[Bro-Dev] conn-size

[Gregor Maier]

Actually there's a bit more to that. Even the original code did a
    Unref(conn_val);
    conn_val = 0;
in Connection::FlipRoles(), i.e. the full conn_val had to be rebuild
after a FlipRole() and only the endpoint pointers were reused as the
original code also kept distinct pointers to the endp_vals.

My code however doesn't keep pointers to the endp_vals, because I now
use a root_analyzer->UpateConnVal() instead of two
root_analyzer->UpdateEndpointVal() calls. As per our general discussion,
I'm doing this so that any child analyzer could now update or extend the
conn_val or the endpoint vals. So, swapping doesn't make sense anymore
unless we keep conn_val (in which case conn_id and the endp_vals have to
be replaced).

(However, I just saw that the conn_size analyzer doesn't flip its
counters. Will fix that.)

(sorry for the first, incorrect e-mail. Should reply to e-mails before
my first coffee)

cu
gregor

On 5/2/11 16:06 , Robin Sommer wrote:
> Gregor,
> 
> the following is part of your conn-size patch:
> 
>     -   RecordVal* tmp_rc = resp_endp;                                                                                                                                       
>     -   resp_endp = orig_endp;                                                                                                                                               
>     -   orig_endp = tmp_rc;                                                                                                                                                  
>     -                                                                                                                                                                        
>     +   // Delete conn_val, will build a new one next time we need one                                                                                                       
>     +   // XXX: breaks the assumption that conn_val never changes!!                
> 
> 
> Why does a swap not work anymore?
> 
> Robin
> 


-- 
Gregor Maier
<gregor at icir.org>  <gregor at icsi.berkeley.edu>
Int. Computer Science Institute (ICSI)
1947 Center St., Ste. 600
Berkeley, CA 94704, USA
http://www.icir.org/gregor/