[JDEV] question on dialback.c

[Dave Smith]

On Tuesday, Mar 11, 2003, at 16:06 America/Denver, Frank Koenen wrote:

> can anyone give me insight on the following lines of code in the 
> function dialback_out_read() in dialback.c:
>
>         /* make sure we're not connecting to ourselves */
>         if( ghash_get( c->d->in_id, xmlnode_get_attrib( x , "id" ) ) 
> != NULL)
>         {
>             log_alert(c->key->server,"hostname maps back to 
> ourselves!");
>             mio_write(m, NULL, "<stream:error>Mirror Mirror on the 
> wall</stream:error>", -1);
>             mio_close(m);
>             break;
>         }
>
>

Heh. Dialback is EEEVIIL. Notice, it even has a talking mirror... ;)

> when i try to add roster items of users from one server to the next, i 
> get the "hostname maps back to ourselves!"
> error. When I enable -D debug mode on the servers to snoop on them... 
> the problem doesn't happen.

Of note, there are numerous race conditions in s2s/dialback that could 
be the cause of this problem.

> I have two servers, one running on 10.10.12.10 and one running on 
> 10.10.12.11, using a test environment and a locally
> configured DNS server to test. The hostnames and reverse lookups are 
> correct.
>
> When I comment out the lines of code above from the dialback.c, 
> everything works okay.
> Any reason why the 'ghash_get( c->d->in_id, xmlnode_get_attrib( x , 
> "id" ) ) != NULL' statement causes a false truth?
> Is it safe to comment this block of logic altogether?

Well, if it works for you, you can certainly try it. As I recall, this 
code is meant to simply protect hapless sys admins from mis-configuring 
their server and setting up a infinite message loop. So, eliminate with 
extreme prejudice :)

There's probably a more concrete answer than this, but frankly it's 
been way too long since I've worked on dialback. :)

Diz