Strange problem with remuxing RTSP streams to Matroska

[Carlos Rafael Giani]

Hello, I am trying to remux a h.264/AAC stream that is transmitted over 
RTSP to Matroska. So far, the result is a Matroska file with one 
elementary stream, but not both. Sometimes, it contains h.264, sometimes 
AAC, never both.

It seems as though somehow the stream starts after the first pad is 
linked, and subsequent linking results in pads that aren't set up 
properly. But I can't see what exactly is wrong.

This can be tested by using gst-rtsp-server's test-mp4 example. Pass an 
mp4 file to it with one h.264 and one AAC stream. I tried the following 
gst-launch line, and it works properly - the resulting Matroska data 
contains both audio and video:


gst-launch-1.0 uridecodebin caps="video/x-h264, parsed=true; audio/mpeg" 
uri="rtsp://127.0.0.1:8554/test" name=d   matroskamux name=e ! filesink 
location=test.mkv  d. ! identity ! e.   d. ! aacparse ! e.


However, I wrote this example for doing it programmatically, and the 
aforementioned error occurs - only one elementary stream is present in 
the Matroska data. (I know the request pads are not released at the end, 
I ignored this for now to keep the example small). So, what is the 
difference? What is the GstParse functionality doing what I am not? 
Anybody can spot anything I missed (especially in the pad_added_cb)?


Here is the test code:


#include <gst/gst.h>
#include <glib-unix.h>


static gboolean sigint(gpointer ptr)
{
     g_main_loop_quit((GMainLoop *)ptr);
     return TRUE;
}


GstElement *pipeline;


static void pad_added_cb(GstElement *decodebin, GstPad *pad, gpointer data)
{
     GstCaps *caps;
     GstStructure *str;
     gchar const *media_type;
     gchar const *req_pad_name;
     gchar const *parser_name = "";
     GstElement *mux;
     GstElement *parser;
     GstPad *req_pad;
     GstPad *parser_pad;

     if (GST_PAD_IS_LINKED(pad))
         return;

     mux = GST_ELEMENT_CAST(data);

     caps = gst_pad_query_caps(pad, NULL);
     str = gst_caps_get_structure(caps, 0);
     media_type = gst_structure_get_name(str);

     if (g_str_has_prefix(media_type, "video/x-h264"))
     {
         req_pad_name = "video_%u";
         parser_name = "identity";
     }
     else if (g_str_has_prefix(media_type, "audio/mpeg"))
     {
         req_pad_name = "audio_%u";
         parser_name = "aacparse";
     }
     else
         return;

     parser = gst_element_factory_make(parser_name, NULL);
     gst_bin_add(GST_BIN(pipeline), parser);
     gst_element_sync_state_with_parent(parser);

     parser_pad = gst_element_get_static_pad(parser, "sink");
     gst_pad_link(pad, parser_pad);
     gst_object_unref(GST_OBJECT(parser_pad));

     parser_pad = gst_element_get_static_pad(parser, "src");
     req_pad = gst_element_get_request_pad(mux, req_pad_name);
     gst_pad_link(parser_pad, req_pad);
     gst_object_unref(GST_OBJECT(parser_pad));
}


int main(int argc, char *argv[])
{
     GMainLoop *loop;
     GstCaps *dec_caps;
     GstElement *uridecodebin, *matroskamux, *filesink;

     gst_init(&argc, &argv);

     if (argc < 2)
         return -1;

     loop = g_main_loop_new(NULL, TRUE);
     g_unix_signal_add(SIGINT, sigint, loop);

     pipeline = gst_pipeline_new(NULL);
     uridecodebin = gst_element_factory_make("uridecodebin", NULL);
     matroskamux = gst_element_factory_make("matroskamux", NULL);
     filesink = gst_element_factory_make("filesink", NULL);

     gst_bin_add_many(GST_BIN(pipeline), uridecodebin, matroskamux, 
filesink, NULL);

     gst_element_link(matroskamux, filesink);

     dec_caps = gst_caps_from_string("video/x-h264, parsed=true; 
audio/mpeg");
     g_object_set(G_OBJECT(uridecodebin), "caps", dec_caps, "uri", 
argv[1], NULL);

     g_object_set(G_OBJECT(filesink), "location", "test.mkv", NULL);

     g_signal_connect(G_OBJECT(uridecodebin), "pad-added", 
G_CALLBACK(pad_added_cb), matroskamux);

     gst_element_set_state(pipeline, GST_STATE_PLAYING);

     g_main_loop_run(loop);

     gst_object_unref(GST_OBJECT(pipeline));

     return 0;
}