Jules Wang <lancelotds@xxxxxxx> wrote: > We add the same two lines as you add. As I`m not quite understand the > code, I suggest you confirm these two lines do cause the problem. Adding a few gf_log in the code, I was able to track down this issue: [2012-09-30 06:59:55.628705] I [socket.c:1809:socket_event_handler] 0-transport: disconnecting now [2012-09-30 06:59:55.628792] E [event.c:250:event_unregister_poll] 0-poll: index not found for fd=-1 (idx_hint=2) [2012-09-30 06:59:55.628826] E [socket.c:2725:fini] 0-XXXmanu: this = 0xbb70ec00 [2012-09-30 06:59:55.628860] E [socket.c:2743:fini] 0-XXXmanu: this = 0xbb70ec00, free(priv = 0xba3c7280) [2012-09-30 06:59:55.628892] E [socket.c:1778:socket_event_handler] 0-XXXmanu: this = 0xbb70f000 [2012-09-30 06:59:55.628920] I [socket.c:1809:socket_event_handler] 0-transport: disconnecting now [2012-09-30 06:59:55.628948] E [event.c:250:event_unregister_poll] 0-poll: index not found for fd=-1 (idx_hint=3) [2012-09-30 06:59:55.628981] E [socket.c:2725:fini] 0-XXXmanu: this = 0xbb70f000 [2012-09-30 06:59:55.629009] E [socket.c:2743:fini] 0-XXXmanu: this = 0xbb70f000, free(priv = 0xba3c7350) [2012-09-30 06:59:58.702383] E [socket.c:1778:socket_event_handler] 0-XXXmanu: this = 0xbb70ec00 socket_event_handler() is called with a this = 0xbb70ec00 that already went through fini(). There is a wrong refcount somewhere. -- Emmanuel Dreyfus http://hcpnet.free.fr/pubz manu@xxxxxxxxxx
