[asterisk-dev] Sequence number sign extension problem?

[Bill Shaw]

Hi all,

I'm developing an IAX hardphone and I've run into a snag that I think is 
a sign extension problem with sequence numbers.  My setup is as follows:

PC development system running Zoiper soft phone - 192.168.1.56
Asterisk 1.6.1 server - 192.168.1.4
My hardphone - 192.168.1.31

I call extension 21 from the softphone,  IAX is configured to pass that 
call through to my hardphone.  The call runs fine for the first 8 
minutes or so,  then * begins vnak'ing every full packet I send.  The 
mini-frames continue for another 8 minutes or so until the vnaks and 
retransmissions swamp the link and sound stops.

I've captured several of these sessions and the vnaks begin abruptly 
when I send seqno 129 and every full packet I send after that gets 
vnak'ed.  I've uploaded a wireshark capture (be aware - 16 minutes of 
phone call = 45 Mb!):
http://b.shaw.home.comcast.net/asterisk/vnak_problem.pcap

The last good full packet I send is around #96430 - ping-pong-ack sequence

The first vnak is at #98057 - lagrq-lagrp-vank sequence

Every full packet after that is vnak'ed.

This is repeatable - happens every time for me.  It smells like a char 
sign extension problem in the IAX code.  Has anyone reported this 
issue?  Has it been fixed in a subsequent release?  Is there a 
workaround?  Maybe a way to reset the sequence number before it 
overflows 129?

One other thing I've noticed about the IAX code can be seen in the same 
2 areas of this capture - * is using a ping-ack-pong-ack sequence 
instead of ping-pong-ack,  and lagrq-ack-lagrp-ack instead of 
lagrq-lagrp-ack as specified in the RFQ.  This isn't a problem for me,  
just curious.  I've implemented ping-pong-ack and lagrq-lagrp-ack in my 
hardphone per the RFQ.

Thanks,

Bill