayName, const
         std::cout << "RTCContext::instance().registerListener() failed" << std::endl;
         return false;
     }
-    //namespace py = boost::python;
-    //namespace np = boost::python::numpy;
-    //Py_Initialize();          // 初始化 Python
-    //np::initialize();
 
     return true;
 }
 bool RTCContext::initRecv(const char* destRoomId, const char* srcUserId, const int16_t destChannelIndex)
 {
-    isRecv_ = true;
+    while (!isOnConsumer_)
+    {
+        std::cout << "wait for OnConsumer" << std::endl;
+        sleep(3);
+    }
     return true;
 
 }
-bool RTCContext::initSend(const char* srcRoomId, const char* destRoomId, const int16_t destChannelIndex, uint8_t channelNum)
+bool RTCContext::initSend(const char* srcRoomId, const char* destRoomId, const int16_t destChannelIndex,
+                          const uint8_t channelNum)
 {
     while (!isOnRoom_)
     {
@@ -295,11 +182,11 @@ bool RTCContext::initSend(const char* srcRoomId, const char* destRoomId, const i
     }
 
     mrtc::MRTCAudioOption option;
-    option.channel = channelNum;
     if (std::string(srcRoomId) != std::string(destRoomId)) {
         strcpy(option.dstRoomId, destRoomId);
     }
     option.channelIndex = destChannelIndex;
+    option.channel = channelNum;
     std::cout << "startCustomAudio" << std::endl;
     int16_t ret2 = rtcEngine_->startCustomAudio(option);
     if (ret2 != 0)
@@ -310,16 +197,12 @@ bool RTCContext::initSend(const char* srcRoomId, const char* destRoomId, const i
     std::cout << "init send succ" << std::endl;
     return true;
 }
-bool RTCContext::initGIL() {
-    isGIL_ = true;
-}
 
