[gst-devel] Segmentation fault in g_type_check_instance_cast

[Joshua Lamorie]

Gidday there,

I'm having some issues with some code that runs fine as a 'standalone'
application, but when I put it into a thread spawned by another
application it has a segfault.

The following is from GDB:

[Switching to Thread 0x7ffff106f710 (LWP 4780)]
IA__g_type_check_instance_cast (type_instance=0x7fffe4008030
[GstAppSrc], iface_type=3825213056) at gtype.c:3973
3973   check = is_instantiatable && iface && type_node_conforms_to_U
(node, iface, TRUE, FALSE);

(gdb) bt
#0  IA__g_type_check_instance_cast (type_instance=0x7fffe4008030
[GstAppSrc], iface_type=3825213056) at gtype.c:3973
#1  0x00007ffff7b3fb64 in ESMRTPThreadMain (pArg=0x7fffec0052d0) at
RTPSourcePlugin.c:283
#2  0x000000346fe64e84 in g_thread_create_proxy (data=0x7fffec005290) at
gthread.c:1893
#3  0x000000346de07761 in start_thread (arg=0x7ffff106f710) at
pthread_create.c:301
#4  0x000000346d2e14fd in clone ()
at ../sysdeps/unix/sysv/linux/x86_64/clone.S:115


Here's some pseudo code of the overall system.

1) Main C++ application loads and initializes itself.
2) Opens a .so containing a few extern C functions/symbols.
3) Calls an Init() function from the .so (which calls gst_init())
4) Calls pthread_create() and sits on an accept() call.
4.1) The accept() returns and a message is read by the C++ app/thread.
4.2) This then calls an Execute() function from the .so.
4.3) The Execute() method creates a thread (either pthread_create or
g_thread_create).
4.3.1) The newly created thread is pretty much copy and pasted from a
functional 'standalone' gstreamer application and initializes itself.
4.3.2) During a call (cast) to GST_APP_SRC(GstElement *) a segfault
occurs somewhere within the inlined functions.

I've put this in gdb and it appears that all of the input arguments are
non-NULL, but deep within the inlined functions I saw some references to
what appear to be global locking variables.  I'm concerned that my
initialization doesn't quite match the standalone case, but I can't
quite figure out what is going on 'behind-the-curtain' in the standalone
case that is different from my multi-threaded example.

Standalone code:

GMainLoop *loop;
int main(int iArgc, char * ppcArgv[]) {
  GstElement * poAppSrc;
  GstElement * poPipeline;
  /* ... snip ... */
  /* initialize gstreamer */
  gst_init(NULL,NULL);
  loop = g_main_loop_new (NULL, TRUE);
  poPipeline = gst_pipeline_new("ESMPipeline");
  poAppSrc = gst_element_factory_make("appsrc", NULL);
  gst_app_src_set_stream_type(GST_APP_SRC(poAppSrc),
    GST_APP_STREAM_TYPE_STREAM);
  /* ... snip ... */
}

Thread code:

void * ESMRTPThreadMain(void * pArg) {
  GstElement * poPipeline = NULL;
  GstElement * poAppSrc = NULL;
  /* ... snip ... */
  poPipeline = gst_pipeline_new("ESMPipeline");
  poAppSrc = gst_element_factory_make("appsrc",NULL);
  if (poAppSrc == NULL) {
    printf("Problem generating appsrc element\n");
    return(NULL);
  }
  /* set stream type */
  gst_app_src_set_stream_type(GST_APP_SRC(poAppSrc),
    GST_APP_STREAM_TYPE_STREAM);
  /* ... snip ... */
}

So what am I doing wrong in the initialization to cause the segfault?

Thanks in advance.

-- 
Joshua Lamorie, P. Eng.
Chief Engineer
Xiphos Technologies Inc.
 
Email: jpl at xiphos.ca
Phone: +1 (514) 847-9474 ext. 227
Fax: +1 (514) 847-9474

The information transmitted is intended only for the person or entity to
which it is addressed and may contain confidential and/or privileged
material. If you have received this in error, please contact the sender
and delete this communication and any copy immediately. Thank you.