我的目标是创建一个简单的自定义接收器，能够从管道接收数据，该管道应该用于不同的应用程序(录音、广播、内部缓冲等)。
在我的第一次尝试中，这个想法是重新传输Http/Udp/等。再次通过Http进行流式传输，因此我使用souphttpsrc、队列和QHttp将数据提供给一个或多个客户端。
应用程序似乎可以工作，因为我可以用我的自定义接收器启动管道(默认情况下，在至少连接一个客户端之前，它只会忽略任何样本)，所以我只需复制客户端响应中的样本。
真正奇怪的是，下载速度真的真的比传入的流媒体速度慢(5-10 KB/s，而不是预期的2-300KB/s)，而且数据似乎完全不可用。我曾尝试在sink之前在管道中插入一个decodebin，速度达到33MBit/s，因此我会排除主线程中负责通过网络发送实际样本的方法的Queued调用引起的性能问题；即使在这种情况下，数据看起来也只是垃圾。
第二，如果我添加一个tee(或多队列)来并行运行一个autovideosink和/或一个autoaudiossink，那么管道在启动时就会被卡住(这不仅仅发生在两个autolink中)，如果我只尝试将两个自定义appsink链接到tee，那么缺陷也是一样的。
调试代码时，即使总线上接收到的最后一条消息是自定义接收器的播放状态更改，管道挂起时似乎永远不会调用newSample()。
谢谢！
这里的代码：
QGst:：FlowReturn MultiHttpSink:：newSample()｛
QGst::SamplePtr sample = pullSample();
char data[sample->buffer()->size()];
qDebug() << "New Sample: size " << sample->buffer()->size();
qDebug() << sample->buffer()->duration().toTime();
qDebug() << sample->buffer()->presentationTimeStamp().toTime();
sample->buffer()->extract(0, &data, sample->buffer()->size());
for (auto it = resources.begin(); it != resources.end(); it++)
QMetaObject::invokeMethod(&*server, "writeToHttpResource", Qt::QueuedConnection, Q_ARG(QPointer<qhttp::server::QHttpResponse>, *it), Q_ARG(QByteArray, data));
return QGst::FlowOk;
}
void Server::setupServer() {
server = new QHttpServer(app);
server->listen(QHostAddress::Any, 8080, [&](QHttpRequest* req, QHttpResponse * res) {
res->setStatusCode(qhttp::ESTATUS_OK); // http status 200
res->addHeader("Content-Type", "video/mp2t");
m_sink.addAudience(res);
});
if (!server->isListening()) {
fprintf(stderr, "failed. can not listen at port 8080!n");
throw std::exception();
}
void Server::setupPipeline() {
/* source pipeline */
QString pipe1Descr = QString(
"souphttpsrc location="%1" ! "
"queue ! tee name=splitter "
"splitter.! decodebin ! autoaudiosink "
"splitter.! decodebin ! autovideosink "
"splitter.! appsink name="mysink" "
).arg("URL");
pipeline1 = QGst::Parse::launch(pipe1Descr).dynamicCast<QGst::Pipeline>();
m_sink.setElement(pipeline1->getElementByName("mysink"));
QGlib::connect(pipeline1->bus(), "message", this, &Server::onBusMessage);
pipeline1->bus()->addSignalWatch();
/* start playing */
pipeline1->setState(QGst::StatePlaying);
}
QString stateString(QGst::State state) {
switch (state) {
case 0:
return "Void Pending";
case 1:
return "Null";
case 2:
return "Ready";
case 3:
return "Playing";
case 4:
return "Paused";
}
return "";
}
QString streamStateString(QGst::StreamStatusType type) {
switch (type) {
case 0:
return "Create";
case 1:
return "Enter";
case 2:
return "Leave";
case 3:
return "Destroy";
case 4:
return "Start";
case 5:
return "Pause";
case 6:
return "Stop";
}
}
void Server::onBusMessage(const QGst::MessagePtr& message) {
switch (message->type()) {
case QGst::MessageEos:
app->quit();
break;
case QGst::MessageError:
qCritical() << message.staticCast<QGst::ErrorMessage>()->error();
qCritical() << message.staticCast<QGst::ErrorMessage>()->debugMessage();
break;
case QGst::MessageStateChanged:
{
