Martin,
I'd suggest adding those fields in a subtree of the status line protocol
item, similar to what is done in the HTTP dissector.
Regards,
Olivier
|-----Original Message-----
|From: Martin Mathieson
|
|Can you point me at examples of this? Is there a convention
|to follow (I
|notice that TCP has a SEQ/ACK tree at the end) ?
|
|I've gotten used to adding hidden fields to my own private
|dissectors, so
|obviously I know they're there :)
|
|
|----- Original Message -----
|From: Ronnie Sahlberg
|
|> Why add these as hidden fields?
|>
|> A lot of other dissectors add flags for duplicate detection
|and also error
|> conditions and
|> make them visible in the dissect tree.
|> This makes it easier for users to see that the fields exist
|and that they
|> can filter on them.
|> (Menu:Match/Selected works)
|>
|> Why not add these fields to the dissect tree as visible fields?
|>
|>
|> ----- Original Message -----
|> From: Martin Mathieson
|>
|> > The attached patches add 2 hidden display filters for SIP - namely:
|> > (1) sip.error (for all responses with code >= 300)
|> > (2) sip.resend (for all packets that appear to have been
|retransmitted).
|> A
|> > field showing a count of these is shown in the SIP stats window.
|> >
|> > These filters make it much easier to zero in on problems in large
|captures
|> > (e.g. bulk call tests) evident from the SIP stats window.
|> >
|> > (1) is straightforward, and needed because the response
|code is regarded
|> as
|> > a string, so its not easy to write your own expression
|> >
|> > (2) is more involved, and I would like to hear from other SIP users
|about
|> > this addition. Some notes:
|> > - Only considers UDP packets
|> > - Hash table contains (call_id, source_addr, dest_addr) -> (cseq,
|> > simplified transaction_state).
|> > An earlier version used proper conversations but the
|call-id is the most
|> > important part of the key. Addresses are still needed for
|when e.g.
|more
|> > than one leg of a proxied call is in the same capture.
|> > - Size of hash table key and value memchunks (as number
|of elements)
|are
|> > settable as a SIP property so as not to waste memory but
|to still allow
|> > large values