1.3.26/1.3.28/1.3.29

Paul R. Ganci wrote:

> Therefore I backed out only that change to the 1.3.25 dccd.c version 
> with the patch attached at the bottom of this message. When I rebuild 
> DCC with this change then the dccd grey lister functions correctly 
> once again. I

Hi Vernon,

I was looking at the following code excerpts and think I see a bug. 
Admittedly my C programming abilities have deteriorated over time from 
lack of use so please show no mercy in shooting this down. However, I 
think the open_srvr_code() found in 1.3.29 is not what is intended.

 From 1.3.25:

static void open_srvr_socs(int *);

               if (get_if_changes(1)) {
                        open_srvr_socs(0);
                        flods_restart("network interfaces changed");
                }

static void
open_srvr_socs(int *retry_secs)
{
                     if (!dcc_udp_bind(dcc_emsg, &sp->udp,
                                          &sp->su, retry_secs)) {

Note that dcc_upd_bind is taking a pointer to int in the last argument. 
That address will point at an int containing the value 0 from the 
open_srvr_socs(0); call in the if statement above it.

Now from 1.3.29:

static int open_srvr_socs(int);

                if (get_if_changes(1)) {
                        int socs = open_srvr_socs(0);
                        if (!socs)
                                bad_stop("failed to open any server 
sockets");
                        else if (socs > 0)
                                flods_restart("network interfaces changed");
                }

static int                              /* # of sockets opened */
open_srvr_socs(int retry_secs)
{
    int *retry_secsp;
   retry_secsp = retry_secs ? &retry_secs : 0;

    if (!dcc_udp_bind(dcc_emsg, &sp->udp,
                                          &sp->su, retry_secsp)) {

Now note that open_srvr_socs() takes an int as an argument and that 
retry_secsp is a pointer to int. When the open_srvr_socs(0) call is made 
retry_secsp address will be null (0) since retry_secs will equal 0. 
Hence you pass a null pointer to dcc_upd_bind() as the last argument 
retry_secsp instead of a pointer to an int containing the value 0.

If my analysis is correct this code might work randomly depending on 
system. Again my analysis could be completely wrong, but I don't think 
it is. Backing out to version 1.3.25 would fix this problem if I am 
correct ... this is what I see. There are probably other simpler fixes, 
but I didn't explore them.

-- 
Paul (ganci@nurdog.com)