QGlib::RefPointer<QGst::StateChangedMessage> msg = message.staticCast<QGst::StateChangedMessage>();
qDebug() << msg->source()->name() <<": State changed from " << stateString(msg->oldState())
<< " -> " << stateString(msg->newState()) << " Transitioning to " << stateString(msg->pendingState());
if(msg->source()->name().compare("pipeline0")==0 ){
if( msg->newState() == QGst::StatePaused ){
need_reset = true;
playing = false;
} else {
need_reset = false;
if( msg->newState() == QGst::StatePlaying ){
playing = true;
}
}
}
break;
}
case QGst::MessageStreamStatus:
{
QGlib::RefPointer<QGst::StreamStatusMessage> msg = message.staticCast<QGst::StreamStatusMessage>();
qDebug() << "Stream Status: " << streamStateString(msg->statusType());
break;
}
default:
qDebug() << "Unhandles Bus Message Type: " << message->typeName();
break;
}
}
void Server::writeToHttpResource(QPointer<qhttp::server::QHttpResponse> res, QByteArray data) {
qDebug() << "Writing data...";
res->write(data);
}
更新
gstreamer的调试日志报告了一些有用的内容。。。
0:00:00.733339246 [335m 8248[00m 0x2747450 [37mDEBUG [00m [00m souphttpsrc gstsouphttpsrc.c:1430:gst_soup_http_src_chunk_allocator:<souphttpsrc0>[00m alloc 4096 bytes <= 18446744073709551615
0:00:00.733350762 [335m 8248[00m 0x274a8f0 [37mDEBUG [00m [00m tee gsttee.c:774:gst_tee_chain:<splitter>[00m received buffer 0x7f69d000fb80
0:00:00.733353629 [335m 8248[00m 0x2747450 [37mDEBUG [00m [00;01;34m GST_MEMORY gstmemory.c:138:gst_memory_init:[00m new memory 0x7f69d0015120, maxsize:4103 offset:0 size:4096
0:00:00.733361430 [335m 8248[00m 0x274a8f0 [37mDEBUG [00m [00;01;35m GST_SCHEDULING gstpad.c:4174:gst_pad_chain_data_unchecked:<mysink2:sink>[00m calling chainfunction &gst_base_sink_chain with buffer buffer: 0x7f69d000fb80, pts 99:99:99.999999999, dts 99:99:99.999999999, dur 99:99:99.999999999, size 1430, offset 15582, offset_end none, flags 0x0
传入网络数据包的大小为4096，但缓冲区的大小在大多数情况下为1430。。。正在调查。。。。我认为我应该以不同的方式访问缓冲区以获得对原始数据的访问
网络数据包的大小是14304096，似乎是缓冲区。。。
0:00:29.316301105 [335m 8248[00m 0x2747450 [37mDEBUG [00m [00m souphttpsrc gstsouphttpsrc.c:1482:gst_soup_http_src_got_chunk_cb:<souphttpsrc0>[00m got chunk of 1430 bytes
为什么转会缓慢且毫无意义，目前仍是一个谜。为什么它也挂着！
更新2
我也尝试过这种方法，结果相同(报告的尺寸相同)：
QGst::FlowReturn MultiHttpSink::newSample() {
QGst::SamplePtr sample = pullSample();
QGst::BufferPtr buffer = sample->buffer();
qDebug() << "New Sample: size " << buffer->size();
QGst::MapInfo map;
if (!buffer->map(map, QGst::MapRead))
return QGst::FlowError;
qDebug() << "New Buffer Map: size " << map.size();
for (auto it = resources.begin(); it != resources.end(); it++)
QMetaObject::invokeMethod(&*server, "writeToHttpResource", Qt::QueuedConnection, Q_ARG(QPointer<qhttp::server::QHttpResponse>, *it), Q_ARG(QByteArray, (char *)map.data()));
buffer->unmap(map);
return QGst::FlowOk;
}
这里是最后一个版本的输出，与之前相同：
... ...
New Sample: size 1430
New Buffer Map: size 1430
QTime("23:34:33.709")
QTime("23:34:33.709")
Writing data...
New Sample: size 4096
New Buffer Map: size 4096
QTime("23:34:33.709")
QTime("23:34:33.709")
Writing data...
... ...
