rtpmanager/rtpssrcdemux: ignore FLOW_EOS as well as NOT_LINKED

Turns out that both can happen if you push on a pad that is being
ripped away, so ignore them both.

Also fixup test to compare a flow-return using int compare, since they
are mostly negative!

Author: havardgraff

subprojects/gst-plugins-good/gst/rtpmanager/gstrtpssrcdemux.c

@@ -709,12 +709,13 @@ gst_rtp_ssrc_demux_chain (GstPad * pad, GstObject * parent, GstBuffer * buf)
   /* push to srcpad */
   ret = gst_pad_push (srcpad, buf);
 
-  if (ret == GST_FLOW_NOT_LINKED) {
-    /* on not-linked, this *could* happen because the ssrc got
+  if (ret == GST_FLOW_NOT_LINKED || ret = GST_FLOW_EOS) {
+    /* on not-linked and eos this *could* happen because the ssrc got
        removed while we were pushing, and since it can also be
        re-created, we can't know for sure if the removal was the
-       real reason, so simply ignore not-linked here */
-    GST_INFO_OBJECT (demux, "Ignoring not-linked flow return");
+       real reason, so simply ignore here */
+    GST_INFO_OBJECT (demux, "Ignoring flow return: %s",
+        gst_flow_get_name (ret));
 
     ret = GST_FLOW_OK;
   }

subprojects/gst-plugins-good/tests/check/elements/rtpssrcdemux.c

@@ -464,7 +464,7 @@ _stress_push_rtp_buffers (gpointer user_data)
 
   gst_harness_set_src_caps_str (h, "application/x-rtp");
   while (running) {
-    fail_unless_equals_uint64 (GST_FLOW_OK,
+    fail_unless_equals_int (GST_FLOW_OK,
         gst_harness_push (h, create_buffer (0, ssrc)));
     g_usleep (G_USEC_PER_SEC / 100);
     ssrc++;
@@ -483,7 +483,7 @@ _stress_push_rtcp_buffers (gpointer user_data)
 
   gst_harness_set_src_caps_str (h, "application/x-rtcp");
   while (running) {
-    fail_unless_equals_uint64 (GST_FLOW_OK,
+    fail_unless_equals_int (GST_FLOW_OK,
         gst_harness_push (h, generate_rtcp_sr_buffer (ssrc)));
     g_usleep (G_USEC_PER_SEC / 90);
     ssrc++;