NuPlayer源码分析（二）NuPlayer的音频buffer处理机制

一、引言：
在上一篇博客，对nuplayer的创建流程有了一个分析，并且在末尾引入了输入输出buffer的处理出处，这篇博客将分析nuplayer对音频数据的处理机制。
先贴出输入输出buffer的核心处理函数：

--------------------------------------------------------------------------
onMessageReceived@frameworks\av\media\libmediaplayerservice\nuplayer\NuPlayerDecoder.cpp：
--------------------------------------------------------------------------
 case kWhatCodecNotify:
 {
   
     if (!isStaleReply(msg)) {
   
         int32_t numInput, numOutput;

         if (!msg->findInt32("input-buffers", &numInput)) {
   
             numInput = INT32_MAX;
         }

         if (!msg->findInt32("output-buffers", &numOutput)) {
   
             numOutput = INT32_MAX;
         }

         if (!mPaused) {
   
             while (numInput-- > 0 && handleAnInputBuffer()) {
   }
         }

         while (numOutput-- > 0 && handleAnOutputBuffer()) {
   }
     }

     requestCodecNotification();
     break;
 }

二、音频输入数据处理：
先看handleAnInputBuffer()函数：

bool NuPlayer::Decoder::handleAnInputBuffer() {
   
    if (mFormatChangePending) {
   
        return false;
    }
    size_t bufferIx = -1;
    /* 从MediaCodec中获取可用的input buffer用于填充数据 */
    status_t res = mCodec->dequeueInputBuffer(&bufferIx);
    ALOGV("[%s] dequeued input: %d",
            mComponentName.c_str(), res == OK ? (int)bufferIx : res);
    if (res != OK) {
   
        if (res != -EAGAIN) {
   
            ALOGE("Failed to dequeue input buffer for %s (err=%d)",
                    mComponentName.c_str(), res);
            handleError(res);
        }
        return false;
    }
	/* 根据index找到对应的buffer并做一些检查工作 */
    CHECK_LT(bufferIx, mInputBuffers.size());
    if (mMediaBuffers[bufferIx] != NULL) {
   
        mMediaBuffers[bufferIx]->release();
        mMediaBuffers.editItemAt(bufferIx) = NULL;
    }
    mInputBufferIsDequeued.editItemAt(bufferIx) = true;
    
	...
	/* 2.调用到父类DecoderBase中 */
    onRequestInputBuffers();
    return true;
}

函数的工作很简单，先是调用MediaCodec拿到可用的输入buffer index，然后做一些检查工作，onRequestInputBuffers()函数是被声明为protected的，所以这里直接调用到父类中去，跟进看一下：

void NuPlayer::DecoderBase::onRequestInputBuffers() {
   
    if (mRequestInputBuffersPending) {
   
        return;
    }
    /* 调回去... */
    doRequestBuffers();
}

好吧，又调回到Decoder…

void NuPlayer::Decoder::doRequestBuffers() {
   
    if (mFormatChangePending) {
   
        return;
    }
    status_t err = OK;
    while (!mDequeuedInputBuffers.empty()) {
   
        size_t bufferIx = *mDequeuedInputBuffers.begin();
        sp<AMessage> msg = new AMessage();
        msg->setSize("buffer-ix", bufferIx);
        /* 1.从MediaSource中找到待写入buffer */
        err = fetchInputData(msg);
        if (err != OK) {
   
            break;
        }
        mDequeuedInputBuffers.erase(mDequeuedInputBuffers.begin());

		/* 2.将source中的buffer拷贝到MediaCodec中 */
        if (!mPendingInputMessages.empty()
                || !onInputBufferFetched(msg)) {
   
            mPendingInputMessages.push_back(msg);
        }
    }

	/* 确认是否还需要持续写入数据 */
    if (err == -EWOULDBLOCK
            && mSource->feedMoreTSData() == OK) {
   
        scheduleRequestBuffers();
    }
}

doRequestBuffers函数的目的是将媒体流中的文件内容写入到MediaCodec中的input buffer中。首先是调用fetchInputData函数找到Source中的待写入buffer，看一下fetchInputData函数：

status_t NuPlayer::Decoder::fetchInputData(sp<AMessage> &reply) {
   
    sp<ABuffer> accessUnit;
    bool dropAccessUnit;
    do {
   
    	/* 1.从source中获取待写入buffer的强指针 */
        status_t err = mSource->dequeueAccessUnit(mIsAudio, &accessUnit);
		/* 异常处理 */
        if (err == -EWOULDBLOCK) {
   
            return err;
        } else if (err != OK) {
   
            if (err == INFO_DISCONTINUITY) {
   
				...
            }

            reply->setInt32("err", err);
            return OK;
        }

        if (!mIsAudio) {
   
            ++mNumFramesTotal;
        }
		
		/* 标志位用于确认找到正确的可写buffer */
        dropAccessUnit =