更新3
这是每个数据块之间的日志：
New Sample: size 1430
New Buffer Map: size 1430
QTime("23:34:33.709")
QTime("23:34:33.709")
0:04:28.221638796 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m basesink gstbasesink.c:3566:gst_base_sink_chain_unlocked:<mysink2>[00m object unref after render 0x7f894400f850
0:04:28.221653137 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00;01;35m GST_SCHEDULING gstpad.c:4180:gst_pad_chain_data_unchecked:<mysink2:sink>[00m called chainfunction &gst_base_sink_chain with buffer 0x7f894400f850, returned ok
0:04:28.221667131 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m tee gsttee.c:778:gst_tee_chain:<splitter>[00m handled buffer ok
0:04:28.221676164 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00;01;35m GST_SCHEDULING gstpad.c:4180:gst_pad_chain_data_unchecked:<splitter:sink>[00m called chainfunction &gst_tee_chain with buffer 0x7f894400f850, returned ok
0:04:28.221687738 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m queue_dataflow gstqueue.c:1482:gst_queue_loop:<queue0>[00m queue is empty
0:04:28.222861204 [335m28768[00m 0xd87450 [37mDEBUG [00m [00m souphttpsrc gstsouphttpsrc.c:1430:gst_soup_http_src_chunk_allocator:<souphttpsrc0>[00m alloc 4096 bytes <= 18446744073709551615
0:04:28.222896720 [335m28768[00m 0xd87450 [37mDEBUG [00m [00;01;34m GST_MEMORY gstmemory.c:138:gst_memory_init:[00m new memory 0x7f8944012fe0, maxsize:4103 offset:0 size:4096
0:04:28.222926663 [335m28768[00m 0xd87450 [37mDEBUG [00m [00m souphttpsrc gstsouphttpsrc.c:1482:gst_soup_http_src_got_chunk_cb:<souphttpsrc0>[00m got chunk of 1430 bytes
0:04:28.222956032 [335m28768[00m 0xd87450 [37mDEBUG [00m [00m basesrc gstbasesrc.c:2316:gst_base_src_do_sync:<souphttpsrc0>[00m no sync needed
0:04:28.222968263 [335m28768[00m 0xd87450 [37mDEBUG [00m [00m basesrc gstbasesrc.c:2520:gst_base_src_get_range:<souphttpsrc0>[00m buffer ok
0:04:28.222978663 [335m28768[00m 0xd87450 [37mDEBUG [00m [00;01;35m GST_SCHEDULING gstpad.c:4174:gst_pad_chain_data_unchecked:<queue0:sink>[00m calling chainfunction &gst_queue_chain with buffer buffer: 0x7f894400f630, pts 99:99:99.999999999, dts 99:99:99.999999999, dur 99:99:99.999999999, size 1430, offset 20789192, offset_end none, flags 0x0
0:04:28.223004500 [335m28768[00m 0xd87450 [37mDEBUG [00m [00;01;35m GST_SCHEDULING gstpad.c:4180:gst_pad_chain_data_unchecked:<queue0:sink>[00m called chainfunction &gst_queue_chain with buffer 0x7f894400f630, returned ok
0:04:28.223013700 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m queue_dataflow gstqueue.c:1494:gst_queue_loop:<queue0>[00m queue is not empty
0:04:28.223031507 [335m28768[00m 0xd87450 [37mDEBUG [00m [00m basesrc gstbasesrc.c:2355:gst_base_src_update_length:<souphttpsrc0>[00m reading offset 20790622, length 4096, size -1, segment.stop -1, maxsize -1
0:04:28.223066379 [335m28768[00m 0xd87450 [37mDEBUG [00m [00m basesrc gstbasesrc.c:2456:gst_base_src_get_range:<souphttpsrc0>[00m calling create offset 20790622 length 4096, time 0
0:04:28.223056071 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00;01;35m GST_SCHEDULING gstpad.c:4174:gst_pad_chain_data_unchecked:<splitter:sink>[00m calling chainfunction &gst_tee_chain with buffer buffer: 0x7f894400f630, pts 99:99:99.999999999, dts 99:99:99.999999999, dur 99:99:99.999999999, size 1430, offset 20789192, offset_end none, flags 0x0
0:04:28.223111831 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m tee gsttee.c:774:gst_tee_chain:<splitter>[00m received buffer 0x7f894400f630
0:04:28.223121952 [335m28768[00m 0xd87450 [37mDEBUG [00m [00m souphttpsrc gstsouphttpsrc.c:1430:gst_soup_http_src_chunk_allocator:<souphttpsrc0>[00m alloc 4096 bytes <= 18446744073709551615
