Re: [Sipwitch-devel] Segmentation fault

[David Sugar]

If I were to make a guess, your backtrace is likely from thread.cpp +734
        
        stack::getInterface((struct sockaddr *)&iface, 
request_address.getAddr());

Where presumably request_address.getAddr() is NULL because at 733:

        request_address.set(uri_host, local_port);

failed to resolve to a valid address.  This I determined because only at
stack.cpp +642 is Socket::getinterface called with a pointer that is
passed, and this is wrapped in stack::getInterface().  Other locations
call with an address for a local object.

I think ultimately, this should be changed:

        request_address.set(uri_host, local_port);
        stack::getInterface((struct sockaddr *)&iface, 
request_address.getAddr());

        if(request_address.getAddr() == NULL) {
                error = SIP_ADDRESS_INCOMPLETE;
                goto invalid;
        }

Clearly the NULL request_address.getAddr() check should be moved up to
occur before stack::getInterface() is called.  The real question though
is then what was originally passed in uri_host??  If you have the core
file still, you may be able to find out...

On 07/04/2010 08:40 PM, John wrote:
> Hi,
> 
> I'm getting a segmentation fault in ucommon-3.1.2 on registration of a
> twinkle client to sipwitch.
> 
> Running
> sipw -gdb -trace -x9 -foreground
> 
> Gives the output below on my sipwitch server
> 
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 0xb620fb70 (LWP 17852)]
> ucc::Socket::getinterface (iface=0xb620f2ac, dest=0x0) at socket.cpp:2824
> 2824          switch(dest->sa_family) {
> (gdb) quit
> 
> Any ideas?
> 
> Thanks.
> 
> Regards,
> John Cahill

dyfet.vcf
Description: Vcard