Is memcmp of struct query incorrect? about tgl HOT 2 OPEN

After digging into majn/telegram-purple#220 some more, I don't believe there to be a problem with memcmp of the query struct's however it does appear that there is a msg_id conflict because the tree comparison function treats msg_ids as unique, but they are only unique per DC (well, per-DC-session), as this is how generate_next_msg_id() creates them. Thus the comparison function should be something like:

static int cmp_query(struct query *a, struct query *b) {
  int ret = 0;
  ret = (a->msg_id - b->msg_id);
  return (ret ? ret : (a->session_id - b->session_id));
}

from tgl.

For the record: I just wrote a fix (the above doesn't work as-is for several reasons) and Eion is currently testing it.

from tgl.