 void RTCContext::destorySend(const int16_t selfChannelIndex)
 {
     rtcEngine_->stopCustomAudio(selfChannelIndex);
 }
-int16_t RTCContext::sendAudioData(uint8_t channelIndex, const void* pData, int32_t nSampleRate, uint64_t nNumberOfChannels,
-                                  uint64_t dataLength)
+int16_t RTCContext::sendAudioData(uint8_t channelIndex, const void* pData, int32_t nSampleRate, uint64_t nNumberOfChannels, uint64_t dataLength)
 {
     std::lock_guard<std::mutex> lock(mutex_);
     if (pData_)
@@ -341,8 +224,7 @@ int16_t RTCContext::sendCustomAudioData(const int16_t channelIndex, void* custom
         return -1;
     }
     std::cout << "customData addr is:" << customData << std::endl;
-    return rtcEngine_->sendCustomAudioData(channelIndex, customData, sampleRate,
-                                           channelNum, dataLen);
+    return rtcEngine_->sendCustomAudioData(channelIndex, customData, sampleRate, channelNum, dataLen);
 }
 mrtc::IMRTCEngine* RTCContext::getRtcEngine() const
 {
@@ -359,12 +241,75 @@ void RTCContext::setpData(void* pData)
     std::lock_guard<std::mutex> lock(mutex_);
     pData_ = pData;
 }
-void RTCContext::setPyCallback(boost::python::object callback) {
-    std::lock_guard<std::mutex> lock(mutex_);
-    pyCallback_ = callback;
+
+void RTCContext::setData(const mrtc::MRTCAudioFrame& frame) {
+    std::lock_guard<std::mutex> lock(dataMutex_);
+    if (dataSize_ == totalSize_) {
+        bottom_ = (bottom_ + 1) % totalSize_;
+        dataSize_--;
+    }
+    RetAudioFrame newFrame;
+    newFrame.dataCount = frame.dataCount;
+    newFrame.sampleRate = frame.sampleRate;
+    newFrame.numChannels = frame.numChannels;
+    newFrame.channelIndex = frame.channelIndex;
+    newFrame.data = std::make_unique<int16_t[]>(frame.dataCount);
+    //std::memcpy(newFrame.data.get(), frame.data, frame.dataCount* sizeof(int16_t));
+    std::memcpy(newFrame.data.get(), frame.data, frame.dataCount);
+    data_[head_] = std::move(newFrame);
+    head_ = (head_ + 1) % totalSize_;
+    dataSize_++;
+
 }
-void RTCContext::setNumpyApi(void **numpyApi) {
-    std::lock_guard<std::mutex> lock(mutex_);
-    numpyApi_ = numpyApi;
-    std::cout << "setNupyApi, numpyApi_:" << numpyApi_[93] << std::endl;
+
+RetAudioFrame RTCContext::getData() {
+    //std::lock_guard<std::mutex> lock(dataMutex_);
+    if (dataSize_ > 0) {
+        RetAudioFrame frame = std::move(data_[bottom_]); // 移动而非拷贝
+        bottom_ = (bottom_ + 1) % totalSize_;
+        dataSize_--;
+        return frame; // 返回值优化(RVO)会生效
+    }
+    return {}; // 返回空对象
 }
+py::array_t<int16_t> RTCContext::getNumpyData() {
+    std::cout << "step1" << std::endl;
+    std::lock_guard<std::mutex> lock(dataMutex_);
+    RetAudioFrame frame = getData();
+    std::cout << "step2" << std::endl;
+    int16_t* dataPtr = frame.data.get();  // 你的数据指针
+    std::cout << "step3" << std::endl;
+    size_t length = frame.dataCount;    // 数据长度
+    std::cout << "step4" << std::endl;
+    if (!dataPtr || length == 0) {
+        return py::array_t<int16_t>({0});  // 返回空数组
+    }
+
+    // 直接构造 pybind11 的 NumPy 数组（自动管理内存）
+    py::array_t<int16_t> result({static_cast<py::ssize_t>(length)});
+    auto buf = result.mutable_unchecked();
+    for (size_t i = 0; i < length; i++) {
+        buf[i] = dataPtr[i];
+    }
+    return result;
+}
+py::list RTCContext::getListData() {
+    std::lock_guard<std::mutex> lock(dataMutex_);
+    RetAudioFrame frame = getData();
+    py::list result;
+    if (frame.data) {
+        for (int i = 0; i < frame.dataCount/sizeof(int16_t); i++) {
+            result.append(frame.data.get()[i]);
+        }
+    }
+    return result;
+}
+int16_t RTCContext::getDataCount() {
+    std::lock_guard<std::mutex> lock(dataMutex_);
+    RetAudioFrame frame = getData();
+    return frame.dataCount;
+}
+int16_t RTCContext::getSize() {
+    std::lock_guard<std::mutex> lock(dataMutex_);
+    return dataSize_;
+}

util/RTCContext.h

@@ -1,9 +1,6 @@
 // RTCContext.h
 #pragma once
 
-//#include "numpyConfig.h"
-#include "numpyStub.h"
-
 #include "IMRTCEngine.hpp"
 #include "MRTCEngineDefine.hpp"
 #include "IMRTCEngineFactory.hpp"
@@ -18,23 +15,26 @@
 #include <sys/stat.h>
 
 #include <Python.h>
-#include <boost/python.hpp>
-#include <boost/python/detail/wrap_python.hpp>
-#include <boost/python/numpy.hpp>
-#include <boost/python/detail/prefix.hpp>
-#include <boost/python/module.hpp>
-#include <boost/python/def.hpp>
-#include <numpy/ndarrayobject.h>
-#include <numpy/arrayobject.h>
+// pybind11 头文件
+#include <pybind11/pybind11.h>
+#include <pybind11/numpy.h>
+#include <pybind11/stl.h>
+namespace py = pybind11;
+//#include <numpy/arrayobject.h>
 
-
-// 必须声明外部变量（关键！）
-//#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
-
-namespace fs = std::filesystem;
 #define ENV_PRODUCT
 //#define SEND_MODE
 
