<html> <head> <meta content="text/html; charset=ISO-8859-1" http-equiv="Content-Type"> </head> <body bgcolor="#FFFFFF" text="#000000"> This code worked for me <blockquote>static void sigint_restore (void) { struct sigaction action; memset (&action, 0, sizeof (action)); action.sa_handler = SIG_DFL; sigaction (SIGINT, &action, NULL); } static void sigint_handler_sighandler (int signum) { g_print ("Caught interrupt -- "); gst_element_send_event (pipeline, gst_event_new_eos ()); sigint_restore (); } static void sigint_setup (void) { struct sigaction action; memset (&action, 0, sizeof (action)); action.sa_handler = sigint_handler_sighandler; sigaction (SIGINT, &action, NULL); } </blockquote> (with sigint_setup being called when I was ready in the main code). However, if I had problems with the pipeline, sending eos was useless. The pipeline problem would mean that the pipeline would never signal that it had stopped. Ian <div class="moz-cite-prefix">On 30/01/2013 22:10, Wes Miller wrote: </div> <blockquote cite="mid:1359583856379-4658253.post@n4.nabble.com" type="cite"> <pre wrap="">I tried using sigaction instead fo signal. No help. It's almost like the SIGKILL kills the gobjects before it get to my program. Which of course makes no sense since gst-launch works perfectly well. I note gst-launch has its own eventg loop (at least it looks like one). Would that make a diffrence? </pre> </blockquote> </body> </html>