环境搭建
CmakeLists.txt 的编写，参考规则就好
-
注意路径
-
用好指令
配置路径
include_directories，add_subdirectory，
add_library，target_link_libraries，set
打印调试
message
引入库文件，注意静态库文件，和动态库文件，静态库文件可以拆分的源码方式有助于缩减包大小。
- 子模块字库的划分可以查分多个CmakeLists.txt
测试环境
一般会帮你创建好native-lib ,那么先引入你用的库，调用一个简单的函数编译通过，证明成功
移植代码
有时候在桌面开发clion简单的demo比较容易，然后移植代码到 手机端。进行功能拆分和封装。
开发思路
先在java端定义好JNI调用函数Native相关调用逻辑
一般都是先准备好库初始化环境，然后调用具体的功能，然后是释放资源停止等相关。
具体功能，比如直播相关视频封装到VideoChannel.cpp 去实现。图片相关放到ImageUtils.cpp 实现。体现功能封装
java如果调用c的具体功能，可以在java端发起一个long类型的指针引用。传给JNI，通过JNI调用C++等cpp，回调给java，这个时候可以类似java在JNI 实现一个Callback的具体逻辑。C++调用Callback的h头文件接口回传给java
设计：自顶向下，java端，JNI层，C++端
视频播放和推流的开发流程模板
public class RtmpClient {
private static final String TAG = "RtmpClient";
static {
System.loadLibrary("native-lib");
}
private LifecycleOwner lifecycleOwner;
private int width;
private int height;
private boolean isConnect;
private VideoChannel videoChannel;
private AudioChannel audioChannel;
public RtmpClient(LifecycleOwner lifecycleOwner) {
this.lifecycleOwner = lifecycleOwner;
nativeInit();
}
public void initVideo(TextureView displayer, int width, int height, int fps, int bitRate) {
this.width = width;
this.height = height;
videoChannel = new VideoChannel(displayer, this, lifecycleOwner);
initVideoEnc(width, height, fps, bitRate);
}
public void initAudio(int sampleRate,int channels){
audioChannel = new AudioChannel(sampleRate, channels, this);
int inputByteNum = initAudioEnc(sampleRate, channels);
audioChannel.setInputByteNum(inputByteNum);
}
public void startLive(String url) {
connect(url);
}
public boolean isConnected() {
return isConnect;
}
public void stopLive() {
isConnect = false;
audioChannel.stop();
disConnect();
Log.e(TAG, "停止直播==========");
}
/**
* JNICall
*
* @param isConnect
*/
public void onPrepare(boolean isConnect) {
this.isConnect = isConnect;
audioChannel.start();
Log.e(TAG, "开始直播==========");
}
public void sendVideo(byte[] buffer) {
nativeSendVideo(buffer);
}
public void sendAudio(byte[] buffer,int len){
nativeSendAudio(buffer, len);
}
public void toggleCamera() {
videoChannel.toggleCamera();
}
public void release() {
videoChannel.release();
audioChannel.release();
releaseVideoEnc();
releaseAudioEnc();
nativeDeInit();
}
public int getWidth() {
return width;
}
public int getHeight() {
return height;
}
private native void nativeInit();
private native void connect(String url);
private native void disConnect();
private native void nativeDeInit();
private native void nativeSendVideo(byte[] buffer);
private native void nativeSendAudio(byte[] buffer,int len);
private native void initVideoEnc(int width, int height, int fps, int bitRate);
private native int initAudioEnc(int sampleRate,int channels);
private native void releaseVideoEnc();
private native void releaseAudioEnc();
}
准备环境 构造nativeinit（）
，链接 startLive（url）
初始化 initVideoEnc(width, height, fps, bitRate);
释放： stopLive()
回调Java：onPrepare(boolean isConnect)
OpenCV 开发模式的模板
public class FaceTracker {
private long mNativeObj = 0;
