[Sumover-dev] Re: Rat Bug

[Piers O'Hanlon]

Hi Andrew,

Thanks for your rat crash report - I managed to track down and fix the
problem - it was a slightly wierd one - but basically it was down to a
bug in the code being triggered thru a circuitous route - For details
see:
https://mediatools.cs.ucl.ac.uk/nets/mmedia/ticket/56

The capture file was useful as I replayed it using tcpreplay and
managed to reproduce the crash (which occurred on Windows only).

Let me know of any other issues.

Thanks,

Piers.

>
> I started capturing before running RAT.  At this point, no packets were being received (I was using multicast, so without RAT running, I had not joined the multicast group).  Once I started RAT, it was about 5 seconds before it crashed.  I left the capture going for a minute anyway.
>
> I have not tried replaying this as I couldn't find a simple way to do it with ethereal.  I am fairly confident though that these are the packets that cause the crash.
>
> I have also used the new RAT in AG3 sessions, but it was this one in the ANL lobby that caused the issue.  I know that there is a bug in AG3 that sometimes gives the wrong tools the wrong addresses, so I guess something else could have been sending on the wrong channel, but there were only RTCP packets in the dump (other than the IGMP join and leave commands), which suggests something isn't getting parsed in there.
>
> I didn't try starting RAT from the command line in that room, so it could be an AG3 option I guess.
>
> Andrew :)
>
> ============================================
> Access Grid Support Centre,
> RSS Group,
> Manchester Computing,
> Kilburn Building,
> University of Manchester,
> Oxford Road,
> Manchester,
> M13 9PL,
> UK
> Tel: +44(0)161-275 0685
> Email: Andrew.Rowley at manchester.ac.uk
>
> > -----Original Message-----
> > From: p.ohanlon at gmail.com [mailto:p.ohanlon at gmail.com] On Behalf Of Piers
> > O'Hanlon
> > Sent: 01 February 2007 19:23
> > To: Andrew.Rowley at manchester.ac.uk
> > Cc: Socrates Varakliotis
> > Subject: Re: Rat Bug
> >
> > HI,
> >
> > Thanks for the info - but I have joined a few AG sessions ok with our UCL
> > RAT..
> >
> > Just to clarify - you started capturing the packets before and after
> > the rat crashed? Have you tried tcpreplay to replay the traffic and
> > crash RAT?
> >
> > How many participants were there in the session (and were they all
> > running AG3 or where there any IG2 nodes present)? Do you have the
> > command arguments used to start it - maybe there's some wierd traffic
> > on the specific port number? Does it crash when you just start rat
> > from the command line?
> >
> > Thanks,
> >
> > Piers.
> >
> > On 31/01/07, Andrew A Rowley <Andrew.Rowley at manchester.ac.uk> wrote:
> > > Hi,
> > >
> > > I was using the new UCL Rat in AG3 today (after disabling the -s
> > argument) and I found that it crashed consistently.  I therefore recorded
> > the attached Ethereal dump to record what was going on.  I think it was
> > probably related to the APP RTCP packets being sent by other AG3 nodes
> > (no-one was actually speaking, so there were no RTP packets) with the AGTk
> > RAT, but I thought you might find this useful in the quest for a more
> > robust RAT!
> > >
> > > Andrew :)
> > >
> > > ============================================
> > > Access Grid Support Centre,
> > > RSS Group,
> > > Manchester Computing,
> > > Kilburn Building,
> > > University of Manchester,
> > > Oxford Road,
> > > Manchester,
> > > M13 9PL,
> > > UK
> > > Tel: +44(0)161-275 0685
> > > Email: Andrew.Rowley at manchester.ac.uk
> > >
> > >
>
>