possible fix for CORE_disconnect segfault, GNUNET_break instead of segfault in other...

diff --git a/src/core/core_api.c b/src/core/core_api.c
index abb7fbc63d8abf139f26428e070a6a3ee1d9cc19..48de0a2147fa55f03f510bb52b08e41e47855d52 100644 (file)
--- a/src/core/core_api.c
+++ b/src/core/core_api.c
@@ -479,6 +479,9 @@ reconnect_later (struct GNUNET_CORE_Handle *h)
     {
       GNUNET_CLIENT_disconnect (h->client, GNUNET_NO);
       h->client = NULL;
+      /* FIXME: is this right? GNUNET_CLIENT_disconnect frees all outstanding transmission handles, so h->cth is invalid! */
+      /* Otherwise a subsequent call to GNUNET_CORE_disconnect tries to cancel this task!!! */
+      h->cth = NULL;
       GNUNET_CONTAINER_multihashmap_iterate (h->peers,
                                             &disconnect_and_free_peer_entry,
                                             h);
@@ -1124,6 +1127,14 @@ main_notify_handler (void *cls,
                  "Received notification about transmission readiness to `%s'.\n",
                  GNUNET_i2s (&smr->peer));
 #endif
+      /* FIXME: pr->pending_head is sometimes NULL here... Safe to just return?  Or does this indicate something is out of sync somewhere else? */
+      if (pr->pending_head == NULL)
+        {
+          GNUNET_break (0);
+          reconnect_later (h);
+          return;
+        }
+
       th = pr->pending_head;
       if (ntohs (smr->smr_id) != th->smr_id)
        {
@@ -1365,7 +1376,7 @@ GNUNET_CORE_connect (const struct GNUNET_CONFIGURATION_Handle *cfg,
 /**
  * Disconnect from the core service.  This function can only 
  * be called *after* all pending 'GNUNET_CORE_notify_transmit_ready'
- * requests have been explicitly cancelled.
+ * requests have been explicitly canceled.
  *
  * @param handle connection to core to disconnect
  */