+// 音频数据帧
+struct RetAudioFrame
+{
+    std::unique_ptr<int16_t[]> data;
+    int dataCount = 0;
+    int sampleRate = 48000;
+    int numChannels = 1;
+    int channelIndex = 0;
+};
+
 class RTCContext :
         public RTCENGINE_NAMESPACE::IMRTCRoomCallBack,
         public RTCENGINE_NAMESPACE::IMRTCConsumerCallBack,
@@ -60,37 +60,22 @@ public:
         static RTCContext instance;
         return instance;
     }
-    static void** numpy_api() {
-        static void** api = [](){
-            // 强制初始化NumPy
-            if (_import_array() < 0) {
-                PyErr_Print();
-                throw std::runtime_error("NumPy initialization failed");
-            }
 
-            void** ptr = reinterpret_cast<void**>(RTC_PLUGINS_ARRAY_API);
-            std::cout << "ptr:" << ptr << std::endl;
-            if (!ptr || !ptr[93]) {
-                std::cerr << "NumPy API corrupt! Expected at 93: "
-                          << (ptr ? (void*)ptr[93] : nullptr) << std::endl;
-                abort();
-            }
-            return ptr;
-        }();
-        return api;
-    }
     RTCContext(const RTCContext&) = delete;
     RTCContext& operator=(const RTCContext&) = delete;
     mrtc::IMRTCEngine* getRtcEngine() const;
     bool init(const char* selfUserId, const char* selfDisplayName, const char* selfRoomId);
     bool initRecv(const char* destRoomId, const char* srcUserId, const int16_t destChannelIndex);
-    bool initSend(const char* srcRoomId, const char* destRoomId, const int16_t destChannelIndex, uint8_t channelNum);
-    bool initGIL();
+    bool initSend(const char* srcRoomId, const char* destRoomId, const int16_t destChannelIndex, const uint8_t channelNum);
+    int16_t getSize();
+    void setData(const mrtc::MRTCAudioFrame& frame);
+    RetAudioFrame getData();
+    py::array_t<int16_t> getNumpyData();
+    py::list getListData();
+    int16_t getDataCount();
 
     void* getpData() const;
     void setpData(void* pData);
-    void setPyCallback(boost::python::object callback);
-    void setNumpyApi(void** numpyApi);
 
     int16_t sendAudioData(uint8_t channelIndex = 0, const void* pData = nullptr, int32_t nSampleRate = 48000, uint64_t nNumberOfChannels = 2, uint64_t dataLength = 0);
     int16_t sendCustomAudioData(const int16_t channelIndex, void* customData, int32_t sampleRate,
@@ -100,17 +85,21 @@ public:
 private:
     RTCContext()
     {
+        data_.resize(totalSize_);
     }
     mutable std::mutex mutex_;
     mrtc::IMRTCEngine * rtcEngine_ = nullptr;
     void* pData_ = nullptr;
     bool isOnRoom_ = false;
-    bool isRecv_ = false;
+    bool isOnConsumer_ = false;
     bool isJoinMultiRoom_ = false;
     bool isMultiRoom_ = false;
-    bool isGIL_ = false;
-    boost::python::object pyCallback_;
-    void ** numpyApi_;
+    std::vector<RetAudioFrame> data_;
+    mutable std::mutex dataMutex_;
+    const int16_t totalSize_ = 100;
+    int16_t dataSize_ = 0;
+    int16_t bottom_ = 0;
+    int16_t head_= 0;
     void onRoom(uint32_t typeId, RTCENGINE_NAMESPACE::MRTCRoomInfo& roomInfo);
     void onConsumer(uint32_t msgId, const char* roomId, const char* peerId, RTCENGINE_NAMESPACE::MRTCConsumerInfo& consumerInfo);
     void onRender(const char* roomId, const char* peerId,

util/numpyStub.h

@@ -1,12 +0,0 @@
-// numpy_interface.h （新建）
-#pragma once
-
-#ifdef IMPLEMENT_NUMPY_API
-// 主模块实现路径
-    #define PY_ARRAY_UNIQUE_SYMBOL RTC_PLUGINS_ARRAY_API
-    #define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
-    #include <numpy/arrayobject.h>
-#else
-// 用户头文件路径
-extern void* RTC_PLUGINS_ARRAY_API[];  // 严格匹配NumPy类型
-#endif