public FaceTracker(String model) {
mNativeObj = nativeCreateObject(model);
}
public void setSurface(Surface surface) {
nativeSetSurface(mNativeObj, surface);
}
public void release() {
nativeDestroyObject(mNativeObj);
mNativeObj = 0;
}
public void start() {
nativeStart(mNativeObj);
}
public void stop() {
nativeStop(mNativeObj);
}
///其他逻辑
public void detect(byte[] inputImage, int width, int height, int rotationdegrees) {
nativeDetect(mNativeObj, inputImage, width, height, rotationdegrees);
}
private static native void nativeDestroyObject(long thiz);
private static native void nativeSetSurface(long thiz, Surface surface);
private static native void nativeStart(long thiz);
private static native void nativeStop(long thiz);
private static native void nativeDetect(long thiz, byte[] inputImage, int width, int height, int rotationDegrees);
}
常用的对应的头文件和Java传递对应C++
include <android/native_window_jni.h>
ANativeWindow * window
对应Java 的Surface
线程同步问题
通常我们做的一些事，都是比较耗时，所以java调用c层可能是不同的子线程，这个时候需要线程同步
我们通常类似java Syncronized，在c中用pthread_mutex_t 枷锁来处理同步问题。
小技巧，通常情况我们会有多出判断 不停的return，看起来比较丑陋，可以运用do while循环优化
如：
pthread_mutex_lock(&mutex);
if(!window) {
pthread_mutex_unlock(&mutex);
return;
}
ANativeWindow_setBuffersGeometry(window,img.cols,img.rows,WINDOW_FORMAT_RGBA_8888);
ANativeWindow_Buffer buffer;
ANativeWindow_lock(window,&buffer,0);
pthread_mutex_unlock(&mutex);
这个里面多个函数需要判断返回值，没有需要跳出返回，用do while优化后，break 处理会好狠多
不需要每个retrun 前面都要写解锁
--->
pthread_mutex_lock(&mutex);
do {
if (!window) {
break;
}
ANativeWindow_setBuffersGeometry(window, img.cols, img.rows, WINDOW_FORMAT_RGBA_8888);
ANativeWindow_Buffer buffer;
if (ANativeWindow_lock(window, &buffer, 0)) {
ANativeWindow_release(window);
window = 0;
break;
}
uint8_t *dstData = static_cast<uint8_t *>(buffer.bits);
uint8_t *srcData = img.data;
int srclineSize = img.cols * 4;
int dstlineSize = buffer.stride * 4;
// 一行一行拷贝，因为需要填充步长
for (int i = 0; i < buffer.height; ++i) {
memcpy(dstData + i * dstlineSize, srcData + i * srclineSize, srclineSize);
}
} while (0);
pthread_mutex_unlock(&mutex);
这样就减少了判断和代码
报错分析
打印，分析日志
debug 不用加断点，出现c++ 问题直接会中断，我们查看调用栈，找问题
- 错误找不到库，大部分是搜索路径问题和编译器路径
1-12 12:49:15.487 9583-9583/? E/AndroidRuntime: FATAL EXCEPTION: main
Process: top.zcwfeng.opencv, PID: 9583
java.lang.UnsatisfiedLinkError: dlopen failed: library "libopencv_java4.so" not found
将apk反编译看一下。分析：我们自己的代码调用libopencv_java4.so
改下libopencv_java4.so 位置，不要放在jni目录改成jniLibs
- 找不到明面上的库
--------- beginning of crash
11-12 12:59:24.802 11388-11388/top.zcwfeng.opencv E/AndroidRuntime: FATAL EXCEPTION: main
Process: top.zcwfeng.opencv, PID: 11388
java.lang.UnsatisfiedLinkError: dlopen failed: library "libc++_shared.so" not found
dlopen failed: library "libc++_shared.so" not found . 是因为libopencv_java4 里面用到了，我们不知道。但是ndk官方开发文档有说明，虽然gradle是自动配置，但是我们需要一个参数
arguments 如下配置
externalNativeBuild {
cmake {
cppFlags ""
arguments "-DANDROID_STL=c++_shared"
abiFilters 'armeabi-v7a'
}
ndk{
abiFilters 'armeabi-v7a'
}
}
网友评论