0:04:28.223139071 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00;01;35m GST_SCHEDULING gstpad.c:4174:gst_pad_chain_data_unchecked:<mysink2:sink>[00m calling chainfunction &gst_base_sink_chain with buffer buffer: 0x7f894400f630, pts 99:99:99.999999999, dts 99:99:99.999999999, dur 99:99:99.999999999, size 1430, offset 20789192, offset_end none, flags 0x0
0:04:28.223152343 [335m28768[00m 0xd87450 [37mDEBUG [00m [00;01;34m GST_MEMORY gstmemory.c:138:gst_memory_init:[00m new memory 0x7f8944015120, maxsize:4103 offset:0 size:4096
0:04:28.223179317 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m basesink gstbasesink.c:3409:gst_base_sink_chain_unlocked:<mysink2>[00m got times start: 99:99:99.999999999, end: 99:99:99.999999999
0:04:28.223208511 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m basesink gstbasesink.c:1958:gst_base_sink_get_sync_times:<mysink2>[00m got times start: 99:99:99.999999999, stop: 99:99:99.999999999, do_sync 0
0:04:28.223222869 [335m28768[00m 0xd87450 [37mDEBUG [00m [00;01;34m GST_MEMORY gstmemory.c:87:_gst_memory_free:[00m free memory 0x7f8944015120
0:04:28.223238437 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00;04m default gstsegment.c:731:gst_segment_to_running_time_full:[00m invalid position (-1)
0:04:28.223281513 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00;04m default gstsegment.c:731:gst_segment_to_running_time_full:[00m invalid position (-1)
0:04:28.223298205 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m basesink gstbasesink.c:3520:gst_base_sink_chain_unlocked:<mysink2>[00m rendering object 0x7f894400f630
0:04:28.223312428 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m basesink gstbasesink.c:946:gst_base_sink_set_last_buffer_unlocked:<mysink2>[00m setting last buffer to 0x7f894400f630
0:04:28.223321131 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00;01;34m GST_MEMORY gstmemory.c:87:_gst_memory_free:[00m free memory 0x7f8944014080
0:04:28.223344396 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00;01;34m GST_CAPS gstpad.c:2641:gst_pad_has_current_caps:<mysink2:sink>[00m check current pad caps (NULL)
0:04:28.223355939 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m appsink gstappsink.c:760:gst_app_sink_render:<mysink2>[00m pushing render buffer 0x7f894400f630 on queue (0)
0:04:28.223369213 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m appsink gstappsink.c:1286:gst_app_sink_pull_sample:<mysink2>[00m trying to grab a buffer
0:04:28.223377762 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m appsink gstappsink.c:706:dequeue_buffer:<mysink2>[00m dequeued buffer 0x7f894400f630
0:04:28.223386131 [335m28768[00m 0xd8a8f0 [37mDEBUG [00m [00m appsink gstappsink.c:1301:gst_app_sink_pull_sample:<mysink2>[00m we have a buffer 0x7f894400f630
New Sample: size 1430
New Buffer Map: size 1430
QTime("23:34:33.709")
QTime("23:34:33.709")
在decodebin之前和tee之后添加队列，tee需要上下文才能运行。
"souphttpsrc location=\"%1\"！"队列！tee name=拆分器"拆分器。！队列！decodebin！自动音频接收器"拆分器。！队列！decodebin！autovideosink"拆分器。！appsink name=\"mylink\">
链接到appsink的最后一个拆分器不需要队列，它将在源插件创建的gsttask的上下文中运行。
我终于发现了错误的bug！
QMetaObject::invokeMethod(&*server, "writeToHttpResource", Qt::QueuedConnection, Q_ARG(QPointer<qhttp::server::QHttpResponse>, *it), Q_ARG(QByteArray, (char *)map.data()));
即使这个代码编译了，它也运行得很糟糕，因为这个：
Q_ARG(QByteArray, (char *)map.data())
在这种情况下，使用QByteArray(char*)是错误的，因为我们不是从字符串初始化它，而是从原始数据块初始化它。按原样编写的数据块一直被解析，寻找新的行组合，随机剪切数据块。
为了避免这种情况，要使用的构造函数是QByteArray(char*ptr，int size)。
这是正确的方法版本：
QMetaObject::invokeMethod(&*server, "writeToHttpResource", Qt::QueuedConnection, Q_ARG(QPointer<GssClientRequest>, *it), Q_ARG(QByteArray, QByteArray((char *) map.data(), map.size())));
最后不是一个很难打败的虫子；)
感谢您的支持！